mbox series

[v6,0/8] Add Rockchip SFC(serial flash controller) support

Message ID 20210608022644.21074-1-jon.lin@rock-chips.com
Headers show
Series Add Rockchip SFC(serial flash controller) support | expand

Message

Jon Lin June 8, 2021, 2:26 a.m. UTC
Changes in v6:
- Add support in device trees for rv1126(Declared in series 5 but not
  submitted)
- Change to use "clk_sfc" "hclk_sfc" as clock lable, since it does not
  affect interpretation and has been widely used
- Support sfc tx_dual, tx_quad(Declared in series 5 but not submitted)
- Simplify the code, such as remove "rockchip_sfc_register_all"(Declared
  in series 5 but not submitted)
- Support SFC ver4 ver5(Declared in series 5 but not submitted)
- Add author Chris Morgan and Jon Lin to spi-rockchip-sfc.c
- Change to use devm_spi_alloc_master and spi_unregister_master

Changes in v5:
- Add support in device trees for rv1126
- Support sfc tx_dual, tx_quad
- Simplify the code, such as remove "rockchip_sfc_register_all"
- Support SFC ver4 ver5

Changes in v4:
- Changing patch back to an "RFC". An engineer from Rockchip
  reached out to me to let me know they are working on this patch for
  upstream, I am submitting this v4 for the community to see however
  I expect Jon Lin (jon.lin@rock-chips.com) will submit new patches
  soon and these are the ones we should pursue for mainlining. Jon's
  patch series should include support for more hardware than this
  series.
- Clean up documentation more and ensure it is correct per
  make dt_binding_check.
- Add support in device trees for rk3036, rk3308, and rv1108.
- Add ahb clock (hclk_sfc) support for rk3036.
- Change rockchip_sfc_wait_fifo_ready() to use a switch statement.
- Change IRQ code to only mark IRQ as handled if it handles the
  specific IRQ (DMA transfer finish) it is supposed to handle.

Changes in v3:
- Changed the name of the clocks to sfc/ahb (from clk-sfc/clk-hsfc).
- Changed the compatible string from rockchip,sfc to
  rockchip,rk3036-sfc. A quick glance at the datasheets suggests this
  driver should work for the PX30, RK180x, RK3036, RK312x, RK3308 and
  RV1108 SoCs, and possibly more. However, I am currently only able
  to test this on a PX30 (an RK3326). The technical reference manuals
  appear to list the same registers for each device.
- Corrected devicetree documentation for formatting and to note these
  changes.
- Replaced the maintainer with Heiko Stuebner and myself, as we will
  take ownership of this going forward.
- Noted that the device (per the reference manual) supports 4 CS, but
  I am only able to test a single CS (CS 0).
- Reordered patches to comply with upstream rules.

Changes in v2:
- Reimplemented driver using spi-mem subsystem.
- Removed power management code as I couldn't get it working properly.
- Added device tree bindings for Odroid Go Advance.

Changes in v1:
hanges made in this new series versus the v8 of the old series:
- Added function to read spi-rx-bus-width from device tree, in the
  event that the SPI chip supports 4x mode but only has 2 pins
  wired (such as the Odroid Go Advance).
- Changed device tree documentation from txt to yaml format.
- Made "reset" message a dev_dbg from a dev_info.
- Changed read and write fifo functions to remove redundant checks.
- Changed the write and read from relaxed to non-relaxed when
  starting the DMA transfer or reading the DMA IRQ.
- Changed from dma_coerce_mask_and_coherent to just
  dma_set_mask_and_coherent.
- Changed name of get_if_type to rockchip_sfc_get_if_type.

Chris Morgan (8):
  dt-bindings: rockchip-sfc: Bindings for Rockchip serial flash
    controller
  spi: rockchip-sfc: add rockchip serial flash controller
  arm64: dts: rockchip: Add SFC to PX30
  clk: rockchip: Add support for hclk_sfc on rk3036
  arm: dts: rockchip: Add SFC to RK3036
  arm: dts: rockchip: Add SFC to RV1108
  arm64: dts: rockchip: Add SFC to RK3308
  arm64: dts: rockchip: Enable SFC for Odroid Go Advance

 .../devicetree/bindings/spi/rockchip-sfc.yaml |  87 +++
 arch/arm/boot/dts/rk3036.dtsi                 |  42 ++
 arch/arm/boot/dts/rv1108.dtsi                 |  37 +
 arch/arm64/boot/dts/rockchip/px30.dtsi        |  38 +
 arch/arm64/boot/dts/rockchip/rk3308.dtsi      |  37 +
 .../boot/dts/rockchip/rk3326-odroid-go2.dts   |  16 +
 drivers/clk/rockchip/clk-rk3036.c             |   2 +-
 drivers/spi/Kconfig                           |   9 +
 drivers/spi/Makefile                          |   1 +
 drivers/spi/spi-rockchip-sfc.c                | 660 ++++++++++++++++++
 include/dt-bindings/clock/rk3036-cru.h        |   1 +
 11 files changed, 929 insertions(+), 1 deletion(-)
 create mode 100644 Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
 create mode 100644 drivers/spi/spi-rockchip-sfc.c

Comments

Johan Jonker June 8, 2021, 6:03 p.m. UTC | #1
Hi Jon,

A look at the build log shows this error:

https://patchwork.ozlabs.org/project/devicetree-bindings/patch/20210608022644.21074-2-jon.lin@rock-chips.com/

dtschema/dtc warnings/errors:
/builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.example.dt.yaml:
spi@ff3a0000: clock-names:0: 'hclk_sfc' was expected
	From schema:
/builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
/builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.example.dt.yaml:
spi@ff3a0000: clock-names:1: 'clk_sfc' was expected
	From schema:
/builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml

Check document with:

make ARCH=arm dt_binding_check
DT_SCHEMA_FILES=Documentation/devicetree/bindings/spi/rockchip-sfc.yaml

make ARCH=arm dtbs_check
DT_SCHEMA_FILES=Documentation/devicetree/bindings/spi/rockchip-sfc.yaml

make ARCH=arm64 dtbs_check
DT_SCHEMA_FILES=Documentation/devicetree/bindings/spi/rockchip-sfc.yaml

Remove any errors before submitting.

===

Johan

On 6/8/21 4:26 AM, Jon Lin wrote:
> From: Chris Morgan <macromorgan@hotmail.com>
> 
> Add bindings for the Rockchip serial flash controller. New device
> specific parameter of rockchip,sfc-no-dma included in documentation.
> 
> Signed-off-by: Chris Morgan <macromorgan@hotmail.com>
> Signed-off-by: Jon Lin <jon.lin@rock-chips.com>
> ---
> 
> Changes in v6:
> - Add support in device trees for rv1126(Declared in series 5 but not
>   submitted)
> - Change to use "clk_sfc" "hclk_sfc" as clock lable, since it does not
>   affect interpretation and has been widely used
> - Support sfc tx_dual, tx_quad(Declared in series 5 but not submitted)
> - Simplify the code, such as remove "rockchip_sfc_register_all"(Declared
>   in series 5 but not submitted)
> - Support SFC ver4 ver5(Declared in series 5 but not submitted)
> - Add author Chris Morgan and Jon Lin to spi-rockchip-sfc.c
> - Change to use devm_spi_alloc_master and spi_unregister_master
> 
> Changes in v5:
> - Add support in device trees for rv1126
> - Support sfc tx_dual, tx_quad
> - Simplify the code, such as remove "rockchip_sfc_register_all"
> - Support SFC ver4 ver5
> 
> Changes in v4:
> - Changing patch back to an "RFC". An engineer from Rockchip
>   reached out to me to let me know they are working on this patch for
>   upstream, I am submitting this v4 for the community to see however
>   I expect Jon Lin (jon.lin@rock-chips.com) will submit new patches
>   soon and these are the ones we should pursue for mainlining. Jon's
>   patch series should include support for more hardware than this
>   series.
> - Clean up documentation more and ensure it is correct per
>   make dt_binding_check.
> - Add support in device trees for rk3036, rk3308, and rv1108.
> - Add ahb clock (hclk_sfc) support for rk3036.
> - Change rockchip_sfc_wait_fifo_ready() to use a switch statement.
> - Change IRQ code to only mark IRQ as handled if it handles the
>   specific IRQ (DMA transfer finish) it is supposed to handle.
> 
> Changes in v3:
> - Changed the name of the clocks to sfc/ahb (from clk-sfc/clk-hsfc).
> - Changed the compatible string from rockchip,sfc to
>   rockchip,rk3036-sfc. A quick glance at the datasheets suggests this
>   driver should work for the PX30, RK180x, RK3036, RK312x, RK3308 and
>   RV1108 SoCs, and possibly more. However, I am currently only able
>   to test this on a PX30 (an RK3326). The technical reference manuals
>   appear to list the same registers for each device.
> - Corrected devicetree documentation for formatting and to note these
>   changes.
> - Replaced the maintainer with Heiko Stuebner and myself, as we will
>   take ownership of this going forward.
> - Noted that the device (per the reference manual) supports 4 CS, but
>   I am only able to test a single CS (CS 0).
> - Reordered patches to comply with upstream rules.
> 
> Changes in v2:
> - Reimplemented driver using spi-mem subsystem.
> - Removed power management code as I couldn't get it working properly.
> - Added device tree bindings for Odroid Go Advance.
> 
> Changes in v1:
> hanges made in this new series versus the v8 of the old series:
> - Added function to read spi-rx-bus-width from device tree, in the
>   event that the SPI chip supports 4x mode but only has 2 pins
>   wired (such as the Odroid Go Advance).
> - Changed device tree documentation from txt to yaml format.
> - Made "reset" message a dev_dbg from a dev_info.
> - Changed read and write fifo functions to remove redundant checks.
> - Changed the write and read from relaxed to non-relaxed when
>   starting the DMA transfer or reading the DMA IRQ.
> - Changed from dma_coerce_mask_and_coherent to just
>   dma_set_mask_and_coherent.
> - Changed name of get_if_type to rockchip_sfc_get_if_type.
> 
>  .../devicetree/bindings/spi/rockchip-sfc.yaml | 87 +++++++++++++++++++
>  1 file changed, 87 insertions(+)
>  create mode 100644 Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
> 
> diff --git a/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml b/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
> new file mode 100644
> index 000000000000..160449713f97
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
> @@ -0,0 +1,87 @@
> +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> +%YAML 1.2
> +---
> +$id: http://devicetree.org/schemas/spi/rockchip-sfc.yaml#
> +$schema: http://devicetree.org/meta-schemas/core.yaml#
> +
> +title: Rockchip Serial Flash Controller (SFC)
> +
> +maintainers:
> +  - Heiko Stuebner <heiko@sntech.de>

> +  - Chris Morgan <macromorgan@hotmail.com>

In the driver spi-rockchip-sfc.c is used:
Chris Morgan <macroalpha82@gmail.com>
Maybe use a consistent email address?
Also Hotmail does strange things to message-ID's and links/content
inside patches.

> +
> +allOf:
> +  - $ref: spi-controller.yaml#
> +
> +properties:
> +  compatible:
> +    oneOf:

> +      - const: rockchip,rk3036-sfc
> +      - items:
> +          - enum:
> +              - rockchip,px30-sfc
> +              - rockchip,rk3308-sfc
> +              - rockchip,rv1108-sfc
> +          - const: rockchip,rk3036-sfc
> +      - const: rockchip,rv1126-sfc

A look at the driver shows this:

+static const struct of_device_id rockchip_sfc_dt_ids[] = {

+	{ .compatible = "rockchip,px30-sfc"},

remove

+	{ .compatible = "rockchip,rk3036-sfc"},

+	{ .compatible = "rockchip,rk3308-sfc"},
+	{ .compatible = "rockchip,rv1126-sfc"},

remove

+	{ /* sentinel */ }
+};

When no '.data = &my_sfc_data' exist then there's no need for additional
compatible strings in the driver.
===
example for I2S:
	{ .compatible = "rockchip,rk3399-i2s", .data = &rk3399_i2s_pins },
===
Compatibility strings are supposed to be SoC orientated.
With rockchip_sfc_get_version() all we need is one fall back string for
now I think. Is rk3568 identical. Please advise.


      - const: rockchip,rk3036-sfc
      - items:
          - enum:
              - rockchip,px30-sfc
              - rockchip,rk3308-sfc
              - rockchip,rk3368-sfc
              - rockchip,rk3568-sfc
              - rockchip,rv1108-sfc
              - rockchip,rv1126-sfc
          - const: rockchip,rk3036-sfc

===

The rk3368 TRM also describes a SFC controller.
For the mainline DT completeness is it possible to add a rk3368 SFC node
as well? ;)

> +
> +  reg:
> +    maxItems: 1
> +
> +  interrupts:
> +    maxItems: 1
> +
> +  clocks:
> +    items:
> +      - description: Bus Clock
> +      - description: Module Clock
> +
> +  clock-names:
> +    items:
> +      - const: hclk_sfc
> +      - const: clk_sfc
> +
> +  power-domains:
> +    maxItems: 1
> +
> +  rockchip,sfc-no-dma:
> +    description: Disable DMA and utilize FIFO mode only
> +    type: boolean
> +
> +required:
> +  - compatible
> +  - reg
> +  - interrupts
> +  - clocks
> +  - clock-names
> +
> +unevaluatedProperties: false
> +
> +examples:
> +  - |
> +    #include <dt-bindings/clock/px30-cru.h>
> +    #include <dt-bindings/interrupt-controller/arm-gic.h>
> +    #include <dt-bindings/power/px30-power.h>
> +
> +    sfc: spi@ff3a0000 {
> +        compatible = "rockchip,px30-sfc","rockchip,rk3036-sfc";
> +        reg = <0xff3a0000 0x4000>;
> +        interrupts = <GIC_SPI 56 IRQ_TYPE_LEVEL_HIGH>;

> +        clocks = <&cru SCLK_SFC>, <&cru HCLK_SFC>;
> +        clock-names = "clk_sfc", "hclk_sfc";

The clocks in the examples and dtsi files must have the same sort order.

> +        pinctrl-0 = <&sfc_clk &sfc_cs &sfc_bus2>;
> +        pinctrl-names = "default";
> +        power-domains = <&power PX30_PD_MMC_NAND>;
> +        #address-cells = <1>;
> +        #size-cells = <0>;
> +

> +        flash@0 {
Rob Herring June 8, 2021, 8:50 p.m. UTC | #2
On Tue, 08 Jun 2021 10:26:37 +0800, Jon Lin wrote:
> From: Chris Morgan <macromorgan@hotmail.com>
> 
> Add bindings for the Rockchip serial flash controller. New device
> specific parameter of rockchip,sfc-no-dma included in documentation.
> 
> Signed-off-by: Chris Morgan <macromorgan@hotmail.com>
> Signed-off-by: Jon Lin <jon.lin@rock-chips.com>
> ---
> 
> Changes in v6:
> - Add support in device trees for rv1126(Declared in series 5 but not
>   submitted)
> - Change to use "clk_sfc" "hclk_sfc" as clock lable, since it does not
>   affect interpretation and has been widely used
> - Support sfc tx_dual, tx_quad(Declared in series 5 but not submitted)
> - Simplify the code, such as remove "rockchip_sfc_register_all"(Declared
>   in series 5 but not submitted)
> - Support SFC ver4 ver5(Declared in series 5 but not submitted)
> - Add author Chris Morgan and Jon Lin to spi-rockchip-sfc.c
> - Change to use devm_spi_alloc_master and spi_unregister_master
> 
> Changes in v5:
> - Add support in device trees for rv1126
> - Support sfc tx_dual, tx_quad
> - Simplify the code, such as remove "rockchip_sfc_register_all"
> - Support SFC ver4 ver5
> 
> Changes in v4:
> - Changing patch back to an "RFC". An engineer from Rockchip
>   reached out to me to let me know they are working on this patch for
>   upstream, I am submitting this v4 for the community to see however
>   I expect Jon Lin (jon.lin@rock-chips.com) will submit new patches
>   soon and these are the ones we should pursue for mainlining. Jon's
>   patch series should include support for more hardware than this
>   series.
> - Clean up documentation more and ensure it is correct per
>   make dt_binding_check.
> - Add support in device trees for rk3036, rk3308, and rv1108.
> - Add ahb clock (hclk_sfc) support for rk3036.
> - Change rockchip_sfc_wait_fifo_ready() to use a switch statement.
> - Change IRQ code to only mark IRQ as handled if it handles the
>   specific IRQ (DMA transfer finish) it is supposed to handle.
> 
> Changes in v3:
> - Changed the name of the clocks to sfc/ahb (from clk-sfc/clk-hsfc).
> - Changed the compatible string from rockchip,sfc to
>   rockchip,rk3036-sfc. A quick glance at the datasheets suggests this
>   driver should work for the PX30, RK180x, RK3036, RK312x, RK3308 and
>   RV1108 SoCs, and possibly more. However, I am currently only able
>   to test this on a PX30 (an RK3326). The technical reference manuals
>   appear to list the same registers for each device.
> - Corrected devicetree documentation for formatting and to note these
>   changes.
> - Replaced the maintainer with Heiko Stuebner and myself, as we will
>   take ownership of this going forward.
> - Noted that the device (per the reference manual) supports 4 CS, but
>   I am only able to test a single CS (CS 0).
> - Reordered patches to comply with upstream rules.
> 
> Changes in v2:
> - Reimplemented driver using spi-mem subsystem.
> - Removed power management code as I couldn't get it working properly.
> - Added device tree bindings for Odroid Go Advance.
> 
> Changes in v1:
> hanges made in this new series versus the v8 of the old series:
> - Added function to read spi-rx-bus-width from device tree, in the
>   event that the SPI chip supports 4x mode but only has 2 pins
>   wired (such as the Odroid Go Advance).
> - Changed device tree documentation from txt to yaml format.
> - Made "reset" message a dev_dbg from a dev_info.
> - Changed read and write fifo functions to remove redundant checks.
> - Changed the write and read from relaxed to non-relaxed when
>   starting the DMA transfer or reading the DMA IRQ.
> - Changed from dma_coerce_mask_and_coherent to just
>   dma_set_mask_and_coherent.
> - Changed name of get_if_type to rockchip_sfc_get_if_type.
> 
>  .../devicetree/bindings/spi/rockchip-sfc.yaml | 87 +++++++++++++++++++
>  1 file changed, 87 insertions(+)
>  create mode 100644 Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
> 

My bot found errors running 'make DT_CHECKER_FLAGS=-m dt_binding_check'
on your patch (DT_CHECKER_FLAGS is new in v5.13):

yamllint warnings/errors:

dtschema/dtc warnings/errors:
/builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.example.dt.yaml: spi@ff3a0000: clock-names:0: 'hclk_sfc' was expected
	From schema: /builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
/builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.example.dt.yaml: spi@ff3a0000: clock-names:1: 'clk_sfc' was expected
	From schema: /builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
\ndoc reference errors (make refcheckdocs):

See https://patchwork.ozlabs.org/patch/1489062

This check can fail if there are any dependencies. The base for a patch
series is generally the most recent rc1.

If you already ran 'make dt_binding_check' and didn't see the above
error(s), then make sure 'yamllint' is installed and dt-schema is up to
date:

pip3 install dtschema --upgrade

Please check and re-submit.
Chris Morgan June 9, 2021, 2:36 a.m. UTC | #3
On Tue, Jun 08, 2021 at 10:26:38AM +0800, Jon Lin wrote:
> From: Chris Morgan <macromorgan@hotmail.com>

> 

> Add the rockchip serial flash controller (SFC) driver.

> 

> Signed-off-by: Chris Morgan <macromorgan@hotmail.com>

> Signed-off-by: Jon Lin <jon.lin@rock-chips.com>


I think I hit "reply" earlier when I meant to hit "reply-all". Whoops.

I wanted to say that for now this driver is not working for me on the
v3 hardware. There appears to be something wrong with alternating
between page programs and erase. I can consistently reproduce this by
writing a 4MB image of random data to my SPI flash chip at offset
c00000 using dd to mtdblock0. I've dumped the contents of what is
written to the controller and it appears at some point it goes from
doing a page program (02) to a dual io read (bb) to a sector erase
(20), then it repeatedly reads the status register (05) and I assume
the register never changes. As a result it just keeps looping over
the status register then eventually times out. I'm still debugging
to try and figure out exactly what is going on though.

I also noticed that reading does not work consistently using dd
from the mtd0 device, however I have a proposed fix for that which
I list below at the appropriate section.

> ---

> 

> Changes in v6: None

> Changes in v5: None

> Changes in v4: None

> Changes in v3: None

> Changes in v2: None

> Changes in v1: None

> 

>  drivers/spi/Kconfig            |   9 +

>  drivers/spi/Makefile           |   1 +

>  drivers/spi/spi-rockchip-sfc.c | 660 +++++++++++++++++++++++++++++++++

>  3 files changed, 670 insertions(+)

>  create mode 100644 drivers/spi/spi-rockchip-sfc.c

> 

> diff --git a/drivers/spi/Kconfig b/drivers/spi/Kconfig

> index e71a4c514f7b..d89e5f3c9107 100644

> --- a/drivers/spi/Kconfig

> +++ b/drivers/spi/Kconfig

> @@ -658,6 +658,15 @@ config SPI_ROCKCHIP

>  	  The main usecase of this controller is to use spi flash as boot

>  	  device.

>  

> +config SPI_ROCKCHIP_SFC

> +	tristate "Rockchip Serial Flash Controller (SFC)"

> +	depends on ARCH_ROCKCHIP || COMPILE_TEST

> +	depends on HAS_IOMEM && HAS_DMA

> +	help

> +	  This enables support for Rockchip serial flash controller. This

> +	  is a specialized controller used to access SPI flash on some

> +	  Rockchip SOCs.

> +

>  config SPI_RB4XX

>  	tristate "Mikrotik RB4XX SPI master"

>  	depends on SPI_MASTER && ATH79

> diff --git a/drivers/spi/Makefile b/drivers/spi/Makefile

> index 13e54c45e9df..699db95c8441 100644

> --- a/drivers/spi/Makefile

> +++ b/drivers/spi/Makefile

> @@ -95,6 +95,7 @@ obj-$(CONFIG_SPI_QCOM_GENI)		+= spi-geni-qcom.o

>  obj-$(CONFIG_SPI_QCOM_QSPI)		+= spi-qcom-qspi.o

>  obj-$(CONFIG_SPI_QUP)			+= spi-qup.o

>  obj-$(CONFIG_SPI_ROCKCHIP)		+= spi-rockchip.o

> +obj-$(CONFIG_SPI_ROCKCHIP_SFC)		+= spi-rockchip-sfc.o

>  obj-$(CONFIG_SPI_RB4XX)			+= spi-rb4xx.o

>  obj-$(CONFIG_MACH_REALTEK_RTL)		+= spi-realtek-rtl.o

>  obj-$(CONFIG_SPI_RPCIF)			+= spi-rpc-if.o

> diff --git a/drivers/spi/spi-rockchip-sfc.c b/drivers/spi/spi-rockchip-sfc.c

> new file mode 100644

> index 000000000000..ec25ad278096

> --- /dev/null

> +++ b/drivers/spi/spi-rockchip-sfc.c

> @@ -0,0 +1,660 @@

> +// SPDX-License-Identifier: GPL-2.0-only

> +/*

> + * Rockchip Serial Flash Controller Driver

> + *

> + * Copyright (c) 2017-2021, Rockchip Inc.

> + * Author: Shawn Lin <shawn.lin@rock-chips.com>

> + *	   Chris Morgan <macroalpha82@gmail.com>

> + *	   Jon Lin <Jon.lin@rock-chips.com>

> + */

> +

> +#include <linux/bitops.h>

> +#include <linux/clk.h>

> +#include <linux/completion.h>

> +#include <linux/dma-mapping.h>

> +#include <linux/iopoll.h>

> +#include <linux/mm.h>

> +#include <linux/module.h>

> +#include <linux/of.h>

> +#include <linux/platform_device.h>

> +#include <linux/slab.h>

> +#include <linux/interrupt.h>

> +#include <linux/spi/spi-mem.h>

> +

> +/* System control */

> +#define SFC_CTRL			0x0

> +#define  SFC_CTRL_PHASE_SEL_NEGETIVE	BIT(1)

> +#define  SFC_CTRL_CMD_BITS_SHIFT	8

> +#define  SFC_CTRL_ADDR_BITS_SHIFT	10

> +#define  SFC_CTRL_DATA_BITS_SHIFT	12

> +

> +/* Interrupt mask */

> +#define SFC_IMR				0x4

> +#define  SFC_IMR_RX_FULL		BIT(0)

> +#define  SFC_IMR_RX_UFLOW		BIT(1)

> +#define  SFC_IMR_TX_OFLOW		BIT(2)

> +#define  SFC_IMR_TX_EMPTY		BIT(3)

> +#define  SFC_IMR_TRAN_FINISH		BIT(4)

> +#define  SFC_IMR_BUS_ERR		BIT(5)

> +#define  SFC_IMR_NSPI_ERR		BIT(6)

> +#define  SFC_IMR_DMA			BIT(7)

> +

> +/* Interrupt clear */

> +#define SFC_ICLR			0x8

> +#define  SFC_ICLR_RX_FULL		BIT(0)

> +#define  SFC_ICLR_RX_UFLOW		BIT(1)

> +#define  SFC_ICLR_TX_OFLOW		BIT(2)

> +#define  SFC_ICLR_TX_EMPTY		BIT(3)

> +#define  SFC_ICLR_TRAN_FINISH		BIT(4)

> +#define  SFC_ICLR_BUS_ERR		BIT(5)

> +#define  SFC_ICLR_NSPI_ERR		BIT(6)

> +#define  SFC_ICLR_DMA			BIT(7)

> +

> +/* FIFO threshold level */

> +#define SFC_FTLR			0xc

> +#define  SFC_FTLR_TX_SHIFT		0

> +#define  SFC_FTLR_TX_MASK		0x1f

> +#define  SFC_FTLR_RX_SHIFT		8

> +#define  SFC_FTLR_RX_MASK		0x1f

> +

> +/* Reset FSM and FIFO */

> +#define SFC_RCVR			0x10

> +#define  SFC_RCVR_RESET			BIT(0)

> +

> +/* Enhanced mode */

> +#define SFC_AX				0x14

> +

> +/* Address Bit number */

> +#define SFC_ABIT			0x18

> +

> +/* Interrupt status */

> +#define SFC_ISR				0x1c

> +#define  SFC_ISR_RX_FULL_SHIFT		BIT(0)

> +#define  SFC_ISR_RX_UFLOW_SHIFT		BIT(1)

> +#define  SFC_ISR_TX_OFLOW_SHIFT		BIT(2)

> +#define  SFC_ISR_TX_EMPTY_SHIFT		BIT(3)

> +#define  SFC_ISR_TX_FINISH_SHIFT	BIT(4)

> +#define  SFC_ISR_BUS_ERR_SHIFT		BIT(5)

> +#define  SFC_ISR_NSPI_ERR_SHIFT		BIT(6)

> +#define  SFC_ISR_DMA_SHIFT		BIT(7)

> +

> +/* FIFO status */

> +#define SFC_FSR				0x20

> +#define  SFC_FSR_TX_IS_FULL		BIT(0)

> +#define  SFC_FSR_TX_IS_EMPTY		BIT(1)

> +#define  SFC_FSR_RX_IS_EMPTY		BIT(2)

> +#define  SFC_FSR_RX_IS_FULL		BIT(3)

> +#define  SFC_FSR_TXLV_MASK		GENMASK(12, 8)

> +#define  SFC_FSR_TXLV_SHIFT		8

> +#define  SFC_FSR_RXLV_MASK		GENMASK(20, 16)

> +#define  SFC_FSR_RXLV_SHIFT		16

> +

> +/* FSM status */

> +#define SFC_SR				0x24

> +#define  SFC_SR_IS_IDLE			0x0

> +#define  SFC_SR_IS_BUSY			0x1

> +

> +/* Raw interrupt status */

> +#define SFC_RISR			0x28

> +#define  SFC_RISR_RX_FULL		BIT(0)

> +#define  SFC_RISR_RX_UNDERFLOW		BIT(1)

> +#define  SFC_RISR_TX_OVERFLOW		BIT(2)

> +#define  SFC_RISR_TX_EMPTY		BIT(3)

> +#define  SFC_RISR_TRAN_FINISH		BIT(4)

> +#define  SFC_RISR_BUS_ERR		BIT(5)

> +#define  SFC_RISR_NSPI_ERR		BIT(6)

> +#define  SFC_RISR_DMA			BIT(7)

> +

> +/* Version */

> +#define SFC_VER				0x2C

> +#define  SFC_VER_3			0x3

> +#define  SFC_VER_4			0x4

> +#define  SFC_VER_5			0x5

> +

> +/* Master trigger */

> +#define SFC_DMA_TRIGGER			0x80

> +

> +/* Src or Dst addr for master */

> +#define SFC_DMA_ADDR			0x84

> +

> +/* Length control register extension 32GB */

> +#define SFC_LEN_CTRL			0x88

> +#define SFC_LEN_CTRL_TRB_SEL		1

> +#define SFC_LEN_EXT			0x8C

> +

> +/* Command */

> +#define SFC_CMD				0x100

> +#define  SFC_CMD_IDX_SHIFT		0

> +#define  SFC_CMD_DUMMY_SHIFT		8

> +#define  SFC_CMD_DIR_SHIFT		12

> +#define  SFC_CMD_DIR_RD			0

> +#define  SFC_CMD_DIR_WR			1

> +#define  SFC_CMD_ADDR_SHIFT		14

> +#define  SFC_CMD_ADDR_0BITS		0

> +#define  SFC_CMD_ADDR_24BITS		1

> +#define  SFC_CMD_ADDR_32BITS		2

> +#define  SFC_CMD_ADDR_XBITS		3

> +#define  SFC_CMD_TRAN_BYTES_SHIFT	16

> +#define  SFC_CMD_CS_SHIFT		30

> +

> +/* Address */

> +#define SFC_ADDR			0x104

> +

> +/* Data */

> +#define SFC_DATA			0x108

> +

> +/* The controller and documentation reports that it supports up to 4 CS

> + * devices (0-3), however I have only been able to test a single CS (CS 0)

> + * due to the configuration of my device.

> + */

> +#define SFC_MAX_CHIPSELECT_NUM		4

> +

> +/* The SFC can transfer max 16KB - 1 at one time

> + * we set it to 15.5KB here for alignment.

> + */

> +#define SFC_MAX_IOSIZE_VER3		(512 * 31)

> +

> +#define SFC_MAX_IOSIZE_VER4		(0xFFFFFFFF)

> +

> +/* DMA is only enabled for large data transmission */

> +#define SFC_DMA_TRANS_THRETHOLD		(0x40)

> +

> +/* Maximum clock values from datasheet suggest keeping clock value under

> + * 150MHz. No minimum or average value is suggested, but the U-boot BSP driver

> + * has a minimum of 10MHz and a default of 80MHz which seems reasonable.

> + */

> +#define SFC_MIN_SPEED_HZ		(10 * 1000 * 1000)

> +#define SFC_DEFAULT_SPEED_HZ		(80 * 1000 * 1000)

> +#define SFC_MAX_SPEED_HZ		(150 * 1000 * 1000)

> +

> +struct rockchip_sfc {

> +	struct device *dev;

> +	void __iomem *regbase;

> +	struct clk *hclk;

> +	struct clk *clk;

> +	u32 frequency;

> +	/* virtual mapped addr for dma_buffer */

> +	void *buffer;

> +	dma_addr_t dma_buffer;

> +	struct completion cp;

> +	bool use_dma;

> +	u32 max_iosize;

> +};

> +

> +static int rockchip_sfc_reset(struct rockchip_sfc *sfc)

> +{

> +	int err;

> +	u32 status;

> +

> +	writel_relaxed(SFC_RCVR_RESET, sfc->regbase + SFC_RCVR);

> +

> +	err = readl_poll_timeout(sfc->regbase + SFC_RCVR, status,

> +				 !(status & SFC_RCVR_RESET), 20,

> +				 jiffies_to_usecs(HZ));

> +	if (err)

> +		dev_err(sfc->dev, "SFC reset never finished\n");

> +

> +	/* Still need to clear the masked interrupt from RISR */

> +	writel_relaxed(0xFFFFFFFF, sfc->regbase + SFC_ICLR);

> +

> +	dev_dbg(sfc->dev, "reset\n");

> +

> +	return err;

> +}

> +

> +static u16 rockchip_sfc_get_version(struct rockchip_sfc *sfc)

> +{

> +	return  (u16)(readl(sfc->regbase + SFC_VER) & 0xffff);

> +}

> +

> +static u32 rockchip_sfc_get_max_iosize(struct rockchip_sfc *sfc)

> +{

> +	if (rockchip_sfc_get_version(sfc) >= SFC_VER_4)

> +		return SFC_MAX_IOSIZE_VER4;

> +	else

> +		return SFC_MAX_IOSIZE_VER3;

> +}

> +

> +static int rockchip_sfc_init(struct rockchip_sfc *sfc)

> +{

> +	writel(0, sfc->regbase + SFC_CTRL);

> +	if (rockchip_sfc_get_version(sfc) >= SFC_VER_4)

> +		writel(SFC_LEN_CTRL_TRB_SEL, sfc->regbase + SFC_LEN_CTRL);

> +

> +	return 0;

> +}

> +

> +static inline int rockchip_sfc_get_fifo_level(struct rockchip_sfc *sfc, int wr)

> +{

> +	u32 fsr = readl_relaxed(sfc->regbase + SFC_FSR);

> +	int level;

> +

> +	if (wr)

> +		level = (fsr & SFC_FSR_TXLV_MASK) >> SFC_FSR_TXLV_SHIFT;

> +	else

> +		level = (fsr & SFC_FSR_RXLV_MASK) >> SFC_FSR_RXLV_SHIFT;

> +

> +	return level;

> +}

> +

> +static int rockchip_sfc_wait_fifo_ready(struct rockchip_sfc *sfc, int wr, u32 timeout)

> +{

> +	unsigned long deadline = jiffies + timeout;

> +	int level;

> +

> +	while (!(level = rockchip_sfc_get_fifo_level(sfc, wr))) {

> +		if (time_after_eq(jiffies, deadline)) {

> +			dev_warn(sfc->dev, "%s fifo timeout\n", wr ? "write" : "read");

> +			return -ETIMEDOUT;

> +		}

> +		udelay(1);

> +	}

> +

> +	return level;

> +}

> +

> +static int rockchip_sfc_xfer_setup(struct rockchip_sfc *sfc,

> +				   struct spi_mem *mem,

> +				   const struct spi_mem_op *op,

> +				   u32 len)

> +{

> +	u32 ctrl = 0, cmd = 0;

> +

> +	/* set CMD */

> +	cmd = op->cmd.opcode;

> +	ctrl |= ((op->cmd.buswidth >> 1) << SFC_CTRL_CMD_BITS_SHIFT);

> +

> +	/* set ADDR */

> +	if (op->addr.nbytes) {

> +		if (op->addr.nbytes == 4) {

> +			cmd |= SFC_CMD_ADDR_32BITS << SFC_CMD_ADDR_SHIFT;

> +		} else if (op->addr.nbytes == 3) {

> +			cmd |= SFC_CMD_ADDR_24BITS << SFC_CMD_ADDR_SHIFT;

> +		} else {

> +			cmd |= SFC_CMD_ADDR_XBITS << SFC_CMD_ADDR_SHIFT;

> +			writel_relaxed(op->addr.nbytes * 8 - 1, sfc->regbase + SFC_ABIT);

> +		}

> +

> +		ctrl |= ((op->addr.buswidth >> 1) << SFC_CTRL_ADDR_BITS_SHIFT);

> +	}

> +

> +	/* set DUMMY */

> +	if (op->dummy.nbytes) {

> +		if (op->dummy.buswidth == 4)

> +			cmd |= op->dummy.nbytes * 2 << SFC_CMD_DUMMY_SHIFT;

> +		else if (op->dummy.buswidth == 2)

> +			cmd |= op->dummy.nbytes * 4 << SFC_CMD_DUMMY_SHIFT;

> +		else

> +			cmd |= op->dummy.nbytes * 8 << SFC_CMD_DUMMY_SHIFT;

> +	}

> +


I have no experience optimizing code or anything, but would it help
here to read the output of rockchip_sfc_get_version() to a variable
in the driver data, and then check that each time we hit this?
I don't know if there is any real-world difference in reading a
variable versus reading a register, so I'm just asking. It's not like
the sfc_version will change once the device is probed. :-)

> +	/* set DATA */

> +	if (rockchip_sfc_get_version(sfc) >= SFC_VER_4) /* Clear it if no data to transfer */

> +		writel(len, sfc->regbase + SFC_LEN_EXT);

> +	else

> +		cmd |= len << SFC_CMD_TRAN_BYTES_SHIFT;

> +	if (len) {

> +		if (op->data.dir == SPI_MEM_DATA_OUT)

> +			cmd |= SFC_CMD_DIR_WR << SFC_CMD_DIR_SHIFT;

> +

> +		ctrl |= ((op->data.buswidth >> 1) << SFC_CTRL_DATA_BITS_SHIFT);

> +	}

> +	if (!len && op->addr.nbytes)

> +		cmd |= SFC_CMD_DIR_WR << SFC_CMD_DIR_SHIFT;

> +

> +	/* set the Controller */

> +	ctrl |= SFC_CTRL_PHASE_SEL_NEGETIVE;

> +	cmd |= mem->spi->chip_select << SFC_CMD_CS_SHIFT;

> +

> +	dev_dbg(sfc->dev, "addr.nbytes=%x(x%d) dummy.nbytes=%x(x%d)\n",

> +		op->addr.nbytes, op->addr.buswidth,

> +		op->dummy.nbytes, op->dummy.buswidth);

> +	dev_dbg(sfc->dev, "ctrl=%x cmd=%x addr=%llx len=%x\n",

> +		ctrl, cmd, op->addr.val, len);

> +

> +	writel(ctrl, sfc->regbase + SFC_CTRL);

> +	writel(cmd, sfc->regbase + SFC_CMD);

> +	if (op->addr.nbytes)

> +		writel(op->addr.val, sfc->regbase + SFC_ADDR);

> +

> +	return 0;

> +}

> +

> +static int rockchip_sfc_write_fifo(struct rockchip_sfc *sfc, const u8 *buf, int len)

> +{

> +	u8 bytes = len & 0x3;

> +	u32 dwords;

> +	int tx_level;

> +	u32 write_words;

> +	u32 tmp = 0;

> +

> +	dwords = len >> 2;

> +	while (dwords) {

> +		tx_level = rockchip_sfc_wait_fifo_ready(sfc, SFC_CMD_DIR_WR, HZ);

> +		if (tx_level < 0)

> +			return tx_level;

> +		write_words = min_t(u32, tx_level, dwords);

> +		iowrite32_rep(sfc->regbase + SFC_DATA, buf, write_words);

> +		buf += write_words << 2;

> +		dwords -= write_words;

> +	}

> +

> +	/* write the rest non word aligned bytes */

> +	if (bytes) {

> +		tx_level = rockchip_sfc_wait_fifo_ready(sfc, SFC_CMD_DIR_WR, HZ);

> +		if (tx_level < 0)

> +			return tx_level;

> +		memcpy(&tmp, buf, bytes);

> +		writel(tmp, sfc->regbase + SFC_DATA);

> +	}

> +

> +	return len;

> +}

> +

> +static int rockchip_sfc_read_fifo(struct rockchip_sfc *sfc, u8 *buf, int len)

> +{

> +	u8 bytes = len & 0x3;

> +	u32 dwords;

> +	u8 read_words;

> +	int rx_level;

> +	int tmp;

> +

> +	/* word aligned access only */

> +	dwords = len >> 2;

> +	while (dwords) {

> +		rx_level = rockchip_sfc_wait_fifo_ready(sfc, SFC_CMD_DIR_RD, HZ);

> +		if (rx_level < 0)

> +			return rx_level;

> +		read_words = min_t(u32, rx_level, dwords);

> +		ioread32_rep(sfc->regbase + SFC_DATA, buf, read_words);

> +		buf += read_words << 2;

> +		dwords -= read_words;

> +	}

> +

> +	/* read the rest non word aligned bytes */

> +	if (bytes) {

> +		rx_level = rockchip_sfc_wait_fifo_ready(sfc, SFC_CMD_DIR_RD, HZ);

> +		if (rx_level < 0)

> +			return rx_level;

> +		tmp = readl_relaxed(sfc->regbase + SFC_DATA);

> +		memcpy(buf, &tmp, bytes);

> +	}

> +

> +	return len;

> +}

> +

> +static int rockchip_sfc_fifo_transfer_dma(struct rockchip_sfc *sfc, dma_addr_t dma_buf, size_t len)

> +{

> +	u32 reg;

> +	int err = 0;

> +

> +	init_completion(&sfc->cp);

> +

> +	writel(0xFFFFFFFF, sfc->regbase + SFC_ICLR);

> +

> +	/* Enable transfer complete interrupt */

> +	reg = readl(sfc->regbase + SFC_IMR);

> +	reg &= ~SFC_IMR_DMA;

> +	writel(reg, sfc->regbase + SFC_IMR);

> +	writel((u32)dma_buf, sfc->regbase + SFC_DMA_ADDR);

> +	writel(0x1, sfc->regbase + SFC_DMA_TRIGGER);

> +

> +	/* Wait for the interrupt. */

> +	if (!wait_for_completion_timeout(&sfc->cp, msecs_to_jiffies(2000))) {

> +		dev_err(sfc->dev, "DMA wait for transfer finish timeout\n");

> +		err = -ETIMEDOUT;

> +	}

> +

> +	writel(0xFFFFFFFF, sfc->regbase + SFC_ICLR);

> +	/* Disable transfer finish interrupt */

> +	reg = readl(sfc->regbase + SFC_IMR);

> +	reg |= SFC_IMR_DMA;

> +	writel(reg, sfc->regbase + SFC_IMR);

> +

> +	return len;

> +}

> +

> +static int rockchip_sfc_xfer_data_poll(struct rockchip_sfc *sfc,

> +				       const struct spi_mem_op *op, u32 len)

> +{

> +	dev_dbg(sfc->dev, "xfer_poll len=%x\n", len);

> +

> +	if (op->data.dir == SPI_MEM_DATA_OUT)

> +		return rockchip_sfc_write_fifo(sfc, op->data.buf.out, len);

> +	else

> +		return rockchip_sfc_read_fifo(sfc, op->data.buf.in, len);

> +}

> +

> +static int rockchip_sfc_xfer_data_dma(struct rockchip_sfc *sfc,

> +				      const struct spi_mem_op *op, u32 len)

> +{

> +	int ret;

> +

> +	dev_dbg(sfc->dev, "xfer_dma len=%x\n", len);

> +

> +	if (op->data.dir == SPI_MEM_DATA_OUT) {

> +		memcpy_toio(sfc->buffer, op->data.buf.out, len);

> +		ret = rockchip_sfc_fifo_transfer_dma(sfc, sfc->dma_buffer, len);

> +	} else {

> +		ret = rockchip_sfc_fifo_transfer_dma(sfc, sfc->dma_buffer, len);

> +		memcpy_fromio(op->data.buf.in, sfc->buffer, len);

> +	}

> +

> +	return ret;

> +}

> +

> +static int rockchip_sfc_xfer_done(struct rockchip_sfc *sfc, u32 timeout_us)

> +{

> +	int ret = 0;

> +	u32 status;

> +

> +	ret = readl_poll_timeout(sfc->regbase + SFC_SR, status,

> +				 !(status & SFC_SR_IS_BUSY),

> +				 20, timeout_us);

> +	if (ret) {

> +		dev_err(sfc->dev, "wait sfc idle timeout\n");

> +		rockchip_sfc_reset(sfc);

> +

> +		ret = -EIO;

> +	}

> +

> +	return ret;

> +}

> +

> +static int rockchip_sfc_exec_mem_op(struct spi_mem *mem, const struct spi_mem_op *op)

> +{

> +	struct rockchip_sfc *sfc = spi_master_get_devdata(mem->spi->master);

> +	u32 len = min_t(u32, op->data.nbytes, sfc->max_iosize);


Is it necessary to use the minimum of these 2 values if we use an
adjust_op_size which I propose below?

> +	int ret;

> +


Again I'm asking this as someone with no experience optimizing code
or with compilers, but would it help to add an unlikely() here? It's
all but guaranteed the first time this code path is called the clock
will need to be set, but each subsequent call should not really have
to care since the clock likely won't be changing (as long as we're on
the same CS). We can also then set the sfc->frequency to what the
max_speed_hz was after successfully setting the rate so we can hit the
faster path after the first run (it looks like it's checking the
uninitalized value but never updating it after it's changed, so it's
always doing the clk_set_rate()).

Also, should we clamp the clock speed between the min and max values?

> +	if (mem->spi->max_speed_hz != sfc->frequency) {

> +		if (clk_set_rate(sfc->clk, mem->spi->max_speed_hz))

> +			return ret;

> +		dev_dbg(sfc->dev, "set_freq=%dHz real_freq=%ldHz\n",

> +			sfc->frequency, clk_get_rate(sfc->clk));

> +	}

> +

> +	rockchip_sfc_xfer_setup(sfc, mem, op, len);

> +	if (len) {

> +		if (likely(sfc->use_dma) && !(len & 0x3) && len >= SFC_DMA_TRANS_THRETHOLD)

> +			ret = rockchip_sfc_xfer_data_dma(sfc, op, len);

> +		else

> +			ret = rockchip_sfc_xfer_data_poll(sfc, op, len);

> +

> +		if (ret != len) {

> +			dev_err(sfc->dev, "xfer data failed ret %d dir %d\n", ret, op->data.dir);

> +

> +			return -EIO;

> +		}

> +	}

> +

> +	return rockchip_sfc_xfer_done(sfc, 100000);

> +}

> +


I'll be honest with you, I had no idea if I needed this function or
not (the rockchip_sfc_get_name() function). I'm pretty sure it's not
needed though, so I assume it can be safely removed.

> +static const char *rockchip_sfc_get_name(struct spi_mem *mem)

> +{

> +	return devm_kasprintf(&mem->spi->dev, GFP_KERNEL, "%s.%d",

> +			      dev_name(&mem->spi->dev), mem->spi->chip_select);

> +}

> +


I had an issue with doing dd as mentioned above reading against mtd0,
and was able to fix it by adding a mem_op of adjust_op_size like this.
Not sure if it works for the v4/v5 hardware though.

static int rockchip_sfc_adjust_op_size(struct spi_mem *mem, struct spi_mem_op *op)
{
	struct rockchip_sfc *sfc = spi_master_get_devdata(mem->spi->master);

	op->data.nbytes = min(op->data.nbytes, sfc->max_iosize);
	return 0;
}

> +static const struct spi_controller_mem_ops rockchip_sfc_mem_ops = {

> +	.exec_op = rockchip_sfc_exec_mem_op,

> +	.get_name = rockchip_sfc_get_name,

> +};

> +

> +static irqreturn_t rockchip_sfc_irq_handler(int irq, void *dev_id)

> +{

> +	struct rockchip_sfc *sfc = dev_id;

> +	u32 reg;

> +

> +	reg = readl(sfc->regbase + SFC_RISR);

> +

> +	/* Clear interrupt */

> +	writel_relaxed(reg, sfc->regbase + SFC_ICLR);

> +

> +	if (reg & SFC_RISR_DMA)

> +		complete(&sfc->cp);

> +


From a previous comment, we should only clear the interrupt if we
handle the specific one in question.

> +	return IRQ_HANDLED;

> +}

> +

> +static int rockchip_sfc_probe(struct platform_device *pdev)

> +{

> +	struct device *dev = &pdev->dev;

> +	struct spi_master *master;

> +	struct resource *res;

> +	struct rockchip_sfc *sfc;

> +	int ret;

> +

> +	master = devm_spi_alloc_master(&pdev->dev, sizeof(*sfc));

> +	if (!master)

> +		return -ENOMEM;

> +

> +	master->mem_ops = &rockchip_sfc_mem_ops;

> +	master->dev.of_node = pdev->dev.of_node;

> +	master->mode_bits = SPI_TX_QUAD | SPI_TX_DUAL | SPI_RX_QUAD | SPI_RX_DUAL;

> +	master->min_speed_hz = SFC_MIN_SPEED_HZ;

> +	master->max_speed_hz = SFC_MAX_SPEED_HZ;

> +	master->num_chipselect = SFC_MAX_CHIPSELECT_NUM;

> +

> +	sfc = spi_master_get_devdata(master);

> +	sfc->dev = dev;

> +

> +	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);

> +	sfc->regbase = devm_ioremap_resource(dev, res);

> +	if (IS_ERR(sfc->regbase))

> +		return PTR_ERR(sfc->regbase);

> +

> +	sfc->clk = devm_clk_get(&pdev->dev, "clk_sfc");

> +	if (IS_ERR(sfc->clk)) {

> +		dev_err(&pdev->dev, "Failed to get sfc interface clk\n");

> +		return PTR_ERR(sfc->clk);

> +	}

> +

> +	sfc->hclk = devm_clk_get(&pdev->dev, "hclk_sfc");

> +	if (IS_ERR(sfc->hclk)) {

> +		dev_err(&pdev->dev, "Failed to get sfc ahb clk\n");

> +		return PTR_ERR(sfc->hclk);

> +	}

> +

> +	sfc->use_dma = !of_property_read_bool(sfc->dev->of_node,

> +					      "rockchip,sfc-no-dma");

> +

> +	if (sfc->use_dma) {

> +		ret = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));

> +		if (ret) {

> +			dev_warn(dev, "Unable to set dma mask\n");

> +			return ret;

> +		}

> +

> +		sfc->buffer = dmam_alloc_coherent(dev, SFC_MAX_IOSIZE_VER3,

> +						  &sfc->dma_buffer,

> +						  GFP_KERNEL);

> +		if (!sfc->buffer)

> +			return -ENOMEM;

> +	}

> +

> +	ret = clk_prepare_enable(sfc->hclk);

> +	if (ret) {

> +		dev_err(&pdev->dev, "Failed to enable ahb clk\n");

> +		goto err_hclk;

> +	}

> +

> +	ret = clk_prepare_enable(sfc->clk);

> +	if (ret) {

> +		dev_err(&pdev->dev, "Failed to enable interface clk\n");

> +		goto err_clk;

> +	}

> +

> +	/* Find the irq */

> +	ret = platform_get_irq(pdev, 0);

> +	if (ret < 0) {

> +		dev_err(dev, "Failed to get the irq\n");

> +		goto err_irq;

> +	}

> +

> +	ret = devm_request_irq(dev, ret, rockchip_sfc_irq_handler,

> +			       0, pdev->name, sfc);

> +	if (ret) {

> +		dev_err(dev, "Failed to request irq\n");

> +

> +		return ret;

> +	}

> +

> +	ret = rockchip_sfc_init(sfc);

> +	if (ret)

> +		goto err_irq;

> +

> +	sfc->max_iosize = rockchip_sfc_get_max_iosize(sfc);

> +

> +	ret = spi_register_master(master);

> +	if (ret)

> +		goto err_irq;

> +

> +	return 0;

> +

> +err_irq:

> +	clk_disable_unprepare(sfc->clk);

> +err_clk:

> +	clk_disable_unprepare(sfc->hclk);

> +err_hclk:

> +	return ret;

> +}

> +

> +static int rockchip_sfc_remove(struct platform_device *pdev)

> +{

> +	struct spi_master *master = platform_get_drvdata(pdev);

> +	struct rockchip_sfc *sfc = platform_get_drvdata(pdev);

> +

> +	spi_unregister_master(master);

> +

> +	clk_disable_unprepare(sfc->clk);

> +	clk_disable_unprepare(sfc->hclk);

> +

> +	return 0;

> +}

> +

> +static const struct of_device_id rockchip_sfc_dt_ids[] = {

> +	{ .compatible = "rockchip,px30-sfc"},

> +	{ .compatible = "rockchip,rk3036-sfc"},

> +	{ .compatible = "rockchip,rk3308-sfc"},

> +	{ .compatible = "rockchip,rv1126-sfc"},

> +	{ /* sentinel */ }

> +};

> +MODULE_DEVICE_TABLE(of, rockchip_sfc_dt_ids);

> +

> +static struct platform_driver rockchip_sfc_driver = {

> +	.driver = {

> +		.name	= "rockchip-sfc",

> +		.of_match_table = rockchip_sfc_dt_ids,

> +	},

> +	.probe	= rockchip_sfc_probe,

> +	.remove	= rockchip_sfc_remove,

> +};

> +module_platform_driver(rockchip_sfc_driver);

> +

> +MODULE_LICENSE("GPL v2");

> +MODULE_DESCRIPTION("Rockchip Serial Flash Controller Driver");

> +MODULE_AUTHOR("Shawn Lin <shawn.lin@rock-chips.com>");

> +MODULE_AUTHOR("Chris Morgan <macroalpha82@gmail.com>");


I know folks will hate on me for this (and for good reason given it
messes up IDs and stuff), but I prefer macromorgan@hotmail.com if my
email is in here, as that's the one I've been using for decades...

> +MODULE_AUTHOR("Jon Lin <Jon.lin@rock-chips.com>");

> -- 

> 2.17.1

> 

> 

>
Jon Lin June 9, 2021, 3:50 a.m. UTC | #4
On 6/9/21 2:03 AM, Johan Jonker wrote:
> Hi Jon,

>

> A look at the build log shows this error:

>

> https://patchwork.ozlabs.org/project/devicetree-bindings/patch/20210608022644.21074-2-jon.lin@rock-chips.com/

>

> dtschema/dtc warnings/errors:

> /builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.example.dt.yaml:

> spi@ff3a0000: clock-names:0: 'hclk_sfc' was expected

> 	From schema:

> /builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml

> /builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.example.dt.yaml:

> spi@ff3a0000: clock-names:1: 'clk_sfc' was expected

> 	From schema:

> /builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml

>

> Check document with:

>

> make ARCH=arm dt_binding_check

> DT_SCHEMA_FILES=Documentation/devicetree/bindings/spi/rockchip-sfc.yaml

>

> make ARCH=arm dtbs_check

> DT_SCHEMA_FILES=Documentation/devicetree/bindings/spi/rockchip-sfc.yaml

>

> make ARCH=arm64 dtbs_check

> DT_SCHEMA_FILES=Documentation/devicetree/bindings/spi/rockchip-sfc.yaml

>

> Remove any errors before submitting.

>

> ===

>

> Johan

>

> On 6/8/21 4:26 AM, Jon Lin wrote:

>> From: Chris Morgan <macromorgan@hotmail.com>

>>

>> Add bindings for the Rockchip serial flash controller. New device

>> specific parameter of rockchip,sfc-no-dma included in documentation.

>>

>> Signed-off-by: Chris Morgan <macromorgan@hotmail.com>

>> Signed-off-by: Jon Lin <jon.lin@rock-chips.com>

>> ---

>>

>> Changes in v6:

>> - Add support in device trees for rv1126(Declared in series 5 but not

>>    submitted)

>> - Change to use "clk_sfc" "hclk_sfc" as clock lable, since it does not

>>    affect interpretation and has been widely used

>> - Support sfc tx_dual, tx_quad(Declared in series 5 but not submitted)

>> - Simplify the code, such as remove "rockchip_sfc_register_all"(Declared

>>    in series 5 but not submitted)

>> - Support SFC ver4 ver5(Declared in series 5 but not submitted)

>> - Add author Chris Morgan and Jon Lin to spi-rockchip-sfc.c

>> - Change to use devm_spi_alloc_master and spi_unregister_master

>>

>> Changes in v5:

>> - Add support in device trees for rv1126

>> - Support sfc tx_dual, tx_quad

>> - Simplify the code, such as remove "rockchip_sfc_register_all"

>> - Support SFC ver4 ver5

>>

>> Changes in v4:

>> - Changing patch back to an "RFC". An engineer from Rockchip

>>    reached out to me to let me know they are working on this patch for

>>    upstream, I am submitting this v4 for the community to see however

>>    I expect Jon Lin (jon.lin@rock-chips.com) will submit new patches

>>    soon and these are the ones we should pursue for mainlining. Jon's

>>    patch series should include support for more hardware than this

>>    series.

>> - Clean up documentation more and ensure it is correct per

>>    make dt_binding_check.

>> - Add support in device trees for rk3036, rk3308, and rv1108.

>> - Add ahb clock (hclk_sfc) support for rk3036.

>> - Change rockchip_sfc_wait_fifo_ready() to use a switch statement.

>> - Change IRQ code to only mark IRQ as handled if it handles the

>>    specific IRQ (DMA transfer finish) it is supposed to handle.

>>

>> Changes in v3:

>> - Changed the name of the clocks to sfc/ahb (from clk-sfc/clk-hsfc).

>> - Changed the compatible string from rockchip,sfc to

>>    rockchip,rk3036-sfc. A quick glance at the datasheets suggests this

>>    driver should work for the PX30, RK180x, RK3036, RK312x, RK3308 and

>>    RV1108 SoCs, and possibly more. However, I am currently only able

>>    to test this on a PX30 (an RK3326). The technical reference manuals

>>    appear to list the same registers for each device.

>> - Corrected devicetree documentation for formatting and to note these

>>    changes.

>> - Replaced the maintainer with Heiko Stuebner and myself, as we will

>>    take ownership of this going forward.

>> - Noted that the device (per the reference manual) supports 4 CS, but

>>    I am only able to test a single CS (CS 0).

>> - Reordered patches to comply with upstream rules.

>>

>> Changes in v2:

>> - Reimplemented driver using spi-mem subsystem.

>> - Removed power management code as I couldn't get it working properly.

>> - Added device tree bindings for Odroid Go Advance.

>>

>> Changes in v1:

>> hanges made in this new series versus the v8 of the old series:

>> - Added function to read spi-rx-bus-width from device tree, in the

>>    event that the SPI chip supports 4x mode but only has 2 pins

>>    wired (such as the Odroid Go Advance).

>> - Changed device tree documentation from txt to yaml format.

>> - Made "reset" message a dev_dbg from a dev_info.

>> - Changed read and write fifo functions to remove redundant checks.

>> - Changed the write and read from relaxed to non-relaxed when

>>    starting the DMA transfer or reading the DMA IRQ.

>> - Changed from dma_coerce_mask_and_coherent to just

>>    dma_set_mask_and_coherent.

>> - Changed name of get_if_type to rockchip_sfc_get_if_type.

>>

>>   .../devicetree/bindings/spi/rockchip-sfc.yaml | 87 +++++++++++++++++++

>>   1 file changed, 87 insertions(+)

>>   create mode 100644 Documentation/devicetree/bindings/spi/rockchip-sfc.yaml

>>

>> diff --git a/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml b/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml

>> new file mode 100644

>> index 000000000000..160449713f97

>> --- /dev/null

>> +++ b/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml

>> @@ -0,0 +1,87 @@

>> +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)

>> +%YAML 1.2

>> +---

>> +$id: http://devicetree.org/schemas/spi/rockchip-sfc.yaml#

>> +$schema: http://devicetree.org/meta-schemas/core.yaml#

>> +

>> +title: Rockchip Serial Flash Controller (SFC)

>> +

>> +maintainers:

>> +  - Heiko Stuebner <heiko@sntech.de>

>> +  - Chris Morgan <macromorgan@hotmail.com>

> In the driver spi-rockchip-sfc.c is used:

> Chris Morgan <macroalpha82@gmail.com>

> Maybe use a consistent email address?

> Also Hotmail does strange things to message-ID's and links/content

> inside patches.

>

>> +

>> +allOf:

>> +  - $ref: spi-controller.yaml#

>> +

>> +properties:

>> +  compatible:

>> +    oneOf:

>> +      - const: rockchip,rk3036-sfc

>> +      - items:

>> +          - enum:

>> +              - rockchip,px30-sfc

>> +              - rockchip,rk3308-sfc

>> +              - rockchip,rv1108-sfc

>> +          - const: rockchip,rk3036-sfc

>> +      - const: rockchip,rv1126-sfc

> A look at the driver shows this:

>

> +static const struct of_device_id rockchip_sfc_dt_ids[] = {

>

> +	{ .compatible = "rockchip,px30-sfc"},

>

> remove

>

> +	{ .compatible = "rockchip,rk3036-sfc"},

>

> +	{ .compatible = "rockchip,rk3308-sfc"},

> +	{ .compatible = "rockchip,rv1126-sfc"},

>

> remove

>

> +	{ /* sentinel */ }

> +};

>

> When no '.data = &my_sfc_data' exist then there's no need for additional

> compatible strings in the driver.

> ===

> example for I2S:

> 	{ .compatible = "rockchip,rk3399-i2s", .data = &rk3399_i2s_pins },

> ===

> Compatibility strings are supposed to be SoC orientated.

> With rockchip_sfc_get_version() all we need is one fall back string for

> now I think. Is rk3568 identical. Please advise.

>

>

>        - const: rockchip,rk3036-sfc

>        - items:

>            - enum:

>                - rockchip,px30-sfc

>                - rockchip,rk3308-sfc

>                - rockchip,rk3368-sfc

>                - rockchip,rk3568-sfc

>                - rockchip,rv1108-sfc

>                - rockchip,rv1126-sfc

>            - const: rockchip,rk3036-sfc

>

> ===

>

> The rk3368 TRM also describes a SFC controller.

> For the mainline DT completeness is it possible to add a rk3368 SFC node

> as well? ;)

>

>> +

>> +  reg:

>> +    maxItems: 1

>> +

>> +  interrupts:

>> +    maxItems: 1

>> +

>> +  clocks:

>> +    items:

>> +      - description: Bus Clock

>> +      - description: Module Clock

>> +

>> +  clock-names:

>> +    items:

>> +      - const: hclk_sfc

>> +      - const: clk_sfc

>> +

>> +  power-domains:

>> +    maxItems: 1

>> +

>> +  rockchip,sfc-no-dma:

>> +    description: Disable DMA and utilize FIFO mode only

>> +    type: boolean

>> +

>> +required:

>> +  - compatible

>> +  - reg

>> +  - interrupts

>> +  - clocks

>> +  - clock-names

>> +

>> +unevaluatedProperties: false

>> +

>> +examples:

>> +  - |

>> +    #include <dt-bindings/clock/px30-cru.h>

>> +    #include <dt-bindings/interrupt-controller/arm-gic.h>

>> +    #include <dt-bindings/power/px30-power.h>

>> +

>> +    sfc: spi@ff3a0000 {

>> +        compatible = "rockchip,px30-sfc","rockchip,rk3036-sfc";

>> +        reg = <0xff3a0000 0x4000>;

>> +        interrupts = <GIC_SPI 56 IRQ_TYPE_LEVEL_HIGH>;

>> +        clocks = <&cru SCLK_SFC>, <&cru HCLK_SFC>;

>> +        clock-names = "clk_sfc", "hclk_sfc";

> The clocks in the examples and dtsi files must have the same sort order.

>

>> +        pinctrl-0 = <&sfc_clk &sfc_cs &sfc_bus2>;

>> +        pinctrl-names = "default";

>> +        power-domains = <&power PX30_PD_MMC_NAND>;

>> +        #address-cells = <1>;

>> +        #size-cells = <0>;

>> +

>> +        flash@0 {

> >From spi-controller.yaml:

>

> patternProperties:

>    "^.*@[0-9a-f]+$":

>      type: object

>

>      properties:

>        compatible:

>          description:

>            Compatible of the SPI device.

>

>        reg:

>          minimum: 0

>

>          maximum: 256

>

>          description:

>            Chip select used by the device.

>

> This allows 257 sub nodes.

> Support up to 4 chip select in the driver.

> #define SFC_MAX_CHIPSELECT_NUM		4

RK SFC design is up to 4 CS.

Sorry, I don't see any specific improvement in this comment, can you 
explain it in detail.
>

>> +            compatible = "jedec,spi-nor";

>> +            reg = <0>;

>> +            spi-max-frequency = <108000000>;

>> +            spi-rx-bus-width = <2>;

>> +            spi-tx-bus-width = <2>;

>> +        };

>> +    };

>> +

>> +...

>>

>

>
Johan Jonker June 9, 2021, 7:13 a.m. UTC | #5
On 6/9/21 5:50 AM, Jon Lin wrote:
> 
> On 6/9/21 2:03 AM, Johan Jonker wrote:
>> Hi Jon,
>>
>> A look at the build log shows this error:
>>
>> https://patchwork.ozlabs.org/project/devicetree-bindings/patch/20210608022644.21074-2-jon.lin@rock-chips.com/
>>
>>
>> dtschema/dtc warnings/errors:
>> /builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.example.dt.yaml:
>>
>> spi@ff3a0000: clock-names:0: 'hclk_sfc' was expected
>>     From schema:
>> /builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
>>
>> /builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.example.dt.yaml:
>>
>> spi@ff3a0000: clock-names:1: 'clk_sfc' was expected
>>     From schema:
>> /builds/robherring/linux-dt-review/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
>>
>>
>> Check document with:
>>
>> make ARCH=arm dt_binding_check
>> DT_SCHEMA_FILES=Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
>>
>> make ARCH=arm dtbs_check
>> DT_SCHEMA_FILES=Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
>>
>> make ARCH=arm64 dtbs_check
>> DT_SCHEMA_FILES=Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
>>
>> Remove any errors before submitting.
>>
>> ===
>>
>> Johan
>>
>> On 6/8/21 4:26 AM, Jon Lin wrote:
>>> From: Chris Morgan <macromorgan@hotmail.com>
>>>
>>> Add bindings for the Rockchip serial flash controller. New device
>>> specific parameter of rockchip,sfc-no-dma included in documentation.
>>>
>>> Signed-off-by: Chris Morgan <macromorgan@hotmail.com>
>>> Signed-off-by: Jon Lin <jon.lin@rock-chips.com>
>>> ---
>>>
>>> Changes in v6:
>>> - Add support in device trees for rv1126(Declared in series 5 but not
>>>    submitted)
>>> - Change to use "clk_sfc" "hclk_sfc" as clock lable, since it does not
>>>    affect interpretation and has been widely used
>>> - Support sfc tx_dual, tx_quad(Declared in series 5 but not submitted)
>>> - Simplify the code, such as remove "rockchip_sfc_register_all"(Declared
>>>    in series 5 but not submitted)
>>> - Support SFC ver4 ver5(Declared in series 5 but not submitted)
>>> - Add author Chris Morgan and Jon Lin to spi-rockchip-sfc.c
>>> - Change to use devm_spi_alloc_master and spi_unregister_master
>>>
>>> Changes in v5:
>>> - Add support in device trees for rv1126
>>> - Support sfc tx_dual, tx_quad
>>> - Simplify the code, such as remove "rockchip_sfc_register_all"
>>> - Support SFC ver4 ver5
>>>
>>> Changes in v4:
>>> - Changing patch back to an "RFC". An engineer from Rockchip
>>>    reached out to me to let me know they are working on this patch for
>>>    upstream, I am submitting this v4 for the community to see however
>>>    I expect Jon Lin (jon.lin@rock-chips.com) will submit new patches
>>>    soon and these are the ones we should pursue for mainlining. Jon's
>>>    patch series should include support for more hardware than this
>>>    series.
>>> - Clean up documentation more and ensure it is correct per
>>>    make dt_binding_check.
>>> - Add support in device trees for rk3036, rk3308, and rv1108.
>>> - Add ahb clock (hclk_sfc) support for rk3036.
>>> - Change rockchip_sfc_wait_fifo_ready() to use a switch statement.
>>> - Change IRQ code to only mark IRQ as handled if it handles the
>>>    specific IRQ (DMA transfer finish) it is supposed to handle.
>>>
>>> Changes in v3:
>>> - Changed the name of the clocks to sfc/ahb (from clk-sfc/clk-hsfc).
>>> - Changed the compatible string from rockchip,sfc to
>>>    rockchip,rk3036-sfc. A quick glance at the datasheets suggests this
>>>    driver should work for the PX30, RK180x, RK3036, RK312x, RK3308 and
>>>    RV1108 SoCs, and possibly more. However, I am currently only able
>>>    to test this on a PX30 (an RK3326). The technical reference manuals
>>>    appear to list the same registers for each device.
>>> - Corrected devicetree documentation for formatting and to note these
>>>    changes.
>>> - Replaced the maintainer with Heiko Stuebner and myself, as we will
>>>    take ownership of this going forward.
>>> - Noted that the device (per the reference manual) supports 4 CS, but
>>>    I am only able to test a single CS (CS 0).
>>> - Reordered patches to comply with upstream rules.
>>>
>>> Changes in v2:
>>> - Reimplemented driver using spi-mem subsystem.
>>> - Removed power management code as I couldn't get it working properly.
>>> - Added device tree bindings for Odroid Go Advance.
>>>
>>> Changes in v1:
>>> hanges made in this new series versus the v8 of the old series:
>>> - Added function to read spi-rx-bus-width from device tree, in the
>>>    event that the SPI chip supports 4x mode but only has 2 pins
>>>    wired (such as the Odroid Go Advance).
>>> - Changed device tree documentation from txt to yaml format.
>>> - Made "reset" message a dev_dbg from a dev_info.
>>> - Changed read and write fifo functions to remove redundant checks.
>>> - Changed the write and read from relaxed to non-relaxed when
>>>    starting the DMA transfer or reading the DMA IRQ.
>>> - Changed from dma_coerce_mask_and_coherent to just
>>>    dma_set_mask_and_coherent.
>>> - Changed name of get_if_type to rockchip_sfc_get_if_type.
>>>
>>>   .../devicetree/bindings/spi/rockchip-sfc.yaml | 87 +++++++++++++++++++
>>>   1 file changed, 87 insertions(+)
>>>   create mode 100644
>>> Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
>>>
>>> diff --git a/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
>>> b/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
>>> new file mode 100644
>>> index 000000000000..160449713f97
>>> --- /dev/null
>>> +++ b/Documentation/devicetree/bindings/spi/rockchip-sfc.yaml
>>> @@ -0,0 +1,87 @@
>>> +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
>>> +%YAML 1.2
>>> +---
>>> +$id: http://devicetree.org/schemas/spi/rockchip-sfc.yaml#
>>> +$schema: http://devicetree.org/meta-schemas/core.yaml#
>>> +
>>> +title: Rockchip Serial Flash Controller (SFC)
>>> +
>>> +maintainers:
>>> +  - Heiko Stuebner <heiko@sntech.de>
>>> +  - Chris Morgan <macromorgan@hotmail.com>
>> In the driver spi-rockchip-sfc.c is used:
>> Chris Morgan <macroalpha82@gmail.com>
>> Maybe use a consistent email address?
>> Also Hotmail does strange things to message-ID's and links/content
>> inside patches.
>>
>>> +
>>> +allOf:
>>> +  - $ref: spi-controller.yaml#
>>> +
>>> +properties:
>>> +  compatible:
>>> +    oneOf:
>>> +      - const: rockchip,rk3036-sfc
>>> +      - items:
>>> +          - enum:
>>> +              - rockchip,px30-sfc
>>> +              - rockchip,rk3308-sfc
>>> +              - rockchip,rv1108-sfc
>>> +          - const: rockchip,rk3036-sfc
>>> +      - const: rockchip,rv1126-sfc
>> A look at the driver shows this:
>>
>> +static const struct of_device_id rockchip_sfc_dt_ids[] = {
>>
>> +    { .compatible = "rockchip,px30-sfc"},
>>
>> remove
>>
>> +    { .compatible = "rockchip,rk3036-sfc"},
>>
>> +    { .compatible = "rockchip,rk3308-sfc"},
>> +    { .compatible = "rockchip,rv1126-sfc"},
>>
>> remove
>>
>> +    { /* sentinel */ }
>> +};
>>
>> When no '.data = &my_sfc_data' exist then there's no need for additional
>> compatible strings in the driver.
>> ===
>> example for I2S:
>>     { .compatible = "rockchip,rk3399-i2s", .data = &rk3399_i2s_pins },
>> ===
>> Compatibility strings are supposed to be SoC orientated.
>> With rockchip_sfc_get_version() all we need is one fall back string for
>> now I think. Is rk3568 identical. Please advise.
>>
>>
>>        - const: rockchip,rk3036-sfc
>>        - items:
>>            - enum:
>>                - rockchip,px30-sfc
>>                - rockchip,rk3308-sfc
>>                - rockchip,rk3368-sfc
>>                - rockchip,rk3568-sfc
>>                - rockchip,rv1108-sfc
>>                - rockchip,rv1126-sfc
>>            - const: rockchip,rk3036-sfc
>>
>> ===
>>
>> The rk3368 TRM also describes a SFC controller.
>> For the mainline DT completeness is it possible to add a rk3368 SFC node
>> as well? ;)
>>
>>> +
>>> +  reg:
>>> +    maxItems: 1
>>> +
>>> +  interrupts:
>>> +    maxItems: 1
>>> +
>>> +  clocks:
>>> +    items:
>>> +      - description: Bus Clock
>>> +      - description: Module Clock
>>> +
>>> +  clock-names:
>>> +    items:
>>> +      - const: hclk_sfc
>>> +      - const: clk_sfc
>>> +
>>> +  power-domains:
>>> +    maxItems: 1
>>> +
>>> +  rockchip,sfc-no-dma:
>>> +    description: Disable DMA and utilize FIFO mode only
>>> +    type: boolean
>>> +
>>> +required:
>>> +  - compatible
>>> +  - reg
>>> +  - interrupts
>>> +  - clocks
>>> +  - clock-names
>>> +
>>> +unevaluatedProperties: false
>>> +
>>> +examples:
>>> +  - |
>>> +    #include <dt-bindings/clock/px30-cru.h>
>>> +    #include <dt-bindings/interrupt-controller/arm-gic.h>
>>> +    #include <dt-bindings/power/px30-power.h>
>>> +
>>> +    sfc: spi@ff3a0000 {
>>> +        compatible = "rockchip,px30-sfc","rockchip,rk3036-sfc";
>>> +        reg = <0xff3a0000 0x4000>;
>>> +        interrupts = <GIC_SPI 56 IRQ_TYPE_LEVEL_HIGH>;
>>> +        clocks = <&cru SCLK_SFC>, <&cru HCLK_SFC>;
>>> +        clock-names = "clk_sfc", "hclk_sfc";
>> The clocks in the examples and dtsi files must have the same sort order.
>>
>>> +        pinctrl-0 = <&sfc_clk &sfc_cs &sfc_bus2>;
>>> +        pinctrl-names = "default";
>>> +        power-domains = <&power PX30_PD_MMC_NAND>;
>>> +        #address-cells = <1>;
>>> +        #size-cells = <0>;
>>> +
>>> +        flash@0 {
>> >From spi-controller.yaml:
>>
>> patternProperties:
>>    "^.*@[0-9a-f]+$":
>>      type: object
>>
>>      properties:
>>        compatible:
>>          description:
>>            Compatible of the SPI device.
>>
>>        reg:
>>          minimum: 0
>>
>>          maximum: 256
>>
>>          description:
>>            Chip select used by the device.
>>
>> This allows 257 sub nodes.
>> Support up to 4 chip select in the driver.
>> #define SFC_MAX_CHIPSELECT_NUM        4

> RK SFC design is up to 4 CS.
> 
> Sorry, I don't see any specific improvement in this comment, can you
> explain it in detail.

In the old text documents it only described the hardware.
With the new YAML format we also check new dts submissions for bogus
properties. So if someone comes up with this example:

flash@88 {
  compatible = "jedec,spi-nor";
  reg = <0x88>;
}

This must give a out of range notification.

Also when someone tries to connect something else then flash that must
be detected. The spi-controller.yaml allows any node name.

 patternProperties:
    "^flash@[0-3]$":
      type: object

      properties:
        compatible:
          const: jedec,spi-nor

        reg:
          minimum: 0
          maximum: 3

===

Example from rockchip,nand-controller.yaml:

patternProperties:
  "^nand@[0-7]$":
    type: object
    properties:
      reg:
        minimum: 0
        maximum: 7

===

>>
>>> +            compatible = "jedec,spi-nor";
>>> +            reg = <0>;
>>> +            spi-max-frequency = <108000000>;
>>> +            spi-rx-bus-width = <2>;
>>> +            spi-tx-bus-width = <2>;
>>> +        };
>>> +    };
>>> +
>>> +...
>>>
>>
>>
> 
>
Jon Lin June 9, 2021, 1:48 p.m. UTC | #6
On 6/9/21 10:36 AM, Chris Morgan wrote:
> On Tue, Jun 08, 2021 at 10:26:38AM +0800, Jon Lin wrote:
>> From: Chris Morgan <macromorgan@hotmail.com>
>>
>> Add the rockchip serial flash controller (SFC) driver.
>>
>> Signed-off-by: Chris Morgan <macromorgan@hotmail.com>
>> Signed-off-by: Jon Lin <jon.lin@rock-chips.com>
> I think I hit "reply" earlier when I meant to hit "reply-all". Whoops.
>
> I wanted to say that for now this driver is not working for me on the
> v3 hardware. There appears to be something wrong with alternating
> between page programs and erase. I can consistently reproduce this by
> writing a 4MB image of random data to my SPI flash chip at offset
> c00000 using dd to mtdblock0. I've dumped the contents of what is
> written to the controller and it appears at some point it goes from
> doing a page program (02) to a dual io read (bb) to a sector erase
> (20), then it repeatedly reads the status register (05) and I assume
> the register never changes. As a result it just keeps looping over
> the status register then eventually times out. I'm still debugging
> to try and figure out exactly what is going on though.
>
> I also noticed that reading does not work consistently using dd
> from the mtd0 device, however I have a proposed fix for that which
> I list below at the appropriate section.
>
Can you enable dev_dbg in rockchip_sfc_xfer_setup, then send me log.

I successfully mount and test spinor jffs2 to confirm read/write/erase 
work work. I'll try more cases further.

>> ---
>>
>> Changes in v6: None
>> Changes in v5: None
>> Changes in v4: None
>> Changes in v3: None
>> Changes in v2: None
>> Changes in v1: None
>>
>>   drivers/spi/Kconfig            |   9 +
>>   drivers/spi/Makefile           |   1 +
>>   drivers/spi/spi-rockchip-sfc.c | 660 +++++++++++++++++++++++++++++++++
>>   3 files changed, 670 insertions(+)
>>   create mode 100644 drivers/spi/spi-rockchip-sfc.c
>>
>> diff --git a/drivers/spi/Kconfig b/drivers/spi/Kconfig
>> index e71a4c514f7b..d89e5f3c9107 100644
>> --- a/drivers/spi/Kconfig
>> +++ b/drivers/spi/Kconfig
>> @@ -658,6 +658,15 @@ config SPI_ROCKCHIP
>>   	  The main usecase of this controller is to use spi flash as boot
>>   	  device.
>>   
>> +config SPI_ROCKCHIP_SFC
>> +	tristate "Rockchip Serial Flash Controller (SFC)"
>> +	depends on ARCH_ROCKCHIP || COMPILE_TEST
>> +	depends on HAS_IOMEM && HAS_DMA
>> +	help
>> +	  This enables support for Rockchip serial flash controller. This
>> +	  is a specialized controller used to access SPI flash on some
>> +	  Rockchip SOCs.
>> +
>>   config SPI_RB4XX
>>   	tristate "Mikrotik RB4XX SPI master"
>>   	depends on SPI_MASTER && ATH79
>> diff --git a/drivers/spi/Makefile b/drivers/spi/Makefile
>> index 13e54c45e9df..699db95c8441 100644
>> --- a/drivers/spi/Makefile
>> +++ b/drivers/spi/Makefile
>> @@ -95,6 +95,7 @@ obj-$(CONFIG_SPI_QCOM_GENI)		+= spi-geni-qcom.o
>>   obj-$(CONFIG_SPI_QCOM_QSPI)		+= spi-qcom-qspi.o
>>   obj-$(CONFIG_SPI_QUP)			+= spi-qup.o
>>   obj-$(CONFIG_SPI_ROCKCHIP)		+= spi-rockchip.o
>> +obj-$(CONFIG_SPI_ROCKCHIP_SFC)		+= spi-rockchip-sfc.o
>>   obj-$(CONFIG_SPI_RB4XX)			+= spi-rb4xx.o
>>   obj-$(CONFIG_MACH_REALTEK_RTL)		+= spi-realtek-rtl.o
>>   obj-$(CONFIG_SPI_RPCIF)			+= spi-rpc-if.o
>> diff --git a/drivers/spi/spi-rockchip-sfc.c b/drivers/spi/spi-rockchip-sfc.c
>> new file mode 100644
>> index 000000000000..ec25ad278096
>> --- /dev/null
>> +++ b/drivers/spi/spi-rockchip-sfc.c
>> @@ -0,0 +1,660 @@
>> +// SPDX-License-Identifier: GPL-2.0-only
>> +/*
>> + * Rockchip Serial Flash Controller Driver
>> + *
>> + * Copyright (c) 2017-2021, Rockchip Inc.
>> + * Author: Shawn Lin <shawn.lin@rock-chips.com>
>> + *	   Chris Morgan <macroalpha82@gmail.com>
>> + *	   Jon Lin <Jon.lin@rock-chips.com>
>> + */
>> +
>> +#include <linux/bitops.h>
>> +#include <linux/clk.h>
>> +#include <linux/completion.h>
>> +#include <linux/dma-mapping.h>
>> +#include <linux/iopoll.h>
>> +#include <linux/mm.h>
>> +#include <linux/module.h>
>> +#include <linux/of.h>
>> +#include <linux/platform_device.h>
>> +#include <linux/slab.h>
>> +#include <linux/interrupt.h>
>> +#include <linux/spi/spi-mem.h>
>> +
>> +/* System control */
>> +#define SFC_CTRL			0x0
>> +#define  SFC_CTRL_PHASE_SEL_NEGETIVE	BIT(1)
>> +#define  SFC_CTRL_CMD_BITS_SHIFT	8
>> +#define  SFC_CTRL_ADDR_BITS_SHIFT	10
>> +#define  SFC_CTRL_DATA_BITS_SHIFT	12
>> +
>> +/* Interrupt mask */
>> +#define SFC_IMR				0x4
>> +#define  SFC_IMR_RX_FULL		BIT(0)
>> +#define  SFC_IMR_RX_UFLOW		BIT(1)
>> +#define  SFC_IMR_TX_OFLOW		BIT(2)
>> +#define  SFC_IMR_TX_EMPTY		BIT(3)
>> +#define  SFC_IMR_TRAN_FINISH		BIT(4)
>> +#define  SFC_IMR_BUS_ERR		BIT(5)
>> +#define  SFC_IMR_NSPI_ERR		BIT(6)
>> +#define  SFC_IMR_DMA			BIT(7)
>> +
>> +/* Interrupt clear */
>> +#define SFC_ICLR			0x8
>> +#define  SFC_ICLR_RX_FULL		BIT(0)
>> +#define  SFC_ICLR_RX_UFLOW		BIT(1)
>> +#define  SFC_ICLR_TX_OFLOW		BIT(2)
>> +#define  SFC_ICLR_TX_EMPTY		BIT(3)
>> +#define  SFC_ICLR_TRAN_FINISH		BIT(4)
>> +#define  SFC_ICLR_BUS_ERR		BIT(5)
>> +#define  SFC_ICLR_NSPI_ERR		BIT(6)
>> +#define  SFC_ICLR_DMA			BIT(7)
>> +
>> +/* FIFO threshold level */
>> +#define SFC_FTLR			0xc
>> +#define  SFC_FTLR_TX_SHIFT		0
>> +#define  SFC_FTLR_TX_MASK		0x1f
>> +#define  SFC_FTLR_RX_SHIFT		8
>> +#define  SFC_FTLR_RX_MASK		0x1f
>> +
>> +/* Reset FSM and FIFO */
>> +#define SFC_RCVR			0x10
>> +#define  SFC_RCVR_RESET			BIT(0)
>> +
>> +/* Enhanced mode */
>> +#define SFC_AX				0x14
>> +
>> +/* Address Bit number */
>> +#define SFC_ABIT			0x18
>> +
>> +/* Interrupt status */
>> +#define SFC_ISR				0x1c
>> +#define  SFC_ISR_RX_FULL_SHIFT		BIT(0)
>> +#define  SFC_ISR_RX_UFLOW_SHIFT		BIT(1)
>> +#define  SFC_ISR_TX_OFLOW_SHIFT		BIT(2)
>> +#define  SFC_ISR_TX_EMPTY_SHIFT		BIT(3)
>> +#define  SFC_ISR_TX_FINISH_SHIFT	BIT(4)
>> +#define  SFC_ISR_BUS_ERR_SHIFT		BIT(5)
>> +#define  SFC_ISR_NSPI_ERR_SHIFT		BIT(6)
>> +#define  SFC_ISR_DMA_SHIFT		BIT(7)
>> +
>> +/* FIFO status */
>> +#define SFC_FSR				0x20
>> +#define  SFC_FSR_TX_IS_FULL		BIT(0)
>> +#define  SFC_FSR_TX_IS_EMPTY		BIT(1)
>> +#define  SFC_FSR_RX_IS_EMPTY		BIT(2)
>> +#define  SFC_FSR_RX_IS_FULL		BIT(3)
>> +#define  SFC_FSR_TXLV_MASK		GENMASK(12, 8)
>> +#define  SFC_FSR_TXLV_SHIFT		8
>> +#define  SFC_FSR_RXLV_MASK		GENMASK(20, 16)
>> +#define  SFC_FSR_RXLV_SHIFT		16
>> +
>> +/* FSM status */
>> +#define SFC_SR				0x24
>> +#define  SFC_SR_IS_IDLE			0x0
>> +#define  SFC_SR_IS_BUSY			0x1
>> +
>> +/* Raw interrupt status */
>> +#define SFC_RISR			0x28
>> +#define  SFC_RISR_RX_FULL		BIT(0)
>> +#define  SFC_RISR_RX_UNDERFLOW		BIT(1)
>> +#define  SFC_RISR_TX_OVERFLOW		BIT(2)
>> +#define  SFC_RISR_TX_EMPTY		BIT(3)
>> +#define  SFC_RISR_TRAN_FINISH		BIT(4)
>> +#define  SFC_RISR_BUS_ERR		BIT(5)
>> +#define  SFC_RISR_NSPI_ERR		BIT(6)
>> +#define  SFC_RISR_DMA			BIT(7)
>> +
>> +/* Version */
>> +#define SFC_VER				0x2C
>> +#define  SFC_VER_3			0x3
>> +#define  SFC_VER_4			0x4
>> +#define  SFC_VER_5			0x5
>> +
>> +/* Master trigger */
>> +#define SFC_DMA_TRIGGER			0x80
>> +
>> +/* Src or Dst addr for master */
>> +#define SFC_DMA_ADDR			0x84
>> +
>> +/* Length control register extension 32GB */
>> +#define SFC_LEN_CTRL			0x88
>> +#define SFC_LEN_CTRL_TRB_SEL		1
>> +#define SFC_LEN_EXT			0x8C
>> +
>> +/* Command */
>> +#define SFC_CMD				0x100
>> +#define  SFC_CMD_IDX_SHIFT		0
>> +#define  SFC_CMD_DUMMY_SHIFT		8
>> +#define  SFC_CMD_DIR_SHIFT		12
>> +#define  SFC_CMD_DIR_RD			0
>> +#define  SFC_CMD_DIR_WR			1
>> +#define  SFC_CMD_ADDR_SHIFT		14
>> +#define  SFC_CMD_ADDR_0BITS		0
>> +#define  SFC_CMD_ADDR_24BITS		1
>> +#define  SFC_CMD_ADDR_32BITS		2
>> +#define  SFC_CMD_ADDR_XBITS		3
>> +#define  SFC_CMD_TRAN_BYTES_SHIFT	16
>> +#define  SFC_CMD_CS_SHIFT		30
>> +
>> +/* Address */
>> +#define SFC_ADDR			0x104
>> +
>> +/* Data */
>> +#define SFC_DATA			0x108
>> +
>> +/* The controller and documentation reports that it supports up to 4 CS
>> + * devices (0-3), however I have only been able to test a single CS (CS 0)
>> + * due to the configuration of my device.
>> + */
>> +#define SFC_MAX_CHIPSELECT_NUM		4
>> +
>> +/* The SFC can transfer max 16KB - 1 at one time
>> + * we set it to 15.5KB here for alignment.
>> + */
>> +#define SFC_MAX_IOSIZE_VER3		(512 * 31)
>> +
>> +#define SFC_MAX_IOSIZE_VER4		(0xFFFFFFFF)
>> +
>> +/* DMA is only enabled for large data transmission */
>> +#define SFC_DMA_TRANS_THRETHOLD		(0x40)
>> +
>> +/* Maximum clock values from datasheet suggest keeping clock value under
>> + * 150MHz. No minimum or average value is suggested, but the U-boot BSP driver
>> + * has a minimum of 10MHz and a default of 80MHz which seems reasonable.
>> + */
>> +#define SFC_MIN_SPEED_HZ		(10 * 1000 * 1000)
>> +#define SFC_DEFAULT_SPEED_HZ		(80 * 1000 * 1000)
>> +#define SFC_MAX_SPEED_HZ		(150 * 1000 * 1000)
>> +
>> +struct rockchip_sfc {
>> +	struct device *dev;
>> +	void __iomem *regbase;
>> +	struct clk *hclk;
>> +	struct clk *clk;
>> +	u32 frequency;
>> +	/* virtual mapped addr for dma_buffer */
>> +	void *buffer;
>> +	dma_addr_t dma_buffer;
>> +	struct completion cp;
>> +	bool use_dma;
>> +	u32 max_iosize;
>> +};
>> +
>> +static int rockchip_sfc_reset(struct rockchip_sfc *sfc)
>> +{
>> +	int err;
>> +	u32 status;
>> +
>> +	writel_relaxed(SFC_RCVR_RESET, sfc->regbase + SFC_RCVR);
>> +
>> +	err = readl_poll_timeout(sfc->regbase + SFC_RCVR, status,
>> +				 !(status & SFC_RCVR_RESET), 20,
>> +				 jiffies_to_usecs(HZ));
>> +	if (err)
>> +		dev_err(sfc->dev, "SFC reset never finished\n");
>> +
>> +	/* Still need to clear the masked interrupt from RISR */
>> +	writel_relaxed(0xFFFFFFFF, sfc->regbase + SFC_ICLR);
>> +
>> +	dev_dbg(sfc->dev, "reset\n");
>> +
>> +	return err;
>> +}
>> +
>> +static u16 rockchip_sfc_get_version(struct rockchip_sfc *sfc)
>> +{
>> +	return  (u16)(readl(sfc->regbase + SFC_VER) & 0xffff);
>> +}
>> +
>> +static u32 rockchip_sfc_get_max_iosize(struct rockchip_sfc *sfc)
>> +{
>> +	if (rockchip_sfc_get_version(sfc) >= SFC_VER_4)
>> +		return SFC_MAX_IOSIZE_VER4;
>> +	else
>> +		return SFC_MAX_IOSIZE_VER3;
>> +}
>> +
>> +static int rockchip_sfc_init(struct rockchip_sfc *sfc)
>> +{
>> +	writel(0, sfc->regbase + SFC_CTRL);
>> +	if (rockchip_sfc_get_version(sfc) >= SFC_VER_4)
>> +		writel(SFC_LEN_CTRL_TRB_SEL, sfc->regbase + SFC_LEN_CTRL);
>> +
>> +	return 0;
>> +}
>> +
>> +static inline int rockchip_sfc_get_fifo_level(struct rockchip_sfc *sfc, int wr)
>> +{
>> +	u32 fsr = readl_relaxed(sfc->regbase + SFC_FSR);
>> +	int level;
>> +
>> +	if (wr)
>> +		level = (fsr & SFC_FSR_TXLV_MASK) >> SFC_FSR_TXLV_SHIFT;
>> +	else
>> +		level = (fsr & SFC_FSR_RXLV_MASK) >> SFC_FSR_RXLV_SHIFT;
>> +
>> +	return level;
>> +}
>> +
>> +static int rockchip_sfc_wait_fifo_ready(struct rockchip_sfc *sfc, int wr, u32 timeout)
>> +{
>> +	unsigned long deadline = jiffies + timeout;
>> +	int level;
>> +
>> +	while (!(level = rockchip_sfc_get_fifo_level(sfc, wr))) {
>> +		if (time_after_eq(jiffies, deadline)) {
>> +			dev_warn(sfc->dev, "%s fifo timeout\n", wr ? "write" : "read");
>> +			return -ETIMEDOUT;
>> +		}
>> +		udelay(1);
>> +	}
>> +
>> +	return level;
>> +}
>> +
>> +static int rockchip_sfc_xfer_setup(struct rockchip_sfc *sfc,
>> +				   struct spi_mem *mem,
>> +				   const struct spi_mem_op *op,
>> +				   u32 len)
>> +{
>> +	u32 ctrl = 0, cmd = 0;
>> +
>> +	/* set CMD */
>> +	cmd = op->cmd.opcode;
>> +	ctrl |= ((op->cmd.buswidth >> 1) << SFC_CTRL_CMD_BITS_SHIFT);
>> +
>> +	/* set ADDR */
>> +	if (op->addr.nbytes) {
>> +		if (op->addr.nbytes == 4) {
>> +			cmd |= SFC_CMD_ADDR_32BITS << SFC_CMD_ADDR_SHIFT;
>> +		} else if (op->addr.nbytes == 3) {
>> +			cmd |= SFC_CMD_ADDR_24BITS << SFC_CMD_ADDR_SHIFT;
>> +		} else {
>> +			cmd |= SFC_CMD_ADDR_XBITS << SFC_CMD_ADDR_SHIFT;
>> +			writel_relaxed(op->addr.nbytes * 8 - 1, sfc->regbase + SFC_ABIT);
>> +		}
>> +
>> +		ctrl |= ((op->addr.buswidth >> 1) << SFC_CTRL_ADDR_BITS_SHIFT);
>> +	}
>> +
>> +	/* set DUMMY */
>> +	if (op->dummy.nbytes) {
>> +		if (op->dummy.buswidth == 4)
>> +			cmd |= op->dummy.nbytes * 2 << SFC_CMD_DUMMY_SHIFT;
>> +		else if (op->dummy.buswidth == 2)
>> +			cmd |= op->dummy.nbytes * 4 << SFC_CMD_DUMMY_SHIFT;
>> +		else
>> +			cmd |= op->dummy.nbytes * 8 << SFC_CMD_DUMMY_SHIFT;
>> +	}
>> +
> I have no experience optimizing code or anything, but would it help
> here to read the output of rockchip_sfc_get_version() to a variable
> in the driver data, and then check that each time we hit this?
> I don't know if there is any real-world difference in reading a
> variable versus reading a register, so I'm just asking. It's not like
> the sfc_version will change once the device is probed. :-)

done.

>> +	/* set DATA */
>> +	if (rockchip_sfc_get_version(sfc) >= SFC_VER_4) /* Clear it if no data to transfer */
>> +		writel(len, sfc->regbase + SFC_LEN_EXT);
>> +	else
>> +		cmd |= len << SFC_CMD_TRAN_BYTES_SHIFT;
>> +	if (len) {
>> +		if (op->data.dir == SPI_MEM_DATA_OUT)
>> +			cmd |= SFC_CMD_DIR_WR << SFC_CMD_DIR_SHIFT;
>> +
>> +		ctrl |= ((op->data.buswidth >> 1) << SFC_CTRL_DATA_BITS_SHIFT);
>> +	}
>> +	if (!len && op->addr.nbytes)
>> +		cmd |= SFC_CMD_DIR_WR << SFC_CMD_DIR_SHIFT;
>> +
>> +	/* set the Controller */
>> +	ctrl |= SFC_CTRL_PHASE_SEL_NEGETIVE;
>> +	cmd |= mem->spi->chip_select << SFC_CMD_CS_SHIFT;
>> +
>> +	dev_dbg(sfc->dev, "addr.nbytes=%x(x%d) dummy.nbytes=%x(x%d)\n",
>> +		op->addr.nbytes, op->addr.buswidth,
>> +		op->dummy.nbytes, op->dummy.buswidth);
>> +	dev_dbg(sfc->dev, "ctrl=%x cmd=%x addr=%llx len=%x\n",
>> +		ctrl, cmd, op->addr.val, len);
>> +
>> +	writel(ctrl, sfc->regbase + SFC_CTRL);
>> +	writel(cmd, sfc->regbase + SFC_CMD);
>> +	if (op->addr.nbytes)
>> +		writel(op->addr.val, sfc->regbase + SFC_ADDR);
>> +
>> +	return 0;
>> +}
>> +
>> +static int rockchip_sfc_write_fifo(struct rockchip_sfc *sfc, const u8 *buf, int len)
>> +{
>> +	u8 bytes = len & 0x3;
>> +	u32 dwords;
>> +	int tx_level;
>> +	u32 write_words;
>> +	u32 tmp = 0;
>> +
>> +	dwords = len >> 2;
>> +	while (dwords) {
>> +		tx_level = rockchip_sfc_wait_fifo_ready(sfc, SFC_CMD_DIR_WR, HZ);
>> +		if (tx_level < 0)
>> +			return tx_level;
>> +		write_words = min_t(u32, tx_level, dwords);
>> +		iowrite32_rep(sfc->regbase + SFC_DATA, buf, write_words);
>> +		buf += write_words << 2;
>> +		dwords -= write_words;
>> +	}
>> +
>> +	/* write the rest non word aligned bytes */
>> +	if (bytes) {
>> +		tx_level = rockchip_sfc_wait_fifo_ready(sfc, SFC_CMD_DIR_WR, HZ);
>> +		if (tx_level < 0)
>> +			return tx_level;
>> +		memcpy(&tmp, buf, bytes);
>> +		writel(tmp, sfc->regbase + SFC_DATA);
>> +	}
>> +
>> +	return len;
>> +}
>> +
>> +static int rockchip_sfc_read_fifo(struct rockchip_sfc *sfc, u8 *buf, int len)
>> +{
>> +	u8 bytes = len & 0x3;
>> +	u32 dwords;
>> +	u8 read_words;
>> +	int rx_level;
>> +	int tmp;
>> +
>> +	/* word aligned access only */
>> +	dwords = len >> 2;
>> +	while (dwords) {
>> +		rx_level = rockchip_sfc_wait_fifo_ready(sfc, SFC_CMD_DIR_RD, HZ);
>> +		if (rx_level < 0)
>> +			return rx_level;
>> +		read_words = min_t(u32, rx_level, dwords);
>> +		ioread32_rep(sfc->regbase + SFC_DATA, buf, read_words);
>> +		buf += read_words << 2;
>> +		dwords -= read_words;
>> +	}
>> +
>> +	/* read the rest non word aligned bytes */
>> +	if (bytes) {
>> +		rx_level = rockchip_sfc_wait_fifo_ready(sfc, SFC_CMD_DIR_RD, HZ);
>> +		if (rx_level < 0)
>> +			return rx_level;
>> +		tmp = readl_relaxed(sfc->regbase + SFC_DATA);
>> +		memcpy(buf, &tmp, bytes);
>> +	}
>> +
>> +	return len;
>> +}
>> +
>> +static int rockchip_sfc_fifo_transfer_dma(struct rockchip_sfc *sfc, dma_addr_t dma_buf, size_t len)
>> +{
>> +	u32 reg;
>> +	int err = 0;
>> +
>> +	init_completion(&sfc->cp);
>> +
>> +	writel(0xFFFFFFFF, sfc->regbase + SFC_ICLR);
>> +
>> +	/* Enable transfer complete interrupt */
>> +	reg = readl(sfc->regbase + SFC_IMR);
>> +	reg &= ~SFC_IMR_DMA;
>> +	writel(reg, sfc->regbase + SFC_IMR);
>> +	writel((u32)dma_buf, sfc->regbase + SFC_DMA_ADDR);
>> +	writel(0x1, sfc->regbase + SFC_DMA_TRIGGER);
>> +
>> +	/* Wait for the interrupt. */
>> +	if (!wait_for_completion_timeout(&sfc->cp, msecs_to_jiffies(2000))) {
>> +		dev_err(sfc->dev, "DMA wait for transfer finish timeout\n");
>> +		err = -ETIMEDOUT;
>> +	}
>> +
>> +	writel(0xFFFFFFFF, sfc->regbase + SFC_ICLR);
>> +	/* Disable transfer finish interrupt */
>> +	reg = readl(sfc->regbase + SFC_IMR);
>> +	reg |= SFC_IMR_DMA;
>> +	writel(reg, sfc->regbase + SFC_IMR);
>> +
>> +	return len;
>> +}
>> +
>> +static int rockchip_sfc_xfer_data_poll(struct rockchip_sfc *sfc,
>> +				       const struct spi_mem_op *op, u32 len)
>> +{
>> +	dev_dbg(sfc->dev, "xfer_poll len=%x\n", len);
>> +
>> +	if (op->data.dir == SPI_MEM_DATA_OUT)
>> +		return rockchip_sfc_write_fifo(sfc, op->data.buf.out, len);
>> +	else
>> +		return rockchip_sfc_read_fifo(sfc, op->data.buf.in, len);
>> +}
>> +
>> +static int rockchip_sfc_xfer_data_dma(struct rockchip_sfc *sfc,
>> +				      const struct spi_mem_op *op, u32 len)
>> +{
>> +	int ret;
>> +
>> +	dev_dbg(sfc->dev, "xfer_dma len=%x\n", len);
>> +
>> +	if (op->data.dir == SPI_MEM_DATA_OUT) {
>> +		memcpy_toio(sfc->buffer, op->data.buf.out, len);
>> +		ret = rockchip_sfc_fifo_transfer_dma(sfc, sfc->dma_buffer, len);
>> +	} else {
>> +		ret = rockchip_sfc_fifo_transfer_dma(sfc, sfc->dma_buffer, len);
>> +		memcpy_fromio(op->data.buf.in, sfc->buffer, len);
>> +	}
>> +
>> +	return ret;
>> +}
>> +
>> +static int rockchip_sfc_xfer_done(struct rockchip_sfc *sfc, u32 timeout_us)
>> +{
>> +	int ret = 0;
>> +	u32 status;
>> +
>> +	ret = readl_poll_timeout(sfc->regbase + SFC_SR, status,
>> +				 !(status & SFC_SR_IS_BUSY),
>> +				 20, timeout_us);
>> +	if (ret) {
>> +		dev_err(sfc->dev, "wait sfc idle timeout\n");
>> +		rockchip_sfc_reset(sfc);
>> +
>> +		ret = -EIO;
>> +	}
>> +
>> +	return ret;
>> +}
>> +
>> +static int rockchip_sfc_exec_mem_op(struct spi_mem *mem, const struct spi_mem_op *op)
>> +{
>> +	struct rockchip_sfc *sfc = spi_master_get_devdata(mem->spi->master);
>> +	u32 len = min_t(u32, op->data.nbytes, sfc->max_iosize);
> Is it necessary to use the minimum of these 2 values if we use an
> adjust_op_size which I propose below?
it necessary if use adjust_op_size, done.
>> +	int ret;
>> +
> Again I'm asking this as someone with no experience optimizing code
> or with compilers, but would it help to add an unlikely() here? It's
> all but guaranteed the first time this code path is called the clock
> will need to be set, but each subsequent call should not really have
> to care since the clock likely won't be changing (as long as we're on
> the same CS). We can also then set the sfc->frequency to what the
> max_speed_hz was after successfully setting the rate so we can hit the
> faster path after the first run (it looks like it's checking the
> uninitalized value but never updating it after it's changed, so it's
> always doing the clk_set_rate()).
>
> Also, should we clamp the clock speed between the min and max values?
done.
>> +	if (mem->spi->max_speed_hz != sfc->frequency) {
>> +		if (clk_set_rate(sfc->clk, mem->spi->max_speed_hz))
>> +			return ret;
>> +		dev_dbg(sfc->dev, "set_freq=%dHz real_freq=%ldHz\n",
>> +			sfc->frequency, clk_get_rate(sfc->clk));
>> +	}
>> +
>> +	rockchip_sfc_xfer_setup(sfc, mem, op, len);
>> +	if (len) {
>> +		if (likely(sfc->use_dma) && !(len & 0x3) && len >= SFC_DMA_TRANS_THRETHOLD)
>> +			ret = rockchip_sfc_xfer_data_dma(sfc, op, len);
>> +		else
>> +			ret = rockchip_sfc_xfer_data_poll(sfc, op, len);
>> +
>> +		if (ret != len) {
>> +			dev_err(sfc->dev, "xfer data failed ret %d dir %d\n", ret, op->data.dir);
>> +
>> +			return -EIO;
>> +		}
>> +	}
>> +
>> +	return rockchip_sfc_xfer_done(sfc, 100000);
>> +}
>> +
> I'll be honest with you, I had no idea if I needed this function or
> not (the rockchip_sfc_get_name() function). I'm pretty sure it's not
> needed though, so I assume it can be safely removed.
>
>> +static const char *rockchip_sfc_get_name(struct spi_mem *mem)
>> +{
>> +	return devm_kasprintf(&mem->spi->dev, GFP_KERNEL, "%s.%d",
>> +			      dev_name(&mem->spi->dev), mem->spi->chip_select);
>> +}
>> +
> I had an issue with doing dd as mentioned above reading against mtd0,
> and was able to fix it by adding a mem_op of adjust_op_size like this.
> Not sure if it works for the v4/v5 hardware though.
>
> static int rockchip_sfc_adjust_op_size(struct spi_mem *mem, struct spi_mem_op *op)
> {
> 	struct rockchip_sfc *sfc = spi_master_get_devdata(mem->spi->master);
>
> 	op->data.nbytes = min(op->data.nbytes, sfc->max_iosize);
> 	return 0;
> }

good idea.

>> +static const struct spi_controller_mem_ops rockchip_sfc_mem_ops = {
>> +	.exec_op = rockchip_sfc_exec_mem_op,
>> +	.get_name = rockchip_sfc_get_name,
>> +};
>> +
>> +static irqreturn_t rockchip_sfc_irq_handler(int irq, void *dev_id)
>> +{
>> +	struct rockchip_sfc *sfc = dev_id;
>> +	u32 reg;
>> +
>> +	reg = readl(sfc->regbase + SFC_RISR);
>> +
>> +	/* Clear interrupt */
>> +	writel_relaxed(reg, sfc->regbase + SFC_ICLR);
>> +
>> +	if (reg & SFC_RISR_DMA)
>> +		complete(&sfc->cp);
>> +
> >From a previous comment, we should only clear the interrupt if we
> handle the specific one in question.
>
Compare with drivers/spi/spi-cadence-quadspi.c, I think the former code 
work well. Can explain more for the reason.

>> +	return IRQ_HANDLED;
>> +}
>> +
>> +static int rockchip_sfc_probe(struct platform_device *pdev)
>> +{
>> +	struct device *dev = &pdev->dev;
>> +	struct spi_master *master;
>> +	struct resource *res;
>> +	struct rockchip_sfc *sfc;
>> +	int ret;
>> +
>> +	master = devm_spi_alloc_master(&pdev->dev, sizeof(*sfc));
>> +	if (!master)
>> +		return -ENOMEM;
>> +
>> +	master->mem_ops = &rockchip_sfc_mem_ops;
>> +	master->dev.of_node = pdev->dev.of_node;
>> +	master->mode_bits = SPI_TX_QUAD | SPI_TX_DUAL | SPI_RX_QUAD | SPI_RX_DUAL;
>> +	master->min_speed_hz = SFC_MIN_SPEED_HZ;
>> +	master->max_speed_hz = SFC_MAX_SPEED_HZ;
>> +	master->num_chipselect = SFC_MAX_CHIPSELECT_NUM;
>> +
>> +	sfc = spi_master_get_devdata(master);
>> +	sfc->dev = dev;
>> +
>> +	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>> +	sfc->regbase = devm_ioremap_resource(dev, res);
>> +	if (IS_ERR(sfc->regbase))
>> +		return PTR_ERR(sfc->regbase);
>> +
>> +	sfc->clk = devm_clk_get(&pdev->dev, "clk_sfc");
>> +	if (IS_ERR(sfc->clk)) {
>> +		dev_err(&pdev->dev, "Failed to get sfc interface clk\n");
>> +		return PTR_ERR(sfc->clk);
>> +	}
>> +
>> +	sfc->hclk = devm_clk_get(&pdev->dev, "hclk_sfc");
>> +	if (IS_ERR(sfc->hclk)) {
>> +		dev_err(&pdev->dev, "Failed to get sfc ahb clk\n");
>> +		return PTR_ERR(sfc->hclk);
>> +	}
>> +
>> +	sfc->use_dma = !of_property_read_bool(sfc->dev->of_node,
>> +					      "rockchip,sfc-no-dma");
>> +
>> +	if (sfc->use_dma) {
>> +		ret = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
>> +		if (ret) {
>> +			dev_warn(dev, "Unable to set dma mask\n");
>> +			return ret;
>> +		}
>> +
>> +		sfc->buffer = dmam_alloc_coherent(dev, SFC_MAX_IOSIZE_VER3,
>> +						  &sfc->dma_buffer,
>> +						  GFP_KERNEL);
>> +		if (!sfc->buffer)
>> +			return -ENOMEM;
>> +	}
>> +
>> +	ret = clk_prepare_enable(sfc->hclk);
>> +	if (ret) {
>> +		dev_err(&pdev->dev, "Failed to enable ahb clk\n");
>> +		goto err_hclk;
>> +	}
>> +
>> +	ret = clk_prepare_enable(sfc->clk);
>> +	if (ret) {
>> +		dev_err(&pdev->dev, "Failed to enable interface clk\n");
>> +		goto err_clk;
>> +	}
>> +
>> +	/* Find the irq */
>> +	ret = platform_get_irq(pdev, 0);
>> +	if (ret < 0) {
>> +		dev_err(dev, "Failed to get the irq\n");
>> +		goto err_irq;
>> +	}
>> +
>> +	ret = devm_request_irq(dev, ret, rockchip_sfc_irq_handler,
>> +			       0, pdev->name, sfc);
>> +	if (ret) {
>> +		dev_err(dev, "Failed to request irq\n");
>> +
>> +		return ret;
>> +	}
>> +
>> +	ret = rockchip_sfc_init(sfc);
>> +	if (ret)
>> +		goto err_irq;
>> +
>> +	sfc->max_iosize = rockchip_sfc_get_max_iosize(sfc);
>> +
>> +	ret = spi_register_master(master);
>> +	if (ret)
>> +		goto err_irq;
>> +
>> +	return 0;
>> +
>> +err_irq:
>> +	clk_disable_unprepare(sfc->clk);
>> +err_clk:
>> +	clk_disable_unprepare(sfc->hclk);
>> +err_hclk:
>> +	return ret;
>> +}
>> +
>> +static int rockchip_sfc_remove(struct platform_device *pdev)
>> +{
>> +	struct spi_master *master = platform_get_drvdata(pdev);
>> +	struct rockchip_sfc *sfc = platform_get_drvdata(pdev);
>> +
>> +	spi_unregister_master(master);
>> +
>> +	clk_disable_unprepare(sfc->clk);
>> +	clk_disable_unprepare(sfc->hclk);
>> +
>> +	return 0;
>> +}
>> +
>> +static const struct of_device_id rockchip_sfc_dt_ids[] = {
>> +	{ .compatible = "rockchip,px30-sfc"},
>> +	{ .compatible = "rockchip,rk3036-sfc"},
>> +	{ .compatible = "rockchip,rk3308-sfc"},
>> +	{ .compatible = "rockchip,rv1126-sfc"},
>> +	{ /* sentinel */ }
>> +};
>> +MODULE_DEVICE_TABLE(of, rockchip_sfc_dt_ids);
>> +
>> +static struct platform_driver rockchip_sfc_driver = {
>> +	.driver = {
>> +		.name	= "rockchip-sfc",
>> +		.of_match_table = rockchip_sfc_dt_ids,
>> +	},
>> +	.probe	= rockchip_sfc_probe,
>> +	.remove	= rockchip_sfc_remove,
>> +};
>> +module_platform_driver(rockchip_sfc_driver);
>> +
>> +MODULE_LICENSE("GPL v2");
>> +MODULE_DESCRIPTION("Rockchip Serial Flash Controller Driver");
>> +MODULE_AUTHOR("Shawn Lin <shawn.lin@rock-chips.com>");
>> +MODULE_AUTHOR("Chris Morgan <macroalpha82@gmail.com>");
> I know folks will hate on me for this (and for good reason given it
> messes up IDs and stuff), but I prefer macromorgan@hotmail.com if my
> email is in here, as that's the one I've been using for decades...
ok.
>> +MODULE_AUTHOR("Jon Lin <Jon.lin@rock-chips.com>");
>> -- 
>> 2.17.1
>>
>>
>>
>
>
Chris Morgan June 9, 2021, 4:57 p.m. UTC | #7
On Wed, Jun 09, 2021 at 09:48:50PM +0800, Jon Lin wrote:
> 
> On 6/9/21 10:36 AM, Chris Morgan wrote:
> > On Tue, Jun 08, 2021 at 10:26:38AM +0800, Jon Lin wrote:
> > > From: Chris Morgan <macromorgan@hotmail.com>
> > > 
> > > Add the rockchip serial flash controller (SFC) driver.
> > > 
> > > Signed-off-by: Chris Morgan <macromorgan@hotmail.com>
> > > Signed-off-by: Jon Lin <jon.lin@rock-chips.com>
> > I think I hit "reply" earlier when I meant to hit "reply-all". Whoops.
> > 
> > I wanted to say that for now this driver is not working for me on the
> > v3 hardware. There appears to be something wrong with alternating
> > between page programs and erase. I can consistently reproduce this by
> > writing a 4MB image of random data to my SPI flash chip at offset
> > c00000 using dd to mtdblock0. I've dumped the contents of what is
> > written to the controller and it appears at some point it goes from
> > doing a page program (02) to a dual io read (bb) to a sector erase
> > (20), then it repeatedly reads the status register (05) and I assume
> > the register never changes. As a result it just keeps looping over
> > the status register then eventually times out. I'm still debugging
> > to try and figure out exactly what is going on though.
> > 
> > I also noticed that reading does not work consistently using dd
> > from the mtd0 device, however I have a proposed fix for that which
> > I list below at the appropriate section.
> > 
> Can you enable dev_dbg in rockchip_sfc_xfer_setup, then send me log.

I can confirm that on the v7 patch I still get some "weirdness" when
I use mtdblock, but when I use flashrom things work entirely as
expected. I will still send you a log, but at this point I wonder if
it's related to an mtdblock issue rather than an issue with this
driver. In retrospect I don't recall testing things this extensively
earlier.

You can reproduce the issue by doing the following:

Create a random 4MB file (dd if=/dev/urandom of=rand.img bs=4M count=1)

Write the 4MB random file to the device using MTD Block drivers
(dd if=rand.img of=/dev/mtdblock0 bs=4096 seek=3072)

Read back the data (dd if=/dev/mtd0 of=rand1.img bs=4096 skip=3072)

Compare checksums (md5sum *.img)
When I do this I get different checksums for the random files. However
when I use flashrom to write instead of MTD Block, it works.

Read the ROM - I can confirm after extensive testing reads seem to work
just fine (dd if=/dev/mtd0 of=rom.img bs=8192)

Copy the random data to the image file
(dd if=rand.img of=rom.img bs=4096 seek=3072)

Flash the file with flashrom (flashrom -p linux_mtd -w rom.img)

Read back the random data
(dd if=/dev/mtd0 of=rand1.img bs=4096 skip=3072)

Compare checksums (md5sum *.img)
At this point the rand.img and rand1.img files have identical
checksums.


As for testing the reading bugs I found on v6 and below, it was a
matter of reading with dd using a block size bigger than the max_iosize
value, but with the addition of the adjust_op_size that seems to be
resolved now completely.

Test case:
Read flash image with varying block sizes
(dd if=/dev/mtd0 of=test1.img bs=4096)
(dd if=/dev/mtd0 of=test2.img bs=8192)
(dd if=/dev/mtd0 of=test3.img bs=16384)

Compare checksums (md5sum *.img)

Prior to v7 the checksum for test3.img would be different than the
other two. v7 now returns the same checksum for all 3 files.

> 
> I successfully mount and test spinor jffs2 to confirm read/write/erase work
> work. I'll try more cases further.
> 
> > > ---
> > > 
> > > Changes in v6: None
> > > Changes in v5: None
> > > Changes in v4: None
> > > Changes in v3: None
> > > Changes in v2: None
> > > Changes in v1: None
> > > 
> > >   drivers/spi/Kconfig            |   9 +
> > >   drivers/spi/Makefile           |   1 +
> > >   drivers/spi/spi-rockchip-sfc.c | 660 +++++++++++++++++++++++++++++++++
> > >   3 files changed, 670 insertions(+)
> > >   create mode 100644 drivers/spi/spi-rockchip-sfc.c
> > > 
> > > diff --git a/drivers/spi/Kconfig b/drivers/spi/Kconfig
> > > index e71a4c514f7b..d89e5f3c9107 100644
> > > --- a/drivers/spi/Kconfig
> > > +++ b/drivers/spi/Kconfig
> > > @@ -658,6 +658,15 @@ config SPI_ROCKCHIP
> > >   	  The main usecase of this controller is to use spi flash as boot
> > >   	  device.
> > > +config SPI_ROCKCHIP_SFC
> > > +	tristate "Rockchip Serial Flash Controller (SFC)"
> > > +	depends on ARCH_ROCKCHIP || COMPILE_TEST
> > > +	depends on HAS_IOMEM && HAS_DMA
> > > +	help
> > > +	  This enables support for Rockchip serial flash controller. This
> > > +	  is a specialized controller used to access SPI flash on some
> > > +	  Rockchip SOCs.
> > > +
> > >   config SPI_RB4XX
> > >   	tristate "Mikrotik RB4XX SPI master"
> > >   	depends on SPI_MASTER && ATH79
> > > diff --git a/drivers/spi/Makefile b/drivers/spi/Makefile
> > > index 13e54c45e9df..699db95c8441 100644
> > > --- a/drivers/spi/Makefile
> > > +++ b/drivers/spi/Makefile
> > > @@ -95,6 +95,7 @@ obj-$(CONFIG_SPI_QCOM_GENI)		+= spi-geni-qcom.o
> > >   obj-$(CONFIG_SPI_QCOM_QSPI)		+= spi-qcom-qspi.o
> > >   obj-$(CONFIG_SPI_QUP)			+= spi-qup.o
> > >   obj-$(CONFIG_SPI_ROCKCHIP)		+= spi-rockchip.o
> > > +obj-$(CONFIG_SPI_ROCKCHIP_SFC)		+= spi-rockchip-sfc.o
> > >   obj-$(CONFIG_SPI_RB4XX)			+= spi-rb4xx.o
> > >   obj-$(CONFIG_MACH_REALTEK_RTL)		+= spi-realtek-rtl.o
> > >   obj-$(CONFIG_SPI_RPCIF)			+= spi-rpc-if.o
> > > diff --git a/drivers/spi/spi-rockchip-sfc.c b/drivers/spi/spi-rockchip-sfc.c
> > > new file mode 100644
> > > index 000000000000..ec25ad278096
> > > --- /dev/null
> > > +++ b/drivers/spi/spi-rockchip-sfc.c
> > > @@ -0,0 +1,660 @@
> > > +// SPDX-License-Identifier: GPL-2.0-only
> > > +/*
> > > + * Rockchip Serial Flash Controller Driver
> > > + *
> > > + * Copyright (c) 2017-2021, Rockchip Inc.
> > > + * Author: Shawn Lin <shawn.lin@rock-chips.com>
> > > + *	   Chris Morgan <macroalpha82@gmail.com>
> > > + *	   Jon Lin <Jon.lin@rock-chips.com>
> > > + */
> > > +
> > > +#include <linux/bitops.h>
> > > +#include <linux/clk.h>
> > > +#include <linux/completion.h>
> > > +#include <linux/dma-mapping.h>
> > > +#include <linux/iopoll.h>
> > > +#include <linux/mm.h>
> > > +#include <linux/module.h>
> > > +#include <linux/of.h>
> > > +#include <linux/platform_device.h>
> > > +#include <linux/slab.h>
> > > +#include <linux/interrupt.h>
> > > +#include <linux/spi/spi-mem.h>
> > > +
> > > +/* System control */
> > > +#define SFC_CTRL			0x0
> > > +#define  SFC_CTRL_PHASE_SEL_NEGETIVE	BIT(1)
> > > +#define  SFC_CTRL_CMD_BITS_SHIFT	8
> > > +#define  SFC_CTRL_ADDR_BITS_SHIFT	10
> > > +#define  SFC_CTRL_DATA_BITS_SHIFT	12
> > > +
> > > +/* Interrupt mask */
> > > +#define SFC_IMR				0x4
> > > +#define  SFC_IMR_RX_FULL		BIT(0)
> > > +#define  SFC_IMR_RX_UFLOW		BIT(1)
> > > +#define  SFC_IMR_TX_OFLOW		BIT(2)
> > > +#define  SFC_IMR_TX_EMPTY		BIT(3)
> > > +#define  SFC_IMR_TRAN_FINISH		BIT(4)
> > > +#define  SFC_IMR_BUS_ERR		BIT(5)
> > > +#define  SFC_IMR_NSPI_ERR		BIT(6)
> > > +#define  SFC_IMR_DMA			BIT(7)
> > > +
> > > +/* Interrupt clear */
> > > +#define SFC_ICLR			0x8
> > > +#define  SFC_ICLR_RX_FULL		BIT(0)
> > > +#define  SFC_ICLR_RX_UFLOW		BIT(1)
> > > +#define  SFC_ICLR_TX_OFLOW		BIT(2)
> > > +#define  SFC_ICLR_TX_EMPTY		BIT(3)
> > > +#define  SFC_ICLR_TRAN_FINISH		BIT(4)
> > > +#define  SFC_ICLR_BUS_ERR		BIT(5)
> > > +#define  SFC_ICLR_NSPI_ERR		BIT(6)
> > > +#define  SFC_ICLR_DMA			BIT(7)
> > > +
> > > +/* FIFO threshold level */
> > > +#define SFC_FTLR			0xc
> > > +#define  SFC_FTLR_TX_SHIFT		0
> > > +#define  SFC_FTLR_TX_MASK		0x1f
> > > +#define  SFC_FTLR_RX_SHIFT		8
> > > +#define  SFC_FTLR_RX_MASK		0x1f
> > > +
> > > +/* Reset FSM and FIFO */
> > > +#define SFC_RCVR			0x10
> > > +#define  SFC_RCVR_RESET			BIT(0)
> > > +
> > > +/* Enhanced mode */
> > > +#define SFC_AX				0x14
> > > +
> > > +/* Address Bit number */
> > > +#define SFC_ABIT			0x18
> > > +
> > > +/* Interrupt status */
> > > +#define SFC_ISR				0x1c
> > > +#define  SFC_ISR_RX_FULL_SHIFT		BIT(0)
> > > +#define  SFC_ISR_RX_UFLOW_SHIFT		BIT(1)
> > > +#define  SFC_ISR_TX_OFLOW_SHIFT		BIT(2)
> > > +#define  SFC_ISR_TX_EMPTY_SHIFT		BIT(3)
> > > +#define  SFC_ISR_TX_FINISH_SHIFT	BIT(4)
> > > +#define  SFC_ISR_BUS_ERR_SHIFT		BIT(5)
> > > +#define  SFC_ISR_NSPI_ERR_SHIFT		BIT(6)
> > > +#define  SFC_ISR_DMA_SHIFT		BIT(7)
> > > +
> > > +/* FIFO status */
> > > +#define SFC_FSR				0x20
> > > +#define  SFC_FSR_TX_IS_FULL		BIT(0)
> > > +#define  SFC_FSR_TX_IS_EMPTY		BIT(1)
> > > +#define  SFC_FSR_RX_IS_EMPTY		BIT(2)
> > > +#define  SFC_FSR_RX_IS_FULL		BIT(3)
> > > +#define  SFC_FSR_TXLV_MASK		GENMASK(12, 8)
> > > +#define  SFC_FSR_TXLV_SHIFT		8
> > > +#define  SFC_FSR_RXLV_MASK		GENMASK(20, 16)
> > > +#define  SFC_FSR_RXLV_SHIFT		16
> > > +
> > > +/* FSM status */
> > > +#define SFC_SR				0x24
> > > +#define  SFC_SR_IS_IDLE			0x0
> > > +#define  SFC_SR_IS_BUSY			0x1
> > > +
> > > +/* Raw interrupt status */
> > > +#define SFC_RISR			0x28
> > > +#define  SFC_RISR_RX_FULL		BIT(0)
> > > +#define  SFC_RISR_RX_UNDERFLOW		BIT(1)
> > > +#define  SFC_RISR_TX_OVERFLOW		BIT(2)
> > > +#define  SFC_RISR_TX_EMPTY		BIT(3)
> > > +#define  SFC_RISR_TRAN_FINISH		BIT(4)
> > > +#define  SFC_RISR_BUS_ERR		BIT(5)
> > > +#define  SFC_RISR_NSPI_ERR		BIT(6)
> > > +#define  SFC_RISR_DMA			BIT(7)
> > > +
> > > +/* Version */
> > > +#define SFC_VER				0x2C
> > > +#define  SFC_VER_3			0x3
> > > +#define  SFC_VER_4			0x4
> > > +#define  SFC_VER_5			0x5
> > > +
> > > +/* Master trigger */
> > > +#define SFC_DMA_TRIGGER			0x80
> > > +
> > > +/* Src or Dst addr for master */
> > > +#define SFC_DMA_ADDR			0x84
> > > +
> > > +/* Length control register extension 32GB */
> > > +#define SFC_LEN_CTRL			0x88
> > > +#define SFC_LEN_CTRL_TRB_SEL		1
> > > +#define SFC_LEN_EXT			0x8C
> > > +
> > > +/* Command */
> > > +#define SFC_CMD				0x100
> > > +#define  SFC_CMD_IDX_SHIFT		0
> > > +#define  SFC_CMD_DUMMY_SHIFT		8
> > > +#define  SFC_CMD_DIR_SHIFT		12
> > > +#define  SFC_CMD_DIR_RD			0
> > > +#define  SFC_CMD_DIR_WR			1
> > > +#define  SFC_CMD_ADDR_SHIFT		14
> > > +#define  SFC_CMD_ADDR_0BITS		0
> > > +#define  SFC_CMD_ADDR_24BITS		1
> > > +#define  SFC_CMD_ADDR_32BITS		2
> > > +#define  SFC_CMD_ADDR_XBITS		3
> > > +#define  SFC_CMD_TRAN_BYTES_SHIFT	16
> > > +#define  SFC_CMD_CS_SHIFT		30
> > > +
> > > +/* Address */
> > > +#define SFC_ADDR			0x104
> > > +
> > > +/* Data */
> > > +#define SFC_DATA			0x108
> > > +
> > > +/* The controller and documentation reports that it supports up to 4 CS
> > > + * devices (0-3), however I have only been able to test a single CS (CS 0)
> > > + * due to the configuration of my device.
> > > + */
> > > +#define SFC_MAX_CHIPSELECT_NUM		4
> > > +
> > > +/* The SFC can transfer max 16KB - 1 at one time
> > > + * we set it to 15.5KB here for alignment.
> > > + */
> > > +#define SFC_MAX_IOSIZE_VER3		(512 * 31)
> > > +
> > > +#define SFC_MAX_IOSIZE_VER4		(0xFFFFFFFF)
> > > +
> > > +/* DMA is only enabled for large data transmission */
> > > +#define SFC_DMA_TRANS_THRETHOLD		(0x40)
> > > +
> > > +/* Maximum clock values from datasheet suggest keeping clock value under
> > > + * 150MHz. No minimum or average value is suggested, but the U-boot BSP driver
> > > + * has a minimum of 10MHz and a default of 80MHz which seems reasonable.
> > > + */
> > > +#define SFC_MIN_SPEED_HZ		(10 * 1000 * 1000)
> > > +#define SFC_DEFAULT_SPEED_HZ		(80 * 1000 * 1000)
> > > +#define SFC_MAX_SPEED_HZ		(150 * 1000 * 1000)
> > > +
> > > +struct rockchip_sfc {
> > > +	struct device *dev;
> > > +	void __iomem *regbase;
> > > +	struct clk *hclk;
> > > +	struct clk *clk;
> > > +	u32 frequency;
> > > +	/* virtual mapped addr for dma_buffer */
> > > +	void *buffer;
> > > +	dma_addr_t dma_buffer;
> > > +	struct completion cp;
> > > +	bool use_dma;
> > > +	u32 max_iosize;
> > > +};
> > > +
> > > +static int rockchip_sfc_reset(struct rockchip_sfc *sfc)
> > > +{
> > > +	int err;
> > > +	u32 status;
> > > +
> > > +	writel_relaxed(SFC_RCVR_RESET, sfc->regbase + SFC_RCVR);
> > > +
> > > +	err = readl_poll_timeout(sfc->regbase + SFC_RCVR, status,
> > > +				 !(status & SFC_RCVR_RESET), 20,
> > > +				 jiffies_to_usecs(HZ));
> > > +	if (err)
> > > +		dev_err(sfc->dev, "SFC reset never finished\n");
> > > +
> > > +	/* Still need to clear the masked interrupt from RISR */
> > > +	writel_relaxed(0xFFFFFFFF, sfc->regbase + SFC_ICLR);
> > > +
> > > +	dev_dbg(sfc->dev, "reset\n");
> > > +
> > > +	return err;
> > > +}
> > > +
> > > +static u16 rockchip_sfc_get_version(struct rockchip_sfc *sfc)
> > > +{
> > > +	return  (u16)(readl(sfc->regbase + SFC_VER) & 0xffff);
> > > +}
> > > +
> > > +static u32 rockchip_sfc_get_max_iosize(struct rockchip_sfc *sfc)
> > > +{
> > > +	if (rockchip_sfc_get_version(sfc) >= SFC_VER_4)
> > > +		return SFC_MAX_IOSIZE_VER4;
> > > +	else
> > > +		return SFC_MAX_IOSIZE_VER3;
> > > +}
> > > +
> > > +static int rockchip_sfc_init(struct rockchip_sfc *sfc)
> > > +{
> > > +	writel(0, sfc->regbase + SFC_CTRL);
> > > +	if (rockchip_sfc_get_version(sfc) >= SFC_VER_4)
> > > +		writel(SFC_LEN_CTRL_TRB_SEL, sfc->regbase + SFC_LEN_CTRL);
> > > +
> > > +	return 0;
> > > +}
> > > +
> > > +static inline int rockchip_sfc_get_fifo_level(struct rockchip_sfc *sfc, int wr)
> > > +{
> > > +	u32 fsr = readl_relaxed(sfc->regbase + SFC_FSR);
> > > +	int level;
> > > +
> > > +	if (wr)
> > > +		level = (fsr & SFC_FSR_TXLV_MASK) >> SFC_FSR_TXLV_SHIFT;
> > > +	else
> > > +		level = (fsr & SFC_FSR_RXLV_MASK) >> SFC_FSR_RXLV_SHIFT;
> > > +
> > > +	return level;
> > > +}
> > > +
> > > +static int rockchip_sfc_wait_fifo_ready(struct rockchip_sfc *sfc, int wr, u32 timeout)
> > > +{
> > > +	unsigned long deadline = jiffies + timeout;
> > > +	int level;
> > > +
> > > +	while (!(level = rockchip_sfc_get_fifo_level(sfc, wr))) {
> > > +		if (time_after_eq(jiffies, deadline)) {
> > > +			dev_warn(sfc->dev, "%s fifo timeout\n", wr ? "write" : "read");
> > > +			return -ETIMEDOUT;
> > > +		}
> > > +		udelay(1);
> > > +	}
> > > +
> > > +	return level;
> > > +}
> > > +
> > > +static int rockchip_sfc_xfer_setup(struct rockchip_sfc *sfc,
> > > +				   struct spi_mem *mem,
> > > +				   const struct spi_mem_op *op,
> > > +				   u32 len)
> > > +{
> > > +	u32 ctrl = 0, cmd = 0;
> > > +
> > > +	/* set CMD */
> > > +	cmd = op->cmd.opcode;
> > > +	ctrl |= ((op->cmd.buswidth >> 1) << SFC_CTRL_CMD_BITS_SHIFT);
> > > +
> > > +	/* set ADDR */
> > > +	if (op->addr.nbytes) {
> > > +		if (op->addr.nbytes == 4) {
> > > +			cmd |= SFC_CMD_ADDR_32BITS << SFC_CMD_ADDR_SHIFT;
> > > +		} else if (op->addr.nbytes == 3) {
> > > +			cmd |= SFC_CMD_ADDR_24BITS << SFC_CMD_ADDR_SHIFT;
> > > +		} else {
> > > +			cmd |= SFC_CMD_ADDR_XBITS << SFC_CMD_ADDR_SHIFT;
> > > +			writel_relaxed(op->addr.nbytes * 8 - 1, sfc->regbase + SFC_ABIT);
> > > +		}
> > > +
> > > +		ctrl |= ((op->addr.buswidth >> 1) << SFC_CTRL_ADDR_BITS_SHIFT);
> > > +	}
> > > +
> > > +	/* set DUMMY */
> > > +	if (op->dummy.nbytes) {
> > > +		if (op->dummy.buswidth == 4)
> > > +			cmd |= op->dummy.nbytes * 2 << SFC_CMD_DUMMY_SHIFT;
> > > +		else if (op->dummy.buswidth == 2)
> > > +			cmd |= op->dummy.nbytes * 4 << SFC_CMD_DUMMY_SHIFT;
> > > +		else
> > > +			cmd |= op->dummy.nbytes * 8 << SFC_CMD_DUMMY_SHIFT;
> > > +	}
> > > +
> > I have no experience optimizing code or anything, but would it help
> > here to read the output of rockchip_sfc_get_version() to a variable
> > in the driver data, and then check that each time we hit this?
> > I don't know if there is any real-world difference in reading a
> > variable versus reading a register, so I'm just asking. It's not like
> > the sfc_version will change once the device is probed. :-)
> 
> done.
> 
> > > +	/* set DATA */
> > > +	if (rockchip_sfc_get_version(sfc) >= SFC_VER_4) /* Clear it if no data to transfer */
> > > +		writel(len, sfc->regbase + SFC_LEN_EXT);
> > > +	else
> > > +		cmd |= len << SFC_CMD_TRAN_BYTES_SHIFT;
> > > +	if (len) {
> > > +		if (op->data.dir == SPI_MEM_DATA_OUT)
> > > +			cmd |= SFC_CMD_DIR_WR << SFC_CMD_DIR_SHIFT;
> > > +
> > > +		ctrl |= ((op->data.buswidth >> 1) << SFC_CTRL_DATA_BITS_SHIFT);
> > > +	}
> > > +	if (!len && op->addr.nbytes)
> > > +		cmd |= SFC_CMD_DIR_WR << SFC_CMD_DIR_SHIFT;
> > > +
> > > +	/* set the Controller */
> > > +	ctrl |= SFC_CTRL_PHASE_SEL_NEGETIVE;
> > > +	cmd |= mem->spi->chip_select << SFC_CMD_CS_SHIFT;
> > > +
> > > +	dev_dbg(sfc->dev, "addr.nbytes=%x(x%d) dummy.nbytes=%x(x%d)\n",
> > > +		op->addr.nbytes, op->addr.buswidth,
> > > +		op->dummy.nbytes, op->dummy.buswidth);
> > > +	dev_dbg(sfc->dev, "ctrl=%x cmd=%x addr=%llx len=%x\n",
> > > +		ctrl, cmd, op->addr.val, len);
> > > +
> > > +	writel(ctrl, sfc->regbase + SFC_CTRL);
> > > +	writel(cmd, sfc->regbase + SFC_CMD);
> > > +	if (op->addr.nbytes)
> > > +		writel(op->addr.val, sfc->regbase + SFC_ADDR);
> > > +
> > > +	return 0;
> > > +}
> > > +
> > > +static int rockchip_sfc_write_fifo(struct rockchip_sfc *sfc, const u8 *buf, int len)
> > > +{
> > > +	u8 bytes = len & 0x3;
> > > +	u32 dwords;
> > > +	int tx_level;
> > > +	u32 write_words;
> > > +	u32 tmp = 0;
> > > +
> > > +	dwords = len >> 2;
> > > +	while (dwords) {
> > > +		tx_level = rockchip_sfc_wait_fifo_ready(sfc, SFC_CMD_DIR_WR, HZ);
> > > +		if (tx_level < 0)
> > > +			return tx_level;
> > > +		write_words = min_t(u32, tx_level, dwords);
> > > +		iowrite32_rep(sfc->regbase + SFC_DATA, buf, write_words);
> > > +		buf += write_words << 2;
> > > +		dwords -= write_words;
> > > +	}
> > > +
> > > +	/* write the rest non word aligned bytes */
> > > +	if (bytes) {
> > > +		tx_level = rockchip_sfc_wait_fifo_ready(sfc, SFC_CMD_DIR_WR, HZ);
> > > +		if (tx_level < 0)
> > > +			return tx_level;
> > > +		memcpy(&tmp, buf, bytes);
> > > +		writel(tmp, sfc->regbase + SFC_DATA);
> > > +	}
> > > +
> > > +	return len;
> > > +}
> > > +
> > > +static int rockchip_sfc_read_fifo(struct rockchip_sfc *sfc, u8 *buf, int len)
> > > +{
> > > +	u8 bytes = len & 0x3;
> > > +	u32 dwords;
> > > +	u8 read_words;
> > > +	int rx_level;
> > > +	int tmp;
> > > +
> > > +	/* word aligned access only */
> > > +	dwords = len >> 2;
> > > +	while (dwords) {
> > > +		rx_level = rockchip_sfc_wait_fifo_ready(sfc, SFC_CMD_DIR_RD, HZ);
> > > +		if (rx_level < 0)
> > > +			return rx_level;
> > > +		read_words = min_t(u32, rx_level, dwords);
> > > +		ioread32_rep(sfc->regbase + SFC_DATA, buf, read_words);
> > > +		buf += read_words << 2;
> > > +		dwords -= read_words;
> > > +	}
> > > +
> > > +	/* read the rest non word aligned bytes */
> > > +	if (bytes) {
> > > +		rx_level = rockchip_sfc_wait_fifo_ready(sfc, SFC_CMD_DIR_RD, HZ);
> > > +		if (rx_level < 0)
> > > +			return rx_level;
> > > +		tmp = readl_relaxed(sfc->regbase + SFC_DATA);
> > > +		memcpy(buf, &tmp, bytes);
> > > +	}
> > > +
> > > +	return len;
> > > +}
> > > +
> > > +static int rockchip_sfc_fifo_transfer_dma(struct rockchip_sfc *sfc, dma_addr_t dma_buf, size_t len)
> > > +{
> > > +	u32 reg;
> > > +	int err = 0;
> > > +
> > > +	init_completion(&sfc->cp);
> > > +
> > > +	writel(0xFFFFFFFF, sfc->regbase + SFC_ICLR);
> > > +
> > > +	/* Enable transfer complete interrupt */
> > > +	reg = readl(sfc->regbase + SFC_IMR);
> > > +	reg &= ~SFC_IMR_DMA;
> > > +	writel(reg, sfc->regbase + SFC_IMR);
> > > +	writel((u32)dma_buf, sfc->regbase + SFC_DMA_ADDR);
> > > +	writel(0x1, sfc->regbase + SFC_DMA_TRIGGER);
> > > +
> > > +	/* Wait for the interrupt. */
> > > +	if (!wait_for_completion_timeout(&sfc->cp, msecs_to_jiffies(2000))) {
> > > +		dev_err(sfc->dev, "DMA wait for transfer finish timeout\n");
> > > +		err = -ETIMEDOUT;
> > > +	}
> > > +
> > > +	writel(0xFFFFFFFF, sfc->regbase + SFC_ICLR);
> > > +	/* Disable transfer finish interrupt */
> > > +	reg = readl(sfc->regbase + SFC_IMR);
> > > +	reg |= SFC_IMR_DMA;
> > > +	writel(reg, sfc->regbase + SFC_IMR);
> > > +
> > > +	return len;
> > > +}
> > > +
> > > +static int rockchip_sfc_xfer_data_poll(struct rockchip_sfc *sfc,
> > > +				       const struct spi_mem_op *op, u32 len)
> > > +{
> > > +	dev_dbg(sfc->dev, "xfer_poll len=%x\n", len);
> > > +
> > > +	if (op->data.dir == SPI_MEM_DATA_OUT)
> > > +		return rockchip_sfc_write_fifo(sfc, op->data.buf.out, len);
> > > +	else
> > > +		return rockchip_sfc_read_fifo(sfc, op->data.buf.in, len);
> > > +}
> > > +
> > > +static int rockchip_sfc_xfer_data_dma(struct rockchip_sfc *sfc,
> > > +				      const struct spi_mem_op *op, u32 len)
> > > +{
> > > +	int ret;
> > > +
> > > +	dev_dbg(sfc->dev, "xfer_dma len=%x\n", len);
> > > +
> > > +	if (op->data.dir == SPI_MEM_DATA_OUT) {
> > > +		memcpy_toio(sfc->buffer, op->data.buf.out, len);
> > > +		ret = rockchip_sfc_fifo_transfer_dma(sfc, sfc->dma_buffer, len);
> > > +	} else {
> > > +		ret = rockchip_sfc_fifo_transfer_dma(sfc, sfc->dma_buffer, len);
> > > +		memcpy_fromio(op->data.buf.in, sfc->buffer, len);
> > > +	}
> > > +
> > > +	return ret;
> > > +}
> > > +
> > > +static int rockchip_sfc_xfer_done(struct rockchip_sfc *sfc, u32 timeout_us)
> > > +{
> > > +	int ret = 0;
> > > +	u32 status;
> > > +
> > > +	ret = readl_poll_timeout(sfc->regbase + SFC_SR, status,
> > > +				 !(status & SFC_SR_IS_BUSY),
> > > +				 20, timeout_us);
> > > +	if (ret) {
> > > +		dev_err(sfc->dev, "wait sfc idle timeout\n");
> > > +		rockchip_sfc_reset(sfc);
> > > +
> > > +		ret = -EIO;
> > > +	}
> > > +
> > > +	return ret;
> > > +}
> > > +
> > > +static int rockchip_sfc_exec_mem_op(struct spi_mem *mem, const struct spi_mem_op *op)
> > > +{
> > > +	struct rockchip_sfc *sfc = spi_master_get_devdata(mem->spi->master);
> > > +	u32 len = min_t(u32, op->data.nbytes, sfc->max_iosize);
> > Is it necessary to use the minimum of these 2 values if we use an
> > adjust_op_size which I propose below?
> it necessary if use adjust_op_size, done.
> > > +	int ret;
> > > +
> > Again I'm asking this as someone with no experience optimizing code
> > or with compilers, but would it help to add an unlikely() here? It's
> > all but guaranteed the first time this code path is called the clock
> > will need to be set, but each subsequent call should not really have
> > to care since the clock likely won't be changing (as long as we're on
> > the same CS). We can also then set the sfc->frequency to what the
> > max_speed_hz was after successfully setting the rate so we can hit the
> > faster path after the first run (it looks like it's checking the
> > uninitalized value but never updating it after it's changed, so it's
> > always doing the clk_set_rate()).
> > 
> > Also, should we clamp the clock speed between the min and max values?
> done.
> > > +	if (mem->spi->max_speed_hz != sfc->frequency) {
> > > +		if (clk_set_rate(sfc->clk, mem->spi->max_speed_hz))
> > > +			return ret;
> > > +		dev_dbg(sfc->dev, "set_freq=%dHz real_freq=%ldHz\n",
> > > +			sfc->frequency, clk_get_rate(sfc->clk));
> > > +	}
> > > +
> > > +	rockchip_sfc_xfer_setup(sfc, mem, op, len);
> > > +	if (len) {
> > > +		if (likely(sfc->use_dma) && !(len & 0x3) && len >= SFC_DMA_TRANS_THRETHOLD)
> > > +			ret = rockchip_sfc_xfer_data_dma(sfc, op, len);
> > > +		else
> > > +			ret = rockchip_sfc_xfer_data_poll(sfc, op, len);
> > > +
> > > +		if (ret != len) {
> > > +			dev_err(sfc->dev, "xfer data failed ret %d dir %d\n", ret, op->data.dir);
> > > +
> > > +			return -EIO;
> > > +		}
> > > +	}
> > > +
> > > +	return rockchip_sfc_xfer_done(sfc, 100000);
> > > +}
> > > +
> > I'll be honest with you, I had no idea if I needed this function or
> > not (the rockchip_sfc_get_name() function). I'm pretty sure it's not
> > needed though, so I assume it can be safely removed.
> > 
> > > +static const char *rockchip_sfc_get_name(struct spi_mem *mem)
> > > +{
> > > +	return devm_kasprintf(&mem->spi->dev, GFP_KERNEL, "%s.%d",
> > > +			      dev_name(&mem->spi->dev), mem->spi->chip_select);
> > > +}
> > > +
> > I had an issue with doing dd as mentioned above reading against mtd0,
> > and was able to fix it by adding a mem_op of adjust_op_size like this.
> > Not sure if it works for the v4/v5 hardware though.
> > 
> > static int rockchip_sfc_adjust_op_size(struct spi_mem *mem, struct spi_mem_op *op)
> > {
> > 	struct rockchip_sfc *sfc = spi_master_get_devdata(mem->spi->master);
> > 
> > 	op->data.nbytes = min(op->data.nbytes, sfc->max_iosize);
> > 	return 0;
> > }
> 
> good idea.
> 
> > > +static const struct spi_controller_mem_ops rockchip_sfc_mem_ops = {
> > > +	.exec_op = rockchip_sfc_exec_mem_op,
> > > +	.get_name = rockchip_sfc_get_name,
> > > +};
> > > +
> > > +static irqreturn_t rockchip_sfc_irq_handler(int irq, void *dev_id)
> > > +{
> > > +	struct rockchip_sfc *sfc = dev_id;
> > > +	u32 reg;
> > > +
> > > +	reg = readl(sfc->regbase + SFC_RISR);
> > > +
> > > +	/* Clear interrupt */
> > > +	writel_relaxed(reg, sfc->regbase + SFC_ICLR);
> > > +
> > > +	if (reg & SFC_RISR_DMA)
> > > +		complete(&sfc->cp);
> > > +
> > >From a previous comment, we should only clear the interrupt if we
> > handle the specific one in question.
> > 
> Compare with drivers/spi/spi-cadence-quadspi.c, I think the former code work
> well. Can explain more for the reason.
> 
> > > +	return IRQ_HANDLED;
> > > +}
> > > +
> > > +static int rockchip_sfc_probe(struct platform_device *pdev)
> > > +{
> > > +	struct device *dev = &pdev->dev;
> > > +	struct spi_master *master;
> > > +	struct resource *res;
> > > +	struct rockchip_sfc *sfc;
> > > +	int ret;
> > > +
> > > +	master = devm_spi_alloc_master(&pdev->dev, sizeof(*sfc));
> > > +	if (!master)
> > > +		return -ENOMEM;
> > > +
> > > +	master->mem_ops = &rockchip_sfc_mem_ops;
> > > +	master->dev.of_node = pdev->dev.of_node;
> > > +	master->mode_bits = SPI_TX_QUAD | SPI_TX_DUAL | SPI_RX_QUAD | SPI_RX_DUAL;
> > > +	master->min_speed_hz = SFC_MIN_SPEED_HZ;
> > > +	master->max_speed_hz = SFC_MAX_SPEED_HZ;
> > > +	master->num_chipselect = SFC_MAX_CHIPSELECT_NUM;
> > > +
> > > +	sfc = spi_master_get_devdata(master);
> > > +	sfc->dev = dev;
> > > +
> > > +	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > > +	sfc->regbase = devm_ioremap_resource(dev, res);
> > > +	if (IS_ERR(sfc->regbase))
> > > +		return PTR_ERR(sfc->regbase);
> > > +
> > > +	sfc->clk = devm_clk_get(&pdev->dev, "clk_sfc");
> > > +	if (IS_ERR(sfc->clk)) {
> > > +		dev_err(&pdev->dev, "Failed to get sfc interface clk\n");
> > > +		return PTR_ERR(sfc->clk);
> > > +	}
> > > +
> > > +	sfc->hclk = devm_clk_get(&pdev->dev, "hclk_sfc");
> > > +	if (IS_ERR(sfc->hclk)) {
> > > +		dev_err(&pdev->dev, "Failed to get sfc ahb clk\n");
> > > +		return PTR_ERR(sfc->hclk);
> > > +	}
> > > +
> > > +	sfc->use_dma = !of_property_read_bool(sfc->dev->of_node,
> > > +					      "rockchip,sfc-no-dma");
> > > +
> > > +	if (sfc->use_dma) {
> > > +		ret = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
> > > +		if (ret) {
> > > +			dev_warn(dev, "Unable to set dma mask\n");
> > > +			return ret;
> > > +		}
> > > +
> > > +		sfc->buffer = dmam_alloc_coherent(dev, SFC_MAX_IOSIZE_VER3,
> > > +						  &sfc->dma_buffer,
> > > +						  GFP_KERNEL);
> > > +		if (!sfc->buffer)
> > > +			return -ENOMEM;
> > > +	}
> > > +
> > > +	ret = clk_prepare_enable(sfc->hclk);
> > > +	if (ret) {
> > > +		dev_err(&pdev->dev, "Failed to enable ahb clk\n");
> > > +		goto err_hclk;
> > > +	}
> > > +
> > > +	ret = clk_prepare_enable(sfc->clk);
> > > +	if (ret) {
> > > +		dev_err(&pdev->dev, "Failed to enable interface clk\n");
> > > +		goto err_clk;
> > > +	}
> > > +
> > > +	/* Find the irq */
> > > +	ret = platform_get_irq(pdev, 0);
> > > +	if (ret < 0) {
> > > +		dev_err(dev, "Failed to get the irq\n");
> > > +		goto err_irq;
> > > +	}
> > > +
> > > +	ret = devm_request_irq(dev, ret, rockchip_sfc_irq_handler,
> > > +			       0, pdev->name, sfc);
> > > +	if (ret) {
> > > +		dev_err(dev, "Failed to request irq\n");
> > > +
> > > +		return ret;
> > > +	}
> > > +
> > > +	ret = rockchip_sfc_init(sfc);
> > > +	if (ret)
> > > +		goto err_irq;
> > > +
> > > +	sfc->max_iosize = rockchip_sfc_get_max_iosize(sfc);
> > > +
> > > +	ret = spi_register_master(master);
> > > +	if (ret)
> > > +		goto err_irq;
> > > +
> > > +	return 0;
> > > +
> > > +err_irq:
> > > +	clk_disable_unprepare(sfc->clk);
> > > +err_clk:
> > > +	clk_disable_unprepare(sfc->hclk);
> > > +err_hclk:
> > > +	return ret;
> > > +}
> > > +
> > > +static int rockchip_sfc_remove(struct platform_device *pdev)
> > > +{
> > > +	struct spi_master *master = platform_get_drvdata(pdev);
> > > +	struct rockchip_sfc *sfc = platform_get_drvdata(pdev);
> > > +
> > > +	spi_unregister_master(master);
> > > +
> > > +	clk_disable_unprepare(sfc->clk);
> > > +	clk_disable_unprepare(sfc->hclk);
> > > +
> > > +	return 0;
> > > +}
> > > +
> > > +static const struct of_device_id rockchip_sfc_dt_ids[] = {
> > > +	{ .compatible = "rockchip,px30-sfc"},
> > > +	{ .compatible = "rockchip,rk3036-sfc"},
> > > +	{ .compatible = "rockchip,rk3308-sfc"},
> > > +	{ .compatible = "rockchip,rv1126-sfc"},
> > > +	{ /* sentinel */ }
> > > +};
> > > +MODULE_DEVICE_TABLE(of, rockchip_sfc_dt_ids);
> > > +
> > > +static struct platform_driver rockchip_sfc_driver = {
> > > +	.driver = {
> > > +		.name	= "rockchip-sfc",
> > > +		.of_match_table = rockchip_sfc_dt_ids,
> > > +	},
> > > +	.probe	= rockchip_sfc_probe,
> > > +	.remove	= rockchip_sfc_remove,
> > > +};
> > > +module_platform_driver(rockchip_sfc_driver);
> > > +
> > > +MODULE_LICENSE("GPL v2");
> > > +MODULE_DESCRIPTION("Rockchip Serial Flash Controller Driver");
> > > +MODULE_AUTHOR("Shawn Lin <shawn.lin@rock-chips.com>");
> > > +MODULE_AUTHOR("Chris Morgan <macroalpha82@gmail.com>");
> > I know folks will hate on me for this (and for good reason given it
> > messes up IDs and stuff), but I prefer macromorgan@hotmail.com if my
> > email is in here, as that's the one I've been using for decades...
> ok.
> > > +MODULE_AUTHOR("Jon Lin <Jon.lin@rock-chips.com>");
> > > -- 
> > > 2.17.1
> > > 
> > > 
> > > 
> > 
> > 
> 
>