From patchwork Wed Nov 22 16:34:22 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Donald Robson X-Patchwork-Id: 746153 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=imgtec.com header.i=@imgtec.com header.b="JOKUnQHX"; dkim=pass (1024-bit key) header.d=IMGTecCRM.onmicrosoft.com header.i=@IMGTecCRM.onmicrosoft.com header.b="eTZRVFLw" Received: from mx08-00376f01.pphosted.com (mx08-00376f01.pphosted.com [91.207.212.86]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id BB84F1B9; Wed, 22 Nov 2023 08:35:48 -0800 (PST) Received: from pps.filterd (m0168888.ppops.net [127.0.0.1]) by mx08-00376f01.pphosted.com (8.17.1.24/8.17.1.24) with ESMTP id 3AM7kZhg006720; Wed, 22 Nov 2023 16:35:05 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=imgtec.com; h= from:to:cc:subject:date:message-id:in-reply-to:references :content-transfer-encoding:content-type:mime-version; s= dk201812; bh=VcTZzRS87l/H8tP3ckp60h9mtsAe2UTSbAIQrqe0Wbg=; b=JOK UnQHXqlxK8pEvbdl6sOFvXLb/6jGL67LTfzKojTZQnx34BZQbxdxLV2v1Vr5w3hY lcgx+NITpAuvVxo1bKKlNnKlHeKDY+TK9FVE8M0z2ZV0UI//CzmI+42hX9nRX71q +eBHrVtOqO+26fiGe1algJqFSr6SLibyCHwoHWSF3JR6LzyKE7ENIq0mlSO8cOzr 7AVorx1YVru+rawu6D9qz6LjataYqXE+p4rQKqqAKjb2zlNEsJyYTpEtyX8ijvCw ZiL/PaPhzh6NsD+vlDhznSYQjwwwZBBXXjQYNcdy8XTh7ZgjHhZu0UU3b5ZXxfqn J0Khvx4sKQKXpcVC4uA== Received: from hhmail05.hh.imgtec.org ([217.156.249.195]) by mx08-00376f01.pphosted.com (PPS) with ESMTPS id 3ug99gsf30-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128 verify=NOT); Wed, 22 Nov 2023 16:35:05 +0000 (GMT) Received: from HHMAIL04.hh.imgtec.org (10.100.10.119) by HHMAIL05.hh.imgtec.org (10.100.10.120) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34; Wed, 22 Nov 2023 16:35:04 +0000 Received: from GBR01-LO4-obe.outbound.protection.outlook.com (104.47.85.105) by email.imgtec.com (10.100.10.121) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34 via Frontend Transport; Wed, 22 Nov 2023 16:35:04 +0000 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=SGucCVn9pp9hlIGNm83VmxGCxdRy1UNPEzArUHJe7KMga8F6DpAOFGgUN/3QbA+Vgc1alRkTbT8WE23aUUbSSpu6+dCA4YiEPYZby/NzMiB8Z9orsHmAAPlsNGsFT1YlTzT2BjJIZTyR68amu4p05/9kRhRGRvIiSi2csndNyJFMedINCMgfJGv8LGulFQ6dS5JLU06/rz2ZS+22mzN9vaVLYIJ037kTgm0vPaRMdPD0AfoeJbFaSiY14SnB/3AlSHqh8nNXkrF8Jni5xZVQ8jM8lJ9Ajlu33iILMT+XypK/lxE/QaZx6SrQ9DAWEob+ycV6r0onCkobXhkmWFX2CA== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=VcTZzRS87l/H8tP3ckp60h9mtsAe2UTSbAIQrqe0Wbg=; b=AY6Iov9B9azG0v7nPQnpFezqt+exB+GlflrTjQ8AQsgJRftIT12Ziq8+zayeGfwKxJD/PbByIESb22TakH44B0k/EtZjgzoyts4i3ddg1DrJaMMAqSS6r2Cq/arGXeW1uM1vCRjrvFGqoG53hu7H8XvcKfBxq8YkXZ+TYUx3R33r56Dhd8lF1owstidX2rCTUTZq9yPLdoxcP35NsCNrNjD9pLwe2M6Q8fYiQMXWaQfa7klhsPdcEg6IZUePQqKZcYQGtEXEjczyk66KwfeWnYf2gt6qzlHq/8MGs9Lz/7D8Zrcy7auuwi0+cGYA3m9gUSNHH+fQAPNnXHIDbJNMLw== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=imgtec.com; dmarc=pass action=none header.from=imgtec.com; dkim=pass header.d=imgtec.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=IMGTecCRM.onmicrosoft.com; s=selector2-IMGTecCRM-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=VcTZzRS87l/H8tP3ckp60h9mtsAe2UTSbAIQrqe0Wbg=; b=eTZRVFLwviV+B+DuZGme75S6sE/eYHzGBu7FZjZ5mNyE/E8/Th7oO9qEPL23nbHKfPVRSBS7IoXFvc9aaBHIqZ6ua40sEEP2zbaNTmPWuZh+A/JF04e8DR7y6FJqlXX1rw25eVGO4p3Pr8FZiQbwsYOu/jzdhKrhDZ++tYNOuUE= Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) by CWLP265MB5329.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1c5::11) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.7025.19; Wed, 22 Nov 2023 16:35:02 +0000 Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34]) by CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34%3]) with mapi id 15.20.7025.017; Wed, 22 Nov 2023 16:35:02 +0000 From: Donald Robson To: CC: , , , , , , , , , , , , , , , , , , , , , , Sarah Walker , Linus Walleij Subject: [PATCH v9 01/20] sizes.h: Add entries between SZ_32G and SZ_64T Date: Wed, 22 Nov 2023 16:34:22 +0000 Message-Id: <58b227d96f27859b453caf0ceaaac81a6616304b.1700668843.git.donald.robson@imgtec.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: References: X-ClientProxiedBy: LO4P265CA0273.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:37a::10) To CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) Precedence: bulk X-Mailing-List: devicetree@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: CWLP265MB5770:EE_|CWLP265MB5329:EE_ X-MS-Office365-Filtering-Correlation-Id: 2b798722-4392-4e1a-a997-08dbeb78fa01 X-MS-Exchange-SenderADCheck: 1 X-MS-Exchange-AntiSpam-Relay: 0 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-Forefront-Antispam-Report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:; IPV:NLI; SFV:NSPM; H:CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM; PTR:; CAT:NONE; SFS:(13230031)(396003)(366004)(136003)(39850400004)(376002)(346002)(230922051799003)(451199024)(1800799012)(186009)(64100799003)(26005)(52116002)(6506007)(6666004)(2616005)(6512007)(6486002)(44832011)(5660300002)(4326008)(8676002)(41300700001)(8936002)(7416002)(2906002)(478600001)(316002)(6916009)(54906003)(66476007)(66556008)(66946007)(86362001)(38100700002)(36756003)(38350700005); DIR:OUT; SFP:1102; X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: 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 X-MS-Exchange-CrossTenant-Network-Message-Id: 2b798722-4392-4e1a-a997-08dbeb78fa01 X-MS-Exchange-CrossTenant-AuthSource: CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 22 Nov 2023 16:35:02.7618 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 0d5fd8bb-e8c2-4e0a-8dd5-2c264f7140fe X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: //wQwP5fAIjvAhiWZ7rKPuvZWMrfi/gtbU27eccpy/6D4llqk5t+Aic+d5rFhP+KIwGiCj+HUuhY+IZi9H+ODyKhOdcQn/ouR8YnNjdEYQM= X-MS-Exchange-Transport-CrossTenantHeadersStamped: CWLP265MB5329 X-OriginatorOrg: imgtec.com X-EXCLAIMER-MD-CONFIG: 15a78312-3e47-46eb-9010-2e54d84a9631 X-Proofpoint-GUID: PqteWj2Qr0dQbSAab5FGMUlzgltzlg7L X-Proofpoint-ORIG-GUID: PqteWj2Qr0dQbSAab5FGMUlzgltzlg7L From: Matt Coster sizes.h has a gap in defines between SZ_32G and SZ_64T. Add the missing defines so they can be used in drivers. Signed-off-by: Matt Coster Signed-off-by: Sarah Walker Signed-off-by: Donald Robson Reviewed-by: Linus Walleij --- include/linux/sizes.h | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/include/linux/sizes.h b/include/linux/sizes.h index 84aa448d8bb3..c3a00b967d18 100644 --- a/include/linux/sizes.h +++ b/include/linux/sizes.h @@ -47,8 +47,17 @@ #define SZ_8G _AC(0x200000000, ULL) #define SZ_16G _AC(0x400000000, ULL) #define SZ_32G _AC(0x800000000, ULL) +#define SZ_64G _AC(0x1000000000, ULL) +#define SZ_128G _AC(0x2000000000, ULL) +#define SZ_256G _AC(0x4000000000, ULL) +#define SZ_512G _AC(0x8000000000, ULL) #define SZ_1T _AC(0x10000000000, ULL) +#define SZ_2T _AC(0x20000000000, ULL) +#define SZ_4T _AC(0x40000000000, ULL) +#define SZ_8T _AC(0x80000000000, ULL) +#define SZ_16T _AC(0x100000000000, ULL) +#define SZ_32T _AC(0x200000000000, ULL) #define SZ_64T _AC(0x400000000000, ULL) #endif /* __LINUX_SIZES_H__ */ From patchwork Wed Nov 22 16:34:23 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Donald Robson X-Patchwork-Id: 746154 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=imgtec.com header.i=@imgtec.com header.b="ceTKru1S"; dkim=pass (1024-bit key) header.d=IMGTecCRM.onmicrosoft.com header.i=@IMGTecCRM.onmicrosoft.com header.b="HTSmNgkL" Received: from mx07-00376f01.pphosted.com (mx07-00376f01.pphosted.com [185.132.180.163]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 7E5C61B8; Wed, 22 Nov 2023 08:35:33 -0800 (PST) Received: from pps.filterd (m0168889.ppops.net [127.0.0.1]) by mx07-00376f01.pphosted.com (8.17.1.24/8.17.1.24) with ESMTP id 3AMCaKJE020106; Wed, 22 Nov 2023 16:35:07 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=imgtec.com; h= from:to:cc:subject:date:message-id:in-reply-to:references :content-transfer-encoding:content-type:mime-version; s= dk201812; bh=Zi8AOeDOdj1x0erlVdaofiiJpR7zAbn10yP0fAu/tQk=; b=ceT Kru1SPdAdKvYWmBDTVlCT6zkpPA93Jo+UQVzlGWS9l9QKLXLWcmGSR/V305yCuWS BzGeya6OagtREAUJl2N9Zl/RxfHltwLwCSP646Talb+BnEFQw/btbGLbVBi8qDoe ZB52ixhW/xi8uhiTZEPOPmyTpQ17HnUcGMpTZg9yTdqAiOFaezSRTwQHBVaJrC0M uanMBAz6zUetL5ZRv4uQoklgrbXIKPxPX1E3yGl9aMKWS/IDVPnY7zfJzmwSWtA8 btFKR8nIoYHTBVxGbjPielZy6s+RKpCiG6J/nVl0V5HsRdiKGSzSP0kMMD1+VGLd FTr4zmZdcidAnBePeAw== Received: from hhmail05.hh.imgtec.org ([217.156.249.195]) by mx07-00376f01.pphosted.com (PPS) with ESMTPS id 3ug99g9m20-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128 verify=NOT); Wed, 22 Nov 2023 16:35:07 +0000 (GMT) Received: from HHMAIL04.hh.imgtec.org (10.100.10.119) by HHMAIL05.hh.imgtec.org (10.100.10.120) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34; Wed, 22 Nov 2023 16:35:06 +0000 Received: from GBR01-CWX-obe.outbound.protection.outlook.com (104.47.85.41) by email.imgtec.com (10.100.10.121) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34 via Frontend Transport; Wed, 22 Nov 2023 16:35:06 +0000 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=XRKZs5dU5WGZ51HxCdLHCVz0640WwHO/687wjlDxp7rBcQoC3OOtVgB+GB6QCmv1FgD+RZZ0GFeNmyMQ3jjk3m0JlRlLNNX2mIK4TNJXx+J6k37LExmLYBjg7ZrExnx9iFEZl4NAxXO5nUQO9tHSo14SU4+A+YQe3bj9iUlVrdQ3RmckB35K7Gyi5spvNltraV7RJKFT+jTFTn4gXdZYJ7JQ4s8p4koe0StNNao4jEIFlKL98Rs6jStU7j+fePTUnl8IgIzQ9KK/RBTs2+wVh/SQmxKL1Cat82+ZEb7dVBLIDmWT4TiWpoJbuoYTYsa1YeUpoyewTiEaapoooOHdpQ== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=Zi8AOeDOdj1x0erlVdaofiiJpR7zAbn10yP0fAu/tQk=; b=ngSfuAkJ92gScSNo/i3BXY8TkFpYztRdjZCehi/rfvEfdEBzn9FA3ZZDeHa2ABl3NZPsWuLySEpLMCP6Ll0odBZrzw4oh2k9XnvdVcT0PWfTXXOB6Y2P0HwSQKQAlcnEtZGZlut/Q1ukXPS3q36l/9TKEkpQmX74vhJ2grVb0VkxB+sUuZX4U10e/n4qoz1W/M3IahI30eeozrZK9fcJ61ObvVhDlXUUl+su0OklU1EoCXZQ1Rttib+9pUi5BMUpPuZI9AQHAzFQ7rTOKSxfrVHnWzUPXo8HrdiZ8GlXhS81qYRbsIPDqQ5B8IxxB8A9r4SNcsGw5bnW6t+Q1af3Fg== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=imgtec.com; dmarc=pass action=none header.from=imgtec.com; dkim=pass header.d=imgtec.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=IMGTecCRM.onmicrosoft.com; s=selector2-IMGTecCRM-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=Zi8AOeDOdj1x0erlVdaofiiJpR7zAbn10yP0fAu/tQk=; b=HTSmNgkLyIZDYy/EHN4HlAvvBsRaEu0MPBFdNEFlNH4rkB7RxL7oW4xpZmM+7pzL4586Gy0k4pGUcxvNbAlfAQlPF+NP5fOH9kD04NDJ++8dZrVbuh5UmzfVop98/LFF28Jgv7N1sYoF10DY6y8By1hLIl9cFgPxGeDETPGOmqY= Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) by CWLP265MB5329.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1c5::11) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.7025.19; Wed, 22 Nov 2023 16:35:05 +0000 Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34]) by CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34%3]) with mapi id 15.20.7025.017; Wed, 22 Nov 2023 16:35:04 +0000 From: Donald Robson To: CC: , , , , , , , , , , , , , , , , , , , , , , Sarah Walker Subject: [PATCH v9 02/20] drm/gpuvm: Helper to get range of unmap from a remap op. Date: Wed, 22 Nov 2023 16:34:23 +0000 Message-Id: <8a0a5b5eeec459d3c60fcdaa5a638ad14a18a59e.1700668843.git.donald.robson@imgtec.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: References: X-ClientProxiedBy: LO4P265CA0273.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:37a::10) To CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) Precedence: bulk X-Mailing-List: devicetree@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: CWLP265MB5770:EE_|CWLP265MB5329:EE_ X-MS-Office365-Filtering-Correlation-Id: e053100b-cd6d-4112-9ab9-08dbeb78fb56 X-MS-Exchange-SenderADCheck: 1 X-MS-Exchange-AntiSpam-Relay: 0 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-Forefront-Antispam-Report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:; IPV:NLI; SFV:NSPM; H:CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM; PTR:; CAT:NONE; SFS:(13230031)(396003)(366004)(136003)(39850400004)(376002)(346002)(230922051799003)(451199024)(1800799012)(186009)(64100799003)(26005)(52116002)(6506007)(6666004)(107886003)(2616005)(6512007)(6486002)(44832011)(5660300002)(4326008)(8676002)(41300700001)(8936002)(7416002)(2906002)(478600001)(316002)(6916009)(66476007)(66556008)(66946007)(86362001)(38100700002)(36756003)(38350700005); DIR:OUT; SFP:1102; X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: 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 X-MS-Exchange-CrossTenant-Network-Message-Id: e053100b-cd6d-4112-9ab9-08dbeb78fb56 X-MS-Exchange-CrossTenant-AuthSource: CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 22 Nov 2023 16:35:04.9473 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 0d5fd8bb-e8c2-4e0a-8dd5-2c264f7140fe X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: +biMBSpXDl5tndv2ajJdcAR2serR5d8GDK3PUhcEKXUD2em5+ldpdL6C5AE2rZvu0ooHhJmYrYwts6rV41Uok4cMbceB0AGPc5g6Xf5Jflc= X-MS-Exchange-Transport-CrossTenantHeadersStamped: CWLP265MB5329 X-OriginatorOrg: imgtec.com X-EXCLAIMER-MD-CONFIG: 15a78312-3e47-46eb-9010-2e54d84a9631 X-Proofpoint-ORIG-GUID: q4jC9LirBfnW3Erusc8LNU0SlP46kHm2 X-Proofpoint-GUID: q4jC9LirBfnW3Erusc8LNU0SlP46kHm2 Determining the start and range of the unmap stage of a remap op is a common piece of code currently implemented by multiple drivers. Add a helper for this. Changes since v7: - Renamed helper to drm_gpuva_op_remap_to_unmap_range() - Improved documentation Changes since v6: - Remove use of __always_inline Signed-off-by: Donald Robson Signed-off-by: Sarah Walker Reviewed-by: Danilo Krummrich --- include/drm/drm_gpuvm.h | 28 ++++++++++++++++++++++++++++ 1 file changed, 28 insertions(+) diff --git a/include/drm/drm_gpuvm.h b/include/drm/drm_gpuvm.h index 8ca10461d8ac..f94fec9a8517 100644 --- a/include/drm/drm_gpuvm.h +++ b/include/drm/drm_gpuvm.h @@ -1213,4 +1213,32 @@ void drm_gpuva_remap(struct drm_gpuva *prev, void drm_gpuva_unmap(struct drm_gpuva_op_unmap *op); +/** + * drm_gpuva_op_remap_to_unmap_range() - Helper to get the start and range of + * the unmap stage of a remap op. + * @op: Remap op. + * @start_addr: Output pointer for the start of the required unmap. + * @range: Output pointer for the length of the required unmap. + * + * The given start address and range will be set such that they represent the + * range of the address space that was previously covered by the mapping being + * re-mapped, but is now empty. + */ +static inline void +drm_gpuva_op_remap_to_unmap_range(const struct drm_gpuva_op_remap *op, + u64 *start_addr, u64 *range) +{ + const u64 va_start = op->prev ? + op->prev->va.addr + op->prev->va.range : + op->unmap->va->va.addr; + const u64 va_end = op->next ? + op->next->va.addr : + op->unmap->va->va.addr + op->unmap->va->va.range; + + if (start_addr) + *start_addr = va_start; + if (range) + *range = va_end - va_start; +} + #endif /* __DRM_GPUVM_H__ */ From patchwork Wed Nov 22 16:34:27 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Donald Robson X-Patchwork-Id: 746152 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=imgtec.com header.i=@imgtec.com header.b="Jzkzf2Hl"; dkim=pass (1024-bit key) header.d=IMGTecCRM.onmicrosoft.com header.i=@IMGTecCRM.onmicrosoft.com header.b="m8W2+gME" Received: from mx08-00376f01.pphosted.com (mx08-00376f01.pphosted.com [91.207.212.86]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 857E2112; Wed, 22 Nov 2023 08:35:53 -0800 (PST) Received: from pps.filterd (m0168888.ppops.net [127.0.0.1]) by mx08-00376f01.pphosted.com (8.17.1.24/8.17.1.24) with ESMTP id 3AMDkLtU020325; Wed, 22 Nov 2023 16:35:14 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=imgtec.com; h= from:to:cc:subject:date:message-id:in-reply-to:references :content-transfer-encoding:content-type:mime-version; s= dk201812; bh=DiX/EB6ywS+jdCDv7ngNPrXTWQBclFOriiTcDYGkLSE=; b=Jzk zf2HlowqctBBXU4P7SzC6BCzdPz6f4NoQEl4sQ1TwI7sGbDc/o/zx+noNXRw3wMw AWaohWC9gk8BR76mx5NgYmoXpiCriKrASplKhzbV+ldZ983Mp34wLdOnYEQ/nJGD MjPdNRJpB+KqTbenscb2Xr84NzGlEZrTXZ6qzICgDtneCB0uA1/Ryymqt3ewrTew bCH5tBnfpoGu4z4j45F5hRJLkxedM1VhRdZj46I/G17GYo3Eg52sMEO79obPnnMl 3bn6VXni1BLYEKw+fcVZcE7GWJocKqurvmQ75PeZFZ1IS7vjnLi+mDg/BG+SrKtT +wEDfbhOiQmvdS3DV/w== Received: from hhmail04.hh.imgtec.org ([217.156.249.195]) by mx08-00376f01.pphosted.com (PPS) with ESMTPS id 3ug99gsf32-3 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128 verify=NOT); Wed, 22 Nov 2023 16:35:13 +0000 (GMT) Received: from HHMAIL04.hh.imgtec.org (10.100.10.119) by HHMAIL04.hh.imgtec.org (10.100.10.119) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34; Wed, 22 Nov 2023 16:35:13 +0000 Received: from GBR01-LO4-obe.outbound.protection.outlook.com (104.47.85.104) by email.imgtec.com (10.100.10.121) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34 via Frontend Transport; Wed, 22 Nov 2023 16:35:13 +0000 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=B0+gTebC6lsCALFib4xbN5r6vcsXHJGMKX2cNdX0mgAAt5ous1ObXME+2ZRUmJLT3Gu00RtgqRTAuPein6VTVmdpi8BcNUZok8m4XrXUHe8Qz6wEtMFPqfx5aNA7KGjcJTtYo9Xyzi77SJ3rpVngmPJrJRUlAGd+o1skooRmiOG1GX8WrNBGJdIe1fvdn3ck0YXzdtu4bztL2LXxYgcF6tabtu+CUJQiz+bcbsWhxm1xdH20z2zmwaQXy1qsJ4596JFWOjdkPOAFEerQMcxAwlhRSfTMLyuuNVAeb/af1Y07NxyllgC9O2URPFLh/rxa4eGCB/nxS4QAb2VKaUNdrw== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=DiX/EB6ywS+jdCDv7ngNPrXTWQBclFOriiTcDYGkLSE=; b=emXgG/cWBVijYXUCsNTpVjqqzy1bW/Y9rfN0+Uu+zR7fbd3UzJ/irmn25s0Pm7jBjAnBptmuARA9VHAp4WlH3vD+MxMdjIKldgYdjtjJRGh9ZYTcyiJlwwNWSxGB48xHquYQzqVRiJ+S4egiyxVm8xwdBCwLm1OSfR5MdHsE5TfHoa2tzKS0IvJ2AN9Te55e/DHFXS466SySkBMGp419Nr6l2Eue94CP/ohfhTlZkrGvuLM8ohC+/ND8z5gN3XeqY8rovRXrBY3aM65X/2Gml6buRSaKuHro9ITzG3ziq3DHl4MI8p7MFyf/Y2QlGF1bhn+n1jMKsh8jaCpPfA0Qww== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=imgtec.com; dmarc=pass action=none header.from=imgtec.com; dkim=pass header.d=imgtec.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=IMGTecCRM.onmicrosoft.com; s=selector2-IMGTecCRM-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=DiX/EB6ywS+jdCDv7ngNPrXTWQBclFOriiTcDYGkLSE=; b=m8W2+gMEx+aUgGiGL1FhY2acCrznDkf9a0ZoCL/NZLp9uaXIPWBP2oBjIBGrNlbhP2ipa1zDVKJeVMyO8Fv4rfkJay+iGO6RO9xj2iXymGOtP91jPaABop0D4YtyzhNlnnjXvC0GA4vEHVEU0LyyI6RcDmLBPq9tTHQXH0hDMgc= Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) by CWLP265MB5329.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1c5::11) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.7025.19; Wed, 22 Nov 2023 16:35:11 +0000 Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34]) by CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34%3]) with mapi id 15.20.7025.017; Wed, 22 Nov 2023 16:35:11 +0000 From: Donald Robson To: CC: , , , , , , , , , , , , , , , , , , , , , , Sarah Walker Subject: [PATCH v9 06/20] drm/imagination: Get GPU resources Date: Wed, 22 Nov 2023 16:34:27 +0000 Message-Id: <579027bd5be4eb3218c9784050ded2326ecbc352.1700668843.git.donald.robson@imgtec.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: References: X-ClientProxiedBy: LO4P265CA0273.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:37a::10) To CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) Precedence: bulk X-Mailing-List: devicetree@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: CWLP265MB5770:EE_|CWLP265MB5329:EE_ X-MS-Office365-Filtering-Correlation-Id: 07321763-3471-475d-485b-08dbeb78ff2f X-MS-Exchange-SenderADCheck: 1 X-MS-Exchange-AntiSpam-Relay: 0 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-Forefront-Antispam-Report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:; IPV:NLI; SFV:NSPM; H:CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM; PTR:; CAT:NONE; SFS:(13230031)(396003)(366004)(136003)(39850400004)(376002)(346002)(230922051799003)(451199024)(1800799012)(186009)(64100799003)(26005)(83380400001)(52116002)(6506007)(6666004)(107886003)(2616005)(6512007)(6486002)(44832011)(5660300002)(4326008)(8676002)(41300700001)(8936002)(7416002)(2906002)(30864003)(478600001)(316002)(6916009)(66476007)(66556008)(66946007)(86362001)(38100700002)(36756003)(38350700005)(134885004); DIR:OUT; SFP:1102; X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: 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 X-MS-Exchange-CrossTenant-Network-Message-Id: 07321763-3471-475d-485b-08dbeb78ff2f X-MS-Exchange-CrossTenant-AuthSource: CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 22 Nov 2023 16:35:11.4227 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 0d5fd8bb-e8c2-4e0a-8dd5-2c264f7140fe X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: oVbw78zdNA2P1U/rNk1Tyq3nH+I8laDz9msGEtq1QHz5KdcgOrtXR23bfHeZj1hO68SBYpB/M93eHDRcK8DlFkM1PQr/TbJcXi9NYKTmyPg= X-MS-Exchange-Transport-CrossTenantHeadersStamped: CWLP265MB5329 X-OriginatorOrg: imgtec.com X-EXCLAIMER-MD-CONFIG: 15a78312-3e47-46eb-9010-2e54d84a9631 X-Proofpoint-GUID: dYQxZjJm2Pc-G7NK0EomfDRmyl_ekpTb X-Proofpoint-ORIG-GUID: dYQxZjJm2Pc-G7NK0EomfDRmyl_ekpTb From: Sarah Walker Acquire clock and register resources, and enable/map as appropriate. Changes since v8: - Corrected license identifiers Changes since v3: - Remove regulator resource (not used on supported platform) - Use devm helpers - Use devm_clk_get_optional() for optional clocks - Don't prepare clocks on resource acquisition - Drop pvr_device_clk_core_get_freq() helper - Drop pvr_device_reg_fini() - Drop NULLing of clocks in pvr_device_clk_init() - Use dev_err_probe() on clock acquisition failure - Remove PVR_CR_READ/WRITE helper macros - Improve documentation for GPU clocks - Remove regs resource (not used in this commit) Co-developed-by: Frank Binns Signed-off-by: Frank Binns Co-developed-by: Matt Coster Signed-off-by: Matt Coster Signed-off-by: Sarah Walker Signed-off-by: Donald Robson Reviewed-by: Maxime Ripard --- drivers/gpu/drm/imagination/Makefile | 1 + drivers/gpu/drm/imagination/pvr_device.c | 147 ++++++++++++++++++++++ drivers/gpu/drm/imagination/pvr_device.h | 152 +++++++++++++++++++++++ drivers/gpu/drm/imagination/pvr_drv.c | 18 ++- 4 files changed, 317 insertions(+), 1 deletion(-) create mode 100644 drivers/gpu/drm/imagination/pvr_device.c diff --git a/drivers/gpu/drm/imagination/Makefile b/drivers/gpu/drm/imagination/Makefile index f12a06ada9ec..d36007f2825c 100644 --- a/drivers/gpu/drm/imagination/Makefile +++ b/drivers/gpu/drm/imagination/Makefile @@ -4,6 +4,7 @@ subdir-ccflags-y := -I$(srctree)/$(src) powervr-y := \ + pvr_device.o \ pvr_drv.o \ obj-$(CONFIG_DRM_POWERVR) += powervr.o diff --git a/drivers/gpu/drm/imagination/pvr_device.c b/drivers/gpu/drm/imagination/pvr_device.c new file mode 100644 index 000000000000..abcdf733f57b --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_device.c @@ -0,0 +1,147 @@ +// SPDX-License-Identifier: GPL-2.0-only OR MIT +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#include "pvr_device.h" + +#include + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/** + * pvr_device_reg_init() - Initialize kernel access to a PowerVR device's + * control registers. + * @pvr_dev: Target PowerVR device. + * + * Sets struct pvr_device->regs. + * + * This method of mapping the device control registers into memory ensures that + * they are unmapped when the driver is detached (i.e. no explicit cleanup is + * required). + * + * Return: + * * 0 on success, or + * * Any error returned by devm_platform_ioremap_resource(). + */ +static int +pvr_device_reg_init(struct pvr_device *pvr_dev) +{ + struct drm_device *drm_dev = from_pvr_device(pvr_dev); + struct platform_device *plat_dev = to_platform_device(drm_dev->dev); + void __iomem *regs; + + pvr_dev->regs = NULL; + + regs = devm_platform_ioremap_resource(plat_dev, 0); + if (IS_ERR(regs)) + return dev_err_probe(drm_dev->dev, PTR_ERR(regs), + "failed to ioremap gpu registers\n"); + + pvr_dev->regs = regs; + + return 0; +} + +/** + * pvr_device_clk_init() - Initialize clocks required by a PowerVR device + * @pvr_dev: Target PowerVR device. + * + * Sets struct pvr_device->core_clk, struct pvr_device->sys_clk and + * struct pvr_device->mem_clk. + * + * Three clocks are required by the PowerVR device: core, sys and mem. On + * return, this function guarantees that the clocks are in one of the following + * states: + * + * * All successfully initialized, + * * Core errored, sys and mem uninitialized, + * * Core deinitialized, sys errored, mem uninitialized, or + * * Core and sys deinitialized, mem errored. + * + * Return: + * * 0 on success, + * * Any error returned by devm_clk_get(), or + * * Any error returned by devm_clk_get_optional(). + */ +static int pvr_device_clk_init(struct pvr_device *pvr_dev) +{ + struct drm_device *drm_dev = from_pvr_device(pvr_dev); + struct clk *core_clk; + struct clk *sys_clk; + struct clk *mem_clk; + + core_clk = devm_clk_get(drm_dev->dev, "core"); + if (IS_ERR(core_clk)) + return dev_err_probe(drm_dev->dev, PTR_ERR(core_clk), + "failed to get core clock\n"); + + sys_clk = devm_clk_get_optional(drm_dev->dev, "sys"); + if (IS_ERR(sys_clk)) + return dev_err_probe(drm_dev->dev, PTR_ERR(core_clk), + "failed to get sys clock\n"); + + mem_clk = devm_clk_get_optional(drm_dev->dev, "mem"); + if (IS_ERR(mem_clk)) + return dev_err_probe(drm_dev->dev, PTR_ERR(core_clk), + "failed to get mem clock\n"); + + pvr_dev->core_clk = core_clk; + pvr_dev->sys_clk = sys_clk; + pvr_dev->mem_clk = mem_clk; + + return 0; +} + +/** + * pvr_device_init() - Initialize a PowerVR device + * @pvr_dev: Target PowerVR device. + * + * If this function returns successfully, the device will have been fully + * initialized. Otherwise, any parts of the device initialized before an error + * occurs will be de-initialized before returning. + * + * NOTE: The initialization steps currently taken are the bare minimum required + * to read from the control registers. The device is unlikely to function + * until further initialization steps are added. [This note should be + * removed when that happens.] + * + * Return: + * * 0 on success, + * * Any error returned by pvr_device_reg_init(), + * * Any error returned by pvr_device_clk_init(), or + * * Any error returned by pvr_device_gpu_init(). + */ +int +pvr_device_init(struct pvr_device *pvr_dev) +{ + int err; + + /* Enable and initialize clocks required for the device to operate. */ + err = pvr_device_clk_init(pvr_dev); + if (err) + return err; + + /* Map the control registers into memory. */ + return pvr_device_reg_init(pvr_dev); +} + +/** + * pvr_device_fini() - Deinitialize a PowerVR device + * @pvr_dev: Target PowerVR device. + */ +void +pvr_device_fini(struct pvr_device *pvr_dev) +{ + /* + * Deinitialization stages are performed in reverse order compared to + * the initialization stages in pvr_device_init(). + */ +} diff --git a/drivers/gpu/drm/imagination/pvr_device.h b/drivers/gpu/drm/imagination/pvr_device.h index d3629164a629..d9d2804f4454 100644 --- a/drivers/gpu/drm/imagination/pvr_device.h +++ b/drivers/gpu/drm/imagination/pvr_device.h @@ -11,9 +11,22 @@ #include #include #include +#include +#include #include +#include +#include +#include #include #include +#include +#include + +/* Forward declaration from . */ +struct clk; + +/* Forward declaration from . */ +struct firmware; /** * struct pvr_device - powervr-specific wrapper for &struct drm_device @@ -26,6 +39,37 @@ struct pvr_device { * from_pvr_device(). */ struct drm_device base; + + /** + * @regs: Device control registers. + * + * These are mapped into memory when the device is initialized; that + * location is where this pointer points. + */ + void __iomem *regs; + + /** + * @core_clk: General core clock. + * + * This is the primary clock used by the entire GPU core. + */ + struct clk *core_clk; + + /** + * @sys_clk: Optional system bus clock. + * + * This may be used on some platforms to provide an independent clock to the SoC Interface + * (SOCIF). If present, this needs to be enabled/disabled together with @core_clk. + */ + struct clk *sys_clk; + + /** + * @mem_clk: Optional memory clock. + * + * This may be used on some platforms to provide an independent clock to the Memory + * Interface (MEMIF). If present, this needs to be enabled/disabled together with @core_clk. + */ + struct clk *mem_clk; }; /** @@ -56,6 +100,114 @@ struct pvr_file { #define to_pvr_file(file) ((file)->driver_priv) +int pvr_device_init(struct pvr_device *pvr_dev); +void pvr_device_fini(struct pvr_device *pvr_dev); + +/** + * PVR_CR_FIELD_GET() - Extract a single field from a PowerVR control register + * @val: Value of the target register. + * @field: Field specifier, as defined in "pvr_rogue_cr_defs.h". + * + * Return: The extracted field. + */ +#define PVR_CR_FIELD_GET(val, field) FIELD_GET(~ROGUE_CR_##field##_CLRMSK, val) + +/** + * pvr_cr_read32() - Read a 32-bit register from a PowerVR device + * @pvr_dev: Target PowerVR device. + * @reg: Target register. + * + * Return: The value of the requested register. + */ +static __always_inline u32 +pvr_cr_read32(struct pvr_device *pvr_dev, u32 reg) +{ + return ioread32(pvr_dev->regs + reg); +} + +/** + * pvr_cr_read64() - Read a 64-bit register from a PowerVR device + * @pvr_dev: Target PowerVR device. + * @reg: Target register. + * + * Return: The value of the requested register. + */ +static __always_inline u64 +pvr_cr_read64(struct pvr_device *pvr_dev, u32 reg) +{ + return ioread64(pvr_dev->regs + reg); +} + +/** + * pvr_cr_write32() - Write to a 32-bit register in a PowerVR device + * @pvr_dev: Target PowerVR device. + * @reg: Target register. + * @val: Value to write. + */ +static __always_inline void +pvr_cr_write32(struct pvr_device *pvr_dev, u32 reg, u32 val) +{ + iowrite32(val, pvr_dev->regs + reg); +} + +/** + * pvr_cr_write64() - Write to a 64-bit register in a PowerVR device + * @pvr_dev: Target PowerVR device. + * @reg: Target register. + * @val: Value to write. + */ +static __always_inline void +pvr_cr_write64(struct pvr_device *pvr_dev, u32 reg, u64 val) +{ + iowrite64(val, pvr_dev->regs + reg); +} + +/** + * pvr_cr_poll_reg32() - Wait for a 32-bit register to match a given value by + * polling + * @pvr_dev: Target PowerVR device. + * @reg_addr: Address of register. + * @reg_value: Expected register value (after masking). + * @reg_mask: Mask of bits valid for comparison with @reg_value. + * @timeout_usec: Timeout length, in us. + * + * Returns: + * * 0 on success, or + * * -%ETIMEDOUT on timeout. + */ +static __always_inline int +pvr_cr_poll_reg32(struct pvr_device *pvr_dev, u32 reg_addr, u32 reg_value, + u32 reg_mask, u64 timeout_usec) +{ + u32 value; + + return readl_poll_timeout(pvr_dev->regs + reg_addr, value, + (value & reg_mask) == reg_value, 0, timeout_usec); +} + +/** + * pvr_cr_poll_reg64() - Wait for a 64-bit register to match a given value by + * polling + * @pvr_dev: Target PowerVR device. + * @reg_addr: Address of register. + * @reg_value: Expected register value (after masking). + * @reg_mask: Mask of bits valid for comparison with @reg_value. + * @timeout_usec: Timeout length, in us. + * + * Returns: + * * 0 on success, or + * * -%ETIMEDOUT on timeout. + */ +static __always_inline int +pvr_cr_poll_reg64(struct pvr_device *pvr_dev, u32 reg_addr, u64 reg_value, + u64 reg_mask, u64 timeout_usec) +{ + u64 value; + + return readq_poll_timeout(pvr_dev->regs + reg_addr, value, + (value & reg_mask) == reg_value, 0, timeout_usec); +} + /** * DOC: IOCTL validation helpers * diff --git a/drivers/gpu/drm/imagination/pvr_drv.c b/drivers/gpu/drm/imagination/pvr_drv.c index 3d4f2ba2c015..597188b31f96 100644 --- a/drivers/gpu/drm/imagination/pvr_drv.c +++ b/drivers/gpu/drm/imagination/pvr_drv.c @@ -464,6 +464,7 @@ pvr_probe(struct platform_device *plat_dev) { struct pvr_device *pvr_dev; struct drm_device *drm_dev; + int err; pvr_dev = devm_drm_dev_alloc(&plat_dev->dev, &pvr_drm_driver, struct pvr_device, base); @@ -474,14 +475,29 @@ pvr_probe(struct platform_device *plat_dev) platform_set_drvdata(plat_dev, drm_dev); - return drm_dev_register(drm_dev, 0); + err = pvr_device_init(pvr_dev); + if (err) + return err; + + err = drm_dev_register(drm_dev, 0); + if (err) + goto err_device_fini; + + return 0; + +err_device_fini: + pvr_device_fini(pvr_dev); + + return err; } static int pvr_remove(struct platform_device *plat_dev) { struct drm_device *drm_dev = platform_get_drvdata(plat_dev); + struct pvr_device *pvr_dev = to_pvr_device(drm_dev); + pvr_device_fini(pvr_dev); drm_dev_unplug(drm_dev); return 0; From patchwork Wed Nov 22 16:34:29 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Donald Robson X-Patchwork-Id: 746151 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=imgtec.com header.i=@imgtec.com header.b="DFA3YzHs"; dkim=pass (1024-bit key) header.d=IMGTecCRM.onmicrosoft.com header.i=@IMGTecCRM.onmicrosoft.com header.b="rvf3Kksu" Received: from mx07-00376f01.pphosted.com (mx07-00376f01.pphosted.com [185.132.180.163]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 416C01B2; Wed, 22 Nov 2023 08:35:51 -0800 (PST) Received: from pps.filterd (m0168889.ppops.net [127.0.0.1]) by mx07-00376f01.pphosted.com (8.17.1.24/8.17.1.24) with ESMTP id 3AM7ePpv020320; Wed, 22 Nov 2023 16:35:20 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=imgtec.com; h= from:to:cc:subject:date:message-id:in-reply-to:references :content-transfer-encoding:content-type:mime-version; s= dk201812; bh=ad0JLZXXdSkZFtFdW/yU414pDjxZHj9TqlE9oP6lZKI=; b=DFA 3YzHsk2bSwu0oIaQy4uBKq5+PxTFhiR4PvbfVW8EiIR2q4vLigQR2HK8El+sRnX1 K1eR/Mvvnpp9yPN23aaqh2Ofrsr2ANcq16q1Xl3TmV6PWXdrZ23/8CXBBkgpWXVM ypxu2OEP9skp20v/4RmsR401xp6X8W49y9Z1FVvpojueYrUPyQOzL4Zzks0tzFxQ 2m54S4dNhZTE6bcYkoQhWPIpJlLOWQ3uk/MGGzBfGgrIqO7ycjmHZ44G8nnHGrxo pWCytxnUv8aLt5V5chXsXJ/U+lmMg92/9ZwmM1UEy5UwaHlaEbZfnkdy1Tqi7GqT gnO8ethIT47LdpY4Hmw== Received: from hhmail04.hh.imgtec.org ([217.156.249.195]) by mx07-00376f01.pphosted.com (PPS) with ESMTPS id 3ug99g9m26-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128 verify=NOT); Wed, 22 Nov 2023 16:35:20 +0000 (GMT) Received: from HHMAIL04.hh.imgtec.org (10.100.10.119) by HHMAIL04.hh.imgtec.org (10.100.10.119) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34; Wed, 22 Nov 2023 16:35:19 +0000 Received: from GBR01-LO4-obe.outbound.protection.outlook.com (104.47.85.104) by email.imgtec.com (10.100.10.121) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34 via Frontend Transport; Wed, 22 Nov 2023 16:35:19 +0000 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=Tbkh0FIvgs1sWLu2wyXkML7umK2lWlK4M6+h77PuCOtsej2MWYxkbaDTuvptXkPnekJCFY1kZ63RIc6Q2i5yVglEMfTGoLbPlN5kiDNU9CbU6ip723QPUci497Rk+8+TEoVJMBGBrRbGiwdkcerdKldLXglp8CnhUx7MBIwnGCIKQtoVuDejt6gfcIJaRSTw+OhuU3NlXef3IuSSPxLRIj+fwS6+spUBBqAsmRNHxpCE5KnZ+PIF8qfYK2n0HrKoiC7UjTNmR6s9MEGoiQHqcU8vazglw9L2KSEYNvg6WTnLYcNKSD2gKTGJ2viUkgjuLaePsIEYrUAAtc9u1jX+vg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=ad0JLZXXdSkZFtFdW/yU414pDjxZHj9TqlE9oP6lZKI=; b=R0Utn0S+ELsu5kIRSc/c3USM6GDVMQeoreEqUekwpHSVu+05aOA6Dzr9EMnOyuTosT7siRyBDXaKTHm4MxMBgSabFcCYPm5xJ4KnhHxqTDPopjF08yw2dLjd/1+ex7j/8V8HEsJ9YfQ7ODVUr9mjRdTT5UsIAxIuUILpvPq16+cD4eUGISd4xZmhcoswXiaO0EGc1SDjNw4jTkE9v8yzc1sxF0dWMn1GZSeh2Zyur2buar70k2p935nEY1RHkUBreZ6apmjyilOlAhyzkWRRSuFYmqxeG3NwOaPEmBS8+p2YVSNGsaDjHypyP3ai/JduQce3ON+trVBuX1WZlKDs8A== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=imgtec.com; dmarc=pass action=none header.from=imgtec.com; dkim=pass header.d=imgtec.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=IMGTecCRM.onmicrosoft.com; s=selector2-IMGTecCRM-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=ad0JLZXXdSkZFtFdW/yU414pDjxZHj9TqlE9oP6lZKI=; b=rvf3KksuGXO/AF+bu8I2cSzV0GMvzh6GHr9ol4dRc0vH3klQevFRXvP2UKNKKlepQIv8qPV4JDaic76U7FFlQZx9a3wq/F+kSQ4VGatzGR/aIPNx3g4UYn4Adrf+3SzHY6toaqxYOpj10ncexn2izlpHHRZy3GCeljxVKxSvV3E= Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) by LO6P265MB6459.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:2df::12) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.7025.18; Wed, 22 Nov 2023 16:35:15 +0000 Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34]) by CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34%3]) with mapi id 15.20.7025.017; Wed, 22 Nov 2023 16:35:15 +0000 From: Donald Robson To: CC: , , , , , , , , , , , , , , , , , , , , , , Sarah Walker Subject: [PATCH v9 08/20] drm/imagination: Add firmware and MMU related headers Date: Wed, 22 Nov 2023 16:34:29 +0000 Message-Id: <23ee233dfbe6f2239328f8201fd6d8c1017cea58.1700668843.git.donald.robson@imgtec.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: References: X-ClientProxiedBy: LO4P265CA0273.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:37a::10) To CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) Precedence: bulk X-Mailing-List: devicetree@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: CWLP265MB5770:EE_|LO6P265MB6459:EE_ X-MS-Office365-Filtering-Correlation-Id: 9138dfb0-10c9-4886-6a0d-08dbeb790177 X-MS-Exchange-SenderADCheck: 1 X-MS-Exchange-AntiSpam-Relay: 0 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-Forefront-Antispam-Report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:; IPV:NLI; SFV:NSPM; H:CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM; PTR:; CAT:NONE; SFS:(13230031)(39850400004)(346002)(396003)(366004)(136003)(376002)(230922051799003)(64100799003)(186009)(451199024)(1800799012)(30864003)(36756003)(7416002)(41300700001)(5660300002)(2906002)(86362001)(38350700005)(26005)(6666004)(478600001)(6506007)(2616005)(6486002)(107886003)(6512007)(52116002)(8936002)(8676002)(4326008)(44832011)(66946007)(316002)(6916009)(66556008)(66476007)(83380400001)(38100700002)(579004); DIR:OUT; SFP:1102; X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: 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 X-MS-Exchange-CrossTenant-Network-Message-Id: 9138dfb0-10c9-4886-6a0d-08dbeb790177 X-MS-Exchange-CrossTenant-AuthSource: CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 22 Nov 2023 16:35:15.2597 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 0d5fd8bb-e8c2-4e0a-8dd5-2c264f7140fe X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: knQylo3iF5caoSR8bk5KfCJCM/nsXDobklUWhB2CG/XLiTI7/2kbDkxQZa5mWuabpOEjsQd1ozzv5xpD983588ZgoRZT8iTGbC1RygKsK/w= X-MS-Exchange-Transport-CrossTenantHeadersStamped: LO6P265MB6459 X-OriginatorOrg: imgtec.com X-EXCLAIMER-MD-CONFIG: 15a78312-3e47-46eb-9010-2e54d84a9631 X-Proofpoint-ORIG-GUID: CLYBKlbWiJTcdeM0sR6MEmw_ps7YyeYb X-Proofpoint-GUID: CLYBKlbWiJTcdeM0sR6MEmw_ps7YyeYb From: Sarah Walker Changes since v8: - Corrected license identifiers Changes since v5: - Split up header commit due to size Signed-off-by: Sarah Walker Signed-off-by: Donald Robson Acked-by: Maxime Ripard --- .../drm/imagination/pvr_rogue_heap_config.h | 113 ++++++ drivers/gpu/drm/imagination/pvr_rogue_meta.h | 356 ++++++++++++++++++ drivers/gpu/drm/imagination/pvr_rogue_mips.h | 335 ++++++++++++++++ .../drm/imagination/pvr_rogue_mips_check.h | 58 +++ .../gpu/drm/imagination/pvr_rogue_mmu_defs.h | 136 +++++++ 5 files changed, 998 insertions(+) create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_heap_config.h create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_meta.h create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_mips.h create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_mips_check.h create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_mmu_defs.h diff --git a/drivers/gpu/drm/imagination/pvr_rogue_heap_config.h b/drivers/gpu/drm/imagination/pvr_rogue_heap_config.h new file mode 100644 index 000000000000..684766006703 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_heap_config.h @@ -0,0 +1,113 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_ROGUE_HEAP_CONFIG_H +#define PVR_ROGUE_HEAP_CONFIG_H + +#include + +/* + * ROGUE Device Virtual Address Space Definitions + * + * This file defines the ROGUE virtual address heaps that are used in + * application memory contexts. It also shows where the Firmware memory heap + * fits into this, but the firmware heap is only ever created in the + * kernel driver and never exposed to userspace. + * + * ROGUE_PDSCODEDATA_HEAP_BASE and ROGUE_USCCODE_HEAP_BASE will be programmed, + * on a global basis, into ROGUE_CR_PDS_EXEC_BASE and ROGUE_CR_USC_CODE_BASE_* + * respectively. Therefore if client drivers use multiple configs they must + * still be consistent with their definitions for these heaps. + * + * Base addresses have to be a multiple of 4MiB. + * Heaps must not start at 0x0000000000, as this is reserved for internal + * use within the driver. + * Range comments, those starting in column 0 below are a section heading of + * sorts and are above the heaps in that range. Often this is the reserved + * size of the heap within the range. + */ + +/* 0x00_0000_0000 ************************************************************/ + +/* 0x00_0000_0000 - 0x00_0040_0000 */ +/* 0 MiB to 4 MiB, size of 4 MiB : RESERVED */ + +/* 0x00_0040_0000 - 0x7F_FFC0_0000 **/ +/* 4 MiB to 512 GiB, size of 512 GiB less 4 MiB : RESERVED **/ + +/* 0x80_0000_0000 ************************************************************/ + +/* 0x80_0000_0000 - 0x9F_FFFF_FFFF **/ +/* 512 GiB to 640 GiB, size of 128 GiB : GENERAL_HEAP **/ +#define ROGUE_GENERAL_HEAP_BASE 0x8000000000ull +#define ROGUE_GENERAL_HEAP_SIZE SZ_128G + +/* 0xA0_0000_0000 - 0xAF_FFFF_FFFF */ +/* 640 GiB to 704 GiB, size of 64 GiB : FREE */ + +/* B0_0000_0000 - 0xB7_FFFF_FFFF */ +/* 704 GiB to 736 GiB, size of 32 GiB : FREE */ + +/* 0xB8_0000_0000 - 0xBF_FFFF_FFFF */ +/* 736 GiB to 768 GiB, size of 32 GiB : RESERVED */ + +/* 0xC0_0000_0000 ************************************************************/ + +/* 0xC0_0000_0000 - 0xD9_FFFF_FFFF */ +/* 768 GiB to 872 GiB, size of 104 GiB : FREE */ + +/* 0xDA_0000_0000 - 0xDA_FFFF_FFFF */ +/* 872 GiB to 876 GiB, size of 4 GiB : PDSCODEDATA_HEAP */ +#define ROGUE_PDSCODEDATA_HEAP_BASE 0xDA00000000ull +#define ROGUE_PDSCODEDATA_HEAP_SIZE SZ_4G + +/* 0xDB_0000_0000 - 0xDB_FFFF_FFFF */ +/* 876 GiB to 880 GiB, size of 256 MiB (reserved 4GiB) : BRN **/ +/* + * The BRN63142 quirk workaround requires Region Header memory to be at the top + * of a 16GiB aligned range. This is so when masked with 0x03FFFFFFFF the + * address will avoid aliasing PB addresses. Start at 879.75GiB. Size of 256MiB. + */ +#define ROGUE_RGNHDR_HEAP_BASE 0xDBF0000000ull +#define ROGUE_RGNHDR_HEAP_SIZE SZ_256M + +/* 0xDC_0000_0000 - 0xDF_FFFF_FFFF */ +/* 880 GiB to 896 GiB, size of 16 GiB : FREE */ + +/* 0xE0_0000_0000 - 0xE0_FFFF_FFFF */ +/* 896 GiB to 900 GiB, size of 4 GiB : USCCODE_HEAP */ +#define ROGUE_USCCODE_HEAP_BASE 0xE000000000ull +#define ROGUE_USCCODE_HEAP_SIZE SZ_4G + +/* 0xE1_0000_0000 - 0xE1_BFFF_FFFF */ +/* 900 GiB to 903 GiB, size of 3 GiB : RESERVED */ + +/* 0xE1_C000_000 - 0xE1_FFFF_FFFF */ +/* 903 GiB to 904 GiB, reserved 1 GiB, : FIRMWARE_HEAP */ +#define ROGUE_FW_HEAP_BASE 0xE1C0000000ull + +/* 0xE2_0000_0000 - 0xE3_FFFF_FFFF */ +/* 904 GiB to 912 GiB, size of 8 GiB : FREE */ + +/* 0xE4_0000_0000 - 0xE7_FFFF_FFFF */ +/* 912 GiB to 968 GiB, size of 16 GiB : TRANSFER_FRAG */ +#define ROGUE_TRANSFER_FRAG_HEAP_BASE 0xE400000000ull +#define ROGUE_TRANSFER_FRAG_HEAP_SIZE SZ_16G + +/* 0xE8_0000_0000 - 0xF1_FFFF_FFFF */ +/* 928 GiB to 968 GiB, size of 40 GiB : RESERVED */ + +/* 0xF2_0000_0000 - 0xF2_001F_FFFF **/ +/* 968 GiB to 969 GiB, size of 2 MiB : VISTEST_HEAP */ +#define ROGUE_VISTEST_HEAP_BASE 0xF200000000ull +#define ROGUE_VISTEST_HEAP_SIZE SZ_2M + +/* 0xF2_4000_0000 - 0xF2_FFFF_FFFF */ +/* 969 GiB to 972 GiB, size of 3 GiB : FREE */ + +/* 0xF3_0000_0000 - 0xFF_FFFF_FFFF */ +/* 972 GiB to 1024 GiB, size of 52 GiB : FREE */ + +/* 0xFF_FFFF_FFFF ************************************************************/ + +#endif /* PVR_ROGUE_HEAP_CONFIG_H */ diff --git a/drivers/gpu/drm/imagination/pvr_rogue_meta.h b/drivers/gpu/drm/imagination/pvr_rogue_meta.h new file mode 100644 index 000000000000..3020e6582daa --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_meta.h @@ -0,0 +1,356 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_ROGUE_META_H +#define PVR_ROGUE_META_H + +/***** The META HW register definitions in the file are updated manually *****/ + +#include +#include + +/* + ****************************************************************************** + * META registers and MACROS + ***************************************************************************** + */ +#define META_CR_CTRLREG_BASE(t) (0x04800000U + (0x1000U * (t))) + +#define META_CR_TXPRIVEXT (0x048000E8) +#define META_CR_TXPRIVEXT_MINIM_EN BIT(7) + +#define META_CR_SYSC_JTAG_THREAD (0x04830030) +#define META_CR_SYSC_JTAG_THREAD_PRIV_EN (0x00000004) + +#define META_CR_PERF_COUNT0 (0x0480FFE0) +#define META_CR_PERF_COUNT1 (0x0480FFE8) +#define META_CR_PERF_COUNT_CTRL_SHIFT (28) +#define META_CR_PERF_COUNT_CTRL_MASK (0xF0000000) +#define META_CR_PERF_COUNT_CTRL_DCACHEHITS (8 << META_CR_PERF_COUNT_CTRL_SHIFT) +#define META_CR_PERF_COUNT_CTRL_ICACHEHITS (9 << META_CR_PERF_COUNT_CTRL_SHIFT) +#define META_CR_PERF_COUNT_CTRL_ICACHEMISS \ + (0xA << META_CR_PERF_COUNT_CTRL_SHIFT) +#define META_CR_PERF_COUNT_CTRL_ICORE (0xD << META_CR_PERF_COUNT_CTRL_SHIFT) +#define META_CR_PERF_COUNT_THR_SHIFT (24) +#define META_CR_PERF_COUNT_THR_MASK (0x0F000000) +#define META_CR_PERF_COUNT_THR_0 (0x1 << META_CR_PERF_COUNT_THR_SHIFT) +#define META_CR_PERF_COUNT_THR_1 (0x2 << META_CR_PERF_COUNT_THR_1) + +#define META_CR_TxVECINT_BHALT (0x04820500) +#define META_CR_PERF_ICORE0 (0x0480FFD0) +#define META_CR_PERF_ICORE1 (0x0480FFD8) +#define META_CR_PERF_ICORE_DCACHEMISS (0x8) + +#define META_CR_PERF_COUNT(ctrl, thr) \ + ((META_CR_PERF_COUNT_CTRL_##ctrl << META_CR_PERF_COUNT_CTRL_SHIFT) | \ + ((thr) << META_CR_PERF_COUNT_THR_SHIFT)) + +#define META_CR_TXUXXRXDT_OFFSET (META_CR_CTRLREG_BASE(0U) + 0x0000FFF0U) +#define META_CR_TXUXXRXRQ_OFFSET (META_CR_CTRLREG_BASE(0U) + 0x0000FFF8U) + +/* Poll for done. */ +#define META_CR_TXUXXRXRQ_DREADY_BIT (0x80000000U) +/* Set for read. */ +#define META_CR_TXUXXRXRQ_RDnWR_BIT (0x00010000U) +#define META_CR_TXUXXRXRQ_TX_S (12) +#define META_CR_TXUXXRXRQ_RX_S (4) +#define META_CR_TXUXXRXRQ_UXX_S (0) + +/* Internal ctrl regs. */ +#define META_CR_TXUIN_ID (0x0) +/* Data unit regs. */ +#define META_CR_TXUD0_ID (0x1) +/* Data unit regs. */ +#define META_CR_TXUD1_ID (0x2) +/* Address unit regs. */ +#define META_CR_TXUA0_ID (0x3) +/* Address unit regs. */ +#define META_CR_TXUA1_ID (0x4) +/* PC registers. */ +#define META_CR_TXUPC_ID (0x5) + +/* Macros to calculate register access values. */ +#define META_CR_CORE_REG(thr, reg_num, unit) \ + (((u32)(thr) << META_CR_TXUXXRXRQ_TX_S) | \ + ((u32)(reg_num) << META_CR_TXUXXRXRQ_RX_S) | \ + ((u32)(unit) << META_CR_TXUXXRXRQ_UXX_S)) + +#define META_CR_THR0_PC META_CR_CORE_REG(0, 0, META_CR_TXUPC_ID) +#define META_CR_THR0_PCX META_CR_CORE_REG(0, 1, META_CR_TXUPC_ID) +#define META_CR_THR0_SP META_CR_CORE_REG(0, 0, META_CR_TXUA0_ID) + +#define META_CR_THR1_PC META_CR_CORE_REG(1, 0, META_CR_TXUPC_ID) +#define META_CR_THR1_PCX META_CR_CORE_REG(1, 1, META_CR_TXUPC_ID) +#define META_CR_THR1_SP META_CR_CORE_REG(1, 0, META_CR_TXUA0_ID) + +#define SP_ACCESS(thread) META_CR_CORE_REG(thread, 0, META_CR_TXUA0_ID) +#define PC_ACCESS(thread) META_CR_CORE_REG(thread, 0, META_CR_TXUPC_ID) + +#define META_CR_COREREG_ENABLE (0x0000000U) +#define META_CR_COREREG_STATUS (0x0000010U) +#define META_CR_COREREG_DEFR (0x00000A0U) +#define META_CR_COREREG_PRIVEXT (0x00000E8U) + +#define META_CR_T0ENABLE_OFFSET \ + (META_CR_CTRLREG_BASE(0U) + META_CR_COREREG_ENABLE) +#define META_CR_T0STATUS_OFFSET \ + (META_CR_CTRLREG_BASE(0U) + META_CR_COREREG_STATUS) +#define META_CR_T0DEFR_OFFSET (META_CR_CTRLREG_BASE(0U) + META_CR_COREREG_DEFR) +#define META_CR_T0PRIVEXT_OFFSET \ + (META_CR_CTRLREG_BASE(0U) + META_CR_COREREG_PRIVEXT) + +#define META_CR_T1ENABLE_OFFSET \ + (META_CR_CTRLREG_BASE(1U) + META_CR_COREREG_ENABLE) +#define META_CR_T1STATUS_OFFSET \ + (META_CR_CTRLREG_BASE(1U) + META_CR_COREREG_STATUS) +#define META_CR_T1DEFR_OFFSET (META_CR_CTRLREG_BASE(1U) + META_CR_COREREG_DEFR) +#define META_CR_T1PRIVEXT_OFFSET \ + (META_CR_CTRLREG_BASE(1U) + META_CR_COREREG_PRIVEXT) + +#define META_CR_TXENABLE_ENABLE_BIT (0x00000001U) /* Set if running */ +#define META_CR_TXSTATUS_PRIV (0x00020000U) +#define META_CR_TXPRIVEXT_MINIM (0x00000080U) + +#define META_MEM_GLOBAL_RANGE_BIT (0x80000000U) + +#define META_CR_TXCLKCTRL (0x048000B0) +#define META_CR_TXCLKCTRL_ALL_ON (0x55111111) +#define META_CR_TXCLKCTRL_ALL_AUTO (0xAA222222) + +#define META_CR_MMCU_LOCAL_EBCTRL (0x04830600) +#define META_CR_MMCU_LOCAL_EBCTRL_ICWIN (0x3 << 14) +#define META_CR_MMCU_LOCAL_EBCTRL_DCWIN (0x3 << 6) +#define META_CR_SYSC_DCPART(n) (0x04830200 + (n) * 0x8) +#define META_CR_SYSC_DCPARTX_CACHED_WRITE_ENABLE (0x1 << 31) +#define META_CR_SYSC_ICPART(n) (0x04830220 + (n) * 0x8) +#define META_CR_SYSC_XCPARTX_LOCAL_ADDR_OFFSET_TOP_HALF (0x8 << 16) +#define META_CR_SYSC_XCPARTX_LOCAL_ADDR_FULL_CACHE (0xF) +#define META_CR_SYSC_XCPARTX_LOCAL_ADDR_HALF_CACHE (0x7) +#define META_CR_MMCU_DCACHE_CTRL (0x04830018) +#define META_CR_MMCU_ICACHE_CTRL (0x04830020) +#define META_CR_MMCU_XCACHE_CTRL_CACHE_HITS_EN (0x1) + +/* + ****************************************************************************** + * META LDR Format + ****************************************************************************** + */ +/* Block header structure. */ +struct rogue_meta_ldr_block_hdr { + u32 dev_id; + u32 sl_code; + u32 sl_data; + u16 pc_ctrl; + u16 crc; +}; + +/* High level data stream block structure. */ +struct rogue_meta_ldr_l1_data_blk { + u16 cmd; + u16 length; + u32 next; + u32 cmd_data[4]; +}; + +/* High level data stream block structure. */ +struct rogue_meta_ldr_l2_data_blk { + u16 tag; + u16 length; + u32 block_data[4]; +}; + +/* Config command structure. */ +struct rogue_meta_ldr_cfg_blk { + u32 type; + u32 block_data[4]; +}; + +/* Block type definitions */ +#define ROGUE_META_LDR_COMMENT_TYPE_MASK (0x0010U) +#define ROGUE_META_LDR_BLK_IS_COMMENT(x) (((x) & ROGUE_META_LDR_COMMENT_TYPE_MASK) != 0U) + +/* + * Command definitions + * Value Name Description + * 0 LoadMem Load memory with binary data. + * 1 LoadCore Load a set of core registers. + * 2 LoadMMReg Load a set of memory mapped registers. + * 3 StartThreads Set each thread PC and SP, then enable threads. + * 4 ZeroMem Zeros a memory region. + * 5 Config Perform a configuration command. + */ +#define ROGUE_META_LDR_CMD_MASK (0x000FU) + +#define ROGUE_META_LDR_CMD_LOADMEM (0x0000U) +#define ROGUE_META_LDR_CMD_LOADCORE (0x0001U) +#define ROGUE_META_LDR_CMD_LOADMMREG (0x0002U) +#define ROGUE_META_LDR_CMD_START_THREADS (0x0003U) +#define ROGUE_META_LDR_CMD_ZEROMEM (0x0004U) +#define ROGUE_META_LDR_CMD_CONFIG (0x0005U) + +/* + * Config Command definitions + * Value Name Description + * 0 Pause Pause for x times 100 instructions + * 1 Read Read a value from register - No value return needed. + * Utilises effects of issuing reads to certain registers + * 2 Write Write to mem location + * 3 MemSet Set mem to value + * 4 MemCheck check mem for specific value. + */ +#define ROGUE_META_LDR_CFG_PAUSE (0x0000) +#define ROGUE_META_LDR_CFG_READ (0x0001) +#define ROGUE_META_LDR_CFG_WRITE (0x0002) +#define ROGUE_META_LDR_CFG_MEMSET (0x0003) +#define ROGUE_META_LDR_CFG_MEMCHECK (0x0004) + +/* + ****************************************************************************** + * ROGUE FW segmented MMU definitions + ****************************************************************************** + */ +/* All threads can access the segment. */ +#define ROGUE_FW_SEGMMU_ALLTHRS (0xf << 8U) +/* Writable. */ +#define ROGUE_FW_SEGMMU_WRITEABLE (0x1U << 1U) +/* All threads can access and writable. */ +#define ROGUE_FW_SEGMMU_ALLTHRS_WRITEABLE \ + (ROGUE_FW_SEGMMU_ALLTHRS | ROGUE_FW_SEGMMU_WRITEABLE) + +/* Direct map region 10 used for mapping GPU memory - max 8MB. */ +#define ROGUE_FW_SEGMMU_DMAP_GPU_ID (10U) +#define ROGUE_FW_SEGMMU_DMAP_GPU_ADDR_START (0x07000000U) +#define ROGUE_FW_SEGMMU_DMAP_GPU_MAX_SIZE (0x00800000U) + +/* Segment IDs. */ +#define ROGUE_FW_SEGMMU_DATA_ID (1U) +#define ROGUE_FW_SEGMMU_BOOTLDR_ID (2U) +#define ROGUE_FW_SEGMMU_TEXT_ID (ROGUE_FW_SEGMMU_BOOTLDR_ID) + +/* + * SLC caching strategy in S7 and volcanic is emitted through the segment MMU. + * All the segments configured through the macro ROGUE_FW_SEGMMU_OUTADDR_TOP are + * CACHED in the SLC. + * The interface has been kept the same to simplify the code changes. + * The bifdm argument is ignored (no longer relevant) in S7 and volcanic. + */ +#define ROGUE_FW_SEGMMU_OUTADDR_TOP_VIVT_SLC(pers, slc_policy, mmu_ctx) \ + ((((u64)((pers) & 0x3)) << 52) | (((u64)((mmu_ctx) & 0xFF)) << 44) | \ + (((u64)((slc_policy) & 0x1)) << 40)) +#define ROGUE_FW_SEGMMU_OUTADDR_TOP_VIVT_SLC_CACHED(mmu_ctx) \ + ROGUE_FW_SEGMMU_OUTADDR_TOP_VIVT_SLC(0x3, 0x0, mmu_ctx) +#define ROGUE_FW_SEGMMU_OUTADDR_TOP_VIVT_SLC_UNCACHED(mmu_ctx) \ + ROGUE_FW_SEGMMU_OUTADDR_TOP_VIVT_SLC(0x0, 0x1, mmu_ctx) + +/* + * To configure the Page Catalog and BIF-DM fed into the BIF for Garten + * accesses through this segment. + */ +#define ROGUE_FW_SEGMMU_OUTADDR_TOP_SLC(pc, bifdm) \ + (((u64)((u64)(pc) & 0xFU) << 44U) | ((u64)((u64)(bifdm) & 0xFU) << 40U)) + +#define ROGUE_FW_SEGMMU_META_BIFDM_ID (0x7U) + +/* META segments have 4kB minimum size. */ +#define ROGUE_FW_SEGMMU_ALIGN (0x1000U) + +/* Segmented MMU registers (n = segment id). */ +#define META_CR_MMCU_SEGMENT_N_BASE(n) (0x04850000U + ((n) * 0x10U)) +#define META_CR_MMCU_SEGMENT_N_LIMIT(n) (0x04850004U + ((n) * 0x10U)) +#define META_CR_MMCU_SEGMENT_N_OUTA0(n) (0x04850008U + ((n) * 0x10U)) +#define META_CR_MMCU_SEGMENT_N_OUTA1(n) (0x0485000CU + ((n) * 0x10U)) + +/* + * The following defines must be recalculated if the Meta MMU segments used + * to access Host-FW data are changed + * Current combinations are: + * - SLC uncached, META cached, FW base address 0x70000000 + * - SLC uncached, META uncached, FW base address 0xF0000000 + * - SLC cached, META cached, FW base address 0x10000000 + * - SLC cached, META uncached, FW base address 0x90000000 + */ +#define ROGUE_FW_SEGMMU_DATA_BASE_ADDRESS (0x10000000U) +#define ROGUE_FW_SEGMMU_DATA_META_CACHED (0x0U) +#define ROGUE_FW_SEGMMU_DATA_META_UNCACHED (META_MEM_GLOBAL_RANGE_BIT) +#define ROGUE_FW_SEGMMU_DATA_META_CACHE_MASK (META_MEM_GLOBAL_RANGE_BIT) +/* + * For non-VIVT SLCs the cacheability of the FW data in the SLC is selected in + * the PTEs for the FW data, not in the Meta Segment MMU, which means these + * defines have no real effect in those cases. + */ +#define ROGUE_FW_SEGMMU_DATA_VIVT_SLC_CACHED (0x0U) +#define ROGUE_FW_SEGMMU_DATA_VIVT_SLC_UNCACHED (0x60000000U) +#define ROGUE_FW_SEGMMU_DATA_VIVT_SLC_CACHE_MASK (0x60000000U) + +/* + ****************************************************************************** + * ROGUE FW Bootloader defaults + ****************************************************************************** + */ +#define ROGUE_FW_BOOTLDR_META_ADDR (0x40000000U) +#define ROGUE_FW_BOOTLDR_DEVV_ADDR_0 (0xC0000000U) +#define ROGUE_FW_BOOTLDR_DEVV_ADDR_1 (0x000000E1) +#define ROGUE_FW_BOOTLDR_DEVV_ADDR \ + ((((u64)ROGUE_FW_BOOTLDR_DEVV_ADDR_1) << 32) | \ + ROGUE_FW_BOOTLDR_DEVV_ADDR_0) +#define ROGUE_FW_BOOTLDR_LIMIT (0x1FFFF000) +#define ROGUE_FW_MAX_BOOTLDR_OFFSET (0x1000) + +/* Bootloader configuration offset is in dwords (512 bytes) */ +#define ROGUE_FW_BOOTLDR_CONF_OFFSET (0x80) + +/* + ****************************************************************************** + * ROGUE META Stack + ****************************************************************************** + */ +#define ROGUE_META_STACK_SIZE (0x1000U) + +/* + ****************************************************************************** + * ROGUE META Core memory + ****************************************************************************** + */ +/* Code and data both map to the same physical memory. */ +#define ROGUE_META_COREMEM_CODE_ADDR (0x80000000U) +#define ROGUE_META_COREMEM_DATA_ADDR (0x82000000U) +#define ROGUE_META_COREMEM_OFFSET_MASK (0x01ffffffU) + +#define ROGUE_META_IS_COREMEM_CODE(a, b) \ + ({ \ + u32 _a = (a), _b = (b); \ + ((_a) >= ROGUE_META_COREMEM_CODE_ADDR) && \ + ((_a) < (ROGUE_META_COREMEM_CODE_ADDR + (_b))); \ + }) +#define ROGUE_META_IS_COREMEM_DATA(a, b) \ + ({ \ + u32 _a = (a), _b = (b); \ + ((_a) >= ROGUE_META_COREMEM_DATA_ADDR) && \ + ((_a) < (ROGUE_META_COREMEM_DATA_ADDR + (_b))); \ + }) +/* + ****************************************************************************** + * 2nd thread + ****************************************************************************** + */ +#define ROGUE_FW_THR1_PC (0x18930000) +#define ROGUE_FW_THR1_SP (0x78890000) + +/* + ****************************************************************************** + * META compatibility + ****************************************************************************** + */ + +#define META_CR_CORE_ID (0x04831000) +#define META_CR_CORE_ID_VER_SHIFT (16U) +#define META_CR_CORE_ID_VER_CLRMSK (0XFF00FFFFU) + +#define ROGUE_CR_META_MTP218_CORE_ID_VALUE 0x19 +#define ROGUE_CR_META_MTP219_CORE_ID_VALUE 0x1E +#define ROGUE_CR_META_LTP218_CORE_ID_VALUE 0x1C +#define ROGUE_CR_META_LTP217_CORE_ID_VALUE 0x1F + +#define ROGUE_FW_PROCESSOR_META "META" + +#endif /* PVR_ROGUE_META_H */ diff --git a/drivers/gpu/drm/imagination/pvr_rogue_mips.h b/drivers/gpu/drm/imagination/pvr_rogue_mips.h new file mode 100644 index 000000000000..41ed618fda3f --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_mips.h @@ -0,0 +1,335 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_ROGUE_MIPS_H +#define PVR_ROGUE_MIPS_H + +#include +#include + +/* Utility defines for memory management. */ +#define ROGUE_MIPSFW_LOG2_PAGE_SIZE_4K (12) +#define ROGUE_MIPSFW_PAGE_SIZE_4K (0x1 << ROGUE_MIPSFW_LOG2_PAGE_SIZE_4K) +#define ROGUE_MIPSFW_PAGE_MASK_4K (ROGUE_MIPSFW_PAGE_SIZE_4K - 1) +#define ROGUE_MIPSFW_LOG2_PAGE_SIZE_64K (16) +#define ROGUE_MIPSFW_PAGE_SIZE_64K (0x1 << ROGUE_MIPSFW_LOG2_PAGE_SIZE_64K) +#define ROGUE_MIPSFW_PAGE_MASK_64K (ROGUE_MIPSFW_PAGE_SIZE_64K - 1) +#define ROGUE_MIPSFW_LOG2_PAGE_SIZE_256K (18) +#define ROGUE_MIPSFW_PAGE_SIZE_256K (0x1 << ROGUE_MIPSFW_LOG2_PAGE_SIZE_256K) +#define ROGUE_MIPSFW_PAGE_MASK_256K (ROGUE_MIPSFW_PAGE_SIZE_256K - 1) +#define ROGUE_MIPSFW_LOG2_PAGE_SIZE_1MB (20) +#define ROGUE_MIPSFW_PAGE_SIZE_1MB (0x1 << ROGUE_MIPSFW_LOG2_PAGE_SIZE_1MB) +#define ROGUE_MIPSFW_PAGE_MASK_1MB (ROGUE_MIPSFW_PAGE_SIZE_1MB - 1) +#define ROGUE_MIPSFW_LOG2_PAGE_SIZE_4MB (22) +#define ROGUE_MIPSFW_PAGE_SIZE_4MB (0x1 << ROGUE_MIPSFW_LOG2_PAGE_SIZE_4MB) +#define ROGUE_MIPSFW_PAGE_MASK_4MB (ROGUE_MIPSFW_PAGE_SIZE_4MB - 1) +#define ROGUE_MIPSFW_LOG2_PTE_ENTRY_SIZE (2) +/* log2 page table sizes dependent on FW heap size and page size (for each OS). */ +#define ROGUE_MIPSFW_LOG2_PAGETABLE_SIZE_4K(pvr_dev) ((pvr_dev)->fw_dev.fw_heap_info.log2_size - \ + ROGUE_MIPSFW_LOG2_PAGE_SIZE_4K + \ + ROGUE_MIPSFW_LOG2_PTE_ENTRY_SIZE) +#define ROGUE_MIPSFW_LOG2_PAGETABLE_SIZE_64K(pvr_dev) ((pvr_dev)->fw_dev.fw_heap_info.log2_size - \ + ROGUE_MIPSFW_LOG2_PAGE_SIZE_64K + \ + ROGUE_MIPSFW_LOG2_PTE_ENTRY_SIZE) +/* Maximum number of page table pages (both Host and MIPS pages). */ +#define ROGUE_MIPSFW_MAX_NUM_PAGETABLE_PAGES (4) +/* Total number of TLB entries. */ +#define ROGUE_MIPSFW_NUMBER_OF_TLB_ENTRIES (16) +/* "Uncached" caching policy. */ +#define ROGUE_MIPSFW_UNCACHED_CACHE_POLICY (2) +/* "Write-back write-allocate" caching policy. */ +#define ROGUE_MIPSFW_WRITEBACK_CACHE_POLICY (3) +/* "Write-through no write-allocate" caching policy. */ +#define ROGUE_MIPSFW_WRITETHROUGH_CACHE_POLICY (1) +/* Cached policy used by MIPS in case of physical bus on 32 bit. */ +#define ROGUE_MIPSFW_CACHED_POLICY (ROGUE_MIPSFW_WRITEBACK_CACHE_POLICY) +/* Cached policy used by MIPS in case of physical bus on more than 32 bit. */ +#define ROGUE_MIPSFW_CACHED_POLICY_ABOVE_32BIT (ROGUE_MIPSFW_WRITETHROUGH_CACHE_POLICY) +/* Total number of Remap entries. */ +#define ROGUE_MIPSFW_NUMBER_OF_REMAP_ENTRIES (2 * ROGUE_MIPSFW_NUMBER_OF_TLB_ENTRIES) + +/* MIPS EntryLo/PTE format. */ + +#define ROGUE_MIPSFW_ENTRYLO_READ_INHIBIT_SHIFT (31U) +#define ROGUE_MIPSFW_ENTRYLO_READ_INHIBIT_CLRMSK (0X7FFFFFFF) +#define ROGUE_MIPSFW_ENTRYLO_READ_INHIBIT_EN (0X80000000) + +#define ROGUE_MIPSFW_ENTRYLO_EXEC_INHIBIT_SHIFT (30U) +#define ROGUE_MIPSFW_ENTRYLO_EXEC_INHIBIT_CLRMSK (0XBFFFFFFF) +#define ROGUE_MIPSFW_ENTRYLO_EXEC_INHIBIT_EN (0X40000000) + +/* Page Frame Number */ +#define ROGUE_MIPSFW_ENTRYLO_PFN_SHIFT (6) +#define ROGUE_MIPSFW_ENTRYLO_PFN_ALIGNSHIFT (12) +/* Mask used for the MIPS Page Table in case of physical bus on 32 bit. */ +#define ROGUE_MIPSFW_ENTRYLO_PFN_MASK (0x03FFFFC0) +#define ROGUE_MIPSFW_ENTRYLO_PFN_SIZE (20) +/* Mask used for the MIPS Page Table in case of physical bus on more than 32 bit. */ +#define ROGUE_MIPSFW_ENTRYLO_PFN_MASK_ABOVE_32BIT (0x3FFFFFC0) +#define ROGUE_MIPSFW_ENTRYLO_PFN_SIZE_ABOVE_32BIT (24) +#define ROGUE_MIPSFW_ADDR_TO_ENTRYLO_PFN_RSHIFT (ROGUE_MIPSFW_ENTRYLO_PFN_ALIGNSHIFT - \ + ROGUE_MIPSFW_ENTRYLO_PFN_SHIFT) + +#define ROGUE_MIPSFW_ENTRYLO_CACHE_POLICY_SHIFT (3U) +#define ROGUE_MIPSFW_ENTRYLO_CACHE_POLICY_CLRMSK (0XFFFFFFC7) + +#define ROGUE_MIPSFW_ENTRYLO_DIRTY_SHIFT (2U) +#define ROGUE_MIPSFW_ENTRYLO_DIRTY_CLRMSK (0XFFFFFFFB) +#define ROGUE_MIPSFW_ENTRYLO_DIRTY_EN (0X00000004) + +#define ROGUE_MIPSFW_ENTRYLO_VALID_SHIFT (1U) +#define ROGUE_MIPSFW_ENTRYLO_VALID_CLRMSK (0XFFFFFFFD) +#define ROGUE_MIPSFW_ENTRYLO_VALID_EN (0X00000002) + +#define ROGUE_MIPSFW_ENTRYLO_GLOBAL_SHIFT (0U) +#define ROGUE_MIPSFW_ENTRYLO_GLOBAL_CLRMSK (0XFFFFFFFE) +#define ROGUE_MIPSFW_ENTRYLO_GLOBAL_EN (0X00000001) + +#define ROGUE_MIPSFW_ENTRYLO_DVG (ROGUE_MIPSFW_ENTRYLO_DIRTY_EN | \ + ROGUE_MIPSFW_ENTRYLO_VALID_EN | \ + ROGUE_MIPSFW_ENTRYLO_GLOBAL_EN) +#define ROGUE_MIPSFW_ENTRYLO_UNCACHED (ROGUE_MIPSFW_UNCACHED_CACHE_POLICY << \ + ROGUE_MIPSFW_ENTRYLO_CACHE_POLICY_SHIFT) +#define ROGUE_MIPSFW_ENTRYLO_DVG_UNCACHED (ROGUE_MIPSFW_ENTRYLO_DVG | \ + ROGUE_MIPSFW_ENTRYLO_UNCACHED) + +/* Remap Range Config Addr Out. */ +/* These defines refer to the upper half of the Remap Range Config register. */ +#define ROGUE_MIPSFW_REMAP_RANGE_ADDR_OUT_MASK (0x0FFFFFF0) +#define ROGUE_MIPSFW_REMAP_RANGE_ADDR_OUT_SHIFT (4) /* wrt upper half of the register. */ +#define ROGUE_MIPSFW_REMAP_RANGE_ADDR_OUT_ALIGNSHIFT (12) +#define ROGUE_MIPSFW_ADDR_TO_RR_ADDR_OUT_RSHIFT (ROGUE_MIPSFW_REMAP_RANGE_ADDR_OUT_ALIGNSHIFT - \ + ROGUE_MIPSFW_REMAP_RANGE_ADDR_OUT_SHIFT) + +/* + * Pages to trampoline problematic physical addresses: + * - ROGUE_MIPSFW_BOOT_REMAP_PHYS_ADDR_IN : 0x1FC0_0000 + * - ROGUE_MIPSFW_DATA_REMAP_PHYS_ADDR_IN : 0x1FC0_1000 + * - ROGUE_MIPSFW_CODE_REMAP_PHYS_ADDR_IN : 0x1FC0_2000 + * - (benign trampoline) : 0x1FC0_3000 + * that would otherwise be erroneously remapped by the MIPS wrapper. + * (see "Firmware virtual layout and remap configuration" section below) + */ + +#define ROGUE_MIPSFW_TRAMPOLINE_LOG2_NUMPAGES (2) +#define ROGUE_MIPSFW_TRAMPOLINE_NUMPAGES BIT(ROGUE_MIPSFW_TRAMPOLINE_LOG2_NUMPAGES) +#define ROGUE_MIPSFW_TRAMPOLINE_SIZE (ROGUE_MIPSFW_TRAMPOLINE_NUMPAGES << \ + ROGUE_MIPSFW_LOG2_PAGE_SIZE_4K) +#define ROGUE_MIPSFW_TRAMPOLINE_LOG2_SEGMENT_SIZE (ROGUE_MIPSFW_TRAMPOLINE_LOG2_NUMPAGES + \ + ROGUE_MIPSFW_LOG2_PAGE_SIZE_4K) + +#define ROGUE_MIPSFW_TRAMPOLINE_TARGET_PHYS_ADDR (ROGUE_MIPSFW_BOOT_REMAP_PHYS_ADDR_IN) +#define ROGUE_MIPSFW_TRAMPOLINE_OFFSET(a) ((a) - ROGUE_MIPSFW_BOOT_REMAP_PHYS_ADDR_IN) + +#define ROGUE_MIPSFW_SENSITIVE_ADDR(a) (ROGUE_MIPSFW_BOOT_REMAP_PHYS_ADDR_IN == \ + (~((1 << ROGUE_MIPSFW_TRAMPOLINE_LOG2_SEGMENT_SIZE) - 1) \ + & (a))) + +/* Firmware virtual layout and remap configuration. */ +/* + * For each remap region we define: + * - the virtual base used by the Firmware to access code/data through that region + * - the microAptivAP physical address correspondent to the virtual base address, + * used as input address and remapped to the actual physical address + * - log2 of size of the region remapped by the MIPS wrapper, i.e. number of bits from + * the bottom of the base input address that survive onto the output address + * (this defines both the alignment and the maximum size of the remapped region) + * - one or more code/data segments within the remapped region. + */ + +/* Boot remap setup. */ +#define ROGUE_MIPSFW_BOOT_REMAP_VIRTUAL_BASE (0xBFC00000) +#define ROGUE_MIPSFW_BOOT_REMAP_PHYS_ADDR_IN (0x1FC00000) +#define ROGUE_MIPSFW_BOOT_REMAP_LOG2_SEGMENT_SIZE (12) +#define ROGUE_MIPSFW_BOOT_NMI_CODE_VIRTUAL_BASE (ROGUE_MIPSFW_BOOT_REMAP_VIRTUAL_BASE) + +/* Data remap setup. */ +#define ROGUE_MIPSFW_DATA_REMAP_VIRTUAL_BASE (0xBFC01000) +#define ROGUE_MIPSFW_DATA_CACHED_REMAP_VIRTUAL_BASE (0x9FC01000) +#define ROGUE_MIPSFW_DATA_REMAP_PHYS_ADDR_IN (0x1FC01000) +#define ROGUE_MIPSFW_DATA_REMAP_LOG2_SEGMENT_SIZE (12) +#define ROGUE_MIPSFW_BOOT_NMI_DATA_VIRTUAL_BASE (ROGUE_MIPSFW_DATA_REMAP_VIRTUAL_BASE) + +/* Code remap setup. */ +#define ROGUE_MIPSFW_CODE_REMAP_VIRTUAL_BASE (0x9FC02000) +#define ROGUE_MIPSFW_CODE_REMAP_PHYS_ADDR_IN (0x1FC02000) +#define ROGUE_MIPSFW_CODE_REMAP_LOG2_SEGMENT_SIZE (12) +#define ROGUE_MIPSFW_EXCEPTIONS_VIRTUAL_BASE (ROGUE_MIPSFW_CODE_REMAP_VIRTUAL_BASE) + +/* Permanent mappings setup. */ +#define ROGUE_MIPSFW_PT_VIRTUAL_BASE (0xCF000000) +#define ROGUE_MIPSFW_REGISTERS_VIRTUAL_BASE (0xCF800000) +#define ROGUE_MIPSFW_STACK_VIRTUAL_BASE (0xCF600000) + +/* Bootloader configuration data. */ +/* + * Bootloader configuration offset (where ROGUE_MIPSFW_BOOT_DATA lives) + * within the bootloader/NMI data page. + */ +#define ROGUE_MIPSFW_BOOTLDR_CONF_OFFSET (0x0) + +/* NMI shared data. */ +/* Base address of the shared data within the bootloader/NMI data page. */ +#define ROGUE_MIPSFW_NMI_SHARED_DATA_BASE (0x100) +/* Size used by Debug dump data. */ +#define ROGUE_MIPSFW_NMI_SHARED_SIZE (0x2B0) +/* Offsets in the NMI shared area in 32-bit words. */ +#define ROGUE_MIPSFW_NMI_SYNC_FLAG_OFFSET (0x0) +#define ROGUE_MIPSFW_NMI_STATE_OFFSET (0x1) +#define ROGUE_MIPSFW_NMI_ERROR_STATE_SET (0x1) + +/* MIPS boot stage. */ +#define ROGUE_MIPSFW_BOOT_STAGE_OFFSET (0x400) + +/* + * MIPS private data in the bootloader data page. + * Memory below this offset is used by the FW only, no interface data allowed. + */ +#define ROGUE_MIPSFW_PRIVATE_DATA_OFFSET (0x800) + +struct rogue_mipsfw_boot_data { + u64 stack_phys_addr; + u64 reg_base; + u64 pt_phys_addr[ROGUE_MIPSFW_MAX_NUM_PAGETABLE_PAGES]; + u32 pt_log2_page_size; + u32 pt_num_pages; + u32 reserved1; + u32 reserved2; +}; + +#define ROGUE_MIPSFW_GET_OFFSET_IN_DWORDS(offset) ((offset) / sizeof(u32)) +#define ROGUE_MIPSFW_GET_OFFSET_IN_QWORDS(offset) ((offset) / sizeof(u64)) + +/* Used for compatibility checks. */ +#define ROGUE_MIPSFW_ARCHTYPE_VER_CLRMSK (0xFFFFE3FFU) +#define ROGUE_MIPSFW_ARCHTYPE_VER_SHIFT (10U) +#define ROGUE_MIPSFW_CORE_ID_VALUE (0x001U) +#define ROGUE_FW_PROCESSOR_MIPS "MIPS" + +/* microAptivAP cache line size. */ +#define ROGUE_MIPSFW_MICROAPTIVEAP_CACHELINE_SIZE (16U) + +/* + * The SOCIF transactions are identified with the top 16 bits of the physical address emitted by + * the MIPS. + */ +#define ROGUE_MIPSFW_WRAPPER_CONFIG_REGBANK_ADDR_ALIGN (16U) + +/* Values to put in the MIPS selectors for performance counters. */ +/* Icache accesses in COUNTER0. */ +#define ROGUE_MIPSFW_PERF_COUNT_CTRL_ICACHE_ACCESSES_C0 (9U) +/* Icache misses in COUNTER1. */ +#define ROGUE_MIPSFW_PERF_COUNT_CTRL_ICACHE_MISSES_C1 (9U) + +/* Dcache accesses in COUNTER0. */ +#define ROGUE_MIPSFW_PERF_COUNT_CTRL_DCACHE_ACCESSES_C0 (10U) +/* Dcache misses in COUNTER1. */ +#define ROGUE_MIPSFW_PERF_COUNT_CTRL_DCACHE_MISSES_C1 (11U) + +/* ITLB instruction accesses in COUNTER0. */ +#define ROGUE_MIPSFW_PERF_COUNT_CTRL_ITLB_INSTR_ACCESSES_C0 (5U) +/* JTLB instruction accesses misses in COUNTER1. */ +#define ROGUE_MIPSFW_PERF_COUNT_CTRL_JTLB_INSTR_MISSES_C1 (7U) + + /* Instructions completed in COUNTER0. */ +#define ROGUE_MIPSFW_PERF_COUNT_CTRL_INSTR_COMPLETED_C0 (1U) +/* JTLB data misses in COUNTER1. */ +#define ROGUE_MIPSFW_PERF_COUNT_CTRL_JTLB_DATA_MISSES_C1 (8U) + +/* Shift for the Event field in the MIPS perf ctrl registers. */ +#define ROGUE_MIPSFW_PERF_COUNT_CTRL_EVENT_SHIFT (5U) + +/* Additional flags for performance counters. See MIPS manual for further reference. */ +#define ROGUE_MIPSFW_PERF_COUNT_CTRL_COUNT_USER_MODE (8U) +#define ROGUE_MIPSFW_PERF_COUNT_CTRL_COUNT_KERNEL_MODE (2U) +#define ROGUE_MIPSFW_PERF_COUNT_CTRL_COUNT_EXL (1U) + +#define ROGUE_MIPSFW_C0_NBHWIRQ 8 + +/* Macros to decode C0_Cause register. */ +#define ROGUE_MIPSFW_C0_CAUSE_EXCCODE(cause) (((cause) & 0x7c) >> 2) +#define ROGUE_MIPSFW_C0_CAUSE_EXCCODE_FWERROR 9 +/* Use only when Coprocessor Unusable exception. */ +#define ROGUE_MIPSFW_C0_CAUSE_UNUSABLE_UNIT(cause) (((cause) >> 28) & 0x3) +#define ROGUE_MIPSFW_C0_CAUSE_PENDING_HWIRQ(cause) (((cause) & 0x3fc00) >> 10) +#define ROGUE_MIPSFW_C0_CAUSE_FDCIPENDING BIT(21) +#define ROGUE_MIPSFW_C0_CAUSE_IV BIT(23) +#define ROGUE_MIPSFW_C0_CAUSE_IC BIT(25) +#define ROGUE_MIPSFW_C0_CAUSE_PCIPENDING BIT(26) +#define ROGUE_MIPSFW_C0_CAUSE_TIPENDING BIT(30) +#define ROGUE_MIPSFW_C0_CAUSE_BRANCH_DELAY BIT(31) + +/* Macros to decode C0_Debug register. */ +#define ROGUE_MIPSFW_C0_DEBUG_EXCCODE(debug) (((debug) >> 10) & 0x1f) +#define ROGUE_MIPSFW_C0_DEBUG_DSS BIT(0) +#define ROGUE_MIPSFW_C0_DEBUG_DBP BIT(1) +#define ROGUE_MIPSFW_C0_DEBUG_DDBL BIT(2) +#define ROGUE_MIPSFW_C0_DEBUG_DDBS BIT(3) +#define ROGUE_MIPSFW_C0_DEBUG_DIB BIT(4) +#define ROGUE_MIPSFW_C0_DEBUG_DINT BIT(5) +#define ROGUE_MIPSFW_C0_DEBUG_DIBIMPR BIT(6) +#define ROGUE_MIPSFW_C0_DEBUG_DDBLIMPR BIT(18) +#define ROGUE_MIPSFW_C0_DEBUG_DDBSIMPR BIT(19) +#define ROGUE_MIPSFW_C0_DEBUG_IEXI BIT(20) +#define ROGUE_MIPSFW_C0_DEBUG_DBUSEP BIT(21) +#define ROGUE_MIPSFW_C0_DEBUG_CACHEEP BIT(22) +#define ROGUE_MIPSFW_C0_DEBUG_MCHECKP BIT(23) +#define ROGUE_MIPSFW_C0_DEBUG_IBUSEP BIT(24) +#define ROGUE_MIPSFW_C0_DEBUG_DM BIT(30) +#define ROGUE_MIPSFW_C0_DEBUG_DBD BIT(31) + +/* Macros to decode TLB entries. */ +#define ROGUE_MIPSFW_TLB_GET_MASK(page_mask) (((page_mask) >> 13) & 0XFFFFU) +/* Page size in KB. */ +#define ROGUE_MIPSFW_TLB_GET_PAGE_SIZE(page_mask) ((((page_mask) | 0x1FFF) + 1) >> 11) +/* Page size in KB. */ +#define ROGUE_MIPSFW_TLB_GET_PAGE_MASK(page_size) ((((page_size) << 11) - 1) & ~0x7FF) +#define ROGUE_MIPSFW_TLB_GET_VPN2(entry_hi) ((entry_hi) >> 13) +#define ROGUE_MIPSFW_TLB_GET_COHERENCY(entry_lo) (((entry_lo) >> 3) & 0x7U) +#define ROGUE_MIPSFW_TLB_GET_PFN(entry_lo) (((entry_lo) >> 6) & 0XFFFFFU) +/* GET_PA uses a non-standard PFN mask for 36 bit addresses. */ +#define ROGUE_MIPSFW_TLB_GET_PA(entry_lo) (((u64)(entry_lo) & \ + ROGUE_MIPSFW_ENTRYLO_PFN_MASK_ABOVE_32BIT) << 6) +#define ROGUE_MIPSFW_TLB_GET_INHIBIT(entry_lo) (((entry_lo) >> 30) & 0x3U) +#define ROGUE_MIPSFW_TLB_GET_DGV(entry_lo) ((entry_lo) & 0x7U) +#define ROGUE_MIPSFW_TLB_GLOBAL BIT(0) +#define ROGUE_MIPSFW_TLB_VALID BIT(1) +#define ROGUE_MIPSFW_TLB_DIRTY BIT(2) +#define ROGUE_MIPSFW_TLB_XI BIT(30) +#define ROGUE_MIPSFW_TLB_RI BIT(31) + +#define ROGUE_MIPSFW_REMAP_GET_REGION_SIZE(region_size_encoding) (1 << (((region_size_encoding) \ + + 1) << 1)) + +struct rogue_mips_tlb_entry { + u32 tlb_page_mask; + u32 tlb_hi; + u32 tlb_lo0; + u32 tlb_lo1; +}; + +struct rogue_mips_remap_entry { + u32 remap_addr_in; /* Always 4k aligned. */ + u32 remap_addr_out; /* Always 4k aligned. */ + u32 remap_region_size; +}; + +struct rogue_mips_state { + u32 error_state; /* This must come first in the structure. */ + u32 error_epc; + u32 status_register; + u32 cause_register; + u32 bad_register; + u32 epc; + u32 sp; + u32 debug; + u32 depc; + u32 bad_instr; + u32 unmapped_address; + struct rogue_mips_tlb_entry tlb[ROGUE_MIPSFW_NUMBER_OF_TLB_ENTRIES]; + struct rogue_mips_remap_entry remap[ROGUE_MIPSFW_NUMBER_OF_REMAP_ENTRIES]; +}; + +#include "pvr_rogue_mips_check.h" + +#endif /* PVR_ROGUE_MIPS_H */ diff --git a/drivers/gpu/drm/imagination/pvr_rogue_mips_check.h b/drivers/gpu/drm/imagination/pvr_rogue_mips_check.h new file mode 100644 index 000000000000..824b4bf33ac1 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_mips_check.h @@ -0,0 +1,58 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_ROGUE_MIPS_CHECK_H +#define PVR_ROGUE_MIPS_CHECK_H + +#include + +static_assert(offsetof(struct rogue_mips_tlb_entry, tlb_page_mask) == 0, + "offsetof(struct rogue_mips_tlb_entry, tlb_page_mask) incorrect"); +static_assert(offsetof(struct rogue_mips_tlb_entry, tlb_hi) == 4, + "offsetof(struct rogue_mips_tlb_entry, tlb_hi) incorrect"); +static_assert(offsetof(struct rogue_mips_tlb_entry, tlb_lo0) == 8, + "offsetof(struct rogue_mips_tlb_entry, tlb_lo0) incorrect"); +static_assert(offsetof(struct rogue_mips_tlb_entry, tlb_lo1) == 12, + "offsetof(struct rogue_mips_tlb_entry, tlb_lo1) incorrect"); +static_assert(sizeof(struct rogue_mips_tlb_entry) == 16, + "struct rogue_mips_tlb_entry is incorrect size"); + +static_assert(offsetof(struct rogue_mips_remap_entry, remap_addr_in) == 0, + "offsetof(struct rogue_mips_remap_entry, remap_addr_in) incorrect"); +static_assert(offsetof(struct rogue_mips_remap_entry, remap_addr_out) == 4, + "offsetof(struct rogue_mips_remap_entry, remap_addr_out) incorrect"); +static_assert(offsetof(struct rogue_mips_remap_entry, remap_region_size) == 8, + "offsetof(struct rogue_mips_remap_entry, remap_region_size) incorrect"); +static_assert(sizeof(struct rogue_mips_remap_entry) == 12, + "struct rogue_mips_remap_entry is incorrect size"); + +static_assert(offsetof(struct rogue_mips_state, error_state) == 0, + "offsetof(struct rogue_mips_state, error_state) incorrect"); +static_assert(offsetof(struct rogue_mips_state, error_epc) == 4, + "offsetof(struct rogue_mips_state, error_epc) incorrect"); +static_assert(offsetof(struct rogue_mips_state, status_register) == 8, + "offsetof(struct rogue_mips_state, status_register) incorrect"); +static_assert(offsetof(struct rogue_mips_state, cause_register) == 12, + "offsetof(struct rogue_mips_state, cause_register) incorrect"); +static_assert(offsetof(struct rogue_mips_state, bad_register) == 16, + "offsetof(struct rogue_mips_state, bad_register) incorrect"); +static_assert(offsetof(struct rogue_mips_state, epc) == 20, + "offsetof(struct rogue_mips_state, epc) incorrect"); +static_assert(offsetof(struct rogue_mips_state, sp) == 24, + "offsetof(struct rogue_mips_state, sp) incorrect"); +static_assert(offsetof(struct rogue_mips_state, debug) == 28, + "offsetof(struct rogue_mips_state, debug) incorrect"); +static_assert(offsetof(struct rogue_mips_state, depc) == 32, + "offsetof(struct rogue_mips_state, depc) incorrect"); +static_assert(offsetof(struct rogue_mips_state, bad_instr) == 36, + "offsetof(struct rogue_mips_state, bad_instr) incorrect"); +static_assert(offsetof(struct rogue_mips_state, unmapped_address) == 40, + "offsetof(struct rogue_mips_state, unmapped_address) incorrect"); +static_assert(offsetof(struct rogue_mips_state, tlb) == 44, + "offsetof(struct rogue_mips_state, tlb) incorrect"); +static_assert(offsetof(struct rogue_mips_state, remap) == 300, + "offsetof(struct rogue_mips_state, remap) incorrect"); +static_assert(sizeof(struct rogue_mips_state) == 684, + "struct rogue_mips_state is incorrect size"); + +#endif /* PVR_ROGUE_MIPS_CHECK_H */ diff --git a/drivers/gpu/drm/imagination/pvr_rogue_mmu_defs.h b/drivers/gpu/drm/imagination/pvr_rogue_mmu_defs.h new file mode 100644 index 000000000000..f361ccdd5405 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_mmu_defs.h @@ -0,0 +1,136 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +/* *** Autogenerated C -- do not edit *** */ + +#ifndef PVR_ROGUE_MMU_DEFS_H +#define PVR_ROGUE_MMU_DEFS_H + +#define ROGUE_MMU_DEFS_REVISION 0 + +#define ROGUE_BIF_DM_ENCODING_VERTEX (0x00000000U) +#define ROGUE_BIF_DM_ENCODING_PIXEL (0x00000001U) +#define ROGUE_BIF_DM_ENCODING_COMPUTE (0x00000002U) +#define ROGUE_BIF_DM_ENCODING_TLA (0x00000003U) +#define ROGUE_BIF_DM_ENCODING_PB_VCE (0x00000004U) +#define ROGUE_BIF_DM_ENCODING_PB_TE (0x00000005U) +#define ROGUE_BIF_DM_ENCODING_META (0x00000007U) +#define ROGUE_BIF_DM_ENCODING_HOST (0x00000008U) +#define ROGUE_BIF_DM_ENCODING_PM_ALIST (0x00000009U) + +#define ROGUE_MMUCTRL_VADDR_PC_INDEX_SHIFT (30U) +#define ROGUE_MMUCTRL_VADDR_PC_INDEX_CLRMSK (0xFFFFFF003FFFFFFFULL) +#define ROGUE_MMUCTRL_VADDR_PD_INDEX_SHIFT (21U) +#define ROGUE_MMUCTRL_VADDR_PD_INDEX_CLRMSK (0xFFFFFFFFC01FFFFFULL) +#define ROGUE_MMUCTRL_VADDR_PT_INDEX_SHIFT (12U) +#define ROGUE_MMUCTRL_VADDR_PT_INDEX_CLRMSK (0xFFFFFFFFFFE00FFFULL) + +#define ROGUE_MMUCTRL_ENTRIES_PC_VALUE (0x00000400U) +#define ROGUE_MMUCTRL_ENTRIES_PD_VALUE (0x00000200U) +#define ROGUE_MMUCTRL_ENTRIES_PT_VALUE (0x00000200U) + +#define ROGUE_MMUCTRL_ENTRY_SIZE_PC_VALUE (0x00000020U) +#define ROGUE_MMUCTRL_ENTRY_SIZE_PD_VALUE (0x00000040U) +#define ROGUE_MMUCTRL_ENTRY_SIZE_PT_VALUE (0x00000040U) + +#define ROGUE_MMUCTRL_PAGE_SIZE_MASK (0x00000007U) +#define ROGUE_MMUCTRL_PAGE_SIZE_4KB (0x00000000U) +#define ROGUE_MMUCTRL_PAGE_SIZE_16KB (0x00000001U) +#define ROGUE_MMUCTRL_PAGE_SIZE_64KB (0x00000002U) +#define ROGUE_MMUCTRL_PAGE_SIZE_256KB (0x00000003U) +#define ROGUE_MMUCTRL_PAGE_SIZE_1MB (0x00000004U) +#define ROGUE_MMUCTRL_PAGE_SIZE_2MB (0x00000005U) + +#define ROGUE_MMUCTRL_PAGE_4KB_RANGE_SHIFT (12U) +#define ROGUE_MMUCTRL_PAGE_4KB_RANGE_CLRMSK (0xFFFFFF0000000FFFULL) + +#define ROGUE_MMUCTRL_PAGE_16KB_RANGE_SHIFT (14U) +#define ROGUE_MMUCTRL_PAGE_16KB_RANGE_CLRMSK (0xFFFFFF0000003FFFULL) + +#define ROGUE_MMUCTRL_PAGE_64KB_RANGE_SHIFT (16U) +#define ROGUE_MMUCTRL_PAGE_64KB_RANGE_CLRMSK (0xFFFFFF000000FFFFULL) + +#define ROGUE_MMUCTRL_PAGE_256KB_RANGE_SHIFT (18U) +#define ROGUE_MMUCTRL_PAGE_256KB_RANGE_CLRMSK (0xFFFFFF000003FFFFULL) + +#define ROGUE_MMUCTRL_PAGE_1MB_RANGE_SHIFT (20U) +#define ROGUE_MMUCTRL_PAGE_1MB_RANGE_CLRMSK (0xFFFFFF00000FFFFFULL) + +#define ROGUE_MMUCTRL_PAGE_2MB_RANGE_SHIFT (21U) +#define ROGUE_MMUCTRL_PAGE_2MB_RANGE_CLRMSK (0xFFFFFF00001FFFFFULL) + +#define ROGUE_MMUCTRL_PT_BASE_4KB_RANGE_SHIFT (12U) +#define ROGUE_MMUCTRL_PT_BASE_4KB_RANGE_CLRMSK (0xFFFFFF0000000FFFULL) + +#define ROGUE_MMUCTRL_PT_BASE_16KB_RANGE_SHIFT (10U) +#define ROGUE_MMUCTRL_PT_BASE_16KB_RANGE_CLRMSK (0xFFFFFF00000003FFULL) + +#define ROGUE_MMUCTRL_PT_BASE_64KB_RANGE_SHIFT (8U) +#define ROGUE_MMUCTRL_PT_BASE_64KB_RANGE_CLRMSK (0xFFFFFF00000000FFULL) + +#define ROGUE_MMUCTRL_PT_BASE_256KB_RANGE_SHIFT (6U) +#define ROGUE_MMUCTRL_PT_BASE_256KB_RANGE_CLRMSK (0xFFFFFF000000003FULL) + +#define ROGUE_MMUCTRL_PT_BASE_1MB_RANGE_SHIFT (5U) +#define ROGUE_MMUCTRL_PT_BASE_1MB_RANGE_CLRMSK (0xFFFFFF000000001FULL) + +#define ROGUE_MMUCTRL_PT_BASE_2MB_RANGE_SHIFT (5U) +#define ROGUE_MMUCTRL_PT_BASE_2MB_RANGE_CLRMSK (0xFFFFFF000000001FULL) + +#define ROGUE_MMUCTRL_PT_DATA_PM_META_PROTECT_SHIFT (62U) +#define ROGUE_MMUCTRL_PT_DATA_PM_META_PROTECT_CLRMSK (0xBFFFFFFFFFFFFFFFULL) +#define ROGUE_MMUCTRL_PT_DATA_PM_META_PROTECT_EN (0x4000000000000000ULL) +#define ROGUE_MMUCTRL_PT_DATA_VP_PAGE_HI_SHIFT (40U) +#define ROGUE_MMUCTRL_PT_DATA_VP_PAGE_HI_CLRMSK (0xC00000FFFFFFFFFFULL) +#define ROGUE_MMUCTRL_PT_DATA_PAGE_SHIFT (12U) +#define ROGUE_MMUCTRL_PT_DATA_PAGE_CLRMSK (0xFFFFFF0000000FFFULL) +#define ROGUE_MMUCTRL_PT_DATA_VP_PAGE_LO_SHIFT (6U) +#define ROGUE_MMUCTRL_PT_DATA_VP_PAGE_LO_CLRMSK (0xFFFFFFFFFFFFF03FULL) +#define ROGUE_MMUCTRL_PT_DATA_ENTRY_PENDING_SHIFT (5U) +#define ROGUE_MMUCTRL_PT_DATA_ENTRY_PENDING_CLRMSK (0xFFFFFFFFFFFFFFDFULL) +#define ROGUE_MMUCTRL_PT_DATA_ENTRY_PENDING_EN (0x0000000000000020ULL) +#define ROGUE_MMUCTRL_PT_DATA_PM_SRC_SHIFT (4U) +#define ROGUE_MMUCTRL_PT_DATA_PM_SRC_CLRMSK (0xFFFFFFFFFFFFFFEFULL) +#define ROGUE_MMUCTRL_PT_DATA_PM_SRC_EN (0x0000000000000010ULL) +#define ROGUE_MMUCTRL_PT_DATA_SLC_BYPASS_CTRL_SHIFT (3U) +#define ROGUE_MMUCTRL_PT_DATA_SLC_BYPASS_CTRL_CLRMSK (0xFFFFFFFFFFFFFFF7ULL) +#define ROGUE_MMUCTRL_PT_DATA_SLC_BYPASS_CTRL_EN (0x0000000000000008ULL) +#define ROGUE_MMUCTRL_PT_DATA_CC_SHIFT (2U) +#define ROGUE_MMUCTRL_PT_DATA_CC_CLRMSK (0xFFFFFFFFFFFFFFFBULL) +#define ROGUE_MMUCTRL_PT_DATA_CC_EN (0x0000000000000004ULL) +#define ROGUE_MMUCTRL_PT_DATA_READ_ONLY_SHIFT (1U) +#define ROGUE_MMUCTRL_PT_DATA_READ_ONLY_CLRMSK (0xFFFFFFFFFFFFFFFDULL) +#define ROGUE_MMUCTRL_PT_DATA_READ_ONLY_EN (0x0000000000000002ULL) +#define ROGUE_MMUCTRL_PT_DATA_VALID_SHIFT (0U) +#define ROGUE_MMUCTRL_PT_DATA_VALID_CLRMSK (0xFFFFFFFFFFFFFFFEULL) +#define ROGUE_MMUCTRL_PT_DATA_VALID_EN (0x0000000000000001ULL) + +#define ROGUE_MMUCTRL_PD_DATA_ENTRY_PENDING_SHIFT (40U) +#define ROGUE_MMUCTRL_PD_DATA_ENTRY_PENDING_CLRMSK (0xFFFFFEFFFFFFFFFFULL) +#define ROGUE_MMUCTRL_PD_DATA_ENTRY_PENDING_EN (0x0000010000000000ULL) +#define ROGUE_MMUCTRL_PD_DATA_PT_BASE_SHIFT (5U) +#define ROGUE_MMUCTRL_PD_DATA_PT_BASE_CLRMSK (0xFFFFFF000000001FULL) +#define ROGUE_MMUCTRL_PD_DATA_PAGE_SIZE_SHIFT (1U) +#define ROGUE_MMUCTRL_PD_DATA_PAGE_SIZE_CLRMSK (0xFFFFFFFFFFFFFFF1ULL) +#define ROGUE_MMUCTRL_PD_DATA_PAGE_SIZE_4KB (0x0000000000000000ULL) +#define ROGUE_MMUCTRL_PD_DATA_PAGE_SIZE_16KB (0x0000000000000002ULL) +#define ROGUE_MMUCTRL_PD_DATA_PAGE_SIZE_64KB (0x0000000000000004ULL) +#define ROGUE_MMUCTRL_PD_DATA_PAGE_SIZE_256KB (0x0000000000000006ULL) +#define ROGUE_MMUCTRL_PD_DATA_PAGE_SIZE_1MB (0x0000000000000008ULL) +#define ROGUE_MMUCTRL_PD_DATA_PAGE_SIZE_2MB (0x000000000000000aULL) +#define ROGUE_MMUCTRL_PD_DATA_VALID_SHIFT (0U) +#define ROGUE_MMUCTRL_PD_DATA_VALID_CLRMSK (0xFFFFFFFFFFFFFFFEULL) +#define ROGUE_MMUCTRL_PD_DATA_VALID_EN (0x0000000000000001ULL) + +#define ROGUE_MMUCTRL_PC_DATA_PD_BASE_SHIFT (4U) +#define ROGUE_MMUCTRL_PC_DATA_PD_BASE_CLRMSK (0x0000000FU) +#define ROGUE_MMUCTRL_PC_DATA_PD_BASE_ALIGNSHIFT (12U) +#define ROGUE_MMUCTRL_PC_DATA_PD_BASE_ALIGNSIZE (4096U) +#define ROGUE_MMUCTRL_PC_DATA_ENTRY_PENDING_SHIFT (1U) +#define ROGUE_MMUCTRL_PC_DATA_ENTRY_PENDING_CLRMSK (0xFFFFFFFDU) +#define ROGUE_MMUCTRL_PC_DATA_ENTRY_PENDING_EN (0x00000002U) +#define ROGUE_MMUCTRL_PC_DATA_VALID_SHIFT (0U) +#define ROGUE_MMUCTRL_PC_DATA_VALID_CLRMSK (0xFFFFFFFEU) +#define ROGUE_MMUCTRL_PC_DATA_VALID_EN (0x00000001U) + +#endif /* PVR_ROGUE_MMU_DEFS_H */ From patchwork Wed Nov 22 16:34:30 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Donald Robson X-Patchwork-Id: 746146 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=imgtec.com header.i=@imgtec.com header.b="yDnp4PjT"; dkim=pass (1024-bit key) header.d=IMGTecCRM.onmicrosoft.com header.i=@IMGTecCRM.onmicrosoft.com header.b="T3h+m0km" Received: from mx07-00376f01.pphosted.com (mx07-00376f01.pphosted.com [185.132.180.163]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 2536E1721; Wed, 22 Nov 2023 08:36:17 -0800 (PST) Received: from pps.filterd (m0168889.ppops.net [127.0.0.1]) by mx07-00376f01.pphosted.com (8.17.1.24/8.17.1.24) with ESMTP id 3AM7nIFe030359; Wed, 22 Nov 2023 16:35:30 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=imgtec.com; h= from:to:cc:subject:date:message-id:in-reply-to:references :content-transfer-encoding:content-type:mime-version; s= dk201812; bh=ZZylZzaf/792UwbVsPxWvcd2cVLaN8nCOD4DjgPPcHc=; b=yDn p4PjTEA+wHGoCJcbz8SNUX3ud6AjRtA3N+UyFLrUztotdEeW4zMFFieuzyOiBS2x q47CoU6JjwYA8i361i86LXnqlsxIUJJ+WFsJ8LUviJiBlxjq3s2taNUeaPV1estK 6lrtFvzsV+LK9IBNTggjqXz+TH5TeMEuQf1cjmFrOLMfFahQnzpbfeePQwxrAQgI PXxro9IljA7OhhkEAki/6s/Jd10ualv32ZnwO/PAYBiqeOO07waDvJWNO0+vwvbq GVKWSfl2M5vfrcgVpRuuU6XqQPM75VkjQfePLGwiU18eZ2wh93V4dD/HEsMbTXXH CAtpHjIrtarhRVuHj3Q== Received: from hhmail05.hh.imgtec.org ([217.156.249.195]) by mx07-00376f01.pphosted.com (PPS) with ESMTPS id 3ug99g9m29-2 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128 verify=NOT); Wed, 22 Nov 2023 16:35:29 +0000 (GMT) Received: from HHMAIL04.hh.imgtec.org (10.100.10.119) by HHMAIL05.hh.imgtec.org (10.100.10.120) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34; Wed, 22 Nov 2023 16:35:28 +0000 Received: from GBR01-LO4-obe.outbound.protection.outlook.com (104.47.85.105) by email.imgtec.com (10.100.10.121) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34 via Frontend Transport; Wed, 22 Nov 2023 16:35:28 +0000 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=FvXR5ueeW78Vit/8J/OwQizGFqGfZj/PxN913KJmjGbMbmbV/O+fyM7THlRsz8yPcRsLF5dCEqRNv7hUa99kJzsfeQpt+SwaiXwy29JbDDwTmsAh5f+B8ojnB+dw37NfUZ8Fqigf0QXuIvhiAy3twIvJjFb2fjMm19ZuWobqRoaUBxL1OjwWCchB/ObbxXXGMViHUiRZvxTuKeaSlYQ0Sr0XG9sZHarvnGjJCjbC1LSzFXLAkG6W2UoxfQWwgtQnaTPedg6JnpFiHMwPjdYtHkeX+alhh2DspNm7KaqX7pQOF9sB8pyMrpt29sAOKY/IF4kVfDiHZRNKF+nuRPrWFg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=ZZylZzaf/792UwbVsPxWvcd2cVLaN8nCOD4DjgPPcHc=; b=PHoGsqcqzmUzclJYqk/3T05tKoeCBWyI1fO6mxa5cgTWFaxMhU37PHLnn4TsU3QNBjolvx0ffGMLZpgyA2u5pAmwnCH/Nonpk4jdzwy68qJzMTuSC/+aGHQG4zPKsX8he7UYmLEcXALaIO1GbuU4royCiRU4nec4hC6RUrPWMs5my24cqckC+leI6j1hwflWXu47Y21z8YZ20geQpiatFFuI8NFV3E6JP4UajtM5g7aYc/6bl+2muZieLecHfDwbyr5EDAI/5gdK22aJh7U7MDxt9srgFpYrFrCyf4EgYSzaIksih9hJAG7vleVSUavhMv/SsH7N9OlVfm9tzfmV3g== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=imgtec.com; dmarc=pass action=none header.from=imgtec.com; dkim=pass header.d=imgtec.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=IMGTecCRM.onmicrosoft.com; s=selector2-IMGTecCRM-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=ZZylZzaf/792UwbVsPxWvcd2cVLaN8nCOD4DjgPPcHc=; b=T3h+m0kmywkSWziPOOTMTs5llpTI6Si9WjSPDwoH+W08SlST8Ylta+ayuxW0Ohvto9aAAfg6sV56OyFGBHcNv89w4rfFf6gKDUQNU1789BxcllhePcDNpOa92ZDKe7hJ/MfCpRb3ejpLXtqQlDcC+6nIZcIX2jSuzGO8Ks8o2v0= Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) by LO6P265MB6459.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:2df::12) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.7025.18; Wed, 22 Nov 2023 16:35:16 +0000 Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34]) by CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34%3]) with mapi id 15.20.7025.017; Wed, 22 Nov 2023 16:35:16 +0000 From: Donald Robson To: CC: , , , , , , , , , , , , , , , , , , , , , , Sarah Walker Subject: [PATCH v9 09/20] drm/imagination: Add FWIF headers Date: Wed, 22 Nov 2023 16:34:30 +0000 Message-Id: X-Mailer: git-send-email 2.25.1 In-Reply-To: References: X-ClientProxiedBy: LO4P265CA0273.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:37a::10) To CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) Precedence: bulk X-Mailing-List: devicetree@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: CWLP265MB5770:EE_|LO6P265MB6459:EE_ X-MS-Office365-Filtering-Correlation-Id: 7fbe8f8a-0bc2-44b7-fe40-08dbeb79024d X-MS-Exchange-SenderADCheck: 1 X-MS-Exchange-AntiSpam-Relay: 0 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-Forefront-Antispam-Report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:; IPV:NLI; SFV:NSPM; H:CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM; PTR:; CAT:NONE; SFS:(13230031)(39850400004)(346002)(396003)(366004)(136003)(376002)(230273577357003)(230173577357003)(230922051799003)(64100799003)(186009)(451199024)(1800799012)(66899024)(30864003)(36756003)(7416002)(41300700001)(5660300002)(2906002)(86362001)(38350700005)(26005)(6666004)(478600001)(6506007)(2616005)(6486002)(107886003)(6512007)(52116002)(8936002)(8676002)(4326008)(44832011)(66946007)(19627235002)(316002)(6916009)(66556008)(66476007)(83380400001)(38100700002)(559001)(579004); DIR:OUT; SFP:1102; X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: 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 X-MS-Exchange-CrossTenant-Network-Message-Id: 7fbe8f8a-0bc2-44b7-fe40-08dbeb79024d X-MS-Exchange-CrossTenant-AuthSource: CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 22 Nov 2023 16:35:16.7412 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 0d5fd8bb-e8c2-4e0a-8dd5-2c264f7140fe X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: QCPObpCIAN1AvaA2XqOb1fMg8bP6KirWU4fEdkBey/G9teYNX4ZMdHc29UevqtbN50R64TNxmrpAMs9+/UyRrt2bqfVI5d9RW0sTtep8Av4= X-MS-Exchange-Transport-CrossTenantHeadersStamped: LO6P265MB6459 X-OriginatorOrg: imgtec.com X-EXCLAIMER-MD-CONFIG: 15a78312-3e47-46eb-9010-2e54d84a9631 X-Proofpoint-ORIG-GUID: 7ato6DeAr63YXsJvahVm8LFAzkUE-gue X-Proofpoint-GUID: 7ato6DeAr63YXsJvahVm8LFAzkUE-gue From: Sarah Walker Changes since v8: - Corrected license identifiers Changes since v7: - Add padding to struct rogue_fwif_ccb_ctl to place read and write offsets in different cache lines Changes since v5: - Split up header commit due to size - Add BRN 71242 to device info Changes since v4: - Add FW header device info Signed-off-by: Sarah Walker Signed-off-by: Donald Robson Acked-by: Maxime Ripard --- drivers/gpu/drm/imagination/pvr_rogue_fwif.h | 2188 +++++++++++++++++ .../drm/imagination/pvr_rogue_fwif_check.h | 493 ++++ .../drm/imagination/pvr_rogue_fwif_client.h | 373 +++ .../imagination/pvr_rogue_fwif_client_check.h | 133 + .../drm/imagination/pvr_rogue_fwif_common.h | 60 + .../drm/imagination/pvr_rogue_fwif_dev_info.h | 113 + .../pvr_rogue_fwif_resetframework.h | 28 + .../drm/imagination/pvr_rogue_fwif_shared.h | 258 ++ .../imagination/pvr_rogue_fwif_shared_check.h | 108 + .../drm/imagination/pvr_rogue_fwif_stream.h | 78 + 10 files changed, 3832 insertions(+) create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_fwif.h create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_fwif_check.h create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_fwif_client.h create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_fwif_client_check.h create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_fwif_common.h create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_fwif_dev_info.h create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_fwif_resetframework.h create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_fwif_shared.h create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_fwif_shared_check.h create mode 100644 drivers/gpu/drm/imagination/pvr_rogue_fwif_stream.h diff --git a/drivers/gpu/drm/imagination/pvr_rogue_fwif.h b/drivers/gpu/drm/imagination/pvr_rogue_fwif.h new file mode 100644 index 000000000000..172886be4c82 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_fwif.h @@ -0,0 +1,2188 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_ROGUE_FWIF_H +#define PVR_ROGUE_FWIF_H + +#include +#include +#include +#include +#include + +#include "pvr_rogue_defs.h" +#include "pvr_rogue_fwif_common.h" +#include "pvr_rogue_fwif_shared.h" + +/* + **************************************************************************** + * Logging type + **************************************************************************** + */ +#define ROGUE_FWIF_LOG_TYPE_NONE 0x00000000U +#define ROGUE_FWIF_LOG_TYPE_TRACE 0x00000001U +#define ROGUE_FWIF_LOG_TYPE_GROUP_MAIN 0x00000002U +#define ROGUE_FWIF_LOG_TYPE_GROUP_MTS 0x00000004U +#define ROGUE_FWIF_LOG_TYPE_GROUP_CLEANUP 0x00000008U +#define ROGUE_FWIF_LOG_TYPE_GROUP_CSW 0x00000010U +#define ROGUE_FWIF_LOG_TYPE_GROUP_BIF 0x00000020U +#define ROGUE_FWIF_LOG_TYPE_GROUP_PM 0x00000040U +#define ROGUE_FWIF_LOG_TYPE_GROUP_RTD 0x00000080U +#define ROGUE_FWIF_LOG_TYPE_GROUP_SPM 0x00000100U +#define ROGUE_FWIF_LOG_TYPE_GROUP_POW 0x00000200U +#define ROGUE_FWIF_LOG_TYPE_GROUP_HWR 0x00000400U +#define ROGUE_FWIF_LOG_TYPE_GROUP_HWP 0x00000800U +#define ROGUE_FWIF_LOG_TYPE_GROUP_RPM 0x00001000U +#define ROGUE_FWIF_LOG_TYPE_GROUP_DMA 0x00002000U +#define ROGUE_FWIF_LOG_TYPE_GROUP_MISC 0x00004000U +#define ROGUE_FWIF_LOG_TYPE_GROUP_DEBUG 0x80000000U +#define ROGUE_FWIF_LOG_TYPE_GROUP_MASK 0x80007FFEU +#define ROGUE_FWIF_LOG_TYPE_MASK 0x80007FFFU + +/* String used in pvrdebug -h output */ +#define ROGUE_FWIF_LOG_GROUPS_STRING_LIST \ + "main,mts,cleanup,csw,bif,pm,rtd,spm,pow,hwr,hwp,rpm,dma,misc,debug" + +/* Table entry to map log group strings to log type value */ +struct rogue_fwif_log_group_map_entry { + const char *log_group_name; + u32 log_group_type; +}; + +/* + **************************************************************************** + * ROGUE FW signature checks + **************************************************************************** + */ +#define ROGUE_FW_SIG_BUFFER_SIZE_MIN (8192) + +#define ROGUE_FWIF_TIMEDIFF_ID ((0x1UL << 28) | ROGUE_CR_TIMER) + +/* + **************************************************************************** + * Trace Buffer + **************************************************************************** + */ + +/* Default size of ROGUE_FWIF_TRACEBUF_SPACE in DWords */ +#define ROGUE_FW_TRACE_BUF_DEFAULT_SIZE_IN_DWORDS 12000U +#define ROGUE_FW_TRACE_BUFFER_ASSERT_SIZE 200U +#define ROGUE_FW_THREAD_NUM 1U +#define ROGUE_FW_THREAD_MAX 2U + +#define ROGUE_FW_POLL_TYPE_SET 0x80000000U + +struct rogue_fwif_file_info_buf { + char path[ROGUE_FW_TRACE_BUFFER_ASSERT_SIZE]; + char info[ROGUE_FW_TRACE_BUFFER_ASSERT_SIZE]; + u32 line_num; + u32 padding; +} __aligned(8); + +struct rogue_fwif_tracebuf_space { + u32 trace_pointer; + + u32 trace_buffer_fw_addr; + + /* To be used by host when reading from trace buffer */ + u32 *trace_buffer; + + struct rogue_fwif_file_info_buf assert_buf; +} __aligned(8); + +/* Total number of FW fault logs stored */ +#define ROGUE_FWIF_FWFAULTINFO_MAX (8U) + +struct rogue_fw_fault_info { + aligned_u64 cr_timer; + aligned_u64 os_timer; + + u32 data __aligned(8); + u32 reserved; + struct rogue_fwif_file_info_buf fault_buf; +} __aligned(8); + +enum rogue_fwif_pow_state { + ROGUE_FWIF_POW_OFF, /* idle and ready to full power down */ + ROGUE_FWIF_POW_ON, /* running HW commands */ + ROGUE_FWIF_POW_FORCED_IDLE, /* forced idle */ + ROGUE_FWIF_POW_IDLE, /* idle waiting for host handshake */ +}; + +/* Firmware HWR states */ +/* The HW state is ok or locked up */ +#define ROGUE_FWIF_HWR_HARDWARE_OK BIT(0) +/* Tells if a HWR reset is in progress */ +#define ROGUE_FWIF_HWR_RESET_IN_PROGRESS BIT(1) +/* A DM unrelated lockup has been detected */ +#define ROGUE_FWIF_HWR_GENERAL_LOCKUP BIT(3) +/* At least one DM is running without being close to a lockup */ +#define ROGUE_FWIF_HWR_DM_RUNNING_OK BIT(4) +/* At least one DM is close to lockup */ +#define ROGUE_FWIF_HWR_DM_STALLING BIT(5) +/* The FW has faulted and needs to restart */ +#define ROGUE_FWIF_HWR_FW_FAULT BIT(6) +/* The FW has requested the host to restart it */ +#define ROGUE_FWIF_HWR_RESTART_REQUESTED BIT(7) + +#define ROGUE_FWIF_PHR_STATE_SHIFT (8U) +/* The FW has requested the host to restart it, per PHR configuration */ +#define ROGUE_FWIF_PHR_RESTART_REQUESTED ((1) << ROGUE_FWIF_PHR_STATE_SHIFT) +/* A PHR triggered GPU reset has just finished */ +#define ROGUE_FWIF_PHR_RESTART_FINISHED ((2) << ROGUE_FWIF_PHR_STATE_SHIFT) +#define ROGUE_FWIF_PHR_RESTART_MASK \ + (ROGUE_FWIF_PHR_RESTART_REQUESTED | ROGUE_FWIF_PHR_RESTART_FINISHED) + +#define ROGUE_FWIF_PHR_MODE_OFF (0UL) +#define ROGUE_FWIF_PHR_MODE_RD_RESET (1UL) +#define ROGUE_FWIF_PHR_MODE_FULL_RESET (2UL) + +/* Firmware per-DM HWR states */ +/* DM is working if all flags are cleared */ +#define ROGUE_FWIF_DM_STATE_WORKING (0) +/* DM is idle and ready for HWR */ +#define ROGUE_FWIF_DM_STATE_READY_FOR_HWR BIT(0) +/* DM need to skip to next cmd before resuming processing */ +#define ROGUE_FWIF_DM_STATE_NEEDS_SKIP BIT(2) +/* DM need partial render cleanup before resuming processing */ +#define ROGUE_FWIF_DM_STATE_NEEDS_PR_CLEANUP BIT(3) +/* DM need to increment Recovery Count once fully recovered */ +#define ROGUE_FWIF_DM_STATE_NEEDS_TRACE_CLEAR BIT(4) +/* DM was identified as locking up and causing HWR */ +#define ROGUE_FWIF_DM_STATE_GUILTY_LOCKUP BIT(5) +/* DM was innocently affected by another lockup which caused HWR */ +#define ROGUE_FWIF_DM_STATE_INNOCENT_LOCKUP BIT(6) +/* DM was identified as over-running and causing HWR */ +#define ROGUE_FWIF_DM_STATE_GUILTY_OVERRUNING BIT(7) +/* DM was innocently affected by another DM over-running which caused HWR */ +#define ROGUE_FWIF_DM_STATE_INNOCENT_OVERRUNING BIT(8) +/* DM was forced into HWR as it delayed more important workloads */ +#define ROGUE_FWIF_DM_STATE_HARD_CONTEXT_SWITCH BIT(9) +/* DM was forced into HWR due to an uncorrected GPU ECC error */ +#define ROGUE_FWIF_DM_STATE_GPU_ECC_HWR BIT(10) + +/* Firmware's connection state */ +enum rogue_fwif_connection_fw_state { + /* Firmware is offline */ + ROGUE_FW_CONNECTION_FW_OFFLINE = 0, + /* Firmware is initialised */ + ROGUE_FW_CONNECTION_FW_READY, + /* Firmware connection is fully established */ + ROGUE_FW_CONNECTION_FW_ACTIVE, + /* Firmware is clearing up connection data*/ + ROGUE_FW_CONNECTION_FW_OFFLOADING, + ROGUE_FW_CONNECTION_FW_STATE_COUNT +}; + +/* OS' connection state */ +enum rogue_fwif_connection_os_state { + /* OS is offline */ + ROGUE_FW_CONNECTION_OS_OFFLINE = 0, + /* OS's KM driver is setup and waiting */ + ROGUE_FW_CONNECTION_OS_READY, + /* OS connection is fully established */ + ROGUE_FW_CONNECTION_OS_ACTIVE, + ROGUE_FW_CONNECTION_OS_STATE_COUNT +}; + +struct rogue_fwif_os_runtime_flags { + unsigned int os_state : 3; + unsigned int fl_ok : 1; + unsigned int fl_grow_pending : 1; + unsigned int isolated_os : 1; + unsigned int reserved : 26; +}; + +#define PVR_SLR_LOG_ENTRIES 10 +/* MAX_CLIENT_CCB_NAME not visible to this header */ +#define PVR_SLR_LOG_STRLEN 30 + +struct rogue_fwif_slr_entry { + aligned_u64 timestamp; + u32 fw_ctx_addr; + u32 num_ufos; + char ccb_name[PVR_SLR_LOG_STRLEN]; + char padding[2]; +} __aligned(8); + +#define MAX_THREAD_NUM 2 + +/* firmware trace control data */ +struct rogue_fwif_tracebuf { + u32 log_type; + struct rogue_fwif_tracebuf_space tracebuf[MAX_THREAD_NUM]; + /* + * Member initialised only when sTraceBuf is actually allocated (in + * ROGUETraceBufferInitOnDemandResources) + */ + u32 tracebuf_size_in_dwords; + /* Compatibility and other flags */ + u32 tracebuf_flags; +} __aligned(8); + +/* firmware system data shared with the Host driver */ +struct rogue_fwif_sysdata { + /* Configuration flags from host */ + u32 config_flags; + /* Extended configuration flags from host */ + u32 config_flags_ext; + enum rogue_fwif_pow_state pow_state; + u32 hw_perf_ridx; + u32 hw_perf_widx; + u32 hw_perf_wrap_count; + /* Constant after setup, needed in FW */ + u32 hw_perf_size; + /* The number of times the FW drops a packet due to buffer full */ + u32 hw_perf_drop_count; + + /* + * ui32HWPerfUt, ui32FirstDropOrdinal, ui32LastDropOrdinal only valid + * when FW is built with ROGUE_HWPERF_UTILIZATION & + * ROGUE_HWPERF_DROP_TRACKING defined in rogue_fw_hwperf.c + */ + /* Buffer utilisation, high watermark of bytes in use */ + u32 hw_perf_ut; + /* The ordinal of the first packet the FW dropped */ + u32 first_drop_ordinal; + /* The ordinal of the last packet the FW dropped */ + u32 last_drop_ordinal; + /* State flags for each Operating System mirrored from Fw coremem */ + struct rogue_fwif_os_runtime_flags + os_runtime_flags_mirror[ROGUE_FW_MAX_NUM_OS]; + + struct rogue_fw_fault_info fault_info[ROGUE_FWIF_FWFAULTINFO_MAX]; + u32 fw_faults; + u32 cr_poll_addr[MAX_THREAD_NUM]; + u32 cr_poll_mask[MAX_THREAD_NUM]; + u32 cr_poll_count[MAX_THREAD_NUM]; + aligned_u64 start_idle_time; + +#if defined(SUPPORT_ROGUE_FW_STATS_FRAMEWORK) +# define ROGUE_FWIF_STATS_FRAMEWORK_LINESIZE (8) +# define ROGUE_FWIF_STATS_FRAMEWORK_MAX \ + (2048 * ROGUE_FWIF_STATS_FRAMEWORK_LINESIZE) + u32 fw_stats_buf[ROGUE_FWIF_STATS_FRAMEWORK_MAX] __aligned(8); +#endif + u32 hwr_state_flags; + u32 hwr_recovery_flags[PVR_FWIF_DM_MAX]; + /* Compatibility and other flags */ + u32 fw_sys_data_flags; + /* Identify whether MC config is P-P or P-S */ + u32 mc_config; +} __aligned(8); + +/* per-os firmware shared data */ +struct rogue_fwif_osdata { + /* Configuration flags from an OS */ + u32 fw_os_config_flags; + /* Markers to signal that the host should perform a full sync check */ + u32 fw_sync_check_mark; + u32 host_sync_check_mark; + + u32 forced_updates_requested; + u8 slr_log_wp; + struct rogue_fwif_slr_entry slr_log_first; + struct rogue_fwif_slr_entry slr_log[PVR_SLR_LOG_ENTRIES]; + aligned_u64 last_forced_update_time; + + /* Interrupt count from Threads > */ + u32 interrupt_count[MAX_THREAD_NUM]; + u32 kccb_cmds_executed; + u32 power_sync_fw_addr; + /* Compatibility and other flags */ + u32 fw_os_data_flags; + u32 padding; +} __aligned(8); + +/* Firmware trace time-stamp field breakup */ + +/* ROGUE_CR_TIMER register read (48 bits) value*/ +#define ROGUE_FWT_TIMESTAMP_TIME_SHIFT (0U) +#define ROGUE_FWT_TIMESTAMP_TIME_CLRMSK (0xFFFF000000000000ull) + +/* Extra debug-info (16 bits) */ +#define ROGUE_FWT_TIMESTAMP_DEBUG_INFO_SHIFT (48U) +#define ROGUE_FWT_TIMESTAMP_DEBUG_INFO_CLRMSK ~ROGUE_FWT_TIMESTAMP_TIME_CLRMSK + +/* Debug-info sub-fields */ +/* + * Bit 0: ROGUE_CR_EVENT_STATUS_MMU_PAGE_FAULT bit from ROGUE_CR_EVENT_STATUS + * register + */ +#define ROGUE_FWT_DEBUG_INFO_MMU_PAGE_FAULT_SHIFT (0U) +#define ROGUE_FWT_DEBUG_INFO_MMU_PAGE_FAULT_SET \ + BIT(ROGUE_FWT_DEBUG_INFO_MMU_PAGE_FAULT_SHIFT) + +/* Bit 1: ROGUE_CR_BIF_MMU_ENTRY_PENDING bit from ROGUE_CR_BIF_MMU_ENTRY register */ +#define ROGUE_FWT_DEBUG_INFO_MMU_ENTRY_PENDING_SHIFT (1U) +#define ROGUE_FWT_DEBUG_INFO_MMU_ENTRY_PENDING_SET \ + BIT(ROGUE_FWT_DEBUG_INFO_MMU_ENTRY_PENDING_SHIFT) + +/* Bit 2: ROGUE_CR_SLAVE_EVENT register is non-zero */ +#define ROGUE_FWT_DEBUG_INFO_SLAVE_EVENTS_SHIFT (2U) +#define ROGUE_FWT_DEBUG_INFO_SLAVE_EVENTS_SET \ + BIT(ROGUE_FWT_DEBUG_INFO_SLAVE_EVENTS_SHIFT) + +/* Bit 3-15: Unused bits */ + +#define ROGUE_FWT_DEBUG_INFO_STR_MAXLEN 64 +#define ROGUE_FWT_DEBUG_INFO_STR_PREPEND " (debug info: " +#define ROGUE_FWT_DEBUG_INFO_STR_APPEND ")" + +/* + ****************************************************************************** + * HWR Data + ****************************************************************************** + */ +enum rogue_hwrtype { + ROGUE_HWRTYPE_UNKNOWNFAILURE = 0, + ROGUE_HWRTYPE_OVERRUN = 1, + ROGUE_HWRTYPE_POLLFAILURE = 2, + ROGUE_HWRTYPE_BIF0FAULT = 3, + ROGUE_HWRTYPE_BIF1FAULT = 4, + ROGUE_HWRTYPE_TEXASBIF0FAULT = 5, + ROGUE_HWRTYPE_MMUFAULT = 6, + ROGUE_HWRTYPE_MMUMETAFAULT = 7, + ROGUE_HWRTYPE_MIPSTLBFAULT = 8, + ROGUE_HWRTYPE_ECCFAULT = 9, + ROGUE_HWRTYPE_MMURISCVFAULT = 10, +}; + +#define ROGUE_FWIF_HWRTYPE_BIF_BANK_GET(hwr_type) \ + (((hwr_type) == ROGUE_HWRTYPE_BIF0FAULT) ? 0 : 1) + +#define ROGUE_FWIF_HWRTYPE_PAGE_FAULT_GET(hwr_type) \ + ((((hwr_type) == ROGUE_HWRTYPE_BIF0FAULT) || \ + ((hwr_type) == ROGUE_HWRTYPE_BIF1FAULT) || \ + ((hwr_type) == ROGUE_HWRTYPE_TEXASBIF0FAULT) || \ + ((hwr_type) == ROGUE_HWRTYPE_MMUFAULT) || \ + ((hwr_type) == ROGUE_HWRTYPE_MMUMETAFAULT) || \ + ((hwr_type) == ROGUE_HWRTYPE_MIPSTLBFAULT) || \ + ((hwr_type) == ROGUE_HWRTYPE_MMURISCVFAULT)) \ + ? true \ + : false) + +struct rogue_bifinfo { + aligned_u64 bif_req_status; + aligned_u64 bif_mmu_status; + aligned_u64 pc_address; /* phys address of the page catalogue */ + aligned_u64 reserved; +}; + +struct rogue_eccinfo { + u32 fault_gpu; +}; + +struct rogue_mmuinfo { + aligned_u64 mmu_status[2]; + aligned_u64 pc_address; /* phys address of the page catalogue */ + aligned_u64 reserved; +}; + +struct rogue_pollinfo { + u32 thread_num; + u32 cr_poll_addr; + u32 cr_poll_mask; + u32 cr_poll_last_value; + aligned_u64 reserved; +} __aligned(8); + +struct rogue_tlbinfo { + u32 bad_addr; + u32 entry_lo; +}; + +struct rogue_hwrinfo { + union { + struct rogue_bifinfo bif_info; + struct rogue_mmuinfo mmu_info; + struct rogue_pollinfo poll_info; + struct rogue_tlbinfo tlb_info; + struct rogue_eccinfo ecc_info; + } hwr_data; + + aligned_u64 cr_timer; + aligned_u64 os_timer; + u32 frame_num; + u32 pid; + u32 active_hwrt_data; + u32 hwr_number; + u32 event_status; + u32 hwr_recovery_flags; + enum rogue_hwrtype hwr_type; + u32 dm; + u32 core_id; + aligned_u64 cr_time_of_kick; + aligned_u64 cr_time_hw_reset_start; + aligned_u64 cr_time_hw_reset_finish; + aligned_u64 cr_time_freelist_ready; + aligned_u64 reserved[2]; +} __aligned(8); + +/* Number of first HWR logs recorded (never overwritten by newer logs) */ +#define ROGUE_FWIF_HWINFO_MAX_FIRST 8U +/* Number of latest HWR logs (older logs are overwritten by newer logs) */ +#define ROGUE_FWIF_HWINFO_MAX_LAST 8U +/* Total number of HWR logs stored in a buffer */ +#define ROGUE_FWIF_HWINFO_MAX \ + (ROGUE_FWIF_HWINFO_MAX_FIRST + ROGUE_FWIF_HWINFO_MAX_LAST) +/* Index of the last log in the HWR log buffer */ +#define ROGUE_FWIF_HWINFO_LAST_INDEX (ROGUE_FWIF_HWINFO_MAX - 1U) + +struct rogue_fwif_hwrinfobuf { + struct rogue_hwrinfo hwr_info[ROGUE_FWIF_HWINFO_MAX]; + u32 hwr_counter; + u32 write_index; + u32 dd_req_count; + u32 hwr_info_buf_flags; /* Compatibility and other flags */ + u32 hwr_dm_locked_up_count[PVR_FWIF_DM_MAX]; + u32 hwr_dm_overran_count[PVR_FWIF_DM_MAX]; + u32 hwr_dm_recovered_count[PVR_FWIF_DM_MAX]; + u32 hwr_dm_false_detect_count[PVR_FWIF_DM_MAX]; +} __aligned(8); + +#define ROGUE_FWIF_CTXSWITCH_PROFILE_FAST_EN (1) +#define ROGUE_FWIF_CTXSWITCH_PROFILE_MEDIUM_EN (2) +#define ROGUE_FWIF_CTXSWITCH_PROFILE_SLOW_EN (3) +#define ROGUE_FWIF_CTXSWITCH_PROFILE_NODELAY_EN (4) + +#define ROGUE_FWIF_CDM_ARBITRATION_TASK_DEMAND_EN (1) +#define ROGUE_FWIF_CDM_ARBITRATION_ROUND_ROBIN_EN (2) + +#define ROGUE_FWIF_ISP_SCHEDMODE_VER1_IPP (1) +#define ROGUE_FWIF_ISP_SCHEDMODE_VER2_ISP (2) +/* + ****************************************************************************** + * ROGUE firmware Init Config Data + ****************************************************************************** + */ + +/* Flag definitions affecting the firmware globally */ +#define ROGUE_FWIF_INICFG_CTXSWITCH_MODE_RAND BIT(0) +#define ROGUE_FWIF_INICFG_CTXSWITCH_SRESET_EN BIT(1) +#define ROGUE_FWIF_INICFG_HWPERF_EN BIT(2) +#define ROGUE_FWIF_INICFG_DM_KILL_MODE_RAND_EN BIT(3) +#define ROGUE_FWIF_INICFG_POW_RASCALDUST BIT(4) +/* Bit 5 is reserved. */ +#define ROGUE_FWIF_INICFG_FBCDC_V3_1_EN BIT(6) +#define ROGUE_FWIF_INICFG_CHECK_MLIST_EN BIT(7) +#define ROGUE_FWIF_INICFG_DISABLE_CLKGATING_EN BIT(8) +/* Bit 9 is reserved. */ +/* Bit 10 is reserved. */ +/* Bit 11 is reserved. */ +#define ROGUE_FWIF_INICFG_REGCONFIG_EN BIT(12) +#define ROGUE_FWIF_INICFG_ASSERT_ON_OUTOFMEMORY BIT(13) +#define ROGUE_FWIF_INICFG_HWP_DISABLE_FILTER BIT(14) +/* Bit 15 is reserved. */ +#define ROGUE_FWIF_INICFG_CTXSWITCH_PROFILE_SHIFT (16) +#define ROGUE_FWIF_INICFG_CTXSWITCH_PROFILE_FAST \ + (ROGUE_FWIF_CTXSWITCH_PROFILE_FAST_EN \ + << ROGUE_FWIF_INICFG_CTXSWITCH_PROFILE_SHIFT) +#define ROGUE_FWIF_INICFG_CTXSWITCH_PROFILE_MEDIUM \ + (ROGUE_FWIF_CTXSWITCH_PROFILE_MEDIUM_EN \ + << ROGUE_FWIF_INICFG_CTXSWITCH_PROFILE_SHIFT) +#define ROGUE_FWIF_INICFG_CTXSWITCH_PROFILE_SLOW \ + (ROGUE_FWIF_CTXSWITCH_PROFILE_SLOW_EN \ + << ROGUE_FWIF_INICFG_CTXSWITCH_PROFILE_SHIFT) +#define ROGUE_FWIF_INICFG_CTXSWITCH_PROFILE_NODELAY \ + (ROGUE_FWIF_CTXSWITCH_PROFILE_NODELAY_EN \ + << ROGUE_FWIF_INICFG_CTXSWITCH_PROFILE_SHIFT) +#define ROGUE_FWIF_INICFG_CTXSWITCH_PROFILE_MASK \ + (7 << ROGUE_FWIF_INICFG_CTXSWITCH_PROFILE_SHIFT) +#define ROGUE_FWIF_INICFG_DISABLE_DM_OVERLAP BIT(19) +#define ROGUE_FWIF_INICFG_ASSERT_ON_HWR_TRIGGER BIT(20) +#define ROGUE_FWIF_INICFG_FABRIC_COHERENCY_ENABLED BIT(21) +#define ROGUE_FWIF_INICFG_VALIDATE_IRQ BIT(22) +#define ROGUE_FWIF_INICFG_DISABLE_PDP_EN BIT(23) +#define ROGUE_FWIF_INICFG_SPU_POWER_STATE_MASK_CHANGE_EN BIT(24) +#define ROGUE_FWIF_INICFG_WORKEST BIT(25) +#define ROGUE_FWIF_INICFG_PDVFS BIT(26) +#define ROGUE_FWIF_INICFG_CDM_ARBITRATION_SHIFT (27) +#define ROGUE_FWIF_INICFG_CDM_ARBITRATION_TASK_DEMAND \ + (ROGUE_FWIF_CDM_ARBITRATION_TASK_DEMAND_EN \ + << ROGUE_FWIF_INICFG_CDM_ARBITRATION_SHIFT) +#define ROGUE_FWIF_INICFG_CDM_ARBITRATION_ROUND_ROBIN \ + (ROGUE_FWIF_CDM_ARBITRATION_ROUND_ROBIN_EN \ + << ROGUE_FWIF_INICFG_CDM_ARBITRATION_SHIFT) +#define ROGUE_FWIF_INICFG_CDM_ARBITRATION_MASK \ + (3 << ROGUE_FWIF_INICFG_CDM_ARBITRATION_SHIFT) +#define ROGUE_FWIF_INICFG_ISPSCHEDMODE_SHIFT (29) +#define ROGUE_FWIF_INICFG_ISPSCHEDMODE_NONE (0) +#define ROGUE_FWIF_INICFG_ISPSCHEDMODE_VER1_IPP \ + (ROGUE_FWIF_ISP_SCHEDMODE_VER1_IPP \ + << ROGUE_FWIF_INICFG_ISPSCHEDMODE_SHIFT) +#define ROGUE_FWIF_INICFG_ISPSCHEDMODE_VER2_ISP \ + (ROGUE_FWIF_ISP_SCHEDMODE_VER2_ISP \ + << ROGUE_FWIF_INICFG_ISPSCHEDMODE_SHIFT) +#define ROGUE_FWIF_INICFG_ISPSCHEDMODE_MASK \ + (ROGUE_FWIF_INICFG_ISPSCHEDMODE_VER1_IPP | \ + ROGUE_FWIF_INICFG_ISPSCHEDMODE_VER2_ISP) +#define ROGUE_FWIF_INICFG_VALIDATE_SOCUSC_TIMER BIT(31) + +#define ROGUE_FWIF_INICFG_ALL (0xFFFFFFFFU) + +/* Extended Flag definitions affecting the firmware globally */ +#define ROGUE_FWIF_INICFG_EXT_TFBC_CONTROL_SHIFT (0) +/* [7] YUV10 override + * [6:4] Quality + * [3] Quality enable + * [2:1] Compression scheme + * [0] Lossy group + */ +#define ROGUE_FWIF_INICFG_EXT_TFBC_CONTROL_MASK (0xFF) +#define ROGUE_FWIF_INICFG_EXT_ALL (ROGUE_FWIF_INICFG_EXT_TFBC_CONTROL_MASK) + +/* Flag definitions affecting only workloads submitted by a particular OS */ +#define ROGUE_FWIF_INICFG_OS_CTXSWITCH_TDM_EN BIT(0) +#define ROGUE_FWIF_INICFG_OS_CTXSWITCH_GEOM_EN BIT(1) +#define ROGUE_FWIF_INICFG_OS_CTXSWITCH_FRAG_EN BIT(2) +#define ROGUE_FWIF_INICFG_OS_CTXSWITCH_CDM_EN BIT(3) + +#define ROGUE_FWIF_INICFG_OS_LOW_PRIO_CS_TDM BIT(4) +#define ROGUE_FWIF_INICFG_OS_LOW_PRIO_CS_GEOM BIT(5) +#define ROGUE_FWIF_INICFG_OS_LOW_PRIO_CS_FRAG BIT(6) +#define ROGUE_FWIF_INICFG_OS_LOW_PRIO_CS_CDM BIT(7) + +#define ROGUE_FWIF_INICFG_OS_ALL (0xFF) + +#define ROGUE_FWIF_INICFG_OS_CTXSWITCH_DM_ALL \ + (ROGUE_FWIF_INICFG_OS_CTXSWITCH_TDM_EN | \ + ROGUE_FWIF_INICFG_OS_CTXSWITCH_GEOM_EN | \ + ROGUE_FWIF_INICFG_OS_CTXSWITCH_FRAG_EN | \ + ROGUE_FWIF_INICFG_OS_CTXSWITCH_CDM_EN) + +#define ROGUE_FWIF_INICFG_OS_CTXSWITCH_CLRMSK \ + ~(ROGUE_FWIF_INICFG_OS_CTXSWITCH_DM_ALL) + +#define ROGUE_FWIF_FILTCFG_TRUNCATE_HALF BIT(3) +#define ROGUE_FWIF_FILTCFG_TRUNCATE_INT BIT(2) +#define ROGUE_FWIF_FILTCFG_NEW_FILTER_MODE BIT(1) + +enum rogue_activepm_conf { + ROGUE_ACTIVEPM_FORCE_OFF = 0, + ROGUE_ACTIVEPM_FORCE_ON = 1, + ROGUE_ACTIVEPM_DEFAULT = 2 +}; + +enum rogue_rd_power_island_conf { + ROGUE_RD_POWER_ISLAND_FORCE_OFF = 0, + ROGUE_RD_POWER_ISLAND_FORCE_ON = 1, + ROGUE_RD_POWER_ISLAND_DEFAULT = 2 +}; + +struct rogue_fw_register_list { + /* Register number */ + u16 reg_num; + /* Indirect register number (or 0 if not used) */ + u16 indirect_reg_num; + /* Start value for indirect register */ + u16 indirect_start_val; + /* End value for indirect register */ + u16 indirect_end_val; +}; + +struct rogue_fwif_dllist_node { + u32 p; + u32 n; +}; + +/* + * This number is used to represent an invalid page catalogue physical address + */ +#define ROGUE_FWIF_INVALID_PC_PHYADDR 0xFFFFFFFFFFFFFFFFLLU + +/* This number is used to represent unallocated page catalog base register */ +#define ROGUE_FW_BIF_INVALID_PCSET 0xFFFFFFFFU + +/* Firmware memory context. */ +struct rogue_fwif_fwmemcontext { + /* device physical address of context's page catalogue */ + aligned_u64 pc_dev_paddr; + /* + * associated page catalog base register (ROGUE_FW_BIF_INVALID_PCSET == + * unallocated) + */ + u32 page_cat_base_reg_set; + /* breakpoint address */ + u32 breakpoint_addr; + /* breakpoint handler address */ + u32 bp_handler_addr; + /* DM and enable control for BP */ + u32 breakpoint_ctl; + /* Compatibility and other flags */ + u32 fw_mem_ctx_flags; + u32 padding; +} __aligned(8); + +/* + * FW context state flags + */ +#define ROGUE_FWIF_CONTEXT_FLAGS_NEED_RESUME (0x00000001U) +#define ROGUE_FWIF_CONTEXT_FLAGS_MC_NEED_RESUME_MASKFULL (0x000000FFU) +#define ROGUE_FWIF_CONTEXT_FLAGS_TDM_HEADER_STALE (0x00000100U) +#define ROGUE_FWIF_CONTEXT_FLAGS_LAST_KICK_SECURE (0x00000200U) + +#define ROGUE_NUM_GEOM_CORES_MAX 4 + +/* + * FW-accessible TA state which must be written out to memory on context store + */ +struct rogue_fwif_geom_ctx_state_per_geom { + /* To store in mid-TA */ + aligned_u64 geom_reg_vdm_call_stack_pointer; + /* Initial value (in case is 'lost' due to a lock-up */ + aligned_u64 geom_reg_vdm_call_stack_pointer_init; + u32 geom_reg_vbs_so_prim[4]; + u16 geom_current_idx; + u16 padding[3]; +} __aligned(8); + +struct rogue_fwif_geom_ctx_state { + /* FW-accessible TA state which must be written out to memory on context store */ + struct rogue_fwif_geom_ctx_state_per_geom geom_core[ROGUE_NUM_GEOM_CORES_MAX]; +} __aligned(8); + +/* + * FW-accessible ISP state which must be written out to memory on context store + */ +struct rogue_fwif_frag_ctx_state { + u32 frag_reg_pm_deallocated_mask_status; + u32 frag_reg_dm_pds_mtilefree_status; + /* Compatibility and other flags */ + u32 ctx_state_flags; + /* + * frag_reg_isp_store should be the last element of the structure as this + * is an array whose size is determined at runtime after detecting the + * ROGUE core + */ + u32 frag_reg_isp_store[]; +} __aligned(8); + +#define ROGUE_FWIF_CTX_USING_BUFFER_A (0) +#define ROGUE_FWIF_CTX_USING_BUFFER_B (1U) + +struct rogue_fwif_compute_ctx_state { + u32 ctx_state_flags; /* Target buffer and other flags */ +}; + +struct rogue_fwif_fwcommoncontext { + /* CCB details for this firmware context */ + u32 ccbctl_fw_addr; /* CCB control */ + u32 ccb_fw_addr; /* CCB base */ + struct rogue_fwif_dma_addr ccb_meta_dma_addr; + + /* Context suspend state */ + /* geom/frag context suspend state, read/written by FW */ + u32 context_state_addr __aligned(8); + + /* Flags e.g. for context switching */ + u32 fw_com_ctx_flags; + u32 priority; + u32 priority_seq_num; + + /* Framework state */ + /* Register updates for Framework */ + u32 rf_cmd_addr __aligned(8); + + /* Statistic updates waiting to be passed back to the host... */ + /* True when some stats are pending */ + bool stats_pending __aligned(4); + /* Number of stores on this context since last update */ + s32 stats_num_stores; + /* Number of OOMs on this context since last update */ + s32 stats_num_out_of_memory; + /* Number of PRs on this context since last update */ + s32 stats_num_partial_renders; + /* Data Master type */ + u32 dm; + /* Device Virtual Address of the signal the context is waiting on */ + aligned_u64 wait_signal_address; + /* List entry for the wait-signal list */ + struct rogue_fwif_dllist_node wait_signal_node __aligned(8); + /* List entry for the buffer stalled list */ + struct rogue_fwif_dllist_node buf_stalled_node __aligned(8); + /* Address of the circular buffer queue pointers */ + aligned_u64 cbuf_queue_ctrl_addr; + + aligned_u64 robustness_address; + /* Max HWR deadline limit in ms */ + u32 max_deadline_ms; + /* Following HWR circular buffer read-offset needs resetting */ + bool read_offset_needs_reset; + + /* List entry for the waiting list */ + struct rogue_fwif_dllist_node waiting_node __aligned(8); + /* List entry for the run list */ + struct rogue_fwif_dllist_node run_node __aligned(8); + /* UFO that last failed (or NULL) */ + struct rogue_fwif_ufo last_failed_ufo; + + /* Memory context */ + u32 fw_mem_context_fw_addr; + + /* References to the host side originators */ + /* the Server Common Context */ + u32 server_common_context_id; + /* associated process ID */ + u32 pid; + + /* True when Geom DM OOM is not allowed */ + bool geom_oom_disabled __aligned(4); +} __aligned(8); + +/* Firmware render context. */ +struct rogue_fwif_fwrendercontext { + /* Geometry firmware context. */ + struct rogue_fwif_fwcommoncontext geom_context; + /* Fragment firmware context. */ + struct rogue_fwif_fwcommoncontext frag_context; + + struct rogue_fwif_static_rendercontext_state static_render_context_state; + + /* Number of commands submitted to the WorkEst FW CCB */ + u32 work_est_ccb_submitted; + + /* Compatibility and other flags */ + u32 fw_render_ctx_flags; +} __aligned(8); + +/* Firmware compute context. */ +struct rogue_fwif_fwcomputecontext { + /* Firmware context for the CDM */ + struct rogue_fwif_fwcommoncontext cdm_context; + + struct rogue_fwif_static_computecontext_state + static_compute_context_state; + + /* Number of commands submitted to the WorkEst FW CCB */ + u32 work_est_ccb_submitted; + + /* Compatibility and other flags */ + u32 compute_ctx_flags; + + u32 wgp_state; + u32 wgp_checksum; + u32 core_mask_a; + u32 core_mask_b; +} __aligned(8); + +/* Firmware TDM context. */ +struct rogue_fwif_fwtdmcontext { + /* Firmware context for the TDM */ + struct rogue_fwif_fwcommoncontext tdm_context; + + /* Number of commands submitted to the WorkEst FW CCB */ + u32 work_est_ccb_submitted; +} __aligned(8); + +/* Firmware TQ3D context. */ +struct rogue_fwif_fwtransfercontext { + /* Firmware context for TQ3D. */ + struct rogue_fwif_fwcommoncontext tq_context; +} __aligned(8); + +/* + ****************************************************************************** + * Defines for CMD_TYPE corruption detection and forward compatibility check + ****************************************************************************** + */ + +/* + * CMD_TYPE 32bit contains: + * 31:16 Reserved for magic value to detect corruption (16 bits) + * 15 Reserved for ROGUE_CCB_TYPE_TASK (1 bit) + * 14:0 Bits available for CMD_TYPEs (15 bits) + */ + +/* Magic value to detect corruption */ +#define ROGUE_CMD_MAGIC_DWORD (0x2ABC) +#define ROGUE_CMD_MAGIC_DWORD_MASK (0xFFFF0000U) +#define ROGUE_CMD_MAGIC_DWORD_SHIFT (16U) +#define ROGUE_CMD_MAGIC_DWORD_SHIFTED \ + (ROGUE_CMD_MAGIC_DWORD << ROGUE_CMD_MAGIC_DWORD_SHIFT) + +/* Kernel CCB control for ROGUE */ +struct rogue_fwif_ccb_ctl { + /* write offset into array of commands (MUST be aligned to 16 bytes!) */ + u32 write_offset; + /* Padding to ensure read and write offsets are in separate cache lines. */ + u8 padding[128 - sizeof(u32)]; + /* read offset into array of commands */ + u32 read_offset; + /* Offset wrapping mask (Total capacity of the CCB - 1) */ + u32 wrap_mask; + /* size of each command in bytes */ + u32 cmd_size; + u32 padding2; +} __aligned(8); + +/* Kernel CCB command structure for ROGUE */ + +#define ROGUE_FWIF_MMUCACHEDATA_FLAGS_PT (0x1U) /* MMU_CTRL_INVAL_PT_EN */ +#define ROGUE_FWIF_MMUCACHEDATA_FLAGS_PD (0x2U) /* MMU_CTRL_INVAL_PD_EN */ +#define ROGUE_FWIF_MMUCACHEDATA_FLAGS_PC (0x4U) /* MMU_CTRL_INVAL_PC_EN */ + +/* + * can't use PM_TLB0 bit from BIFPM_CTRL reg because it collides with PT + * bit from BIF_CTRL reg + */ +#define ROGUE_FWIF_MMUCACHEDATA_FLAGS_PMTLB (0x10) +/* BIF_CTRL_INVAL_TLB1_EN */ +#define ROGUE_FWIF_MMUCACHEDATA_FLAGS_TLB \ + (ROGUE_FWIF_MMUCACHEDATA_FLAGS_PMTLB | 0x8) +/* MMU_CTRL_INVAL_ALL_CONTEXTS_EN */ +#define ROGUE_FWIF_MMUCACHEDATA_FLAGS_CTX_ALL (0x800) + +/* indicates FW should interrupt the host */ +#define ROGUE_FWIF_MMUCACHEDATA_FLAGS_INTERRUPT (0x4000000U) + +struct rogue_fwif_mmucachedata { + u32 cache_flags; + u32 mmu_cache_sync_fw_addr; + u32 mmu_cache_sync_update_value; +}; + +#define ROGUE_FWIF_BPDATA_FLAGS_ENABLE BIT(0) +#define ROGUE_FWIF_BPDATA_FLAGS_WRITE BIT(1) +#define ROGUE_FWIF_BPDATA_FLAGS_CTL BIT(2) +#define ROGUE_FWIF_BPDATA_FLAGS_REGS BIT(3) + +struct rogue_fwif_bpdata { + /* Memory context */ + u32 fw_mem_context_fw_addr; + /* Breakpoint address */ + u32 bp_addr; + /* Breakpoint handler */ + u32 bp_handler_addr; + /* Breakpoint control */ + u32 bp_dm; + u32 bp_data_flags; + /* Number of temporary registers to overallocate */ + u32 temp_regs; + /* Number of shared registers to overallocate */ + u32 shared_regs; + /* DM associated with the breakpoint */ + u32 dm; +}; + +#define ROGUE_FWIF_KCCB_CMD_KICK_DATA_MAX_NUM_CLEANUP_CTLS \ + (ROGUE_FWIF_PRBUFFER_MAXSUPPORTED + 1U) /* +1 is RTDATASET cleanup */ + +struct rogue_fwif_kccb_cmd_kick_data { + /* address of the firmware context */ + u32 context_fw_addr; + /* Client CCB woff update */ + u32 client_woff_update; + /* Client CCB wrap mask update after CCCB growth */ + u32 client_wrap_mask_update; + /* number of CleanupCtl pointers attached */ + u32 num_cleanup_ctl; + /* CleanupCtl structures associated with command */ + u32 cleanup_ctl_fw_addr + [ROGUE_FWIF_KCCB_CMD_KICK_DATA_MAX_NUM_CLEANUP_CTLS]; + /* + * offset to the CmdHeader which houses the workload estimation kick + * data. + */ + u32 work_est_cmd_header_offset; +}; + +struct rogue_fwif_kccb_cmd_combined_geom_frag_kick_data { + struct rogue_fwif_kccb_cmd_kick_data geom_cmd_kick_data; + struct rogue_fwif_kccb_cmd_kick_data frag_cmd_kick_data; +}; + +struct rogue_fwif_kccb_cmd_force_update_data { + /* address of the firmware context */ + u32 context_fw_addr; + /* Client CCB fence offset */ + u32 ccb_fence_offset; +}; + +enum rogue_fwif_cleanup_type { + /* FW common context cleanup */ + ROGUE_FWIF_CLEANUP_FWCOMMONCONTEXT, + /* FW HW RT data cleanup */ + ROGUE_FWIF_CLEANUP_HWRTDATA, + /* FW freelist cleanup */ + ROGUE_FWIF_CLEANUP_FREELIST, + /* FW ZS Buffer cleanup */ + ROGUE_FWIF_CLEANUP_ZSBUFFER, +}; + +struct rogue_fwif_cleanup_request { + /* Cleanup type */ + enum rogue_fwif_cleanup_type cleanup_type; + union { + /* FW common context to cleanup */ + u32 context_fw_addr; + /* HW RT to cleanup */ + u32 hwrt_data_fw_addr; + /* Freelist to cleanup */ + u32 freelist_fw_addr; + /* ZS Buffer to cleanup */ + u32 zs_buffer_fw_addr; + } cleanup_data; +}; + +enum rogue_fwif_power_type { + ROGUE_FWIF_POW_OFF_REQ = 1, + ROGUE_FWIF_POW_FORCED_IDLE_REQ, + ROGUE_FWIF_POW_NUM_UNITS_CHANGE, + ROGUE_FWIF_POW_APM_LATENCY_CHANGE +}; + +enum rogue_fwif_power_force_idle_type { + ROGUE_FWIF_POWER_FORCE_IDLE = 1, + ROGUE_FWIF_POWER_CANCEL_FORCED_IDLE, + ROGUE_FWIF_POWER_HOST_TIMEOUT, +}; + +struct rogue_fwif_power_request { + /* Type of power request */ + enum rogue_fwif_power_type pow_type; + union { + /* Number of active Dusts */ + u32 num_of_dusts; + /* If the operation is mandatory */ + bool forced __aligned(4); + /* + * Type of Request. Consolidating Force Idle, Cancel Forced + * Idle, Host Timeout + */ + enum rogue_fwif_power_force_idle_type pow_request_type; + } power_req_data; +}; + +struct rogue_fwif_slcflushinvaldata { + /* Context to fence on (only useful when bDMContext == TRUE) */ + u32 context_fw_addr; + /* Invalidate the cache as well as flushing */ + bool inval __aligned(4); + /* The data to flush/invalidate belongs to a specific DM context */ + bool dm_context __aligned(4); + /* Optional address of range (only useful when bDMContext == FALSE) */ + aligned_u64 address; + /* Optional size of range (only useful when bDMContext == FALSE) */ + aligned_u64 size; +}; + +enum rogue_fwif_hwperf_update_config { + ROGUE_FWIF_HWPERF_CTRL_TOGGLE = 0, + ROGUE_FWIF_HWPERF_CTRL_SET = 1, + ROGUE_FWIF_HWPERF_CTRL_EMIT_FEATURES_EV = 2 +}; + +struct rogue_fwif_hwperf_ctrl { + enum rogue_fwif_hwperf_update_config opcode; /* Control operation code */ + aligned_u64 mask; /* Mask of events to toggle */ +}; + +struct rogue_fwif_hwperf_config_enable_blks { + /* Number of ROGUE_HWPERF_CONFIG_MUX_CNTBLK in the array */ + u32 num_blocks; + /* Address of the ROGUE_HWPERF_CONFIG_MUX_CNTBLK array */ + u32 block_configs_fw_addr; +}; + +struct rogue_fwif_hwperf_config_da_blks { + /* Number of ROGUE_HWPERF_CONFIG_CNTBLK in the array */ + u32 num_blocks; + /* Address of the ROGUE_HWPERF_CONFIG_CNTBLK array */ + u32 block_configs_fw_addr; +}; + +struct rogue_fwif_coreclkspeedchange_data { + u32 new_clock_speed; /* New clock speed */ +}; + +#define ROGUE_FWIF_HWPERF_CTRL_BLKS_MAX 16 + +struct rogue_fwif_hwperf_ctrl_blks { + bool enable; + /* Number of block IDs in the array */ + u32 num_blocks; + /* Array of ROGUE_HWPERF_CNTBLK_ID values */ + u16 block_ids[ROGUE_FWIF_HWPERF_CTRL_BLKS_MAX]; +}; + +struct rogue_fwif_hwperf_select_custom_cntrs { + u16 custom_block; + u16 num_counters; + u32 custom_counter_ids_fw_addr; +}; + +struct rogue_fwif_zsbuffer_backing_data { + u32 zs_buffer_fw_addr; /* ZS-Buffer FW address */ + + bool done __aligned(4); /* action backing/unbacking succeeded */ +}; + +struct rogue_fwif_freelist_gs_data { + /* Freelist FW address */ + u32 freelist_fw_addr; + /* Amount of the Freelist change */ + u32 delta_pages; + /* New amount of pages on the freelist (including ready pages) */ + u32 new_pages; + /* Number of ready pages to be held in reserve until OOM */ + u32 ready_pages; +}; + +#define MAX_FREELISTS_SIZE 3 +#define MAX_HW_GEOM_FRAG_CONTEXTS_SIZE 3 + +#define ROGUE_FWIF_MAX_FREELISTS_TO_RECONSTRUCT \ + (MAX_HW_GEOM_FRAG_CONTEXTS_SIZE * MAX_FREELISTS_SIZE * 2U) +#define ROGUE_FWIF_FREELISTS_RECONSTRUCTION_FAILED_FLAG 0x80000000U + +struct rogue_fwif_freelists_reconstruction_data { + u32 freelist_count; + u32 freelist_ids[ROGUE_FWIF_MAX_FREELISTS_TO_RECONSTRUCT]; +}; + +struct rogue_fwif_write_offset_update_data { + /* + * Context to that may need to be resumed following write offset update + */ + u32 context_fw_addr; +} __aligned(8); + +/* + ****************************************************************************** + * Proactive DVFS Structures + ****************************************************************************** + */ +#define NUM_OPP_VALUES 16 + +struct pdvfs_opp { + u32 volt; /* V */ + u32 freq; /* Hz */ +} __aligned(8); + +struct rogue_fwif_pdvfs_opp { + struct pdvfs_opp opp_values[NUM_OPP_VALUES]; + u32 min_opp_point; + u32 max_opp_point; +} __aligned(8); + +struct rogue_fwif_pdvfs_max_freq_data { + u32 max_opp_point; +} __aligned(8); + +struct rogue_fwif_pdvfs_min_freq_data { + u32 min_opp_point; +} __aligned(8); + +/* + ****************************************************************************** + * Register configuration structures + ****************************************************************************** + */ + +#define ROGUE_FWIF_REG_CFG_MAX_SIZE 512 + +enum rogue_fwif_regdata_cmd_type { + ROGUE_FWIF_REGCFG_CMD_ADD = 101, + ROGUE_FWIF_REGCFG_CMD_CLEAR = 102, + ROGUE_FWIF_REGCFG_CMD_ENABLE = 103, + ROGUE_FWIF_REGCFG_CMD_DISABLE = 104 +}; + +enum rogue_fwif_reg_cfg_type { + /* Sidekick power event */ + ROGUE_FWIF_REG_CFG_TYPE_PWR_ON = 0, + /* Rascal / dust power event */ + ROGUE_FWIF_REG_CFG_TYPE_DUST_CHANGE, + /* Geometry kick */ + ROGUE_FWIF_REG_CFG_TYPE_GEOM, + /* Fragment kick */ + ROGUE_FWIF_REG_CFG_TYPE_FRAG, + /* Compute kick */ + ROGUE_FWIF_REG_CFG_TYPE_CDM, + /* TLA kick */ + ROGUE_FWIF_REG_CFG_TYPE_TLA, + /* TDM kick */ + ROGUE_FWIF_REG_CFG_TYPE_TDM, + /* Applies to all types. Keep as last element */ + ROGUE_FWIF_REG_CFG_TYPE_ALL +}; + +struct rogue_fwif_reg_cfg_rec { + u64 sddr; + u64 mask; + u64 value; +}; + +struct rogue_fwif_regconfig_data { + enum rogue_fwif_regdata_cmd_type cmd_type; + enum rogue_fwif_reg_cfg_type reg_config_type; + struct rogue_fwif_reg_cfg_rec reg_config __aligned(8); +}; + +struct rogue_fwif_reg_cfg { + /* + * PDump WRW command write granularity is 32 bits. + * Add padding to ensure array size is 32 bit granular. + */ + u8 num_regs_type[ALIGN((u32)ROGUE_FWIF_REG_CFG_TYPE_ALL, + sizeof(u32))] __aligned(8); + struct rogue_fwif_reg_cfg_rec + reg_configs[ROGUE_FWIF_REG_CFG_MAX_SIZE] __aligned(8); +} __aligned(8); + +enum rogue_fwif_os_state_change { + ROGUE_FWIF_OS_ONLINE = 1, + ROGUE_FWIF_OS_OFFLINE +}; + +struct rogue_fwif_os_state_change_data { + u32 osid; + enum rogue_fwif_os_state_change new_os_state; +} __aligned(8); + +enum rogue_fwif_counter_dump_request { + ROGUE_FWIF_PWR_COUNTER_DUMP_START = 1, + ROGUE_FWIF_PWR_COUNTER_DUMP_STOP, + ROGUE_FWIF_PWR_COUNTER_DUMP_SAMPLE, +}; + +struct rogue_fwif_counter_dump_data { + enum rogue_fwif_counter_dump_request counter_dump_request; +} __aligned(8); + +enum rogue_fwif_kccb_cmd_type { + /* Common commands */ + ROGUE_FWIF_KCCB_CMD_KICK = 101U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + ROGUE_FWIF_KCCB_CMD_MMUCACHE = 102U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + ROGUE_FWIF_KCCB_CMD_BP = 103U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* SLC flush and invalidation request */ + ROGUE_FWIF_KCCB_CMD_SLCFLUSHINVAL = 105U | + ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* + * Requests cleanup of a FW resource (type specified in the command + * data) + */ + ROGUE_FWIF_KCCB_CMD_CLEANUP = 106U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Power request */ + ROGUE_FWIF_KCCB_CMD_POW = 107U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Backing for on-demand ZS-Buffer done */ + ROGUE_FWIF_KCCB_CMD_ZSBUFFER_BACKING_UPDATE = + 108U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Unbacking for on-demand ZS-Buffer done */ + ROGUE_FWIF_KCCB_CMD_ZSBUFFER_UNBACKING_UPDATE = + 109U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Freelist Grow done */ + ROGUE_FWIF_KCCB_CMD_FREELIST_GROW_UPDATE = + 110U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Freelists Reconstruction done */ + ROGUE_FWIF_KCCB_CMD_FREELISTS_RECONSTRUCTION_UPDATE = + 112U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* + * Informs the firmware that the host has added more data to a CDM2 + * Circular Buffer + */ + ROGUE_FWIF_KCCB_CMD_NOTIFY_WRITE_OFFSET_UPDATE = + 114U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Health check request */ + ROGUE_FWIF_KCCB_CMD_HEALTH_CHECK = 115U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Forcing signalling of all unmet UFOs for a given CCB offset */ + ROGUE_FWIF_KCCB_CMD_FORCE_UPDATE = 116U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + + /* There is a geometry and a fragment command in this single kick */ + ROGUE_FWIF_KCCB_CMD_COMBINED_GEOM_FRAG_KICK = 117U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Informs the FW that a Guest OS has come online / offline. */ + ROGUE_FWIF_KCCB_CMD_OS_ONLINE_STATE_CONFIGURE = 118U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + + /* Commands only permitted to the native or host OS */ + ROGUE_FWIF_KCCB_CMD_REGCONFIG = 200U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + + /* Configure HWPerf events (to be generated) and HWPerf buffer address (if required) */ + ROGUE_FWIF_KCCB_CMD_HWPERF_UPDATE_CONFIG = 201U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + + /* Enable or disable multiple HWPerf blocks (reusing existing configuration) */ + ROGUE_FWIF_KCCB_CMD_HWPERF_CTRL_BLKS = 203U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Core clock speed change event */ + ROGUE_FWIF_KCCB_CMD_CORECLKSPEEDCHANGE = 204U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + + /* + * Ask the firmware to update its cached ui32LogType value from the (shared) + * tracebuf control structure + */ + ROGUE_FWIF_KCCB_CMD_LOGTYPE_UPDATE = 206U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Set a maximum frequency/OPP point */ + ROGUE_FWIF_KCCB_CMD_PDVFS_LIMIT_MAX_FREQ = 207U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* + * Changes the relative scheduling priority for a particular OSid. It can + * only be serviced for the Host DDK + */ + ROGUE_FWIF_KCCB_CMD_OSID_PRIORITY_CHANGE = 208U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Set or clear firmware state flags */ + ROGUE_FWIF_KCCB_CMD_STATEFLAGS_CTRL = 209U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + + /* Set a minimum frequency/OPP point */ + ROGUE_FWIF_KCCB_CMD_PDVFS_LIMIT_MIN_FREQ = 212U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Configure Periodic Hardware Reset behaviour */ + ROGUE_FWIF_KCCB_CMD_PHR_CFG = 213U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + + /* Configure Safety Firmware Watchdog */ + ROGUE_FWIF_KCCB_CMD_WDG_CFG = 215U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Controls counter dumping in the FW */ + ROGUE_FWIF_KCCB_CMD_COUNTER_DUMP = 216U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Configure, clear and enable multiple HWPerf blocks */ + ROGUE_FWIF_KCCB_CMD_HWPERF_CONFIG_ENABLE_BLKS = 217U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Configure the custom counters for HWPerf */ + ROGUE_FWIF_KCCB_CMD_HWPERF_SELECT_CUSTOM_CNTRS = 218U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + + /* Configure directly addressable counters for HWPerf */ + ROGUE_FWIF_KCCB_CMD_HWPERF_CONFIG_BLKS = 220U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, +}; + +#define ROGUE_FWIF_LAST_ALLOWED_GUEST_KCCB_CMD \ + (ROGUE_FWIF_KCCB_CMD_REGCONFIG - 1) + +/* Kernel CCB command packet */ +struct rogue_fwif_kccb_cmd { + /* Command type */ + enum rogue_fwif_kccb_cmd_type cmd_type; + /* Compatibility and other flags */ + u32 kccb_flags; + + /* + * NOTE: Make sure that uCmdData is the last member of this struct + * This is to calculate actual command size for device mem copy. + * (Refer ROGUEGetCmdMemCopySize()) + */ + union { + /* Data for Kick command */ + struct rogue_fwif_kccb_cmd_kick_data cmd_kick_data; + /* Data for combined geom/frag Kick command */ + struct rogue_fwif_kccb_cmd_combined_geom_frag_kick_data + combined_geom_frag_cmd_kick_data; + /* Data for MMU cache command */ + struct rogue_fwif_mmucachedata mmu_cache_data; + /* Data for Breakpoint Commands */ + struct rogue_fwif_bpdata bp_data; + /* Data for SLC Flush/Inval commands */ + struct rogue_fwif_slcflushinvaldata slc_flush_inval_data; + /* Data for cleanup commands */ + struct rogue_fwif_cleanup_request cleanup_data; + /* Data for power request commands */ + struct rogue_fwif_power_request pow_data; + /* Data for HWPerf control command */ + struct rogue_fwif_hwperf_ctrl hw_perf_ctrl; + /* + * Data for HWPerf configure, clear and enable performance + * counter block command + */ + struct rogue_fwif_hwperf_config_enable_blks + hw_perf_cfg_enable_blks; + /* + * Data for HWPerf enable or disable performance counter block + * commands + */ + struct rogue_fwif_hwperf_ctrl_blks hw_perf_ctrl_blks; + /* Data for HWPerf configure the custom counters to read */ + struct rogue_fwif_hwperf_select_custom_cntrs + hw_perf_select_cstm_cntrs; + /* Data for HWPerf configure Directly Addressable blocks */ + struct rogue_fwif_hwperf_config_da_blks hw_perf_cfg_da_blks; + /* Data for core clock speed change */ + struct rogue_fwif_coreclkspeedchange_data + core_clk_speed_change_data; + /* Feedback for Z/S Buffer backing/unbacking */ + struct rogue_fwif_zsbuffer_backing_data zs_buffer_backing_data; + /* Feedback for Freelist grow/shrink */ + struct rogue_fwif_freelist_gs_data free_list_gs_data; + /* Feedback for Freelists reconstruction*/ + struct rogue_fwif_freelists_reconstruction_data + free_lists_reconstruction_data; + /* Data for custom register configuration */ + struct rogue_fwif_regconfig_data reg_config_data; + /* Data for informing the FW about the write offset update */ + struct rogue_fwif_write_offset_update_data + write_offset_update_data; + /* Data for setting the max frequency/OPP */ + struct rogue_fwif_pdvfs_max_freq_data pdvfs_max_freq_data; + /* Data for setting the min frequency/OPP */ + struct rogue_fwif_pdvfs_min_freq_data pdvfs_min_freq_data; + /* Data for updating the Guest Online states */ + struct rogue_fwif_os_state_change_data cmd_os_online_state_data; + /* Dev address for TBI buffer allocated on demand */ + u32 tbi_buffer_fw_addr; + /* Data for dumping of register ranges */ + struct rogue_fwif_counter_dump_data counter_dump_config_data; + /* Data for signalling all unmet fences for a given CCB */ + struct rogue_fwif_kccb_cmd_force_update_data force_update_data; + } cmd_data __aligned(8); +} __aligned(8); + +PVR_FW_STRUCT_SIZE_ASSERT(struct rogue_fwif_kccb_cmd); + +/* + ****************************************************************************** + * Firmware CCB command structure for ROGUE + ****************************************************************************** + */ + +struct rogue_fwif_fwccb_cmd_zsbuffer_backing_data { + u32 zs_buffer_id; +}; + +struct rogue_fwif_fwccb_cmd_freelist_gs_data { + u32 freelist_id; +}; + +struct rogue_fwif_fwccb_cmd_freelists_reconstruction_data { + u32 freelist_count; + u32 hwr_counter; + u32 freelist_ids[ROGUE_FWIF_MAX_FREELISTS_TO_RECONSTRUCT]; +}; + +/* 1 if a page fault happened */ +#define ROGUE_FWIF_FWCCB_CMD_CONTEXT_RESET_FLAG_PF BIT(0) +/* 1 if applicable to all contexts */ +#define ROGUE_FWIF_FWCCB_CMD_CONTEXT_RESET_FLAG_ALL_CTXS BIT(1) + +struct rogue_fwif_fwccb_cmd_context_reset_data { + /* Context affected by the reset */ + u32 server_common_context_id; + /* Reason for reset */ + enum rogue_context_reset_reason reset_reason; + /* Data Master affected by the reset */ + u32 dm; + /* Job ref running at the time of reset */ + u32 reset_job_ref; + /* ROGUE_FWIF_FWCCB_CMD_CONTEXT_RESET_FLAG bitfield */ + u32 flags; + /* At what page catalog address */ + aligned_u64 pc_address; + /* Page fault address (only when applicable) */ + aligned_u64 fault_address; +}; + +struct rogue_fwif_fwccb_cmd_fw_pagefault_data { + /* Page fault address */ + u64 fw_fault_addr; +}; + +enum rogue_fwif_fwccb_cmd_type { + /* Requests ZSBuffer to be backed with physical pages */ + ROGUE_FWIF_FWCCB_CMD_ZSBUFFER_BACKING = 101U | + ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Requests ZSBuffer to be unbacked */ + ROGUE_FWIF_FWCCB_CMD_ZSBUFFER_UNBACKING = 102U | + ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Requests an on-demand freelist grow/shrink */ + ROGUE_FWIF_FWCCB_CMD_FREELIST_GROW = 103U | + ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Requests freelists reconstruction */ + ROGUE_FWIF_FWCCB_CMD_FREELISTS_RECONSTRUCTION = + 104U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Notifies host of a HWR event on a context */ + ROGUE_FWIF_FWCCB_CMD_CONTEXT_RESET_NOTIFICATION = + 105U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Requests an on-demand debug dump */ + ROGUE_FWIF_FWCCB_CMD_DEBUG_DUMP = 106U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + /* Requests an on-demand update on process stats */ + ROGUE_FWIF_FWCCB_CMD_UPDATE_STATS = 107U | + ROGUE_CMD_MAGIC_DWORD_SHIFTED, + + ROGUE_FWIF_FWCCB_CMD_CORE_CLK_RATE_CHANGE = + 108U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + ROGUE_FWIF_FWCCB_CMD_REQUEST_GPU_RESTART = + 109U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, + + /* Notifies host of a FW pagefault */ + ROGUE_FWIF_FWCCB_CMD_CONTEXT_FW_PF_NOTIFICATION = + 112U | ROGUE_CMD_MAGIC_DWORD_SHIFTED, +}; + +enum rogue_fwif_fwccb_cmd_update_stats_type { + /* + * PVRSRVStatsUpdateRenderContextStats should increase the value of the + * ui32TotalNumPartialRenders stat + */ + ROGUE_FWIF_FWCCB_CMD_UPDATE_NUM_PARTIAL_RENDERS = 1, + /* + * PVRSRVStatsUpdateRenderContextStats should increase the value of the + * ui32TotalNumOutOfMemory stat + */ + ROGUE_FWIF_FWCCB_CMD_UPDATE_NUM_OUT_OF_MEMORY, + /* + * PVRSRVStatsUpdateRenderContextStats should increase the value of the + * ui32NumGeomStores stat + */ + ROGUE_FWIF_FWCCB_CMD_UPDATE_NUM_GEOM_STORES, + /* + * PVRSRVStatsUpdateRenderContextStats should increase the value of the + * ui32NumFragStores stat + */ + ROGUE_FWIF_FWCCB_CMD_UPDATE_NUM_FRAG_STORES, + /* + * PVRSRVStatsUpdateRenderContextStats should increase the value of the + * ui32NumCDMStores stat + */ + ROGUE_FWIF_FWCCB_CMD_UPDATE_NUM_CDM_STORES, + /* + * PVRSRVStatsUpdateRenderContextStats should increase the value of the + * ui32NumTDMStores stat + */ + ROGUE_FWIF_FWCCB_CMD_UPDATE_NUM_TDM_STORES +}; + +struct rogue_fwif_fwccb_cmd_update_stats_data { + /* Element to update */ + enum rogue_fwif_fwccb_cmd_update_stats_type element_to_update; + /* The pid of the process whose stats are being updated */ + u32 pid_owner; + /* Adjustment to be made to the statistic */ + s32 adjustment_value; +}; + +struct rogue_fwif_fwccb_cmd_core_clk_rate_change_data { + u32 core_clk_rate; +} __aligned(8); + +struct rogue_fwif_fwccb_cmd { + /* Command type */ + enum rogue_fwif_fwccb_cmd_type cmd_type; + /* Compatibility and other flags */ + u32 fwccb_flags; + + union { + /* Data for Z/S-Buffer on-demand (un)backing*/ + struct rogue_fwif_fwccb_cmd_zsbuffer_backing_data + cmd_zs_buffer_backing; + /* Data for on-demand freelist grow/shrink */ + struct rogue_fwif_fwccb_cmd_freelist_gs_data cmd_free_list_gs; + /* Data for freelists reconstruction */ + struct rogue_fwif_fwccb_cmd_freelists_reconstruction_data + cmd_freelists_reconstruction; + /* Data for context reset notification */ + struct rogue_fwif_fwccb_cmd_context_reset_data + cmd_context_reset_notification; + /* Data for updating process stats */ + struct rogue_fwif_fwccb_cmd_update_stats_data + cmd_update_stats_data; + struct rogue_fwif_fwccb_cmd_core_clk_rate_change_data + cmd_core_clk_rate_change; + struct rogue_fwif_fwccb_cmd_fw_pagefault_data cmd_fw_pagefault; + } cmd_data __aligned(8); +} __aligned(8); + +PVR_FW_STRUCT_SIZE_ASSERT(struct rogue_fwif_fwccb_cmd); + +/* + ****************************************************************************** + * Workload estimation Firmware CCB command structure for ROGUE + ****************************************************************************** + */ +struct rogue_fwif_workest_fwccb_cmd { + /* Index for return data array */ + u16 return_data_index; + /* The cycles the workload took on the hardware */ + u32 cycles_taken; +}; + +/* + ****************************************************************************** + * Client CCB commands for ROGUE + ****************************************************************************** + */ + +/* + * Required memory alignment for 64-bit variables accessible by Meta + * (The gcc meta aligns 64-bit variables to 64-bit; therefore, memory shared + * between the host and meta that contains 64-bit variables has to maintain + * this alignment) + */ +#define ROGUE_FWIF_FWALLOC_ALIGN sizeof(u64) + +#define ROGUE_CCB_TYPE_TASK BIT(15) +#define ROGUE_CCB_FWALLOC_ALIGN(size) \ + (((size) + (ROGUE_FWIF_FWALLOC_ALIGN - 1)) & \ + ~(ROGUE_FWIF_FWALLOC_ALIGN - 1)) + +#define ROGUE_FWIF_CCB_CMD_TYPE_GEOM \ + (201U | ROGUE_CMD_MAGIC_DWORD_SHIFTED | ROGUE_CCB_TYPE_TASK) +#define ROGUE_FWIF_CCB_CMD_TYPE_TQ_3D \ + (202U | ROGUE_CMD_MAGIC_DWORD_SHIFTED | ROGUE_CCB_TYPE_TASK) +#define ROGUE_FWIF_CCB_CMD_TYPE_FRAG \ + (203U | ROGUE_CMD_MAGIC_DWORD_SHIFTED | ROGUE_CCB_TYPE_TASK) +#define ROGUE_FWIF_CCB_CMD_TYPE_FRAG_PR \ + (204U | ROGUE_CMD_MAGIC_DWORD_SHIFTED | ROGUE_CCB_TYPE_TASK) +#define ROGUE_FWIF_CCB_CMD_TYPE_CDM \ + (205U | ROGUE_CMD_MAGIC_DWORD_SHIFTED | ROGUE_CCB_TYPE_TASK) +#define ROGUE_FWIF_CCB_CMD_TYPE_TQ_TDM \ + (206U | ROGUE_CMD_MAGIC_DWORD_SHIFTED | ROGUE_CCB_TYPE_TASK) +#define ROGUE_FWIF_CCB_CMD_TYPE_FBSC_INVALIDATE \ + (207U | ROGUE_CMD_MAGIC_DWORD_SHIFTED | ROGUE_CCB_TYPE_TASK) +#define ROGUE_FWIF_CCB_CMD_TYPE_TQ_2D \ + (208U | ROGUE_CMD_MAGIC_DWORD_SHIFTED | ROGUE_CCB_TYPE_TASK) +#define ROGUE_FWIF_CCB_CMD_TYPE_PRE_TIMESTAMP \ + (209U | ROGUE_CMD_MAGIC_DWORD_SHIFTED | ROGUE_CCB_TYPE_TASK) +#define ROGUE_FWIF_CCB_CMD_TYPE_NULL \ + (210U | ROGUE_CMD_MAGIC_DWORD_SHIFTED | ROGUE_CCB_TYPE_TASK) +#define ROGUE_FWIF_CCB_CMD_TYPE_ABORT \ + (211U | ROGUE_CMD_MAGIC_DWORD_SHIFTED | ROGUE_CCB_TYPE_TASK) + +/* Leave a gap between CCB specific commands and generic commands */ +#define ROGUE_FWIF_CCB_CMD_TYPE_FENCE (212U | ROGUE_CMD_MAGIC_DWORD_SHIFTED) +#define ROGUE_FWIF_CCB_CMD_TYPE_UPDATE (213U | ROGUE_CMD_MAGIC_DWORD_SHIFTED) +#define ROGUE_FWIF_CCB_CMD_TYPE_RMW_UPDATE \ + (214U | ROGUE_CMD_MAGIC_DWORD_SHIFTED) +#define ROGUE_FWIF_CCB_CMD_TYPE_FENCE_PR (215U | ROGUE_CMD_MAGIC_DWORD_SHIFTED) +#define ROGUE_FWIF_CCB_CMD_TYPE_PRIORITY (216U | ROGUE_CMD_MAGIC_DWORD_SHIFTED) +/* + * Pre and Post timestamp commands are supposed to sandwich the DM cmd. The + * padding code with the CCB wrap upsets the FW if we don't have the task type + * bit cleared for POST_TIMESTAMPs. That's why we have 2 different cmd types. + */ +#define ROGUE_FWIF_CCB_CMD_TYPE_POST_TIMESTAMP \ + (217U | ROGUE_CMD_MAGIC_DWORD_SHIFTED) +#define ROGUE_FWIF_CCB_CMD_TYPE_UNFENCED_UPDATE \ + (218U | ROGUE_CMD_MAGIC_DWORD_SHIFTED) +#define ROGUE_FWIF_CCB_CMD_TYPE_UNFENCED_RMW_UPDATE \ + (219U | ROGUE_CMD_MAGIC_DWORD_SHIFTED) + +#define ROGUE_FWIF_CCB_CMD_TYPE_PADDING (221U | ROGUE_CMD_MAGIC_DWORD_SHIFTED) + +struct rogue_fwif_workest_kick_data { + /* Index for the KM Workload estimation return data array */ + u16 return_data_index __aligned(8); + /* Predicted time taken to do the work in cycles */ + u32 cycles_prediction __aligned(8); + /* Deadline for the workload */ + aligned_u64 deadline; +}; + +struct rogue_fwif_ccb_cmd_header { + u32 cmd_type; + u32 cmd_size; + /* + * external job reference - provided by client and used in debug for + * tracking submitted work + */ + u32 ext_job_ref; + /* + * internal job reference - generated by services and used in debug for + * tracking submitted work + */ + u32 int_job_ref; + /* Workload Estimation - Workload Estimation Data */ + struct rogue_fwif_workest_kick_data work_est_kick_data __aligned(8); +}; + +/* + ****************************************************************************** + * Client CCB commands which are only required by the kernel + ****************************************************************************** + */ +struct rogue_fwif_cmd_priority { + s32 priority; +}; + +/* + ****************************************************************************** + * Signature and Checksums Buffer + ****************************************************************************** + */ +struct rogue_fwif_sigbuf_ctl { + /* Ptr to Signature Buffer memory */ + u32 buffer_fw_addr; + /* Amount of space left for storing regs in the buffer */ + u32 left_size_in_regs; +} __aligned(8); + +struct rogue_fwif_counter_dump_ctl { + /* Ptr to counter dump buffer */ + u32 buffer_fw_addr; + /* Amount of space for storing in the buffer */ + u32 size_in_dwords; +} __aligned(8); + +struct rogue_fwif_firmware_gcov_ctl { + /* Ptr to firmware gcov buffer */ + u32 buffer_fw_addr; + /* Amount of space for storing in the buffer */ + u32 size; +} __aligned(8); + +/* + ***************************************************************************** + * ROGUE Compatibility checks + ***************************************************************************** + */ + +/* + * WARNING: Whenever the layout of ROGUE_FWIF_COMPCHECKS_BVNC changes, the + * following define should be increased by 1 to indicate to the compatibility + * logic that layout has changed. + */ +#define ROGUE_FWIF_COMPCHECKS_LAYOUT_VERSION 3 + +struct rogue_fwif_compchecks_bvnc { + /* WARNING: This field must be defined as first one in this structure */ + u32 layout_version; + aligned_u64 bvnc; +} __aligned(8); + +struct rogue_fwif_init_options { + u8 os_count_support; + u8 padding[7]; +} __aligned(8); + +#define ROGUE_FWIF_COMPCHECKS_BVNC_DECLARE_AND_INIT(name) \ + struct rogue_fwif_compchecks_bvnc(name) = { \ + ROGUE_FWIF_COMPCHECKS_LAYOUT_VERSION, \ + 0, \ + } + +static inline void rogue_fwif_compchecks_bvnc_init(struct rogue_fwif_compchecks_bvnc *compchecks) +{ + compchecks->layout_version = ROGUE_FWIF_COMPCHECKS_LAYOUT_VERSION; + compchecks->bvnc = 0; +} + +struct rogue_fwif_compchecks { + /* hardware BVNC (from the ROGUE registers) */ + struct rogue_fwif_compchecks_bvnc hw_bvnc; + /* firmware BVNC */ + struct rogue_fwif_compchecks_bvnc fw_bvnc; + /* identifier of the FW processor version */ + u32 fw_processor_version; + /* software DDK version */ + u32 ddk_version; + /* software DDK build no. */ + u32 ddk_build; + /* build options bit-field */ + u32 build_options; + /* initialisation options bit-field */ + struct rogue_fwif_init_options init_options; + /* Information is valid */ + bool updated __aligned(4); + u32 padding; +} __aligned(8); + +/* + ****************************************************************************** + * Updated configuration post FW data init. + ****************************************************************************** + */ +struct rogue_fwif_runtime_cfg { + /* APM latency in ms before signalling IDLE to the host */ + u32 active_pm_latency_ms; + /* Compatibility and other flags */ + u32 runtime_cfg_flags; + /* + * If set, APM latency does not reset to system default each GPU power + * transition + */ + bool active_pm_latency_persistant __aligned(4); + /* Core clock speed, currently only used to calculate timer ticks */ + u32 core_clock_speed; + /* Last number of dusts change requested by the host */ + u32 default_dusts_num_init; + /* Periodic Hardware Reset configuration values */ + u32 phr_mode; + /* New number of milliseconds C/S is allowed to last */ + u32 hcs_deadline_ms; + /* The watchdog period in microseconds */ + u32 wdg_period_us; + /* Array of priorities per OS */ + u32 osid_priority[ROGUE_FW_MAX_NUM_OS]; + /* On-demand allocated HWPerf buffer address, to be passed to the FW */ + u32 hwperf_buf_fw_addr; + + bool padding __aligned(4); +}; + +/* + ***************************************************************************** + * Control data for ROGUE + ***************************************************************************** + */ + +#define ROGUE_FWIF_HWR_DEBUG_DUMP_ALL (99999U) + +enum rogue_fwif_tpu_dm { + ROGUE_FWIF_TPU_DM_PDM = 0, + ROGUE_FWIF_TPU_DM_VDM = 1, + ROGUE_FWIF_TPU_DM_CDM = 2, + ROGUE_FWIF_TPU_DM_TDM = 3, + ROGUE_FWIF_TPU_DM_LAST +}; + +enum rogue_fwif_gpio_val_mode { + /* No GPIO validation */ + ROGUE_FWIF_GPIO_VAL_OFF = 0, + /* + * Simple test case that initiates by sending data via the GPIO and then + * sends back any data received over the GPIO + */ + ROGUE_FWIF_GPIO_VAL_GENERAL = 1, + /* + * More complex test case that writes and reads data across the entire + * GPIO AP address range. + */ + ROGUE_FWIF_GPIO_VAL_AP = 2, + /* Validates the GPIO Testbench. */ + ROGUE_FWIF_GPIO_VAL_TESTBENCH = 5, + /* Send and then receive each byte in the range 0-255. */ + ROGUE_FWIF_GPIO_VAL_LOOPBACK = 6, + /* Send and then receive each power-of-2 byte in the range 0-255. */ + ROGUE_FWIF_GPIO_VAL_LOOPBACK_LITE = 7, + ROGUE_FWIF_GPIO_VAL_LAST +}; + +enum fw_perf_conf { + FW_PERF_CONF_NONE = 0, + FW_PERF_CONF_ICACHE = 1, + FW_PERF_CONF_DCACHE = 2, + FW_PERF_CONF_JTLB_INSTR = 5, + FW_PERF_CONF_INSTRUCTIONS = 6 +}; + +enum fw_boot_stage { + FW_BOOT_STAGE_TLB_INIT_FAILURE = -2, + FW_BOOT_STAGE_NOT_AVAILABLE = -1, + FW_BOOT_NOT_STARTED = 0, + FW_BOOT_BLDR_STARTED = 1, + FW_BOOT_CACHE_DONE, + FW_BOOT_TLB_DONE, + FW_BOOT_MAIN_STARTED, + FW_BOOT_ALIGNCHECKS_DONE, + FW_BOOT_INIT_DONE, +}; + +/* + * Kernel CCB return slot responses. Usage of bit-fields instead of bare + * integers allows FW to possibly pack-in several responses for each single kCCB + * command. + */ +/* Command executed (return status from FW) */ +#define ROGUE_FWIF_KCCB_RTN_SLOT_CMD_EXECUTED BIT(0) +/* A cleanup was requested but resource busy */ +#define ROGUE_FWIF_KCCB_RTN_SLOT_CLEANUP_BUSY BIT(1) +/* Poll failed in FW for a HW operation to complete */ +#define ROGUE_FWIF_KCCB_RTN_SLOT_POLL_FAILURE BIT(2) +/* Reset value of a kCCB return slot (set by host) */ +#define ROGUE_FWIF_KCCB_RTN_SLOT_NO_RESPONSE 0x0U + +struct rogue_fwif_connection_ctl { + /* Fw-Os connection states */ + enum rogue_fwif_connection_fw_state connection_fw_state; + enum rogue_fwif_connection_os_state connection_os_state; + u32 alive_fw_token; + u32 alive_os_token; +} __aligned(8); + +struct rogue_fwif_osinit { + /* Kernel CCB */ + u32 kernel_ccbctl_fw_addr; + u32 kernel_ccb_fw_addr; + u32 kernel_ccb_rtn_slots_fw_addr; + + /* Firmware CCB */ + u32 firmware_ccbctl_fw_addr; + u32 firmware_ccb_fw_addr; + + /* Workload Estimation Firmware CCB */ + u32 work_est_firmware_ccbctl_fw_addr; + u32 work_est_firmware_ccb_fw_addr; + + u32 rogue_fwif_hwr_info_buf_ctl_fw_addr; + + u32 hwr_debug_dump_limit; + + u32 fw_os_data_fw_addr; + + /* Compatibility checks to be populated by the Firmware */ + struct rogue_fwif_compchecks rogue_comp_checks; +} __aligned(8); + +/* BVNC Features */ +struct rogue_hwperf_bvnc_block { + /* Counter block ID, see ROGUE_HWPERF_CNTBLK_ID */ + u16 block_id; + + /* Number of counters in this block type */ + u16 num_counters; + + /* Number of blocks of this type */ + u16 num_blocks; + + u16 reserved; +}; + +#define ROGUE_HWPERF_MAX_BVNC_LEN (24) + +#define ROGUE_HWPERF_MAX_BVNC_BLOCK_LEN (16U) + +/* BVNC Features */ +struct rogue_hwperf_bvnc { + /* BVNC string */ + char bvnc_string[ROGUE_HWPERF_MAX_BVNC_LEN]; + /* See ROGUE_HWPERF_FEATURE_FLAGS */ + u32 bvnc_km_feature_flags; + /* Number of blocks described in aBvncBlocks */ + u16 num_bvnc_blocks; + /* Number of GPU cores present */ + u16 bvnc_gpu_cores; + /* Supported Performance Blocks for BVNC */ + struct rogue_hwperf_bvnc_block + bvnc_blocks[ROGUE_HWPERF_MAX_BVNC_BLOCK_LEN]; +}; + +PVR_FW_STRUCT_SIZE_ASSERT(struct rogue_hwperf_bvnc); + +struct rogue_fwif_sysinit { + /* Fault read address */ + aligned_u64 fault_phys_addr; + + /* PDS execution base */ + aligned_u64 pds_exec_base; + /* UCS execution base */ + aligned_u64 usc_exec_base; + /* FBCDC bindless texture state table base */ + aligned_u64 fbcdc_state_table_base; + aligned_u64 fbcdc_large_state_table_base; + /* Texture state base */ + aligned_u64 texture_heap_base; + + /* Event filter for Firmware events */ + u64 hw_perf_filter; + + aligned_u64 slc3_fence_dev_addr; + + u32 tpu_trilinear_frac_mask[ROGUE_FWIF_TPU_DM_LAST] __aligned(8); + + /* Signature and Checksum Buffers for DMs */ + struct rogue_fwif_sigbuf_ctl sigbuf_ctl[PVR_FWIF_DM_MAX]; + + struct rogue_fwif_pdvfs_opp pdvfs_opp_info; + + struct rogue_fwif_dma_addr coremem_data_store; + + struct rogue_fwif_counter_dump_ctl counter_dump_ctl; + + u32 filter_flags; + + u32 runtime_cfg_fw_addr; + + u32 trace_buf_ctl_fw_addr; + u32 fw_sys_data_fw_addr; + + u32 gpu_util_fw_cb_ctl_fw_addr; + u32 reg_cfg_fw_addr; + u32 hwperf_ctl_fw_addr; + + u32 align_checks; + + /* Core clock speed at FW boot time */ + u32 initial_core_clock_speed; + + /* APM latency in ms before signalling IDLE to the host */ + u32 active_pm_latency_ms; + + /* Flag to be set by the Firmware after successful start */ + bool firmware_started __aligned(4); + + /* Host/FW Trace synchronisation Partition Marker */ + u32 marker_val; + + /* Firmware initialization complete time */ + u32 firmware_started_timestamp; + + u32 jones_disable_mask; + + /* Firmware performance counter config */ + enum fw_perf_conf firmware_perf; + + /* + * FW Pointer to memory containing core clock rate in Hz. + * Firmware (PDVFS) updates the memory when running on non primary FW + * thread to communicate to host driver. + */ + u32 core_clock_rate_fw_addr; + + enum rogue_fwif_gpio_val_mode gpio_validation_mode; + + /* Used in HWPerf for decoding BVNC Features */ + struct rogue_hwperf_bvnc bvnc_km_feature_flags; + + /* Value to write into ROGUE_CR_TFBC_COMPRESSION_CONTROL */ + u32 tfbc_compression_control; +} __aligned(8); + +/* + ***************************************************************************** + * Timer correlation shared data and defines + ***************************************************************************** + */ + +struct rogue_fwif_time_corr { + aligned_u64 os_timestamp; + aligned_u64 os_mono_timestamp; + aligned_u64 cr_timestamp; + + /* + * Utility variable used to convert CR timer deltas to OS timer deltas + * (nS), where the deltas are relative to the timestamps above: + * deltaOS = (deltaCR * K) >> decimal_shift, see full explanation below + */ + aligned_u64 cr_delta_to_os_delta_kns; + + u32 core_clock_speed; + u32 reserved; +} __aligned(8); + +/* + * The following macros are used to help converting FW timestamps to the Host + * time domain. On the FW the ROGUE_CR_TIMER counter is used to keep track of + * time; it increments by 1 every 256 GPU clock ticks, so the general + * formula to perform the conversion is: + * + * [ GPU clock speed in Hz, if (scale == 10^9) then deltaOS is in nS, + * otherwise if (scale == 10^6) then deltaOS is in uS ] + * + * deltaCR * 256 256 * scale + * deltaOS = --------------- * scale = deltaCR * K [ K = --------------- ] + * GPUclockspeed GPUclockspeed + * + * The actual K is multiplied by 2^20 (and deltaCR * K is divided by 2^20) + * to get some better accuracy and to avoid returning 0 in the integer + * division 256000000/GPUfreq if GPUfreq is greater than 256MHz. + * This is the same as keeping K as a decimal number. + * + * The maximum deltaOS is slightly more than 5hrs for all GPU frequencies + * (deltaCR * K is more or less a constant), and it's relative to the base + * OS timestamp sampled as a part of the timer correlation data. + * This base is refreshed on GPU power-on, DVFS transition and periodic + * frequency calibration (executed every few seconds if the FW is doing + * some work), so as long as the GPU is doing something and one of these + * events is triggered then deltaCR * K will not overflow and deltaOS will be + * correct. + */ + +#define ROGUE_FWIF_CRDELTA_TO_OSDELTA_ACCURACY_SHIFT (20) + +#define ROGUE_FWIF_GET_DELTA_OSTIME_NS(delta_cr, k) \ + (((delta_cr) * (k)) >> ROGUE_FWIF_CRDELTA_TO_OSDELTA_ACCURACY_SHIFT) + +/* + ****************************************************************************** + * GPU Utilisation + ****************************************************************************** + */ + +/* See rogue_common.h for a list of GPU states */ +#define ROGUE_FWIF_GPU_UTIL_TIME_MASK \ + (0xFFFFFFFFFFFFFFFFull & ~ROGUE_FWIF_GPU_UTIL_STATE_MASK) + +#define ROGUE_FWIF_GPU_UTIL_GET_TIME(word) \ + ((word)(&ROGUE_FWIF_GPU_UTIL_TIME_MASK)) +#define ROGUE_FWIF_GPU_UTIL_GET_STATE(word) \ + ((word)(&ROGUE_FWIF_GPU_UTIL_STATE_MASK)) + +/* + * The OS timestamps computed by the FW are approximations of the real time, + * which means they could be slightly behind or ahead the real timer on the + * Host. In some cases we can perform subtractions between FW approximated + * timestamps and real OS timestamps, so we need a form of protection against + * negative results if for instance the FW one is a bit ahead of time. + */ +#define ROGUE_FWIF_GPU_UTIL_GET_PERIOD(newtime, oldtime) \ + (((newtime) > (oldtime)) ? ((newtime) - (oldtime)) : 0U) + +#define ROGUE_FWIF_GPU_UTIL_MAKE_WORD(time, state) \ + (ROGUE_FWIF_GPU_UTIL_GET_TIME(time) | \ + ROGUE_FWIF_GPU_UTIL_GET_STATE(state)) + +/* + * The timer correlation array must be big enough to ensure old entries won't be + * overwritten before all the HWPerf events linked to those entries are + * processed by the MISR. The update frequency of this array depends on how fast + * the system can change state (basically how small the APM latency is) and + * perform DVFS transitions. + * + * The minimum size is 2 (not 1) to avoid race conditions between the FW reading + * an entry while the Host is updating it. With 2 entries in the worst case the + * FW will read old data, which is still quite ok if the Host is updating the + * timer correlation at that time. + */ +#define ROGUE_FWIF_TIME_CORR_ARRAY_SIZE 256U +#define ROGUE_FWIF_TIME_CORR_CURR_INDEX(seqcount) \ + ((seqcount) % ROGUE_FWIF_TIME_CORR_ARRAY_SIZE) + +/* Make sure the timer correlation array size is a power of 2 */ +static_assert((ROGUE_FWIF_TIME_CORR_ARRAY_SIZE & + (ROGUE_FWIF_TIME_CORR_ARRAY_SIZE - 1U)) == 0U, + "ROGUE_FWIF_TIME_CORR_ARRAY_SIZE must be a power of two"); + +struct rogue_fwif_gpu_util_fwcb { + struct rogue_fwif_time_corr time_corr[ROGUE_FWIF_TIME_CORR_ARRAY_SIZE]; + u32 time_corr_seq_count; + + /* Compatibility and other flags */ + u32 gpu_util_flags; + + /* Last GPU state + OS time of the last state update */ + aligned_u64 last_word; + + /* Counters for the amount of time the GPU was active/idle/blocked */ + aligned_u64 stats_counters[PVR_FWIF_GPU_UTIL_STATE_NUM]; +} __aligned(8); + +struct rogue_fwif_rta_ctl { + /* Render number */ + u32 render_target_index; + /* index in RTA */ + u32 current_render_target; + /* total active RTs */ + u32 active_render_targets; + /* total active RTs from the first TA kick, for OOM */ + u32 cumul_active_render_targets; + /* Array of valid RT indices */ + u32 valid_render_targets_fw_addr; + /* Array of number of occurred partial renders per render target */ + u32 rta_num_partial_renders_fw_addr; + /* Number of render targets in the array */ + u32 max_rts; + /* Compatibility and other flags */ + u32 rta_ctl_flags; +} __aligned(8); + +struct rogue_fwif_freelist { + aligned_u64 freelist_dev_addr; + aligned_u64 current_dev_addr; + u32 current_stack_top; + u32 max_pages; + u32 grow_pages; + /* HW pages */ + u32 current_pages; + u32 allocated_page_count; + u32 allocated_mmu_page_count; + u32 freelist_id; + + bool grow_pending __aligned(4); + /* Pages that should be used only when OOM is reached */ + u32 ready_pages; + /* Compatibility and other flags */ + u32 freelist_flags; + /* PM Global PB on which Freelist is loaded */ + u32 pm_global_pb; + u32 padding; +} __aligned(8); + +/* + ****************************************************************************** + * HWRTData + ****************************************************************************** + */ + +/* HWRTData flags */ +/* Deprecated flags 1:0 */ +#define HWRTDATA_HAS_LAST_GEOM BIT(2) +#define HWRTDATA_PARTIAL_RENDERED BIT(3) +#define HWRTDATA_DISABLE_TILE_REORDERING BIT(4) +#define HWRTDATA_NEED_BRN65101_BLIT BIT(5) +#define HWRTDATA_FIRST_BRN65101_STRIP BIT(6) +#define HWRTDATA_NEED_BRN67182_2ND_RENDER BIT(7) + +enum rogue_fwif_rtdata_state { + ROGUE_FWIF_RTDATA_STATE_NONE = 0, + ROGUE_FWIF_RTDATA_STATE_KICK_GEOM, + ROGUE_FWIF_RTDATA_STATE_KICK_GEOM_FIRST, + ROGUE_FWIF_RTDATA_STATE_GEOM_FINISHED, + ROGUE_FWIF_RTDATA_STATE_KICK_FRAG, + ROGUE_FWIF_RTDATA_STATE_FRAG_FINISHED, + ROGUE_FWIF_RTDATA_STATE_FRAG_CONTEXT_STORED, + ROGUE_FWIF_RTDATA_STATE_GEOM_OUTOFMEM, + ROGUE_FWIF_RTDATA_STATE_PARTIALRENDERFINISHED, + /* + * In case of HWR, we can't set the RTDATA state to NONE, as this will + * cause any TA to become a first TA. To ensure all related TA's are + * skipped, we use the HWR state + */ + ROGUE_FWIF_RTDATA_STATE_HWR, + ROGUE_FWIF_RTDATA_STATE_UNKNOWN = 0x7FFFFFFFU +}; + +struct rogue_fwif_hwrtdata_common { + bool geom_caches_need_zeroing __aligned(4); + + u32 screen_pixel_max; + aligned_u64 multi_sample_ctl; + u64 flipped_multi_sample_ctl; + u32 tpc_stride; + u32 tpc_size; + u32 te_screen; + u32 mtile_stride; + u32 teaa; + u32 te_mtile1; + u32 te_mtile2; + u32 isp_merge_lower_x; + u32 isp_merge_lower_y; + u32 isp_merge_upper_x; + u32 isp_merge_upper_y; + u32 isp_merge_scale_x; + u32 isp_merge_scale_y; + u32 rgn_header_size; + u32 isp_mtile_size; + u32 padding; +} __aligned(8); + +struct rogue_fwif_hwrtdata { + /* MList Data Store */ + aligned_u64 pm_mlist_dev_addr; + + aligned_u64 vce_cat_base[4]; + aligned_u64 vce_last_cat_base[4]; + aligned_u64 te_cat_base[4]; + aligned_u64 te_last_cat_base[4]; + aligned_u64 alist_cat_base; + aligned_u64 alist_last_cat_base; + + aligned_u64 pm_alist_stack_pointer; + u32 pm_mlist_stack_pointer; + + u32 hwrt_data_common_fw_addr; + + u32 hwrt_data_flags; + enum rogue_fwif_rtdata_state state; + + u32 freelists_fw_addr[MAX_FREELISTS_SIZE] __aligned(8); + u32 freelist_hwr_snapshot[MAX_FREELISTS_SIZE]; + + aligned_u64 vheap_table_dev_addr; + + struct rogue_fwif_rta_ctl rta_ctl; + + aligned_u64 tail_ptrs_dev_addr; + aligned_u64 macrotile_array_dev_addr; + aligned_u64 rgn_header_dev_addr; + aligned_u64 rtc_dev_addr; + + u32 owner_geom_not_used_by_host __aligned(8); + + bool geom_caches_need_zeroing __aligned(4); + + struct rogue_fwif_cleanup_ctl cleanup_state __aligned(64); +} __aligned(8); + +/* + ****************************************************************************** + * Sync checkpoints + ****************************************************************************** + */ + +#define PVR_SYNC_CHECKPOINT_UNDEF 0x000 +#define PVR_SYNC_CHECKPOINT_ACTIVE 0xac1 /* Checkpoint has not signaled. */ +#define PVR_SYNC_CHECKPOINT_SIGNALED 0x519 /* Checkpoint has signaled. */ +#define PVR_SYNC_CHECKPOINT_ERRORED 0xeff /* Checkpoint has been errored. */ + +#include "pvr_rogue_fwif_check.h" + +#endif /* PVR_ROGUE_FWIF_H */ diff --git a/drivers/gpu/drm/imagination/pvr_rogue_fwif_check.h b/drivers/gpu/drm/imagination/pvr_rogue_fwif_check.h new file mode 100644 index 000000000000..51dc37e78f41 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_fwif_check.h @@ -0,0 +1,493 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_ROGUE_FWIF_CHECK_H +#define PVR_ROGUE_FWIF_CHECK_H + +#include + +#define OFFSET_CHECK(type, member, offset) \ + static_assert(offsetof(type, member) == (offset), \ + "offsetof(" #type ", " #member ") incorrect") + +#define SIZE_CHECK(type, size) \ + static_assert(sizeof(type) == (size), #type " is incorrect size") + +OFFSET_CHECK(struct rogue_fwif_file_info_buf, path, 0); +OFFSET_CHECK(struct rogue_fwif_file_info_buf, info, 200); +OFFSET_CHECK(struct rogue_fwif_file_info_buf, line_num, 400); +SIZE_CHECK(struct rogue_fwif_file_info_buf, 408); + +OFFSET_CHECK(struct rogue_fwif_tracebuf_space, trace_pointer, 0); +OFFSET_CHECK(struct rogue_fwif_tracebuf_space, trace_buffer_fw_addr, 4); +OFFSET_CHECK(struct rogue_fwif_tracebuf_space, trace_buffer, 8); +OFFSET_CHECK(struct rogue_fwif_tracebuf_space, assert_buf, 16); +SIZE_CHECK(struct rogue_fwif_tracebuf_space, 424); + +OFFSET_CHECK(struct rogue_fwif_tracebuf, log_type, 0); +OFFSET_CHECK(struct rogue_fwif_tracebuf, tracebuf, 8); +OFFSET_CHECK(struct rogue_fwif_tracebuf, tracebuf_size_in_dwords, 856); +OFFSET_CHECK(struct rogue_fwif_tracebuf, tracebuf_flags, 860); +SIZE_CHECK(struct rogue_fwif_tracebuf, 864); + +OFFSET_CHECK(struct rogue_fw_fault_info, cr_timer, 0); +OFFSET_CHECK(struct rogue_fw_fault_info, os_timer, 8); +OFFSET_CHECK(struct rogue_fw_fault_info, data, 16); +OFFSET_CHECK(struct rogue_fw_fault_info, reserved, 20); +OFFSET_CHECK(struct rogue_fw_fault_info, fault_buf, 24); +SIZE_CHECK(struct rogue_fw_fault_info, 432); + +OFFSET_CHECK(struct rogue_fwif_sysdata, config_flags, 0); +OFFSET_CHECK(struct rogue_fwif_sysdata, config_flags_ext, 4); +OFFSET_CHECK(struct rogue_fwif_sysdata, pow_state, 8); +OFFSET_CHECK(struct rogue_fwif_sysdata, hw_perf_ridx, 12); +OFFSET_CHECK(struct rogue_fwif_sysdata, hw_perf_widx, 16); +OFFSET_CHECK(struct rogue_fwif_sysdata, hw_perf_wrap_count, 20); +OFFSET_CHECK(struct rogue_fwif_sysdata, hw_perf_size, 24); +OFFSET_CHECK(struct rogue_fwif_sysdata, hw_perf_drop_count, 28); +OFFSET_CHECK(struct rogue_fwif_sysdata, hw_perf_ut, 32); +OFFSET_CHECK(struct rogue_fwif_sysdata, first_drop_ordinal, 36); +OFFSET_CHECK(struct rogue_fwif_sysdata, last_drop_ordinal, 40); +OFFSET_CHECK(struct rogue_fwif_sysdata, os_runtime_flags_mirror, 44); +OFFSET_CHECK(struct rogue_fwif_sysdata, fault_info, 80); +OFFSET_CHECK(struct rogue_fwif_sysdata, fw_faults, 3536); +OFFSET_CHECK(struct rogue_fwif_sysdata, cr_poll_addr, 3540); +OFFSET_CHECK(struct rogue_fwif_sysdata, cr_poll_mask, 3548); +OFFSET_CHECK(struct rogue_fwif_sysdata, cr_poll_count, 3556); +OFFSET_CHECK(struct rogue_fwif_sysdata, start_idle_time, 3568); +OFFSET_CHECK(struct rogue_fwif_sysdata, hwr_state_flags, 3576); +OFFSET_CHECK(struct rogue_fwif_sysdata, hwr_recovery_flags, 3580); +OFFSET_CHECK(struct rogue_fwif_sysdata, fw_sys_data_flags, 3616); +OFFSET_CHECK(struct rogue_fwif_sysdata, mc_config, 3620); +SIZE_CHECK(struct rogue_fwif_sysdata, 3624); + +OFFSET_CHECK(struct rogue_fwif_slr_entry, timestamp, 0); +OFFSET_CHECK(struct rogue_fwif_slr_entry, fw_ctx_addr, 8); +OFFSET_CHECK(struct rogue_fwif_slr_entry, num_ufos, 12); +OFFSET_CHECK(struct rogue_fwif_slr_entry, ccb_name, 16); +SIZE_CHECK(struct rogue_fwif_slr_entry, 48); + +OFFSET_CHECK(struct rogue_fwif_osdata, fw_os_config_flags, 0); +OFFSET_CHECK(struct rogue_fwif_osdata, fw_sync_check_mark, 4); +OFFSET_CHECK(struct rogue_fwif_osdata, host_sync_check_mark, 8); +OFFSET_CHECK(struct rogue_fwif_osdata, forced_updates_requested, 12); +OFFSET_CHECK(struct rogue_fwif_osdata, slr_log_wp, 16); +OFFSET_CHECK(struct rogue_fwif_osdata, slr_log_first, 24); +OFFSET_CHECK(struct rogue_fwif_osdata, slr_log, 72); +OFFSET_CHECK(struct rogue_fwif_osdata, last_forced_update_time, 552); +OFFSET_CHECK(struct rogue_fwif_osdata, interrupt_count, 560); +OFFSET_CHECK(struct rogue_fwif_osdata, kccb_cmds_executed, 568); +OFFSET_CHECK(struct rogue_fwif_osdata, power_sync_fw_addr, 572); +OFFSET_CHECK(struct rogue_fwif_osdata, fw_os_data_flags, 576); +SIZE_CHECK(struct rogue_fwif_osdata, 584); + +OFFSET_CHECK(struct rogue_bifinfo, bif_req_status, 0); +OFFSET_CHECK(struct rogue_bifinfo, bif_mmu_status, 8); +OFFSET_CHECK(struct rogue_bifinfo, pc_address, 16); +OFFSET_CHECK(struct rogue_bifinfo, reserved, 24); +SIZE_CHECK(struct rogue_bifinfo, 32); + +OFFSET_CHECK(struct rogue_eccinfo, fault_gpu, 0); +SIZE_CHECK(struct rogue_eccinfo, 4); + +OFFSET_CHECK(struct rogue_mmuinfo, mmu_status, 0); +OFFSET_CHECK(struct rogue_mmuinfo, pc_address, 16); +OFFSET_CHECK(struct rogue_mmuinfo, reserved, 24); +SIZE_CHECK(struct rogue_mmuinfo, 32); + +OFFSET_CHECK(struct rogue_pollinfo, thread_num, 0); +OFFSET_CHECK(struct rogue_pollinfo, cr_poll_addr, 4); +OFFSET_CHECK(struct rogue_pollinfo, cr_poll_mask, 8); +OFFSET_CHECK(struct rogue_pollinfo, cr_poll_last_value, 12); +OFFSET_CHECK(struct rogue_pollinfo, reserved, 16); +SIZE_CHECK(struct rogue_pollinfo, 24); + +OFFSET_CHECK(struct rogue_tlbinfo, bad_addr, 0); +OFFSET_CHECK(struct rogue_tlbinfo, entry_lo, 4); +SIZE_CHECK(struct rogue_tlbinfo, 8); + +OFFSET_CHECK(struct rogue_hwrinfo, hwr_data, 0); +OFFSET_CHECK(struct rogue_hwrinfo, cr_timer, 32); +OFFSET_CHECK(struct rogue_hwrinfo, os_timer, 40); +OFFSET_CHECK(struct rogue_hwrinfo, frame_num, 48); +OFFSET_CHECK(struct rogue_hwrinfo, pid, 52); +OFFSET_CHECK(struct rogue_hwrinfo, active_hwrt_data, 56); +OFFSET_CHECK(struct rogue_hwrinfo, hwr_number, 60); +OFFSET_CHECK(struct rogue_hwrinfo, event_status, 64); +OFFSET_CHECK(struct rogue_hwrinfo, hwr_recovery_flags, 68); +OFFSET_CHECK(struct rogue_hwrinfo, hwr_type, 72); +OFFSET_CHECK(struct rogue_hwrinfo, dm, 76); +OFFSET_CHECK(struct rogue_hwrinfo, core_id, 80); +OFFSET_CHECK(struct rogue_hwrinfo, cr_time_of_kick, 88); +OFFSET_CHECK(struct rogue_hwrinfo, cr_time_hw_reset_start, 96); +OFFSET_CHECK(struct rogue_hwrinfo, cr_time_hw_reset_finish, 104); +OFFSET_CHECK(struct rogue_hwrinfo, cr_time_freelist_ready, 112); +OFFSET_CHECK(struct rogue_hwrinfo, reserved, 120); +SIZE_CHECK(struct rogue_hwrinfo, 136); + +OFFSET_CHECK(struct rogue_fwif_hwrinfobuf, hwr_info, 0); +OFFSET_CHECK(struct rogue_fwif_hwrinfobuf, hwr_counter, 2176); +OFFSET_CHECK(struct rogue_fwif_hwrinfobuf, write_index, 2180); +OFFSET_CHECK(struct rogue_fwif_hwrinfobuf, dd_req_count, 2184); +OFFSET_CHECK(struct rogue_fwif_hwrinfobuf, hwr_info_buf_flags, 2188); +OFFSET_CHECK(struct rogue_fwif_hwrinfobuf, hwr_dm_locked_up_count, 2192); +OFFSET_CHECK(struct rogue_fwif_hwrinfobuf, hwr_dm_overran_count, 2228); +OFFSET_CHECK(struct rogue_fwif_hwrinfobuf, hwr_dm_recovered_count, 2264); +OFFSET_CHECK(struct rogue_fwif_hwrinfobuf, hwr_dm_false_detect_count, 2300); +SIZE_CHECK(struct rogue_fwif_hwrinfobuf, 2336); + +OFFSET_CHECK(struct rogue_fwif_fwmemcontext, pc_dev_paddr, 0); +OFFSET_CHECK(struct rogue_fwif_fwmemcontext, page_cat_base_reg_set, 8); +OFFSET_CHECK(struct rogue_fwif_fwmemcontext, breakpoint_addr, 12); +OFFSET_CHECK(struct rogue_fwif_fwmemcontext, bp_handler_addr, 16); +OFFSET_CHECK(struct rogue_fwif_fwmemcontext, breakpoint_ctl, 20); +OFFSET_CHECK(struct rogue_fwif_fwmemcontext, fw_mem_ctx_flags, 24); +SIZE_CHECK(struct rogue_fwif_fwmemcontext, 32); + +OFFSET_CHECK(struct rogue_fwif_geom_ctx_state_per_geom, geom_reg_vdm_call_stack_pointer, 0); +OFFSET_CHECK(struct rogue_fwif_geom_ctx_state_per_geom, geom_reg_vdm_call_stack_pointer_init, 8); +OFFSET_CHECK(struct rogue_fwif_geom_ctx_state_per_geom, geom_reg_vbs_so_prim, 16); +OFFSET_CHECK(struct rogue_fwif_geom_ctx_state_per_geom, geom_current_idx, 32); +SIZE_CHECK(struct rogue_fwif_geom_ctx_state_per_geom, 40); + +OFFSET_CHECK(struct rogue_fwif_geom_ctx_state, geom_core, 0); +SIZE_CHECK(struct rogue_fwif_geom_ctx_state, 160); + +OFFSET_CHECK(struct rogue_fwif_frag_ctx_state, frag_reg_pm_deallocated_mask_status, 0); +OFFSET_CHECK(struct rogue_fwif_frag_ctx_state, frag_reg_dm_pds_mtilefree_status, 4); +OFFSET_CHECK(struct rogue_fwif_frag_ctx_state, ctx_state_flags, 8); +OFFSET_CHECK(struct rogue_fwif_frag_ctx_state, frag_reg_isp_store, 12); +SIZE_CHECK(struct rogue_fwif_frag_ctx_state, 16); + +OFFSET_CHECK(struct rogue_fwif_compute_ctx_state, ctx_state_flags, 0); +SIZE_CHECK(struct rogue_fwif_compute_ctx_state, 4); + +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, ccbctl_fw_addr, 0); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, ccb_fw_addr, 4); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, ccb_meta_dma_addr, 8); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, context_state_addr, 24); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, fw_com_ctx_flags, 28); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, priority, 32); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, priority_seq_num, 36); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, rf_cmd_addr, 40); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, stats_pending, 44); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, stats_num_stores, 48); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, stats_num_out_of_memory, 52); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, stats_num_partial_renders, 56); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, dm, 60); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, wait_signal_address, 64); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, wait_signal_node, 72); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, buf_stalled_node, 80); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, cbuf_queue_ctrl_addr, 88); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, robustness_address, 96); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, max_deadline_ms, 104); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, read_offset_needs_reset, 108); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, waiting_node, 112); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, run_node, 120); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, last_failed_ufo, 128); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, fw_mem_context_fw_addr, 136); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, server_common_context_id, 140); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, pid, 144); +OFFSET_CHECK(struct rogue_fwif_fwcommoncontext, geom_oom_disabled, 148); +SIZE_CHECK(struct rogue_fwif_fwcommoncontext, 152); + +OFFSET_CHECK(struct rogue_fwif_ccb_ctl, write_offset, 0); +OFFSET_CHECK(struct rogue_fwif_ccb_ctl, padding, 4); +OFFSET_CHECK(struct rogue_fwif_ccb_ctl, read_offset, 128); +OFFSET_CHECK(struct rogue_fwif_ccb_ctl, wrap_mask, 132); +OFFSET_CHECK(struct rogue_fwif_ccb_ctl, cmd_size, 136); +OFFSET_CHECK(struct rogue_fwif_ccb_ctl, padding2, 140); +SIZE_CHECK(struct rogue_fwif_ccb_ctl, 144); + +OFFSET_CHECK(struct rogue_fwif_kccb_cmd_kick_data, context_fw_addr, 0); +OFFSET_CHECK(struct rogue_fwif_kccb_cmd_kick_data, client_woff_update, 4); +OFFSET_CHECK(struct rogue_fwif_kccb_cmd_kick_data, client_wrap_mask_update, 8); +OFFSET_CHECK(struct rogue_fwif_kccb_cmd_kick_data, num_cleanup_ctl, 12); +OFFSET_CHECK(struct rogue_fwif_kccb_cmd_kick_data, cleanup_ctl_fw_addr, 16); +OFFSET_CHECK(struct rogue_fwif_kccb_cmd_kick_data, work_est_cmd_header_offset, 28); +SIZE_CHECK(struct rogue_fwif_kccb_cmd_kick_data, 32); + +OFFSET_CHECK(struct rogue_fwif_kccb_cmd_combined_geom_frag_kick_data, geom_cmd_kick_data, 0); +OFFSET_CHECK(struct rogue_fwif_kccb_cmd_combined_geom_frag_kick_data, frag_cmd_kick_data, 32); +SIZE_CHECK(struct rogue_fwif_kccb_cmd_combined_geom_frag_kick_data, 64); + +OFFSET_CHECK(struct rogue_fwif_kccb_cmd_force_update_data, context_fw_addr, 0); +OFFSET_CHECK(struct rogue_fwif_kccb_cmd_force_update_data, ccb_fence_offset, 4); +SIZE_CHECK(struct rogue_fwif_kccb_cmd_force_update_data, 8); + +OFFSET_CHECK(struct rogue_fwif_cleanup_request, cleanup_type, 0); +OFFSET_CHECK(struct rogue_fwif_cleanup_request, cleanup_data, 4); +SIZE_CHECK(struct rogue_fwif_cleanup_request, 8); + +OFFSET_CHECK(struct rogue_fwif_power_request, pow_type, 0); +OFFSET_CHECK(struct rogue_fwif_power_request, power_req_data, 4); +SIZE_CHECK(struct rogue_fwif_power_request, 8); + +OFFSET_CHECK(struct rogue_fwif_slcflushinvaldata, context_fw_addr, 0); +OFFSET_CHECK(struct rogue_fwif_slcflushinvaldata, inval, 4); +OFFSET_CHECK(struct rogue_fwif_slcflushinvaldata, dm_context, 8); +OFFSET_CHECK(struct rogue_fwif_slcflushinvaldata, address, 16); +OFFSET_CHECK(struct rogue_fwif_slcflushinvaldata, size, 24); +SIZE_CHECK(struct rogue_fwif_slcflushinvaldata, 32); + +OFFSET_CHECK(struct rogue_fwif_hwperf_ctrl, opcode, 0); +OFFSET_CHECK(struct rogue_fwif_hwperf_ctrl, mask, 8); +SIZE_CHECK(struct rogue_fwif_hwperf_ctrl, 16); + +OFFSET_CHECK(struct rogue_fwif_hwperf_config_enable_blks, num_blocks, 0); +OFFSET_CHECK(struct rogue_fwif_hwperf_config_enable_blks, block_configs_fw_addr, 4); +SIZE_CHECK(struct rogue_fwif_hwperf_config_enable_blks, 8); + +OFFSET_CHECK(struct rogue_fwif_hwperf_config_da_blks, num_blocks, 0); +OFFSET_CHECK(struct rogue_fwif_hwperf_config_da_blks, block_configs_fw_addr, 4); +SIZE_CHECK(struct rogue_fwif_hwperf_config_da_blks, 8); + +OFFSET_CHECK(struct rogue_fwif_coreclkspeedchange_data, new_clock_speed, 0); +SIZE_CHECK(struct rogue_fwif_coreclkspeedchange_data, 4); + +OFFSET_CHECK(struct rogue_fwif_hwperf_ctrl_blks, enable, 0); +OFFSET_CHECK(struct rogue_fwif_hwperf_ctrl_blks, num_blocks, 4); +OFFSET_CHECK(struct rogue_fwif_hwperf_ctrl_blks, block_ids, 8); +SIZE_CHECK(struct rogue_fwif_hwperf_ctrl_blks, 40); + +OFFSET_CHECK(struct rogue_fwif_hwperf_select_custom_cntrs, custom_block, 0); +OFFSET_CHECK(struct rogue_fwif_hwperf_select_custom_cntrs, num_counters, 2); +OFFSET_CHECK(struct rogue_fwif_hwperf_select_custom_cntrs, custom_counter_ids_fw_addr, 4); +SIZE_CHECK(struct rogue_fwif_hwperf_select_custom_cntrs, 8); + +OFFSET_CHECK(struct rogue_fwif_zsbuffer_backing_data, zs_buffer_fw_addr, 0); +OFFSET_CHECK(struct rogue_fwif_zsbuffer_backing_data, done, 4); +SIZE_CHECK(struct rogue_fwif_zsbuffer_backing_data, 8); + +OFFSET_CHECK(struct rogue_fwif_freelist_gs_data, freelist_fw_addr, 0); +OFFSET_CHECK(struct rogue_fwif_freelist_gs_data, delta_pages, 4); +OFFSET_CHECK(struct rogue_fwif_freelist_gs_data, new_pages, 8); +OFFSET_CHECK(struct rogue_fwif_freelist_gs_data, ready_pages, 12); +SIZE_CHECK(struct rogue_fwif_freelist_gs_data, 16); + +OFFSET_CHECK(struct rogue_fwif_freelists_reconstruction_data, freelist_count, 0); +OFFSET_CHECK(struct rogue_fwif_freelists_reconstruction_data, freelist_ids, 4); +SIZE_CHECK(struct rogue_fwif_freelists_reconstruction_data, 76); + +OFFSET_CHECK(struct rogue_fwif_write_offset_update_data, context_fw_addr, 0); +SIZE_CHECK(struct rogue_fwif_write_offset_update_data, 8); + +OFFSET_CHECK(struct rogue_fwif_kccb_cmd, cmd_type, 0); +OFFSET_CHECK(struct rogue_fwif_kccb_cmd, kccb_flags, 4); +OFFSET_CHECK(struct rogue_fwif_kccb_cmd, cmd_data, 8); +SIZE_CHECK(struct rogue_fwif_kccb_cmd, 88); + +OFFSET_CHECK(struct rogue_fwif_fwccb_cmd_context_reset_data, server_common_context_id, 0); +OFFSET_CHECK(struct rogue_fwif_fwccb_cmd_context_reset_data, reset_reason, 4); +OFFSET_CHECK(struct rogue_fwif_fwccb_cmd_context_reset_data, dm, 8); +OFFSET_CHECK(struct rogue_fwif_fwccb_cmd_context_reset_data, reset_job_ref, 12); +OFFSET_CHECK(struct rogue_fwif_fwccb_cmd_context_reset_data, flags, 16); +OFFSET_CHECK(struct rogue_fwif_fwccb_cmd_context_reset_data, pc_address, 24); +OFFSET_CHECK(struct rogue_fwif_fwccb_cmd_context_reset_data, fault_address, 32); +SIZE_CHECK(struct rogue_fwif_fwccb_cmd_context_reset_data, 40); + +OFFSET_CHECK(struct rogue_fwif_fwccb_cmd_fw_pagefault_data, fw_fault_addr, 0); +SIZE_CHECK(struct rogue_fwif_fwccb_cmd_fw_pagefault_data, 8); + +OFFSET_CHECK(struct rogue_fwif_fwccb_cmd, cmd_type, 0); +OFFSET_CHECK(struct rogue_fwif_fwccb_cmd, fwccb_flags, 4); +OFFSET_CHECK(struct rogue_fwif_fwccb_cmd, cmd_data, 8); +SIZE_CHECK(struct rogue_fwif_fwccb_cmd, 88); + +OFFSET_CHECK(struct rogue_fwif_ccb_cmd_header, cmd_type, 0); +OFFSET_CHECK(struct rogue_fwif_ccb_cmd_header, cmd_size, 4); +OFFSET_CHECK(struct rogue_fwif_ccb_cmd_header, ext_job_ref, 8); +OFFSET_CHECK(struct rogue_fwif_ccb_cmd_header, int_job_ref, 12); +OFFSET_CHECK(struct rogue_fwif_ccb_cmd_header, work_est_kick_data, 16); +SIZE_CHECK(struct rogue_fwif_ccb_cmd_header, 40); + +OFFSET_CHECK(struct rogue_fwif_runtime_cfg, active_pm_latency_ms, 0); +OFFSET_CHECK(struct rogue_fwif_runtime_cfg, runtime_cfg_flags, 4); +OFFSET_CHECK(struct rogue_fwif_runtime_cfg, active_pm_latency_persistant, 8); +OFFSET_CHECK(struct rogue_fwif_runtime_cfg, core_clock_speed, 12); +OFFSET_CHECK(struct rogue_fwif_runtime_cfg, default_dusts_num_init, 16); +OFFSET_CHECK(struct rogue_fwif_runtime_cfg, phr_mode, 20); +OFFSET_CHECK(struct rogue_fwif_runtime_cfg, hcs_deadline_ms, 24); +OFFSET_CHECK(struct rogue_fwif_runtime_cfg, wdg_period_us, 28); +OFFSET_CHECK(struct rogue_fwif_runtime_cfg, osid_priority, 32); +OFFSET_CHECK(struct rogue_fwif_runtime_cfg, hwperf_buf_fw_addr, 64); +OFFSET_CHECK(struct rogue_fwif_runtime_cfg, padding, 68); +SIZE_CHECK(struct rogue_fwif_runtime_cfg, 72); + +OFFSET_CHECK(struct rogue_fwif_connection_ctl, connection_fw_state, 0); +OFFSET_CHECK(struct rogue_fwif_connection_ctl, connection_os_state, 4); +OFFSET_CHECK(struct rogue_fwif_connection_ctl, alive_fw_token, 8); +OFFSET_CHECK(struct rogue_fwif_connection_ctl, alive_os_token, 12); +SIZE_CHECK(struct rogue_fwif_connection_ctl, 16); + +OFFSET_CHECK(struct rogue_fwif_compchecks_bvnc, layout_version, 0); +OFFSET_CHECK(struct rogue_fwif_compchecks_bvnc, bvnc, 8); +SIZE_CHECK(struct rogue_fwif_compchecks_bvnc, 16); + +OFFSET_CHECK(struct rogue_fwif_init_options, os_count_support, 0); +SIZE_CHECK(struct rogue_fwif_init_options, 8); + +OFFSET_CHECK(struct rogue_fwif_compchecks, hw_bvnc, 0); +OFFSET_CHECK(struct rogue_fwif_compchecks, fw_bvnc, 16); +OFFSET_CHECK(struct rogue_fwif_compchecks, fw_processor_version, 32); +OFFSET_CHECK(struct rogue_fwif_compchecks, ddk_version, 36); +OFFSET_CHECK(struct rogue_fwif_compchecks, ddk_build, 40); +OFFSET_CHECK(struct rogue_fwif_compchecks, build_options, 44); +OFFSET_CHECK(struct rogue_fwif_compchecks, init_options, 48); +OFFSET_CHECK(struct rogue_fwif_compchecks, updated, 56); +SIZE_CHECK(struct rogue_fwif_compchecks, 64); + +OFFSET_CHECK(struct rogue_fwif_osinit, kernel_ccbctl_fw_addr, 0); +OFFSET_CHECK(struct rogue_fwif_osinit, kernel_ccb_fw_addr, 4); +OFFSET_CHECK(struct rogue_fwif_osinit, kernel_ccb_rtn_slots_fw_addr, 8); +OFFSET_CHECK(struct rogue_fwif_osinit, firmware_ccbctl_fw_addr, 12); +OFFSET_CHECK(struct rogue_fwif_osinit, firmware_ccb_fw_addr, 16); +OFFSET_CHECK(struct rogue_fwif_osinit, work_est_firmware_ccbctl_fw_addr, 20); +OFFSET_CHECK(struct rogue_fwif_osinit, work_est_firmware_ccb_fw_addr, 24); +OFFSET_CHECK(struct rogue_fwif_osinit, rogue_fwif_hwr_info_buf_ctl_fw_addr, 28); +OFFSET_CHECK(struct rogue_fwif_osinit, hwr_debug_dump_limit, 32); +OFFSET_CHECK(struct rogue_fwif_osinit, fw_os_data_fw_addr, 36); +OFFSET_CHECK(struct rogue_fwif_osinit, rogue_comp_checks, 40); +SIZE_CHECK(struct rogue_fwif_osinit, 104); + +OFFSET_CHECK(struct rogue_fwif_sigbuf_ctl, buffer_fw_addr, 0); +OFFSET_CHECK(struct rogue_fwif_sigbuf_ctl, left_size_in_regs, 4); +SIZE_CHECK(struct rogue_fwif_sigbuf_ctl, 8); + +OFFSET_CHECK(struct pdvfs_opp, volt, 0); +OFFSET_CHECK(struct pdvfs_opp, freq, 4); +SIZE_CHECK(struct pdvfs_opp, 8); + +OFFSET_CHECK(struct rogue_fwif_pdvfs_opp, opp_values, 0); +OFFSET_CHECK(struct rogue_fwif_pdvfs_opp, min_opp_point, 128); +OFFSET_CHECK(struct rogue_fwif_pdvfs_opp, max_opp_point, 132); +SIZE_CHECK(struct rogue_fwif_pdvfs_opp, 136); + +OFFSET_CHECK(struct rogue_fwif_counter_dump_ctl, buffer_fw_addr, 0); +OFFSET_CHECK(struct rogue_fwif_counter_dump_ctl, size_in_dwords, 4); +SIZE_CHECK(struct rogue_fwif_counter_dump_ctl, 8); + +OFFSET_CHECK(struct rogue_hwperf_bvnc, bvnc_string, 0); +OFFSET_CHECK(struct rogue_hwperf_bvnc, bvnc_km_feature_flags, 24); +OFFSET_CHECK(struct rogue_hwperf_bvnc, num_bvnc_blocks, 28); +OFFSET_CHECK(struct rogue_hwperf_bvnc, bvnc_gpu_cores, 30); +OFFSET_CHECK(struct rogue_hwperf_bvnc, bvnc_blocks, 32); +SIZE_CHECK(struct rogue_hwperf_bvnc, 160); + +OFFSET_CHECK(struct rogue_fwif_sysinit, fault_phys_addr, 0); +OFFSET_CHECK(struct rogue_fwif_sysinit, pds_exec_base, 8); +OFFSET_CHECK(struct rogue_fwif_sysinit, usc_exec_base, 16); +OFFSET_CHECK(struct rogue_fwif_sysinit, fbcdc_state_table_base, 24); +OFFSET_CHECK(struct rogue_fwif_sysinit, fbcdc_large_state_table_base, 32); +OFFSET_CHECK(struct rogue_fwif_sysinit, texture_heap_base, 40); +OFFSET_CHECK(struct rogue_fwif_sysinit, hw_perf_filter, 48); +OFFSET_CHECK(struct rogue_fwif_sysinit, slc3_fence_dev_addr, 56); +OFFSET_CHECK(struct rogue_fwif_sysinit, tpu_trilinear_frac_mask, 64); +OFFSET_CHECK(struct rogue_fwif_sysinit, sigbuf_ctl, 80); +OFFSET_CHECK(struct rogue_fwif_sysinit, pdvfs_opp_info, 152); +OFFSET_CHECK(struct rogue_fwif_sysinit, coremem_data_store, 288); +OFFSET_CHECK(struct rogue_fwif_sysinit, counter_dump_ctl, 304); +OFFSET_CHECK(struct rogue_fwif_sysinit, filter_flags, 312); +OFFSET_CHECK(struct rogue_fwif_sysinit, runtime_cfg_fw_addr, 316); +OFFSET_CHECK(struct rogue_fwif_sysinit, trace_buf_ctl_fw_addr, 320); +OFFSET_CHECK(struct rogue_fwif_sysinit, fw_sys_data_fw_addr, 324); +OFFSET_CHECK(struct rogue_fwif_sysinit, gpu_util_fw_cb_ctl_fw_addr, 328); +OFFSET_CHECK(struct rogue_fwif_sysinit, reg_cfg_fw_addr, 332); +OFFSET_CHECK(struct rogue_fwif_sysinit, hwperf_ctl_fw_addr, 336); +OFFSET_CHECK(struct rogue_fwif_sysinit, align_checks, 340); +OFFSET_CHECK(struct rogue_fwif_sysinit, initial_core_clock_speed, 344); +OFFSET_CHECK(struct rogue_fwif_sysinit, active_pm_latency_ms, 348); +OFFSET_CHECK(struct rogue_fwif_sysinit, firmware_started, 352); +OFFSET_CHECK(struct rogue_fwif_sysinit, marker_val, 356); +OFFSET_CHECK(struct rogue_fwif_sysinit, firmware_started_timestamp, 360); +OFFSET_CHECK(struct rogue_fwif_sysinit, jones_disable_mask, 364); +OFFSET_CHECK(struct rogue_fwif_sysinit, firmware_perf, 368); +OFFSET_CHECK(struct rogue_fwif_sysinit, core_clock_rate_fw_addr, 372); +OFFSET_CHECK(struct rogue_fwif_sysinit, gpio_validation_mode, 376); +OFFSET_CHECK(struct rogue_fwif_sysinit, bvnc_km_feature_flags, 380); +OFFSET_CHECK(struct rogue_fwif_sysinit, tfbc_compression_control, 540); +SIZE_CHECK(struct rogue_fwif_sysinit, 544); + +OFFSET_CHECK(struct rogue_fwif_gpu_util_fwcb, time_corr, 0); +OFFSET_CHECK(struct rogue_fwif_gpu_util_fwcb, time_corr_seq_count, 10240); +OFFSET_CHECK(struct rogue_fwif_gpu_util_fwcb, gpu_util_flags, 10244); +OFFSET_CHECK(struct rogue_fwif_gpu_util_fwcb, last_word, 10248); +OFFSET_CHECK(struct rogue_fwif_gpu_util_fwcb, stats_counters, 10256); +SIZE_CHECK(struct rogue_fwif_gpu_util_fwcb, 10280); + +OFFSET_CHECK(struct rogue_fwif_rta_ctl, render_target_index, 0); +OFFSET_CHECK(struct rogue_fwif_rta_ctl, current_render_target, 4); +OFFSET_CHECK(struct rogue_fwif_rta_ctl, active_render_targets, 8); +OFFSET_CHECK(struct rogue_fwif_rta_ctl, cumul_active_render_targets, 12); +OFFSET_CHECK(struct rogue_fwif_rta_ctl, valid_render_targets_fw_addr, 16); +OFFSET_CHECK(struct rogue_fwif_rta_ctl, rta_num_partial_renders_fw_addr, 20); +OFFSET_CHECK(struct rogue_fwif_rta_ctl, max_rts, 24); +OFFSET_CHECK(struct rogue_fwif_rta_ctl, rta_ctl_flags, 28); +SIZE_CHECK(struct rogue_fwif_rta_ctl, 32); + +OFFSET_CHECK(struct rogue_fwif_freelist, freelist_dev_addr, 0); +OFFSET_CHECK(struct rogue_fwif_freelist, current_dev_addr, 8); +OFFSET_CHECK(struct rogue_fwif_freelist, current_stack_top, 16); +OFFSET_CHECK(struct rogue_fwif_freelist, max_pages, 20); +OFFSET_CHECK(struct rogue_fwif_freelist, grow_pages, 24); +OFFSET_CHECK(struct rogue_fwif_freelist, current_pages, 28); +OFFSET_CHECK(struct rogue_fwif_freelist, allocated_page_count, 32); +OFFSET_CHECK(struct rogue_fwif_freelist, allocated_mmu_page_count, 36); +OFFSET_CHECK(struct rogue_fwif_freelist, freelist_id, 40); +OFFSET_CHECK(struct rogue_fwif_freelist, grow_pending, 44); +OFFSET_CHECK(struct rogue_fwif_freelist, ready_pages, 48); +OFFSET_CHECK(struct rogue_fwif_freelist, freelist_flags, 52); +OFFSET_CHECK(struct rogue_fwif_freelist, pm_global_pb, 56); +SIZE_CHECK(struct rogue_fwif_freelist, 64); + +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, geom_caches_need_zeroing, 0); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, screen_pixel_max, 4); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, multi_sample_ctl, 8); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, flipped_multi_sample_ctl, 16); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, tpc_stride, 24); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, tpc_size, 28); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, te_screen, 32); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, mtile_stride, 36); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, teaa, 40); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, te_mtile1, 44); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, te_mtile2, 48); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, isp_merge_lower_x, 52); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, isp_merge_lower_y, 56); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, isp_merge_upper_x, 60); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, isp_merge_upper_y, 64); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, isp_merge_scale_x, 68); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, isp_merge_scale_y, 72); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, rgn_header_size, 76); +OFFSET_CHECK(struct rogue_fwif_hwrtdata_common, isp_mtile_size, 80); +SIZE_CHECK(struct rogue_fwif_hwrtdata_common, 88); + +OFFSET_CHECK(struct rogue_fwif_hwrtdata, pm_mlist_dev_addr, 0); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, vce_cat_base, 8); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, vce_last_cat_base, 40); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, te_cat_base, 72); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, te_last_cat_base, 104); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, alist_cat_base, 136); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, alist_last_cat_base, 144); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, pm_alist_stack_pointer, 152); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, pm_mlist_stack_pointer, 160); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, hwrt_data_common_fw_addr, 164); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, hwrt_data_flags, 168); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, state, 172); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, freelists_fw_addr, 176); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, freelist_hwr_snapshot, 188); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, vheap_table_dev_addr, 200); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, rta_ctl, 208); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, tail_ptrs_dev_addr, 240); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, macrotile_array_dev_addr, 248); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, rgn_header_dev_addr, 256); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, rtc_dev_addr, 264); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, owner_geom_not_used_by_host, 272); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, geom_caches_need_zeroing, 276); +OFFSET_CHECK(struct rogue_fwif_hwrtdata, cleanup_state, 320); +SIZE_CHECK(struct rogue_fwif_hwrtdata, 384); + +OFFSET_CHECK(struct rogue_fwif_sync_checkpoint, state, 0); +OFFSET_CHECK(struct rogue_fwif_sync_checkpoint, fw_ref_count, 4); +SIZE_CHECK(struct rogue_fwif_sync_checkpoint, 8); + +#endif /* PVR_ROGUE_FWIF_CHECK_H */ diff --git a/drivers/gpu/drm/imagination/pvr_rogue_fwif_client.h b/drivers/gpu/drm/imagination/pvr_rogue_fwif_client.h new file mode 100644 index 000000000000..6e224400083a --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_fwif_client.h @@ -0,0 +1,373 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_ROGUE_FWIF_CLIENT_H +#define PVR_ROGUE_FWIF_CLIENT_H + +#include +#include +#include +#include + +#include "pvr_rogue_fwif_shared.h" + +/* + * Page size used for Parameter Management. + */ +#define ROGUE_PM_PAGE_SIZE SZ_4K + +/* + * Minimum/Maximum PB size. + * + * Base page size is dependent on core: + * S6/S6XT/S7 = 50 pages + * S8XE = 40 pages + * S8XE with BRN66011 fixed = 25 pages + * + * Minimum PB = Base Pages + (NUM_TE_PIPES-1)*16K + (NUM_VCE_PIPES-1)*64K + + * IF_PM_PREALLOC(NUM_TE_PIPES*16K + NUM_VCE_PIPES*16K) + * + * Maximum PB size must ensure that no PM address space can be fully used, + * because if the full address space was used it would wrap and corrupt itself. + * Since there are two freelists (local is always minimum sized) this can be + * described as following three conditions being met: + * + * (Minimum PB + Maximum PB) < ALIST PM address space size (16GB) + * (Minimum PB + Maximum PB) < TE PM address space size (16GB) / NUM_TE_PIPES + * (Minimum PB + Maximum PB) < VCE PM address space size (16GB) / NUM_VCE_PIPES + * + * Since the max of NUM_TE_PIPES and NUM_VCE_PIPES is 4, we have a hard limit + * of 4GB minus the Minimum PB. For convenience we take the smaller power-of-2 + * value of 2GB. This is far more than any current applications use. + */ +#define ROGUE_PM_MAX_FREELIST_SIZE SZ_2G + +/* + * Flags supported by the geometry DM command i.e. &struct rogue_fwif_cmd_geom. + */ + +#define ROGUE_GEOM_FLAGS_FIRSTKICK BIT_MASK(0) +#define ROGUE_GEOM_FLAGS_LASTKICK BIT_MASK(1) +/* Use single core in a multi core setup. */ +#define ROGUE_GEOM_FLAGS_SINGLE_CORE BIT_MASK(3) + +/* + * Flags supported by the fragment DM command i.e. &struct rogue_fwif_cmd_frag. + */ + +/* Use single core in a multi core setup. */ +#define ROGUE_FRAG_FLAGS_SINGLE_CORE BIT_MASK(3) +/* Indicates whether this render produces visibility results. */ +#define ROGUE_FRAG_FLAGS_GET_VIS_RESULTS BIT_MASK(5) +/* Indicates whether a depth buffer is present. */ +#define ROGUE_FRAG_FLAGS_DEPTHBUFFER BIT_MASK(7) +/* Indicates whether a stencil buffer is present. */ +#define ROGUE_FRAG_FLAGS_STENCILBUFFER BIT_MASK(8) +/* Disable pixel merging for this render. */ +#define ROGUE_FRAG_FLAGS_DISABLE_PIXELMERGE BIT_MASK(15) +/* Indicates whether a scratch buffer is present. */ +#define ROGUE_FRAG_FLAGS_SCRATCHBUFFER BIT_MASK(19) +/* Disallow compute overlapped with this render. */ +#define ROGUE_FRAG_FLAGS_PREVENT_CDM_OVERLAP BIT_MASK(26) + +/* + * Flags supported by the compute DM command i.e. &struct rogue_fwif_cmd_compute. + */ + +#define ROGUE_COMPUTE_FLAG_PREVENT_ALL_OVERLAP BIT_MASK(2) +/*!< Use single core in a multi core setup. */ +#define ROGUE_COMPUTE_FLAG_SINGLE_CORE BIT_MASK(5) + +/* + * Flags supported by the transfer DM command i.e. &struct rogue_fwif_cmd_transfer. + */ + +/*!< Use single core in a multi core setup. */ +#define ROGUE_TRANSFER_FLAGS_SINGLE_CORE BIT_MASK(1) + +/* + ************************************************ + * Parameter/HWRTData control structures. + ************************************************ + */ + +/* + * Configuration registers which need to be loaded by the firmware before a geometry + * job can be started. + */ +struct rogue_fwif_geom_regs { + u64 vdm_ctrl_stream_base; + u64 tpu_border_colour_table; + + /* Only used when feature VDM_DRAWINDIRECT present. */ + u64 vdm_draw_indirect0; + /* Only used when feature VDM_DRAWINDIRECT present. */ + u32 vdm_draw_indirect1; + + u32 ppp_ctrl; + u32 te_psg; + /* Only used when BRN 49927 present. */ + u32 tpu; + + u32 vdm_context_resume_task0_size; + /* Only used when feature VDM_OBJECT_LEVEL_LLS present. */ + u32 vdm_context_resume_task3_size; + + /* Only used when BRN 56279 or BRN 67381 present. */ + u32 pds_ctrl; + + u32 view_idx; + + /* Only used when feature TESSELLATION present */ + u32 pds_coeff_free_prog; + + u32 padding; +}; + +/* Only used when BRN 44455 or BRN 63027 present. */ +struct rogue_fwif_dummy_rgnhdr_init_geom_regs { + u64 te_psgregion_addr; +}; + +/* + * Represents a geometry command that can be used to tile a whole scene's objects as + * per TA behavior. + */ +struct rogue_fwif_cmd_geom { + /* + * rogue_fwif_cmd_geom_frag_shared field must always be at the beginning of the + * struct. + * + * The command struct (rogue_fwif_cmd_geom) is shared between Client and + * Firmware. Kernel is unable to perform read/write operations on the + * command struct, the SHARED region is the only exception from this rule. + * This region must be the first member so that Kernel can easily access it. + * For more info, see rogue_fwif_cmd_geom_frag_shared definition. + */ + struct rogue_fwif_cmd_geom_frag_shared cmd_shared; + + struct rogue_fwif_geom_regs regs __aligned(8); + u32 flags __aligned(8); + + /* + * Holds the geometry/fragment fence value to allow the fragment partial render command + * to go through. + */ + struct rogue_fwif_ufo partial_render_geom_frag_fence; + + /* Only used when BRN 44455 or BRN 63027 present. */ + struct rogue_fwif_dummy_rgnhdr_init_geom_regs dummy_rgnhdr_init_geom_regs __aligned(8); + + /* Only used when BRN 61484 or BRN 66333 present. */ + u32 brn61484_66333_live_rt; + + u32 padding; +}; + +/* + * Configuration registers which need to be loaded by the firmware before ISP + * can be started. + */ +struct rogue_fwif_frag_regs { + u32 usc_pixel_output_ctrl; + +#define ROGUE_MAXIMUM_OUTPUT_REGISTERS_PER_PIXEL 8U + u32 usc_clear_register[ROGUE_MAXIMUM_OUTPUT_REGISTERS_PER_PIXEL]; + + u32 isp_bgobjdepth; + u32 isp_bgobjvals; + u32 isp_aa; + /* Only used when feature S7_TOP_INFRASTRUCTURE present. */ + u32 isp_xtp_pipe_enable; + + u32 isp_ctl; + + /* Only used when BRN 49927 present. */ + u32 tpu; + + u32 event_pixel_pds_info; + + /* Only used when feature CLUSTER_GROUPING present. */ + u32 pixel_phantom; + + u32 view_idx; + + u32 event_pixel_pds_data; + + /* Only used when BRN 65101 present. */ + u32 brn65101_event_pixel_pds_data; + + /* Only used when feature GPU_MULTICORE_SUPPORT or BRN 47217 present. */ + u32 isp_oclqry_stride; + + /* Only used when feature ZLS_SUBTILE present. */ + u32 isp_zls_pixels; + + /* Only used when feature ISP_ZLS_D24_S8_PACKING_OGL_MODE present. */ + u32 rgx_cr_blackpearl_fix; + + /* All values below the ALIGN(8) must be 64 bit. */ + aligned_u64 isp_scissor_base; + u64 isp_dbias_base; + u64 isp_oclqry_base; + u64 isp_zlsctl; + u64 isp_zload_store_base; + u64 isp_stencil_load_store_base; + + /* + * Only used when feature FBCDC_ALGORITHM present and value < 3 or feature + * FB_CDC_V4 present. Additionally, BRNs 48754, 60227, 72310 and 72311 must + * not be present. + */ + u64 fb_cdc_zls; + +#define ROGUE_PBE_WORDS_REQUIRED_FOR_RENDERS 3U + u64 pbe_word[8U][ROGUE_PBE_WORDS_REQUIRED_FOR_RENDERS]; + u64 tpu_border_colour_table; + u64 pds_bgnd[3U]; + + /* Only used when BRN 65101 present. */ + u64 pds_bgnd_brn65101[3U]; + + u64 pds_pr_bgnd[3U]; + + /* Only used when BRN 62850 or 62865 present. */ + u64 isp_dummy_stencil_store_base; + + /* Only used when BRN 66193 present. */ + u64 isp_dummy_depth_store_base; + + /* Only used when BRN 67182 present. */ + u32 rgnhdr_single_rt_size; + /* Only used when BRN 67182 present. */ + u32 rgnhdr_scratch_offset; +}; + +struct rogue_fwif_cmd_frag { + struct rogue_fwif_cmd_geom_frag_shared cmd_shared __aligned(8); + + struct rogue_fwif_frag_regs regs __aligned(8); + /* command control flags. */ + u32 flags; + /* Stride IN BYTES for Z-Buffer in case of RTAs. */ + u32 zls_stride; + /* Stride IN BYTES for S-Buffer in case of RTAs. */ + u32 sls_stride; + + /* Only used if feature GPU_MULTICORE_SUPPORT present. */ + u32 execute_count; +}; + +/* + * Configuration registers which need to be loaded by the firmware before CDM + * can be started. + */ +struct rogue_fwif_compute_regs { + u64 tpu_border_colour_table; + + /* Only used when feature CDM_USER_MODE_QUEUE present. */ + u64 cdm_cb_queue; + + /* Only used when feature CDM_USER_MODE_QUEUE present. */ + u64 cdm_cb_base; + /* Only used when feature CDM_USER_MODE_QUEUE present. */ + u64 cdm_cb; + + /* Only used when feature CDM_USER_MODE_QUEUE is not present. */ + u64 cdm_ctrl_stream_base; + + u64 cdm_context_state_base_addr; + + /* Only used when BRN 49927 is present. */ + u32 tpu; + u32 cdm_resume_pds1; + + /* Only used when feature COMPUTE_MORTON_CAPABLE present. */ + u32 cdm_item; + + /* Only used when feature CLUSTER_GROUPING present. */ + u32 compute_cluster; + + /* Only used when feature TPU_DM_GLOBAL_REGISTERS present. */ + u32 tpu_tag_cdm_ctrl; + + u32 padding; +}; + +struct rogue_fwif_cmd_compute { + /* Common command attributes */ + struct rogue_fwif_cmd_common common __aligned(8); + + /* CDM registers */ + struct rogue_fwif_compute_regs regs; + + /* Control flags */ + u32 flags __aligned(8); + + /* Only used when feature UNIFIED_STORE_VIRTUAL_PARTITIONING present. */ + u32 num_temp_regions; + + /* Only used when feature CDM_USER_MODE_QUEUE present. */ + u32 stream_start_offset; + + /* Only used when feature GPU_MULTICORE_SUPPORT present. */ + u32 execute_count; +}; + +struct rogue_fwif_transfer_regs { + /* + * All 32 bit values should be added in the top section. This then requires only a + * single RGXFW_ALIGN to align all the 64 bit values in the second section. + */ + u32 isp_bgobjvals; + + u32 usc_pixel_output_ctrl; + u32 usc_clear_register0; + u32 usc_clear_register1; + u32 usc_clear_register2; + u32 usc_clear_register3; + + u32 isp_mtile_size; + u32 isp_render_origin; + u32 isp_ctl; + + /* Only used when feature S7_TOP_INFRASTRUCTURE present. */ + u32 isp_xtp_pipe_enable; + u32 isp_aa; + + u32 event_pixel_pds_info; + + u32 event_pixel_pds_code; + u32 event_pixel_pds_data; + + u32 isp_render; + u32 isp_rgn; + + /* Only used when feature GPU_MULTICORE_SUPPORT present. */ + u32 frag_screen; + + /* All values below the aligned_u64 must be 64 bit. */ + aligned_u64 pds_bgnd0_base; + u64 pds_bgnd1_base; + u64 pds_bgnd3_sizeinfo; + + u64 isp_mtile_base; +#define ROGUE_PBE_WORDS_REQUIRED_FOR_TQS 3 + /* TQ_MAX_RENDER_TARGETS * PBE_STATE_SIZE */ + u64 pbe_wordx_mrty[3U * ROGUE_PBE_WORDS_REQUIRED_FOR_TQS]; +}; + +struct rogue_fwif_cmd_transfer { + /* Common command attributes */ + struct rogue_fwif_cmd_common common __aligned(8); + + struct rogue_fwif_transfer_regs regs __aligned(8); + + u32 flags; + + u32 padding; +}; + +#include "pvr_rogue_fwif_client_check.h" + +#endif /* PVR_ROGUE_FWIF_CLIENT_H */ diff --git a/drivers/gpu/drm/imagination/pvr_rogue_fwif_client_check.h b/drivers/gpu/drm/imagination/pvr_rogue_fwif_client_check.h new file mode 100644 index 000000000000..54aa4474163e --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_fwif_client_check.h @@ -0,0 +1,133 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_ROGUE_FWIF_CLIENT_CHECK_H +#define PVR_ROGUE_FWIF_CLIENT_CHECK_H + +#include + +#define OFFSET_CHECK(type, member, offset) \ + static_assert(offsetof(type, member) == (offset), \ + "offsetof(" #type ", " #member ") incorrect") + +#define SIZE_CHECK(type, size) \ + static_assert(sizeof(type) == (size), #type " is incorrect size") + +OFFSET_CHECK(struct rogue_fwif_geom_regs, vdm_ctrl_stream_base, 0); +OFFSET_CHECK(struct rogue_fwif_geom_regs, tpu_border_colour_table, 8); +OFFSET_CHECK(struct rogue_fwif_geom_regs, vdm_draw_indirect0, 16); +OFFSET_CHECK(struct rogue_fwif_geom_regs, vdm_draw_indirect1, 24); +OFFSET_CHECK(struct rogue_fwif_geom_regs, ppp_ctrl, 28); +OFFSET_CHECK(struct rogue_fwif_geom_regs, te_psg, 32); +OFFSET_CHECK(struct rogue_fwif_geom_regs, tpu, 36); +OFFSET_CHECK(struct rogue_fwif_geom_regs, vdm_context_resume_task0_size, 40); +OFFSET_CHECK(struct rogue_fwif_geom_regs, vdm_context_resume_task3_size, 44); +OFFSET_CHECK(struct rogue_fwif_geom_regs, pds_ctrl, 48); +OFFSET_CHECK(struct rogue_fwif_geom_regs, view_idx, 52); +OFFSET_CHECK(struct rogue_fwif_geom_regs, pds_coeff_free_prog, 56); +SIZE_CHECK(struct rogue_fwif_geom_regs, 64); + +OFFSET_CHECK(struct rogue_fwif_dummy_rgnhdr_init_geom_regs, te_psgregion_addr, 0); +SIZE_CHECK(struct rogue_fwif_dummy_rgnhdr_init_geom_regs, 8); + +OFFSET_CHECK(struct rogue_fwif_cmd_geom, cmd_shared, 0); +OFFSET_CHECK(struct rogue_fwif_cmd_geom, regs, 16); +OFFSET_CHECK(struct rogue_fwif_cmd_geom, flags, 80); +OFFSET_CHECK(struct rogue_fwif_cmd_geom, partial_render_geom_frag_fence, 84); +OFFSET_CHECK(struct rogue_fwif_cmd_geom, dummy_rgnhdr_init_geom_regs, 96); +OFFSET_CHECK(struct rogue_fwif_cmd_geom, brn61484_66333_live_rt, 104); +SIZE_CHECK(struct rogue_fwif_cmd_geom, 112); + +OFFSET_CHECK(struct rogue_fwif_frag_regs, usc_pixel_output_ctrl, 0); +OFFSET_CHECK(struct rogue_fwif_frag_regs, usc_clear_register, 4); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_bgobjdepth, 36); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_bgobjvals, 40); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_aa, 44); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_xtp_pipe_enable, 48); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_ctl, 52); +OFFSET_CHECK(struct rogue_fwif_frag_regs, tpu, 56); +OFFSET_CHECK(struct rogue_fwif_frag_regs, event_pixel_pds_info, 60); +OFFSET_CHECK(struct rogue_fwif_frag_regs, pixel_phantom, 64); +OFFSET_CHECK(struct rogue_fwif_frag_regs, view_idx, 68); +OFFSET_CHECK(struct rogue_fwif_frag_regs, event_pixel_pds_data, 72); +OFFSET_CHECK(struct rogue_fwif_frag_regs, brn65101_event_pixel_pds_data, 76); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_oclqry_stride, 80); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_zls_pixels, 84); +OFFSET_CHECK(struct rogue_fwif_frag_regs, rgx_cr_blackpearl_fix, 88); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_scissor_base, 96); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_dbias_base, 104); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_oclqry_base, 112); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_zlsctl, 120); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_zload_store_base, 128); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_stencil_load_store_base, 136); +OFFSET_CHECK(struct rogue_fwif_frag_regs, fb_cdc_zls, 144); +OFFSET_CHECK(struct rogue_fwif_frag_regs, pbe_word, 152); +OFFSET_CHECK(struct rogue_fwif_frag_regs, tpu_border_colour_table, 344); +OFFSET_CHECK(struct rogue_fwif_frag_regs, pds_bgnd, 352); +OFFSET_CHECK(struct rogue_fwif_frag_regs, pds_bgnd_brn65101, 376); +OFFSET_CHECK(struct rogue_fwif_frag_regs, pds_pr_bgnd, 400); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_dummy_stencil_store_base, 424); +OFFSET_CHECK(struct rogue_fwif_frag_regs, isp_dummy_depth_store_base, 432); +OFFSET_CHECK(struct rogue_fwif_frag_regs, rgnhdr_single_rt_size, 440); +OFFSET_CHECK(struct rogue_fwif_frag_regs, rgnhdr_scratch_offset, 444); +SIZE_CHECK(struct rogue_fwif_frag_regs, 448); + +OFFSET_CHECK(struct rogue_fwif_cmd_frag, cmd_shared, 0); +OFFSET_CHECK(struct rogue_fwif_cmd_frag, regs, 16); +OFFSET_CHECK(struct rogue_fwif_cmd_frag, flags, 464); +OFFSET_CHECK(struct rogue_fwif_cmd_frag, zls_stride, 468); +OFFSET_CHECK(struct rogue_fwif_cmd_frag, sls_stride, 472); +OFFSET_CHECK(struct rogue_fwif_cmd_frag, execute_count, 476); +SIZE_CHECK(struct rogue_fwif_cmd_frag, 480); + +OFFSET_CHECK(struct rogue_fwif_compute_regs, tpu_border_colour_table, 0); +OFFSET_CHECK(struct rogue_fwif_compute_regs, cdm_cb_queue, 8); +OFFSET_CHECK(struct rogue_fwif_compute_regs, cdm_cb_base, 16); +OFFSET_CHECK(struct rogue_fwif_compute_regs, cdm_cb, 24); +OFFSET_CHECK(struct rogue_fwif_compute_regs, cdm_ctrl_stream_base, 32); +OFFSET_CHECK(struct rogue_fwif_compute_regs, cdm_context_state_base_addr, 40); +OFFSET_CHECK(struct rogue_fwif_compute_regs, tpu, 48); +OFFSET_CHECK(struct rogue_fwif_compute_regs, cdm_resume_pds1, 52); +OFFSET_CHECK(struct rogue_fwif_compute_regs, cdm_item, 56); +OFFSET_CHECK(struct rogue_fwif_compute_regs, compute_cluster, 60); +OFFSET_CHECK(struct rogue_fwif_compute_regs, tpu_tag_cdm_ctrl, 64); +SIZE_CHECK(struct rogue_fwif_compute_regs, 72); + +OFFSET_CHECK(struct rogue_fwif_cmd_compute, common, 0); +OFFSET_CHECK(struct rogue_fwif_cmd_compute, regs, 8); +OFFSET_CHECK(struct rogue_fwif_cmd_compute, flags, 80); +OFFSET_CHECK(struct rogue_fwif_cmd_compute, num_temp_regions, 84); +OFFSET_CHECK(struct rogue_fwif_cmd_compute, stream_start_offset, 88); +OFFSET_CHECK(struct rogue_fwif_cmd_compute, execute_count, 92); +SIZE_CHECK(struct rogue_fwif_cmd_compute, 96); + +OFFSET_CHECK(struct rogue_fwif_transfer_regs, isp_bgobjvals, 0); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, usc_pixel_output_ctrl, 4); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, usc_clear_register0, 8); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, usc_clear_register1, 12); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, usc_clear_register2, 16); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, usc_clear_register3, 20); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, isp_mtile_size, 24); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, isp_render_origin, 28); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, isp_ctl, 32); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, isp_xtp_pipe_enable, 36); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, isp_aa, 40); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, event_pixel_pds_info, 44); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, event_pixel_pds_code, 48); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, event_pixel_pds_data, 52); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, isp_render, 56); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, isp_rgn, 60); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, frag_screen, 64); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, pds_bgnd0_base, 72); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, pds_bgnd1_base, 80); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, pds_bgnd3_sizeinfo, 88); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, isp_mtile_base, 96); +OFFSET_CHECK(struct rogue_fwif_transfer_regs, pbe_wordx_mrty, 104); +SIZE_CHECK(struct rogue_fwif_transfer_regs, 176); + +OFFSET_CHECK(struct rogue_fwif_cmd_transfer, common, 0); +OFFSET_CHECK(struct rogue_fwif_cmd_transfer, regs, 8); +OFFSET_CHECK(struct rogue_fwif_cmd_transfer, flags, 184); +SIZE_CHECK(struct rogue_fwif_cmd_transfer, 192); + +#endif /* PVR_ROGUE_FWIF_CLIENT_CHECK_H */ diff --git a/drivers/gpu/drm/imagination/pvr_rogue_fwif_common.h b/drivers/gpu/drm/imagination/pvr_rogue_fwif_common.h new file mode 100644 index 000000000000..6ebb95ba98a6 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_fwif_common.h @@ -0,0 +1,60 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_ROGUE_FWIF_COMMON_H +#define PVR_ROGUE_FWIF_COMMON_H + +#include + +/* + * This macro represents a mask of LSBs that must be zero on data structure + * sizes and offsets to ensure they are 8-byte granular on types shared between + * the FW and host driver. + */ +#define PVR_FW_ALIGNMENT_LSB 7U + +/* Macro to test structure size alignment. */ +#define PVR_FW_STRUCT_SIZE_ASSERT(_a) \ + static_assert((sizeof(_a) & PVR_FW_ALIGNMENT_LSB) == 0U, \ + "Size of " #_a " is not properly aligned") + +/* The master definition for data masters known to the firmware. */ + +#define PVR_FWIF_DM_GP (0) +/* Either TDM or 2D DM is present. */ +/* When the 'tla' feature is present in the hw (as per @pvr_device_features). */ +#define PVR_FWIF_DM_2D (1) +/* + * When the 'fastrender_dm' feature is present in the hw (as per + * @pvr_device_features). + */ +#define PVR_FWIF_DM_TDM (1) + +#define PVR_FWIF_DM_GEOM (2) +#define PVR_FWIF_DM_FRAG (3) +#define PVR_FWIF_DM_CDM (4) +#define PVR_FWIF_DM_RAY (5) +#define PVR_FWIF_DM_GEOM2 (6) +#define PVR_FWIF_DM_GEOM3 (7) +#define PVR_FWIF_DM_GEOM4 (8) + +#define PVR_FWIF_DM_LAST PVR_FWIF_DM_GEOM4 + +/* Maximum number of DM in use: GP, 2D/TDM, GEOM, 3D, CDM, RAY, GEOM2, GEOM3, GEOM4 */ +#define PVR_FWIF_DM_MAX (PVR_FWIF_DM_LAST + 1U) + +/* GPU Utilisation states */ +#define PVR_FWIF_GPU_UTIL_STATE_IDLE 0U +#define PVR_FWIF_GPU_UTIL_STATE_ACTIVE 1U +#define PVR_FWIF_GPU_UTIL_STATE_BLOCKED 2U +#define PVR_FWIF_GPU_UTIL_STATE_NUM 3U +#define PVR_FWIF_GPU_UTIL_STATE_MASK 0x3ULL + +/* + * Maximum amount of register writes that can be done by the register + * programmer (FW or META DMA). This is not a HW limitation, it is only + * a protection against malformed inputs to the register programmer. + */ +#define PVR_MAX_NUM_REGISTER_PROGRAMMER_WRITES 128U + +#endif /* PVR_ROGUE_FWIF_COMMON_H */ diff --git a/drivers/gpu/drm/imagination/pvr_rogue_fwif_dev_info.h b/drivers/gpu/drm/imagination/pvr_rogue_fwif_dev_info.h new file mode 100644 index 000000000000..168277bce948 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_fwif_dev_info.h @@ -0,0 +1,113 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef __PVR_ROGUE_FWIF_DEV_INFO_H__ +#define __PVR_ROGUE_FWIF_DEV_INFO_H__ + +enum { + PVR_FW_HAS_BRN_44079 = 0, + PVR_FW_HAS_BRN_47217, + PVR_FW_HAS_BRN_48492, + PVR_FW_HAS_BRN_48545, + PVR_FW_HAS_BRN_49927, + PVR_FW_HAS_BRN_50767, + PVR_FW_HAS_BRN_51764, + PVR_FW_HAS_BRN_62269, + PVR_FW_HAS_BRN_63142, + PVR_FW_HAS_BRN_63553, + PVR_FW_HAS_BRN_66011, + PVR_FW_HAS_BRN_71242, + + PVR_FW_HAS_BRN_MAX +}; + +enum { + PVR_FW_HAS_ERN_35421 = 0, + PVR_FW_HAS_ERN_38020, + PVR_FW_HAS_ERN_38748, + PVR_FW_HAS_ERN_42064, + PVR_FW_HAS_ERN_42290, + PVR_FW_HAS_ERN_42606, + PVR_FW_HAS_ERN_47025, + PVR_FW_HAS_ERN_57596, + + PVR_FW_HAS_ERN_MAX +}; + +enum { + PVR_FW_HAS_FEATURE_AXI_ACELITE = 0, + PVR_FW_HAS_FEATURE_CDM_CONTROL_STREAM_FORMAT, + PVR_FW_HAS_FEATURE_CLUSTER_GROUPING, + PVR_FW_HAS_FEATURE_COMMON_STORE_SIZE_IN_DWORDS, + PVR_FW_HAS_FEATURE_COMPUTE, + PVR_FW_HAS_FEATURE_COMPUTE_MORTON_CAPABLE, + PVR_FW_HAS_FEATURE_COMPUTE_OVERLAP, + PVR_FW_HAS_FEATURE_COREID_PER_OS, + PVR_FW_HAS_FEATURE_DYNAMIC_DUST_POWER, + PVR_FW_HAS_FEATURE_ECC_RAMS, + PVR_FW_HAS_FEATURE_FBCDC, + PVR_FW_HAS_FEATURE_FBCDC_ALGORITHM, + PVR_FW_HAS_FEATURE_FBCDC_ARCHITECTURE, + PVR_FW_HAS_FEATURE_FBC_MAX_DEFAULT_DESCRIPTORS, + PVR_FW_HAS_FEATURE_FBC_MAX_LARGE_DESCRIPTORS, + PVR_FW_HAS_FEATURE_FB_CDC_V4, + PVR_FW_HAS_FEATURE_GPU_MULTICORE_SUPPORT, + PVR_FW_HAS_FEATURE_GPU_VIRTUALISATION, + PVR_FW_HAS_FEATURE_GS_RTA_SUPPORT, + PVR_FW_HAS_FEATURE_IRQ_PER_OS, + PVR_FW_HAS_FEATURE_ISP_MAX_TILES_IN_FLIGHT, + PVR_FW_HAS_FEATURE_ISP_SAMPLES_PER_PIXEL, + PVR_FW_HAS_FEATURE_ISP_ZLS_D24_S8_PACKING_OGL_MODE, + PVR_FW_HAS_FEATURE_LAYOUT_MARS, + PVR_FW_HAS_FEATURE_MAX_PARTITIONS, + PVR_FW_HAS_FEATURE_META, + PVR_FW_HAS_FEATURE_META_COREMEM_SIZE, + PVR_FW_HAS_FEATURE_MIPS, + PVR_FW_HAS_FEATURE_NUM_CLUSTERS, + PVR_FW_HAS_FEATURE_NUM_ISP_IPP_PIPES, + PVR_FW_HAS_FEATURE_NUM_OSIDS, + PVR_FW_HAS_FEATURE_NUM_RASTER_PIPES, + PVR_FW_HAS_FEATURE_PBE2_IN_XE, + PVR_FW_HAS_FEATURE_PBVNC_COREID_REG, + PVR_FW_HAS_FEATURE_PERFBUS, + PVR_FW_HAS_FEATURE_PERF_COUNTER_BATCH, + PVR_FW_HAS_FEATURE_PHYS_BUS_WIDTH, + PVR_FW_HAS_FEATURE_RISCV_FW_PROCESSOR, + PVR_FW_HAS_FEATURE_ROGUEXE, + PVR_FW_HAS_FEATURE_S7_TOP_INFRASTRUCTURE, + PVR_FW_HAS_FEATURE_SIMPLE_INTERNAL_PARAMETER_FORMAT, + PVR_FW_HAS_FEATURE_SIMPLE_INTERNAL_PARAMETER_FORMAT_V2, + PVR_FW_HAS_FEATURE_SIMPLE_PARAMETER_FORMAT_VERSION, + PVR_FW_HAS_FEATURE_SLC_BANKS, + PVR_FW_HAS_FEATURE_SLC_CACHE_LINE_SIZE_BITS, + PVR_FW_HAS_FEATURE_SLC_SIZE_CONFIGURABLE, + PVR_FW_HAS_FEATURE_SLC_SIZE_IN_KILOBYTES, + PVR_FW_HAS_FEATURE_SOC_TIMER, + PVR_FW_HAS_FEATURE_SYS_BUS_SECURE_RESET, + PVR_FW_HAS_FEATURE_TESSELLATION, + PVR_FW_HAS_FEATURE_TILE_REGION_PROTECTION, + PVR_FW_HAS_FEATURE_TILE_SIZE_X, + PVR_FW_HAS_FEATURE_TILE_SIZE_Y, + PVR_FW_HAS_FEATURE_TLA, + PVR_FW_HAS_FEATURE_TPU_CEM_DATAMASTER_GLOBAL_REGISTERS, + PVR_FW_HAS_FEATURE_TPU_DM_GLOBAL_REGISTERS, + PVR_FW_HAS_FEATURE_TPU_FILTERING_MODE_CONTROL, + PVR_FW_HAS_FEATURE_USC_MIN_OUTPUT_REGISTERS_PER_PIX, + PVR_FW_HAS_FEATURE_VDM_DRAWINDIRECT, + PVR_FW_HAS_FEATURE_VDM_OBJECT_LEVEL_LLS, + PVR_FW_HAS_FEATURE_VIRTUAL_ADDRESS_SPACE_BITS, + PVR_FW_HAS_FEATURE_WATCHDOG_TIMER, + PVR_FW_HAS_FEATURE_WORKGROUP_PROTECTION, + PVR_FW_HAS_FEATURE_XE_ARCHITECTURE, + PVR_FW_HAS_FEATURE_XE_MEMORY_HIERARCHY, + PVR_FW_HAS_FEATURE_XE_TPU2, + PVR_FW_HAS_FEATURE_XPU_MAX_REGBANKS_ADDR_WIDTH, + PVR_FW_HAS_FEATURE_XPU_MAX_SLAVES, + PVR_FW_HAS_FEATURE_XPU_REGISTER_BROADCAST, + PVR_FW_HAS_FEATURE_XT_TOP_INFRASTRUCTURE, + PVR_FW_HAS_FEATURE_ZLS_SUBTILE, + + PVR_FW_HAS_FEATURE_MAX +}; + +#endif /* __PVR_ROGUE_FWIF_DEV_INFO_H__ */ diff --git a/drivers/gpu/drm/imagination/pvr_rogue_fwif_resetframework.h b/drivers/gpu/drm/imagination/pvr_rogue_fwif_resetframework.h new file mode 100644 index 000000000000..1db1f4c532bc --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_fwif_resetframework.h @@ -0,0 +1,28 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_ROGUE_FWIF_RESETFRAMEWORK_H +#define PVR_ROGUE_FWIF_RESETFRAMEWORK_H + +#include +#include + +#include "pvr_rogue_fwif_shared.h" + +struct rogue_fwif_rf_registers { + union { + u64 cdmreg_cdm_cb_base; + u64 cdmreg_cdm_ctrl_stream_base; + }; + u64 cdmreg_cdm_cb_queue; + u64 cdmreg_cdm_cb; +}; + +struct rogue_fwif_rf_cmd { + /* THIS MUST BE THE LAST MEMBER OF THE CONTAINING STRUCTURE */ + struct rogue_fwif_rf_registers fw_registers __aligned(8); +}; + +#define ROGUE_FWIF_RF_CMD_SIZE sizeof(struct rogue_fwif_rf_cmd) + +#endif /* PVR_ROGUE_FWIF_RESETFRAMEWORK_H */ diff --git a/drivers/gpu/drm/imagination/pvr_rogue_fwif_shared.h b/drivers/gpu/drm/imagination/pvr_rogue_fwif_shared.h new file mode 100644 index 000000000000..6c09c15bf9bd --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_fwif_shared.h @@ -0,0 +1,258 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_ROGUE_FWIF_SHARED_H +#define PVR_ROGUE_FWIF_SHARED_H + +#include +#include + +#define ROGUE_FWIF_NUM_RTDATAS 2U +#define ROGUE_FWIF_NUM_GEOMDATAS 1U +#define ROGUE_FWIF_NUM_RTDATA_FREELISTS 2U +#define ROGUE_NUM_GEOM_CORES 1U + +#define ROGUE_NUM_GEOM_CORES_SIZE 2U + +/* + * Maximum number of UFOs in a CCB command. + * The number is based on having 32 sync prims (as originally), plus 32 sync + * checkpoints. + * Once the use of sync prims is no longer supported, we will retain + * the same total (64) as the number of sync checkpoints which may be + * supporting a fence is not visible to the client driver and has to + * allow for the number of different timelines involved in fence merges. + */ +#define ROGUE_FWIF_CCB_CMD_MAX_UFOS (32U + 32U) + +/* + * This is a generic limit imposed on any DM (GEOMETRY,FRAGMENT,CDM,TDM,2D,TRANSFER) + * command passed through the bridge. + * Just across the bridge in the server, any incoming kick command size is + * checked against this maximum limit. + * In case the incoming command size is larger than the specified limit, + * the bridge call is retired with error. + */ +#define ROGUE_FWIF_DM_INDEPENDENT_KICK_CMD_SIZE (1024U) + +#define ROGUE_FWIF_PRBUFFER_START (0) +#define ROGUE_FWIF_PRBUFFER_ZSBUFFER (0) +#define ROGUE_FWIF_PRBUFFER_MSAABUFFER (1) +#define ROGUE_FWIF_PRBUFFER_MAXSUPPORTED (2) + +struct rogue_fwif_dma_addr { + aligned_u64 dev_addr; + u32 fw_addr; + u32 padding; +} __aligned(8); + +struct rogue_fwif_ufo { + u32 addr; + u32 value; +}; + +#define ROGUE_FWIF_UFO_ADDR_IS_SYNC_CHECKPOINT (1) + +struct rogue_fwif_sync_checkpoint { + u32 state; + u32 fw_ref_count; +}; + +struct rogue_fwif_cleanup_ctl { + /* Number of commands received by the FW */ + u32 submitted_commands; + /* Number of commands executed by the FW */ + u32 executed_commands; +} __aligned(8); + +/* + * Used to share frame numbers across UM-KM-FW, + * frame number is set in UM, + * frame number is required in both KM for HTB and FW for FW trace. + * + * May be used to house Kick flags in the future. + */ +struct rogue_fwif_cmd_common { + /* associated frame number */ + u32 frame_num; +}; + +/* + * Geometry and fragment commands require set of firmware addresses that are stored in the Kernel. + * Client has handle(s) to Kernel containers storing these addresses, instead of raw addresses. We + * have to patch/write these addresses in KM to prevent UM from controlling FW addresses directly. + * Typedefs for geometry and fragment commands are shared between Client and Firmware (both + * single-BVNC). Kernel is implemented in a multi-BVNC manner, so it can't use geometry|fragment + * CMD type definitions directly. Therefore we have a SHARED block that is shared between UM-KM-FW + * across all BVNC configurations. + */ +struct rogue_fwif_cmd_geom_frag_shared { + /* Common command attributes */ + struct rogue_fwif_cmd_common cmn; + + /* + * RTData associated with this command, this is used for context + * selection and for storing out HW-context, when TA is switched out for + * continuing later + */ + u32 hwrt_data_fw_addr; + + /* Supported PR Buffers like Z/S/MSAA Scratch */ + u32 pr_buffer_fw_addr[ROGUE_FWIF_PRBUFFER_MAXSUPPORTED]; +}; + +/* + * Client Circular Command Buffer (CCCB) control structure. + * This is shared between the Server and the Firmware and holds byte offsets + * into the CCCB as well as the wrapping mask to aid wrap around. A given + * snapshot of this queue with Cmd 1 running on the GPU might be: + * + * Roff Doff Woff + * [..........|-1----------|=2===|=3===|=4===|~5~~~~|~6~~~~|~7~~~~|..........] + * < runnable commands >< !ready to run > + * + * Cmd 1 : Currently executing on the GPU data master. + * Cmd 2,3,4: Fence dependencies met, commands runnable. + * Cmd 5... : Fence dependency not met yet. + */ +struct rogue_fwif_cccb_ctl { + /* Host write offset into CCB. This must be aligned to 16 bytes. */ + u32 write_offset; + /* + * Firmware read offset into CCB. Points to the command that is runnable + * on GPU, if R!=W + */ + u32 read_offset; + /* + * Firmware fence dependency offset. Points to commands not ready, i.e. + * fence dependencies are not met. + */ + u32 dep_offset; + /* Offset wrapping mask, total capacity in bytes of the CCB-1 */ + u32 wrap_mask; + + /* Only used if SUPPORT_AGP is present. */ + u32 read_offset2; + + /* Only used if SUPPORT_AGP4 is present. */ + u32 read_offset3; + /* Only used if SUPPORT_AGP4 is present. */ + u32 read_offset4; + + u32 padding; +} __aligned(8); + +#define ROGUE_FW_LOCAL_FREELIST (0) +#define ROGUE_FW_GLOBAL_FREELIST (1) +#define ROGUE_FW_FREELIST_TYPE_LAST ROGUE_FW_GLOBAL_FREELIST +#define ROGUE_FW_MAX_FREELISTS (ROGUE_FW_FREELIST_TYPE_LAST + 1U) + +struct rogue_fwif_geom_registers_caswitch { + u64 geom_reg_vdm_context_state_base_addr; + u64 geom_reg_vdm_context_state_resume_addr; + u64 geom_reg_ta_context_state_base_addr; + + struct { + u64 geom_reg_vdm_context_store_task0; + u64 geom_reg_vdm_context_store_task1; + u64 geom_reg_vdm_context_store_task2; + + /* VDM resume state update controls */ + u64 geom_reg_vdm_context_resume_task0; + u64 geom_reg_vdm_context_resume_task1; + u64 geom_reg_vdm_context_resume_task2; + + u64 geom_reg_vdm_context_store_task3; + u64 geom_reg_vdm_context_store_task4; + + u64 geom_reg_vdm_context_resume_task3; + u64 geom_reg_vdm_context_resume_task4; + } geom_state[2]; +}; + +#define ROGUE_FWIF_GEOM_REGISTERS_CSWITCH_SIZE \ + sizeof(struct rogue_fwif_geom_registers_caswitch) + +struct rogue_fwif_cdm_registers_cswitch { + u64 cdmreg_cdm_context_pds0; + u64 cdmreg_cdm_context_pds1; + u64 cdmreg_cdm_terminate_pds; + u64 cdmreg_cdm_terminate_pds1; + + /* CDM resume controls */ + u64 cdmreg_cdm_resume_pds0; + u64 cdmreg_cdm_context_pds0_b; + u64 cdmreg_cdm_resume_pds0_b; +}; + +struct rogue_fwif_static_rendercontext_state { + /* Geom registers for ctx switch */ + struct rogue_fwif_geom_registers_caswitch ctxswitch_regs[ROGUE_NUM_GEOM_CORES_SIZE] + __aligned(8); +}; + +#define ROGUE_FWIF_STATIC_RENDERCONTEXT_SIZE \ + sizeof(struct rogue_fwif_static_rendercontext_state) + +struct rogue_fwif_static_computecontext_state { + /* CDM registers for ctx switch */ + struct rogue_fwif_cdm_registers_cswitch ctxswitch_regs __aligned(8); +}; + +#define ROGUE_FWIF_STATIC_COMPUTECONTEXT_SIZE \ + sizeof(struct rogue_fwif_static_computecontext_state) + +enum rogue_fwif_prbuffer_state { + ROGUE_FWIF_PRBUFFER_UNBACKED = 0, + ROGUE_FWIF_PRBUFFER_BACKED, + ROGUE_FWIF_PRBUFFER_BACKING_PENDING, + ROGUE_FWIF_PRBUFFER_UNBACKING_PENDING, +}; + +struct rogue_fwif_prbuffer { + /* Buffer ID*/ + u32 buffer_id; + /* Needs On-demand Z/S/MSAA Buffer allocation */ + bool on_demand __aligned(4); + /* Z/S/MSAA -Buffer state */ + enum rogue_fwif_prbuffer_state state; + /* Cleanup state */ + struct rogue_fwif_cleanup_ctl cleanup_sate; + /* Compatibility and other flags */ + u32 prbuffer_flags; +} __aligned(8); + +/* Last reset reason for a context. */ +enum rogue_context_reset_reason { + /* No reset reason recorded */ + ROGUE_CONTEXT_RESET_REASON_NONE = 0, + /* Caused a reset due to locking up */ + ROGUE_CONTEXT_RESET_REASON_GUILTY_LOCKUP = 1, + /* Affected by another context locking up */ + ROGUE_CONTEXT_RESET_REASON_INNOCENT_LOCKUP = 2, + /* Overran the global deadline */ + ROGUE_CONTEXT_RESET_REASON_GUILTY_OVERRUNING = 3, + /* Affected by another context overrunning */ + ROGUE_CONTEXT_RESET_REASON_INNOCENT_OVERRUNING = 4, + /* Forced reset to ensure scheduling requirements */ + ROGUE_CONTEXT_RESET_REASON_HARD_CONTEXT_SWITCH = 5, + /* FW Safety watchdog triggered */ + ROGUE_CONTEXT_RESET_REASON_FW_WATCHDOG = 12, + /* FW page fault (no HWR) */ + ROGUE_CONTEXT_RESET_REASON_FW_PAGEFAULT = 13, + /* FW execution error (GPU reset requested) */ + ROGUE_CONTEXT_RESET_REASON_FW_EXEC_ERR = 14, + /* Host watchdog detected FW error */ + ROGUE_CONTEXT_RESET_REASON_HOST_WDG_FW_ERR = 15, + /* Geometry DM OOM event is not allowed */ + ROGUE_CONTEXT_GEOM_OOM_DISABLED = 16, +}; + +struct rogue_context_reset_reason_data { + enum rogue_context_reset_reason reset_reason; + u32 reset_ext_job_ref; +}; + +#include "pvr_rogue_fwif_shared_check.h" + +#endif /* PVR_ROGUE_FWIF_SHARED_H */ diff --git a/drivers/gpu/drm/imagination/pvr_rogue_fwif_shared_check.h b/drivers/gpu/drm/imagination/pvr_rogue_fwif_shared_check.h new file mode 100644 index 000000000000..597ed54bbd3a --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_fwif_shared_check.h @@ -0,0 +1,108 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_ROGUE_FWIF_SHARED_CHECK_H +#define PVR_ROGUE_FWIF_SHARED_CHECK_H + +#include + +#define OFFSET_CHECK(type, member, offset) \ + static_assert(offsetof(type, member) == (offset), \ + "offsetof(" #type ", " #member ") incorrect") + +#define SIZE_CHECK(type, size) \ + static_assert(sizeof(type) == (size), #type " is incorrect size") + +OFFSET_CHECK(struct rogue_fwif_dma_addr, dev_addr, 0); +OFFSET_CHECK(struct rogue_fwif_dma_addr, fw_addr, 8); +SIZE_CHECK(struct rogue_fwif_dma_addr, 16); + +OFFSET_CHECK(struct rogue_fwif_ufo, addr, 0); +OFFSET_CHECK(struct rogue_fwif_ufo, value, 4); +SIZE_CHECK(struct rogue_fwif_ufo, 8); + +OFFSET_CHECK(struct rogue_fwif_cleanup_ctl, submitted_commands, 0); +OFFSET_CHECK(struct rogue_fwif_cleanup_ctl, executed_commands, 4); +SIZE_CHECK(struct rogue_fwif_cleanup_ctl, 8); + +OFFSET_CHECK(struct rogue_fwif_cccb_ctl, write_offset, 0); +OFFSET_CHECK(struct rogue_fwif_cccb_ctl, read_offset, 4); +OFFSET_CHECK(struct rogue_fwif_cccb_ctl, dep_offset, 8); +OFFSET_CHECK(struct rogue_fwif_cccb_ctl, wrap_mask, 12); +OFFSET_CHECK(struct rogue_fwif_cccb_ctl, read_offset2, 16); +OFFSET_CHECK(struct rogue_fwif_cccb_ctl, read_offset3, 20); +OFFSET_CHECK(struct rogue_fwif_cccb_ctl, read_offset4, 24); +SIZE_CHECK(struct rogue_fwif_cccb_ctl, 32); + +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_reg_vdm_context_state_base_addr, 0); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_reg_vdm_context_state_resume_addr, 8); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_reg_ta_context_state_base_addr, 16); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[0].geom_reg_vdm_context_store_task0, 24); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[0].geom_reg_vdm_context_store_task1, 32); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[0].geom_reg_vdm_context_store_task2, 40); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[0].geom_reg_vdm_context_resume_task0, 48); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[0].geom_reg_vdm_context_resume_task1, 56); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[0].geom_reg_vdm_context_resume_task2, 64); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[0].geom_reg_vdm_context_store_task3, 72); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[0].geom_reg_vdm_context_store_task4, 80); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[0].geom_reg_vdm_context_resume_task3, 88); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[0].geom_reg_vdm_context_resume_task4, 96); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[1].geom_reg_vdm_context_store_task0, 104); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[1].geom_reg_vdm_context_store_task1, 112); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[1].geom_reg_vdm_context_store_task2, 120); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[1].geom_reg_vdm_context_resume_task0, 128); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[1].geom_reg_vdm_context_resume_task1, 136); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[1].geom_reg_vdm_context_resume_task2, 144); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[1].geom_reg_vdm_context_store_task3, 152); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[1].geom_reg_vdm_context_store_task4, 160); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[1].geom_reg_vdm_context_resume_task3, 168); +OFFSET_CHECK(struct rogue_fwif_geom_registers_caswitch, + geom_state[1].geom_reg_vdm_context_resume_task4, 176); +SIZE_CHECK(struct rogue_fwif_geom_registers_caswitch, 184); + +OFFSET_CHECK(struct rogue_fwif_cdm_registers_cswitch, cdmreg_cdm_context_pds0, 0); +OFFSET_CHECK(struct rogue_fwif_cdm_registers_cswitch, cdmreg_cdm_context_pds1, 8); +OFFSET_CHECK(struct rogue_fwif_cdm_registers_cswitch, cdmreg_cdm_terminate_pds, 16); +OFFSET_CHECK(struct rogue_fwif_cdm_registers_cswitch, cdmreg_cdm_terminate_pds1, 24); +OFFSET_CHECK(struct rogue_fwif_cdm_registers_cswitch, cdmreg_cdm_resume_pds0, 32); +OFFSET_CHECK(struct rogue_fwif_cdm_registers_cswitch, cdmreg_cdm_context_pds0_b, 40); +OFFSET_CHECK(struct rogue_fwif_cdm_registers_cswitch, cdmreg_cdm_resume_pds0_b, 48); +SIZE_CHECK(struct rogue_fwif_cdm_registers_cswitch, 56); + +OFFSET_CHECK(struct rogue_fwif_static_rendercontext_state, ctxswitch_regs, 0); +SIZE_CHECK(struct rogue_fwif_static_rendercontext_state, 368); + +OFFSET_CHECK(struct rogue_fwif_static_computecontext_state, ctxswitch_regs, 0); +SIZE_CHECK(struct rogue_fwif_static_computecontext_state, 56); + +OFFSET_CHECK(struct rogue_fwif_cmd_common, frame_num, 0); +SIZE_CHECK(struct rogue_fwif_cmd_common, 4); + +OFFSET_CHECK(struct rogue_fwif_cmd_geom_frag_shared, cmn, 0); +OFFSET_CHECK(struct rogue_fwif_cmd_geom_frag_shared, hwrt_data_fw_addr, 4); +OFFSET_CHECK(struct rogue_fwif_cmd_geom_frag_shared, pr_buffer_fw_addr, 8); +SIZE_CHECK(struct rogue_fwif_cmd_geom_frag_shared, 16); + +#endif /* PVR_ROGUE_FWIF_SHARED_CHECK_H */ diff --git a/drivers/gpu/drm/imagination/pvr_rogue_fwif_stream.h b/drivers/gpu/drm/imagination/pvr_rogue_fwif_stream.h new file mode 100644 index 000000000000..1c2c4ebedc25 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_rogue_fwif_stream.h @@ -0,0 +1,78 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_ROGUE_FWIF_STREAM_H +#define PVR_ROGUE_FWIF_STREAM_H + +/** + * DOC: Streams + * + * Commands are submitted to the kernel driver in the form of streams. + * + * A command stream has the following layout : + * - A 64-bit header containing: + * * A u32 containing the length of the main stream inclusive of the length of the header. + * * A u32 for padding. + * - The main stream data. + * - The extension stream (optional), which is composed of: + * * One or more headers. + * * The extension stream data, corresponding to the extension headers. + * + * The main stream provides the base command data. This has a fixed layout based on the features + * supported by a given GPU. + * + * The extension stream provides the command parameters that are required for BRNs & ERNs for the + * current GPU. This stream is comprised of one or more headers, followed by data for each given + * BRN/ERN. + * + * Each header is a u32 containing a bitmask of quirks & enhancements in the extension stream, a + * "type" field determining the set of quirks & enhancements the bitmask represents, and a + * continuation bit determining whether any more headers are present. The headers are then followed + * by command data; this is specific to each quirk/enhancement. All unused / reserved bits in the + * header must be set to 0. + * + * All parameters and headers in the main and extension streams must be naturally aligned. + * + * If a parameter appears in both the main and extension streams, then the extension parameter is + * used. + */ + +/* + * Stream extension header definition + */ +#define PVR_STREAM_EXTHDR_TYPE_SHIFT 29U +#define PVR_STREAM_EXTHDR_TYPE_MASK (7U << PVR_STREAM_EXTHDR_TYPE_SHIFT) +#define PVR_STREAM_EXTHDR_TYPE_MAX 8U +#define PVR_STREAM_EXTHDR_CONTINUATION BIT(28U) + +#define PVR_STREAM_EXTHDR_DATA_MASK ~(PVR_STREAM_EXTHDR_TYPE_MASK | PVR_STREAM_EXTHDR_CONTINUATION) + +/* + * Stream extension header - Geometry 0 + */ +#define PVR_STREAM_EXTHDR_TYPE_GEOM0 0U + +#define PVR_STREAM_EXTHDR_GEOM0_BRN49927 BIT(0U) + +#define PVR_STREAM_EXTHDR_GEOM0_VALID PVR_STREAM_EXTHDR_GEOM0_BRN49927 + +/* + * Stream extension header - Fragment 0 + */ +#define PVR_STREAM_EXTHDR_TYPE_FRAG0 0U + +#define PVR_STREAM_EXTHDR_FRAG0_BRN47217 BIT(0U) +#define PVR_STREAM_EXTHDR_FRAG0_BRN49927 BIT(1U) + +#define PVR_STREAM_EXTHDR_FRAG0_VALID PVR_STREAM_EXTHDR_FRAG0_BRN49927 + +/* + * Stream extension header - Compute 0 + */ +#define PVR_STREAM_EXTHDR_TYPE_COMPUTE0 0U + +#define PVR_STREAM_EXTHDR_COMPUTE0_BRN49927 BIT(0U) + +#define PVR_STREAM_EXTHDR_COMPUTE0_VALID PVR_STREAM_EXTHDR_COMPUTE0_BRN49927 + +#endif /* PVR_ROGUE_FWIF_STREAM_H */ From patchwork Wed Nov 22 16:34:31 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Donald Robson X-Patchwork-Id: 746148 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=imgtec.com header.i=@imgtec.com header.b="xzaw7u3A"; dkim=pass (1024-bit key) header.d=IMGTecCRM.onmicrosoft.com header.i=@IMGTecCRM.onmicrosoft.com header.b="EpuP9uoQ" Received: from mx07-00376f01.pphosted.com (mx07-00376f01.pphosted.com [185.132.180.163]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 43A08D5C; Wed, 22 Nov 2023 08:35:59 -0800 (PST) Received: from pps.filterd (m0168889.ppops.net [127.0.0.1]) by mx07-00376f01.pphosted.com (8.17.1.24/8.17.1.24) with ESMTP id 3AMCaKJF020106; Wed, 22 Nov 2023 16:35:28 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=imgtec.com; h= from:to:cc:subject:date:message-id:in-reply-to:references :content-transfer-encoding:content-type:mime-version; s= dk201812; bh=KVS9RmGljGo6v4Gif8xEoY8nF8L5OG7mLMGh73yT2+k=; b=xza w7u3Aao3UumDitmtjOpi8je+imyWGWlhn76ApPC/h1MtTfCpFF3hjjxF1cwBm6PJ jpxxkNb7yFbWAZu35LDXfDSreIAhBHjBzlQ7ZJSBiNbTUGtvxk1rLqYT4y/v+qtR 6xa56jOBUojnJmxBLZsDCLy+IzP8ZBbfYtqcLE84tQIUwcyW1ESmV9KI0C/PwAQ3 9rqaYH03tXHuk//Ro+Op4WSo6XR/Mn0AntPmZiSJY/MoVIt2bIOLNRm+ZWFzQXkh EBFDjukrChkmkbcbgRfN6cSfPHvBdW8X2eevrVxqB7c4IzhUI+bik5z1HmMlQMA3 z87YOmHH0fGeCeanDEw== Received: from hhmail04.hh.imgtec.org ([217.156.249.195]) by mx07-00376f01.pphosted.com (PPS) with ESMTPS id 3ug99g9m28-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128 verify=NOT); Wed, 22 Nov 2023 16:35:27 +0000 (GMT) Received: from HHMAIL04.hh.imgtec.org (10.100.10.119) by HHMAIL04.hh.imgtec.org (10.100.10.119) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34; Wed, 22 Nov 2023 16:35:27 +0000 Received: from GBR01-LO4-obe.outbound.protection.outlook.com (104.47.85.105) by email.imgtec.com (10.100.10.121) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34 via Frontend Transport; Wed, 22 Nov 2023 16:35:26 +0000 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=QcKESi/wcrT4GkOZqSU91VcOfTCQcZCb7/xt4XsitdCqN9A+XnlOe2Y87zJ+8zxnVIIuFgf+i8d2wsQfYAbe+8CT9v4xrzrDc2jidGYGJRnGvdxntrPhD1C2As91y2WqJ5w3gCBmuzcEIAyeSYykPHBrvy686Z6gDK9MzopTnzVKx/UhM67myj/PdCkYnERp8UFJ8qUcwhehpuOqzPpH/eOTgoo4v65spu2bXy6ZuX1ki+18KUHPdehJhkrJSveceU6SvmkYpSKbZc655/nJ8EiuvuH41tWvh8KSx/y5oPwq6jcg/12QlUwav8ieW41Dob2De/FhelKWVqE2jrCVTg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=KVS9RmGljGo6v4Gif8xEoY8nF8L5OG7mLMGh73yT2+k=; b=V/vpTS0ydozmEglC6LosjpEA2ogRKe7rDXjZSCRHAn39M8d9diFd0YQBpa7k8RT/Ysy9gmWr4etQZeplwp/b8SieL4Q2zu+QnXa5PK6jewrLAELtL5fBKcoY43RaGAWBItGmIRvZsHyxk3Bd96Sov72+oofV8tuXfa937jAJKZ2HObeBGU8Qo+cQ7brwIa6R4Tze7aLQB4PiCd6zcgnTsc6hPHMFxMAcuFk19K8gxk2ISNCu79rdDfnnRtEIgbcwR7AnJWxmW9zNcwWPcBWd2ZWsx4i6oAfroKBK1Ez1QmQVHYsX2N9lOliAJQlj/L3DUHi6HiRIrW4atBKzJYUY4A== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=imgtec.com; dmarc=pass action=none header.from=imgtec.com; dkim=pass header.d=imgtec.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=IMGTecCRM.onmicrosoft.com; s=selector2-IMGTecCRM-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=KVS9RmGljGo6v4Gif8xEoY8nF8L5OG7mLMGh73yT2+k=; b=EpuP9uoQKeysvKhMul7gQPLwpc38oV95enKTlPz1L0TGKJLi7bFIeeoDtz3xvDXE+2Qu13Ruv4frSSTAw3PXbp/8bl7l7TOEY6mCZo9C0+sDIiPxJaGSnLZcHxBY4fFAnGKG6lMPzaat0BS3aZ6+r5sOKYj8hwCF6sBP123HVr0= Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) by LO6P265MB6459.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:2df::12) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.7025.18; Wed, 22 Nov 2023 16:35:18 +0000 Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34]) by CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34%3]) with mapi id 15.20.7025.017; Wed, 22 Nov 2023 16:35:18 +0000 From: Donald Robson To: CC: , , , , , , , , , , , , , , , , , , , , , , Sarah Walker Subject: [PATCH v9 10/20] drm/imagination: Add GPU ID parsing and firmware loading Date: Wed, 22 Nov 2023 16:34:31 +0000 Message-Id: <1ff76f7a5b45c742279c78910f8491b8a5e7f6e6.1700668843.git.donald.robson@imgtec.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: References: X-ClientProxiedBy: LO4P265CA0273.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:37a::10) To CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) Precedence: bulk X-Mailing-List: devicetree@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: CWLP265MB5770:EE_|LO6P265MB6459:EE_ X-MS-Office365-Filtering-Correlation-Id: 11cbd294-2389-4bfa-1bce-08dbeb79032e X-MS-Exchange-SenderADCheck: 1 X-MS-Exchange-AntiSpam-Relay: 0 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-Forefront-Antispam-Report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:; IPV:NLI; SFV:NSPM; H:CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM; PTR:; CAT:NONE; SFS:(13230031)(39850400004)(346002)(396003)(366004)(136003)(376002)(230922051799003)(64100799003)(186009)(451199024)(1800799012)(30864003)(36756003)(7416002)(41300700001)(5660300002)(2906002)(86362001)(38350700005)(26005)(6666004)(478600001)(6506007)(2616005)(6486002)(107886003)(6512007)(52116002)(8936002)(8676002)(4326008)(44832011)(66946007)(316002)(6916009)(66556008)(66476007)(83380400001)(38100700002)(559001)(579004); DIR:OUT; SFP:1102; X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: 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 X-MS-Exchange-CrossTenant-Network-Message-Id: 11cbd294-2389-4bfa-1bce-08dbeb79032e X-MS-Exchange-CrossTenant-AuthSource: CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 22 Nov 2023 16:35:18.1511 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 0d5fd8bb-e8c2-4e0a-8dd5-2c264f7140fe X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: CASSwvLvRIQVSIM+HQCpMKDc5NDx2cWy4FKsEpgxYuoM4AYKVItZheE4sTVNLGFrYXgoPFzFJPdue0Su1pEJOsoa6OYctSVHMgs+4V4lAVI= X-MS-Exchange-Transport-CrossTenantHeadersStamped: LO6P265MB6459 X-OriginatorOrg: imgtec.com X-EXCLAIMER-MD-CONFIG: 15a78312-3e47-46eb-9010-2e54d84a9631 X-Proofpoint-ORIG-GUID: 7fb2uokeA_vCpgjJaurk8Q6sYUM4RCsR X-Proofpoint-GUID: 7fb2uokeA_vCpgjJaurk8Q6sYUM4RCsR From: Sarah Walker Read the GPU ID register at probe time and select the correct features/quirks/enhancements. Use the GPU ID to form the firmware file name and load the firmware. The features/quirks/enhancements arrays are currently hardcoded in the driver for the supported GPUs. We are looking at moving this information to the firmware image. Changes since v8: - Corrected license identifiers Changes since v7: - Fix kerneldoc for pvr_device_info_set_enhancements() Changes since v5: - Add BRN 71242 to device info Changes since v4: - Retrieve device information from firmware header - Pull forward firmware header parsing from FW infrastructure patch - Use devm_add_action_or_reset to release firmware Changes since v3: - Use drm_dev_{enter,exit} Co-developed-by: Frank Binns Signed-off-by: Frank Binns Co-developed-by: Matt Coster Signed-off-by: Matt Coster Co-developed-by: Donald Robson Signed-off-by: Donald Robson Signed-off-by: Sarah Walker --- drivers/gpu/drm/imagination/Makefile | 2 + drivers/gpu/drm/imagination/pvr_device.c | 323 ++++++++++- drivers/gpu/drm/imagination/pvr_device.h | 220 ++++++++ drivers/gpu/drm/imagination/pvr_device_info.c | 254 +++++++++ drivers/gpu/drm/imagination/pvr_device_info.h | 186 +++++++ drivers/gpu/drm/imagination/pvr_drv.c | 521 +++++++++++++++++- drivers/gpu/drm/imagination/pvr_drv.h | 107 ++++ drivers/gpu/drm/imagination/pvr_fw.c | 145 +++++ drivers/gpu/drm/imagination/pvr_fw.h | 34 ++ drivers/gpu/drm/imagination/pvr_fw_info.h | 135 +++++ 10 files changed, 1925 insertions(+), 2 deletions(-) create mode 100644 drivers/gpu/drm/imagination/pvr_device_info.c create mode 100644 drivers/gpu/drm/imagination/pvr_device_info.h create mode 100644 drivers/gpu/drm/imagination/pvr_fw.c create mode 100644 drivers/gpu/drm/imagination/pvr_fw.h create mode 100644 drivers/gpu/drm/imagination/pvr_fw_info.h diff --git a/drivers/gpu/drm/imagination/Makefile b/drivers/gpu/drm/imagination/Makefile index d36007f2825c..0b863e9f51b8 100644 --- a/drivers/gpu/drm/imagination/Makefile +++ b/drivers/gpu/drm/imagination/Makefile @@ -5,6 +5,8 @@ subdir-ccflags-y := -I$(srctree)/$(src) powervr-y := \ pvr_device.o \ + pvr_device_info.o \ pvr_drv.o \ + pvr_fw.o obj-$(CONFIG_DRM_POWERVR) += powervr.o diff --git a/drivers/gpu/drm/imagination/pvr_device.c b/drivers/gpu/drm/imagination/pvr_device.c index abcdf733f57b..05e382cacb27 100644 --- a/drivers/gpu/drm/imagination/pvr_device.c +++ b/drivers/gpu/drm/imagination/pvr_device.c @@ -2,19 +2,31 @@ /* Copyright (c) 2023 Imagination Technologies Ltd. */ #include "pvr_device.h" +#include "pvr_device_info.h" + +#include "pvr_fw.h" +#include "pvr_rogue_cr_defs.h" #include +#include #include #include #include #include #include +#include #include +#include #include +#include #include #include #include +#include + +/* Major number for the supported version of the firmware. */ +#define PVR_FW_VERSION_MAJOR 1 /** * pvr_device_reg_init() - Initialize kernel access to a PowerVR device's @@ -100,6 +112,209 @@ static int pvr_device_clk_init(struct pvr_device *pvr_dev) return 0; } +/** + * pvr_build_firmware_filename() - Construct a PowerVR firmware filename + * @pvr_dev: Target PowerVR device. + * @base: First part of the filename. + * @major: Major version number. + * + * A PowerVR firmware filename consists of three parts separated by underscores + * (``'_'``) along with a '.fw' file suffix. The first part is the exact value + * of @base, the second part is the hardware version string derived from @pvr_fw + * and the final part is the firmware version number constructed from @major with + * a 'v' prefix, e.g. powervr/rogue_4.40.2.51_v1.fw. + * + * The returned string will have been slab allocated and must be freed with + * kfree(). + * + * Return: + * * The constructed filename on success, or + * * Any error returned by kasprintf(). + */ +static char * +pvr_build_firmware_filename(struct pvr_device *pvr_dev, const char *base, + u8 major) +{ + struct pvr_gpu_id *gpu_id = &pvr_dev->gpu_id; + + return kasprintf(GFP_KERNEL, "%s_%d.%d.%d.%d_v%d.fw", base, gpu_id->b, + gpu_id->v, gpu_id->n, gpu_id->c, major); +} + +static void +pvr_release_firmware(void *data) +{ + struct pvr_device *pvr_dev = data; + + release_firmware(pvr_dev->fw_dev.firmware); +} + +/** + * pvr_request_firmware() - Load firmware for a PowerVR device + * @pvr_dev: Target PowerVR device. + * + * See pvr_build_firmware_filename() for details on firmware file naming. + * + * Return: + * * 0 on success, + * * Any error returned by pvr_build_firmware_filename(), or + * * Any error returned by request_firmware(). + */ +static int +pvr_request_firmware(struct pvr_device *pvr_dev) +{ + struct drm_device *drm_dev = &pvr_dev->base; + char *filename; + const struct firmware *fw; + int err; + + filename = pvr_build_firmware_filename(pvr_dev, "powervr/rogue", + PVR_FW_VERSION_MAJOR); + if (IS_ERR(filename)) + return PTR_ERR(filename); + + /* + * This function takes a copy of &filename, meaning we can free our + * instance before returning. + */ + err = request_firmware(&fw, filename, pvr_dev->base.dev); + if (err) { + drm_err(drm_dev, "failed to load firmware %s (err=%d)\n", + filename, err); + goto err_free_filename; + } + + drm_info(drm_dev, "loaded firmware %s\n", filename); + kfree(filename); + + pvr_dev->fw_dev.firmware = fw; + + return devm_add_action_or_reset(drm_dev->dev, pvr_release_firmware, pvr_dev); + +err_free_filename: + kfree(filename); + + return err; +} + +/** + * pvr_load_gpu_id() - Load a PowerVR device's GPU ID (BVNC) from control registers. + * + * Sets struct pvr_dev.gpu_id. + * + * @pvr_dev: Target PowerVR device. + */ +static void +pvr_load_gpu_id(struct pvr_device *pvr_dev) +{ + struct pvr_gpu_id *gpu_id = &pvr_dev->gpu_id; + u64 bvnc; + + /* + * Try reading the BVNC using the newer (cleaner) method first. If the + * B value is zero, fall back to the older method. + */ + bvnc = pvr_cr_read64(pvr_dev, ROGUE_CR_CORE_ID__PBVNC); + + gpu_id->b = PVR_CR_FIELD_GET(bvnc, CORE_ID__PBVNC__BRANCH_ID); + if (gpu_id->b != 0) { + gpu_id->v = PVR_CR_FIELD_GET(bvnc, CORE_ID__PBVNC__VERSION_ID); + gpu_id->n = PVR_CR_FIELD_GET(bvnc, CORE_ID__PBVNC__NUMBER_OF_SCALABLE_UNITS); + gpu_id->c = PVR_CR_FIELD_GET(bvnc, CORE_ID__PBVNC__CONFIG_ID); + } else { + u32 core_rev = pvr_cr_read32(pvr_dev, ROGUE_CR_CORE_REVISION); + u32 core_id = pvr_cr_read32(pvr_dev, ROGUE_CR_CORE_ID); + u16 core_id_config = PVR_CR_FIELD_GET(core_id, CORE_ID_CONFIG); + + gpu_id->b = PVR_CR_FIELD_GET(core_rev, CORE_REVISION_MAJOR); + gpu_id->v = PVR_CR_FIELD_GET(core_rev, CORE_REVISION_MINOR); + gpu_id->n = FIELD_GET(0xFF00, core_id_config); + gpu_id->c = FIELD_GET(0x00FF, core_id_config); + } +} + +/** + * pvr_set_dma_info() - Set PowerVR device DMA information + * @pvr_dev: Target PowerVR device. + * + * Sets the DMA mask and max segment size for the PowerVR device. + * + * Return: + * * 0 on success, + * * Any error returned by PVR_FEATURE_VALUE(), or + * * Any error returned by dma_set_mask(). + */ + +static int +pvr_set_dma_info(struct pvr_device *pvr_dev) +{ + struct drm_device *drm_dev = from_pvr_device(pvr_dev); + u16 phys_bus_width; + int err; + + err = PVR_FEATURE_VALUE(pvr_dev, phys_bus_width, &phys_bus_width); + if (err) { + drm_err(drm_dev, "Failed to get device physical bus width\n"); + return err; + } + + err = dma_set_mask(drm_dev->dev, DMA_BIT_MASK(phys_bus_width)); + if (err) { + drm_err(drm_dev, "Failed to set DMA mask (err=%d)\n", err); + return err; + } + + dma_set_max_seg_size(drm_dev->dev, UINT_MAX); + + return 0; +} + +/** + * pvr_device_gpu_init() - GPU-specific initialization for a PowerVR device + * @pvr_dev: Target PowerVR device. + * + * The following steps are taken to ensure the device is ready: + * + * 1. Read the hardware version information from control registers, + * 2. Initialise the hardware feature information, + * 3. Setup the device DMA information, + * 4. Setup the device-scoped memory context, and + * 5. Load firmware into the device. + * + * Return: + * * 0 on success, + * * -%ENODEV if the GPU is not supported, + * * Any error returned by pvr_set_dma_info(), + * * Any error returned by pvr_memory_context_init(), or + * * Any error returned by pvr_request_firmware(). + */ +static int +pvr_device_gpu_init(struct pvr_device *pvr_dev) +{ + int err; + + pvr_load_gpu_id(pvr_dev); + + err = pvr_request_firmware(pvr_dev); + if (err) + return err; + + err = pvr_fw_validate_init_device_info(pvr_dev); + if (err) + return err; + + if (PVR_HAS_FEATURE(pvr_dev, meta)) + pvr_dev->fw_dev.processor_type = PVR_FW_PROCESSOR_TYPE_META; + else if (PVR_HAS_FEATURE(pvr_dev, mips)) + pvr_dev->fw_dev.processor_type = PVR_FW_PROCESSOR_TYPE_MIPS; + else if (PVR_HAS_FEATURE(pvr_dev, riscv_fw_processor)) + pvr_dev->fw_dev.processor_type = PVR_FW_PROCESSOR_TYPE_RISCV; + else + return -EINVAL; + + return pvr_set_dma_info(pvr_dev); +} + /** * pvr_device_init() - Initialize a PowerVR device * @pvr_dev: Target PowerVR device. @@ -130,7 +345,12 @@ pvr_device_init(struct pvr_device *pvr_dev) return err; /* Map the control registers into memory. */ - return pvr_device_reg_init(pvr_dev); + err = pvr_device_reg_init(pvr_dev); + if (err) + return err; + + /* Perform GPU-specific initialization steps. */ + return pvr_device_gpu_init(pvr_dev); } /** @@ -145,3 +365,104 @@ pvr_device_fini(struct pvr_device *pvr_dev) * the initialization stages in pvr_device_init(). */ } + +bool +pvr_device_has_uapi_quirk(struct pvr_device *pvr_dev, u32 quirk) +{ + switch (quirk) { + case 47217: + return PVR_HAS_QUIRK(pvr_dev, 47217); + case 48545: + return PVR_HAS_QUIRK(pvr_dev, 48545); + case 49927: + return PVR_HAS_QUIRK(pvr_dev, 49927); + case 51764: + return PVR_HAS_QUIRK(pvr_dev, 51764); + case 62269: + return PVR_HAS_QUIRK(pvr_dev, 62269); + default: + return false; + }; +} + +bool +pvr_device_has_uapi_enhancement(struct pvr_device *pvr_dev, u32 enhancement) +{ + switch (enhancement) { + case 35421: + return PVR_HAS_ENHANCEMENT(pvr_dev, 35421); + case 42064: + return PVR_HAS_ENHANCEMENT(pvr_dev, 42064); + default: + return false; + }; +} + +/** + * pvr_device_has_feature() - Look up device feature based on feature definition + * @pvr_dev: Device pointer. + * @feature: Feature to look up. Should be one of %PVR_FEATURE_*. + * + * Returns: + * * %true if feature is present on device, or + * * %false if feature is not present on device. + */ +bool +pvr_device_has_feature(struct pvr_device *pvr_dev, u32 feature) +{ + switch (feature) { + case PVR_FEATURE_CLUSTER_GROUPING: + return PVR_HAS_FEATURE(pvr_dev, cluster_grouping); + + case PVR_FEATURE_COMPUTE_MORTON_CAPABLE: + return PVR_HAS_FEATURE(pvr_dev, compute_morton_capable); + + case PVR_FEATURE_FB_CDC_V4: + return PVR_HAS_FEATURE(pvr_dev, fb_cdc_v4); + + case PVR_FEATURE_GPU_MULTICORE_SUPPORT: + return PVR_HAS_FEATURE(pvr_dev, gpu_multicore_support); + + case PVR_FEATURE_ISP_ZLS_D24_S8_PACKING_OGL_MODE: + return PVR_HAS_FEATURE(pvr_dev, isp_zls_d24_s8_packing_ogl_mode); + + case PVR_FEATURE_S7_TOP_INFRASTRUCTURE: + return PVR_HAS_FEATURE(pvr_dev, s7_top_infrastructure); + + case PVR_FEATURE_TESSELLATION: + return PVR_HAS_FEATURE(pvr_dev, tessellation); + + case PVR_FEATURE_TPU_DM_GLOBAL_REGISTERS: + return PVR_HAS_FEATURE(pvr_dev, tpu_dm_global_registers); + + case PVR_FEATURE_VDM_DRAWINDIRECT: + return PVR_HAS_FEATURE(pvr_dev, vdm_drawindirect); + + case PVR_FEATURE_VDM_OBJECT_LEVEL_LLS: + return PVR_HAS_FEATURE(pvr_dev, vdm_object_level_lls); + + case PVR_FEATURE_ZLS_SUBTILE: + return PVR_HAS_FEATURE(pvr_dev, zls_subtile); + + /* Derived features. */ + case PVR_FEATURE_CDM_USER_MODE_QUEUE: { + u8 cdm_control_stream_format = 0; + + PVR_FEATURE_VALUE(pvr_dev, cdm_control_stream_format, &cdm_control_stream_format); + return (cdm_control_stream_format >= 2 && cdm_control_stream_format <= 4); + } + + case PVR_FEATURE_REQUIRES_FB_CDC_ZLS_SETUP: + if (PVR_HAS_FEATURE(pvr_dev, fbcdc_algorithm)) { + u8 fbcdc_algorithm = 0; + + PVR_FEATURE_VALUE(pvr_dev, fbcdc_algorithm, &fbcdc_algorithm); + return (fbcdc_algorithm < 3 || PVR_HAS_FEATURE(pvr_dev, fb_cdc_v4)); + } + return false; + + default: + WARN(true, "Looking up undefined feature %u\n", feature); + return false; + } +} diff --git a/drivers/gpu/drm/imagination/pvr_device.h b/drivers/gpu/drm/imagination/pvr_device.h index d9d2804f4454..36fe67bed27f 100644 --- a/drivers/gpu/drm/imagination/pvr_device.h +++ b/drivers/gpu/drm/imagination/pvr_device.h @@ -4,6 +4,9 @@ #ifndef PVR_DEVICE_H #define PVR_DEVICE_H +#include "pvr_device_info.h" +#include "pvr_fw.h" + #include #include #include @@ -28,6 +31,26 @@ struct clk; /* Forward declaration from . */ struct firmware; +/** + * struct pvr_gpu_id - Hardware GPU ID information for a PowerVR device + * @b: Branch ID. + * @v: Version ID. + * @n: Number of scalable units. + * @c: Config ID. + */ +struct pvr_gpu_id { + u16 b, v, n, c; +}; + +/** + * struct pvr_fw_version - Firmware version information + * @major: Major version number. + * @minor: Minor version number. + */ +struct pvr_fw_version { + u16 major, minor; +}; + /** * struct pvr_device - powervr-specific wrapper for &struct drm_device */ @@ -40,6 +63,35 @@ struct pvr_device { */ struct drm_device base; + /** @gpu_id: GPU ID detected at runtime. */ + struct pvr_gpu_id gpu_id; + + /** + * @features: Hardware feature information. + * + * Do not access this member directly, instead use PVR_HAS_FEATURE() + * or PVR_FEATURE_VALUE() macros. + */ + struct pvr_device_features features; + + /** + * @quirks: Hardware quirk information. + * + * Do not access this member directly, instead use PVR_HAS_QUIRK(). + */ + struct pvr_device_quirks quirks; + + /** + * @enhancements: Hardware enhancement information. + * + * Do not access this member directly, instead use + * PVR_HAS_ENHANCEMENT(). + */ + struct pvr_device_enhancements enhancements; + + /** @fw_version: Firmware version detected at runtime. */ + struct pvr_fw_version fw_version; + /** * @regs: Device control registers. * @@ -70,6 +122,9 @@ struct pvr_device { * Interface (MEMIF). If present, this needs to be enabled/disabled together with @core_clk. */ struct clk *mem_clk; + + /** @fw_dev: Firmware related data. */ + struct pvr_fw_device fw_dev; }; /** @@ -92,6 +147,76 @@ struct pvr_file { struct pvr_device *pvr_dev; }; +/** + * PVR_HAS_FEATURE() - Tests whether a PowerVR device has a given feature + * @pvr_dev: [IN] Target PowerVR device. + * @feature: [IN] Hardware feature name. + * + * Feature names are derived from those found in &struct pvr_device_features by + * dropping the 'has_' prefix, which is applied by this macro. + * + * Return: + * * true if the named feature is present in the hardware + * * false if the named feature is not present in the hardware + */ +#define PVR_HAS_FEATURE(pvr_dev, feature) ((pvr_dev)->features.has_##feature) + +/** + * PVR_FEATURE_VALUE() - Gets a PowerVR device feature value + * @pvr_dev: [IN] Target PowerVR device. + * @feature: [IN] Feature name. + * @value_out: [OUT] Feature value. + * + * This macro will get a feature value for those features that have values. + * If the feature is not present, nothing will be stored to @value_out. + * + * Feature names are derived from those found in &struct pvr_device_features by + * dropping the 'has_' prefix. + * + * Return: + * * 0 on success, or + * * -%EINVAL if the named feature is not present in the hardware + */ +#define PVR_FEATURE_VALUE(pvr_dev, feature, value_out) \ + ({ \ + struct pvr_device *_pvr_dev = pvr_dev; \ + int _ret = -EINVAL; \ + if (_pvr_dev->features.has_##feature) { \ + *(value_out) = _pvr_dev->features.feature; \ + _ret = 0; \ + } \ + _ret; \ + }) + +/** + * PVR_HAS_QUIRK() - Tests whether a physical device has a given quirk + * @pvr_dev: [IN] Target PowerVR device. + * @quirk: [IN] Hardware quirk name. + * + * Quirk numbers are derived from those found in #pvr_device_quirks by + * dropping the 'has_brn' prefix, which is applied by this macro. + * + * Returns + * * true if the quirk is present in the hardware, or + * * false if the quirk is not present in the hardware. + */ +#define PVR_HAS_QUIRK(pvr_dev, quirk) ((pvr_dev)->quirks.has_brn##quirk) + +/** + * PVR_HAS_ENHANCEMENT() - Tests whether a physical device has a given + * enhancement + * @pvr_dev: [IN] Target PowerVR device. + * @enhancement: [IN] Hardware enhancement name. + * + * Enhancement numbers are derived from those found in #pvr_device_enhancements + * by dropping the 'has_ern' prefix, which is applied by this macro. + * + * Returns + * * true if the enhancement is present in the hardware, or + * * false if the enhancement is not present in the hardware. + */ +#define PVR_HAS_ENHANCEMENT(pvr_dev, enhancement) ((pvr_dev)->enhancements.has_ern##enhancement) + #define from_pvr_device(pvr_dev) (&(pvr_dev)->base) #define to_pvr_device(drm_dev) container_of_const(drm_dev, struct pvr_device, base) @@ -100,9 +225,77 @@ struct pvr_file { #define to_pvr_file(file) ((file)->driver_priv) +/** + * PVR_PACKED_BVNC() - Packs B, V, N and C values into a 64-bit unsigned integer + * @b: Branch ID. + * @v: Version ID. + * @n: Number of scalable units. + * @c: Config ID. + * + * The packed layout is as follows: + * + * +--------+--------+--------+-------+ + * | 63..48 | 47..32 | 31..16 | 15..0 | + * +========+========+========+=======+ + * | B | V | N | C | + * +--------+--------+--------+-------+ + * + * pvr_gpu_id_to_packed_bvnc() should be used instead of this macro when a + * &struct pvr_gpu_id is available in order to ensure proper type checking. + * + * Return: Packed BVNC. + */ +/* clang-format off */ +#define PVR_PACKED_BVNC(b, v, n, c) \ + ((((u64)(b) & GENMASK_ULL(15, 0)) << 48) | \ + (((u64)(v) & GENMASK_ULL(15, 0)) << 32) | \ + (((u64)(n) & GENMASK_ULL(15, 0)) << 16) | \ + (((u64)(c) & GENMASK_ULL(15, 0)) << 0)) +/* clang-format on */ + +/** + * pvr_gpu_id_to_packed_bvnc() - Packs B, V, N and C values into a 64-bit + * unsigned integer + * @gpu_id: GPU ID. + * + * The packed layout is as follows: + * + * +--------+--------+--------+-------+ + * | 63..48 | 47..32 | 31..16 | 15..0 | + * +========+========+========+=======+ + * | B | V | N | C | + * +--------+--------+--------+-------+ + * + * This should be used in preference to PVR_PACKED_BVNC() when a &struct + * pvr_gpu_id is available in order to ensure proper type checking. + * + * Return: Packed BVNC. + */ +static __always_inline u64 +pvr_gpu_id_to_packed_bvnc(struct pvr_gpu_id *gpu_id) +{ + return PVR_PACKED_BVNC(gpu_id->b, gpu_id->v, gpu_id->n, gpu_id->c); +} + +static __always_inline void +packed_bvnc_to_pvr_gpu_id(u64 bvnc, struct pvr_gpu_id *gpu_id) +{ + gpu_id->b = (bvnc & GENMASK_ULL(63, 48)) >> 48; + gpu_id->v = (bvnc & GENMASK_ULL(47, 32)) >> 32; + gpu_id->n = (bvnc & GENMASK_ULL(31, 16)) >> 16; + gpu_id->c = bvnc & GENMASK_ULL(15, 0); +} + int pvr_device_init(struct pvr_device *pvr_dev); void pvr_device_fini(struct pvr_device *pvr_dev); +bool +pvr_device_has_uapi_quirk(struct pvr_device *pvr_dev, u32 quirk); +bool +pvr_device_has_uapi_enhancement(struct pvr_device *pvr_dev, u32 enhancement); +bool +pvr_device_has_feature(struct pvr_device *pvr_dev, u32 feature); + /** * PVR_CR_FIELD_GET() - Extract a single field from a PowerVR control register * @val: Value of the target register. @@ -208,6 +401,29 @@ pvr_cr_poll_reg64(struct pvr_device *pvr_dev, u32 reg_addr, u64 reg_value, (value & reg_mask) == reg_value, 0, timeout_usec); } +/** + * pvr_round_up_to_cacheline_size() - Round up a provided size to be cacheline + * aligned + * @pvr_dev: Target PowerVR device. + * @size: Initial size, in bytes. + * + * Returns: + * * Size aligned to cacheline size. + */ +static __always_inline size_t +pvr_round_up_to_cacheline_size(struct pvr_device *pvr_dev, size_t size) +{ + u16 slc_cacheline_size_bits = 0; + u16 slc_cacheline_size_bytes; + + WARN_ON(!PVR_HAS_FEATURE(pvr_dev, slc_cache_line_size_bits)); + PVR_FEATURE_VALUE(pvr_dev, slc_cache_line_size_bits, + &slc_cacheline_size_bits); + slc_cacheline_size_bytes = slc_cacheline_size_bits / 8; + + return round_up(size, slc_cacheline_size_bytes); +} + /** * DOC: IOCTL validation helpers * @@ -302,4 +518,8 @@ pvr_ioctl_union_padding_check(void *instance, size_t union_offset, __union_size, __member_size); \ }) +#define PVR_FW_PROCESSOR_TYPE_META 0 +#define PVR_FW_PROCESSOR_TYPE_MIPS 1 +#define PVR_FW_PROCESSOR_TYPE_RISCV 2 + #endif /* PVR_DEVICE_H */ diff --git a/drivers/gpu/drm/imagination/pvr_device_info.c b/drivers/gpu/drm/imagination/pvr_device_info.c new file mode 100644 index 000000000000..11e6bef52ecd --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_device_info.c @@ -0,0 +1,254 @@ +// SPDX-License-Identifier: GPL-2.0-only OR MIT +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#include "pvr_device.h" +#include "pvr_device_info.h" +#include "pvr_rogue_fwif_dev_info.h" + +#include + +#include +#include +#include +#include + +#define QUIRK_MAPPING(quirk) \ + [PVR_FW_HAS_BRN_##quirk] = offsetof(struct pvr_device, quirks.has_brn##quirk) + +static const uintptr_t quirks_mapping[] = { + QUIRK_MAPPING(44079), + QUIRK_MAPPING(47217), + QUIRK_MAPPING(48492), + QUIRK_MAPPING(48545), + QUIRK_MAPPING(49927), + QUIRK_MAPPING(50767), + QUIRK_MAPPING(51764), + QUIRK_MAPPING(62269), + QUIRK_MAPPING(63142), + QUIRK_MAPPING(63553), + QUIRK_MAPPING(66011), + QUIRK_MAPPING(71242), +}; + +#undef QUIRK_MAPPING + +#define ENHANCEMENT_MAPPING(enhancement) \ + [PVR_FW_HAS_ERN_##enhancement] = offsetof(struct pvr_device, \ + enhancements.has_ern##enhancement) + +static const uintptr_t enhancements_mapping[] = { + ENHANCEMENT_MAPPING(35421), + ENHANCEMENT_MAPPING(38020), + ENHANCEMENT_MAPPING(38748), + ENHANCEMENT_MAPPING(42064), + ENHANCEMENT_MAPPING(42290), + ENHANCEMENT_MAPPING(42606), + ENHANCEMENT_MAPPING(47025), + ENHANCEMENT_MAPPING(57596), +}; + +#undef ENHANCEMENT_MAPPING + +static void pvr_device_info_set_common(struct pvr_device *pvr_dev, const u64 *bitmask, + u32 bitmask_size, const uintptr_t *mapping, u32 mapping_max) +{ + const u32 mapping_max_size = (mapping_max + 63) >> 6; + const u32 nr_bits = min(bitmask_size * 64, mapping_max); + + /* Warn if any unsupported values in the bitmask. */ + if (bitmask_size > mapping_max_size) { + if (mapping == quirks_mapping) + drm_warn(from_pvr_device(pvr_dev), "Unsupported quirks in firmware image"); + else + drm_warn(from_pvr_device(pvr_dev), + "Unsupported enhancements in firmware image"); + } else if (bitmask_size == mapping_max_size && (mapping_max & 63)) { + u64 invalid_mask = ~0ull << (mapping_max & 63); + + if (bitmask[bitmask_size - 1] & invalid_mask) { + if (mapping == quirks_mapping) + drm_warn(from_pvr_device(pvr_dev), + "Unsupported quirks in firmware image"); + else + drm_warn(from_pvr_device(pvr_dev), + "Unsupported enhancements in firmware image"); + } + } + + for (u32 i = 0; i < nr_bits; i++) { + if (bitmask[i >> 6] & BIT_ULL(i & 63)) + *(bool *)((u8 *)pvr_dev + mapping[i]) = true; + } +} + +/** + * pvr_device_info_set_quirks() - Set device quirks from device information in firmware + * @pvr_dev: Device pointer. + * @quirks: Pointer to quirks mask in device information. + * @quirks_size: Size of quirks mask, in u64s. + */ +void pvr_device_info_set_quirks(struct pvr_device *pvr_dev, const u64 *quirks, u32 quirks_size) +{ + BUILD_BUG_ON(ARRAY_SIZE(quirks_mapping) != PVR_FW_HAS_BRN_MAX); + + pvr_device_info_set_common(pvr_dev, quirks, quirks_size, quirks_mapping, + ARRAY_SIZE(quirks_mapping)); +} + +/** + * pvr_device_info_set_enhancements() - Set device enhancements from device information in firmware + * @pvr_dev: Device pointer. + * @enhancements: Pointer to enhancements mask in device information. + * @enhancements_size: Size of enhancements mask, in u64s. + */ +void pvr_device_info_set_enhancements(struct pvr_device *pvr_dev, const u64 *enhancements, + u32 enhancements_size) +{ + BUILD_BUG_ON(ARRAY_SIZE(enhancements_mapping) != PVR_FW_HAS_ERN_MAX); + + pvr_device_info_set_common(pvr_dev, enhancements, enhancements_size, + enhancements_mapping, ARRAY_SIZE(enhancements_mapping)); +} + +#define FEATURE_MAPPING(fw_feature, feature) \ + [PVR_FW_HAS_FEATURE_##fw_feature] = { \ + .flag_offset = offsetof(struct pvr_device, features.has_##feature), \ + .value_offset = 0 \ + } + +#define FEATURE_MAPPING_VALUE(fw_feature, feature) \ + [PVR_FW_HAS_FEATURE_##fw_feature] = { \ + .flag_offset = offsetof(struct pvr_device, features.has_##feature), \ + .value_offset = offsetof(struct pvr_device, features.feature) \ + } + +static const struct { + uintptr_t flag_offset; + uintptr_t value_offset; +} features_mapping[] = { + FEATURE_MAPPING(AXI_ACELITE, axi_acelite), + FEATURE_MAPPING_VALUE(CDM_CONTROL_STREAM_FORMAT, cdm_control_stream_format), + FEATURE_MAPPING(CLUSTER_GROUPING, cluster_grouping), + FEATURE_MAPPING_VALUE(COMMON_STORE_SIZE_IN_DWORDS, common_store_size_in_dwords), + FEATURE_MAPPING(COMPUTE, compute), + FEATURE_MAPPING(COMPUTE_MORTON_CAPABLE, compute_morton_capable), + FEATURE_MAPPING(COMPUTE_OVERLAP, compute_overlap), + FEATURE_MAPPING(COREID_PER_OS, coreid_per_os), + FEATURE_MAPPING(DYNAMIC_DUST_POWER, dynamic_dust_power), + FEATURE_MAPPING_VALUE(ECC_RAMS, ecc_rams), + FEATURE_MAPPING_VALUE(FBCDC, fbcdc), + FEATURE_MAPPING_VALUE(FBCDC_ALGORITHM, fbcdc_algorithm), + FEATURE_MAPPING_VALUE(FBCDC_ARCHITECTURE, fbcdc_architecture), + FEATURE_MAPPING_VALUE(FBC_MAX_DEFAULT_DESCRIPTORS, fbc_max_default_descriptors), + FEATURE_MAPPING_VALUE(FBC_MAX_LARGE_DESCRIPTORS, fbc_max_large_descriptors), + FEATURE_MAPPING(FB_CDC_V4, fb_cdc_v4), + FEATURE_MAPPING(GPU_MULTICORE_SUPPORT, gpu_multicore_support), + FEATURE_MAPPING(GPU_VIRTUALISATION, gpu_virtualisation), + FEATURE_MAPPING(GS_RTA_SUPPORT, gs_rta_support), + FEATURE_MAPPING(IRQ_PER_OS, irq_per_os), + FEATURE_MAPPING_VALUE(ISP_MAX_TILES_IN_FLIGHT, isp_max_tiles_in_flight), + FEATURE_MAPPING_VALUE(ISP_SAMPLES_PER_PIXEL, isp_samples_per_pixel), + FEATURE_MAPPING(ISP_ZLS_D24_S8_PACKING_OGL_MODE, isp_zls_d24_s8_packing_ogl_mode), + FEATURE_MAPPING_VALUE(LAYOUT_MARS, layout_mars), + FEATURE_MAPPING_VALUE(MAX_PARTITIONS, max_partitions), + FEATURE_MAPPING_VALUE(META, meta), + FEATURE_MAPPING_VALUE(META_COREMEM_SIZE, meta_coremem_size), + FEATURE_MAPPING(MIPS, mips), + FEATURE_MAPPING_VALUE(NUM_CLUSTERS, num_clusters), + FEATURE_MAPPING_VALUE(NUM_ISP_IPP_PIPES, num_isp_ipp_pipes), + FEATURE_MAPPING_VALUE(NUM_OSIDS, num_osids), + FEATURE_MAPPING_VALUE(NUM_RASTER_PIPES, num_raster_pipes), + FEATURE_MAPPING(PBE2_IN_XE, pbe2_in_xe), + FEATURE_MAPPING(PBVNC_COREID_REG, pbvnc_coreid_reg), + FEATURE_MAPPING(PERFBUS, perfbus), + FEATURE_MAPPING(PERF_COUNTER_BATCH, perf_counter_batch), + FEATURE_MAPPING_VALUE(PHYS_BUS_WIDTH, phys_bus_width), + FEATURE_MAPPING(RISCV_FW_PROCESSOR, riscv_fw_processor), + FEATURE_MAPPING(ROGUEXE, roguexe), + FEATURE_MAPPING(S7_TOP_INFRASTRUCTURE, s7_top_infrastructure), + FEATURE_MAPPING(SIMPLE_INTERNAL_PARAMETER_FORMAT, simple_internal_parameter_format), + FEATURE_MAPPING(SIMPLE_INTERNAL_PARAMETER_FORMAT_V2, simple_internal_parameter_format_v2), + FEATURE_MAPPING_VALUE(SIMPLE_PARAMETER_FORMAT_VERSION, simple_parameter_format_version), + FEATURE_MAPPING_VALUE(SLC_BANKS, slc_banks), + FEATURE_MAPPING_VALUE(SLC_CACHE_LINE_SIZE_BITS, slc_cache_line_size_bits), + FEATURE_MAPPING(SLC_SIZE_CONFIGURABLE, slc_size_configurable), + FEATURE_MAPPING_VALUE(SLC_SIZE_IN_KILOBYTES, slc_size_in_kilobytes), + FEATURE_MAPPING(SOC_TIMER, soc_timer), + FEATURE_MAPPING(SYS_BUS_SECURE_RESET, sys_bus_secure_reset), + FEATURE_MAPPING(TESSELLATION, tessellation), + FEATURE_MAPPING(TILE_REGION_PROTECTION, tile_region_protection), + FEATURE_MAPPING_VALUE(TILE_SIZE_X, tile_size_x), + FEATURE_MAPPING_VALUE(TILE_SIZE_Y, tile_size_y), + FEATURE_MAPPING(TLA, tla), + FEATURE_MAPPING(TPU_CEM_DATAMASTER_GLOBAL_REGISTERS, tpu_cem_datamaster_global_registers), + FEATURE_MAPPING(TPU_DM_GLOBAL_REGISTERS, tpu_dm_global_registers), + FEATURE_MAPPING(TPU_FILTERING_MODE_CONTROL, tpu_filtering_mode_control), + FEATURE_MAPPING_VALUE(USC_MIN_OUTPUT_REGISTERS_PER_PIX, usc_min_output_registers_per_pix), + FEATURE_MAPPING(VDM_DRAWINDIRECT, vdm_drawindirect), + FEATURE_MAPPING(VDM_OBJECT_LEVEL_LLS, vdm_object_level_lls), + FEATURE_MAPPING_VALUE(VIRTUAL_ADDRESS_SPACE_BITS, virtual_address_space_bits), + FEATURE_MAPPING(WATCHDOG_TIMER, watchdog_timer), + FEATURE_MAPPING(WORKGROUP_PROTECTION, workgroup_protection), + FEATURE_MAPPING_VALUE(XE_ARCHITECTURE, xe_architecture), + FEATURE_MAPPING(XE_MEMORY_HIERARCHY, xe_memory_hierarchy), + FEATURE_MAPPING(XE_TPU2, xe_tpu2), + FEATURE_MAPPING_VALUE(XPU_MAX_REGBANKS_ADDR_WIDTH, xpu_max_regbanks_addr_width), + FEATURE_MAPPING_VALUE(XPU_MAX_SLAVES, xpu_max_slaves), + FEATURE_MAPPING_VALUE(XPU_REGISTER_BROADCAST, xpu_register_broadcast), + FEATURE_MAPPING(XT_TOP_INFRASTRUCTURE, xt_top_infrastructure), + FEATURE_MAPPING(ZLS_SUBTILE, zls_subtile), +}; + +#undef FEATURE_MAPPING_VALUE +#undef FEATURE_MAPPING + +/** + * pvr_device_info_set_features() - Set device features from device information in firmware + * @pvr_dev: Device pointer. + * @features: Pointer to features mask in device information. + * @features_size: Size of features mask, in u64s. + * @feature_param_size: Size of feature parameters, in u64s. + * + * Returns: + * * 0 on success, or + * * -%EINVAL on malformed stream. + */ +int pvr_device_info_set_features(struct pvr_device *pvr_dev, const u64 *features, u32 features_size, + u32 feature_param_size) +{ + const u32 mapping_max = ARRAY_SIZE(features_mapping); + const u32 mapping_max_size = (mapping_max + 63) >> 6; + const u32 nr_bits = min(features_size * 64, mapping_max); + const u64 *feature_params = features + features_size; + u32 param_idx = 0; + + BUILD_BUG_ON(ARRAY_SIZE(features_mapping) != PVR_FW_HAS_FEATURE_MAX); + + /* Verify no unsupported values in the bitmask. */ + if (features_size > mapping_max_size) { + drm_warn(from_pvr_device(pvr_dev), "Unsupported features in firmware image"); + } else if (features_size == mapping_max_size && (mapping_max & 63)) { + u64 invalid_mask = ~0ull << (mapping_max & 63); + + if (features[features_size - 1] & invalid_mask) + drm_warn(from_pvr_device(pvr_dev), + "Unsupported features in firmware image"); + } + + for (u32 i = 0; i < nr_bits; i++) { + if (features[i >> 6] & BIT_ULL(i & 63)) { + *(bool *)((u8 *)pvr_dev + features_mapping[i].flag_offset) = true; + + if (features_mapping[i].value_offset) { + if (param_idx >= feature_param_size) + return -EINVAL; + + *(u64 *)((u8 *)pvr_dev + features_mapping[i].value_offset) = + feature_params[param_idx]; + param_idx++; + } + } + } + + return 0; +} diff --git a/drivers/gpu/drm/imagination/pvr_device_info.h b/drivers/gpu/drm/imagination/pvr_device_info.h new file mode 100644 index 000000000000..f61fb988b553 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_device_info.h @@ -0,0 +1,186 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_DEVICE_INFO_H +#define PVR_DEVICE_INFO_H + +#include + +struct pvr_device; + +/* + * struct pvr_device_features - Hardware feature information + */ +struct pvr_device_features { + bool has_axi_acelite; + bool has_cdm_control_stream_format; + bool has_cluster_grouping; + bool has_common_store_size_in_dwords; + bool has_compute; + bool has_compute_morton_capable; + bool has_compute_overlap; + bool has_coreid_per_os; + bool has_dynamic_dust_power; + bool has_ecc_rams; + bool has_fb_cdc_v4; + bool has_fbc_max_default_descriptors; + bool has_fbc_max_large_descriptors; + bool has_fbcdc; + bool has_fbcdc_algorithm; + bool has_fbcdc_architecture; + bool has_gpu_multicore_support; + bool has_gpu_virtualisation; + bool has_gs_rta_support; + bool has_irq_per_os; + bool has_isp_max_tiles_in_flight; + bool has_isp_samples_per_pixel; + bool has_isp_zls_d24_s8_packing_ogl_mode; + bool has_layout_mars; + bool has_max_partitions; + bool has_meta; + bool has_meta_coremem_size; + bool has_mips; + bool has_num_clusters; + bool has_num_isp_ipp_pipes; + bool has_num_osids; + bool has_num_raster_pipes; + bool has_pbe2_in_xe; + bool has_pbvnc_coreid_reg; + bool has_perfbus; + bool has_perf_counter_batch; + bool has_phys_bus_width; + bool has_riscv_fw_processor; + bool has_roguexe; + bool has_s7_top_infrastructure; + bool has_simple_internal_parameter_format; + bool has_simple_internal_parameter_format_v2; + bool has_simple_parameter_format_version; + bool has_slc_banks; + bool has_slc_cache_line_size_bits; + bool has_slc_size_configurable; + bool has_slc_size_in_kilobytes; + bool has_soc_timer; + bool has_sys_bus_secure_reset; + bool has_tessellation; + bool has_tile_region_protection; + bool has_tile_size_x; + bool has_tile_size_y; + bool has_tla; + bool has_tpu_cem_datamaster_global_registers; + bool has_tpu_dm_global_registers; + bool has_tpu_filtering_mode_control; + bool has_usc_min_output_registers_per_pix; + bool has_vdm_drawindirect; + bool has_vdm_object_level_lls; + bool has_virtual_address_space_bits; + bool has_watchdog_timer; + bool has_workgroup_protection; + bool has_xe_architecture; + bool has_xe_memory_hierarchy; + bool has_xe_tpu2; + bool has_xpu_max_regbanks_addr_width; + bool has_xpu_max_slaves; + bool has_xpu_register_broadcast; + bool has_xt_top_infrastructure; + bool has_zls_subtile; + + u64 cdm_control_stream_format; + u64 common_store_size_in_dwords; + u64 ecc_rams; + u64 fbc_max_default_descriptors; + u64 fbc_max_large_descriptors; + u64 fbcdc; + u64 fbcdc_algorithm; + u64 fbcdc_architecture; + u64 isp_max_tiles_in_flight; + u64 isp_samples_per_pixel; + u64 layout_mars; + u64 max_partitions; + u64 meta; + u64 meta_coremem_size; + u64 num_clusters; + u64 num_isp_ipp_pipes; + u64 num_osids; + u64 num_raster_pipes; + u64 phys_bus_width; + u64 simple_parameter_format_version; + u64 slc_banks; + u64 slc_cache_line_size_bits; + u64 slc_size_in_kilobytes; + u64 tile_size_x; + u64 tile_size_y; + u64 usc_min_output_registers_per_pix; + u64 virtual_address_space_bits; + u64 xe_architecture; + u64 xpu_max_regbanks_addr_width; + u64 xpu_max_slaves; + u64 xpu_register_broadcast; +}; + +/* + * struct pvr_device_quirks - Hardware quirk information + */ +struct pvr_device_quirks { + bool has_brn44079; + bool has_brn47217; + bool has_brn48492; + bool has_brn48545; + bool has_brn49927; + bool has_brn50767; + bool has_brn51764; + bool has_brn62269; + bool has_brn63142; + bool has_brn63553; + bool has_brn66011; + bool has_brn71242; +}; + +/* + * struct pvr_device_enhancements - Hardware enhancement information + */ +struct pvr_device_enhancements { + bool has_ern35421; + bool has_ern38020; + bool has_ern38748; + bool has_ern42064; + bool has_ern42290; + bool has_ern42606; + bool has_ern47025; + bool has_ern57596; +}; + +void pvr_device_info_set_quirks(struct pvr_device *pvr_dev, const u64 *bitmask, + u32 bitmask_len); +void pvr_device_info_set_enhancements(struct pvr_device *pvr_dev, const u64 *bitmask, + u32 bitmask_len); +int pvr_device_info_set_features(struct pvr_device *pvr_dev, const u64 *features, u32 features_size, + u32 feature_param_size); + +/* + * Meta cores + * + * These are the values for the 'meta' feature when the feature is present + * (as per &struct pvr_device_features)/ + */ +#define PVR_META_MTP218 (1) +#define PVR_META_MTP219 (2) +#define PVR_META_LTP218 (3) +#define PVR_META_LTP217 (4) + +enum { + PVR_FEATURE_CDM_USER_MODE_QUEUE, + PVR_FEATURE_CLUSTER_GROUPING, + PVR_FEATURE_COMPUTE_MORTON_CAPABLE, + PVR_FEATURE_FB_CDC_V4, + PVR_FEATURE_GPU_MULTICORE_SUPPORT, + PVR_FEATURE_ISP_ZLS_D24_S8_PACKING_OGL_MODE, + PVR_FEATURE_REQUIRES_FB_CDC_ZLS_SETUP, + PVR_FEATURE_S7_TOP_INFRASTRUCTURE, + PVR_FEATURE_TESSELLATION, + PVR_FEATURE_TPU_DM_GLOBAL_REGISTERS, + PVR_FEATURE_VDM_DRAWINDIRECT, + PVR_FEATURE_VDM_OBJECT_LEVEL_LLS, + PVR_FEATURE_ZLS_SUBTILE, +}; + +#endif /* PVR_DEVICE_INFO_H */ diff --git a/drivers/gpu/drm/imagination/pvr_drv.c b/drivers/gpu/drm/imagination/pvr_drv.c index 597188b31f96..1e4408f33e94 100644 --- a/drivers/gpu/drm/imagination/pvr_drv.c +++ b/drivers/gpu/drm/imagination/pvr_drv.c @@ -3,6 +3,9 @@ #include "pvr_device.h" #include "pvr_drv.h" +#include "pvr_rogue_defs.h" +#include "pvr_rogue_fwif_client.h" +#include "pvr_rogue_fwif_shared.h" #include @@ -87,6 +90,382 @@ pvr_ioctl_get_bo_mmap_offset(struct drm_device *drm_dev, void *raw_args, return -ENOTTY; } +static __always_inline u64 +pvr_fw_version_packed(u32 major, u32 minor) +{ + return ((u64)major << 32) | minor; +} + +static u32 +rogue_get_common_store_partition_space_size(struct pvr_device *pvr_dev) +{ + u32 max_partitions = 0; + u32 tile_size_x = 0; + u32 tile_size_y = 0; + + PVR_FEATURE_VALUE(pvr_dev, tile_size_x, &tile_size_x); + PVR_FEATURE_VALUE(pvr_dev, tile_size_y, &tile_size_y); + PVR_FEATURE_VALUE(pvr_dev, max_partitions, &max_partitions); + + if (tile_size_x == 16 && tile_size_y == 16) { + u32 usc_min_output_registers_per_pix = 0; + + PVR_FEATURE_VALUE(pvr_dev, usc_min_output_registers_per_pix, + &usc_min_output_registers_per_pix); + + return tile_size_x * tile_size_y * max_partitions * + usc_min_output_registers_per_pix; + } + + return max_partitions * 1024; +} + +static u32 +rogue_get_common_store_alloc_region_size(struct pvr_device *pvr_dev) +{ + u32 common_store_size_in_dwords = 512 * 4 * 4; + u32 alloc_region_size; + + PVR_FEATURE_VALUE(pvr_dev, common_store_size_in_dwords, &common_store_size_in_dwords); + + alloc_region_size = common_store_size_in_dwords - (256U * 4U) - + rogue_get_common_store_partition_space_size(pvr_dev); + + if (PVR_HAS_QUIRK(pvr_dev, 44079)) { + u32 common_store_split_point = (768U * 4U * 4U); + + return min(common_store_split_point - (256U * 4U), alloc_region_size); + } + + return alloc_region_size; +} + +static inline u32 +rogue_get_num_phantoms(struct pvr_device *pvr_dev) +{ + u32 num_clusters = 1; + + PVR_FEATURE_VALUE(pvr_dev, num_clusters, &num_clusters); + + return ROGUE_REQ_NUM_PHANTOMS(num_clusters); +} + +static inline u32 +rogue_get_max_coeffs(struct pvr_device *pvr_dev) +{ + u32 max_coeff_additional_portion = ROGUE_MAX_VERTEX_SHARED_REGISTERS; + u32 pending_allocation_shared_regs = 2U * 1024U; + u32 pending_allocation_coeff_regs = 0U; + u32 num_phantoms = rogue_get_num_phantoms(pvr_dev); + u32 tiles_in_flight = 0; + u32 max_coeff_pixel_portion; + + PVR_FEATURE_VALUE(pvr_dev, isp_max_tiles_in_flight, &tiles_in_flight); + max_coeff_pixel_portion = DIV_ROUND_UP(tiles_in_flight, num_phantoms); + max_coeff_pixel_portion *= ROGUE_MAX_PIXEL_SHARED_REGISTERS; + + /* + * Compute tasks on cores with BRN48492 and without compute overlap may lock + * up without two additional lines of coeffs. + */ + if (PVR_HAS_QUIRK(pvr_dev, 48492) && !PVR_HAS_FEATURE(pvr_dev, compute_overlap)) + pending_allocation_coeff_regs = 2U * 1024U; + + if (PVR_HAS_ENHANCEMENT(pvr_dev, 38748)) + pending_allocation_shared_regs = 0; + + if (PVR_HAS_ENHANCEMENT(pvr_dev, 38020)) + max_coeff_additional_portion += ROGUE_MAX_COMPUTE_SHARED_REGISTERS; + + return rogue_get_common_store_alloc_region_size(pvr_dev) + pending_allocation_coeff_regs - + (max_coeff_pixel_portion + max_coeff_additional_portion + + pending_allocation_shared_regs); +} + +static inline u32 +rogue_get_cdm_max_local_mem_size_regs(struct pvr_device *pvr_dev) +{ + u32 available_coeffs_in_dwords = rogue_get_max_coeffs(pvr_dev); + + if (PVR_HAS_QUIRK(pvr_dev, 48492) && PVR_HAS_FEATURE(pvr_dev, roguexe) && + !PVR_HAS_FEATURE(pvr_dev, compute_overlap)) { + /* Driver must not use the 2 reserved lines. */ + available_coeffs_in_dwords -= ROGUE_CSRM_LINE_SIZE_IN_DWORDS * 2; + } + + /* + * The maximum amount of local memory available to a kernel is the minimum + * of the total number of coefficient registers available and the max common + * store allocation size which can be made by the CDM. + * + * If any coeff lines are reserved for tessellation or pixel then we need to + * subtract those too. + */ + return min(available_coeffs_in_dwords, (u32)ROGUE_MAX_PER_KERNEL_LOCAL_MEM_SIZE_REGS); +} + +/** + * pvr_dev_query_gpu_info_get() + * @pvr_dev: Device pointer. + * @args: [IN] Device query arguments containing a pointer to a userspace + * struct drm_pvr_dev_query_gpu_info. + * + * If the query object pointer is NULL, the size field is updated with the + * expected size of the query object. + * + * Returns: + * * 0 on success, or if size is requested using a NULL pointer, or + * * -%E2BIG if the indicated length of the allocation is less than is + * required to contain the copied data, or + * * -%EFAULT if local memory could not be copied to userspace. + */ +static int +pvr_dev_query_gpu_info_get(struct pvr_device *pvr_dev, + struct drm_pvr_ioctl_dev_query_args *args) +{ + struct drm_pvr_dev_query_gpu_info gpu_info = {0}; + int err; + + if (!args->pointer) { + args->size = sizeof(struct drm_pvr_dev_query_gpu_info); + return 0; + } + + gpu_info.gpu_id = + pvr_gpu_id_to_packed_bvnc(&pvr_dev->gpu_id); + gpu_info.num_phantoms = rogue_get_num_phantoms(pvr_dev); + + err = PVR_UOBJ_SET(args->pointer, args->size, gpu_info); + if (err < 0) + return err; + + if (args->size > sizeof(gpu_info)) + args->size = sizeof(gpu_info); + return 0; +} + +/** + * pvr_dev_query_runtime_info_get() + * @pvr_dev: Device pointer. + * @args: [IN] Device query arguments containing a pointer to a userspace + * struct drm_pvr_dev_query_runtime_info. + * + * If the query object pointer is NULL, the size field is updated with the + * expected size of the query object. + * + * Returns: + * * 0 on success, or if size is requested using a NULL pointer, or + * * -%E2BIG if the indicated length of the allocation is less than is + * required to contain the copied data, or + * * -%EFAULT if local memory could not be copied to userspace. + */ +static int +pvr_dev_query_runtime_info_get(struct pvr_device *pvr_dev, + struct drm_pvr_ioctl_dev_query_args *args) +{ + struct drm_pvr_dev_query_runtime_info runtime_info = {0}; + int err; + + if (!args->pointer) { + args->size = sizeof(struct drm_pvr_dev_query_runtime_info); + return 0; + } + + runtime_info.free_list_min_pages = 0; /* FIXME */ + runtime_info.free_list_max_pages = + ROGUE_PM_MAX_FREELIST_SIZE / ROGUE_PM_PAGE_SIZE; + runtime_info.common_store_alloc_region_size = + rogue_get_common_store_alloc_region_size(pvr_dev); + runtime_info.common_store_partition_space_size = + rogue_get_common_store_partition_space_size(pvr_dev); + runtime_info.max_coeffs = rogue_get_max_coeffs(pvr_dev); + runtime_info.cdm_max_local_mem_size_regs = + rogue_get_cdm_max_local_mem_size_regs(pvr_dev); + + err = PVR_UOBJ_SET(args->pointer, args->size, runtime_info); + if (err < 0) + return err; + + if (args->size > sizeof(runtime_info)) + args->size = sizeof(runtime_info); + return 0; +} + +/** + * pvr_dev_query_quirks_get() - Unpack array of quirks at the address given + * in a struct drm_pvr_dev_query_quirks, or gets the amount of space required + * for it. + * @pvr_dev: Device pointer. + * @args: [IN] Device query arguments containing a pointer to a userspace + * struct drm_pvr_dev_query_query_quirks. + * + * If the query object pointer is NULL, the size field is updated with the + * expected size of the query object. + * If the userspace pointer in the query object is NULL, or the count is + * short, no data is copied. + * The count field will be updated to that copied, or if either pointer is + * NULL, that which would have been copied. + * The size field in the query object will be updated to the size copied. + * + * Returns: + * * 0 on success, or if size/count is requested using a NULL pointer, or + * * -%EINVAL if args contained non-zero reserved fields, or + * * -%E2BIG if the indicated length of the allocation is less than is + * required to contain the copied data, or + * * -%EFAULT if local memory could not be copied to userspace. + */ +static int +pvr_dev_query_quirks_get(struct pvr_device *pvr_dev, + struct drm_pvr_ioctl_dev_query_args *args) +{ + /* + * @FIXME - hardcoding of numbers here is intended as an + * intermediate step so the UAPI can be fixed, but requires a + * a refactor in the future to store them in a more appropriate + * location + */ + static const u32 umd_quirks_musthave[] = { + 47217, + 49927, + 62269, + }; + static const u32 umd_quirks[] = { + 48545, + 51764, + }; + struct drm_pvr_dev_query_quirks query; + u32 out[ARRAY_SIZE(umd_quirks_musthave) + ARRAY_SIZE(umd_quirks)]; + size_t out_musthave_count = 0; + size_t out_count = 0; + int err; + + if (!args->pointer) { + args->size = sizeof(struct drm_pvr_dev_query_quirks); + return 0; + } + + err = PVR_UOBJ_GET(query, args->size, args->pointer); + + if (err < 0) + return err; + if (query._padding_c) + return -EINVAL; + + for (int i = 0; i < ARRAY_SIZE(umd_quirks_musthave); i++) { + if (pvr_device_has_uapi_quirk(pvr_dev, umd_quirks_musthave[i])) { + out[out_count++] = umd_quirks_musthave[i]; + out_musthave_count++; + } + } + + for (int i = 0; i < ARRAY_SIZE(umd_quirks); i++) { + if (pvr_device_has_uapi_quirk(pvr_dev, umd_quirks[i])) + out[out_count++] = umd_quirks[i]; + } + + if (!query.quirks) + goto copy_out; + if (query.count < out_count) + return -E2BIG; + + if (copy_to_user(u64_to_user_ptr(query.quirks), out, + out_count * sizeof(u32))) { + return -EFAULT; + } + + query.musthave_count = out_musthave_count; + +copy_out: + query.count = out_count; + err = PVR_UOBJ_SET(args->pointer, args->size, query); + if (err < 0) + return err; + + args->size = sizeof(query); + return 0; +} + +/** + * pvr_dev_query_enhancements_get() - Unpack array of enhancements at the + * address given in a struct drm_pvr_dev_query_enhancements, or gets the amount + * of space required for it. + * @pvr_dev: Device pointer. + * @args: [IN] Device query arguments containing a pointer to a userspace + * struct drm_pvr_dev_query_enhancements. + * + * If the query object pointer is NULL, the size field is updated with the + * expected size of the query object. + * If the userspace pointer in the query object is NULL, or the count is + * short, no data is copied. + * The count field will be updated to that copied, or if either pointer is + * NULL, that which would have been copied. + * The size field in the query object will be updated to the size copied. + * + * Returns: + * * 0 on success, or if size/count is requested using a NULL pointer, or + * * -%EINVAL if args contained non-zero reserved fields, or + * * -%E2BIG if the indicated length of the allocation is less than is + * required to contain the copied data, or + * * -%EFAULT if local memory could not be copied to userspace. + */ +static int +pvr_dev_query_enhancements_get(struct pvr_device *pvr_dev, + struct drm_pvr_ioctl_dev_query_args *args) +{ + /* + * @FIXME - hardcoding of numbers here is intended as an + * intermediate step so the UAPI can be fixed, but requires a + * a refactor in the future to store them in a more appropriate + * location + */ + const u32 umd_enhancements[] = { + 35421, + 42064, + }; + struct drm_pvr_dev_query_enhancements query; + u32 out[ARRAY_SIZE(umd_enhancements)]; + size_t out_idx = 0; + int err; + + if (!args->pointer) { + args->size = sizeof(struct drm_pvr_dev_query_enhancements); + return 0; + } + + err = PVR_UOBJ_GET(query, args->size, args->pointer); + + if (err < 0) + return err; + if (query._padding_a) + return -EINVAL; + if (query._padding_c) + return -EINVAL; + + for (int i = 0; i < ARRAY_SIZE(umd_enhancements); i++) { + if (pvr_device_has_uapi_enhancement(pvr_dev, umd_enhancements[i])) + out[out_idx++] = umd_enhancements[i]; + } + + if (!query.enhancements) + goto copy_out; + if (query.count < out_idx) + return -E2BIG; + + if (copy_to_user(u64_to_user_ptr(query.enhancements), out, + out_idx * sizeof(u32))) { + return -EFAULT; + } + +copy_out: + query.count = out_idx; + err = PVR_UOBJ_SET(args->pointer, args->size, query); + if (err < 0) + return err; + + args->size = sizeof(query); + return 0; +} + /** * pvr_ioctl_dev_query() - IOCTL to copy information about a device * @drm_dev: [IN] DRM device. @@ -111,7 +490,41 @@ static int pvr_ioctl_dev_query(struct drm_device *drm_dev, void *raw_args, struct drm_file *file) { - return -ENOTTY; + struct pvr_device *pvr_dev = to_pvr_device(drm_dev); + struct drm_pvr_ioctl_dev_query_args *args = raw_args; + int idx; + int ret = -EINVAL; + + if (!drm_dev_enter(drm_dev, &idx)) + return -EIO; + + switch ((enum drm_pvr_dev_query)args->type) { + case DRM_PVR_DEV_QUERY_GPU_INFO_GET: + ret = pvr_dev_query_gpu_info_get(pvr_dev, args); + break; + + case DRM_PVR_DEV_QUERY_RUNTIME_INFO_GET: + ret = pvr_dev_query_runtime_info_get(pvr_dev, args); + break; + + case DRM_PVR_DEV_QUERY_QUIRKS_GET: + ret = pvr_dev_query_quirks_get(pvr_dev, args); + break; + + case DRM_PVR_DEV_QUERY_ENHANCEMENTS_GET: + ret = pvr_dev_query_enhancements_get(pvr_dev, args); + break; + + case DRM_PVR_DEV_QUERY_HEAP_INFO_GET: + return -EINVAL; + + case DRM_PVR_DEV_QUERY_STATIC_DATA_AREAS_GET: + return -EINVAL; + } + + drm_dev_exit(idx); + + return ret; } /** @@ -349,6 +762,112 @@ pvr_ioctl_submit_jobs(struct drm_device *drm_dev, void *raw_args, return -ENOTTY; } +int +pvr_get_uobj(u64 usr_ptr, u32 usr_stride, u32 min_stride, u32 obj_size, void *out) +{ + if (usr_stride < min_stride) + return -EINVAL; + + return copy_struct_from_user(out, obj_size, u64_to_user_ptr(usr_ptr), usr_stride); +} + +int +pvr_set_uobj(u64 usr_ptr, u32 usr_stride, u32 min_stride, u32 obj_size, const void *in) +{ + if (usr_stride < min_stride) + return -EINVAL; + + if (copy_to_user(u64_to_user_ptr(usr_ptr), in, min_t(u32, usr_stride, obj_size))) + return -EFAULT; + + if (usr_stride > obj_size && + clear_user(u64_to_user_ptr(usr_ptr + obj_size), usr_stride - obj_size)) { + return -EFAULT; + } + + return 0; +} + +int +pvr_get_uobj_array(const struct drm_pvr_obj_array *in, u32 min_stride, u32 obj_size, void **out) +{ + int ret = 0; + void *out_alloc; + + if (in->stride < min_stride) + return -EINVAL; + + if (!in->count) + return 0; + + out_alloc = kvmalloc_array(in->count, obj_size, GFP_KERNEL); + if (!out_alloc) + return -ENOMEM; + + if (obj_size == in->stride) { + if (copy_from_user(out_alloc, u64_to_user_ptr(in->array), + (unsigned long)obj_size * in->count)) + ret = -EFAULT; + } else { + void __user *in_ptr = u64_to_user_ptr(in->array); + void *out_ptr = out_alloc; + + for (u32 i = 0; i < in->count; i++) { + ret = copy_struct_from_user(out_ptr, obj_size, in_ptr, in->stride); + if (ret) + break; + + out_ptr += obj_size; + in_ptr += in->stride; + } + } + + if (ret) { + kvfree(out_alloc); + return ret; + } + + *out = out_alloc; + return 0; +} + +int +pvr_set_uobj_array(const struct drm_pvr_obj_array *out, u32 min_stride, u32 obj_size, + const void *in) +{ + if (out->stride < min_stride) + return -EINVAL; + + if (!out->count) + return 0; + + if (obj_size == out->stride) { + if (copy_to_user(u64_to_user_ptr(out->array), in, + (unsigned long)obj_size * out->count)) + return -EFAULT; + } else { + u32 cpy_elem_size = min_t(u32, out->stride, obj_size); + void __user *out_ptr = u64_to_user_ptr(out->array); + const void *in_ptr = in; + + for (u32 i = 0; i < out->count; i++) { + if (copy_to_user(out_ptr, in_ptr, cpy_elem_size)) + return -EFAULT; + + out_ptr += obj_size; + in_ptr += out->stride; + } + + if (out->stride > obj_size && + clear_user(u64_to_user_ptr(out->array + obj_size), + out->stride - obj_size)) { + return -EFAULT; + } + } + + return 0; +} + #define DRM_PVR_IOCTL(_name, _func, _flags) \ DRM_IOCTL_DEF_DRV(PVR_##_name, pvr_ioctl_##_func, _flags) diff --git a/drivers/gpu/drm/imagination/pvr_drv.h b/drivers/gpu/drm/imagination/pvr_drv.h index c29d11af88fe..378fe477b759 100644 --- a/drivers/gpu/drm/imagination/pvr_drv.h +++ b/drivers/gpu/drm/imagination/pvr_drv.h @@ -19,4 +19,111 @@ #define PVR_DRIVER_MINOR 0 #define PVR_DRIVER_PATCHLEVEL 0 +int pvr_get_uobj(u64 usr_ptr, u32 usr_size, u32 min_size, u32 obj_size, void *out); +int pvr_set_uobj(u64 usr_ptr, u32 usr_size, u32 min_size, u32 obj_size, const void *in); +int pvr_get_uobj_array(const struct drm_pvr_obj_array *in, u32 min_stride, u32 obj_size, + void **out); +int pvr_set_uobj_array(const struct drm_pvr_obj_array *out, u32 min_stride, u32 obj_size, + const void *in); + +#define PVR_UOBJ_MIN_SIZE_INTERNAL(_typename, _last_mandatory_field) \ + (offsetof(_typename, _last_mandatory_field) + \ + sizeof(((_typename *)NULL)->_last_mandatory_field)) + +/* NOLINTBEGIN(bugprone-macro-parentheses) */ +#define PVR_UOBJ_DECL(_typename, _last_mandatory_field) \ + , _typename : PVR_UOBJ_MIN_SIZE_INTERNAL(_typename, _last_mandatory_field) +/* NOLINTEND(bugprone-macro-parentheses) */ + +/** + * DOC: PVR user objects. + * + * Macros used to aid copying structured and array data to and from + * userspace. Objects can differ in size, provided the minimum size + * allowed is specified (using the last mandatory field in the struct). + * All types used with PVR_UOBJ_GET/SET macros must be listed here under + * PVR_UOBJ_MIN_SIZE, with the last mandatory struct field specified. + */ + +/** + * PVR_UOBJ_MIN_SIZE() - Fetch the minimum copy size of a compatible type object. + * @_obj_name: The name of the object. Cannot be a typename - this is deduced. + * + * This cannot fail. Using the macro with an incompatible type will result in a + * compiler error. + * + * To add compatibility for a type, list it within the macro in an orderly + * fashion. The second argument is the name of the last mandatory field of the + * struct type, which is used to calculate the size. See also PVR_UOBJ_DECL(). + * + * Return: The minimum copy size. + */ +#define PVR_UOBJ_MIN_SIZE(_obj_name) _Generic(_obj_name \ + PVR_UOBJ_DECL(struct drm_pvr_job, hwrt) \ + PVR_UOBJ_DECL(struct drm_pvr_sync_op, value) \ + PVR_UOBJ_DECL(struct drm_pvr_dev_query_gpu_info, num_phantoms) \ + PVR_UOBJ_DECL(struct drm_pvr_dev_query_runtime_info, cdm_max_local_mem_size_regs) \ + PVR_UOBJ_DECL(struct drm_pvr_dev_query_quirks, _padding_c) \ + PVR_UOBJ_DECL(struct drm_pvr_dev_query_enhancements, _padding_c) \ + PVR_UOBJ_DECL(struct drm_pvr_heap, page_size_log2) \ + PVR_UOBJ_DECL(struct drm_pvr_dev_query_heap_info, heaps) \ + PVR_UOBJ_DECL(struct drm_pvr_static_data_area, offset) \ + PVR_UOBJ_DECL(struct drm_pvr_dev_query_static_data_areas, static_data_areas) \ + ) + +/** + * PVR_UOBJ_GET() - Copies from _src_usr_ptr to &_dest_obj. + * @_dest_obj: The destination container object in kernel space. + * @_usr_size: The size of the source container in user space. + * @_src_usr_ptr: __u64 raw pointer to the source container in user space. + * + * Return: Error code. See pvr_get_uobj(). + */ +#define PVR_UOBJ_GET(_dest_obj, _usr_size, _src_usr_ptr) \ + pvr_get_uobj(_src_usr_ptr, _usr_size, \ + PVR_UOBJ_MIN_SIZE(_dest_obj), \ + sizeof(_dest_obj), &(_dest_obj)) + +/** + * PVR_UOBJ_SET() - Copies from &_src_obj to _dest_usr_ptr. + * @_dest_usr_ptr: __u64 raw pointer to the destination container in user space. + * @_usr_size: The size of the destination container in user space. + * @_src_obj: The source container object in kernel space. + * + * Return: Error code. See pvr_set_uobj(). + */ +#define PVR_UOBJ_SET(_dest_usr_ptr, _usr_size, _src_obj) \ + pvr_set_uobj(_dest_usr_ptr, _usr_size, \ + PVR_UOBJ_MIN_SIZE(_src_obj), \ + sizeof(_src_obj), &(_src_obj)) + +/** + * PVR_UOBJ_GET_ARRAY() - Copies from @_src_drm_pvr_obj_array.array to + * alloced memory and returns a pointer in _dest_array. + * @_dest_array: The destination C array object in kernel space. + * @_src_drm_pvr_obj_array: The &struct drm_pvr_obj_array containing a __u64 raw + * pointer to the source C array in user space and the size of each array + * element in user space (the 'stride'). + * + * Return: Error code. See pvr_get_uobj_array(). + */ +#define PVR_UOBJ_GET_ARRAY(_dest_array, _src_drm_pvr_obj_array) \ + pvr_get_uobj_array(_src_drm_pvr_obj_array, \ + PVR_UOBJ_MIN_SIZE((_dest_array)[0]), \ + sizeof((_dest_array)[0]), (void **)&(_dest_array)) + +/** + * PVR_UOBJ_SET_ARRAY() - Copies from _src_array to @_dest_drm_pvr_obj_array.array. + * @_dest_drm_pvr_obj_array: The &struct drm_pvr_obj_array containing a __u64 raw + * pointer to the destination C array in user space and the size of each array + * element in user space (the 'stride'). + * @_src_array: The source C array object in kernel space. + * + * Return: Error code. See pvr_set_uobj_array(). + */ +#define PVR_UOBJ_SET_ARRAY(_dest_drm_pvr_obj_array, _src_array) \ + pvr_set_uobj_array(_dest_drm_pvr_obj_array, \ + PVR_UOBJ_MIN_SIZE((_src_array)[0]), \ + sizeof((_src_array)[0]), _src_array) + #endif /* PVR_DRV_H */ diff --git a/drivers/gpu/drm/imagination/pvr_fw.c b/drivers/gpu/drm/imagination/pvr_fw.c new file mode 100644 index 000000000000..a9377c735c9c --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_fw.c @@ -0,0 +1,145 @@ +// SPDX-License-Identifier: GPL-2.0-only OR MIT +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#include "pvr_device.h" +#include "pvr_device_info.h" +#include "pvr_fw.h" + +#include +#include +#include + +#define FW_MAX_SUPPORTED_MAJOR_VERSION 1 + +/** + * pvr_fw_validate() - Parse firmware header and check compatibility + * @pvr_dev: Device pointer. + * + * Returns: + * * 0 on success, or + * * -EINVAL if firmware is incompatible. + */ +static int +pvr_fw_validate(struct pvr_device *pvr_dev) +{ + struct drm_device *drm_dev = from_pvr_device(pvr_dev); + const struct firmware *firmware = pvr_dev->fw_dev.firmware; + const struct pvr_fw_layout_entry *layout_entries; + const struct pvr_fw_info_header *header; + const u8 *fw = firmware->data; + u32 fw_offset = firmware->size - SZ_4K; + u32 layout_table_size; + u32 entry; + + if (firmware->size < SZ_4K || (firmware->size % FW_BLOCK_SIZE)) + return -EINVAL; + + header = (const struct pvr_fw_info_header *)&fw[fw_offset]; + + if (header->info_version != PVR_FW_INFO_VERSION) { + drm_err(drm_dev, "Unsupported fw info version %u\n", + header->info_version); + return -EINVAL; + } + + if (header->header_len != sizeof(struct pvr_fw_info_header) || + header->layout_entry_size != sizeof(struct pvr_fw_layout_entry) || + header->layout_entry_num > PVR_FW_INFO_MAX_NUM_ENTRIES) { + drm_err(drm_dev, "FW info format mismatch\n"); + return -EINVAL; + } + + if (!(header->flags & PVR_FW_FLAGS_OPEN_SOURCE) || + header->fw_version_major > FW_MAX_SUPPORTED_MAJOR_VERSION || + header->fw_version_major == 0) { + drm_err(drm_dev, "Unsupported FW version %u.%u (build: %u%s)\n", + header->fw_version_major, header->fw_version_minor, + header->fw_version_build, + (header->flags & PVR_FW_FLAGS_OPEN_SOURCE) ? " OS" : ""); + return -EINVAL; + } + + if (pvr_gpu_id_to_packed_bvnc(&pvr_dev->gpu_id) != header->bvnc) { + struct pvr_gpu_id fw_gpu_id; + + packed_bvnc_to_pvr_gpu_id(header->bvnc, &fw_gpu_id); + drm_err(drm_dev, "FW built for incorrect GPU ID %i.%i.%i.%i (expected %i.%i.%i.%i)\n", + fw_gpu_id.b, fw_gpu_id.v, fw_gpu_id.n, fw_gpu_id.c, + pvr_dev->gpu_id.b, pvr_dev->gpu_id.v, pvr_dev->gpu_id.n, pvr_dev->gpu_id.c); + return -EINVAL; + } + + fw_offset += header->header_len; + layout_table_size = + header->layout_entry_size * header->layout_entry_num; + if ((fw_offset + layout_table_size) > firmware->size) + return -EINVAL; + + layout_entries = (const struct pvr_fw_layout_entry *)&fw[fw_offset]; + for (entry = 0; entry < header->layout_entry_num; entry++) { + u32 start_addr = layout_entries[entry].base_addr; + u32 end_addr = start_addr + layout_entries[entry].alloc_size; + + if (start_addr >= end_addr) + return -EINVAL; + } + + fw_offset = (firmware->size - SZ_4K) - header->device_info_size; + + drm_info(drm_dev, "FW version v%u.%u (build %u OS)\n", header->fw_version_major, + header->fw_version_minor, header->fw_version_build); + + pvr_dev->fw_version.major = header->fw_version_major; + pvr_dev->fw_version.minor = header->fw_version_minor; + + pvr_dev->fw_dev.header = header; + pvr_dev->fw_dev.layout_entries = layout_entries; + + return 0; +} + +static int +pvr_fw_get_device_info(struct pvr_device *pvr_dev) +{ + const struct firmware *firmware = pvr_dev->fw_dev.firmware; + struct pvr_fw_device_info_header *header; + const u8 *fw = firmware->data; + const u64 *dev_info; + u32 fw_offset; + + fw_offset = (firmware->size - SZ_4K) - pvr_dev->fw_dev.header->device_info_size; + + header = (struct pvr_fw_device_info_header *)&fw[fw_offset]; + dev_info = (u64 *)(header + 1); + + pvr_device_info_set_quirks(pvr_dev, dev_info, header->brn_mask_size); + dev_info += header->brn_mask_size; + + pvr_device_info_set_enhancements(pvr_dev, dev_info, header->ern_mask_size); + dev_info += header->ern_mask_size; + + return pvr_device_info_set_features(pvr_dev, dev_info, header->feature_mask_size, + header->feature_param_size); +} + +/** + * pvr_fw_validate_init_device_info() - Validate firmware and initialise device information + * @pvr_dev: Target PowerVR device. + * + * This function must be called before querying device information. + * + * Returns: + * * 0 on success, or + * * -%EINVAL if firmware validation fails. + */ +int +pvr_fw_validate_init_device_info(struct pvr_device *pvr_dev) +{ + int err; + + err = pvr_fw_validate(pvr_dev); + if (err) + return err; + + return pvr_fw_get_device_info(pvr_dev); +} diff --git a/drivers/gpu/drm/imagination/pvr_fw.h b/drivers/gpu/drm/imagination/pvr_fw.h new file mode 100644 index 000000000000..8331344e83ea --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_fw.h @@ -0,0 +1,34 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_FW_H +#define PVR_FW_H + +#include "pvr_fw_info.h" + +#include + +/* Forward declarations from "pvr_device.h". */ +struct pvr_device; +struct pvr_file; + +struct pvr_fw_device { + /** @firmware: Handle to the firmware loaded into the device. */ + const struct firmware *firmware; + + /** @header: Pointer to firmware header. */ + const struct pvr_fw_info_header *header; + + /** @layout_entries: Pointer to firmware layout. */ + const struct pvr_fw_layout_entry *layout_entries; + + /** + * @processor_type: FW processor type for this device. Must be one of + * %PVR_FW_PROCESSOR_TYPE_*. + */ + u16 processor_type; +}; + +int pvr_fw_validate_init_device_info(struct pvr_device *pvr_dev); + +#endif /* PVR_FW_H */ diff --git a/drivers/gpu/drm/imagination/pvr_fw_info.h b/drivers/gpu/drm/imagination/pvr_fw_info.h new file mode 100644 index 000000000000..ad5d44a3067a --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_fw_info.h @@ -0,0 +1,135 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_FW_INFO_H +#define PVR_FW_INFO_H + +#include +#include +#include + +/* + * Firmware binary block unit in bytes. + * Raw data stored in FW binary will be aligned to this size. + */ +#define FW_BLOCK_SIZE SZ_4K + +/* Maximum number of entries in firmware layout table. */ +#define PVR_FW_INFO_MAX_NUM_ENTRIES 8 + +enum pvr_fw_section_id { + META_CODE = 0, + META_PRIVATE_DATA, + META_COREMEM_CODE, + META_COREMEM_DATA, + MIPS_CODE, + MIPS_EXCEPTIONS_CODE, + MIPS_BOOT_CODE, + MIPS_PRIVATE_DATA, + MIPS_BOOT_DATA, + MIPS_STACK, + RISCV_UNCACHED_CODE, + RISCV_CACHED_CODE, + RISCV_PRIVATE_DATA, + RISCV_COREMEM_CODE, + RISCV_COREMEM_DATA, +}; + +enum pvr_fw_section_type { + NONE = 0, + FW_CODE, + FW_DATA, + FW_COREMEM_CODE, + FW_COREMEM_DATA, +}; + +/* + * FW binary format with FW info attached: + * + * Contents Offset + * +-----------------+ + * | | 0 + * | | + * | Original binary | + * | file | + * | (.ldr/.elf) | + * | | + * | | + * +-----------------+ + * | Device info | FILE_SIZE - 4K - device_info_size + * +-----------------+ + * | FW info header | FILE_SIZE - 4K + * +-----------------+ + * | | + * | FW layout table | + * | | + * +-----------------+ + * FILE_SIZE + */ + +#define PVR_FW_INFO_VERSION 3 + +#define PVR_FW_FLAGS_OPEN_SOURCE BIT(0) + +/** struct pvr_fw_info_header - Firmware header */ +struct pvr_fw_info_header { + /** @info_version: FW info header version. */ + u32 info_version; + /** @header_len: Header length. */ + u32 header_len; + /** @layout_entry_num: Number of entries in the layout table. */ + u32 layout_entry_num; + /** @layout_entry_size: Size of an entry in the layout table. */ + u32 layout_entry_size; + /** @bvnc: GPU ID supported by firmware. */ + aligned_u64 bvnc; + /** @fw_page_size: Page size of processor on which firmware executes. */ + u32 fw_page_size; + /** @flags: Compatibility flags. */ + u32 flags; + /** @fw_version_major: Firmware major version number. */ + u16 fw_version_major; + /** @fw_version_minor: Firmware minor version number. */ + u16 fw_version_minor; + /** @fw_version_build: Firmware build number. */ + u32 fw_version_build; + /** @device_info_size: Size of device info structure. */ + u32 device_info_size; + /** @padding: Padding. */ + u32 padding; +}; + +/** + * struct pvr_fw_layout_entry - Entry in firmware layout table, describing a + * section of the firmware image + */ +struct pvr_fw_layout_entry { + /** @id: Section ID. */ + enum pvr_fw_section_id id; + /** @type: Section type. */ + enum pvr_fw_section_type type; + /** @base_addr: Base address of section in FW address space. */ + u32 base_addr; + /** @max_size: Maximum size of section, in bytes. */ + u32 max_size; + /** @alloc_size: Allocation size of section, in bytes. */ + u32 alloc_size; + /** @alloc_offset: Allocation offset of section. */ + u32 alloc_offset; +}; + +/** + * struct pvr_fw_device_info_header - Device information header. + */ +struct pvr_fw_device_info_header { + /* BRN Mask size (in u64s). */ + u64 brn_mask_size; + /* ERN Mask size (in u64s). */ + u64 ern_mask_size; + /* Feature Mask size (in u64s). */ + u64 feature_mask_size; + /* Feature Parameter size (in u64s). */ + u64 feature_param_size; +}; + +#endif /* PVR_FW_INFO_H */ From patchwork Wed Nov 22 16:34:33 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Donald Robson X-Patchwork-Id: 746150 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=imgtec.com header.i=@imgtec.com header.b="vJvgX8XL"; dkim=pass (1024-bit key) header.d=IMGTecCRM.onmicrosoft.com header.i=@IMGTecCRM.onmicrosoft.com header.b="FOOIHI77" Received: from mx07-00376f01.pphosted.com (mx07-00376f01.pphosted.com [185.132.180.163]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 1D8B5D53; Wed, 22 Nov 2023 08:35:57 -0800 (PST) Received: from pps.filterd (m0168889.ppops.net [127.0.0.1]) by mx07-00376f01.pphosted.com (8.17.1.24/8.17.1.24) with ESMTP id 3AM7nIFd030359; Wed, 22 Nov 2023 16:35:29 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=imgtec.com; h= from:to:cc:subject:date:message-id:in-reply-to:references :content-transfer-encoding:content-type:mime-version; s= dk201812; bh=nb3bB0GBBq2k+t711y9wbdFIHseSOdel5Z5NkhEN6bg=; b=vJv gX8XLBLWWcrFkUexFBabFMNNjzF75H4HasnIZwmL0d7DXGi5G+iGi+okSiwUpOKh sjDzVhrH0MC3MbWsmUfFJsJWwWwG7qqgT8A2aUKBY89zU7lp0lP7qC0X+pPwOsWU NRhrOYspJsjatSK4B/qDAHwbC4mttBG2KTH/NYdBqHfIREfSwgCbR2F6geibfnlO 0ka8NCrhwAmZz8uK6IpZzxSPqMvpohK0Sy8sNDceesMqg3FA+Bci+0PbBWb91aur vCsEqaBLQVTV47HEWaBDpunbKyceU082kPEgvn0pQbMn87BJAaJclH+tzK45vCDN adacA6u557qwTnFzULg== Received: from hhmail05.hh.imgtec.org ([217.156.249.195]) by mx07-00376f01.pphosted.com (PPS) with ESMTPS id 3ug99g9m29-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128 verify=NOT); Wed, 22 Nov 2023 16:35:28 +0000 (GMT) Received: from HHMAIL04.hh.imgtec.org (10.100.10.119) by HHMAIL05.hh.imgtec.org (10.100.10.120) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34; Wed, 22 Nov 2023 16:35:27 +0000 Received: from GBR01-LO4-obe.outbound.protection.outlook.com (104.47.85.105) by email.imgtec.com (10.100.10.121) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34 via Frontend Transport; Wed, 22 Nov 2023 16:35:27 +0000 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=ARH1ErKeRfVdBZK7mnRaIFm2amJ8yfwLA2MhWJ65LTm/L72vMJc9iS+4SNjvaLfXHejprvu+/fEfAX1AZSCj9sBXqDKnkizbm2vlRa5I/5liSq2sqFn9Pihcma1VXmEitgNk5CkkPlCvxgdlpml7J8yVo5WjArP/i4rPA8aTCUsnc/hIyjowGYEIWXGbo5hpme4TfG9JBOZDQKVnMdNEUjmIHX15EjSH2qAGiCEHi+YYzuDgPd+6Fghv4xyqRf8HM7NXTnCpbCYU+qbJYpJE6CSNJOysA19XMBb7TRkWACg+/EylhoZJqLJmnKg6A3hDG86ZF95ptrsYqaLNXf6QGw== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=nb3bB0GBBq2k+t711y9wbdFIHseSOdel5Z5NkhEN6bg=; b=EJ0SDN7Gk8gHjWtl/+tdeAOrqbLKq9GMQNCWvuQUv+U2dWiXKZTpREJ0FGOQVrZL5YKUrpaJAxHMtcvVxZdL07L5XgYBzvAo60kl+h5esXZPCfC7fFWuK6NFD+X87eIiROmV30rlmo5y1by0mQEKuhaDU1yA7k40MYcs+iHnr6+JX3m0sFSYSl7kjiPeSmdW3bHue833tykR8Ym5Nt942z26Qz/ra7IUZ6x5j5se81FnVCfW+FyZEtfWZ9D/ps51Hxj26/njZ+VEFF7M+sd5k4Ew0PiuRuRU2ZMwxOy1tX6sP0Vb4ovNfyVdwABq2+DOYqzQsGnV8qsJSHgaFEw/FQ== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=imgtec.com; dmarc=pass action=none header.from=imgtec.com; dkim=pass header.d=imgtec.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=IMGTecCRM.onmicrosoft.com; s=selector2-IMGTecCRM-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=nb3bB0GBBq2k+t711y9wbdFIHseSOdel5Z5NkhEN6bg=; b=FOOIHI77JlFqdOvzA/dbj0GyXxvjb+l7T+hA6lD/Nqpom0ZQJFmYotZvv+lK48syWX7Wi41dRUVi27AEHPXGrOlqGJPSgL+lY/RWEVk8SaRWbatygq6oS/UXs6q5v9PyQoHf35en4rawh9NKbyzkD98Q7Ez3WNr1ukRYdrBsXbQ= Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) by LO6P265MB6459.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:2df::12) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.7025.18; Wed, 22 Nov 2023 16:35:21 +0000 Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34]) by CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34%3]) with mapi id 15.20.7025.017; Wed, 22 Nov 2023 16:35:21 +0000 From: Donald Robson To: CC: , , , , , , , , , , , , , , , , , , , , , , Sarah Walker Subject: [PATCH v9 12/20] drm/imagination: Implement power management Date: Wed, 22 Nov 2023 16:34:33 +0000 Message-Id: X-Mailer: git-send-email 2.25.1 In-Reply-To: References: X-ClientProxiedBy: LO4P265CA0273.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:37a::10) To CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) Precedence: bulk X-Mailing-List: devicetree@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: CWLP265MB5770:EE_|LO6P265MB6459:EE_ X-MS-Office365-Filtering-Correlation-Id: 254bb5aa-3cf4-4211-a50f-08dbeb7904fc X-MS-Exchange-SenderADCheck: 1 X-MS-Exchange-AntiSpam-Relay: 0 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-Forefront-Antispam-Report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:; IPV:NLI; SFV:NSPM; H:CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM; PTR:; CAT:NONE; SFS:(13230031)(39850400004)(346002)(396003)(366004)(136003)(376002)(230922051799003)(64100799003)(186009)(451199024)(1800799012)(30864003)(36756003)(7416002)(41300700001)(5660300002)(2906002)(86362001)(38350700005)(26005)(6666004)(478600001)(6506007)(2616005)(6486002)(107886003)(6512007)(52116002)(8936002)(8676002)(4326008)(44832011)(66946007)(316002)(6916009)(66556008)(66476007)(83380400001)(38100700002); DIR:OUT; SFP:1102; X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: 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 X-MS-Exchange-CrossTenant-Network-Message-Id: 254bb5aa-3cf4-4211-a50f-08dbeb7904fc X-MS-Exchange-CrossTenant-AuthSource: CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 22 Nov 2023 16:35:21.1469 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 0d5fd8bb-e8c2-4e0a-8dd5-2c264f7140fe X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: jZnUXmnR6Mi4Dt+u+oUtGDOviOcRFDMXy55I8VOgPA9v23Jh5rsaSCj/sY0/WAvHGm0mMZktOO4LheiH/AVSw4ghME8yg7x2qLGlZUceHLE= X-MS-Exchange-Transport-CrossTenantHeadersStamped: LO6P265MB6459 X-OriginatorOrg: imgtec.com X-EXCLAIMER-MD-CONFIG: 15a78312-3e47-46eb-9010-2e54d84a9631 X-Proofpoint-ORIG-GUID: eeWb9vj-U3KAmQQ85izXVAo9-t7LEifx X-Proofpoint-GUID: eeWb9vj-U3KAmQQ85izXVAo9-t7LEifx From: Sarah Walker Add power management to the driver, using runtime pm. The power off sequence depends on firmware commands which are not implemented in this patch. Changes since v8: - Corrected license identifiers Changes since v5: - Use RUNTIME_PM_OPS() to declare PM callbacks - Add Kconfig dependency on CONFIG_PM Changes since v4: - Suspend runtime PM before unplugging device on rmmod Changes since v3: - Don't power device when calling pvr_device_gpu_fini() - Documentation for pvr_dev->lost has been improved - pvr_power_init() renamed to pvr_watchdog_init() - Use drm_dev_{enter,exit} Changes since v2: - Use runtime PM - Implement watchdog Signed-off-by: Sarah Walker Signed-off-by: Donald Robson Reviewed-by: Maxime Ripard --- drivers/gpu/drm/imagination/Kconfig | 1 + drivers/gpu/drm/imagination/Makefile | 1 + drivers/gpu/drm/imagination/pvr_device.c | 23 +- drivers/gpu/drm/imagination/pvr_device.h | 22 ++ drivers/gpu/drm/imagination/pvr_drv.c | 20 +- drivers/gpu/drm/imagination/pvr_power.c | 271 +++++++++++++++++++++++ drivers/gpu/drm/imagination/pvr_power.h | 39 ++++ 7 files changed, 374 insertions(+), 3 deletions(-) create mode 100644 drivers/gpu/drm/imagination/pvr_power.c create mode 100644 drivers/gpu/drm/imagination/pvr_power.h diff --git a/drivers/gpu/drm/imagination/Kconfig b/drivers/gpu/drm/imagination/Kconfig index 6f82edf89144..0abd1b9bf3be 100644 --- a/drivers/gpu/drm/imagination/Kconfig +++ b/drivers/gpu/drm/imagination/Kconfig @@ -5,6 +5,7 @@ config DRM_POWERVR tristate "Imagination Technologies PowerVR (Series 6 and later) & IMG Graphics" depends on ARM64 depends on DRM + depends on PM select DRM_GEM_SHMEM_HELPER select DRM_SCHED select DRM_GPUVM diff --git a/drivers/gpu/drm/imagination/Makefile b/drivers/gpu/drm/imagination/Makefile index 678c3dbb4326..d9e00a0db6b2 100644 --- a/drivers/gpu/drm/imagination/Makefile +++ b/drivers/gpu/drm/imagination/Makefile @@ -10,6 +10,7 @@ powervr-y := \ pvr_fw.o \ pvr_gem.o \ pvr_mmu.o \ + pvr_power.o \ pvr_vm.o obj-$(CONFIG_DRM_POWERVR) += powervr.o diff --git a/drivers/gpu/drm/imagination/pvr_device.c b/drivers/gpu/drm/imagination/pvr_device.c index 201ae780494f..e16282325178 100644 --- a/drivers/gpu/drm/imagination/pvr_device.c +++ b/drivers/gpu/drm/imagination/pvr_device.c @@ -5,6 +5,7 @@ #include "pvr_device_info.h" #include "pvr_fw.h" +#include "pvr_power.h" #include "pvr_rogue_cr_defs.h" #include "pvr_vm.h" @@ -361,6 +362,8 @@ pvr_device_gpu_fini(struct pvr_device *pvr_dev) int pvr_device_init(struct pvr_device *pvr_dev) { + struct drm_device *drm_dev = from_pvr_device(pvr_dev); + struct device *dev = drm_dev->dev; int err; /* Enable and initialize clocks required for the device to operate. */ @@ -368,13 +371,29 @@ pvr_device_init(struct pvr_device *pvr_dev) if (err) return err; + /* Explicitly power the GPU so we can access control registers before the FW is booted. */ + err = pm_runtime_resume_and_get(dev); + if (err) + return err; + /* Map the control registers into memory. */ err = pvr_device_reg_init(pvr_dev); if (err) - return err; + goto err_pm_runtime_put; /* Perform GPU-specific initialization steps. */ - return pvr_device_gpu_init(pvr_dev); + err = pvr_device_gpu_init(pvr_dev); + if (err) + goto err_pm_runtime_put; + + pm_runtime_put(dev); + + return 0; + +err_pm_runtime_put: + pm_runtime_put_sync_suspend(dev); + + return err; } /** diff --git a/drivers/gpu/drm/imagination/pvr_device.h b/drivers/gpu/drm/imagination/pvr_device.h index bfc853ffd58f..771ba879f02d 100644 --- a/drivers/gpu/drm/imagination/pvr_device.h +++ b/drivers/gpu/drm/imagination/pvr_device.h @@ -141,6 +141,28 @@ struct pvr_device { * before submitting the next job. */ atomic_t mmu_flush_cache_flags; + + struct { + /** @work: Work item for watchdog callback. */ + struct delayed_work work; + + /** @old_kccb_cmds_executed: KCCB command execution count at last watchdog poll. */ + u32 old_kccb_cmds_executed; + + /** @kccb_stall_count: Number of watchdog polls KCCB has been stalled for. */ + u32 kccb_stall_count; + } watchdog; + + /** + * @lost: %true if the device has been lost. + * + * This variable is set if the device has become irretrievably unavailable, e.g. if the + * firmware processor has stopped responding and can not be revived via a hard reset. + */ + bool lost; + + /** @sched_wq: Workqueue for schedulers. */ + struct workqueue_struct *sched_wq; }; /** diff --git a/drivers/gpu/drm/imagination/pvr_drv.c b/drivers/gpu/drm/imagination/pvr_drv.c index 6d3bc886e1c3..12e136217c45 100644 --- a/drivers/gpu/drm/imagination/pvr_drv.c +++ b/drivers/gpu/drm/imagination/pvr_drv.c @@ -5,6 +5,7 @@ #include "pvr_drv.h" #include "pvr_gem.h" #include "pvr_mmu.h" +#include "pvr_power.h" #include "pvr_rogue_defs.h" #include "pvr_rogue_fwif_client.h" #include "pvr_rogue_fwif_shared.h" @@ -1265,9 +1266,16 @@ pvr_probe(struct platform_device *plat_dev) platform_set_drvdata(plat_dev, drm_dev); + devm_pm_runtime_enable(&plat_dev->dev); + pm_runtime_mark_last_busy(&plat_dev->dev); + + pm_runtime_set_autosuspend_delay(&plat_dev->dev, 50); + pm_runtime_use_autosuspend(&plat_dev->dev); + pvr_watchdog_init(pvr_dev); + err = pvr_device_init(pvr_dev); if (err) - return err; + goto err_watchdog_fini; err = drm_dev_register(drm_dev, 0); if (err) @@ -1278,6 +1286,9 @@ pvr_probe(struct platform_device *plat_dev) err_device_fini: pvr_device_fini(pvr_dev); +err_watchdog_fini: + pvr_watchdog_fini(pvr_dev); + return err; } @@ -1287,8 +1298,10 @@ pvr_remove(struct platform_device *plat_dev) struct drm_device *drm_dev = platform_get_drvdata(plat_dev); struct pvr_device *pvr_dev = to_pvr_device(drm_dev); + pm_runtime_suspend(drm_dev->dev); pvr_device_fini(pvr_dev); drm_dev_unplug(drm_dev); + pvr_watchdog_fini(pvr_dev); return 0; } @@ -1299,11 +1312,16 @@ static const struct of_device_id dt_match[] = { }; MODULE_DEVICE_TABLE(of, dt_match); +static const struct dev_pm_ops pvr_pm_ops = { + RUNTIME_PM_OPS(pvr_power_device_suspend, pvr_power_device_resume, pvr_power_device_idle) +}; + static struct platform_driver pvr_driver = { .probe = pvr_probe, .remove = pvr_remove, .driver = { .name = PVR_DRIVER_NAME, + .pm = &pvr_pm_ops, .of_match_table = dt_match, }, }; diff --git a/drivers/gpu/drm/imagination/pvr_power.c b/drivers/gpu/drm/imagination/pvr_power.c new file mode 100644 index 000000000000..88f14a4d31ab --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_power.c @@ -0,0 +1,271 @@ +// SPDX-License-Identifier: GPL-2.0-only OR MIT +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#include "pvr_device.h" +#include "pvr_fw.h" +#include "pvr_power.h" +#include "pvr_rogue_fwif.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#define POWER_SYNC_TIMEOUT_US (1000000) /* 1s */ + +#define WATCHDOG_TIME_MS (500) + +static int +pvr_power_send_command(struct pvr_device *pvr_dev, struct rogue_fwif_kccb_cmd *pow_cmd) +{ + /* TODO: implement */ + return -ENODEV; +} + +static int +pvr_power_request_idle(struct pvr_device *pvr_dev) +{ + struct rogue_fwif_kccb_cmd pow_cmd; + + /* Send FORCED_IDLE request to FW. */ + pow_cmd.cmd_type = ROGUE_FWIF_KCCB_CMD_POW; + pow_cmd.cmd_data.pow_data.pow_type = ROGUE_FWIF_POW_FORCED_IDLE_REQ; + pow_cmd.cmd_data.pow_data.power_req_data.pow_request_type = ROGUE_FWIF_POWER_FORCE_IDLE; + + return pvr_power_send_command(pvr_dev, &pow_cmd); +} + +static int +pvr_power_request_pwr_off(struct pvr_device *pvr_dev) +{ + struct rogue_fwif_kccb_cmd pow_cmd; + + /* Send POW_OFF request to firmware. */ + pow_cmd.cmd_type = ROGUE_FWIF_KCCB_CMD_POW; + pow_cmd.cmd_data.pow_data.pow_type = ROGUE_FWIF_POW_OFF_REQ; + pow_cmd.cmd_data.pow_data.power_req_data.forced = true; + + return pvr_power_send_command(pvr_dev, &pow_cmd); +} + +static int +pvr_power_fw_disable(struct pvr_device *pvr_dev, bool hard_reset) +{ + if (!hard_reset) { + int err; + + cancel_delayed_work_sync(&pvr_dev->watchdog.work); + + err = pvr_power_request_idle(pvr_dev); + if (err) + return err; + + err = pvr_power_request_pwr_off(pvr_dev); + if (err) + return err; + } + + /* TODO: stop firmware */ + return -ENODEV; +} + +static int +pvr_power_fw_enable(struct pvr_device *pvr_dev) +{ + int err; + + /* TODO: start firmware */ + err = -ENODEV; + if (err) + return err; + + queue_delayed_work(pvr_dev->sched_wq, &pvr_dev->watchdog.work, + msecs_to_jiffies(WATCHDOG_TIME_MS)); + + return 0; +} + +bool +pvr_power_is_idle(struct pvr_device *pvr_dev) +{ + /* TODO: implement */ + return true; +} + +static bool +pvr_watchdog_kccb_stalled(struct pvr_device *pvr_dev) +{ + /* TODO: implement */ + return false; +} + +static void +pvr_watchdog_worker(struct work_struct *work) +{ + struct pvr_device *pvr_dev = container_of(work, struct pvr_device, + watchdog.work.work); + bool stalled; + + if (pvr_dev->lost) + return; + + if (pm_runtime_get_if_in_use(from_pvr_device(pvr_dev)->dev) <= 0) + goto out_requeue; + + stalled = pvr_watchdog_kccb_stalled(pvr_dev); + + if (stalled) { + drm_err(from_pvr_device(pvr_dev), "FW stalled, trying hard reset"); + + pvr_power_reset(pvr_dev, true); + /* Device may be lost at this point. */ + } + + pm_runtime_put(from_pvr_device(pvr_dev)->dev); + +out_requeue: + if (!pvr_dev->lost) { + queue_delayed_work(pvr_dev->sched_wq, &pvr_dev->watchdog.work, + msecs_to_jiffies(WATCHDOG_TIME_MS)); + } +} + +/** + * pvr_watchdog_init() - Initialise watchdog for device + * @pvr_dev: Target PowerVR device. + * + * Returns: + * * 0 on success, or + * * -%ENOMEM on out of memory. + */ +int +pvr_watchdog_init(struct pvr_device *pvr_dev) +{ + INIT_DELAYED_WORK(&pvr_dev->watchdog.work, pvr_watchdog_worker); + + return 0; +} + +int +pvr_power_device_suspend(struct device *dev) +{ + struct platform_device *plat_dev = to_platform_device(dev); + struct drm_device *drm_dev = platform_get_drvdata(plat_dev); + struct pvr_device *pvr_dev = to_pvr_device(drm_dev); + int idx; + + if (!drm_dev_enter(drm_dev, &idx)) + return -EIO; + + clk_disable_unprepare(pvr_dev->mem_clk); + clk_disable_unprepare(pvr_dev->sys_clk); + clk_disable_unprepare(pvr_dev->core_clk); + + drm_dev_exit(idx); + + return 0; +} + +int +pvr_power_device_resume(struct device *dev) +{ + struct platform_device *plat_dev = to_platform_device(dev); + struct drm_device *drm_dev = platform_get_drvdata(plat_dev); + struct pvr_device *pvr_dev = to_pvr_device(drm_dev); + int idx; + int err; + + if (!drm_dev_enter(drm_dev, &idx)) + return -EIO; + + err = clk_prepare_enable(pvr_dev->core_clk); + if (err) + goto err_drm_dev_exit; + + err = clk_prepare_enable(pvr_dev->sys_clk); + if (err) + goto err_core_clk_disable; + + err = clk_prepare_enable(pvr_dev->mem_clk); + if (err) + goto err_sys_clk_disable; + + drm_dev_exit(idx); + + return 0; + +err_sys_clk_disable: + clk_disable_unprepare(pvr_dev->sys_clk); + +err_core_clk_disable: + clk_disable_unprepare(pvr_dev->core_clk); + +err_drm_dev_exit: + drm_dev_exit(idx); + + return err; +} + +int +pvr_power_device_idle(struct device *dev) +{ + struct platform_device *plat_dev = to_platform_device(dev); + struct drm_device *drm_dev = platform_get_drvdata(plat_dev); + struct pvr_device *pvr_dev = to_pvr_device(drm_dev); + + return pvr_power_is_idle(pvr_dev) ? 0 : -EBUSY; +} + +/** + * pvr_power_reset() - Reset the GPU + * @pvr_dev: Device pointer + * @hard_reset: %true for hard reset, %false for soft reset + * + * If @hard_reset is %false and the FW processor fails to respond during the reset process, this + * function will attempt a hard reset. + * + * If a hard reset fails then the GPU device is reported as lost. + * + * Returns: + * * 0 on success, or + * * Any error code returned by pvr_power_get, pvr_power_fw_disable or pvr_power_fw_enable(). + */ +int +pvr_power_reset(struct pvr_device *pvr_dev, bool hard_reset) +{ + /* TODO: Implement hard reset. */ + int err; + + /* + * Take a power reference during the reset. This should prevent any interference with the + * power state during reset. + */ + WARN_ON(pvr_power_get(pvr_dev)); + + err = pvr_power_fw_disable(pvr_dev, false); + if (err) + goto err_power_put; + + err = pvr_power_fw_enable(pvr_dev); + +err_power_put: + pvr_power_put(pvr_dev); + + return err; +} + +/** + * pvr_watchdog_fini() - Shutdown watchdog for device + * @pvr_dev: Target PowerVR device. + */ +void +pvr_watchdog_fini(struct pvr_device *pvr_dev) +{ + cancel_delayed_work_sync(&pvr_dev->watchdog.work); +} diff --git a/drivers/gpu/drm/imagination/pvr_power.h b/drivers/gpu/drm/imagination/pvr_power.h new file mode 100644 index 000000000000..360980f454d7 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_power.h @@ -0,0 +1,39 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_POWER_H +#define PVR_POWER_H + +#include "pvr_device.h" + +#include +#include + +int pvr_watchdog_init(struct pvr_device *pvr_dev); +void pvr_watchdog_fini(struct pvr_device *pvr_dev); + +bool pvr_power_is_idle(struct pvr_device *pvr_dev); + +int pvr_power_device_suspend(struct device *dev); +int pvr_power_device_resume(struct device *dev); +int pvr_power_device_idle(struct device *dev); + +int pvr_power_reset(struct pvr_device *pvr_dev, bool hard_reset); + +static __always_inline int +pvr_power_get(struct pvr_device *pvr_dev) +{ + struct drm_device *drm_dev = from_pvr_device(pvr_dev); + + return pm_runtime_resume_and_get(drm_dev->dev); +} + +static __always_inline int +pvr_power_put(struct pvr_device *pvr_dev) +{ + struct drm_device *drm_dev = from_pvr_device(pvr_dev); + + return pm_runtime_put(drm_dev->dev); +} + +#endif /* PVR_POWER_H */ From patchwork Wed Nov 22 16:34:36 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Donald Robson X-Patchwork-Id: 746147 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=imgtec.com header.i=@imgtec.com header.b="ehx7M3av"; dkim=pass (1024-bit key) header.d=IMGTecCRM.onmicrosoft.com header.i=@IMGTecCRM.onmicrosoft.com header.b="lYBLSmqL" Received: from mx07-00376f01.pphosted.com (mx07-00376f01.pphosted.com [185.132.180.163]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 49D5BD72; Wed, 22 Nov 2023 08:36:01 -0800 (PST) Received: from pps.filterd (m0168889.ppops.net [127.0.0.1]) by mx07-00376f01.pphosted.com (8.17.1.24/8.17.1.24) with ESMTP id 3AMCaKJG020106; Wed, 22 Nov 2023 16:35:29 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=imgtec.com; h= from:to:cc:subject:date:message-id:in-reply-to:references :content-transfer-encoding:content-type:mime-version; s= dk201812; bh=GxmfN8rVvd3YfGfz59gK/3JJGHhPU1f5MEvCTSmel8o=; b=ehx 7M3avAvtfnqHOe6XsO3/+Y+93XICn6SIIfQDQiw0sVcfEFBKhNEFmt7+AZdldBTq r8dx2FiWvih60q5OdvoZzUAaiThEPXzh7mKtkbhUJwWQOyHbJ7AfVfqLESnfJcFk hw2lQktcdLmA86XWELj0YvsuseYRkkkq3Zwo6L4+FxbR99/uk31tVYeiN5mkj/Q1 ma9p2E6Cizd38TTVvEe3bIfthvHCQvDY4cgS1tJUfEKMGacPns1j3Gzww7AaR8ys m3UQd75J3F78aBNYT4rnym1oiZdlTAfVpfTxwwEDSyzFibk7d147tN2vmgFZ3DNq xp9dSiemPZHRWw/OHrQ== Received: from hhmail04.hh.imgtec.org ([217.156.249.195]) by mx07-00376f01.pphosted.com (PPS) with ESMTPS id 3ug99g9m28-2 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128 verify=NOT); Wed, 22 Nov 2023 16:35:28 +0000 (GMT) Received: from HHMAIL04.hh.imgtec.org (10.100.10.119) by HHMAIL04.hh.imgtec.org (10.100.10.119) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34; Wed, 22 Nov 2023 16:35:27 +0000 Received: from GBR01-LO4-obe.outbound.protection.outlook.com (104.47.85.105) by email.imgtec.com (10.100.10.121) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34 via Frontend Transport; Wed, 22 Nov 2023 16:35:27 +0000 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=H12vd9WxMKLCDAcy7R6V1W32ibzDmRj1ZcUYB2NzfODQWS3PuIEuyQncgBK7MbTVDOXkbSf+IapAybCvgDsp6FmXtE+ncCqEuv20zJ3dUzV95m3ypJ0RYpBhOOnK+RrboM2Y+FVsPsFEb/h30jTC2e6xclzUD83Z52gRZFLxTj8pgtIm3OSu19n40UURPcGlWhZ+7jYAkjxKf5y5x3V7tzSPJlf2agQ/57Rb5MMGqSLaUpl1UKp2msm/cX2P4ym4pc0G04nacvJEmtCXwwr1Fz7XRJSCjveZRNfY7r85THtvgQeZgmWPalLyvxyRZVXtqydpE5blkmthexKD7i9vkg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=GxmfN8rVvd3YfGfz59gK/3JJGHhPU1f5MEvCTSmel8o=; b=JERmKX9QndMMrUYDNUtM+dVwElJSWn64wyapajtil8fX8hjXDghVznZ+9RXkVfeowGvHeQU4FHUVtXz+eQd864BIteTYBWgLRywHJXOwu2qwEbrwx0oeAhgbGqL4bNOgAnfikVuW4/CcaDHB052dJr2GBsT82tWouCVUj5avpCaMa7azkdQgi6qZFvbHWJhrfDm9eWFqZHhM+sIET8EE66jyjdQgLuEXVjzmSdzppfPyzwR9GUoe0ecZES2sm0tXReo/i3pc2xok8brPaNswHguDj+pwDWSFhu6m9+0csr+hnbFCXXQBWeVFKYQ1dkOfCmbEABKC1KMB1LIMQiq6Tg== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=imgtec.com; dmarc=pass action=none header.from=imgtec.com; dkim=pass header.d=imgtec.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=IMGTecCRM.onmicrosoft.com; s=selector2-IMGTecCRM-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=GxmfN8rVvd3YfGfz59gK/3JJGHhPU1f5MEvCTSmel8o=; b=lYBLSmqLxzPQNaGQnsC1bP6dOfLnA4EDkUb3Kgq5NvH27QzLHRC+YkNp+PYuVmonwFt31kAe+oO694/OW7FnHeMCn8skEEnsqI3cnPBuwCo7VvpYxK4bo85nGQGE/M0iu0ZQK4zmbpcMjSJnG5ypSRHo4lNDnJ1/1Fls8vprzEU= Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) by CWLP265MB6864.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1fe::14) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.7025.19; Wed, 22 Nov 2023 16:35:24 +0000 Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34]) by CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34%3]) with mapi id 15.20.7025.017; Wed, 22 Nov 2023 16:35:24 +0000 From: Donald Robson To: CC: , , , , , , , , , , , , , , , , , , , , , , Sarah Walker Subject: [PATCH v9 15/20] drm/imagination: Implement free list and HWRT create and destroy ioctls Date: Wed, 22 Nov 2023 16:34:36 +0000 Message-Id: <919358c5887a7628da588c455a5bb7e3ea4b47ae.1700668843.git.donald.robson@imgtec.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: References: X-ClientProxiedBy: LO4P265CA0273.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:37a::10) To CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) Precedence: bulk X-Mailing-List: devicetree@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: CWLP265MB5770:EE_|CWLP265MB6864:EE_ X-MS-Office365-Filtering-Correlation-Id: 56d4ccd3-67c9-4779-8cbf-08dbeb79072a X-MS-Exchange-SenderADCheck: 1 X-MS-Exchange-AntiSpam-Relay: 0 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-Forefront-Antispam-Report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:; IPV:NLI; SFV:NSPM; H:CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM; PTR:; CAT:NONE; SFS:(13230031)(136003)(366004)(376002)(346002)(39850400004)(396003)(230922051799003)(64100799003)(451199024)(186009)(1800799012)(52116002)(5660300002)(38350700005)(6666004)(6506007)(6512007)(2616005)(8936002)(8676002)(4326008)(83380400001)(2906002)(7416002)(478600001)(30864003)(66476007)(6916009)(66556008)(107886003)(316002)(26005)(66946007)(6486002)(44832011)(38100700002)(41300700001)(36756003)(86362001)(559001)(579004); DIR:OUT; SFP:1102; X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: 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 X-MS-Exchange-CrossTenant-Network-Message-Id: 56d4ccd3-67c9-4779-8cbf-08dbeb79072a X-MS-Exchange-CrossTenant-AuthSource: CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 22 Nov 2023 16:35:24.8132 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 0d5fd8bb-e8c2-4e0a-8dd5-2c264f7140fe X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: +Is7kBCnJ5P9IgBpQnB4Yac3eQTAWe4oUuxfL1KSIwxtsHVyZrc0+HqsEA0phdRNMVF8B0oUtYA9Zo1YnEOTSs0FsqALyD0u1Naa3YdtaQw= X-MS-Exchange-Transport-CrossTenantHeadersStamped: CWLP265MB6864 X-OriginatorOrg: imgtec.com X-EXCLAIMER-MD-CONFIG: 15a78312-3e47-46eb-9010-2e54d84a9631 X-Proofpoint-ORIG-GUID: duFyLtQKil9fVgmZtD8ypigIe6iEp4dS X-Proofpoint-GUID: duFyLtQKil9fVgmZtD8ypigIe6iEp4dS From: Sarah Walker Implement ioctls to create and destroy free lists and HWRT datasets. Free lists are used for GPU-side memory allocation during geometry processing. HWRT datasets are the FW-side structures representing render targets. Changes since v8: - Corrected license identifiers Changes since v6: - Fix out-of-bounds shift in get_cr_multisamplectl_val() Changes since v4: - Remove use of drm_gem_shmem_get_pages() Changes since v3: - Support free list grow requests from FW - Use drm_dev_{enter,exit} Co-developed-by: Boris Brezillon Signed-off-by: Boris Brezillon Co-developed-by: Donald Robson Signed-off-by: Donald Robson Signed-off-by: Sarah Walker --- drivers/gpu/drm/imagination/Makefile | 2 + drivers/gpu/drm/imagination/pvr_ccb.c | 10 + drivers/gpu/drm/imagination/pvr_device.h | 24 + drivers/gpu/drm/imagination/pvr_drv.c | 112 +++- drivers/gpu/drm/imagination/pvr_free_list.c | 625 ++++++++++++++++++++ drivers/gpu/drm/imagination/pvr_free_list.h | 195 ++++++ drivers/gpu/drm/imagination/pvr_hwrt.c | 549 +++++++++++++++++ drivers/gpu/drm/imagination/pvr_hwrt.h | 165 ++++++ 8 files changed, 1678 insertions(+), 4 deletions(-) create mode 100644 drivers/gpu/drm/imagination/pvr_free_list.c create mode 100644 drivers/gpu/drm/imagination/pvr_free_list.h create mode 100644 drivers/gpu/drm/imagination/pvr_hwrt.c create mode 100644 drivers/gpu/drm/imagination/pvr_hwrt.h diff --git a/drivers/gpu/drm/imagination/Makefile b/drivers/gpu/drm/imagination/Makefile index 71dc36cc6b9d..2bd501018d5d 100644 --- a/drivers/gpu/drm/imagination/Makefile +++ b/drivers/gpu/drm/imagination/Makefile @@ -8,12 +8,14 @@ powervr-y := \ pvr_device.o \ pvr_device_info.o \ pvr_drv.o \ + pvr_free_list.o \ pvr_fw.o \ pvr_fw_meta.o \ pvr_fw_mips.o \ pvr_fw_startstop.o \ pvr_fw_trace.o \ pvr_gem.o \ + pvr_hwrt.o \ pvr_mmu.o \ pvr_power.o \ pvr_vm.o \ diff --git a/drivers/gpu/drm/imagination/pvr_ccb.c b/drivers/gpu/drm/imagination/pvr_ccb.c index 48f06f58f3f1..4deeac7ed40a 100644 --- a/drivers/gpu/drm/imagination/pvr_ccb.c +++ b/drivers/gpu/drm/imagination/pvr_ccb.c @@ -4,6 +4,7 @@ #include "pvr_ccb.h" #include "pvr_device.h" #include "pvr_drv.h" +#include "pvr_free_list.h" #include "pvr_fw.h" #include "pvr_gem.h" #include "pvr_power.h" @@ -139,6 +140,15 @@ process_fwccb_command(struct pvr_device *pvr_dev, struct rogue_fwif_fwccb_cmd *c pvr_power_reset(pvr_dev, false); break; + case ROGUE_FWIF_FWCCB_CMD_FREELISTS_RECONSTRUCTION: + pvr_free_list_process_reconstruct_req(pvr_dev, + &cmd->cmd_data.cmd_freelists_reconstruction); + break; + + case ROGUE_FWIF_FWCCB_CMD_FREELIST_GROW: + pvr_free_list_process_grow_req(pvr_dev, &cmd->cmd_data.cmd_free_list_gs); + break; + default: drm_info(from_pvr_device(pvr_dev), "Received unknown FWCCB command %x\n", cmd->cmd_type); diff --git a/drivers/gpu/drm/imagination/pvr_device.h b/drivers/gpu/drm/imagination/pvr_device.h index 8853249f4884..f5b82b793566 100644 --- a/drivers/gpu/drm/imagination/pvr_device.h +++ b/drivers/gpu/drm/imagination/pvr_device.h @@ -152,6 +152,14 @@ struct pvr_device { */ atomic_t mmu_flush_cache_flags; + /** + * @free_list_ids: Array of free lists belonging to this device. Array members + * are of type "struct pvr_free_list *". + * + * This array is used to allocate IDs used by the firmware. + */ + struct xarray free_list_ids; + struct { /** @work: Work item for watchdog callback. */ struct delayed_work work; @@ -247,6 +255,22 @@ struct pvr_file { */ struct pvr_device *pvr_dev; + /** + * @free_list_handles: Array of free lists belonging to this file. Array + * members are of type "struct pvr_free_list *". + * + * This array is used to allocate handles returned to userspace. + */ + struct xarray free_list_handles; + + /** + * @hwrt_handles: Array of HWRT datasets belonging to this file. Array + * members are of type "struct pvr_hwrt_dataset *". + * + * This array is used to allocate handles returned to userspace. + */ + struct xarray hwrt_handles; + /** * @vm_ctx_handles: Array of VM contexts belonging to this file. Array * members are of type "struct pvr_vm_context *". diff --git a/drivers/gpu/drm/imagination/pvr_drv.c b/drivers/gpu/drm/imagination/pvr_drv.c index 33b38c0d79c8..f485e2cc60f9 100644 --- a/drivers/gpu/drm/imagination/pvr_drv.c +++ b/drivers/gpu/drm/imagination/pvr_drv.c @@ -3,7 +3,9 @@ #include "pvr_device.h" #include "pvr_drv.h" +#include "pvr_free_list.h" #include "pvr_gem.h" +#include "pvr_hwrt.h" #include "pvr_mmu.h" #include "pvr_power.h" #include "pvr_rogue_defs.h" @@ -711,7 +713,41 @@ static int pvr_ioctl_create_free_list(struct drm_device *drm_dev, void *raw_args, struct drm_file *file) { - return -ENOTTY; + struct drm_pvr_ioctl_create_free_list_args *args = raw_args; + struct pvr_file *pvr_file = to_pvr_file(file); + struct pvr_free_list *free_list; + int idx; + int err; + + if (!drm_dev_enter(drm_dev, &idx)) + return -EIO; + + free_list = pvr_free_list_create(pvr_file, args); + if (IS_ERR(free_list)) { + err = PTR_ERR(free_list); + goto err_drm_dev_exit; + } + + /* Allocate object handle for userspace. */ + err = xa_alloc(&pvr_file->free_list_handles, + &args->handle, + free_list, + xa_limit_32b, + GFP_KERNEL); + if (err < 0) + goto err_cleanup; + + drm_dev_exit(idx); + + return 0; + +err_cleanup: + pvr_free_list_put(free_list); + +err_drm_dev_exit: + drm_dev_exit(idx); + + return err; } /** @@ -731,7 +767,19 @@ static int pvr_ioctl_destroy_free_list(struct drm_device *drm_dev, void *raw_args, struct drm_file *file) { - return -ENOTTY; + struct drm_pvr_ioctl_destroy_free_list_args *args = raw_args; + struct pvr_file *pvr_file = to_pvr_file(file); + struct pvr_free_list *free_list; + + if (args->_padding_4) + return -EINVAL; + + free_list = xa_erase(&pvr_file->free_list_handles, args->handle); + if (!free_list) + return -EINVAL; + + pvr_free_list_put(free_list); + return 0; } /** @@ -751,7 +799,41 @@ static int pvr_ioctl_create_hwrt_dataset(struct drm_device *drm_dev, void *raw_args, struct drm_file *file) { - return -ENOTTY; + struct drm_pvr_ioctl_create_hwrt_dataset_args *args = raw_args; + struct pvr_file *pvr_file = to_pvr_file(file); + struct pvr_hwrt_dataset *hwrt; + int idx; + int err; + + if (!drm_dev_enter(drm_dev, &idx)) + return -EIO; + + hwrt = pvr_hwrt_dataset_create(pvr_file, args); + if (IS_ERR(hwrt)) { + err = PTR_ERR(hwrt); + goto err_drm_dev_exit; + } + + /* Allocate object handle for userspace. */ + err = xa_alloc(&pvr_file->hwrt_handles, + &args->handle, + hwrt, + xa_limit_32b, + GFP_KERNEL); + if (err < 0) + goto err_cleanup; + + drm_dev_exit(idx); + + return 0; + +err_cleanup: + pvr_hwrt_dataset_put(hwrt); + +err_drm_dev_exit: + drm_dev_exit(idx); + + return err; } /** @@ -771,7 +853,19 @@ static int pvr_ioctl_destroy_hwrt_dataset(struct drm_device *drm_dev, void *raw_args, struct drm_file *file) { - return -ENOTTY; + struct drm_pvr_ioctl_destroy_hwrt_dataset_args *args = raw_args; + struct pvr_file *pvr_file = to_pvr_file(file); + struct pvr_hwrt_dataset *hwrt; + + if (args->_padding_4) + return -EINVAL; + + hwrt = xa_erase(&pvr_file->hwrt_handles, args->handle); + if (!hwrt) + return -EINVAL; + + pvr_hwrt_dataset_put(hwrt); + return 0; } /** @@ -1195,6 +1289,8 @@ pvr_drm_driver_open(struct drm_device *drm_dev, struct drm_file *file) */ pvr_file->pvr_dev = pvr_dev; + xa_init_flags(&pvr_file->free_list_handles, XA_FLAGS_ALLOC1); + xa_init_flags(&pvr_file->hwrt_handles, XA_FLAGS_ALLOC1); xa_init_flags(&pvr_file->vm_ctx_handles, XA_FLAGS_ALLOC1); /* @@ -1223,6 +1319,8 @@ pvr_drm_driver_postclose(__always_unused struct drm_device *drm_dev, struct pvr_file *pvr_file = to_pvr_file(file); /* Drop references on any remaining objects. */ + pvr_destroy_free_lists_for_file(pvr_file); + pvr_destroy_hwrt_datasets_for_file(pvr_file); pvr_destroy_vm_contexts_for_file(pvr_file); kfree(pvr_file); @@ -1281,6 +1379,8 @@ pvr_probe(struct platform_device *plat_dev) if (err) goto err_device_fini; + xa_init_flags(&pvr_dev->free_list_ids, XA_FLAGS_ALLOC1); + return 0; err_device_fini: @@ -1298,6 +1398,10 @@ pvr_remove(struct platform_device *plat_dev) struct drm_device *drm_dev = platform_get_drvdata(plat_dev); struct pvr_device *pvr_dev = to_pvr_device(drm_dev); + WARN_ON(!xa_empty(&pvr_dev->free_list_ids)); + + xa_destroy(&pvr_dev->free_list_ids); + pm_runtime_suspend(drm_dev->dev); pvr_device_fini(pvr_dev); drm_dev_unplug(drm_dev); diff --git a/drivers/gpu/drm/imagination/pvr_free_list.c b/drivers/gpu/drm/imagination/pvr_free_list.c new file mode 100644 index 000000000000..c61fd417edcb --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_free_list.c @@ -0,0 +1,625 @@ +// SPDX-License-Identifier: GPL-2.0-only OR MIT +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#include "pvr_free_list.h" +#include "pvr_gem.h" +#include "pvr_hwrt.h" +#include "pvr_rogue_fwif.h" +#include "pvr_vm.h" + +#include +#include +#include +#include + +#define FREE_LIST_ENTRY_SIZE sizeof(u32) + +#define FREE_LIST_ALIGNMENT \ + ((ROGUE_BIF_PM_FREELIST_BASE_ADDR_ALIGNSIZE / FREE_LIST_ENTRY_SIZE) - 1) + +#define FREE_LIST_MIN_PAGES 50 +#define FREE_LIST_MIN_PAGES_BRN66011 40 +#define FREE_LIST_MIN_PAGES_ROGUEXE 25 + +/** + * pvr_get_free_list_min_pages() - Get minimum free list size for this device + * @pvr_dev: Device pointer. + * + * Returns: + * * Minimum free list size, in PM physical pages. + */ +u32 +pvr_get_free_list_min_pages(struct pvr_device *pvr_dev) +{ + u32 value; + + if (PVR_HAS_FEATURE(pvr_dev, roguexe)) { + if (PVR_HAS_QUIRK(pvr_dev, 66011)) + value = FREE_LIST_MIN_PAGES_BRN66011; + else + value = FREE_LIST_MIN_PAGES_ROGUEXE; + } else { + value = FREE_LIST_MIN_PAGES; + } + + return value; +} + +static int +free_list_create_kernel_structure(struct pvr_file *pvr_file, + struct drm_pvr_ioctl_create_free_list_args *args, + struct pvr_free_list *free_list) +{ + struct pvr_gem_object *free_list_obj; + struct pvr_vm_context *vm_ctx; + u64 free_list_size; + int err; + + if (args->grow_threshold > 100 || + args->initial_num_pages > args->max_num_pages || + args->grow_num_pages > args->max_num_pages || + args->max_num_pages == 0 || + (args->initial_num_pages < args->max_num_pages && !args->grow_num_pages) || + (args->initial_num_pages == args->max_num_pages && args->grow_num_pages)) + return -EINVAL; + + if ((args->initial_num_pages & FREE_LIST_ALIGNMENT) || + (args->max_num_pages & FREE_LIST_ALIGNMENT) || + (args->grow_num_pages & FREE_LIST_ALIGNMENT)) + return -EINVAL; + + vm_ctx = pvr_vm_context_lookup(pvr_file, args->vm_context_handle); + if (!vm_ctx) + return -EINVAL; + + free_list_obj = pvr_vm_find_gem_object(vm_ctx, args->free_list_gpu_addr, + NULL, &free_list_size); + if (!free_list_obj) { + err = -EINVAL; + goto err_put_vm_context; + } + + if ((free_list_obj->flags & DRM_PVR_BO_ALLOW_CPU_USERSPACE_ACCESS) || + !(free_list_obj->flags & DRM_PVR_BO_PM_FW_PROTECT) || + free_list_size < (args->max_num_pages * FREE_LIST_ENTRY_SIZE)) { + err = -EINVAL; + goto err_put_free_list_obj; + } + + free_list->pvr_dev = pvr_file->pvr_dev; + free_list->current_pages = 0; + free_list->max_pages = args->max_num_pages; + free_list->grow_pages = args->grow_num_pages; + free_list->grow_threshold = args->grow_threshold; + free_list->obj = free_list_obj; + free_list->free_list_gpu_addr = args->free_list_gpu_addr; + free_list->initial_num_pages = args->initial_num_pages; + + pvr_vm_context_put(vm_ctx); + + return 0; + +err_put_free_list_obj: + pvr_gem_object_put(free_list_obj); + +err_put_vm_context: + pvr_vm_context_put(vm_ctx); + + return err; +} + +static void +free_list_destroy_kernel_structure(struct pvr_free_list *free_list) +{ + WARN_ON(!list_empty(&free_list->hwrt_list)); + + pvr_gem_object_put(free_list->obj); +} + +/** + * calculate_free_list_ready_pages_locked() - Function to work out the number of free + * list pages to reserve for growing within + * the FW without having to wait for the + * host to progress a grow request + * @free_list: Pointer to free list. + * @pages: Total pages currently in free list. + * + * If the threshold or grow size means less than the alignment size (4 pages on + * Rogue), then the feature is not used. + * + * Caller must hold &free_list->lock. + * + * Return: number of pages to reserve. + */ +static u32 +calculate_free_list_ready_pages_locked(struct pvr_free_list *free_list, u32 pages) +{ + u32 ready_pages; + + lockdep_assert_held(&free_list->lock); + + ready_pages = ((pages * free_list->grow_threshold) / 100); + + /* The number of pages must be less than the grow size. */ + ready_pages = min(ready_pages, free_list->grow_pages); + + /* + * The number of pages must be a multiple of the free list align size. + */ + ready_pages &= ~FREE_LIST_ALIGNMENT; + + return ready_pages; +} + +static u32 +calculate_free_list_ready_pages(struct pvr_free_list *free_list, u32 pages) +{ + u32 ret; + + mutex_lock(&free_list->lock); + + ret = calculate_free_list_ready_pages_locked(free_list, pages); + + mutex_unlock(&free_list->lock); + + return ret; +} + +static void +free_list_fw_init(void *cpu_ptr, void *priv) +{ + struct rogue_fwif_freelist *fw_data = cpu_ptr; + struct pvr_free_list *free_list = priv; + u32 ready_pages; + + /* Fill out FW structure */ + ready_pages = calculate_free_list_ready_pages(free_list, + free_list->initial_num_pages); + + fw_data->max_pages = free_list->max_pages; + fw_data->current_pages = free_list->initial_num_pages - ready_pages; + fw_data->grow_pages = free_list->grow_pages; + fw_data->ready_pages = ready_pages; + fw_data->freelist_id = free_list->fw_id; + fw_data->grow_pending = false; + fw_data->current_stack_top = fw_data->current_pages - 1; + fw_data->freelist_dev_addr = free_list->free_list_gpu_addr; + fw_data->current_dev_addr = (fw_data->freelist_dev_addr + + ((fw_data->max_pages - fw_data->current_pages) * + FREE_LIST_ENTRY_SIZE)) & + ~((u64)ROGUE_BIF_PM_FREELIST_BASE_ADDR_ALIGNSIZE - 1); +} + +static int +free_list_create_fw_structure(struct pvr_file *pvr_file, + struct drm_pvr_ioctl_create_free_list_args *args, + struct pvr_free_list *free_list) +{ + struct pvr_device *pvr_dev = pvr_file->pvr_dev; + + /* + * Create and map the FW structure so we can initialise it. This is not + * accessed on the CPU side post-initialisation so the mapping lifetime + * is only for this function. + */ + free_list->fw_data = pvr_fw_object_create_and_map(pvr_dev, sizeof(*free_list->fw_data), + PVR_BO_FW_FLAGS_DEVICE_UNCACHED, + free_list_fw_init, free_list, + &free_list->fw_obj); + if (IS_ERR(free_list->fw_data)) + return PTR_ERR(free_list->fw_data); + + return 0; +} + +static void +free_list_destroy_fw_structure(struct pvr_free_list *free_list) +{ + pvr_fw_object_unmap_and_destroy(free_list->fw_obj); +} + +static int +pvr_free_list_insert_pages_locked(struct pvr_free_list *free_list, + struct sg_table *sgt, u32 offset, u32 num_pages) +{ + struct sg_dma_page_iter dma_iter; + u32 *page_list; + + lockdep_assert_held(&free_list->lock); + + page_list = pvr_gem_object_vmap(free_list->obj); + if (IS_ERR(page_list)) + return PTR_ERR(page_list); + + offset /= FREE_LIST_ENTRY_SIZE; + /* clang-format off */ + for_each_sgtable_dma_page(sgt, &dma_iter, 0) { + dma_addr_t dma_addr = sg_page_iter_dma_address(&dma_iter); + u64 dma_pfn = dma_addr >> + ROGUE_BIF_PM_PHYSICAL_PAGE_ALIGNSHIFT; + u32 dma_addr_offset; + + BUILD_BUG_ON(ROGUE_BIF_PM_PHYSICAL_PAGE_SIZE > PAGE_SIZE); + + for (dma_addr_offset = 0; dma_addr_offset < PAGE_SIZE; + dma_addr_offset += ROGUE_BIF_PM_PHYSICAL_PAGE_SIZE) { + WARN_ON_ONCE(dma_pfn >> 32); + + page_list[offset++] = (u32)dma_pfn; + dma_pfn++; + + num_pages--; + if (!num_pages) + break; + } + + if (!num_pages) + break; + }; + /* clang-format on */ + + /* Make sure our free_list update is flushed. */ + wmb(); + + pvr_gem_object_vunmap(free_list->obj); + + return 0; +} + +static int +pvr_free_list_insert_node_locked(struct pvr_free_list_node *free_list_node) +{ + struct pvr_free_list *free_list = free_list_node->free_list; + struct sg_table *sgt; + u32 start_page; + u32 offset; + int err; + + lockdep_assert_held(&free_list->lock); + + start_page = free_list->max_pages - free_list->current_pages - + free_list_node->num_pages; + offset = (start_page * FREE_LIST_ENTRY_SIZE) & + ~((u64)ROGUE_BIF_PM_FREELIST_BASE_ADDR_ALIGNSIZE - 1); + + sgt = drm_gem_shmem_get_pages_sgt(&free_list_node->mem_obj->base); + if (WARN_ON(IS_ERR(sgt))) + return PTR_ERR(sgt); + + err = pvr_free_list_insert_pages_locked(free_list, sgt, + offset, free_list_node->num_pages); + if (!err) + free_list->current_pages += free_list_node->num_pages; + + return err; +} + +static int +pvr_free_list_grow(struct pvr_free_list *free_list, u32 num_pages) +{ + struct pvr_device *pvr_dev = free_list->pvr_dev; + struct pvr_free_list_node *free_list_node; + int err; + + mutex_lock(&free_list->lock); + + if (num_pages & FREE_LIST_ALIGNMENT) { + err = -EINVAL; + goto err_unlock; + } + + free_list_node = kzalloc(sizeof(*free_list_node), GFP_KERNEL); + if (!free_list_node) { + err = -ENOMEM; + goto err_unlock; + } + + free_list_node->num_pages = num_pages; + free_list_node->free_list = free_list; + + free_list_node->mem_obj = pvr_gem_object_create(pvr_dev, + num_pages << + ROGUE_BIF_PM_PHYSICAL_PAGE_ALIGNSHIFT, + PVR_BO_FW_FLAGS_DEVICE_CACHED); + if (IS_ERR(free_list_node->mem_obj)) { + err = PTR_ERR(free_list_node->mem_obj); + goto err_free; + } + + err = pvr_free_list_insert_node_locked(free_list_node); + if (err) + goto err_destroy_gem_object; + + list_add_tail(&free_list_node->node, &free_list->mem_block_list); + + /* + * Reserve a number ready pages to allow the FW to process OOM quickly + * and asynchronously request a grow. + */ + free_list->ready_pages = + calculate_free_list_ready_pages_locked(free_list, + free_list->current_pages); + free_list->current_pages -= free_list->ready_pages; + + mutex_unlock(&free_list->lock); + + return 0; + +err_destroy_gem_object: + pvr_gem_object_put(free_list_node->mem_obj); + +err_free: + kfree(free_list_node); + +err_unlock: + mutex_unlock(&free_list->lock); + + return err; +} + +void pvr_free_list_process_grow_req(struct pvr_device *pvr_dev, + struct rogue_fwif_fwccb_cmd_freelist_gs_data *req) +{ + struct pvr_free_list *free_list = pvr_free_list_lookup_id(pvr_dev, req->freelist_id); + struct rogue_fwif_kccb_cmd resp_cmd = { + .cmd_type = ROGUE_FWIF_KCCB_CMD_FREELIST_GROW_UPDATE, + }; + struct rogue_fwif_freelist_gs_data *resp = &resp_cmd.cmd_data.free_list_gs_data; + u32 grow_pages = 0; + + /* If we don't have a freelist registered for this ID, we can't do much. */ + if (WARN_ON(!free_list)) + return; + + /* Since the FW made the request, it has already consumed the ready pages, + * update the host struct. + */ + free_list->current_pages += free_list->ready_pages; + free_list->ready_pages = 0; + + /* If the grow succeeds, update the grow_pages argument. */ + if (!pvr_free_list_grow(free_list, free_list->grow_pages)) + grow_pages = free_list->grow_pages; + + /* Now prepare the response and send it back to the FW. */ + pvr_fw_object_get_fw_addr(free_list->fw_obj, &resp->freelist_fw_addr); + resp->delta_pages = grow_pages; + resp->new_pages = free_list->current_pages + free_list->ready_pages; + resp->ready_pages = free_list->ready_pages; + pvr_free_list_put(free_list); + + WARN_ON(pvr_kccb_send_cmd(pvr_dev, &resp_cmd, NULL)); +} + +static void +pvr_free_list_free_node(struct pvr_free_list_node *free_list_node) +{ + pvr_gem_object_put(free_list_node->mem_obj); + + kfree(free_list_node); +} + +/** + * pvr_free_list_create() - Create a new free list and return an object pointer + * @pvr_file: Pointer to pvr_file structure. + * @args: Creation arguments from userspace. + * + * Return: + * * Pointer to new free_list, or + * * ERR_PTR(-%ENOMEM) on out of memory. + */ +struct pvr_free_list * +pvr_free_list_create(struct pvr_file *pvr_file, + struct drm_pvr_ioctl_create_free_list_args *args) +{ + struct pvr_free_list *free_list; + int err; + + /* Create and fill out the kernel structure */ + free_list = kzalloc(sizeof(*free_list), GFP_KERNEL); + + if (!free_list) + return ERR_PTR(-ENOMEM); + + kref_init(&free_list->ref_count); + INIT_LIST_HEAD(&free_list->mem_block_list); + INIT_LIST_HEAD(&free_list->hwrt_list); + mutex_init(&free_list->lock); + + err = free_list_create_kernel_structure(pvr_file, args, free_list); + if (err < 0) + goto err_free; + + /* Allocate global object ID for firmware. */ + err = xa_alloc(&pvr_file->pvr_dev->free_list_ids, + &free_list->fw_id, + free_list, + xa_limit_32b, + GFP_KERNEL); + if (err) + goto err_destroy_kernel_structure; + + err = free_list_create_fw_structure(pvr_file, args, free_list); + if (err < 0) + goto err_free_fw_id; + + err = pvr_free_list_grow(free_list, args->initial_num_pages); + if (err < 0) + goto err_fw_struct_cleanup; + + return free_list; + +err_fw_struct_cleanup: + WARN_ON(pvr_fw_structure_cleanup(free_list->pvr_dev, + ROGUE_FWIF_CLEANUP_FREELIST, + free_list->fw_obj, 0)); + +err_free_fw_id: + xa_erase(&free_list->pvr_dev->free_list_ids, free_list->fw_id); + +err_destroy_kernel_structure: + free_list_destroy_kernel_structure(free_list); + +err_free: + mutex_destroy(&free_list->lock); + kfree(free_list); + + return ERR_PTR(err); +} + +static void +pvr_free_list_release(struct kref *ref_count) +{ + struct pvr_free_list *free_list = + container_of(ref_count, struct pvr_free_list, ref_count); + struct list_head *pos, *n; + int err; + + xa_erase(&free_list->pvr_dev->free_list_ids, free_list->fw_id); + + err = pvr_fw_structure_cleanup(free_list->pvr_dev, + ROGUE_FWIF_CLEANUP_FREELIST, + free_list->fw_obj, 0); + if (err == -EBUSY) { + /* Flush the FWCCB to process any HWR or freelist reconstruction + * request that might keep the freelist busy, and try again. + */ + pvr_fwccb_process(free_list->pvr_dev); + err = pvr_fw_structure_cleanup(free_list->pvr_dev, + ROGUE_FWIF_CLEANUP_FREELIST, + free_list->fw_obj, 0); + } + + WARN_ON(err); + + /* clang-format off */ + list_for_each_safe(pos, n, &free_list->mem_block_list) { + struct pvr_free_list_node *free_list_node = + container_of(pos, struct pvr_free_list_node, node); + + list_del(pos); + pvr_free_list_free_node(free_list_node); + } + /* clang-format on */ + + free_list_destroy_kernel_structure(free_list); + free_list_destroy_fw_structure(free_list); + mutex_destroy(&free_list->lock); + kfree(free_list); +} + +/** + * pvr_destroy_free_lists_for_file: Destroy any free lists associated with the + * given file. + * @pvr_file: Pointer to pvr_file structure. + * + * Removes all free lists associated with @pvr_file from the device free_list + * list and drops initial references. Free lists will then be destroyed once + * all outstanding references are dropped. + */ +void pvr_destroy_free_lists_for_file(struct pvr_file *pvr_file) +{ + struct pvr_free_list *free_list; + unsigned long handle; + + xa_for_each(&pvr_file->free_list_handles, handle, free_list) { + (void)free_list; + pvr_free_list_put(xa_erase(&pvr_file->free_list_handles, handle)); + } +} + +/** + * pvr_free_list_put() - Release reference on free list + * @free_list: Pointer to list to release reference on + */ +void +pvr_free_list_put(struct pvr_free_list *free_list) +{ + if (free_list) + kref_put(&free_list->ref_count, pvr_free_list_release); +} + +void pvr_free_list_add_hwrt(struct pvr_free_list *free_list, struct pvr_hwrt_data *hwrt_data) +{ + mutex_lock(&free_list->lock); + + list_add_tail(&hwrt_data->freelist_node, &free_list->hwrt_list); + + mutex_unlock(&free_list->lock); +} + +void pvr_free_list_remove_hwrt(struct pvr_free_list *free_list, struct pvr_hwrt_data *hwrt_data) +{ + mutex_lock(&free_list->lock); + + list_del(&hwrt_data->freelist_node); + + mutex_unlock(&free_list->lock); +} + +static void +pvr_free_list_reconstruct(struct pvr_device *pvr_dev, u32 freelist_id) +{ + struct pvr_free_list *free_list = pvr_free_list_lookup_id(pvr_dev, freelist_id); + struct pvr_free_list_node *free_list_node; + struct rogue_fwif_freelist *fw_data; + struct pvr_hwrt_data *hwrt_data; + + if (!free_list) + return; + + mutex_lock(&free_list->lock); + + /* Rebuild the free list based on the memory block list. */ + free_list->current_pages = 0; + + list_for_each_entry(free_list_node, &free_list->mem_block_list, node) + WARN_ON(pvr_free_list_insert_node_locked(free_list_node)); + + /* + * Remove the ready pages, which are reserved to allow the FW to process OOM quickly and + * asynchronously request a grow. + */ + free_list->current_pages -= free_list->ready_pages; + + fw_data = free_list->fw_data; + fw_data->current_stack_top = fw_data->current_pages - 1; + fw_data->allocated_page_count = 0; + fw_data->allocated_mmu_page_count = 0; + + /* Reset the state of any associated HWRTs. */ + list_for_each_entry(hwrt_data, &free_list->hwrt_list, freelist_node) { + struct rogue_fwif_hwrtdata *hwrt_fw_data = pvr_fw_object_vmap(hwrt_data->fw_obj); + + if (!WARN_ON(IS_ERR(hwrt_fw_data))) { + hwrt_fw_data->state = ROGUE_FWIF_RTDATA_STATE_HWR; + hwrt_fw_data->hwrt_data_flags &= ~HWRTDATA_HAS_LAST_GEOM; + } + + pvr_fw_object_vunmap(hwrt_data->fw_obj); + } + + mutex_unlock(&free_list->lock); + + pvr_free_list_put(free_list); +} + +void +pvr_free_list_process_reconstruct_req(struct pvr_device *pvr_dev, + struct rogue_fwif_fwccb_cmd_freelists_reconstruction_data *req) +{ + struct rogue_fwif_kccb_cmd resp_cmd = { + .cmd_type = ROGUE_FWIF_KCCB_CMD_FREELISTS_RECONSTRUCTION_UPDATE, + }; + struct rogue_fwif_freelists_reconstruction_data *resp = + &resp_cmd.cmd_data.free_lists_reconstruction_data; + + for (u32 i = 0; i < req->freelist_count; i++) + pvr_free_list_reconstruct(pvr_dev, req->freelist_ids[i]); + + resp->freelist_count = req->freelist_count; + memcpy(resp->freelist_ids, req->freelist_ids, + req->freelist_count * sizeof(resp->freelist_ids[0])); + + WARN_ON(pvr_kccb_send_cmd(pvr_dev, &resp_cmd, NULL)); +} diff --git a/drivers/gpu/drm/imagination/pvr_free_list.h b/drivers/gpu/drm/imagination/pvr_free_list.h new file mode 100644 index 000000000000..bfb4f5fc622c --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_free_list.h @@ -0,0 +1,195 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_FREE_LIST_H +#define PVR_FREE_LIST_H + +#include +#include +#include +#include +#include +#include +#include + +#include "pvr_device.h" + +/* Forward declaration from pvr_gem.h. */ +struct pvr_fw_object; + +/* Forward declaration from pvr_gem.h. */ +struct pvr_gem_object; + +/* Forward declaration from pvr_hwrt.h. */ +struct pvr_hwrt_data; + +/** + * struct pvr_free_list_node - structure representing an allocation in the free + * list + */ +struct pvr_free_list_node { + /** @node: List node for &pvr_free_list.mem_block_list. */ + struct list_head node; + + /** @free_list: Pointer to owning free list. */ + struct pvr_free_list *free_list; + + /** @num_pages: Number of pages in this node. */ + u32 num_pages; + + /** @mem_obj: GEM object representing the pages in this node. */ + struct pvr_gem_object *mem_obj; +}; + +/** + * struct pvr_free_list - structure representing a free list + */ +struct pvr_free_list { + /** @ref_count: Reference count of object. */ + struct kref ref_count; + + /** @pvr_dev: Pointer to device that owns this object. */ + struct pvr_device *pvr_dev; + + /** @obj: GEM object representing the free list. */ + struct pvr_gem_object *obj; + + /** @fw_obj: FW object representing the FW-side structure. */ + struct pvr_fw_object *fw_obj; + + /** @fw_data: Pointer to CPU mapping of the FW-side structure. */ + struct rogue_fwif_freelist *fw_data; + + /** + * @lock: Mutex protecting modification of the free list. Must be held when accessing any + * of the members below. + */ + struct mutex lock; + + /** @fw_id: Firmware ID for this object. */ + u32 fw_id; + + /** @current_pages: Current number of pages in free list. */ + u32 current_pages; + + /** @max_pages: Maximum number of pages in free list. */ + u32 max_pages; + + /** @grow_pages: Pages to grow free list by per request. */ + u32 grow_pages; + + /** + * @grow_threshold: Percentage of FL memory used that should trigger a + * new grow request. + */ + u32 grow_threshold; + + /** + * @ready_pages: Number of pages reserved for FW to use while a grow + * request is being processed. + */ + u32 ready_pages; + + /** @mem_block_list: List of memory blocks in this free list. */ + struct list_head mem_block_list; + + /** @hwrt_list: List of HWRTs using this free list. */ + struct list_head hwrt_list; + + /** @initial_num_pages: Initial number of pages in free list. */ + u32 initial_num_pages; + + /** @free_list_gpu_addr: Address of free list in GPU address space. */ + u64 free_list_gpu_addr; +}; + +struct pvr_free_list * +pvr_free_list_create(struct pvr_file *pvr_file, + struct drm_pvr_ioctl_create_free_list_args *args); + +void +pvr_destroy_free_lists_for_file(struct pvr_file *pvr_file); + +u32 +pvr_get_free_list_min_pages(struct pvr_device *pvr_dev); + +static __always_inline struct pvr_free_list * +pvr_free_list_get(struct pvr_free_list *free_list) +{ + if (free_list) + kref_get(&free_list->ref_count); + + return free_list; +} + +/** + * pvr_free_list_lookup() - Lookup free list pointer from handle and file + * @pvr_file: Pointer to pvr_file structure. + * @handle: Object handle. + * + * Takes reference on free list object. Call pvr_free_list_put() to release. + * + * Returns: + * * The requested object on success, or + * * %NULL on failure (object does not exist in list, is not a free list, or + * does not belong to @pvr_file) + */ +static __always_inline struct pvr_free_list * +pvr_free_list_lookup(struct pvr_file *pvr_file, u32 handle) +{ + struct pvr_free_list *free_list; + + xa_lock(&pvr_file->free_list_handles); + free_list = pvr_free_list_get(xa_load(&pvr_file->free_list_handles, handle)); + xa_unlock(&pvr_file->free_list_handles); + + return free_list; +} + +/** + * pvr_free_list_lookup_id() - Lookup free list pointer from FW ID + * @pvr_dev: Device pointer. + * @id: FW object ID. + * + * Takes reference on free list object. Call pvr_free_list_put() to release. + * + * Returns: + * * The requested object on success, or + * * %NULL on failure (object does not exist in list, or is not a free list) + */ +static __always_inline struct pvr_free_list * +pvr_free_list_lookup_id(struct pvr_device *pvr_dev, u32 id) +{ + struct pvr_free_list *free_list; + + xa_lock(&pvr_dev->free_list_ids); + + /* Contexts are removed from the ctx_ids set in the context release path, + * meaning the ref_count reached zero before they get removed. We need + * to make sure we're not trying to acquire a context that's being + * destroyed. + */ + free_list = xa_load(&pvr_dev->free_list_ids, id); + if (free_list && !kref_get_unless_zero(&free_list->ref_count)) + free_list = NULL; + xa_unlock(&pvr_dev->free_list_ids); + + return free_list; +} + +void +pvr_free_list_put(struct pvr_free_list *free_list); + +void +pvr_free_list_add_hwrt(struct pvr_free_list *free_list, struct pvr_hwrt_data *hwrt_data); +void +pvr_free_list_remove_hwrt(struct pvr_free_list *free_list, struct pvr_hwrt_data *hwrt_data); + +void pvr_free_list_process_grow_req(struct pvr_device *pvr_dev, + struct rogue_fwif_fwccb_cmd_freelist_gs_data *req); + +void +pvr_free_list_process_reconstruct_req(struct pvr_device *pvr_dev, + struct rogue_fwif_fwccb_cmd_freelists_reconstruction_data *req); + +#endif /* PVR_FREE_LIST_H */ diff --git a/drivers/gpu/drm/imagination/pvr_hwrt.c b/drivers/gpu/drm/imagination/pvr_hwrt.c new file mode 100644 index 000000000000..c4213c18489e --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_hwrt.c @@ -0,0 +1,549 @@ +// SPDX-License-Identifier: GPL-2.0-only OR MIT +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#include "pvr_free_list.h" +#include "pvr_hwrt.h" +#include "pvr_gem.h" +#include "pvr_rogue_cr_defs_client.h" +#include "pvr_rogue_fwif.h" + +#include +#include +#include +#include +#include +#include + +static_assert(ROGUE_FWIF_NUM_RTDATAS == 2); +static_assert(ROGUE_FWIF_NUM_GEOMDATAS == 1); +static_assert(ROGUE_FWIF_NUM_RTDATA_FREELISTS == 2); + +/* + * struct pvr_rt_mtile_info - Render target macrotile information + */ +struct pvr_rt_mtile_info { + u32 mtile_x[3]; + u32 mtile_y[3]; + u32 tile_max_x; + u32 tile_max_y; + u32 tile_size_x; + u32 tile_size_y; + u32 num_tiles_x; + u32 num_tiles_y; +}; + +/* Size of Shadow Render Target Cache entry */ +#define SRTC_ENTRY_SIZE sizeof(u32) +/* Size of Renders Accumulation Array entry */ +#define RAA_ENTRY_SIZE sizeof(u32) + +static int +hwrt_init_kernel_structure(struct pvr_file *pvr_file, + struct drm_pvr_ioctl_create_hwrt_dataset_args *args, + struct pvr_hwrt_dataset *hwrt) +{ + struct pvr_device *pvr_dev = pvr_file->pvr_dev; + int err; + int i; + + hwrt->pvr_dev = pvr_dev; + hwrt->max_rts = args->layers; + + /* Get pointers to the free lists */ + for (i = 0; i < ARRAY_SIZE(hwrt->free_lists); i++) { + hwrt->free_lists[i] = pvr_free_list_lookup(pvr_file, args->free_list_handles[i]); + if (!hwrt->free_lists[i]) { + err = -EINVAL; + goto err_put_free_lists; + } + } + + if (hwrt->free_lists[ROGUE_FW_LOCAL_FREELIST]->current_pages < + pvr_get_free_list_min_pages(pvr_dev)) { + err = -EINVAL; + goto err_put_free_lists; + } + + return 0; + +err_put_free_lists: + for (i = 0; i < ARRAY_SIZE(hwrt->free_lists); i++) { + pvr_free_list_put(hwrt->free_lists[i]); + hwrt->free_lists[i] = NULL; + } + + return err; +} + +static void +hwrt_fini_kernel_structure(struct pvr_hwrt_dataset *hwrt) +{ + int i; + + for (i = 0; i < ARRAY_SIZE(hwrt->free_lists); i++) { + pvr_free_list_put(hwrt->free_lists[i]); + hwrt->free_lists[i] = NULL; + } +} + +static void +hwrt_fini_common_fw_structure(struct pvr_hwrt_dataset *hwrt) +{ + pvr_fw_object_destroy(hwrt->common_fw_obj); +} + +static int +get_cr_isp_mtile_size_val(struct pvr_device *pvr_dev, u32 samples, + struct pvr_rt_mtile_info *info, u32 *value_out) +{ + u32 x = info->mtile_x[0]; + u32 y = info->mtile_y[0]; + u32 samples_per_pixel; + int err; + + err = PVR_FEATURE_VALUE(pvr_dev, isp_samples_per_pixel, &samples_per_pixel); + if (err) + return err; + + if (samples_per_pixel == 1) { + if (samples >= 4) + x <<= 1; + if (samples >= 2) + y <<= 1; + } else if (samples_per_pixel == 2) { + if (samples >= 8) + x <<= 1; + if (samples >= 4) + y <<= 1; + } else if (samples_per_pixel == 4) { + if (samples >= 8) + y <<= 1; + } else { + WARN(true, "Unsupported ISP samples per pixel value"); + return -EINVAL; + } + + *value_out = ((x << ROGUE_CR_ISP_MTILE_SIZE_X_SHIFT) & ~ROGUE_CR_ISP_MTILE_SIZE_X_CLRMSK) | + ((y << ROGUE_CR_ISP_MTILE_SIZE_Y_SHIFT) & ~ROGUE_CR_ISP_MTILE_SIZE_Y_CLRMSK); + + return 0; +} + +static int +get_cr_multisamplectl_val(u32 samples, bool y_flip, u64 *value_out) +{ + static const struct { + u8 x[8]; + u8 y[8]; + } sample_positions[4] = { + /* 1 sample */ + { + .x = { 8 }, + .y = { 8 }, + }, + /* 2 samples */ + { + .x = { 12, 4 }, + .y = { 12, 4 }, + }, + /* 4 samples */ + { + .x = { 6, 14, 2, 10 }, + .y = { 2, 6, 10, 14 }, + }, + /* 8 samples */ + { + .x = { 9, 7, 13, 5, 3, 1, 11, 15 }, + .y = { 5, 11, 9, 3, 13, 7, 15, 1 }, + }, + }; + const int idx = fls(samples) - 1; + u64 value = 0; + + if (idx < 0 || idx > 3) + return -EINVAL; + + for (u32 i = 0; i < 8; i++) { + value |= ((u64)sample_positions[idx].x[i]) << (i * 8); + if (y_flip) + value |= (((u64)(16 - sample_positions[idx].y[i]) & 0xf)) << (i * 8 + 4); + else + value |= ((u64)sample_positions[idx].y[i]) << (i * 8 + 4); + } + + *value_out = value; + + return 0; +} + +static int +get_cr_te_aa_val(struct pvr_device *pvr_dev, u32 samples, u32 *value_out) +{ + u32 samples_per_pixel; + u32 value = 0; + int err = 0; + + err = PVR_FEATURE_VALUE(pvr_dev, isp_samples_per_pixel, &samples_per_pixel); + if (err) + return err; + + switch (samples_per_pixel) { + case 1: + if (samples >= 2) + value |= ROGUE_CR_TE_AA_Y_EN; + if (samples >= 4) + value |= ROGUE_CR_TE_AA_X_EN; + break; + case 2: + if (samples >= 2) + value |= ROGUE_CR_TE_AA_X2_EN; + if (samples >= 4) + value |= ROGUE_CR_TE_AA_Y_EN; + if (samples >= 8) + value |= ROGUE_CR_TE_AA_X_EN; + break; + case 4: + if (samples >= 2) + value |= ROGUE_CR_TE_AA_X2_EN; + if (samples >= 4) + value |= ROGUE_CR_TE_AA_Y2_EN; + if (samples >= 8) + value |= ROGUE_CR_TE_AA_Y_EN; + break; + default: + WARN(true, "Unsupported ISP samples per pixel value"); + return -EINVAL; + } + + *value_out = value; + + return 0; +} + +static void +hwrtdata_common_init(void *cpu_ptr, void *priv) +{ + struct pvr_hwrt_dataset *hwrt = priv; + + memcpy(cpu_ptr, &hwrt->common, sizeof(hwrt->common)); +} + +static int +hwrt_init_common_fw_structure(struct pvr_file *pvr_file, + struct drm_pvr_ioctl_create_hwrt_dataset_args *args, + struct pvr_hwrt_dataset *hwrt) +{ + struct drm_pvr_create_hwrt_geom_data_args *geom_data_args = &args->geom_data_args; + struct pvr_device *pvr_dev = pvr_file->pvr_dev; + struct pvr_rt_mtile_info info; + int err; + + err = PVR_FEATURE_VALUE(pvr_dev, tile_size_x, &info.tile_size_x); + if (WARN_ON(err)) + return err; + + err = PVR_FEATURE_VALUE(pvr_dev, tile_size_y, &info.tile_size_y); + if (WARN_ON(err)) + return err; + + info.num_tiles_x = DIV_ROUND_UP(args->width, info.tile_size_x); + info.num_tiles_y = DIV_ROUND_UP(args->height, info.tile_size_y); + + if (PVR_HAS_FEATURE(pvr_dev, simple_parameter_format_version)) { + u32 parameter_format; + + err = PVR_FEATURE_VALUE(pvr_dev, simple_parameter_format_version, + ¶meter_format); + if (WARN_ON(err)) + return err; + + WARN_ON(parameter_format != 2); + + /* + * Set up 16 macrotiles with a multiple of 2x2 tiles per macrotile, which is + * aligned to a tile group. + */ + info.mtile_x[0] = DIV_ROUND_UP(info.num_tiles_x, 8) * 2; + info.mtile_y[0] = DIV_ROUND_UP(info.num_tiles_y, 8) * 2; + info.mtile_x[1] = 0; + info.mtile_y[1] = 0; + info.mtile_x[2] = 0; + info.mtile_y[2] = 0; + info.tile_max_x = round_up(info.num_tiles_x, 2) - 1; + info.tile_max_y = round_up(info.num_tiles_y, 2) - 1; + } else { + /* Set up 16 macrotiles with a multiple of 4x4 tiles per macrotile. */ + info.mtile_x[0] = round_up(DIV_ROUND_UP(info.num_tiles_x, 4), 4); + info.mtile_y[0] = round_up(DIV_ROUND_UP(info.num_tiles_y, 4), 4); + info.mtile_x[1] = info.mtile_x[0] * 2; + info.mtile_y[1] = info.mtile_y[0] * 2; + info.mtile_x[2] = info.mtile_x[0] * 3; + info.mtile_y[2] = info.mtile_y[0] * 3; + info.tile_max_x = info.num_tiles_x - 1; + info.tile_max_y = info.num_tiles_y - 1; + } + + hwrt->common.geom_caches_need_zeroing = false; + + hwrt->common.isp_merge_lower_x = args->isp_merge_lower_x; + hwrt->common.isp_merge_lower_y = args->isp_merge_lower_y; + hwrt->common.isp_merge_upper_x = args->isp_merge_upper_x; + hwrt->common.isp_merge_upper_y = args->isp_merge_upper_y; + hwrt->common.isp_merge_scale_x = args->isp_merge_scale_x; + hwrt->common.isp_merge_scale_y = args->isp_merge_scale_y; + + err = get_cr_multisamplectl_val(args->samples, false, + &hwrt->common.multi_sample_ctl); + if (err) + return err; + + err = get_cr_multisamplectl_val(args->samples, true, + &hwrt->common.flipped_multi_sample_ctl); + if (err) + return err; + + hwrt->common.mtile_stride = info.mtile_x[0] * info.mtile_y[0]; + + err = get_cr_te_aa_val(pvr_dev, args->samples, &hwrt->common.teaa); + if (err) + return err; + + hwrt->common.screen_pixel_max = + (((args->width - 1) << ROGUE_CR_PPP_SCREEN_PIXXMAX_SHIFT) & + ~ROGUE_CR_PPP_SCREEN_PIXXMAX_CLRMSK) | + (((args->height - 1) << ROGUE_CR_PPP_SCREEN_PIXYMAX_SHIFT) & + ~ROGUE_CR_PPP_SCREEN_PIXYMAX_CLRMSK); + + hwrt->common.te_screen = + ((info.tile_max_x << ROGUE_CR_TE_SCREEN_XMAX_SHIFT) & + ~ROGUE_CR_TE_SCREEN_XMAX_CLRMSK) | + ((info.tile_max_y << ROGUE_CR_TE_SCREEN_YMAX_SHIFT) & + ~ROGUE_CR_TE_SCREEN_YMAX_CLRMSK); + hwrt->common.te_mtile1 = + ((info.mtile_x[0] << ROGUE_CR_TE_MTILE1_X1_SHIFT) & ~ROGUE_CR_TE_MTILE1_X1_CLRMSK) | + ((info.mtile_x[1] << ROGUE_CR_TE_MTILE1_X2_SHIFT) & ~ROGUE_CR_TE_MTILE1_X2_CLRMSK) | + ((info.mtile_x[2] << ROGUE_CR_TE_MTILE1_X3_SHIFT) & ~ROGUE_CR_TE_MTILE1_X3_CLRMSK); + hwrt->common.te_mtile2 = + ((info.mtile_y[0] << ROGUE_CR_TE_MTILE2_Y1_SHIFT) & ~ROGUE_CR_TE_MTILE2_Y1_CLRMSK) | + ((info.mtile_y[1] << ROGUE_CR_TE_MTILE2_Y2_SHIFT) & ~ROGUE_CR_TE_MTILE2_Y2_CLRMSK) | + ((info.mtile_y[2] << ROGUE_CR_TE_MTILE2_Y3_SHIFT) & ~ROGUE_CR_TE_MTILE2_Y3_CLRMSK); + + err = get_cr_isp_mtile_size_val(pvr_dev, args->samples, &info, + &hwrt->common.isp_mtile_size); + if (err) + return err; + + hwrt->common.tpc_stride = geom_data_args->tpc_stride; + hwrt->common.tpc_size = geom_data_args->tpc_size; + + hwrt->common.rgn_header_size = args->region_header_size; + + err = pvr_fw_object_create(pvr_dev, sizeof(struct rogue_fwif_hwrtdata_common), + PVR_BO_FW_FLAGS_DEVICE_UNCACHED, hwrtdata_common_init, hwrt, + &hwrt->common_fw_obj); + + return err; +} + +static void +hwrt_fw_data_init(void *cpu_ptr, void *priv) +{ + struct pvr_hwrt_data *hwrt_data = priv; + + memcpy(cpu_ptr, &hwrt_data->data, sizeof(hwrt_data->data)); +} + +static int +hwrt_data_init_fw_structure(struct pvr_file *pvr_file, + struct pvr_hwrt_dataset *hwrt, + struct drm_pvr_ioctl_create_hwrt_dataset_args *args, + struct drm_pvr_create_hwrt_rt_data_args *rt_data_args, + struct pvr_hwrt_data *hwrt_data) +{ + struct drm_pvr_create_hwrt_geom_data_args *geom_data_args = &args->geom_data_args; + struct pvr_device *pvr_dev = pvr_file->pvr_dev; + struct rogue_fwif_rta_ctl *rta_ctl; + int free_list_i; + int err; + + pvr_fw_object_get_fw_addr(hwrt->common_fw_obj, + &hwrt_data->data.hwrt_data_common_fw_addr); + + for (free_list_i = 0; free_list_i < ARRAY_SIZE(hwrt->free_lists); free_list_i++) { + pvr_fw_object_get_fw_addr(hwrt->free_lists[free_list_i]->fw_obj, + &hwrt_data->data.freelists_fw_addr[free_list_i]); + } + + hwrt_data->data.tail_ptrs_dev_addr = geom_data_args->tpc_dev_addr; + hwrt_data->data.vheap_table_dev_addr = geom_data_args->vheap_table_dev_addr; + hwrt_data->data.rtc_dev_addr = geom_data_args->rtc_dev_addr; + + hwrt_data->data.pm_mlist_dev_addr = rt_data_args->pm_mlist_dev_addr; + hwrt_data->data.macrotile_array_dev_addr = rt_data_args->macrotile_array_dev_addr; + hwrt_data->data.rgn_header_dev_addr = rt_data_args->region_header_dev_addr; + + rta_ctl = &hwrt_data->data.rta_ctl; + + rta_ctl->render_target_index = 0; + rta_ctl->active_render_targets = 0; + rta_ctl->valid_render_targets_fw_addr = 0; + rta_ctl->rta_num_partial_renders_fw_addr = 0; + rta_ctl->max_rts = args->layers; + + if (args->layers > 1) { + err = pvr_fw_object_create(pvr_dev, args->layers * SRTC_ENTRY_SIZE, + PVR_BO_FW_FLAGS_DEVICE_UNCACHED, + NULL, NULL, &hwrt_data->srtc_obj); + if (err) + return err; + pvr_fw_object_get_fw_addr(hwrt_data->srtc_obj, + &rta_ctl->valid_render_targets_fw_addr); + + err = pvr_fw_object_create(pvr_dev, args->layers * RAA_ENTRY_SIZE, + PVR_BO_FW_FLAGS_DEVICE_UNCACHED, + NULL, NULL, &hwrt_data->raa_obj); + if (err) + goto err_put_shadow_rt_cache; + pvr_fw_object_get_fw_addr(hwrt_data->raa_obj, + &rta_ctl->rta_num_partial_renders_fw_addr); + } + + err = pvr_fw_object_create(pvr_dev, sizeof(struct rogue_fwif_hwrtdata), + PVR_BO_FW_FLAGS_DEVICE_UNCACHED, + hwrt_fw_data_init, hwrt_data, &hwrt_data->fw_obj); + if (err) + goto err_put_raa_obj; + + pvr_free_list_add_hwrt(hwrt->free_lists[0], hwrt_data); + + return 0; + +err_put_raa_obj: + if (args->layers > 1) + pvr_fw_object_destroy(hwrt_data->raa_obj); + +err_put_shadow_rt_cache: + if (args->layers > 1) + pvr_fw_object_destroy(hwrt_data->srtc_obj); + + return err; +} + +static void +hwrt_data_fini_fw_structure(struct pvr_hwrt_dataset *hwrt, int hwrt_nr) +{ + struct pvr_hwrt_data *hwrt_data = &hwrt->data[hwrt_nr]; + + pvr_free_list_remove_hwrt(hwrt->free_lists[0], hwrt_data); + + if (hwrt->max_rts > 1) { + pvr_fw_object_destroy(hwrt_data->raa_obj); + pvr_fw_object_destroy(hwrt_data->srtc_obj); + } + + pvr_fw_object_destroy(hwrt_data->fw_obj); +} + +/** + * pvr_hwrt_dataset_create() - Create a new HWRT dataset + * @pvr_file: Pointer to pvr_file structure. + * @args: Creation arguments from userspace. + * + * Return: + * * Pointer to new HWRT, or + * * ERR_PTR(-%ENOMEM) on out of memory. + */ +struct pvr_hwrt_dataset * +pvr_hwrt_dataset_create(struct pvr_file *pvr_file, + struct drm_pvr_ioctl_create_hwrt_dataset_args *args) +{ + struct pvr_hwrt_dataset *hwrt; + int err; + + /* Create and fill out the kernel structure */ + hwrt = kzalloc(sizeof(*hwrt), GFP_KERNEL); + + if (!hwrt) + return ERR_PTR(-ENOMEM); + + kref_init(&hwrt->ref_count); + + err = hwrt_init_kernel_structure(pvr_file, args, hwrt); + if (err < 0) + goto err_free; + + err = hwrt_init_common_fw_structure(pvr_file, args, hwrt); + if (err < 0) + goto err_free; + + for (int i = 0; i < ARRAY_SIZE(hwrt->data); i++) { + err = hwrt_data_init_fw_structure(pvr_file, hwrt, args, + &args->rt_data_args[i], + &hwrt->data[i]); + if (err < 0) { + i--; + /* Destroy already created structures. */ + for (; i >= 0; i--) + hwrt_data_fini_fw_structure(hwrt, i); + goto err_free; + } + + hwrt->data[i].hwrt_dataset = hwrt; + } + + return hwrt; + +err_free: + pvr_hwrt_dataset_put(hwrt); + + return ERR_PTR(err); +} + +static void +pvr_hwrt_dataset_release(struct kref *ref_count) +{ + struct pvr_hwrt_dataset *hwrt = + container_of(ref_count, struct pvr_hwrt_dataset, ref_count); + + for (int i = ARRAY_SIZE(hwrt->data) - 1; i >= 0; i--) { + WARN_ON(pvr_fw_structure_cleanup(hwrt->pvr_dev, ROGUE_FWIF_CLEANUP_HWRTDATA, + hwrt->data[i].fw_obj, 0)); + hwrt_data_fini_fw_structure(hwrt, i); + } + + hwrt_fini_common_fw_structure(hwrt); + hwrt_fini_kernel_structure(hwrt); + + kfree(hwrt); +} + +/** + * pvr_destroy_hwrt_datasets_for_file: Destroy any HWRT datasets associated + * with the given file. + * @pvr_file: Pointer to pvr_file structure. + * + * Removes all HWRT datasets associated with @pvr_file from the device + * hwrt_dataset list and drops initial references. HWRT datasets will then be + * destroyed once all outstanding references are dropped. + */ +void pvr_destroy_hwrt_datasets_for_file(struct pvr_file *pvr_file) +{ + struct pvr_hwrt_dataset *hwrt; + unsigned long handle; + + xa_for_each(&pvr_file->hwrt_handles, handle, hwrt) { + (void)hwrt; + pvr_hwrt_dataset_put(xa_erase(&pvr_file->hwrt_handles, handle)); + } +} + +/** + * pvr_hwrt_dataset_put() - Release reference on HWRT dataset + * @hwrt: Pointer to HWRT dataset to release reference on + */ +void +pvr_hwrt_dataset_put(struct pvr_hwrt_dataset *hwrt) +{ + if (hwrt) + kref_put(&hwrt->ref_count, pvr_hwrt_dataset_release); +} diff --git a/drivers/gpu/drm/imagination/pvr_hwrt.h b/drivers/gpu/drm/imagination/pvr_hwrt.h new file mode 100644 index 000000000000..76992948d047 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_hwrt.h @@ -0,0 +1,165 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_HWRT_H +#define PVR_HWRT_H + +#include +#include +#include +#include +#include +#include + +#include "pvr_device.h" +#include "pvr_rogue_fwif_shared.h" + +/* Forward declaration from pvr_free_list.h. */ +struct pvr_free_list; + +/* Forward declaration from pvr_gem.h. */ +struct pvr_fw_object; + +/** + * struct pvr_hwrt_data - structure representing HWRT data + */ +struct pvr_hwrt_data { + /** @fw_obj: FW object representing the FW-side structure. */ + struct pvr_fw_object *fw_obj; + + /** @data: Local copy of FW-side structure. */ + struct rogue_fwif_hwrtdata data; + + /** @freelist_node: List node connecting this HWRT to the local freelist. */ + struct list_head freelist_node; + + /** + * @srtc_obj: FW object representing shadow render target cache. + * + * Only valid if @max_rts > 1. + */ + struct pvr_fw_object *srtc_obj; + + /** + * @raa_obj: FW object representing renders accumulation array. + * + * Only valid if @max_rts > 1. + */ + struct pvr_fw_object *raa_obj; + + /** @hwrt_dataset: Back pointer to owning HWRT dataset. */ + struct pvr_hwrt_dataset *hwrt_dataset; +}; + +/** + * struct pvr_hwrt_dataset - structure representing a HWRT data set. + */ +struct pvr_hwrt_dataset { + /** @ref_count: Reference count of object. */ + struct kref ref_count; + + /** @pvr_dev: Pointer to device that owns this object. */ + struct pvr_device *pvr_dev; + + /** @common_fw_obj: FW object representing common FW-side structure. */ + struct pvr_fw_object *common_fw_obj; + + struct rogue_fwif_hwrtdata_common common; + + /** @data: HWRT data structures belonging to this set. */ + struct pvr_hwrt_data data[ROGUE_FWIF_NUM_RTDATAS]; + + /** @free_lists: Free lists used by HWRT data set. */ + struct pvr_free_list *free_lists[ROGUE_FWIF_NUM_RTDATA_FREELISTS]; + + /** @max_rts: Maximum render targets for this HWRT data set. */ + u16 max_rts; +}; + +struct pvr_hwrt_dataset * +pvr_hwrt_dataset_create(struct pvr_file *pvr_file, + struct drm_pvr_ioctl_create_hwrt_dataset_args *args); + +void +pvr_destroy_hwrt_datasets_for_file(struct pvr_file *pvr_file); + +/** + * pvr_hwrt_dataset_lookup() - Lookup HWRT dataset pointer from handle + * @pvr_file: Pointer to pvr_file structure. + * @handle: Object handle. + * + * Takes reference on dataset object. Call pvr_hwrt_dataset_put() to release. + * + * Returns: + * * The requested object on success, or + * * %NULL on failure (object does not exist in list, or is not a HWRT + * dataset) + */ +static __always_inline struct pvr_hwrt_dataset * +pvr_hwrt_dataset_lookup(struct pvr_file *pvr_file, u32 handle) +{ + struct pvr_hwrt_dataset *hwrt; + + xa_lock(&pvr_file->hwrt_handles); + hwrt = xa_load(&pvr_file->hwrt_handles, handle); + + if (hwrt) + kref_get(&hwrt->ref_count); + + xa_unlock(&pvr_file->hwrt_handles); + + return hwrt; +} + +void +pvr_hwrt_dataset_put(struct pvr_hwrt_dataset *hwrt); + +/** + * pvr_hwrt_data_lookup() - Lookup HWRT data pointer from handle and index + * @pvr_file: Pointer to pvr_file structure. + * @handle: Object handle. + * @index: Index of RT data within dataset. + * + * Takes reference on dataset object. Call pvr_hwrt_data_put() to release. + * + * Returns: + * * The requested object on success, or + * * %NULL on failure (object does not exist in list, or is not a HWRT + * dataset, or index is out of range) + */ +static __always_inline struct pvr_hwrt_data * +pvr_hwrt_data_lookup(struct pvr_file *pvr_file, u32 handle, u32 index) +{ + struct pvr_hwrt_dataset *hwrt_dataset = pvr_hwrt_dataset_lookup(pvr_file, handle); + + if (hwrt_dataset) { + if (index < ARRAY_SIZE(hwrt_dataset->data)) + return &hwrt_dataset->data[index]; + + pvr_hwrt_dataset_put(hwrt_dataset); + } + + return NULL; +} + +/** + * pvr_hwrt_data_put() - Release reference on HWRT data + * @hwrt: Pointer to HWRT data to release reference on + */ +static __always_inline void +pvr_hwrt_data_put(struct pvr_hwrt_data *hwrt) +{ + if (hwrt) + pvr_hwrt_dataset_put(hwrt->hwrt_dataset); +} + +static __always_inline struct pvr_hwrt_data * +pvr_hwrt_data_get(struct pvr_hwrt_data *hwrt) +{ + if (hwrt) + kref_get(&hwrt->hwrt_dataset->ref_count); + + return hwrt; +} + +#endif /* PVR_HWRT_H */ From patchwork Wed Nov 22 16:34:38 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Donald Robson X-Patchwork-Id: 746145 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=imgtec.com header.i=@imgtec.com header.b="dGLAm9d5"; dkim=pass (1024-bit key) header.d=IMGTecCRM.onmicrosoft.com header.i=@IMGTecCRM.onmicrosoft.com header.b="E7NQPPMW" Received: from mx08-00376f01.pphosted.com (mx08-00376f01.pphosted.com [91.207.212.86]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 8F5041733; Wed, 22 Nov 2023 08:36:20 -0800 (PST) Received: from pps.filterd (m0168888.ppops.net [127.0.0.1]) by mx08-00376f01.pphosted.com (8.17.1.24/8.17.1.24) with ESMTP id 3AMG01mC020623; Wed, 22 Nov 2023 16:35:36 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=imgtec.com; h= from:to:cc:subject:date:message-id:in-reply-to:references :content-transfer-encoding:content-type:mime-version; s= dk201812; bh=0VftBz/02CZjA0gXu84nRn1v6yz3NqW6a7K70Ppl4EQ=; b=dGL Am9d535MltjoSGbn+3Fo5q+DSx6708qDcMmvTHIL9gcjX6bVRRf8V/vBtMmk/Vi7 jiywhQCk6sBGCuDcg94HKQNqmLtZC+MjLVJV1LIVLXtHtZCv94tBXrPZBiyWM787 zKcJpvTP4HZbR8iFmPQkTkkYrGE7jkVnfMTPzITBAN1T3pxuj6o40xqBGcW3rSmA PXlDBaTc/2wy44CuR5v2oD2KrvYHmkjlW7c3NlRmNaChkZFD5bosPoquRsF3bqUS 8G99jk37+hjc37HTJcMpbqir7H0zl5ioKGw0O3/S9KyieWtQ50bYgN/+khIacfKF Y1Cnr0uf1GW3jrE6VzQ== Received: from hhmail05.hh.imgtec.org ([217.156.249.195]) by mx08-00376f01.pphosted.com (PPS) with ESMTPS id 3ug99gsf3d-4 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128 verify=NOT); Wed, 22 Nov 2023 16:35:35 +0000 (GMT) Received: from HHMAIL04.hh.imgtec.org (10.100.10.119) by HHMAIL05.hh.imgtec.org (10.100.10.120) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34; Wed, 22 Nov 2023 16:35:33 +0000 Received: from GBR01-LO4-obe.outbound.protection.outlook.com (104.47.85.105) by email.imgtec.com (10.100.10.121) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34 via Frontend Transport; Wed, 22 Nov 2023 16:35:33 +0000 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=V2XXqR2Wj4b4g557zUeEo/JuUdkm4gtzZ3qkRg9K/LfQBoALIshyAk4GeioAylxjnra/eY+aS6Sxq7Ld15uPq3nTwANbXORXu+KM4QJHdhp9i4PHjYMpXfZJxgzOup8ijS0R/HM9NiEwFEHGX+pDcnhuZ9v95EenfE4CsTrSoTuWxGa9vbNMjZqwGjeaui5BlKDmewjxNsUxZcFH4nwzpOPudO7rT2HLLRhejDHXB9OOtQaJ6YDfW0FxAFDU2+cXV425hzv4UX7RNKyOQeypwWAY4wgk1mwdxq7VchbemKzWOtkPy4dVQR/zg6e9lGsBr4ubkZJPW05SlchzA1ZSMA== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=0VftBz/02CZjA0gXu84nRn1v6yz3NqW6a7K70Ppl4EQ=; b=R7ifmWa0d9IMyHMUOefiy/ui/+guVFQ7cEuIyiI9Gqf0nXQYVRelOBBgYd+knCCYpedShcC87uULvN+Cr7SYhsT00vwBb+L9TJ70FOGhQ3DGZ7btT231dodFjqUDz/p/SAqTnItxFw2UHtFGZDekU4l0rtT+y3HCn9YhhwkatrpXspkDl9MUpUUsWuhcltGJD3NS36s6d42kTZWJ6RkiBy0m/Hfx6AO5bmSEAOZyvcWw6U/yalhTUBX7HARSrK9vOsqfW39RJ0hvj6QDU1VIS6WEWdouh1sz6ZuL5YZ+11nzlBB0tRQboH+FF0R9oQX5GJxi6pe6WMT4JVrPlNXrjg== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=imgtec.com; dmarc=pass action=none header.from=imgtec.com; dkim=pass header.d=imgtec.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=IMGTecCRM.onmicrosoft.com; s=selector2-IMGTecCRM-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=0VftBz/02CZjA0gXu84nRn1v6yz3NqW6a7K70Ppl4EQ=; b=E7NQPPMWci9MI70D3AK6A3BE1/uWDvt7obwRWvpHnJ3MnZKPKpnxcmGs8+AxJIKPYjOocPutx9NdJb5EqRWev1732kpCl0Eial3zkY5aCdzln1p9G+aWV/5cQR9iGiNqnD9BW1U3PX5Aj/TRwQO8Sz3vNQGNdmo/wxIKGnF81iY= Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) by CWLP265MB6864.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1fe::14) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.7025.19; Wed, 22 Nov 2023 16:35:27 +0000 Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34]) by CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34%3]) with mapi id 15.20.7025.017; Wed, 22 Nov 2023 16:35:27 +0000 From: Donald Robson To: CC: , , , , , , , , , , , , , , , , , , , , , , Sarah Walker Subject: [PATCH v9 17/20] drm/imagination: Implement job submission and scheduling Date: Wed, 22 Nov 2023 16:34:38 +0000 Message-Id: X-Mailer: git-send-email 2.25.1 In-Reply-To: References: X-ClientProxiedBy: LO4P265CA0273.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:37a::10) To CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) Precedence: bulk X-Mailing-List: devicetree@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: CWLP265MB5770:EE_|CWLP265MB6864:EE_ X-MS-Office365-Filtering-Correlation-Id: d01752ab-6639-4dc6-2e8e-08dbeb790890 X-MS-Exchange-SenderADCheck: 1 X-MS-Exchange-AntiSpam-Relay: 0 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-Forefront-Antispam-Report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:; IPV:NLI; SFV:NSPM; H:CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM; PTR:; CAT:NONE; SFS:(13230031)(136003)(366004)(376002)(346002)(39850400004)(396003)(230922051799003)(64100799003)(451199024)(186009)(1800799012)(52116002)(5660300002)(38350700005)(6666004)(6506007)(6512007)(2616005)(8936002)(8676002)(4326008)(83380400001)(2906002)(7416002)(478600001)(30864003)(66476007)(6916009)(66556008)(107886003)(316002)(26005)(66946007)(6486002)(966005)(44832011)(38100700002)(41300700001)(36756003)(86362001)(559001)(579004); DIR:OUT; SFP:1102; X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: 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 X-MS-Exchange-CrossTenant-Network-Message-Id: d01752ab-6639-4dc6-2e8e-08dbeb790890 X-MS-Exchange-CrossTenant-AuthSource: CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 22 Nov 2023 16:35:27.1801 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 0d5fd8bb-e8c2-4e0a-8dd5-2c264f7140fe X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: PyLAxA6yMzBoFC613dTUSN4lWHgNRp2icA60z4o8DcZ+oAzncavW/CCP8iGvmMy2zTx464BM58Xo8tY1urq3LXAwAHgGjR7sJNmYi4OaO50= X-MS-Exchange-Transport-CrossTenantHeadersStamped: CWLP265MB6864 X-OriginatorOrg: imgtec.com X-EXCLAIMER-MD-CONFIG: 15a78312-3e47-46eb-9010-2e54d84a9631 X-Proofpoint-GUID: o2arVab-XtYKkCIph77TPo7XRd9koXS8 X-Proofpoint-ORIG-GUID: o2arVab-XtYKkCIph77TPo7XRd9koXS8 From: Sarah Walker Implement job submission ioctl. Job scheduling is implemented using drm_sched. Jobs are submitted in a stream format. This is intended to allow the UAPI data format to be independent of the actual FWIF structures in use, which vary depending on the GPU in use. The stream formats are documented at: https://gitlab.freedesktop.org/mesa/mesa/-/blob/f8d2b42ae65c2f16f36a43e0ae39d288431e4263/src/imagination/csbgen/rogue_kmd_stream.xml Changes since v8: - Updated for upstreamed DRM scheduler changes - Removed workaround code for the pending_list previously being updated after run_job() returned - Fixed null deref in pvr_queue_cleanup_fw_context() for bad stream ptr given to create_context ioctl - Corrected license identifiers Changes since v7: - Updated for v8 "DRM scheduler changes for XE" patchset Changes since v6: - Fix fence handling in pvr_sync_signal_array_add() - Add handling for SUBMIT_JOB_FRAG_CMD_DISABLE_PIXELMERGE flag - Fix missing dma_resv locking in job submit path Changes since v5: - Fix leak in job creation error path Changes since v4: - Use a regular workqueue for job scheduling Changes since v3: - Support partial render jobs - Add job timeout handler - Split sync handling out of job code - Use drm_dev_{enter,exit} Changes since v2: - Use drm_sched for job scheduling Co-developed-by: Boris Brezillon Signed-off-by: Boris Brezillon Co-developed-by: Donald Robson Signed-off-by: Donald Robson Signed-off-by: Sarah Walker --- drivers/gpu/drm/imagination/Kconfig | 1 + drivers/gpu/drm/imagination/Makefile | 3 + drivers/gpu/drm/imagination/pvr_context.c | 125 +- drivers/gpu/drm/imagination/pvr_context.h | 44 + drivers/gpu/drm/imagination/pvr_device.c | 31 + drivers/gpu/drm/imagination/pvr_device.h | 21 + drivers/gpu/drm/imagination/pvr_drv.c | 40 +- drivers/gpu/drm/imagination/pvr_job.c | 788 +++++++++ drivers/gpu/drm/imagination/pvr_job.h | 161 ++ drivers/gpu/drm/imagination/pvr_power.c | 28 + drivers/gpu/drm/imagination/pvr_queue.c | 1432 +++++++++++++++++ drivers/gpu/drm/imagination/pvr_queue.h | 169 ++ drivers/gpu/drm/imagination/pvr_stream_defs.c | 226 +++ drivers/gpu/drm/imagination/pvr_sync.c | 289 ++++ drivers/gpu/drm/imagination/pvr_sync.h | 84 + 15 files changed, 3438 insertions(+), 4 deletions(-) create mode 100644 drivers/gpu/drm/imagination/pvr_job.c create mode 100644 drivers/gpu/drm/imagination/pvr_job.h create mode 100644 drivers/gpu/drm/imagination/pvr_queue.c create mode 100644 drivers/gpu/drm/imagination/pvr_queue.h create mode 100644 drivers/gpu/drm/imagination/pvr_sync.c create mode 100644 drivers/gpu/drm/imagination/pvr_sync.h diff --git a/drivers/gpu/drm/imagination/Kconfig b/drivers/gpu/drm/imagination/Kconfig index 0abd1b9bf3be..3bfa2ac212dc 100644 --- a/drivers/gpu/drm/imagination/Kconfig +++ b/drivers/gpu/drm/imagination/Kconfig @@ -6,6 +6,7 @@ config DRM_POWERVR depends on ARM64 depends on DRM depends on PM + select DRM_EXEC select DRM_GEM_SHMEM_HELPER select DRM_SCHED select DRM_GPUVM diff --git a/drivers/gpu/drm/imagination/Makefile b/drivers/gpu/drm/imagination/Makefile index 7f7bea8c60c4..7e3515c99caa 100644 --- a/drivers/gpu/drm/imagination/Makefile +++ b/drivers/gpu/drm/imagination/Makefile @@ -18,10 +18,13 @@ powervr-y := \ pvr_fw_trace.o \ pvr_gem.o \ pvr_hwrt.o \ + pvr_job.o \ pvr_mmu.o \ pvr_power.o \ + pvr_queue.o \ pvr_stream.o \ pvr_stream_defs.o \ + pvr_sync.o \ pvr_vm.o \ pvr_vm_mips.o diff --git a/drivers/gpu/drm/imagination/pvr_context.c b/drivers/gpu/drm/imagination/pvr_context.c index 6cec5aa5a759..eded5e955cc0 100644 --- a/drivers/gpu/drm/imagination/pvr_context.c +++ b/drivers/gpu/drm/imagination/pvr_context.c @@ -6,10 +6,12 @@ #include "pvr_device.h" #include "pvr_drv.h" #include "pvr_gem.h" +#include "pvr_job.h" #include "pvr_power.h" #include "pvr_rogue_fwif.h" #include "pvr_rogue_fwif_common.h" #include "pvr_rogue_fwif_resetframework.h" +#include "pvr_stream.h" #include "pvr_stream_defs.h" #include "pvr_vm.h" @@ -164,6 +166,116 @@ ctx_fw_data_init(void *cpu_ptr, void *priv) memcpy(cpu_ptr, ctx->data, ctx->data_size); } +/** + * pvr_context_destroy_queues() - Destroy all queues attached to a context. + * @ctx: Context to destroy queues on. + * + * Should be called when the last reference to a context object is dropped. + * It releases all resources attached to the queues bound to this context. + */ +static void pvr_context_destroy_queues(struct pvr_context *ctx) +{ + switch (ctx->type) { + case DRM_PVR_CTX_TYPE_RENDER: + pvr_queue_destroy(ctx->queues.fragment); + pvr_queue_destroy(ctx->queues.geometry); + break; + case DRM_PVR_CTX_TYPE_COMPUTE: + pvr_queue_destroy(ctx->queues.compute); + break; + case DRM_PVR_CTX_TYPE_TRANSFER_FRAG: + pvr_queue_destroy(ctx->queues.transfer); + break; + } +} + +/** + * pvr_context_create_queues() - Create all queues attached to a context. + * @ctx: Context to create queues on. + * @args: Context creation arguments passed by userspace. + * @fw_ctx_map: CPU mapping of the FW context object. + * + * Return: + * * 0 on success, or + * * A negative error code otherwise. + */ +static int pvr_context_create_queues(struct pvr_context *ctx, + struct drm_pvr_ioctl_create_context_args *args, + void *fw_ctx_map) +{ + int err; + + switch (ctx->type) { + case DRM_PVR_CTX_TYPE_RENDER: + ctx->queues.geometry = pvr_queue_create(ctx, DRM_PVR_JOB_TYPE_GEOMETRY, + args, fw_ctx_map); + if (IS_ERR(ctx->queues.geometry)) { + err = PTR_ERR(ctx->queues.geometry); + ctx->queues.geometry = NULL; + goto err_destroy_queues; + } + + ctx->queues.fragment = pvr_queue_create(ctx, DRM_PVR_JOB_TYPE_FRAGMENT, + args, fw_ctx_map); + if (IS_ERR(ctx->queues.fragment)) { + err = PTR_ERR(ctx->queues.fragment); + ctx->queues.fragment = NULL; + goto err_destroy_queues; + } + return 0; + + case DRM_PVR_CTX_TYPE_COMPUTE: + ctx->queues.compute = pvr_queue_create(ctx, DRM_PVR_JOB_TYPE_COMPUTE, + args, fw_ctx_map); + if (IS_ERR(ctx->queues.compute)) { + err = PTR_ERR(ctx->queues.compute); + ctx->queues.compute = NULL; + goto err_destroy_queues; + } + return 0; + + case DRM_PVR_CTX_TYPE_TRANSFER_FRAG: + ctx->queues.transfer = pvr_queue_create(ctx, DRM_PVR_JOB_TYPE_TRANSFER_FRAG, + args, fw_ctx_map); + if (IS_ERR(ctx->queues.transfer)) { + err = PTR_ERR(ctx->queues.transfer); + ctx->queues.transfer = NULL; + goto err_destroy_queues; + } + return 0; + } + + return -EINVAL; + +err_destroy_queues: + pvr_context_destroy_queues(ctx); + return err; +} + +/** + * pvr_context_kill_queues() - Kill queues attached to context. + * @ctx: Context to kill queues on. + * + * Killing the queues implies making them unusable for future jobs, while still + * letting the currently submitted jobs a chance to finish. Queue resources will + * stay around until pvr_context_destroy_queues() is called. + */ +static void pvr_context_kill_queues(struct pvr_context *ctx) +{ + switch (ctx->type) { + case DRM_PVR_CTX_TYPE_RENDER: + pvr_queue_kill(ctx->queues.fragment); + pvr_queue_kill(ctx->queues.geometry); + break; + case DRM_PVR_CTX_TYPE_COMPUTE: + pvr_queue_kill(ctx->queues.compute); + break; + case DRM_PVR_CTX_TYPE_TRANSFER_FRAG: + pvr_queue_kill(ctx->queues.transfer); + break; + } +} + /** * pvr_context_create() - Create a context. * @pvr_file: File to attach the created context to. @@ -214,10 +326,14 @@ int pvr_context_create(struct pvr_file *pvr_file, struct drm_pvr_ioctl_create_co goto err_put_vm; } - err = init_fw_objs(ctx, args, ctx->data); + err = pvr_context_create_queues(ctx, args, ctx->data); if (err) goto err_free_ctx_data; + err = init_fw_objs(ctx, args, ctx->data); + if (err) + goto err_destroy_queues; + err = pvr_fw_object_create(pvr_dev, ctx_size, PVR_BO_FW_FLAGS_DEVICE_UNCACHED, ctx_fw_data_init, ctx, &ctx->fw_obj); if (err) @@ -243,6 +359,9 @@ int pvr_context_create(struct pvr_file *pvr_file, struct drm_pvr_ioctl_create_co err_destroy_fw_obj: pvr_fw_object_destroy(ctx->fw_obj); +err_destroy_queues: + pvr_context_destroy_queues(ctx); + err_free_ctx_data: kfree(ctx->data); @@ -262,6 +381,7 @@ pvr_context_release(struct kref *ref_count) struct pvr_device *pvr_dev = ctx->pvr_dev; xa_erase(&pvr_dev->ctx_ids, ctx->ctx_id); + pvr_context_destroy_queues(ctx); pvr_fw_object_destroy(ctx->fw_obj); kfree(ctx->data); pvr_vm_context_put(ctx->vm_ctx); @@ -299,6 +419,9 @@ pvr_context_destroy(struct pvr_file *pvr_file, u32 handle) if (!ctx) return -EINVAL; + /* Make sure nothing can be queued to the queues after that point. */ + pvr_context_kill_queues(ctx); + /* Release the reference held by the handle set. */ pvr_context_put(ctx); diff --git a/drivers/gpu/drm/imagination/pvr_context.h b/drivers/gpu/drm/imagination/pvr_context.h index 502dfad113e4..0c7b97dfa6ba 100644 --- a/drivers/gpu/drm/imagination/pvr_context.h +++ b/drivers/gpu/drm/imagination/pvr_context.h @@ -15,6 +15,7 @@ #include "pvr_cccb.h" #include "pvr_device.h" +#include "pvr_queue.h" /* Forward declaration from pvr_gem.h. */ struct pvr_fw_object; @@ -58,8 +59,51 @@ struct pvr_context { /** @ctx_id: FW context ID. */ u32 ctx_id; + + /** + * @faulty: Set to 1 when the context queues had unfinished job when + * a GPU reset happened. + * + * In that case, the context is in an inconsistent state and can't be + * used anymore. + */ + atomic_t faulty; + + /** @queues: Union containing all kind of queues. */ + union { + struct { + /** @geometry: Geometry queue. */ + struct pvr_queue *geometry; + + /** @fragment: Fragment queue. */ + struct pvr_queue *fragment; + }; + + /** @compute: Compute queue. */ + struct pvr_queue *compute; + + /** @compute: Transfer queue. */ + struct pvr_queue *transfer; + } queues; }; +static __always_inline struct pvr_queue * +pvr_context_get_queue_for_job(struct pvr_context *ctx, enum drm_pvr_job_type type) +{ + switch (type) { + case DRM_PVR_JOB_TYPE_GEOMETRY: + return ctx->type == DRM_PVR_CTX_TYPE_RENDER ? ctx->queues.geometry : NULL; + case DRM_PVR_JOB_TYPE_FRAGMENT: + return ctx->type == DRM_PVR_CTX_TYPE_RENDER ? ctx->queues.fragment : NULL; + case DRM_PVR_JOB_TYPE_COMPUTE: + return ctx->type == DRM_PVR_CTX_TYPE_COMPUTE ? ctx->queues.compute : NULL; + case DRM_PVR_JOB_TYPE_TRANSFER_FRAG: + return ctx->type == DRM_PVR_CTX_TYPE_TRANSFER_FRAG ? ctx->queues.transfer : NULL; + } + + return NULL; +} + /** * pvr_context_get() - Take additional reference on context. * @ctx: Context pointer. diff --git a/drivers/gpu/drm/imagination/pvr_device.c b/drivers/gpu/drm/imagination/pvr_device.c index 2d6db4715f85..199f812d1ee7 100644 --- a/drivers/gpu/drm/imagination/pvr_device.c +++ b/drivers/gpu/drm/imagination/pvr_device.c @@ -6,7 +6,9 @@ #include "pvr_fw.h" #include "pvr_power.h" +#include "pvr_queue.h" #include "pvr_rogue_cr_defs.h" +#include "pvr_stream.h" #include "pvr_vm.h" #include @@ -117,6 +119,32 @@ static int pvr_device_clk_init(struct pvr_device *pvr_dev) return 0; } +/** + * pvr_device_process_active_queues() - Process all queue related events. + * @pvr_dev: PowerVR device to check + * + * This is called any time we receive a FW event. It iterates over all + * active queues and calls pvr_queue_process() on them. + */ +void pvr_device_process_active_queues(struct pvr_device *pvr_dev) +{ + struct pvr_queue *queue, *tmp_queue; + LIST_HEAD(active_queues); + + mutex_lock(&pvr_dev->queues.lock); + + /* Move all active queues to a temporary list. Queues that remain + * active after we're done processing them are re-inserted to + * the queues.active list by pvr_queue_process(). + */ + list_splice_init(&pvr_dev->queues.active, &active_queues); + + list_for_each_entry_safe(queue, tmp_queue, &active_queues, node) + pvr_queue_process(queue); + + mutex_unlock(&pvr_dev->queues.lock); +} + static irqreturn_t pvr_device_irq_thread_handler(int irq, void *data) { struct pvr_device *pvr_dev = data; @@ -132,6 +160,7 @@ static irqreturn_t pvr_device_irq_thread_handler(int irq, void *data) if (pvr_dev->fw_dev.booted) { pvr_fwccb_process(pvr_dev); pvr_kccb_wake_up_waiters(pvr_dev); + pvr_device_process_active_queues(pvr_dev); } pm_runtime_mark_last_busy(from_pvr_device(pvr_dev)->dev); @@ -398,6 +427,8 @@ pvr_device_gpu_init(struct pvr_device *pvr_dev) else return -EINVAL; + pvr_stream_create_musthave_masks(pvr_dev); + err = pvr_set_dma_info(pvr_dev); if (err) return err; diff --git a/drivers/gpu/drm/imagination/pvr_device.h b/drivers/gpu/drm/imagination/pvr_device.h index 299f3e022c82..b6e9a7e8072e 100644 --- a/drivers/gpu/drm/imagination/pvr_device.h +++ b/drivers/gpu/drm/imagination/pvr_device.h @@ -173,6 +173,26 @@ struct pvr_device { */ struct xarray free_list_ids; + /** + * @job_ids: Array of jobs belonging to this device. Array members + * are of type "struct pvr_job *". + */ + struct xarray job_ids; + + /** + * @queues: Queue-related fields. + */ + struct { + /** @active: Active queue list. */ + struct list_head active; + + /** @idle: Idle queue list. */ + struct list_head idle; + + /** @lock: Lock protecting access to the active/idle lists. */ + struct mutex lock; + } queues; + struct { /** @work: Work item for watchdog callback. */ struct delayed_work work; @@ -442,6 +462,7 @@ packed_bvnc_to_pvr_gpu_id(u64 bvnc, struct pvr_gpu_id *gpu_id) int pvr_device_init(struct pvr_device *pvr_dev); void pvr_device_fini(struct pvr_device *pvr_dev); +void pvr_device_reset(struct pvr_device *pvr_dev); bool pvr_device_has_uapi_quirk(struct pvr_device *pvr_dev, u32 quirk); diff --git a/drivers/gpu/drm/imagination/pvr_drv.c b/drivers/gpu/drm/imagination/pvr_drv.c index e76363f7dbf7..a6f1d52aaebb 100644 --- a/drivers/gpu/drm/imagination/pvr_drv.c +++ b/drivers/gpu/drm/imagination/pvr_drv.c @@ -7,6 +7,7 @@ #include "pvr_free_list.h" #include "pvr_gem.h" #include "pvr_hwrt.h" +#include "pvr_job.h" #include "pvr_mmu.h" #include "pvr_power.h" #include "pvr_rogue_defs.h" @@ -32,6 +33,8 @@ #include #include #include +#include +#include /** * DOC: PowerVR (Series 6 and later) and IMG Graphics Driver @@ -397,7 +400,8 @@ pvr_dev_query_runtime_info_get(struct pvr_device *pvr_dev, return 0; } - runtime_info.free_list_min_pages = 0; /* FIXME */ + runtime_info.free_list_min_pages = + pvr_get_free_list_min_pages(pvr_dev); runtime_info.free_list_max_pages = ROGUE_PM_MAX_FREELIST_SIZE / ROGUE_PM_PAGE_SIZE; runtime_info.common_store_alloc_region_size = @@ -1137,7 +1141,20 @@ static int pvr_ioctl_submit_jobs(struct drm_device *drm_dev, void *raw_args, struct drm_file *file) { - return -ENOTTY; + struct drm_pvr_ioctl_submit_jobs_args *args = raw_args; + struct pvr_device *pvr_dev = to_pvr_device(drm_dev); + struct pvr_file *pvr_file = to_pvr_file(file); + int idx; + int err; + + if (!drm_dev_enter(drm_dev, &idx)) + return -EIO; + + err = pvr_submit_jobs(pvr_dev, pvr_file, args); + + drm_dev_exit(idx); + + return err; } int @@ -1353,7 +1370,8 @@ pvr_drm_driver_postclose(__always_unused struct drm_device *drm_dev, DEFINE_DRM_GEM_FOPS(pvr_drm_driver_fops); static struct drm_driver pvr_drm_driver = { - .driver_features = DRIVER_GEM | DRIVER_GEM_GPUVA | DRIVER_RENDER, + .driver_features = DRIVER_GEM | DRIVER_GEM_GPUVA | DRIVER_RENDER | + DRIVER_SYNCOBJ | DRIVER_SYNCOBJ_TIMELINE, .open = pvr_drm_driver_open, .postclose = pvr_drm_driver_postclose, .ioctls = pvr_drm_driver_ioctls, @@ -1386,8 +1404,15 @@ pvr_probe(struct platform_device *plat_dev) drm_dev = &pvr_dev->base; platform_set_drvdata(plat_dev, drm_dev); + + init_rwsem(&pvr_dev->reset_sem); + pvr_context_device_init(pvr_dev); + err = pvr_queue_device_init(pvr_dev); + if (err) + goto err_context_fini; + devm_pm_runtime_enable(&plat_dev->dev); pm_runtime_mark_last_busy(&plat_dev->dev); @@ -1404,6 +1429,7 @@ pvr_probe(struct platform_device *plat_dev) goto err_device_fini; xa_init_flags(&pvr_dev->free_list_ids, XA_FLAGS_ALLOC1); + xa_init_flags(&pvr_dev->job_ids, XA_FLAGS_ALLOC1); return 0; @@ -1413,6 +1439,11 @@ pvr_probe(struct platform_device *plat_dev) err_watchdog_fini: pvr_watchdog_fini(pvr_dev); + pvr_queue_device_fini(pvr_dev); + +err_context_fini: + pvr_context_device_fini(pvr_dev); + return err; } @@ -1422,14 +1453,17 @@ pvr_remove(struct platform_device *plat_dev) struct drm_device *drm_dev = platform_get_drvdata(plat_dev); struct pvr_device *pvr_dev = to_pvr_device(drm_dev); + WARN_ON(!xa_empty(&pvr_dev->job_ids)); WARN_ON(!xa_empty(&pvr_dev->free_list_ids)); + xa_destroy(&pvr_dev->job_ids); xa_destroy(&pvr_dev->free_list_ids); pm_runtime_suspend(drm_dev->dev); pvr_device_fini(pvr_dev); drm_dev_unplug(drm_dev); pvr_watchdog_fini(pvr_dev); + pvr_queue_device_fini(pvr_dev); pvr_context_device_fini(pvr_dev); return 0; diff --git a/drivers/gpu/drm/imagination/pvr_job.c b/drivers/gpu/drm/imagination/pvr_job.c new file mode 100644 index 000000000000..9d0812710295 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_job.c @@ -0,0 +1,788 @@ +// SPDX-License-Identifier: GPL-2.0-only OR MIT +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#include "pvr_context.h" +#include "pvr_device.h" +#include "pvr_drv.h" +#include "pvr_gem.h" +#include "pvr_hwrt.h" +#include "pvr_job.h" +#include "pvr_mmu.h" +#include "pvr_power.h" +#include "pvr_rogue_fwif.h" +#include "pvr_rogue_fwif_client.h" +#include "pvr_stream.h" +#include "pvr_stream_defs.h" +#include "pvr_sync.h" + +#include +#include +#include +#include + +static void pvr_job_release(struct kref *kref) +{ + struct pvr_job *job = container_of(kref, struct pvr_job, ref_count); + + xa_erase(&job->pvr_dev->job_ids, job->id); + + pvr_hwrt_data_put(job->hwrt); + pvr_context_put(job->ctx); + + WARN_ON(job->paired_job); + + pvr_queue_job_cleanup(job); + pvr_job_release_pm_ref(job); + + kfree(job->cmd); + kfree(job); +} + +/** + * pvr_job_put() - Release reference on job + * @job: Target job. + */ +void +pvr_job_put(struct pvr_job *job) +{ + if (job) + kref_put(&job->ref_count, pvr_job_release); +} + +/** + * pvr_job_process_stream() - Build job FW structure from stream + * @pvr_dev: Device pointer. + * @cmd_defs: Stream definition. + * @stream: Pointer to command stream. + * @stream_size: Size of command stream, in bytes. + * @job: Pointer to job. + * + * Caller is responsible for freeing the output structure. + * + * Returns: + * * 0 on success, + * * -%ENOMEM on out of memory, or + * * -%EINVAL on malformed stream. + */ +static int +pvr_job_process_stream(struct pvr_device *pvr_dev, const struct pvr_stream_cmd_defs *cmd_defs, + void *stream, u32 stream_size, struct pvr_job *job) +{ + int err; + + job->cmd = kzalloc(cmd_defs->dest_size, GFP_KERNEL); + if (!job->cmd) + return -ENOMEM; + + job->cmd_len = cmd_defs->dest_size; + + err = pvr_stream_process(pvr_dev, cmd_defs, stream, stream_size, job->cmd); + if (err) + kfree(job->cmd); + + return err; +} + +static int pvr_fw_cmd_init(struct pvr_device *pvr_dev, struct pvr_job *job, + const struct pvr_stream_cmd_defs *stream_def, + u64 stream_userptr, u32 stream_len) +{ + void *stream; + int err; + + stream = kzalloc(stream_len, GFP_KERNEL); + if (!stream) + return -ENOMEM; + + if (copy_from_user(stream, u64_to_user_ptr(stream_userptr), stream_len)) { + err = -EFAULT; + goto err_free_stream; + } + + err = pvr_job_process_stream(pvr_dev, stream_def, stream, stream_len, job); + +err_free_stream: + kfree(stream); + + return err; +} + +static u32 +convert_geom_flags(u32 in_flags) +{ + u32 out_flags = 0; + + if (in_flags & DRM_PVR_SUBMIT_JOB_GEOM_CMD_FIRST) + out_flags |= ROGUE_GEOM_FLAGS_FIRSTKICK; + if (in_flags & DRM_PVR_SUBMIT_JOB_GEOM_CMD_LAST) + out_flags |= ROGUE_GEOM_FLAGS_LASTKICK; + if (in_flags & DRM_PVR_SUBMIT_JOB_GEOM_CMD_SINGLE_CORE) + out_flags |= ROGUE_GEOM_FLAGS_SINGLE_CORE; + + return out_flags; +} + +static u32 +convert_frag_flags(u32 in_flags) +{ + u32 out_flags = 0; + + if (in_flags & DRM_PVR_SUBMIT_JOB_FRAG_CMD_SINGLE_CORE) + out_flags |= ROGUE_FRAG_FLAGS_SINGLE_CORE; + if (in_flags & DRM_PVR_SUBMIT_JOB_FRAG_CMD_DEPTHBUFFER) + out_flags |= ROGUE_FRAG_FLAGS_DEPTHBUFFER; + if (in_flags & DRM_PVR_SUBMIT_JOB_FRAG_CMD_STENCILBUFFER) + out_flags |= ROGUE_FRAG_FLAGS_STENCILBUFFER; + if (in_flags & DRM_PVR_SUBMIT_JOB_FRAG_CMD_PREVENT_CDM_OVERLAP) + out_flags |= ROGUE_FRAG_FLAGS_PREVENT_CDM_OVERLAP; + if (in_flags & DRM_PVR_SUBMIT_JOB_FRAG_CMD_SCRATCHBUFFER) + out_flags |= ROGUE_FRAG_FLAGS_SCRATCHBUFFER; + if (in_flags & DRM_PVR_SUBMIT_JOB_FRAG_CMD_GET_VIS_RESULTS) + out_flags |= ROGUE_FRAG_FLAGS_GET_VIS_RESULTS; + if (in_flags & DRM_PVR_SUBMIT_JOB_FRAG_CMD_DISABLE_PIXELMERGE) + out_flags |= ROGUE_FRAG_FLAGS_DISABLE_PIXELMERGE; + + return out_flags; +} + +static int +pvr_geom_job_fw_cmd_init(struct pvr_job *job, + struct drm_pvr_job *args) +{ + struct rogue_fwif_cmd_geom *cmd; + int err; + + if (args->flags & ~DRM_PVR_SUBMIT_JOB_GEOM_CMD_FLAGS_MASK) + return -EINVAL; + + if (job->ctx->type != DRM_PVR_CTX_TYPE_RENDER) + return -EINVAL; + + if (!job->hwrt) + return -EINVAL; + + job->fw_ccb_cmd_type = ROGUE_FWIF_CCB_CMD_TYPE_GEOM; + err = pvr_fw_cmd_init(job->pvr_dev, job, &pvr_cmd_geom_stream, + args->cmd_stream, args->cmd_stream_len); + if (err) + return err; + + cmd = job->cmd; + cmd->cmd_shared.cmn.frame_num = 0; + cmd->flags = convert_geom_flags(args->flags); + pvr_fw_object_get_fw_addr(job->hwrt->fw_obj, &cmd->cmd_shared.hwrt_data_fw_addr); + return 0; +} + +static int +pvr_frag_job_fw_cmd_init(struct pvr_job *job, + struct drm_pvr_job *args) +{ + struct rogue_fwif_cmd_frag *cmd; + int err; + + if (args->flags & ~DRM_PVR_SUBMIT_JOB_FRAG_CMD_FLAGS_MASK) + return -EINVAL; + + if (job->ctx->type != DRM_PVR_CTX_TYPE_RENDER) + return -EINVAL; + + if (!job->hwrt) + return -EINVAL; + + job->fw_ccb_cmd_type = (args->flags & DRM_PVR_SUBMIT_JOB_FRAG_CMD_PARTIAL_RENDER) ? + ROGUE_FWIF_CCB_CMD_TYPE_FRAG_PR : + ROGUE_FWIF_CCB_CMD_TYPE_FRAG; + err = pvr_fw_cmd_init(job->pvr_dev, job, &pvr_cmd_frag_stream, + args->cmd_stream, args->cmd_stream_len); + if (err) + return err; + + cmd = job->cmd; + cmd->cmd_shared.cmn.frame_num = 0; + cmd->flags = convert_frag_flags(args->flags); + pvr_fw_object_get_fw_addr(job->hwrt->fw_obj, &cmd->cmd_shared.hwrt_data_fw_addr); + return 0; +} + +static u32 +convert_compute_flags(u32 in_flags) +{ + u32 out_flags = 0; + + if (in_flags & DRM_PVR_SUBMIT_JOB_COMPUTE_CMD_PREVENT_ALL_OVERLAP) + out_flags |= ROGUE_COMPUTE_FLAG_PREVENT_ALL_OVERLAP; + if (in_flags & DRM_PVR_SUBMIT_JOB_COMPUTE_CMD_SINGLE_CORE) + out_flags |= ROGUE_COMPUTE_FLAG_SINGLE_CORE; + + return out_flags; +} + +static int +pvr_compute_job_fw_cmd_init(struct pvr_job *job, + struct drm_pvr_job *args) +{ + struct rogue_fwif_cmd_compute *cmd; + int err; + + if (args->flags & ~DRM_PVR_SUBMIT_JOB_COMPUTE_CMD_FLAGS_MASK) + return -EINVAL; + + if (job->ctx->type != DRM_PVR_CTX_TYPE_COMPUTE) + return -EINVAL; + + job->fw_ccb_cmd_type = ROGUE_FWIF_CCB_CMD_TYPE_CDM; + err = pvr_fw_cmd_init(job->pvr_dev, job, &pvr_cmd_compute_stream, + args->cmd_stream, args->cmd_stream_len); + if (err) + return err; + + cmd = job->cmd; + cmd->common.frame_num = 0; + cmd->flags = convert_compute_flags(args->flags); + return 0; +} + +static u32 +convert_transfer_flags(u32 in_flags) +{ + u32 out_flags = 0; + + if (in_flags & DRM_PVR_SUBMIT_JOB_TRANSFER_CMD_SINGLE_CORE) + out_flags |= ROGUE_TRANSFER_FLAGS_SINGLE_CORE; + + return out_flags; +} + +static int +pvr_transfer_job_fw_cmd_init(struct pvr_job *job, + struct drm_pvr_job *args) +{ + struct rogue_fwif_cmd_transfer *cmd; + int err; + + if (args->flags & ~DRM_PVR_SUBMIT_JOB_TRANSFER_CMD_FLAGS_MASK) + return -EINVAL; + + if (job->ctx->type != DRM_PVR_CTX_TYPE_TRANSFER_FRAG) + return -EINVAL; + + job->fw_ccb_cmd_type = ROGUE_FWIF_CCB_CMD_TYPE_TQ_3D; + err = pvr_fw_cmd_init(job->pvr_dev, job, &pvr_cmd_transfer_stream, + args->cmd_stream, args->cmd_stream_len); + if (err) + return err; + + cmd = job->cmd; + cmd->common.frame_num = 0; + cmd->flags = convert_transfer_flags(args->flags); + return 0; +} + +static int +pvr_job_fw_cmd_init(struct pvr_job *job, + struct drm_pvr_job *args) +{ + switch (args->type) { + case DRM_PVR_JOB_TYPE_GEOMETRY: + return pvr_geom_job_fw_cmd_init(job, args); + + case DRM_PVR_JOB_TYPE_FRAGMENT: + return pvr_frag_job_fw_cmd_init(job, args); + + case DRM_PVR_JOB_TYPE_COMPUTE: + return pvr_compute_job_fw_cmd_init(job, args); + + case DRM_PVR_JOB_TYPE_TRANSFER_FRAG: + return pvr_transfer_job_fw_cmd_init(job, args); + + default: + return -EINVAL; + } +} + +/** + * struct pvr_job_data - Helper container for pairing jobs with the + * sync_ops supplied for them by the user. + */ +struct pvr_job_data { + /** @job: Pointer to the job. */ + struct pvr_job *job; + + /** @sync_ops: Pointer to the sync_ops associated with @job. */ + struct drm_pvr_sync_op *sync_ops; + + /** @sync_op_count: Number of members of @sync_ops. */ + u32 sync_op_count; +}; + +/** + * prepare_job_syncs() - Prepare all sync objects for a single job. + * @pvr_file: PowerVR file. + * @job_data: Precreated job and sync_ops array. + * @signal_array: xarray to receive signal sync objects. + * + * Returns: + * * 0 on success, or + * * Any error code returned by pvr_sync_signal_array_collect_ops(), + * pvr_sync_add_deps_to_job(), drm_sched_job_add_resv_dependencies() or + * pvr_sync_signal_array_update_fences(). + */ +static int +prepare_job_syncs(struct pvr_file *pvr_file, + struct pvr_job_data *job_data, + struct xarray *signal_array) +{ + struct dma_fence *done_fence; + int err = pvr_sync_signal_array_collect_ops(signal_array, + from_pvr_file(pvr_file), + job_data->sync_op_count, + job_data->sync_ops); + + if (err) + return err; + + err = pvr_sync_add_deps_to_job(pvr_file, &job_data->job->base, + job_data->sync_op_count, + job_data->sync_ops, signal_array); + if (err) + return err; + + if (job_data->job->hwrt) { + /* The geometry job writes the HWRT region headers, which are + * then read by the fragment job. + */ + struct drm_gem_object *obj = + gem_from_pvr_gem(job_data->job->hwrt->fw_obj->gem); + enum dma_resv_usage usage = + dma_resv_usage_rw(job_data->job->type == + DRM_PVR_JOB_TYPE_GEOMETRY); + + dma_resv_lock(obj->resv, NULL); + err = drm_sched_job_add_resv_dependencies(&job_data->job->base, + obj->resv, usage); + dma_resv_unlock(obj->resv); + if (err) + return err; + } + + /* We need to arm the job to get the job done fence. */ + done_fence = pvr_queue_job_arm(job_data->job); + + err = pvr_sync_signal_array_update_fences(signal_array, + job_data->sync_op_count, + job_data->sync_ops, + done_fence); + return err; +} + +/** + * prepare_job_syncs_for_each() - Prepare all sync objects for an array of jobs. + * @file: PowerVR file. + * @job_data: Array of precreated jobs and their sync_ops. + * @job_count: Number of jobs. + * @signal_array: xarray to receive signal sync objects. + * + * Returns: + * * 0 on success, or + * * Any error code returned by pvr_vm_bind_job_prepare_syncs(). + */ +static int +prepare_job_syncs_for_each(struct pvr_file *pvr_file, + struct pvr_job_data *job_data, + u32 *job_count, + struct xarray *signal_array) +{ + for (u32 i = 0; i < *job_count; i++) { + int err = prepare_job_syncs(pvr_file, &job_data[i], + signal_array); + + if (err) { + *job_count = i; + return err; + } + } + + return 0; +} + +static struct pvr_job * +create_job(struct pvr_device *pvr_dev, + struct pvr_file *pvr_file, + struct drm_pvr_job *args) +{ + struct pvr_job *job = NULL; + int err; + + if (!args->cmd_stream || !args->cmd_stream_len) + return ERR_PTR(-EINVAL); + + if (args->type != DRM_PVR_JOB_TYPE_GEOMETRY && + args->type != DRM_PVR_JOB_TYPE_FRAGMENT && + (args->hwrt.set_handle || args->hwrt.data_index)) + return ERR_PTR(-EINVAL); + + job = kzalloc(sizeof(*job), GFP_KERNEL); + if (!job) + return ERR_PTR(-ENOMEM); + + kref_init(&job->ref_count); + job->type = args->type; + job->pvr_dev = pvr_dev; + + err = xa_alloc(&pvr_dev->job_ids, &job->id, job, xa_limit_32b, GFP_KERNEL); + if (err) + goto err_put_job; + + job->ctx = pvr_context_lookup(pvr_file, args->context_handle); + if (!job->ctx) { + err = -EINVAL; + goto err_put_job; + } + + if (args->hwrt.set_handle) { + job->hwrt = pvr_hwrt_data_lookup(pvr_file, args->hwrt.set_handle, + args->hwrt.data_index); + if (!job->hwrt) { + err = -EINVAL; + goto err_put_job; + } + } + + err = pvr_job_fw_cmd_init(job, args); + if (err) + goto err_put_job; + + err = pvr_queue_job_init(job); + if (err) + goto err_put_job; + + return job; + +err_put_job: + pvr_job_put(job); + return ERR_PTR(err); +} + +/** + * pvr_job_data_fini() - Cleanup all allocs used to set up job submission. + * @job_data: Job data array. + * @job_count: Number of members of @job_data. + */ +static void +pvr_job_data_fini(struct pvr_job_data *job_data, u32 job_count) +{ + for (u32 i = 0; i < job_count; i++) { + pvr_job_put(job_data[i].job); + kvfree(job_data[i].sync_ops); + } +} + +/** + * pvr_job_data_init() - Init an array of created jobs, associating them with + * the appropriate sync_ops args, which will be copied in. + * @pvr_dev: Target PowerVR device. + * @pvr_file: Pointer to PowerVR file structure. + * @job_args: Job args array copied from user. + * @job_count: Number of members of @job_args. + * @job_data_out: Job data array. + */ +static int pvr_job_data_init(struct pvr_device *pvr_dev, + struct pvr_file *pvr_file, + struct drm_pvr_job *job_args, + u32 *job_count, + struct pvr_job_data *job_data_out) +{ + int err = 0, i = 0; + + for (; i < *job_count; i++) { + job_data_out[i].job = + create_job(pvr_dev, pvr_file, &job_args[i]); + err = PTR_ERR_OR_ZERO(job_data_out[i].job); + + if (err) { + *job_count = i; + job_data_out[i].job = NULL; + goto err_cleanup; + } + + err = PVR_UOBJ_GET_ARRAY(job_data_out[i].sync_ops, + &job_args[i].sync_ops); + if (err) { + *job_count = i; + + /* Ensure the job created above is also cleaned up. */ + i++; + goto err_cleanup; + } + + job_data_out[i].sync_op_count = job_args[i].sync_ops.count; + } + + return 0; + +err_cleanup: + pvr_job_data_fini(job_data_out, i); + + return err; +} + +static void +push_jobs(struct pvr_job_data *job_data, u32 job_count) +{ + for (u32 i = 0; i < job_count; i++) + pvr_queue_job_push(job_data[i].job); +} + +static int +prepare_fw_obj_resv(struct drm_exec *exec, struct pvr_fw_object *fw_obj) +{ + return drm_exec_prepare_obj(exec, gem_from_pvr_gem(fw_obj->gem), 1); +} + +static int +jobs_lock_all_objs(struct drm_exec *exec, struct pvr_job_data *job_data, + u32 job_count) +{ + for (u32 i = 0; i < job_count; i++) { + struct pvr_job *job = job_data[i].job; + + /* Grab a lock on a the context, to guard against + * concurrent submission to the same queue. + */ + int err = drm_exec_lock_obj(exec, + gem_from_pvr_gem(job->ctx->fw_obj->gem)); + + if (err) + return err; + + if (job->hwrt) { + err = prepare_fw_obj_resv(exec, + job->hwrt->fw_obj); + if (err) + return err; + } + } + + return 0; +} + +static int +prepare_job_resvs_for_each(struct drm_exec *exec, struct pvr_job_data *job_data, + u32 job_count) +{ + drm_exec_until_all_locked(exec) { + int err = jobs_lock_all_objs(exec, job_data, job_count); + + drm_exec_retry_on_contention(exec); + if (err) + return err; + } + + return 0; +} + +static void +update_job_resvs(struct pvr_job *job) +{ + if (job->hwrt) { + enum dma_resv_usage usage = job->type == DRM_PVR_JOB_TYPE_GEOMETRY ? + DMA_RESV_USAGE_WRITE : DMA_RESV_USAGE_READ; + struct drm_gem_object *obj = gem_from_pvr_gem(job->hwrt->fw_obj->gem); + + dma_resv_add_fence(obj->resv, &job->base.s_fence->finished, usage); + } +} + +static void +update_job_resvs_for_each(struct pvr_job_data *job_data, u32 job_count) +{ + for (u32 i = 0; i < job_count; i++) + update_job_resvs(job_data[i].job); +} + +static bool can_combine_jobs(struct pvr_job *a, struct pvr_job *b) +{ + struct pvr_job *geom_job = a, *frag_job = b; + struct dma_fence *fence; + unsigned long index; + + /* Geometry and fragment jobs can be combined if they are queued to the + * same context and targeting the same HWRT. + */ + if (a->type != DRM_PVR_JOB_TYPE_GEOMETRY || + b->type != DRM_PVR_JOB_TYPE_FRAGMENT || + a->ctx != b->ctx || + a->hwrt != b->hwrt) + return false; + + xa_for_each(&frag_job->base.dependencies, index, fence) { + /* We combine when we see an explicit geom -> frag dep. */ + if (&geom_job->base.s_fence->scheduled == fence) + return true; + } + + return false; +} + +static struct dma_fence * +get_last_queued_job_scheduled_fence(struct pvr_queue *queue, + struct pvr_job_data *job_data, + u32 cur_job_pos) +{ + /* We iterate over the current job array in reverse order to grab the + * last to-be-queued job targeting the same queue. + */ + for (u32 i = cur_job_pos; i > 0; i--) { + struct pvr_job *job = job_data[i - 1].job; + + if (job->ctx == queue->ctx && job->type == queue->type) + return dma_fence_get(&job->base.s_fence->scheduled); + } + + /* If we didn't find any, we just return the last queued job scheduled + * fence attached to the queue. + */ + return dma_fence_get(queue->last_queued_job_scheduled_fence); +} + +static int +pvr_jobs_link_geom_frag(struct pvr_job_data *job_data, u32 *job_count) +{ + for (u32 i = 0; i < *job_count - 1; i++) { + struct pvr_job *geom_job = job_data[i].job; + struct pvr_job *frag_job = job_data[i + 1].job; + struct pvr_queue *frag_queue; + struct dma_fence *f; + + if (!can_combine_jobs(job_data[i].job, job_data[i + 1].job)) + continue; + + /* The fragment job will be submitted by the geometry queue. We + * need to make sure it comes after all the other fragment jobs + * queued before it. + */ + frag_queue = pvr_context_get_queue_for_job(frag_job->ctx, + frag_job->type); + f = get_last_queued_job_scheduled_fence(frag_queue, job_data, + i); + if (f) { + int err = drm_sched_job_add_dependency(&geom_job->base, + f); + if (err) { + *job_count = i; + return err; + } + } + + /* The KCCB slot will be reserved by the geometry job, so we can + * drop the KCCB fence on the fragment job. + */ + pvr_kccb_fence_put(frag_job->kccb_fence); + frag_job->kccb_fence = NULL; + + geom_job->paired_job = frag_job; + frag_job->paired_job = geom_job; + + /* Skip the fragment job we just paired to the geometry job. */ + i++; + } + + return 0; +} + +/** + * pvr_submit_jobs() - Submit jobs to the GPU + * @pvr_dev: Target PowerVR device. + * @pvr_file: Pointer to PowerVR file structure. + * @args: Ioctl args. + * @job_count: Number of jobs in @jobs_args. On error this will be updated + * with the index into @jobs_args where the error occurred. + * + * This initial implementation is entirely synchronous; on return the GPU will + * be idle. This will not be the case for future implementations. + * + * Returns: + * * 0 on success, + * * -%EFAULT if arguments can not be copied from user space, or + * * -%EINVAL on invalid arguments, or + * * Any other error. + */ +int +pvr_submit_jobs(struct pvr_device *pvr_dev, struct pvr_file *pvr_file, + struct drm_pvr_ioctl_submit_jobs_args *args) +{ + struct pvr_job_data *job_data = NULL; + struct drm_pvr_job *job_args; + struct xarray signal_array; + u32 jobs_alloced = 0; + struct drm_exec exec; + int err; + + if (!args->jobs.count) + return -EINVAL; + + err = PVR_UOBJ_GET_ARRAY(job_args, &args->jobs); + if (err) + return err; + + job_data = kvmalloc_array(args->jobs.count, sizeof(*job_data), + GFP_KERNEL | __GFP_ZERO); + if (!job_data) { + err = -ENOMEM; + goto out_free; + } + + err = pvr_job_data_init(pvr_dev, pvr_file, job_args, &args->jobs.count, + job_data); + if (err) + goto out_free; + + jobs_alloced = args->jobs.count; + + /* + * Flush MMU if needed - this has been deferred until now to avoid + * overuse of this expensive operation. + */ + err = pvr_mmu_flush_exec(pvr_dev, false); + if (err) + goto out_job_data_cleanup; + + drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT | DRM_EXEC_IGNORE_DUPLICATES); + + xa_init_flags(&signal_array, XA_FLAGS_ALLOC); + + err = prepare_job_syncs_for_each(pvr_file, job_data, &args->jobs.count, + &signal_array); + if (err) + goto out_exec_fini; + + err = prepare_job_resvs_for_each(&exec, job_data, args->jobs.count); + if (err) + goto out_exec_fini; + + err = pvr_jobs_link_geom_frag(job_data, &args->jobs.count); + if (err) + goto out_exec_fini; + + /* Anything after that point must succeed because we start exposing job + * finished fences to the outside world. + */ + update_job_resvs_for_each(job_data, args->jobs.count); + push_jobs(job_data, args->jobs.count); + pvr_sync_signal_array_push_fences(&signal_array); + err = 0; + +out_exec_fini: + drm_exec_fini(&exec); + pvr_sync_signal_array_cleanup(&signal_array); + +out_job_data_cleanup: + pvr_job_data_fini(job_data, jobs_alloced); + +out_free: + kvfree(job_data); + kvfree(job_args); + + return err; +} diff --git a/drivers/gpu/drm/imagination/pvr_job.h b/drivers/gpu/drm/imagination/pvr_job.h new file mode 100644 index 000000000000..0ca003c5c475 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_job.h @@ -0,0 +1,161 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_JOB_H +#define PVR_JOB_H + +#include + +#include +#include + +#include +#include + +#include "pvr_power.h" + +/* Forward declaration from "pvr_context.h". */ +struct pvr_context; + +/* Forward declarations from "pvr_device.h". */ +struct pvr_device; +struct pvr_file; + +/* Forward declarations from "pvr_hwrt.h". */ +struct pvr_hwrt_data; + +/* Forward declaration from "pvr_queue.h". */ +struct pvr_queue; + +struct pvr_job { + /** @base: drm_sched_job object. */ + struct drm_sched_job base; + + /** @ref_count: Refcount for job. */ + struct kref ref_count; + + /** @type: Type of job. */ + enum drm_pvr_job_type type; + + /** @id: Job ID number. */ + u32 id; + + /** + * @paired_job: Job paired to this job. + * + * This field is only meaningful for geometry and fragment jobs. + * + * Paired jobs are executed on the same context, and need to be submitted + * atomically to the FW, to make sure the partial render logic has a + * fragment job to execute when the Parameter Manager runs out of memory. + * + * The geometry job should point to the fragment job it's paired with, + * and the fragment job should point to the geometry job it's paired with. + */ + struct pvr_job *paired_job; + + /** @cccb_fence: Fence used to wait for CCCB space. */ + struct dma_fence *cccb_fence; + + /** @kccb_fence: Fence used to wait for KCCB space. */ + struct dma_fence *kccb_fence; + + /** @done_fence: Fence to signal when the job is done. */ + struct dma_fence *done_fence; + + /** @pvr_dev: Device pointer. */ + struct pvr_device *pvr_dev; + + /** @ctx: Pointer to owning context. */ + struct pvr_context *ctx; + + /** @cmd: Command data. Format depends on @type. */ + void *cmd; + + /** @cmd_len: Length of command data, in bytes. */ + u32 cmd_len; + + /** + * @fw_ccb_cmd_type: Firmware CCB command type. Must be one of %ROGUE_FWIF_CCB_CMD_TYPE_*. + */ + u32 fw_ccb_cmd_type; + + /** @hwrt: HWRT object. Will be NULL for compute and transfer jobs. */ + struct pvr_hwrt_data *hwrt; + + /** + * @has_pm_ref: True if the job has a power ref, thus forcing the GPU to stay on until + * the job is done. + */ + bool has_pm_ref; +}; + +/** + * pvr_job_get() - Take additional reference on job. + * @job: Job pointer. + * + * Call pvr_job_put() to release. + * + * Returns: + * * The requested job on success, or + * * %NULL if no job pointer passed. + */ +static __always_inline struct pvr_job * +pvr_job_get(struct pvr_job *job) +{ + if (job) + kref_get(&job->ref_count); + + return job; +} + +void pvr_job_put(struct pvr_job *job); + +/** + * pvr_job_release_pm_ref() - Release the PM ref if the job acquired it. + * @job: The job to release the PM ref on. + */ +static __always_inline void +pvr_job_release_pm_ref(struct pvr_job *job) +{ + if (job->has_pm_ref) { + pvr_power_put(job->pvr_dev); + job->has_pm_ref = false; + } +} + +/** + * pvr_job_get_pm_ref() - Get a PM ref and attach it to the job. + * @job: The job to attach the PM ref to. + * + * Return: + * * 0 on success, or + * * Any error returned by pvr_power_get() otherwise. + */ +static __always_inline int +pvr_job_get_pm_ref(struct pvr_job *job) +{ + int err; + + if (job->has_pm_ref) + return 0; + + err = pvr_power_get(job->pvr_dev); + if (!err) + job->has_pm_ref = true; + + return err; +} + +int pvr_job_wait_first_non_signaled_native_dep(struct pvr_job *job); + +bool pvr_job_non_native_deps_done(struct pvr_job *job); + +int pvr_job_fits_in_cccb(struct pvr_job *job, unsigned long native_dep_count); + +void pvr_job_submit(struct pvr_job *job); + +int pvr_submit_jobs(struct pvr_device *pvr_dev, struct pvr_file *pvr_file, + struct drm_pvr_ioctl_submit_jobs_args *args); + +#endif /* PVR_JOB_H */ diff --git a/drivers/gpu/drm/imagination/pvr_power.c b/drivers/gpu/drm/imagination/pvr_power.c index 5f7cb7beb879..ba7816fd28ec 100644 --- a/drivers/gpu/drm/imagination/pvr_power.c +++ b/drivers/gpu/drm/imagination/pvr_power.c @@ -5,6 +5,7 @@ #include "pvr_fw.h" #include "pvr_fw_startstop.h" #include "pvr_power.h" +#include "pvr_queue.h" #include "pvr_rogue_fwif.h" #include @@ -155,6 +156,21 @@ pvr_watchdog_kccb_stalled(struct pvr_device *pvr_dev) pvr_dev->watchdog.kccb_stall_count = 0; return true; } + } else if (pvr_dev->watchdog.old_kccb_cmds_executed == kccb_cmds_executed) { + bool has_active_contexts; + + mutex_lock(&pvr_dev->queues.lock); + has_active_contexts = list_empty(&pvr_dev->queues.active); + mutex_unlock(&pvr_dev->queues.lock); + + if (has_active_contexts) { + /* Send a HEALTH_CHECK command so we can verify FW is still alive. */ + struct rogue_fwif_kccb_cmd health_check_cmd; + + health_check_cmd.cmd_type = ROGUE_FWIF_KCCB_CMD_HEALTH_CHECK; + + pvr_kccb_send_cmd_powered(pvr_dev, &health_check_cmd, NULL); + } } else { pvr_dev->watchdog.old_kccb_cmds_executed = kccb_cmds_executed; pvr_dev->watchdog.kccb_stall_count = 0; @@ -318,6 +334,7 @@ pvr_power_device_idle(struct device *dev) int pvr_power_reset(struct pvr_device *pvr_dev, bool hard_reset) { + bool queues_disabled = false; int err; /* @@ -337,6 +354,11 @@ pvr_power_reset(struct pvr_device *pvr_dev, bool hard_reset) disable_irq(pvr_dev->irq); do { + if (hard_reset) { + pvr_queue_device_pre_reset(pvr_dev); + queues_disabled = true; + } + err = pvr_power_fw_disable(pvr_dev, hard_reset); if (!err) { if (hard_reset) { @@ -372,6 +394,9 @@ pvr_power_reset(struct pvr_device *pvr_dev, bool hard_reset) } } while (err); + if (queues_disabled) + pvr_queue_device_post_reset(pvr_dev); + enable_irq(pvr_dev->irq); up_write(&pvr_dev->reset_sem); @@ -386,6 +411,9 @@ pvr_power_reset(struct pvr_device *pvr_dev, bool hard_reset) /* Leave IRQs disabled if the device is lost. */ + if (queues_disabled) + pvr_queue_device_post_reset(pvr_dev); + err_up_write: up_write(&pvr_dev->reset_sem); diff --git a/drivers/gpu/drm/imagination/pvr_queue.c b/drivers/gpu/drm/imagination/pvr_queue.c new file mode 100644 index 000000000000..d65c3fbedf5a --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_queue.c @@ -0,0 +1,1432 @@ +// SPDX-License-Identifier: GPL-2.0-only OR MIT +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#include +#include + +#include "pvr_cccb.h" +#include "pvr_context.h" +#include "pvr_device.h" +#include "pvr_drv.h" +#include "pvr_job.h" +#include "pvr_queue.h" +#include "pvr_vm.h" + +#include "pvr_rogue_fwif_client.h" + +#define MAX_DEADLINE_MS 30000 + +#define CTX_COMPUTE_CCCB_SIZE_LOG2 15 +#define CTX_FRAG_CCCB_SIZE_LOG2 15 +#define CTX_GEOM_CCCB_SIZE_LOG2 15 +#define CTX_TRANSFER_CCCB_SIZE_LOG2 15 + +static int get_xfer_ctx_state_size(struct pvr_device *pvr_dev) +{ + u32 num_isp_store_registers; + + if (PVR_HAS_FEATURE(pvr_dev, xe_memory_hierarchy)) { + num_isp_store_registers = 1; + } else { + int err; + + err = PVR_FEATURE_VALUE(pvr_dev, num_isp_ipp_pipes, &num_isp_store_registers); + if (WARN_ON(err)) + return err; + } + + return sizeof(struct rogue_fwif_frag_ctx_state) + + (num_isp_store_registers * + sizeof(((struct rogue_fwif_frag_ctx_state *)0)->frag_reg_isp_store[0])); +} + +static int get_frag_ctx_state_size(struct pvr_device *pvr_dev) +{ + u32 num_isp_store_registers; + int err; + + if (PVR_HAS_FEATURE(pvr_dev, xe_memory_hierarchy)) { + err = PVR_FEATURE_VALUE(pvr_dev, num_raster_pipes, &num_isp_store_registers); + if (WARN_ON(err)) + return err; + + if (PVR_HAS_FEATURE(pvr_dev, gpu_multicore_support)) { + u32 xpu_max_slaves; + + err = PVR_FEATURE_VALUE(pvr_dev, xpu_max_slaves, &xpu_max_slaves); + if (WARN_ON(err)) + return err; + + num_isp_store_registers *= (1 + xpu_max_slaves); + } + } else { + err = PVR_FEATURE_VALUE(pvr_dev, num_isp_ipp_pipes, &num_isp_store_registers); + if (WARN_ON(err)) + return err; + } + + return sizeof(struct rogue_fwif_frag_ctx_state) + + (num_isp_store_registers * + sizeof(((struct rogue_fwif_frag_ctx_state *)0)->frag_reg_isp_store[0])); +} + +static int get_ctx_state_size(struct pvr_device *pvr_dev, enum drm_pvr_job_type type) +{ + switch (type) { + case DRM_PVR_JOB_TYPE_GEOMETRY: + return sizeof(struct rogue_fwif_geom_ctx_state); + case DRM_PVR_JOB_TYPE_FRAGMENT: + return get_frag_ctx_state_size(pvr_dev); + case DRM_PVR_JOB_TYPE_COMPUTE: + return sizeof(struct rogue_fwif_compute_ctx_state); + case DRM_PVR_JOB_TYPE_TRANSFER_FRAG: + return get_xfer_ctx_state_size(pvr_dev); + } + + WARN(1, "Invalid queue type"); + return -EINVAL; +} + +static u32 get_ctx_offset(enum drm_pvr_job_type type) +{ + switch (type) { + case DRM_PVR_JOB_TYPE_GEOMETRY: + return offsetof(struct rogue_fwif_fwrendercontext, geom_context); + case DRM_PVR_JOB_TYPE_FRAGMENT: + return offsetof(struct rogue_fwif_fwrendercontext, frag_context); + case DRM_PVR_JOB_TYPE_COMPUTE: + return offsetof(struct rogue_fwif_fwcomputecontext, cdm_context); + case DRM_PVR_JOB_TYPE_TRANSFER_FRAG: + return offsetof(struct rogue_fwif_fwtransfercontext, tq_context); + } + + return 0; +} + +static const char * +pvr_queue_fence_get_driver_name(struct dma_fence *f) +{ + return PVR_DRIVER_NAME; +} + +static void pvr_queue_fence_release(struct dma_fence *f) +{ + struct pvr_queue_fence *fence = container_of(f, struct pvr_queue_fence, base); + + pvr_context_put(fence->queue->ctx); + dma_fence_free(f); +} + +static const char * +pvr_queue_job_fence_get_timeline_name(struct dma_fence *f) +{ + struct pvr_queue_fence *fence = container_of(f, struct pvr_queue_fence, base); + + switch (fence->queue->type) { + case DRM_PVR_JOB_TYPE_GEOMETRY: + return "geometry"; + + case DRM_PVR_JOB_TYPE_FRAGMENT: + return "fragment"; + + case DRM_PVR_JOB_TYPE_COMPUTE: + return "compute"; + + case DRM_PVR_JOB_TYPE_TRANSFER_FRAG: + return "transfer"; + } + + WARN(1, "Invalid queue type"); + return "invalid"; +} + +static const char * +pvr_queue_cccb_fence_get_timeline_name(struct dma_fence *f) +{ + struct pvr_queue_fence *fence = container_of(f, struct pvr_queue_fence, base); + + switch (fence->queue->type) { + case DRM_PVR_JOB_TYPE_GEOMETRY: + return "geometry-cccb"; + + case DRM_PVR_JOB_TYPE_FRAGMENT: + return "fragment-cccb"; + + case DRM_PVR_JOB_TYPE_COMPUTE: + return "compute-cccb"; + + case DRM_PVR_JOB_TYPE_TRANSFER_FRAG: + return "transfer-cccb"; + } + + WARN(1, "Invalid queue type"); + return "invalid"; +} + +static const struct dma_fence_ops pvr_queue_job_fence_ops = { + .get_driver_name = pvr_queue_fence_get_driver_name, + .get_timeline_name = pvr_queue_job_fence_get_timeline_name, + .release = pvr_queue_fence_release, +}; + +/** + * to_pvr_queue_job_fence() - Return a pvr_queue_fence object if the fence is + * backed by a UFO. + * @f: The dma_fence to turn into a pvr_queue_fence. + * + * Return: + * * A non-NULL pvr_queue_fence object if the dma_fence is backed by a UFO, or + * * NULL otherwise. + */ +static struct pvr_queue_fence * +to_pvr_queue_job_fence(struct dma_fence *f) +{ + struct drm_sched_fence *sched_fence = to_drm_sched_fence(f); + + if (sched_fence) + f = sched_fence->parent; + + if (f && f->ops == &pvr_queue_job_fence_ops) + return container_of(f, struct pvr_queue_fence, base); + + return NULL; +} + +static const struct dma_fence_ops pvr_queue_cccb_fence_ops = { + .get_driver_name = pvr_queue_fence_get_driver_name, + .get_timeline_name = pvr_queue_cccb_fence_get_timeline_name, + .release = pvr_queue_fence_release, +}; + +/** + * pvr_queue_fence_put() - Put wrapper for pvr_queue_fence objects. + * @f: The dma_fence object to put. + * + * If the pvr_queue_fence has been initialized, we call dma_fence_put(), + * otherwise we free the object with dma_fence_free(). This allows us + * to do the right thing before and after pvr_queue_fence_init() had been + * called. + */ +static void pvr_queue_fence_put(struct dma_fence *f) +{ + if (!f) + return; + + if (WARN_ON(f->ops && + f->ops != &pvr_queue_cccb_fence_ops && + f->ops != &pvr_queue_job_fence_ops)) + return; + + /* If the fence hasn't been initialized yet, free the object directly. */ + if (f->ops) + dma_fence_put(f); + else + dma_fence_free(f); +} + +/** + * pvr_queue_fence_alloc() - Allocate a pvr_queue_fence fence object + * + * Call this function to allocate job CCCB and done fences. This only + * allocates the objects. Initialization happens when the underlying + * dma_fence object is to be returned to drm_sched (in prepare_job() or + * run_job()). + * + * Return: + * * A valid pointer if the allocation succeeds, or + * * NULL if the allocation fails. + */ +static struct dma_fence * +pvr_queue_fence_alloc(void) +{ + struct pvr_queue_fence *fence; + + fence = kzalloc(sizeof(*fence), GFP_KERNEL); + if (!fence) + return NULL; + + return &fence->base; +} + +/** + * pvr_queue_fence_init() - Initializes a pvr_queue_fence object. + * @f: The fence to initialize + * @queue: The queue this fence belongs to. + * @fence_ops: The fence operations. + * @fence_ctx: The fence context. + * + * Wrapper around dma_fence_init() that takes care of initializing the + * pvr_queue_fence::queue field too. + */ +static void +pvr_queue_fence_init(struct dma_fence *f, + struct pvr_queue *queue, + const struct dma_fence_ops *fence_ops, + struct pvr_queue_fence_ctx *fence_ctx) +{ + struct pvr_queue_fence *fence = container_of(f, struct pvr_queue_fence, base); + + pvr_context_get(queue->ctx); + fence->queue = queue; + dma_fence_init(&fence->base, fence_ops, + &fence_ctx->lock, fence_ctx->id, + atomic_inc_return(&fence_ctx->seqno)); +} + +/** + * pvr_queue_cccb_fence_init() - Initializes a CCCB fence object. + * @fence: The fence to initialize. + * @queue: The queue this fence belongs to. + * + * Initializes a fence that can be used to wait for CCCB space. + * + * Should be called in the ::prepare_job() path, so the fence returned to + * drm_sched is valid. + */ +static void +pvr_queue_cccb_fence_init(struct dma_fence *fence, struct pvr_queue *queue) +{ + pvr_queue_fence_init(fence, queue, &pvr_queue_cccb_fence_ops, + &queue->cccb_fence_ctx.base); +} + +/** + * pvr_queue_job_fence_init() - Initializes a job done fence object. + * @fence: The fence to initialize. + * @queue: The queue this fence belongs to. + * + * Initializes a fence that will be signaled when the GPU is done executing + * a job. + * + * Should be called *before* the ::run_job() path, so the fence is initialised + * before being placed in the pending_list. + */ +static void +pvr_queue_job_fence_init(struct dma_fence *fence, struct pvr_queue *queue) +{ + pvr_queue_fence_init(fence, queue, &pvr_queue_job_fence_ops, + &queue->job_fence_ctx); +} + +/** + * pvr_queue_fence_ctx_init() - Queue fence context initialization. + * @fence_ctx: The context to initialize + */ +static void +pvr_queue_fence_ctx_init(struct pvr_queue_fence_ctx *fence_ctx) +{ + spin_lock_init(&fence_ctx->lock); + fence_ctx->id = dma_fence_context_alloc(1); + atomic_set(&fence_ctx->seqno, 0); +} + +static u32 ufo_cmds_size(u32 elem_count) +{ + /* We can pass at most ROGUE_FWIF_CCB_CMD_MAX_UFOS per UFO-related command. */ + u32 full_cmd_count = elem_count / ROGUE_FWIF_CCB_CMD_MAX_UFOS; + u32 remaining_elems = elem_count % ROGUE_FWIF_CCB_CMD_MAX_UFOS; + u32 size = full_cmd_count * + pvr_cccb_get_size_of_cmd_with_hdr(ROGUE_FWIF_CCB_CMD_MAX_UFOS * + sizeof(struct rogue_fwif_ufo)); + + if (remaining_elems) { + size += pvr_cccb_get_size_of_cmd_with_hdr(remaining_elems * + sizeof(struct rogue_fwif_ufo)); + } + + return size; +} + +static u32 job_cmds_size(struct pvr_job *job, u32 ufo_wait_count) +{ + /* One UFO cmd for the fence signaling, one UFO cmd per native fence native, + * and a command for the job itself. + */ + return ufo_cmds_size(1) + ufo_cmds_size(ufo_wait_count) + + pvr_cccb_get_size_of_cmd_with_hdr(job->cmd_len); +} + +/** + * job_count_remaining_native_deps() - Count the number of non-signaled native dependencies. + * @job: Job to operate on. + * + * Returns: Number of non-signaled native deps remaining. + */ +static unsigned long job_count_remaining_native_deps(struct pvr_job *job) +{ + unsigned long remaining_count = 0; + struct dma_fence *fence = NULL; + unsigned long index; + + xa_for_each(&job->base.dependencies, index, fence) { + struct pvr_queue_fence *jfence; + + jfence = to_pvr_queue_job_fence(fence); + if (!jfence) + continue; + + if (!dma_fence_is_signaled(&jfence->base)) + remaining_count++; + } + + return remaining_count; +} + +/** + * pvr_queue_get_job_cccb_fence() - Get the CCCB fence attached to a job. + * @queue: The queue this job will be submitted to. + * @job: The job to get the CCCB fence on. + * + * The CCCB fence is a synchronization primitive allowing us to delay job + * submission until there's enough space in the CCCB to submit the job. + * + * Return: + * * NULL if there's enough space in the CCCB to submit this job, or + * * A valid dma_fence object otherwise. + */ +static struct dma_fence * +pvr_queue_get_job_cccb_fence(struct pvr_queue *queue, struct pvr_job *job) +{ + struct pvr_queue_fence *cccb_fence; + unsigned int native_deps_remaining; + + /* If the fence is NULL, that means we already checked that we had + * enough space in the cccb for our job. + */ + if (!job->cccb_fence) + return NULL; + + mutex_lock(&queue->cccb_fence_ctx.job_lock); + + /* Count remaining native dependencies and check if the job fits in the CCCB. */ + native_deps_remaining = job_count_remaining_native_deps(job); + if (pvr_cccb_cmdseq_fits(&queue->cccb, job_cmds_size(job, native_deps_remaining))) { + pvr_queue_fence_put(job->cccb_fence); + job->cccb_fence = NULL; + goto out_unlock; + } + + /* There should be no job attached to the CCCB fence context: + * drm_sched_entity guarantees that jobs are submitted one at a time. + */ + if (WARN_ON(queue->cccb_fence_ctx.job)) + pvr_job_put(queue->cccb_fence_ctx.job); + + queue->cccb_fence_ctx.job = pvr_job_get(job); + + /* Initialize the fence before returning it. */ + cccb_fence = container_of(job->cccb_fence, struct pvr_queue_fence, base); + if (!WARN_ON(cccb_fence->queue)) + pvr_queue_cccb_fence_init(job->cccb_fence, queue); + +out_unlock: + mutex_unlock(&queue->cccb_fence_ctx.job_lock); + + return dma_fence_get(job->cccb_fence); +} + +/** + * pvr_queue_get_job_kccb_fence() - Get the KCCB fence attached to a job. + * @queue: The queue this job will be submitted to. + * @job: The job to get the KCCB fence on. + * + * The KCCB fence is a synchronization primitive allowing us to delay job + * submission until there's enough space in the KCCB to submit the job. + * + * Return: + * * NULL if there's enough space in the KCCB to submit this job, or + * * A valid dma_fence object otherwise. + */ +static struct dma_fence * +pvr_queue_get_job_kccb_fence(struct pvr_queue *queue, struct pvr_job *job) +{ + struct pvr_device *pvr_dev = queue->ctx->pvr_dev; + struct dma_fence *kccb_fence = NULL; + + /* If the fence is NULL, that means we already checked that we had + * enough space in the KCCB for our job. + */ + if (!job->kccb_fence) + return NULL; + + if (!WARN_ON(job->kccb_fence->ops)) { + kccb_fence = pvr_kccb_reserve_slot(pvr_dev, job->kccb_fence); + job->kccb_fence = NULL; + } + + return kccb_fence; +} + +static struct dma_fence * +pvr_queue_get_paired_frag_job_dep(struct pvr_queue *queue, struct pvr_job *job) +{ + struct pvr_job *frag_job = job->type == DRM_PVR_JOB_TYPE_GEOMETRY ? + job->paired_job : NULL; + struct dma_fence *f; + unsigned long index; + + if (!frag_job) + return NULL; + + xa_for_each(&frag_job->base.dependencies, index, f) { + /* Skip already signaled fences. */ + if (dma_fence_is_signaled(f)) + continue; + + /* Skip our own fence. */ + if (f == &job->base.s_fence->scheduled) + continue; + + return dma_fence_get(f); + } + + return frag_job->base.sched->ops->prepare_job(&frag_job->base, &queue->entity); +} + +/** + * pvr_queue_prepare_job() - Return the next internal dependencies expressed as a dma_fence. + * @sched_job: The job to query the next internal dependency on + * @s_entity: The entity this job is queue on. + * + * After iterating over drm_sched_job::dependencies, drm_sched let the driver return + * its own internal dependencies. We use this function to return our internal dependencies. + */ +static struct dma_fence * +pvr_queue_prepare_job(struct drm_sched_job *sched_job, + struct drm_sched_entity *s_entity) +{ + struct pvr_job *job = container_of(sched_job, struct pvr_job, base); + struct pvr_queue *queue = container_of(s_entity, struct pvr_queue, entity); + struct dma_fence *internal_dep = NULL; + + /* + * Initialize the done_fence, so we can signal it. This must be done + * here because otherwise by the time of run_job() the job will end up + * in the pending list without a valid fence. + */ + if (job->type == DRM_PVR_JOB_TYPE_FRAGMENT && job->paired_job) { + /* + * This will be called on a paired fragment job after being + * submitted to firmware. We can tell if this is the case and + * bail early from whether run_job() has been called on the + * geometry job, which would issue a pm ref. + */ + if (job->paired_job->has_pm_ref) + return NULL; + + /* + * In this case we need to use the job's own ctx to initialise + * the done_fence. The other steps are done in the ctx of the + * paired geometry job. + */ + pvr_queue_job_fence_init(job->done_fence, + job->ctx->queues.fragment); + } else { + pvr_queue_job_fence_init(job->done_fence, queue); + } + + /* CCCB fence is used to make sure we have enough space in the CCCB to + * submit our commands. + */ + internal_dep = pvr_queue_get_job_cccb_fence(queue, job); + + /* KCCB fence is used to make sure we have a KCCB slot to queue our + * CMD_KICK. + */ + if (!internal_dep) + internal_dep = pvr_queue_get_job_kccb_fence(queue, job); + + /* Any extra internal dependency should be added here, using the following + * pattern: + * + * if (!internal_dep) + * internal_dep = pvr_queue_get_job_xxxx_fence(queue, job); + */ + + /* The paired job fence should come last, when everything else is ready. */ + if (!internal_dep) + internal_dep = pvr_queue_get_paired_frag_job_dep(queue, job); + + return internal_dep; +} + +/** + * pvr_queue_update_active_state_locked() - Update the queue active state. + * @queue: Queue to update the state on. + * + * Locked version of pvr_queue_update_active_state(). Must be called with + * pvr_device::queue::lock held. + */ +static void pvr_queue_update_active_state_locked(struct pvr_queue *queue) +{ + struct pvr_device *pvr_dev = queue->ctx->pvr_dev; + + lockdep_assert_held(&pvr_dev->queues.lock); + + /* The queue is temporary out of any list when it's being reset, + * we don't want a call to pvr_queue_update_active_state_locked() + * to re-insert it behind our back. + */ + if (list_empty(&queue->node)) + return; + + if (!atomic_read(&queue->in_flight_job_count)) + list_move_tail(&queue->node, &pvr_dev->queues.idle); + else + list_move_tail(&queue->node, &pvr_dev->queues.active); +} + +/** + * pvr_queue_update_active_state() - Update the queue active state. + * @queue: Queue to update the state on. + * + * Active state is based on the in_flight_job_count value. + * + * Updating the active state implies moving the queue in or out of the + * active queue list, which also defines whether the queue is checked + * or not when a FW event is received. + * + * This function should be called any time a job is submitted or it done + * fence is signaled. + */ +static void pvr_queue_update_active_state(struct pvr_queue *queue) +{ + struct pvr_device *pvr_dev = queue->ctx->pvr_dev; + + mutex_lock(&pvr_dev->queues.lock); + pvr_queue_update_active_state_locked(queue); + mutex_unlock(&pvr_dev->queues.lock); +} + +static void pvr_queue_submit_job_to_cccb(struct pvr_job *job) +{ + struct pvr_queue *queue = container_of(job->base.sched, struct pvr_queue, scheduler); + struct rogue_fwif_ufo ufos[ROGUE_FWIF_CCB_CMD_MAX_UFOS]; + struct pvr_cccb *cccb = &queue->cccb; + struct pvr_queue_fence *jfence; + struct dma_fence *fence; + unsigned long index; + u32 ufo_count = 0; + + /* We need to add the queue to the active list before updating the CCCB, + * otherwise we might miss the FW event informing us that something + * happened on this queue. + */ + atomic_inc(&queue->in_flight_job_count); + pvr_queue_update_active_state(queue); + + xa_for_each(&job->base.dependencies, index, fence) { + jfence = to_pvr_queue_job_fence(fence); + if (!jfence) + continue; + + /* Skip the partial render fence, we will place it at the end. */ + if (job->type == DRM_PVR_JOB_TYPE_FRAGMENT && job->paired_job && + &job->paired_job->base.s_fence->scheduled == fence) + continue; + + if (dma_fence_is_signaled(&jfence->base)) + continue; + + pvr_fw_object_get_fw_addr(jfence->queue->timeline_ufo.fw_obj, + &ufos[ufo_count].addr); + ufos[ufo_count++].value = jfence->base.seqno; + + if (ufo_count == ARRAY_SIZE(ufos)) { + pvr_cccb_write_command_with_header(cccb, ROGUE_FWIF_CCB_CMD_TYPE_FENCE_PR, + sizeof(ufos), ufos, 0, 0); + ufo_count = 0; + } + } + + /* Partial render fence goes last. */ + if (job->type == DRM_PVR_JOB_TYPE_FRAGMENT && job->paired_job) { + jfence = to_pvr_queue_job_fence(job->paired_job->done_fence); + if (!WARN_ON(!jfence)) { + pvr_fw_object_get_fw_addr(jfence->queue->timeline_ufo.fw_obj, + &ufos[ufo_count].addr); + ufos[ufo_count++].value = job->paired_job->done_fence->seqno; + } + } + + if (ufo_count) { + pvr_cccb_write_command_with_header(cccb, ROGUE_FWIF_CCB_CMD_TYPE_FENCE_PR, + sizeof(ufos[0]) * ufo_count, ufos, 0, 0); + } + + if (job->type == DRM_PVR_JOB_TYPE_GEOMETRY && job->paired_job) { + struct rogue_fwif_cmd_geom *cmd = job->cmd; + + /* Reference value for the partial render test is the current queue fence + * seqno minus one. + */ + pvr_fw_object_get_fw_addr(queue->timeline_ufo.fw_obj, + &cmd->partial_render_geom_frag_fence.addr); + cmd->partial_render_geom_frag_fence.value = job->done_fence->seqno - 1; + } + + /* Submit job to FW */ + pvr_cccb_write_command_with_header(cccb, job->fw_ccb_cmd_type, job->cmd_len, job->cmd, + job->id, job->id); + + /* Signal the job fence. */ + pvr_fw_object_get_fw_addr(queue->timeline_ufo.fw_obj, &ufos[0].addr); + ufos[0].value = job->done_fence->seqno; + pvr_cccb_write_command_with_header(cccb, ROGUE_FWIF_CCB_CMD_TYPE_UPDATE, + sizeof(ufos[0]), ufos, 0, 0); +} + +/** + * pvr_queue_run_job() - Submit a job to the FW. + * @sched_job: The job to submit. + * + * This function is called when all non-native dependencies have been met and + * when the commands resulting from this job are guaranteed to fit in the CCCB. + */ +static struct dma_fence *pvr_queue_run_job(struct drm_sched_job *sched_job) +{ + struct pvr_job *job = container_of(sched_job, struct pvr_job, base); + struct pvr_device *pvr_dev = job->pvr_dev; + int err; + + /* The fragment job is issued along the geometry job when we use combined + * geom+frag kicks. When we get there, we should simply return the + * done_fence that's been initialized earlier. + */ + if (job->paired_job && job->type == DRM_PVR_JOB_TYPE_FRAGMENT && + job->done_fence->ops) { + return dma_fence_get(job->done_fence); + } + + /* The only kind of jobs that can be paired are geometry and fragment, and + * we bail out early if we see a fragment job that's paired with a geomtry + * job. + * Paired jobs must also target the same context and point to the same + * HWRT. + */ + if (WARN_ON(job->paired_job && + (job->type != DRM_PVR_JOB_TYPE_GEOMETRY || + job->paired_job->type != DRM_PVR_JOB_TYPE_FRAGMENT || + job->hwrt != job->paired_job->hwrt || + job->ctx != job->paired_job->ctx))) + return ERR_PTR(-EINVAL); + + err = pvr_job_get_pm_ref(job); + if (WARN_ON(err)) + return ERR_PTR(err); + + if (job->paired_job) { + err = pvr_job_get_pm_ref(job->paired_job); + if (WARN_ON(err)) + return ERR_PTR(err); + } + + /* Submit our job to the CCCB */ + pvr_queue_submit_job_to_cccb(job); + + if (job->paired_job) { + struct pvr_job *geom_job = job; + struct pvr_job *frag_job = job->paired_job; + struct pvr_queue *geom_queue = job->ctx->queues.geometry; + struct pvr_queue *frag_queue = job->ctx->queues.fragment; + + /* Submit the fragment job along the geometry job and send a combined kick. */ + pvr_queue_submit_job_to_cccb(frag_job); + pvr_cccb_send_kccb_combined_kick(pvr_dev, + &geom_queue->cccb, &frag_queue->cccb, + pvr_context_get_fw_addr(geom_job->ctx) + + geom_queue->ctx_offset, + pvr_context_get_fw_addr(frag_job->ctx) + + frag_queue->ctx_offset, + job->hwrt, + frag_job->fw_ccb_cmd_type == + ROGUE_FWIF_CCB_CMD_TYPE_FRAG_PR); + } else { + struct pvr_queue *queue = container_of(job->base.sched, + struct pvr_queue, scheduler); + + pvr_cccb_send_kccb_kick(pvr_dev, &queue->cccb, + pvr_context_get_fw_addr(job->ctx) + queue->ctx_offset, + job->hwrt); + } + + return dma_fence_get(job->done_fence); +} + +static void pvr_queue_stop(struct pvr_queue *queue, struct pvr_job *bad_job) +{ + drm_sched_stop(&queue->scheduler, bad_job ? &bad_job->base : NULL); +} + +static void pvr_queue_start(struct pvr_queue *queue) +{ + struct pvr_job *job; + + /* Make sure we CPU-signal the UFO object, so other queues don't get + * blocked waiting on it. + */ + *queue->timeline_ufo.value = atomic_read(&queue->job_fence_ctx.seqno); + + list_for_each_entry(job, &queue->scheduler.pending_list, base.list) { + if (dma_fence_is_signaled(job->done_fence)) { + /* Jobs might have completed after drm_sched_stop() was called. + * In that case, re-assign the parent field to the done_fence. + */ + WARN_ON(job->base.s_fence->parent); + job->base.s_fence->parent = dma_fence_get(job->done_fence); + } else { + /* If we had unfinished jobs, flag the entity as guilty so no + * new job can be submitted. + */ + atomic_set(&queue->ctx->faulty, 1); + } + } + + drm_sched_start(&queue->scheduler, true); +} + +/** + * pvr_queue_timedout_job() - Handle a job timeout event. + * @s_job: The job this timeout occurred on. + * + * FIXME: We don't do anything here to unblock the situation, we just stop+start + * the scheduler, and re-assign parent fences in the middle. + * + * Return: + * * DRM_GPU_SCHED_STAT_NOMINAL. + */ +static enum drm_gpu_sched_stat +pvr_queue_timedout_job(struct drm_sched_job *s_job) +{ + struct drm_gpu_scheduler *sched = s_job->sched; + struct pvr_queue *queue = container_of(sched, struct pvr_queue, scheduler); + struct pvr_device *pvr_dev = queue->ctx->pvr_dev; + struct pvr_job *job; + u32 job_count = 0; + + dev_err(sched->dev, "Job timeout\n"); + + /* Before we stop the scheduler, make sure the queue is out of any list, so + * any call to pvr_queue_update_active_state_locked() that might happen + * until the scheduler is really stopped doesn't end up re-inserting the + * queue in the active list. This would cause + * pvr_queue_signal_done_fences() and drm_sched_stop() to race with each + * other when accessing the pending_list, since drm_sched_stop() doesn't + * grab the job_list_lock when modifying the list (it's assuming the + * only other accessor is the scheduler, and it's safe to not grab the + * lock since it's stopped). + */ + mutex_lock(&pvr_dev->queues.lock); + list_del_init(&queue->node); + mutex_unlock(&pvr_dev->queues.lock); + + drm_sched_stop(sched, s_job); + + /* Re-assign job parent fences. */ + list_for_each_entry(job, &sched->pending_list, base.list) { + job->base.s_fence->parent = dma_fence_get(job->done_fence); + job_count++; + } + WARN_ON(atomic_read(&queue->in_flight_job_count) != job_count); + + /* Re-insert the queue in the proper list, and kick a queue processing + * operation if there were jobs pending. + */ + mutex_lock(&pvr_dev->queues.lock); + if (!job_count) { + list_move_tail(&queue->node, &pvr_dev->queues.idle); + } else { + atomic_set(&queue->in_flight_job_count, job_count); + list_move_tail(&queue->node, &pvr_dev->queues.active); + pvr_queue_process(queue); + } + mutex_unlock(&pvr_dev->queues.lock); + + drm_sched_start(sched, true); + + return DRM_GPU_SCHED_STAT_NOMINAL; +} + +/** + * pvr_queue_free_job() - Release the reference the scheduler had on a job object. + * @sched_job: Job object to free. + */ +static void pvr_queue_free_job(struct drm_sched_job *sched_job) +{ + struct pvr_job *job = container_of(sched_job, struct pvr_job, base); + + drm_sched_job_cleanup(sched_job); + job->paired_job = NULL; + pvr_job_put(job); +} + +static const struct drm_sched_backend_ops pvr_queue_sched_ops = { + .prepare_job = pvr_queue_prepare_job, + .run_job = pvr_queue_run_job, + .timedout_job = pvr_queue_timedout_job, + .free_job = pvr_queue_free_job, +}; + +/** + * pvr_queue_fence_is_ufo_backed() - Check if a dma_fence is backed by a UFO object + * @f: Fence to test. + * + * A UFO-backed fence is a fence that can be signaled or waited upon FW-side. + * pvr_job::done_fence objects are backed by the timeline UFO attached to the queue + * they are pushed to, but those fences are not directly exposed to the outside + * world, so we also need to check if the fence we're being passed is a + * drm_sched_fence that was coming from our driver. + */ +bool pvr_queue_fence_is_ufo_backed(struct dma_fence *f) +{ + struct drm_sched_fence *sched_fence = f ? to_drm_sched_fence(f) : NULL; + + if (sched_fence && + sched_fence->sched->ops == &pvr_queue_sched_ops) + return true; + + if (f && f->ops == &pvr_queue_job_fence_ops) + return true; + + return false; +} + +/** + * pvr_queue_signal_done_fences() - Signal done fences. + * @queue: Queue to check. + * + * Signal done fences of jobs whose seqno is less than the current value of + * the UFO object attached to the queue. + */ +static void +pvr_queue_signal_done_fences(struct pvr_queue *queue) +{ + struct pvr_job *job, *tmp_job; + u32 cur_seqno; + + spin_lock(&queue->scheduler.job_list_lock); + cur_seqno = *queue->timeline_ufo.value; + list_for_each_entry_safe(job, tmp_job, &queue->scheduler.pending_list, base.list) { + if ((int)(cur_seqno - lower_32_bits(job->done_fence->seqno)) < 0) + break; + + if (!dma_fence_is_signaled(job->done_fence)) { + dma_fence_signal(job->done_fence); + pvr_job_release_pm_ref(job); + atomic_dec(&queue->in_flight_job_count); + } + } + spin_unlock(&queue->scheduler.job_list_lock); +} + +/** + * pvr_queue_check_job_waiting_for_cccb_space() - Check if the job waiting for CCCB space + * can be unblocked + * pushed to the CCCB + * @queue: Queue to check + * + * If we have a job waiting for CCCB, and this job now fits in the CCCB, we signal + * its CCCB fence, which should kick drm_sched. + */ +static void +pvr_queue_check_job_waiting_for_cccb_space(struct pvr_queue *queue) +{ + struct pvr_queue_fence *cccb_fence; + u32 native_deps_remaining; + struct pvr_job *job; + + mutex_lock(&queue->cccb_fence_ctx.job_lock); + job = queue->cccb_fence_ctx.job; + if (!job) + goto out_unlock; + + /* If we have a job attached to the CCCB fence context, its CCCB fence + * shouldn't be NULL. + */ + if (WARN_ON(!job->cccb_fence)) { + job = NULL; + goto out_unlock; + } + + /* If we get there, CCCB fence has to be initialized. */ + cccb_fence = container_of(job->cccb_fence, struct pvr_queue_fence, base); + if (WARN_ON(!cccb_fence->queue)) { + job = NULL; + goto out_unlock; + } + + /* Evict signaled dependencies before checking for CCCB space. + * If the job fits, signal the CCCB fence, this should unblock + * the drm_sched_entity. + */ + native_deps_remaining = job_count_remaining_native_deps(job); + if (!pvr_cccb_cmdseq_fits(&queue->cccb, job_cmds_size(job, native_deps_remaining))) { + job = NULL; + goto out_unlock; + } + + dma_fence_signal(job->cccb_fence); + pvr_queue_fence_put(job->cccb_fence); + job->cccb_fence = NULL; + queue->cccb_fence_ctx.job = NULL; + +out_unlock: + mutex_unlock(&queue->cccb_fence_ctx.job_lock); + + pvr_job_put(job); +} + +/** + * pvr_queue_process() - Process events that happened on a queue. + * @queue: Queue to check + * + * Signal job fences and check if jobs waiting for CCCB space can be unblocked. + */ +void pvr_queue_process(struct pvr_queue *queue) +{ + lockdep_assert_held(&queue->ctx->pvr_dev->queues.lock); + + pvr_queue_check_job_waiting_for_cccb_space(queue); + pvr_queue_signal_done_fences(queue); + pvr_queue_update_active_state_locked(queue); +} + +static u32 get_dm_type(struct pvr_queue *queue) +{ + switch (queue->type) { + case DRM_PVR_JOB_TYPE_GEOMETRY: + return PVR_FWIF_DM_GEOM; + case DRM_PVR_JOB_TYPE_TRANSFER_FRAG: + case DRM_PVR_JOB_TYPE_FRAGMENT: + return PVR_FWIF_DM_FRAG; + case DRM_PVR_JOB_TYPE_COMPUTE: + return PVR_FWIF_DM_CDM; + } + + return ~0; +} + +/** + * init_fw_context() - Initializes the queue part of a FW context. + * @queue: Queue object to initialize the FW context for. + * @fw_ctx_map: The FW context CPU mapping. + * + * FW contexts are containing various states, one of them being a per-queue state + * that needs to be initialized for each queue being exposed by a context. This + * function takes care of that. + */ +static void init_fw_context(struct pvr_queue *queue, void *fw_ctx_map) +{ + struct pvr_context *ctx = queue->ctx; + struct pvr_fw_object *fw_mem_ctx_obj = pvr_vm_get_fw_mem_context(ctx->vm_ctx); + struct rogue_fwif_fwcommoncontext *cctx_fw; + struct pvr_cccb *cccb = &queue->cccb; + + cctx_fw = fw_ctx_map + queue->ctx_offset; + cctx_fw->ccbctl_fw_addr = cccb->ctrl_fw_addr; + cctx_fw->ccb_fw_addr = cccb->cccb_fw_addr; + + cctx_fw->dm = get_dm_type(queue); + cctx_fw->priority = ctx->priority; + cctx_fw->priority_seq_num = 0; + cctx_fw->max_deadline_ms = MAX_DEADLINE_MS; + cctx_fw->pid = task_tgid_nr(current); + cctx_fw->server_common_context_id = ctx->ctx_id; + + pvr_fw_object_get_fw_addr(fw_mem_ctx_obj, &cctx_fw->fw_mem_context_fw_addr); + + pvr_fw_object_get_fw_addr(queue->reg_state_obj, &cctx_fw->context_state_addr); +} + +/** + * pvr_queue_cleanup_fw_context() - Wait for the FW context to be idle and clean it up. + * @queue: Queue on FW context to clean up. + * + * Return: + * * 0 on success, + * * Any error returned by pvr_fw_structure_cleanup() otherwise. + */ +static int pvr_queue_cleanup_fw_context(struct pvr_queue *queue) +{ + if (!queue->ctx->fw_obj) + return 0; + + return pvr_fw_structure_cleanup(queue->ctx->pvr_dev, + ROGUE_FWIF_CLEANUP_FWCOMMONCONTEXT, + queue->ctx->fw_obj, queue->ctx_offset); +} + +/** + * pvr_queue_job_init() - Initialize queue related fields in a pvr_job object. + * @job: The job to initialize. + * + * Bind the job to a queue and allocate memory to guarantee pvr_queue_job_arm() + * and pvr_queue_job_push() can't fail. We also make sure the context type is + * valid and the job can fit in the CCCB. + * + * Return: + * * 0 on success, or + * * An error code if something failed. + */ +int pvr_queue_job_init(struct pvr_job *job) +{ + /* Fragment jobs need at least one native fence wait on the geometry job fence. */ + u32 min_native_dep_count = job->type == DRM_PVR_JOB_TYPE_FRAGMENT ? 1 : 0; + struct pvr_queue *queue; + int err; + + if (atomic_read(&job->ctx->faulty)) + return -EIO; + + queue = pvr_context_get_queue_for_job(job->ctx, job->type); + if (!queue) + return -EINVAL; + + if (!pvr_cccb_cmdseq_can_fit(&queue->cccb, job_cmds_size(job, min_native_dep_count))) + return -E2BIG; + + err = drm_sched_job_init(&job->base, &queue->entity, 1, THIS_MODULE); + if (err) + return err; + + job->cccb_fence = pvr_queue_fence_alloc(); + job->kccb_fence = pvr_kccb_fence_alloc(); + job->done_fence = pvr_queue_fence_alloc(); + if (!job->cccb_fence || !job->kccb_fence || !job->done_fence) + return -ENOMEM; + + return 0; +} + +/** + * pvr_queue_job_arm() - Arm a job object. + * @job: The job to arm. + * + * Initializes fences and return the drm_sched finished fence so it can + * be exposed to the outside world. Once this function is called, you should + * make sure the job is pushed using pvr_queue_job_push(), or guarantee that + * no one grabbed a reference to the returned fence. The latter can happen if + * we do multi-job submission, and something failed when creating/initializing + * a job. In that case, we know the fence didn't leave the driver, and we + * can thus guarantee nobody will wait on an dead fence object. + * + * Return: + * * A dma_fence object. + */ +struct dma_fence *pvr_queue_job_arm(struct pvr_job *job) +{ + drm_sched_job_arm(&job->base); + + return &job->base.s_fence->finished; +} + +/** + * pvr_queue_job_cleanup() - Cleanup fence/scheduler related fields in the job object. + * @job: The job to cleanup. + * + * Should be called in the job release path. + */ +void pvr_queue_job_cleanup(struct pvr_job *job) +{ + pvr_queue_fence_put(job->done_fence); + pvr_queue_fence_put(job->cccb_fence); + pvr_kccb_fence_put(job->kccb_fence); + + if (job->base.s_fence) + drm_sched_job_cleanup(&job->base); +} + +/** + * pvr_queue_job_push() - Push a job to its queue. + * @job: The job to push. + * + * Must be called after pvr_queue_job_init() and after all dependencies + * have been added to the job. This will effectively queue the job to + * the drm_sched_entity attached to the queue. We grab a reference on + * the job object, so the caller is free to drop its reference when it's + * done accessing the job object. + */ +void pvr_queue_job_push(struct pvr_job *job) +{ + struct pvr_queue *queue = container_of(job->base.sched, struct pvr_queue, scheduler); + + /* Keep track of the last queued job scheduled fence for combined submit. */ + dma_fence_put(queue->last_queued_job_scheduled_fence); + queue->last_queued_job_scheduled_fence = dma_fence_get(&job->base.s_fence->scheduled); + + pvr_job_get(job); + drm_sched_entity_push_job(&job->base); +} + +static void reg_state_init(void *cpu_ptr, void *priv) +{ + struct pvr_queue *queue = priv; + + if (queue->type == DRM_PVR_JOB_TYPE_GEOMETRY) { + struct rogue_fwif_geom_ctx_state *geom_ctx_state_fw = cpu_ptr; + + geom_ctx_state_fw->geom_core[0].geom_reg_vdm_call_stack_pointer_init = + queue->callstack_addr; + } +} + +/** + * pvr_queue_create() - Create a queue object. + * @ctx: The context this queue will be attached to. + * @type: The type of jobs being pushed to this queue. + * @args: The arguments passed to the context creation function. + * @fw_ctx_map: CPU mapping of the FW context object. + * + * Create a queue object that will be used to queue and track jobs. + * + * Return: + * * A valid pointer to a pvr_queue object, or + * * An error pointer if the creation/initialization failed. + */ +struct pvr_queue *pvr_queue_create(struct pvr_context *ctx, + enum drm_pvr_job_type type, + struct drm_pvr_ioctl_create_context_args *args, + void *fw_ctx_map) +{ + static const struct { + u32 cccb_size; + const char *name; + } props[] = { + [DRM_PVR_JOB_TYPE_GEOMETRY] = { + .cccb_size = CTX_GEOM_CCCB_SIZE_LOG2, + .name = "geometry", + }, + [DRM_PVR_JOB_TYPE_FRAGMENT] = { + .cccb_size = CTX_FRAG_CCCB_SIZE_LOG2, + .name = "fragment" + }, + [DRM_PVR_JOB_TYPE_COMPUTE] = { + .cccb_size = CTX_COMPUTE_CCCB_SIZE_LOG2, + .name = "compute" + }, + [DRM_PVR_JOB_TYPE_TRANSFER_FRAG] = { + .cccb_size = CTX_TRANSFER_CCCB_SIZE_LOG2, + .name = "transfer_frag" + }, + }; + struct pvr_device *pvr_dev = ctx->pvr_dev; + struct drm_gpu_scheduler *sched; + struct pvr_queue *queue; + int ctx_state_size, err; + void *cpu_map; + + if (WARN_ON(type >= sizeof(props))) + return ERR_PTR(-EINVAL); + + switch (ctx->type) { + case DRM_PVR_CTX_TYPE_RENDER: + if (type != DRM_PVR_JOB_TYPE_GEOMETRY && + type != DRM_PVR_JOB_TYPE_FRAGMENT) + return ERR_PTR(-EINVAL); + break; + case DRM_PVR_CTX_TYPE_COMPUTE: + if (type != DRM_PVR_JOB_TYPE_COMPUTE) + return ERR_PTR(-EINVAL); + break; + case DRM_PVR_CTX_TYPE_TRANSFER_FRAG: + if (type != DRM_PVR_JOB_TYPE_TRANSFER_FRAG) + return ERR_PTR(-EINVAL); + break; + default: + return ERR_PTR(-EINVAL); + } + + ctx_state_size = get_ctx_state_size(pvr_dev, type); + if (ctx_state_size < 0) + return ERR_PTR(ctx_state_size); + + queue = kzalloc(sizeof(*queue), GFP_KERNEL); + if (!queue) + return ERR_PTR(-ENOMEM); + + queue->type = type; + queue->ctx_offset = get_ctx_offset(type); + queue->ctx = ctx; + queue->callstack_addr = args->callstack_addr; + sched = &queue->scheduler; + INIT_LIST_HEAD(&queue->node); + mutex_init(&queue->cccb_fence_ctx.job_lock); + pvr_queue_fence_ctx_init(&queue->cccb_fence_ctx.base); + pvr_queue_fence_ctx_init(&queue->job_fence_ctx); + + err = pvr_cccb_init(pvr_dev, &queue->cccb, props[type].cccb_size, props[type].name); + if (err) + goto err_free_queue; + + err = pvr_fw_object_create(pvr_dev, ctx_state_size, + PVR_BO_FW_FLAGS_DEVICE_UNCACHED, + reg_state_init, queue, &queue->reg_state_obj); + if (err) + goto err_cccb_fini; + + init_fw_context(queue, fw_ctx_map); + + if (type != DRM_PVR_JOB_TYPE_GEOMETRY && type != DRM_PVR_JOB_TYPE_FRAGMENT && + args->callstack_addr) { + err = -EINVAL; + goto err_release_reg_state; + } + + cpu_map = pvr_fw_object_create_and_map(pvr_dev, sizeof(*queue->timeline_ufo.value), + PVR_BO_FW_FLAGS_DEVICE_UNCACHED, + NULL, NULL, &queue->timeline_ufo.fw_obj); + if (IS_ERR(cpu_map)) { + err = PTR_ERR(cpu_map); + goto err_release_reg_state; + } + + queue->timeline_ufo.value = cpu_map; + + err = drm_sched_init(&queue->scheduler, + &pvr_queue_sched_ops, + pvr_dev->sched_wq, 1, 64 * 1024, 1, + msecs_to_jiffies(500), + pvr_dev->sched_wq, NULL, "pvr-queue", + pvr_dev->base.dev); + if (err) + goto err_release_ufo; + + err = drm_sched_entity_init(&queue->entity, + DRM_SCHED_PRIORITY_MIN, + &sched, 1, &ctx->faulty); + if (err) + goto err_sched_fini; + + mutex_lock(&pvr_dev->queues.lock); + list_add_tail(&queue->node, &pvr_dev->queues.idle); + mutex_unlock(&pvr_dev->queues.lock); + + return queue; + +err_sched_fini: + drm_sched_fini(&queue->scheduler); + +err_release_ufo: + pvr_fw_object_unmap_and_destroy(queue->timeline_ufo.fw_obj); + +err_release_reg_state: + pvr_fw_object_destroy(queue->reg_state_obj); + +err_cccb_fini: + pvr_cccb_fini(&queue->cccb); + +err_free_queue: + mutex_destroy(&queue->cccb_fence_ctx.job_lock); + kfree(queue); + + return ERR_PTR(err); +} + +void pvr_queue_device_pre_reset(struct pvr_device *pvr_dev) +{ + struct pvr_queue *queue; + + mutex_lock(&pvr_dev->queues.lock); + list_for_each_entry(queue, &pvr_dev->queues.idle, node) + pvr_queue_stop(queue, NULL); + list_for_each_entry(queue, &pvr_dev->queues.active, node) + pvr_queue_stop(queue, NULL); + mutex_unlock(&pvr_dev->queues.lock); +} + +void pvr_queue_device_post_reset(struct pvr_device *pvr_dev) +{ + struct pvr_queue *queue; + + mutex_lock(&pvr_dev->queues.lock); + list_for_each_entry(queue, &pvr_dev->queues.active, node) + pvr_queue_start(queue); + list_for_each_entry(queue, &pvr_dev->queues.idle, node) + pvr_queue_start(queue); + mutex_unlock(&pvr_dev->queues.lock); +} + +/** + * pvr_queue_kill() - Kill a queue. + * @queue: The queue to kill. + * + * Kill the queue so no new jobs can be pushed. Should be called when the + * context handle is destroyed. The queue object might last longer if jobs + * are still in flight and holding a reference to the context this queue + * belongs to. + */ +void pvr_queue_kill(struct pvr_queue *queue) +{ + drm_sched_entity_destroy(&queue->entity); + dma_fence_put(queue->last_queued_job_scheduled_fence); + queue->last_queued_job_scheduled_fence = NULL; +} + +/** + * pvr_queue_destroy() - Destroy a queue. + * @queue: The queue to destroy. + * + * Cleanup the queue and free the resources attached to it. Should be + * called from the context release function. + */ +void pvr_queue_destroy(struct pvr_queue *queue) +{ + if (!queue) + return; + + mutex_lock(&queue->ctx->pvr_dev->queues.lock); + list_del_init(&queue->node); + mutex_unlock(&queue->ctx->pvr_dev->queues.lock); + + drm_sched_fini(&queue->scheduler); + drm_sched_entity_fini(&queue->entity); + + if (WARN_ON(queue->last_queued_job_scheduled_fence)) + dma_fence_put(queue->last_queued_job_scheduled_fence); + + pvr_queue_cleanup_fw_context(queue); + + pvr_fw_object_unmap_and_destroy(queue->timeline_ufo.fw_obj); + pvr_fw_object_destroy(queue->reg_state_obj); + pvr_cccb_fini(&queue->cccb); + mutex_destroy(&queue->cccb_fence_ctx.job_lock); + kfree(queue); +} + +/** + * pvr_queue_device_init() - Device-level initialization of queue related fields. + * @pvr_dev: The device to initialize. + * + * Initializes all fields related to queue management in pvr_device. + * + * Return: + * * 0 on success, or + * * An error code on failure. + */ +int pvr_queue_device_init(struct pvr_device *pvr_dev) +{ + int err; + + INIT_LIST_HEAD(&pvr_dev->queues.active); + INIT_LIST_HEAD(&pvr_dev->queues.idle); + err = drmm_mutex_init(from_pvr_device(pvr_dev), &pvr_dev->queues.lock); + if (err) + return err; + + pvr_dev->sched_wq = alloc_workqueue("powervr-sched", WQ_UNBOUND, 0); + if (!pvr_dev->sched_wq) + return -ENOMEM; + + return 0; +} + +/** + * pvr_queue_device_fini() - Device-level cleanup of queue related fields. + * @pvr_dev: The device to cleanup. + * + * Cleanup/free all queue-related resources attached to a pvr_device object. + */ +void pvr_queue_device_fini(struct pvr_device *pvr_dev) +{ + destroy_workqueue(pvr_dev->sched_wq); +} diff --git a/drivers/gpu/drm/imagination/pvr_queue.h b/drivers/gpu/drm/imagination/pvr_queue.h new file mode 100644 index 000000000000..b5ce2c742150 --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_queue.h @@ -0,0 +1,169 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_QUEUE_H +#define PVR_QUEUE_H + +#include + +#include "pvr_cccb.h" +#include "pvr_device.h" + +struct pvr_context; +struct pvr_queue; + +/** + * struct pvr_queue_fence_ctx - Queue fence context + * + * Used to implement dma_fence_ops for pvr_job::{done,cccb}_fence. + */ +struct pvr_queue_fence_ctx { + /** @id: Fence context ID allocated with dma_fence_context_alloc(). */ + u64 id; + + /** @seqno: Sequence number incremented each time a fence is created. */ + atomic_t seqno; + + /** @lock: Lock used to synchronize access to fences allocated by this context. */ + spinlock_t lock; +}; + +/** + * struct pvr_queue_cccb_fence_ctx - CCCB fence context + * + * Context used to manage fences controlling access to the CCCB. No fences are + * issued if there's enough space in the CCCB to push job commands. + */ +struct pvr_queue_cccb_fence_ctx { + /** @base: Base queue fence context. */ + struct pvr_queue_fence_ctx base; + + /** + * @job: Job waiting for CCCB space. + * + * Thanks to the serializationg done at the drm_sched_entity level, + * there's no more than one job waiting for CCCB at a given time. + * + * This field is NULL if no jobs are currently waiting for CCCB space. + * + * Must be accessed with @job_lock held. + */ + struct pvr_job *job; + + /** @lock: Lock protecting access to the job object. */ + struct mutex job_lock; +}; + +/** + * struct pvr_queue_fence - Queue fence object + */ +struct pvr_queue_fence { + /** @base: Base dma_fence. */ + struct dma_fence base; + + /** @queue: Queue that created this fence. */ + struct pvr_queue *queue; +}; + +/** + * struct pvr_queue - Job queue + * + * Used to queue and track execution of pvr_job objects. + */ +struct pvr_queue { + /** @scheduler: Single entity scheduler use to push jobs to this queue. */ + struct drm_gpu_scheduler scheduler; + + /** @entity: Scheduling entity backing this queue. */ + struct drm_sched_entity entity; + + /** @type: Type of jobs queued to this queue. */ + enum drm_pvr_job_type type; + + /** @ctx: Context object this queue is bound to. */ + struct pvr_context *ctx; + + /** @node: Used to add the queue to the active/idle queue list. */ + struct list_head node; + + /** + * @in_flight_job_count: Number of jobs submitted to the CCCB that + * have not been processed yet. + */ + atomic_t in_flight_job_count; + + /** + * @cccb_fence_ctx: CCCB fence context. + * + * Used to control access to the CCCB is full, such that we don't + * end up trying to push commands to the CCCB if there's not enough + * space to receive all commands needed for a job to complete. + */ + struct pvr_queue_cccb_fence_ctx cccb_fence_ctx; + + /** @job_fence_ctx: Job fence context object. */ + struct pvr_queue_fence_ctx job_fence_ctx; + + /** @timeline_ufo: Timeline UFO for the context queue. */ + struct { + /** @fw_obj: FW object representing the UFO value. */ + struct pvr_fw_object *fw_obj; + + /** @value: CPU mapping of the UFO value. */ + u32 *value; + } timeline_ufo; + + /** + * last_queued_job_scheduled_fence: The scheduled fence of the last + * job queued to this queue. + * + * We use it to insert frag -> geom dependencies when issuing combined + * geom+frag jobs, to guarantee that the fragment job that's part of + * the combined operation comes after all fragment jobs that were queued + * before it. + */ + struct dma_fence *last_queued_job_scheduled_fence; + + /** @cccb: Client Circular Command Buffer. */ + struct pvr_cccb cccb; + + /** @reg_state_obj: FW object representing the register state of this queue. */ + struct pvr_fw_object *reg_state_obj; + + /** @ctx_offset: Offset of the queue context in the FW context object. */ + u32 ctx_offset; + + /** @callstack_addr: Initial call stack address for register state object. */ + u64 callstack_addr; +}; + +bool pvr_queue_fence_is_ufo_backed(struct dma_fence *f); + +int pvr_queue_job_init(struct pvr_job *job); + +void pvr_queue_job_cleanup(struct pvr_job *job); + +void pvr_queue_job_push(struct pvr_job *job); + +struct dma_fence *pvr_queue_job_arm(struct pvr_job *job); + +struct pvr_queue *pvr_queue_create(struct pvr_context *ctx, + enum drm_pvr_job_type type, + struct drm_pvr_ioctl_create_context_args *args, + void *fw_ctx_map); + +void pvr_queue_kill(struct pvr_queue *queue); + +void pvr_queue_destroy(struct pvr_queue *queue); + +void pvr_queue_process(struct pvr_queue *queue); + +void pvr_queue_device_pre_reset(struct pvr_device *pvr_dev); + +void pvr_queue_device_post_reset(struct pvr_device *pvr_dev); + +int pvr_queue_device_init(struct pvr_device *pvr_dev); + +void pvr_queue_device_fini(struct pvr_device *pvr_dev); + +#endif /* PVR_QUEUE_H */ diff --git a/drivers/gpu/drm/imagination/pvr_stream_defs.c b/drivers/gpu/drm/imagination/pvr_stream_defs.c index 59265a7accce..f8bd1a8c01db 100644 --- a/drivers/gpu/drm/imagination/pvr_stream_defs.c +++ b/drivers/gpu/drm/imagination/pvr_stream_defs.c @@ -43,6 +43,232 @@ * existing parameters, to preserve order. As parameters are naturally aligned, care must be taken * with respect to implicit padding in the stream; padding should be minimised as much as possible. */ +static const struct pvr_stream_def rogue_fwif_cmd_geom_stream[] = { + PVR_STREAM_DEF(rogue_fwif_cmd_geom, regs.vdm_ctrl_stream_base, 64), + PVR_STREAM_DEF(rogue_fwif_cmd_geom, regs.tpu_border_colour_table, 64), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_geom, regs.vdm_draw_indirect0, 64, + PVR_FEATURE_VDM_DRAWINDIRECT), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_geom, regs.vdm_draw_indirect1, 32, + PVR_FEATURE_VDM_DRAWINDIRECT), + PVR_STREAM_DEF(rogue_fwif_cmd_geom, regs.ppp_ctrl, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_geom, regs.te_psg, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_geom, regs.vdm_context_resume_task0_size, 32), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_geom, regs.vdm_context_resume_task3_size, 32, + PVR_FEATURE_VDM_OBJECT_LEVEL_LLS), + PVR_STREAM_DEF(rogue_fwif_cmd_geom, regs.view_idx, 32), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_geom, regs.pds_coeff_free_prog, 32, + PVR_FEATURE_TESSELLATION), +}; + +static const struct pvr_stream_def rogue_fwif_cmd_geom_stream_brn49927[] = { + PVR_STREAM_DEF(rogue_fwif_cmd_geom, regs.tpu, 32), +}; + +static const struct pvr_stream_ext_def cmd_geom_ext_streams_0[] = { + { + .stream = rogue_fwif_cmd_geom_stream_brn49927, + .stream_len = ARRAY_SIZE(rogue_fwif_cmd_geom_stream_brn49927), + .header_mask = PVR_STREAM_EXTHDR_GEOM0_BRN49927, + .quirk = 49927, + }, +}; + +static const struct pvr_stream_ext_header cmd_geom_ext_headers[] = { + { + .ext_streams = cmd_geom_ext_streams_0, + .ext_streams_num = ARRAY_SIZE(cmd_geom_ext_streams_0), + .valid_mask = PVR_STREAM_EXTHDR_GEOM0_VALID, + }, +}; + +const struct pvr_stream_cmd_defs pvr_cmd_geom_stream = { + .type = PVR_STREAM_TYPE_GEOM, + + .main_stream = rogue_fwif_cmd_geom_stream, + .main_stream_len = ARRAY_SIZE(rogue_fwif_cmd_geom_stream), + + .ext_nr_headers = ARRAY_SIZE(cmd_geom_ext_headers), + .ext_headers = cmd_geom_ext_headers, + + .dest_size = sizeof(struct rogue_fwif_cmd_geom), +}; + +static const struct pvr_stream_def rogue_fwif_cmd_frag_stream[] = { + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.isp_scissor_base, 64), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.isp_dbias_base, 64), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.isp_oclqry_base, 64), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.isp_zlsctl, 64), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.isp_zload_store_base, 64), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.isp_stencil_load_store_base, 64), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_frag, regs.fb_cdc_zls, 64, + PVR_FEATURE_REQUIRES_FB_CDC_ZLS_SETUP), + PVR_STREAM_DEF_ARRAY(rogue_fwif_cmd_frag, regs.pbe_word), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.tpu_border_colour_table, 64), + PVR_STREAM_DEF_ARRAY(rogue_fwif_cmd_frag, regs.pds_bgnd), + PVR_STREAM_DEF_ARRAY(rogue_fwif_cmd_frag, regs.pds_pr_bgnd), + PVR_STREAM_DEF_ARRAY(rogue_fwif_cmd_frag, regs.usc_clear_register), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.usc_pixel_output_ctrl, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.isp_bgobjdepth, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.isp_bgobjvals, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.isp_aa, 32), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_frag, regs.isp_xtp_pipe_enable, 32, + PVR_FEATURE_S7_TOP_INFRASTRUCTURE), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.isp_ctl, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.event_pixel_pds_info, 32), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_frag, regs.pixel_phantom, 32, + PVR_FEATURE_CLUSTER_GROUPING), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.view_idx, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.event_pixel_pds_data, 32), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_frag, regs.isp_oclqry_stride, 32, + PVR_FEATURE_GPU_MULTICORE_SUPPORT), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_frag, regs.isp_zls_pixels, 32, + PVR_FEATURE_ZLS_SUBTILE), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_frag, regs.rgx_cr_blackpearl_fix, 32, + PVR_FEATURE_ISP_ZLS_D24_S8_PACKING_OGL_MODE), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, zls_stride, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_frag, sls_stride, 32), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_frag, execute_count, 32, + PVR_FEATURE_GPU_MULTICORE_SUPPORT), +}; + +static const struct pvr_stream_def rogue_fwif_cmd_frag_stream_brn47217[] = { + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.isp_oclqry_stride, 32), +}; + +static const struct pvr_stream_def rogue_fwif_cmd_frag_stream_brn49927[] = { + PVR_STREAM_DEF(rogue_fwif_cmd_frag, regs.tpu, 32), +}; + +static const struct pvr_stream_ext_def cmd_frag_ext_streams_0[] = { + { + .stream = rogue_fwif_cmd_frag_stream_brn47217, + .stream_len = ARRAY_SIZE(rogue_fwif_cmd_frag_stream_brn47217), + .header_mask = PVR_STREAM_EXTHDR_FRAG0_BRN47217, + .quirk = 47217, + }, + { + .stream = rogue_fwif_cmd_frag_stream_brn49927, + .stream_len = ARRAY_SIZE(rogue_fwif_cmd_frag_stream_brn49927), + .header_mask = PVR_STREAM_EXTHDR_FRAG0_BRN49927, + .quirk = 49927, + }, +}; + +static const struct pvr_stream_ext_header cmd_frag_ext_headers[] = { + { + .ext_streams = cmd_frag_ext_streams_0, + .ext_streams_num = ARRAY_SIZE(cmd_frag_ext_streams_0), + .valid_mask = PVR_STREAM_EXTHDR_FRAG0_VALID, + }, +}; + +const struct pvr_stream_cmd_defs pvr_cmd_frag_stream = { + .type = PVR_STREAM_TYPE_FRAG, + + .main_stream = rogue_fwif_cmd_frag_stream, + .main_stream_len = ARRAY_SIZE(rogue_fwif_cmd_frag_stream), + + .ext_nr_headers = ARRAY_SIZE(cmd_frag_ext_headers), + .ext_headers = cmd_frag_ext_headers, + + .dest_size = sizeof(struct rogue_fwif_cmd_frag), +}; + +static const struct pvr_stream_def rogue_fwif_cmd_compute_stream[] = { + PVR_STREAM_DEF(rogue_fwif_cmd_compute, regs.tpu_border_colour_table, 64), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_compute, regs.cdm_cb_queue, 64, + PVR_FEATURE_CDM_USER_MODE_QUEUE), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_compute, regs.cdm_cb_base, 64, + PVR_FEATURE_CDM_USER_MODE_QUEUE), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_compute, regs.cdm_cb, 64, + PVR_FEATURE_CDM_USER_MODE_QUEUE), + PVR_STREAM_DEF_NOT_FEATURE(rogue_fwif_cmd_compute, regs.cdm_ctrl_stream_base, 64, + PVR_FEATURE_CDM_USER_MODE_QUEUE), + PVR_STREAM_DEF(rogue_fwif_cmd_compute, regs.cdm_context_state_base_addr, 64), + PVR_STREAM_DEF(rogue_fwif_cmd_compute, regs.cdm_resume_pds1, 32), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_compute, regs.cdm_item, 32, + PVR_FEATURE_COMPUTE_MORTON_CAPABLE), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_compute, regs.compute_cluster, 32, + PVR_FEATURE_CLUSTER_GROUPING), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_compute, regs.tpu_tag_cdm_ctrl, 32, + PVR_FEATURE_TPU_DM_GLOBAL_REGISTERS), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_compute, stream_start_offset, 32, + PVR_FEATURE_CDM_USER_MODE_QUEUE), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_compute, execute_count, 32, + PVR_FEATURE_GPU_MULTICORE_SUPPORT), +}; + +static const struct pvr_stream_def rogue_fwif_cmd_compute_stream_brn49927[] = { + PVR_STREAM_DEF(rogue_fwif_cmd_compute, regs.tpu, 32), +}; + +static const struct pvr_stream_ext_def cmd_compute_ext_streams_0[] = { + { + .stream = rogue_fwif_cmd_compute_stream_brn49927, + .stream_len = ARRAY_SIZE(rogue_fwif_cmd_compute_stream_brn49927), + .header_mask = PVR_STREAM_EXTHDR_COMPUTE0_BRN49927, + .quirk = 49927, + }, +}; + +static const struct pvr_stream_ext_header cmd_compute_ext_headers[] = { + { + .ext_streams = cmd_compute_ext_streams_0, + .ext_streams_num = ARRAY_SIZE(cmd_compute_ext_streams_0), + .valid_mask = PVR_STREAM_EXTHDR_COMPUTE0_VALID, + }, +}; + +const struct pvr_stream_cmd_defs pvr_cmd_compute_stream = { + .type = PVR_STREAM_TYPE_COMPUTE, + + .main_stream = rogue_fwif_cmd_compute_stream, + .main_stream_len = ARRAY_SIZE(rogue_fwif_cmd_compute_stream), + + .ext_nr_headers = ARRAY_SIZE(cmd_compute_ext_headers), + .ext_headers = cmd_compute_ext_headers, + + .dest_size = sizeof(struct rogue_fwif_cmd_compute), +}; + +static const struct pvr_stream_def rogue_fwif_cmd_transfer_stream[] = { + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.pds_bgnd0_base, 64), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.pds_bgnd1_base, 64), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.pds_bgnd3_sizeinfo, 64), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.isp_mtile_base, 64), + PVR_STREAM_DEF_ARRAY(rogue_fwif_cmd_transfer, regs.pbe_wordx_mrty), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.isp_bgobjvals, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.usc_pixel_output_ctrl, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.usc_clear_register0, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.usc_clear_register1, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.usc_clear_register2, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.usc_clear_register3, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.isp_mtile_size, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.isp_render_origin, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.isp_ctl, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.isp_aa, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.event_pixel_pds_info, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.event_pixel_pds_code, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.event_pixel_pds_data, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.isp_render, 32), + PVR_STREAM_DEF(rogue_fwif_cmd_transfer, regs.isp_rgn, 32), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_transfer, regs.isp_xtp_pipe_enable, 32, + PVR_FEATURE_S7_TOP_INFRASTRUCTURE), + PVR_STREAM_DEF_FEATURE(rogue_fwif_cmd_transfer, regs.frag_screen, 32, + PVR_FEATURE_GPU_MULTICORE_SUPPORT), +}; + +const struct pvr_stream_cmd_defs pvr_cmd_transfer_stream = { + .type = PVR_STREAM_TYPE_TRANSFER, + + .main_stream = rogue_fwif_cmd_transfer_stream, + .main_stream_len = ARRAY_SIZE(rogue_fwif_cmd_transfer_stream), + + .ext_nr_headers = 0, + + .dest_size = sizeof(struct rogue_fwif_cmd_transfer), +}; + static const struct pvr_stream_def rogue_fwif_static_render_context_state_stream[] = { PVR_STREAM_DEF(rogue_fwif_geom_registers_caswitch, geom_reg_vdm_context_state_base_addr, 64), diff --git a/drivers/gpu/drm/imagination/pvr_sync.c b/drivers/gpu/drm/imagination/pvr_sync.c new file mode 100644 index 000000000000..129f646d14ba --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_sync.c @@ -0,0 +1,289 @@ +// SPDX-License-Identifier: GPL-2.0-only OR MIT +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#include + +#include +#include +#include +#include + +#include "pvr_device.h" +#include "pvr_queue.h" +#include "pvr_sync.h" + +static int +pvr_check_sync_op(const struct drm_pvr_sync_op *sync_op) +{ + u8 handle_type; + + if (sync_op->flags & ~DRM_PVR_SYNC_OP_FLAGS_MASK) + return -EINVAL; + + handle_type = sync_op->flags & DRM_PVR_SYNC_OP_FLAG_HANDLE_TYPE_MASK; + if (handle_type != DRM_PVR_SYNC_OP_FLAG_HANDLE_TYPE_SYNCOBJ && + handle_type != DRM_PVR_SYNC_OP_FLAG_HANDLE_TYPE_TIMELINE_SYNCOBJ) + return -EINVAL; + + if (handle_type == DRM_PVR_SYNC_OP_FLAG_HANDLE_TYPE_SYNCOBJ && + sync_op->value != 0) + return -EINVAL; + + return 0; +} + +static void +pvr_sync_signal_free(struct pvr_sync_signal *sig_sync) +{ + if (!sig_sync) + return; + + drm_syncobj_put(sig_sync->syncobj); + dma_fence_chain_free(sig_sync->chain); + dma_fence_put(sig_sync->fence); + kfree(sig_sync); +} + +void +pvr_sync_signal_array_cleanup(struct xarray *array) +{ + struct pvr_sync_signal *sig_sync; + unsigned long i; + + xa_for_each(array, i, sig_sync) + pvr_sync_signal_free(sig_sync); + + xa_destroy(array); +} + +static struct pvr_sync_signal * +pvr_sync_signal_array_add(struct xarray *array, struct drm_file *file, u32 handle, u64 point) +{ + struct pvr_sync_signal *sig_sync; + struct dma_fence *cur_fence; + int err; + u32 id; + + sig_sync = kzalloc(sizeof(*sig_sync), GFP_KERNEL); + if (!sig_sync) + return ERR_PTR(-ENOMEM); + + sig_sync->handle = handle; + sig_sync->point = point; + + if (point > 0) { + sig_sync->chain = dma_fence_chain_alloc(); + if (!sig_sync->chain) { + err = -ENOMEM; + goto err_free_sig_sync; + } + } + + sig_sync->syncobj = drm_syncobj_find(file, handle); + if (!sig_sync->syncobj) { + err = -EINVAL; + goto err_free_sig_sync; + } + + /* Retrieve the current fence attached to that point. It's + * perfectly fine to get a NULL fence here, it just means there's + * no fence attached to that point yet. + */ + if (!drm_syncobj_find_fence(file, handle, point, 0, &cur_fence)) + sig_sync->fence = cur_fence; + + err = xa_alloc(array, &id, sig_sync, xa_limit_32b, GFP_KERNEL); + if (err) + goto err_free_sig_sync; + + return sig_sync; + +err_free_sig_sync: + pvr_sync_signal_free(sig_sync); + return ERR_PTR(err); +} + +static struct pvr_sync_signal * +pvr_sync_signal_array_search(struct xarray *array, u32 handle, u64 point) +{ + struct pvr_sync_signal *sig_sync; + unsigned long i; + + xa_for_each(array, i, sig_sync) { + if (handle == sig_sync->handle && point == sig_sync->point) + return sig_sync; + } + + return NULL; +} + +static struct pvr_sync_signal * +pvr_sync_signal_array_get(struct xarray *array, struct drm_file *file, u32 handle, u64 point) +{ + struct pvr_sync_signal *sig_sync; + + sig_sync = pvr_sync_signal_array_search(array, handle, point); + if (sig_sync) + return sig_sync; + + return pvr_sync_signal_array_add(array, file, handle, point); +} + +int +pvr_sync_signal_array_collect_ops(struct xarray *array, + struct drm_file *file, + u32 sync_op_count, + const struct drm_pvr_sync_op *sync_ops) +{ + for (u32 i = 0; i < sync_op_count; i++) { + struct pvr_sync_signal *sig_sync; + int ret; + + if (!(sync_ops[i].flags & DRM_PVR_SYNC_OP_FLAG_SIGNAL)) + continue; + + ret = pvr_check_sync_op(&sync_ops[i]); + if (ret) + return ret; + + sig_sync = pvr_sync_signal_array_get(array, file, + sync_ops[i].handle, + sync_ops[i].value); + if (IS_ERR(sig_sync)) + return PTR_ERR(sig_sync); + } + + return 0; +} + +int +pvr_sync_signal_array_update_fences(struct xarray *array, + u32 sync_op_count, + const struct drm_pvr_sync_op *sync_ops, + struct dma_fence *done_fence) +{ + for (u32 i = 0; i < sync_op_count; i++) { + struct dma_fence *old_fence; + struct pvr_sync_signal *sig_sync; + + if (!(sync_ops[i].flags & DRM_PVR_SYNC_OP_FLAG_SIGNAL)) + continue; + + sig_sync = pvr_sync_signal_array_search(array, sync_ops[i].handle, + sync_ops[i].value); + if (WARN_ON(!sig_sync)) + return -EINVAL; + + old_fence = sig_sync->fence; + sig_sync->fence = dma_fence_get(done_fence); + dma_fence_put(old_fence); + + if (WARN_ON(!sig_sync->fence)) + return -EINVAL; + } + + return 0; +} + +void +pvr_sync_signal_array_push_fences(struct xarray *array) +{ + struct pvr_sync_signal *sig_sync; + unsigned long i; + + xa_for_each(array, i, sig_sync) { + if (sig_sync->chain) { + drm_syncobj_add_point(sig_sync->syncobj, sig_sync->chain, + sig_sync->fence, sig_sync->point); + sig_sync->chain = NULL; + } else { + drm_syncobj_replace_fence(sig_sync->syncobj, sig_sync->fence); + } + } +} + +static int +pvr_sync_add_dep_to_job(struct drm_sched_job *job, struct dma_fence *f) +{ + struct dma_fence_unwrap iter; + u32 native_fence_count = 0; + struct dma_fence *uf; + int err = 0; + + dma_fence_unwrap_for_each(uf, &iter, f) { + if (pvr_queue_fence_is_ufo_backed(uf)) + native_fence_count++; + } + + /* No need to unwrap the fence if it's fully non-native. */ + if (!native_fence_count) + return drm_sched_job_add_dependency(job, f); + + dma_fence_unwrap_for_each(uf, &iter, f) { + /* There's no dma_fence_unwrap_stop() helper cleaning up the refs + * owned by dma_fence_unwrap(), so let's just iterate over all + * entries without doing anything when something failed. + */ + if (err) + continue; + + if (pvr_queue_fence_is_ufo_backed(uf)) { + struct drm_sched_fence *s_fence = to_drm_sched_fence(uf); + + /* If this is a native dependency, we wait for the scheduled fence, + * and we will let pvr_queue_run_job() issue FW waits. + */ + err = drm_sched_job_add_dependency(job, + dma_fence_get(&s_fence->scheduled)); + } else { + err = drm_sched_job_add_dependency(job, dma_fence_get(uf)); + } + } + + dma_fence_put(f); + return err; +} + +int +pvr_sync_add_deps_to_job(struct pvr_file *pvr_file, struct drm_sched_job *job, + u32 sync_op_count, + const struct drm_pvr_sync_op *sync_ops, + struct xarray *signal_array) +{ + int err = 0; + + if (!sync_op_count) + return 0; + + for (u32 i = 0; i < sync_op_count; i++) { + struct pvr_sync_signal *sig_sync; + struct dma_fence *fence; + + if (sync_ops[i].flags & DRM_PVR_SYNC_OP_FLAG_SIGNAL) + continue; + + err = pvr_check_sync_op(&sync_ops[i]); + if (err) + return err; + + sig_sync = pvr_sync_signal_array_search(signal_array, sync_ops[i].handle, + sync_ops[i].value); + if (sig_sync) { + if (WARN_ON(!sig_sync->fence)) + return -EINVAL; + + fence = dma_fence_get(sig_sync->fence); + } else { + err = drm_syncobj_find_fence(from_pvr_file(pvr_file), sync_ops[i].handle, + sync_ops[i].value, 0, &fence); + if (err) + return err; + } + + err = pvr_sync_add_dep_to_job(job, fence); + if (err) + return err; + } + + return 0; +} diff --git a/drivers/gpu/drm/imagination/pvr_sync.h b/drivers/gpu/drm/imagination/pvr_sync.h new file mode 100644 index 000000000000..db6ccfda104a --- /dev/null +++ b/drivers/gpu/drm/imagination/pvr_sync.h @@ -0,0 +1,84 @@ +/* SPDX-License-Identifier: GPL-2.0-only OR MIT */ +/* Copyright (c) 2023 Imagination Technologies Ltd. */ + +#ifndef PVR_SYNC_H +#define PVR_SYNC_H + +#include + +/* Forward declaration from . */ +struct xarray; + +/* Forward declaration from . */ +struct drm_file; + +/* Forward declaration from . */ +struct drm_sched_job; + +/* Forward declaration from "pvr_device.h". */ +struct pvr_file; + +/** + * struct pvr_sync_signal - Object encoding a syncobj signal operation + * + * The job submission logic collects all signal operations in an array of + * pvr_sync_signal objects. This array also serves as a cache to get the + * latest dma_fence when multiple jobs are submitted at once, and one job + * signals a syncobj point that's later waited on by a subsequent job. + */ +struct pvr_sync_signal { + /** @handle: Handle of the syncobj to signal. */ + u32 handle; + + /** + * @point: Point to signal in the syncobj. + * + * Only relevant for timeline syncobjs. + */ + u64 point; + + /** @syncobj: Syncobj retrieved from the handle. */ + struct drm_syncobj *syncobj; + + /** + * @chain: Chain object used to link the new fence with the + * existing timeline syncobj. + * + * Should be zero when manipulating a regular syncobj. + */ + struct dma_fence_chain *chain; + + /** + * @fence: New fence object to attach to the syncobj. + * + * This pointer starts with the current fence bound to + * the pair. + */ + struct dma_fence *fence; +}; + +void +pvr_sync_signal_array_cleanup(struct xarray *array); + +int +pvr_sync_signal_array_collect_ops(struct xarray *array, + struct drm_file *file, + u32 sync_op_count, + const struct drm_pvr_sync_op *sync_ops); + +int +pvr_sync_signal_array_update_fences(struct xarray *array, + u32 sync_op_count, + const struct drm_pvr_sync_op *sync_ops, + struct dma_fence *done_fence); + +void +pvr_sync_signal_array_push_fences(struct xarray *array); + +int +pvr_sync_add_deps_to_job(struct pvr_file *pvr_file, struct drm_sched_job *job, + u32 sync_op_count, + const struct drm_pvr_sync_op *sync_ops, + struct xarray *signal_array); + +#endif /* PVR_SYNC_H */ From patchwork Wed Nov 22 16:34:41 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Donald Robson X-Patchwork-Id: 746149 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=imgtec.com header.i=@imgtec.com header.b="BGe0t36C"; dkim=pass (1024-bit key) header.d=IMGTecCRM.onmicrosoft.com header.i=@IMGTecCRM.onmicrosoft.com header.b="EfTBpFV5" Received: from mx08-00376f01.pphosted.com (mx08-00376f01.pphosted.com [91.207.212.86]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id C9F2610CE; Wed, 22 Nov 2023 08:36:03 -0800 (PST) Received: from pps.filterd (m0168888.ppops.net [127.0.0.1]) by mx08-00376f01.pphosted.com (8.17.1.24/8.17.1.24) with ESMTP id 3AMG01mA020623; Wed, 22 Nov 2023 16:35:33 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=imgtec.com; h= from:to:cc:subject:date:message-id:in-reply-to:references :content-transfer-encoding:content-type:mime-version; s= dk201812; bh=29VnZyKrx6FzxdItxhHmlg0RxGYggTpD6CaxDzKxDAU=; b=BGe 0t36Cw4iQnMZCJNUslNHDEqHCV8yZm5Je+Bh+FQK7qFhc5ZUHxaOMx8b/nzIMBja WQmpv7lhSVx+D+mPBkwlndlM9fMGVzOxfdP+L9T0JLcLUw08QT2Q5/cKN7EPTz3I UvLUDLpn66iS2k+e26tCvFdcwlnbHK72di65YCDGXO04gIOVFyAeNIBn25p8Qh6G 45SEzKJswR6T39NvaIkbLLRMYt7pG2y4heptGGJN65E9m2F/b/zd9JL1i6xQUVcM RACt9U1L2xybkw/xLXwFxqSfkL4qkwakcsE2ZYX/feVFxQq4zfXCqoX4muMOfUVg 0Qh4soOhUT0gOP13ITw== Received: from hhmail05.hh.imgtec.org ([217.156.249.195]) by mx08-00376f01.pphosted.com (PPS) with ESMTPS id 3ug99gsf3d-2 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128 verify=NOT); Wed, 22 Nov 2023 16:35:33 +0000 (GMT) Received: from HHMAIL04.hh.imgtec.org (10.100.10.119) by HHMAIL05.hh.imgtec.org (10.100.10.120) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34; Wed, 22 Nov 2023 16:35:32 +0000 Received: from GBR01-LO4-obe.outbound.protection.outlook.com (104.47.85.105) by email.imgtec.com (10.100.10.121) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.34 via Frontend Transport; Wed, 22 Nov 2023 16:35:32 +0000 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=i3lhwbZOo3QMN8gr73bLwNVlr9ts+MBcTYqgjRWHYDceXl21ObvVMHYyeVxuFwHyetGyhWOICLW8Ra0TK1ASymcxqujgeROtOZfNzMx2rOyJsAMyl4hw2VWDYpAV5b+saC07hUjgKOJDr3QwhYNb4B/7sJoEhaKh2JBk6IB1TtiWhVzFJAInRYWl7MUcxoqu8tCipcTzNvv+tU0o71BJ5gdvpaBWhl3K6rKKP72+NJxg337koHcPv5nqG5ma0EZOVjz58rVVRyvoJ5Rm1WWxyu6jwGsWrt+K050WVj8b511NFD/VDy16Crq9LzVOi6w0TRtEawlAWFeAC5Bu7dXopg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=29VnZyKrx6FzxdItxhHmlg0RxGYggTpD6CaxDzKxDAU=; b=VSz/pKy2e7KMKblDOxeWDXpOHctUxrDJxmnSYI3LQ3aCB6KV56YvZBzGuA/YDb855phvEPfZsc0ePq+ymY9ElnAIGG0Y09eKXcLUC7bU9yJGG2KXXnWY/QKa/dotpHn4VRwnoVdxxwBGwcp6pcIUEq8hho2T/yGImL5/wEO9SXdpcwKKo4SKfbXQqb2cFy4woxnWwOic+NVRQukNWJpjI1a3VP2/qsM1Z7URCBfHJU4kf3MfzFRJZ7TsqHBXjftNSk11BY9ej+BfT0t3RwnVF1tzbAyZZd0YE+hMZ/x2SRX15rWBeFvz5gvl/eoc8mR1v7DvvpRJCgqz2v0isZ2jZg== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=imgtec.com; dmarc=pass action=none header.from=imgtec.com; dkim=pass header.d=imgtec.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=IMGTecCRM.onmicrosoft.com; s=selector2-IMGTecCRM-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=29VnZyKrx6FzxdItxhHmlg0RxGYggTpD6CaxDzKxDAU=; b=EfTBpFV5W65NI5imUdkKTRYcbXSN0gfu06Mc9IfqzUwrrhzVZ2F30LY6XhmBpYdS6rtOOsxEo0dvIOejW4FtIQb84oe1i7GLOgC3k0aESzvhcdkqkXRTYlSCMsYYSFmlBbt+OlVz4m8TGdLrpGUlXn9ExcNfUEqXeTjffCdg7fA= Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) by CWLP265MB6864.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1fe::14) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.7025.19; Wed, 22 Nov 2023 16:35:29 +0000 Received: from CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34]) by CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM ([fe80::a85a:76f7:c085:2b34%3]) with mapi id 15.20.7025.017; Wed, 22 Nov 2023 16:35:29 +0000 From: Donald Robson To: CC: , , , , , , , , , , , , , , , , , , , , , , Sarah Walker Subject: [PATCH v9 20/20] drm/imagination: Add driver documentation Date: Wed, 22 Nov 2023 16:34:41 +0000 Message-Id: <76a7b18cfbe93066efcee3311ae795176ce7c65d.1700668843.git.donald.robson@imgtec.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: References: X-ClientProxiedBy: LO4P265CA0273.GBRP265.PROD.OUTLOOK.COM (2603:10a6:600:37a::10) To CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM (2603:10a6:400:1a0::8) Precedence: bulk X-Mailing-List: devicetree@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: CWLP265MB5770:EE_|CWLP265MB6864:EE_ X-MS-Office365-Filtering-Correlation-Id: ccae1df2-facc-4d20-47f3-08dbeb7909cd X-MS-Exchange-SenderADCheck: 1 X-MS-Exchange-AntiSpam-Relay: 0 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-Forefront-Antispam-Report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:; IPV:NLI; SFV:NSPM; H:CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM; PTR:; CAT:NONE; SFS:(13230031)(136003)(366004)(376002)(346002)(39850400004)(396003)(230922051799003)(64100799003)(451199024)(186009)(1800799012)(52116002)(5660300002)(38350700005)(6666004)(6506007)(6512007)(2616005)(8936002)(8676002)(4326008)(2906002)(7416002)(478600001)(66476007)(6916009)(66556008)(107886003)(316002)(26005)(66946007)(6486002)(44832011)(38100700002)(41300700001)(36756003)(86362001); DIR:OUT; SFP:1102; X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: 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 X-MS-Exchange-CrossTenant-Network-Message-Id: ccae1df2-facc-4d20-47f3-08dbeb7909cd X-MS-Exchange-CrossTenant-AuthSource: CWLP265MB5770.GBRP265.PROD.OUTLOOK.COM X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 22 Nov 2023 16:35:29.2162 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 0d5fd8bb-e8c2-4e0a-8dd5-2c264f7140fe X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: r1bnNmhfgSicV0RTifU0N01oHeBs23nX6XNL8ErteBHy9igLxwWQGaH3cLZw7rj1nVo6ppBunZ1axDyt1pA6CW/nDFUrwcqWe5plFeAlk20= X-MS-Exchange-Transport-CrossTenantHeadersStamped: CWLP265MB6864 X-OriginatorOrg: imgtec.com X-EXCLAIMER-MD-CONFIG: 15a78312-3e47-46eb-9010-2e54d84a9631 X-Proofpoint-GUID: 6SN7h277s9s_anW70ov5GjkJmZOQChtn X-Proofpoint-ORIG-GUID: 6SN7h277s9s_anW70ov5GjkJmZOQChtn From: Sarah Walker Add documentation for the UAPI. Changes since v5: - Remove obsolete VM documentation Co-developed-by: Matt Coster Signed-off-by: Matt Coster Co-developed-by: Donald Robson Signed-off-by: Donald Robson Signed-off-by: Sarah Walker Reviewed-by: Maxime Ripard --- Documentation/gpu/drivers.rst | 2 + Documentation/gpu/imagination/index.rst | 13 ++ Documentation/gpu/imagination/uapi.rst | 174 ++++++++++++++++++++++++ MAINTAINERS | 1 + 4 files changed, 190 insertions(+) create mode 100644 Documentation/gpu/imagination/index.rst create mode 100644 Documentation/gpu/imagination/uapi.rst diff --git a/Documentation/gpu/drivers.rst b/Documentation/gpu/drivers.rst index 45a12e552091..cc6535f5f28c 100644 --- a/Documentation/gpu/drivers.rst +++ b/Documentation/gpu/drivers.rst @@ -3,9 +3,11 @@ GPU Driver Documentation ======================== .. toctree:: + :maxdepth: 3 amdgpu/index i915 + imagination/index mcde meson pl111 diff --git a/Documentation/gpu/imagination/index.rst b/Documentation/gpu/imagination/index.rst new file mode 100644 index 000000000000..dc9579e758c3 --- /dev/null +++ b/Documentation/gpu/imagination/index.rst @@ -0,0 +1,13 @@ +======================================= +drm/imagination PowerVR Graphics Driver +======================================= + +.. kernel-doc:: drivers/gpu/drm/imagination/pvr_drv.c + :doc: PowerVR Graphics Driver + +Contents +======== +.. toctree:: + :maxdepth: 2 + + uapi diff --git a/Documentation/gpu/imagination/uapi.rst b/Documentation/gpu/imagination/uapi.rst new file mode 100644 index 000000000000..2227ea7e6222 --- /dev/null +++ b/Documentation/gpu/imagination/uapi.rst @@ -0,0 +1,174 @@ +==== +UAPI +==== +The sources associated with this section can be found in ``pvr_drm.h``. + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: PowerVR UAPI + +OBJECT ARRAYS +============= +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_obj_array + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: DRM_PVR_OBJ_ARRAY + +IOCTLS +====== +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: PowerVR IOCTL interface + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: PVR_IOCTL + +DEV_QUERY +--------- +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: PowerVR IOCTL DEV_QUERY interface + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_dev_query + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_ioctl_dev_query_args + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_dev_query_gpu_info + drm_pvr_dev_query_runtime_info + drm_pvr_dev_query_hwrt_info + drm_pvr_dev_query_quirks + drm_pvr_dev_query_enhancements + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_heap_id + drm_pvr_heap + drm_pvr_dev_query_heap_info + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: Flags for DRM_PVR_DEV_QUERY_HEAP_INFO_GET. + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_static_data_area_usage + drm_pvr_static_data_area + drm_pvr_dev_query_static_data_areas + +CREATE_BO +--------- +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: PowerVR IOCTL CREATE_BO interface + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_ioctl_create_bo_args + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: Flags for CREATE_BO + +GET_BO_MMAP_OFFSET +------------------ +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: PowerVR IOCTL GET_BO_MMAP_OFFSET interface + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_ioctl_get_bo_mmap_offset_args + +CREATE_VM_CONTEXT and DESTROY_VM_CONTEXT +---------------------------------------- +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: PowerVR IOCTL CREATE_VM_CONTEXT and DESTROY_VM_CONTEXT interfaces + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_ioctl_create_vm_context_args + drm_pvr_ioctl_destroy_vm_context_args + +VM_MAP and VM_UNMAP +------------------- +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: PowerVR IOCTL VM_MAP and VM_UNMAP interfaces + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_ioctl_vm_map_args + drm_pvr_ioctl_vm_unmap_args + +CREATE_CONTEXT and DESTROY_CONTEXT +---------------------------------- +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: PowerVR IOCTL CREATE_CONTEXT and DESTROY_CONTEXT interfaces + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_ioctl_create_context_args + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_ctx_priority + drm_pvr_ctx_type + drm_pvr_static_render_context_state + drm_pvr_static_render_context_state_format + drm_pvr_reset_framework + drm_pvr_reset_framework_format + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_ioctl_destroy_context_args + +CREATE_FREE_LIST and DESTROY_FREE_LIST +-------------------------------------- +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: PowerVR IOCTL CREATE_FREE_LIST and DESTROY_FREE_LIST interfaces + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_ioctl_create_free_list_args + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_ioctl_destroy_free_list_args + +CREATE_HWRT_DATASET and DESTROY_HWRT_DATASET +-------------------------------------- +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: PowerVR IOCTL CREATE_HWRT_DATASET and DESTROY_HWRT_DATASET interfaces + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_ioctl_create_hwrt_dataset_args + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_create_hwrt_geom_data_args + drm_pvr_create_hwrt_rt_data_args + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_ioctl_destroy_hwrt_dataset_args + +SUBMIT_JOBS +----------- +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: PowerVR IOCTL SUBMIT_JOBS interface + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: Flags for the drm_pvr_sync_op object. + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_ioctl_submit_jobs_args + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: Flags for SUBMIT_JOB ioctl geometry command. + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: Flags for SUBMIT_JOB ioctl fragment command. + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: Flags for SUBMIT_JOB ioctl compute command. + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :doc: Flags for SUBMIT_JOB ioctl transfer command. + +.. kernel-doc:: include/uapi/drm/pvr_drm.h + :identifiers: drm_pvr_sync_op + drm_pvr_job_type + drm_pvr_hwrt_data_ref + drm_pvr_job + +Internal notes +============== +.. kernel-doc:: drivers/gpu/drm/imagination/pvr_device.h + :doc: IOCTL validation helpers + +.. kernel-doc:: drivers/gpu/drm/imagination/pvr_device.h + :identifiers: PVR_STATIC_ASSERT_64BIT_ALIGNED PVR_IOCTL_UNION_PADDING_CHECK + pvr_ioctl_union_padding_check diff --git a/MAINTAINERS b/MAINTAINERS index debbc1956b8e..a6d8a15a23dd 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -10398,6 +10398,7 @@ M: Donald Robson M: Matt Coster S: Supported F: Documentation/devicetree/bindings/gpu/img,powervr.yaml +F: Documentation/gpu/imagination/ F: drivers/gpu/drm/imagination/ F: include/uapi/drm/pvr_drm.h