| Message ID | 20200626155832.2323789-2-ardb@kernel.org |
|---|---|
| State | Accepted |
| Commit | 1583052d111f8ea43f9954c5e749164fd2b954af |
| Headers | show |
| Series | [v3,1/2] arm64/acpi: disallow AML memory opregions to access kernel memory | expand |
On Fri, 26 Jun 2020 17:58:31 +0200 Ard Biesheuvel <ardb@kernel.org> wrote: > AML uses SystemMemory opregions to allow AML handlers to access MMIO > registers of, e.g., GPIO controllers, or access reserved regions of > memory that are owned by the firmware. > > Currently, we also allow AML access to memory that is owned by the > kernel and mapped via the linear region, which does not seem to be > supported by a valid use case, and exposes the kernel's internal > state to AML methods that may be buggy and exploitable. > > On arm64, ACPI support requires booting in EFI mode, and so we can cross > reference the requested region against the EFI memory map, rather than > just do a minimal check on the first page. So let's only permit regions > to be remapped by the ACPI core if > - they don't appear in the EFI memory map at all (which is the case for > most MMIO), or > - they are covered by a single region in the EFI memory map, which is not > of a type that describes memory that is given to the kernel at boot. > > Reported-by: Jason A. Donenfeld <Jason@zx2c4.com> > Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Hi Ard, Ran into a problem with this one. See below > --- > arch/arm64/include/asm/acpi.h | 15 +---- > arch/arm64/kernel/acpi.c | 66 ++++++++++++++++++++ > 2 files changed, 67 insertions(+), 14 deletions(-) > > diff --git a/arch/arm64/include/asm/acpi.h b/arch/arm64/include/asm/acpi.h > index a45366c3909b..bd68e1b7f29f 100644 > --- a/arch/arm64/include/asm/acpi.h > +++ b/arch/arm64/include/asm/acpi.h > @@ -47,20 +47,7 @@ > pgprot_t __acpi_get_mem_attribute(phys_addr_t addr); > > /* ACPI table mapping after acpi_permanent_mmap is set */ > -static inline void __iomem *acpi_os_ioremap(acpi_physical_address phys, > - acpi_size size) > -{ > - /* For normal memory we already have a cacheable mapping. */ > - if (memblock_is_map_memory(phys)) > - return (void __iomem *)__phys_to_virt(phys); > - > - /* > - * We should still honor the memory's attribute here because > - * crash dump kernel possibly excludes some ACPI (reclaim) > - * regions from memblock list. > - */ > - return __ioremap(phys, size, __acpi_get_mem_attribute(phys)); > -} > +void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size); > #define acpi_os_ioremap acpi_os_ioremap > > typedef u64 phys_cpuid_t; > diff --git a/arch/arm64/kernel/acpi.c b/arch/arm64/kernel/acpi.c > index a7586a4db142..01b861e225b0 100644 > --- a/arch/arm64/kernel/acpi.c > +++ b/arch/arm64/kernel/acpi.c > @@ -261,6 +261,72 @@ pgprot_t __acpi_get_mem_attribute(phys_addr_t addr) > return __pgprot(PROT_DEVICE_nGnRnE); > } > > +void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size) > +{ > + efi_memory_desc_t *md, *region = NULL; > + pgprot_t prot; > + > + if (WARN_ON_ONCE(!efi_enabled(EFI_MEMMAP))) > + return NULL; > + > + for_each_efi_memory_desc(md) { > + u64 end = md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT); > + > + if (phys < md->phys_addr || phys >= end) > + continue; > + > + if (phys + size > end) { > + pr_warn(FW_BUG "requested region covers multiple EFI memory regions\n"); > + return NULL; > + } > + region = md; > + break; > + } > + > + /* > + * It is fine for AML to remap regions that are not represented in the > + * EFI memory map at all, as it only describes normal memory, and MMIO > + * regions that require a virtual mapping to make them accessible to > + * the EFI runtime services. > + */ > + prot = __pgprot(PROT_DEVICE_nGnRnE); > + if (region) { > + switch (region->type) { > + case EFI_LOADER_CODE: > + case EFI_LOADER_DATA: Unfortunately this seems to have broken overriding of ACPI tables from an initrd. My particular test environment is qemu + EDK2. It only has obvious visible affect on tables that are used late in the boot such as PPTT as they get dropped before they are used. These are read after ACPICA is initialized and acpi_reallocate_root_table() has been called. The back trace is: acpi_os_ioremap+0xfc/0x288 acpi_os_map_iomem+0xc4/0x188 acpi_os_map_memory+0x18/0x28 acpi_tb_acquire_table+0x54/0x8c acpi_tb_validate_table+0x34/0x5c acpi_tb_validate_temp_table+0x34/0x40 acpi_tb_verify_temp_table+0x48/0x250 acpi_reallocate_root_table+0x12c/0x160 Seems that the table is in a region of type EFI_LOADER_DATA. I don't really know enough about this area to be sure what the right fix is or even whether this is a kernel issue, or one that should be fixed elsewhere in the stack. For now I'm just carry a hack that treats EFI_LOADER_DATA in the same fashion as EFI_ACPI_RECLAIM_MEMORY below. What's the right way to fix this? Jonathan > + case EFI_BOOT_SERVICES_CODE: > + case EFI_BOOT_SERVICES_DATA: > + case EFI_CONVENTIONAL_MEMORY: > + case EFI_PERSISTENT_MEMORY: > + pr_warn(FW_BUG "requested region covers kernel memory @ %pa\n", &phys); > + return NULL; > + > + case EFI_ACPI_RECLAIM_MEMORY: > + /* > + * ACPI reclaim memory is used to pass firmware tables > + * and other data that is intended for consumption by > + * the OS only, which may decide it wants to reclaim > + * that memory and use it for something else. We never > + * do that, but we usually add it to the linear map > + * anyway, in which case we should use the existing > + * mapping. > + */ > + if (memblock_is_map_memory(phys)) > + return (void __iomem *)__phys_to_virt(phys); > + /* fall through */ > + > + default: > + if (region->attribute & EFI_MEMORY_WB) > + prot = PAGE_KERNEL; > + else if (region->attribute & EFI_MEMORY_WT) > + prot = __pgprot(PROT_NORMAL_WT); > + else if (region->attribute & EFI_MEMORY_WC) > + prot = __pgprot(PROT_NORMAL_NC); > + } > + } > + return __ioremap(phys, size, prot); > +} > + > /* > * Claim Synchronous External Aborts as a firmware first notification. > *
On Mon, 28 Sep 2020 at 18:02, Jonathan Cameron <Jonathan.Cameron@huawei.com> wrote: > > On Fri, 26 Jun 2020 17:58:31 +0200 > Ard Biesheuvel <ardb@kernel.org> wrote: > > > AML uses SystemMemory opregions to allow AML handlers to access MMIO > > registers of, e.g., GPIO controllers, or access reserved regions of > > memory that are owned by the firmware. > > > > Currently, we also allow AML access to memory that is owned by the > > kernel and mapped via the linear region, which does not seem to be > > supported by a valid use case, and exposes the kernel's internal > > state to AML methods that may be buggy and exploitable. > > > > On arm64, ACPI support requires booting in EFI mode, and so we can cross > > reference the requested region against the EFI memory map, rather than > > just do a minimal check on the first page. So let's only permit regions > > to be remapped by the ACPI core if > > - they don't appear in the EFI memory map at all (which is the case for > > most MMIO), or > > - they are covered by a single region in the EFI memory map, which is not > > of a type that describes memory that is given to the kernel at boot. > > > > Reported-by: Jason A. Donenfeld <Jason@zx2c4.com> > > Signed-off-by: Ard Biesheuvel <ardb@kernel.org> > > Hi Ard, > > Ran into a problem with this one. See below > > > --- > > arch/arm64/include/asm/acpi.h | 15 +---- > > arch/arm64/kernel/acpi.c | 66 ++++++++++++++++++++ > > 2 files changed, 67 insertions(+), 14 deletions(-) > > > > diff --git a/arch/arm64/include/asm/acpi.h b/arch/arm64/include/asm/acpi.h > > index a45366c3909b..bd68e1b7f29f 100644 > > --- a/arch/arm64/include/asm/acpi.h > > +++ b/arch/arm64/include/asm/acpi.h > > @@ -47,20 +47,7 @@ > > pgprot_t __acpi_get_mem_attribute(phys_addr_t addr); > > > > /* ACPI table mapping after acpi_permanent_mmap is set */ > > -static inline void __iomem *acpi_os_ioremap(acpi_physical_address phys, > > - acpi_size size) > > -{ > > - /* For normal memory we already have a cacheable mapping. */ > > - if (memblock_is_map_memory(phys)) > > - return (void __iomem *)__phys_to_virt(phys); > > - > > - /* > > - * We should still honor the memory's attribute here because > > - * crash dump kernel possibly excludes some ACPI (reclaim) > > - * regions from memblock list. > > - */ > > - return __ioremap(phys, size, __acpi_get_mem_attribute(phys)); > > -} > > +void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size); > > #define acpi_os_ioremap acpi_os_ioremap > > > > typedef u64 phys_cpuid_t; > > diff --git a/arch/arm64/kernel/acpi.c b/arch/arm64/kernel/acpi.c > > index a7586a4db142..01b861e225b0 100644 > > --- a/arch/arm64/kernel/acpi.c > > +++ b/arch/arm64/kernel/acpi.c > > @@ -261,6 +261,72 @@ pgprot_t __acpi_get_mem_attribute(phys_addr_t addr) > > return __pgprot(PROT_DEVICE_nGnRnE); > > } > > > > +void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size) > > +{ > > + efi_memory_desc_t *md, *region = NULL; > > + pgprot_t prot; > > + > > + if (WARN_ON_ONCE(!efi_enabled(EFI_MEMMAP))) > > + return NULL; > > + > > + for_each_efi_memory_desc(md) { > > + u64 end = md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT); > > + > > + if (phys < md->phys_addr || phys >= end) > > + continue; > > + > > + if (phys + size > end) { > > + pr_warn(FW_BUG "requested region covers multiple EFI memory regions\n"); > > + return NULL; > > + } > > + region = md; > > + break; > > + } > > + > > + /* > > + * It is fine for AML to remap regions that are not represented in the > > + * EFI memory map at all, as it only describes normal memory, and MMIO > > + * regions that require a virtual mapping to make them accessible to > > + * the EFI runtime services. > > + */ > > + prot = __pgprot(PROT_DEVICE_nGnRnE); > > + if (region) { > > + switch (region->type) { > > + case EFI_LOADER_CODE: > > + case EFI_LOADER_DATA: > > Unfortunately this seems to have broken overriding of ACPI tables from an initrd. > My particular test environment is qemu + EDK2. > > It only has obvious visible affect on tables that are used late in the boot such as PPTT > as they get dropped before they are used. > > These are read after ACPICA is initialized and acpi_reallocate_root_table() > has been called. The back trace is: > > acpi_os_ioremap+0xfc/0x288 > acpi_os_map_iomem+0xc4/0x188 > acpi_os_map_memory+0x18/0x28 > acpi_tb_acquire_table+0x54/0x8c > acpi_tb_validate_table+0x34/0x5c > acpi_tb_validate_temp_table+0x34/0x40 > acpi_tb_verify_temp_table+0x48/0x250 > acpi_reallocate_root_table+0x12c/0x160 > > Seems that the table is in a region of type EFI_LOADER_DATA. > > I don't really know enough about this area to be sure what the right fix is or > even whether this is a kernel issue, or one that should be fixed elsewhere in > the stack. > > For now I'm just carry a hack that treats EFI_LOADER_DATA in the same fashion as > EFI_ACPI_RECLAIM_MEMORY below. > > What's the right way to fix this? > Hi Jonathan, That is an excellent question. The purpose of this change is to ensure that firmware cannot manipulate the internal state of the kernel. So as long as we can ensure that this memory is not claimed by the kernel's memory subsystem, we should be fine. Since this is an obvious debug feature, what we could do is reserve this memory permanently in some way, and make the test take this into account. Do you have a full stack trace? How early does this run?
On Mon, 28 Sep 2020 18:49:35 +0200 Ard Biesheuvel <ardb@kernel.org> wrote: > On Mon, 28 Sep 2020 at 18:02, Jonathan Cameron > <Jonathan.Cameron@huawei.com> wrote: > > > > On Fri, 26 Jun 2020 17:58:31 +0200 > > Ard Biesheuvel <ardb@kernel.org> wrote: > > > > > AML uses SystemMemory opregions to allow AML handlers to access MMIO > > > registers of, e.g., GPIO controllers, or access reserved regions of > > > memory that are owned by the firmware. > > > > > > Currently, we also allow AML access to memory that is owned by the > > > kernel and mapped via the linear region, which does not seem to be > > > supported by a valid use case, and exposes the kernel's internal > > > state to AML methods that may be buggy and exploitable. > > > > > > On arm64, ACPI support requires booting in EFI mode, and so we can cross > > > reference the requested region against the EFI memory map, rather than > > > just do a minimal check on the first page. So let's only permit regions > > > to be remapped by the ACPI core if > > > - they don't appear in the EFI memory map at all (which is the case for > > > most MMIO), or > > > - they are covered by a single region in the EFI memory map, which is not > > > of a type that describes memory that is given to the kernel at boot. > > > > > > Reported-by: Jason A. Donenfeld <Jason@zx2c4.com> > > > Signed-off-by: Ard Biesheuvel <ardb@kernel.org> > > > > Hi Ard, > > > > Ran into a problem with this one. See below > > > > > --- > > > arch/arm64/include/asm/acpi.h | 15 +---- > > > arch/arm64/kernel/acpi.c | 66 ++++++++++++++++++++ > > > 2 files changed, 67 insertions(+), 14 deletions(-) > > > > > > diff --git a/arch/arm64/include/asm/acpi.h b/arch/arm64/include/asm/acpi.h > > > index a45366c3909b..bd68e1b7f29f 100644 > > > --- a/arch/arm64/include/asm/acpi.h > > > +++ b/arch/arm64/include/asm/acpi.h > > > @@ -47,20 +47,7 @@ > > > pgprot_t __acpi_get_mem_attribute(phys_addr_t addr); > > > > > > /* ACPI table mapping after acpi_permanent_mmap is set */ > > > -static inline void __iomem *acpi_os_ioremap(acpi_physical_address phys, > > > - acpi_size size) > > > -{ > > > - /* For normal memory we already have a cacheable mapping. */ > > > - if (memblock_is_map_memory(phys)) > > > - return (void __iomem *)__phys_to_virt(phys); > > > - > > > - /* > > > - * We should still honor the memory's attribute here because > > > - * crash dump kernel possibly excludes some ACPI (reclaim) > > > - * regions from memblock list. > > > - */ > > > - return __ioremap(phys, size, __acpi_get_mem_attribute(phys)); > > > -} > > > +void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size); > > > #define acpi_os_ioremap acpi_os_ioremap > > > > > > typedef u64 phys_cpuid_t; > > > diff --git a/arch/arm64/kernel/acpi.c b/arch/arm64/kernel/acpi.c > > > index a7586a4db142..01b861e225b0 100644 > > > --- a/arch/arm64/kernel/acpi.c > > > +++ b/arch/arm64/kernel/acpi.c > > > @@ -261,6 +261,72 @@ pgprot_t __acpi_get_mem_attribute(phys_addr_t addr) > > > return __pgprot(PROT_DEVICE_nGnRnE); > > > } > > > > > > +void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size) > > > +{ > > > + efi_memory_desc_t *md, *region = NULL; > > > + pgprot_t prot; > > > + > > > + if (WARN_ON_ONCE(!efi_enabled(EFI_MEMMAP))) > > > + return NULL; > > > + > > > + for_each_efi_memory_desc(md) { > > > + u64 end = md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT); > > > + > > > + if (phys < md->phys_addr || phys >= end) > > > + continue; > > > + > > > + if (phys + size > end) { > > > + pr_warn(FW_BUG "requested region covers multiple EFI memory regions\n"); > > > + return NULL; > > > + } > > > + region = md; > > > + break; > > > + } > > > + > > > + /* > > > + * It is fine for AML to remap regions that are not represented in the > > > + * EFI memory map at all, as it only describes normal memory, and MMIO > > > + * regions that require a virtual mapping to make them accessible to > > > + * the EFI runtime services. > > > + */ > > > + prot = __pgprot(PROT_DEVICE_nGnRnE); > > > + if (region) { > > > + switch (region->type) { > > > + case EFI_LOADER_CODE: > > > + case EFI_LOADER_DATA: > > > > Unfortunately this seems to have broken overriding of ACPI tables from an initrd. > > My particular test environment is qemu + EDK2. > > > > It only has obvious visible affect on tables that are used late in the boot such as PPTT > > as they get dropped before they are used. > > > > These are read after ACPICA is initialized and acpi_reallocate_root_table() > > has been called. The back trace is: > > > > acpi_os_ioremap+0xfc/0x288 > > acpi_os_map_iomem+0xc4/0x188 > > acpi_os_map_memory+0x18/0x28 > > acpi_tb_acquire_table+0x54/0x8c > > acpi_tb_validate_table+0x34/0x5c > > acpi_tb_validate_temp_table+0x34/0x40 > > acpi_tb_verify_temp_table+0x48/0x250 > > acpi_reallocate_root_table+0x12c/0x160 > > > > Seems that the table is in a region of type EFI_LOADER_DATA. > > > > I don't really know enough about this area to be sure what the right fix is or > > even whether this is a kernel issue, or one that should be fixed elsewhere in > > the stack. > > > > For now I'm just carry a hack that treats EFI_LOADER_DATA in the same fashion as > > EFI_ACPI_RECLAIM_MEMORY below. > > > > What's the right way to fix this? > > > > Hi Jonathan, > > That is an excellent question. > > The purpose of this change is to ensure that firmware cannot > manipulate the internal state of the kernel. So as long as we can > ensure that this memory is not claimed by the kernel's memory > subsystem, we should be fine. > > Since this is an obvious debug feature, what we could do is reserve > this memory permanently in some way, and make the test take this into > account. Whilst it is a debug feature, I wonder if it gets shipped in production hardware. If not, could be we cynical and just drop the check if the relevant config option is enabled? Perhaps just don't release the EFI_LOADER_DATA for other use? (if this option is enabled only) > > Do you have a full stack trace? How early does this run? For the place where it first occurs, ie the trace above, the acpi_reallocate_root_table() is the call from acpi_early_init() from start_kernel(). We hit the table a lot during later calls though and hence would run into the same problem. Jonathan
On Mon, 28 Sep 2020 at 19:18, Jonathan Cameron <Jonathan.Cameron@huawei.com> wrote: > > On Mon, 28 Sep 2020 18:49:35 +0200 > Ard Biesheuvel <ardb@kernel.org> wrote: > > > On Mon, 28 Sep 2020 at 18:02, Jonathan Cameron > > <Jonathan.Cameron@huawei.com> wrote: > > > > > > On Fri, 26 Jun 2020 17:58:31 +0200 > > > Ard Biesheuvel <ardb@kernel.org> wrote: > > > > > > > AML uses SystemMemory opregions to allow AML handlers to access MMIO > > > > registers of, e.g., GPIO controllers, or access reserved regions of > > > > memory that are owned by the firmware. > > > > > > > > Currently, we also allow AML access to memory that is owned by the > > > > kernel and mapped via the linear region, which does not seem to be > > > > supported by a valid use case, and exposes the kernel's internal > > > > state to AML methods that may be buggy and exploitable. > > > > > > > > On arm64, ACPI support requires booting in EFI mode, and so we can cross > > > > reference the requested region against the EFI memory map, rather than > > > > just do a minimal check on the first page. So let's only permit regions > > > > to be remapped by the ACPI core if > > > > - they don't appear in the EFI memory map at all (which is the case for > > > > most MMIO), or > > > > - they are covered by a single region in the EFI memory map, which is not > > > > of a type that describes memory that is given to the kernel at boot. > > > > > > > > Reported-by: Jason A. Donenfeld <Jason@zx2c4.com> > > > > Signed-off-by: Ard Biesheuvel <ardb@kernel.org> > > > > > > Hi Ard, > > > > > > Ran into a problem with this one. See below > > > > > > > --- > > > > arch/arm64/include/asm/acpi.h | 15 +---- > > > > arch/arm64/kernel/acpi.c | 66 ++++++++++++++++++++ > > > > 2 files changed, 67 insertions(+), 14 deletions(-) > > > > > > > > diff --git a/arch/arm64/include/asm/acpi.h b/arch/arm64/include/asm/acpi.h > > > > index a45366c3909b..bd68e1b7f29f 100644 > > > > --- a/arch/arm64/include/asm/acpi.h > > > > +++ b/arch/arm64/include/asm/acpi.h > > > > @@ -47,20 +47,7 @@ > > > > pgprot_t __acpi_get_mem_attribute(phys_addr_t addr); > > > > > > > > /* ACPI table mapping after acpi_permanent_mmap is set */ > > > > -static inline void __iomem *acpi_os_ioremap(acpi_physical_address phys, > > > > - acpi_size size) > > > > -{ > > > > - /* For normal memory we already have a cacheable mapping. */ > > > > - if (memblock_is_map_memory(phys)) > > > > - return (void __iomem *)__phys_to_virt(phys); > > > > - > > > > - /* > > > > - * We should still honor the memory's attribute here because > > > > - * crash dump kernel possibly excludes some ACPI (reclaim) > > > > - * regions from memblock list. > > > > - */ > > > > - return __ioremap(phys, size, __acpi_get_mem_attribute(phys)); > > > > -} > > > > +void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size); > > > > #define acpi_os_ioremap acpi_os_ioremap > > > > > > > > typedef u64 phys_cpuid_t; > > > > diff --git a/arch/arm64/kernel/acpi.c b/arch/arm64/kernel/acpi.c > > > > index a7586a4db142..01b861e225b0 100644 > > > > --- a/arch/arm64/kernel/acpi.c > > > > +++ b/arch/arm64/kernel/acpi.c > > > > @@ -261,6 +261,72 @@ pgprot_t __acpi_get_mem_attribute(phys_addr_t addr) > > > > return __pgprot(PROT_DEVICE_nGnRnE); > > > > } > > > > > > > > +void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size) > > > > +{ > > > > + efi_memory_desc_t *md, *region = NULL; > > > > + pgprot_t prot; > > > > + > > > > + if (WARN_ON_ONCE(!efi_enabled(EFI_MEMMAP))) > > > > + return NULL; > > > > + > > > > + for_each_efi_memory_desc(md) { > > > > + u64 end = md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT); > > > > + > > > > + if (phys < md->phys_addr || phys >= end) > > > > + continue; > > > > + > > > > + if (phys + size > end) { > > > > + pr_warn(FW_BUG "requested region covers multiple EFI memory regions\n"); > > > > + return NULL; > > > > + } > > > > + region = md; > > > > + break; > > > > + } > > > > + > > > > + /* > > > > + * It is fine for AML to remap regions that are not represented in the > > > > + * EFI memory map at all, as it only describes normal memory, and MMIO > > > > + * regions that require a virtual mapping to make them accessible to > > > > + * the EFI runtime services. > > > > + */ > > > > + prot = __pgprot(PROT_DEVICE_nGnRnE); > > > > + if (region) { > > > > + switch (region->type) { > > > > + case EFI_LOADER_CODE: > > > > + case EFI_LOADER_DATA: > > > > > > Unfortunately this seems to have broken overriding of ACPI tables from an initrd. > > > My particular test environment is qemu + EDK2. > > > > > > It only has obvious visible affect on tables that are used late in the boot such as PPTT > > > as they get dropped before they are used. > > > > > > These are read after ACPICA is initialized and acpi_reallocate_root_table() > > > has been called. The back trace is: > > > > > > acpi_os_ioremap+0xfc/0x288 > > > acpi_os_map_iomem+0xc4/0x188 > > > acpi_os_map_memory+0x18/0x28 > > > acpi_tb_acquire_table+0x54/0x8c > > > acpi_tb_validate_table+0x34/0x5c > > > acpi_tb_validate_temp_table+0x34/0x40 > > > acpi_tb_verify_temp_table+0x48/0x250 > > > acpi_reallocate_root_table+0x12c/0x160 > > > > > > Seems that the table is in a region of type EFI_LOADER_DATA. > > > > > > I don't really know enough about this area to be sure what the right fix is or > > > even whether this is a kernel issue, or one that should be fixed elsewhere in > > > the stack. > > > > > > For now I'm just carry a hack that treats EFI_LOADER_DATA in the same fashion as > > > EFI_ACPI_RECLAIM_MEMORY below. > > > > > > What's the right way to fix this? > > > > > > > Hi Jonathan, > > > > That is an excellent question. > > > > The purpose of this change is to ensure that firmware cannot > > manipulate the internal state of the kernel. So as long as we can > > ensure that this memory is not claimed by the kernel's memory > > subsystem, we should be fine. > > > > Since this is an obvious debug feature, what we could do is reserve > > this memory permanently in some way, and make the test take this into > > account. > > Whilst it is a debug feature, I wonder if it gets shipped in production > hardware. If not, could be we cynical and just drop the check if the > relevant config option is enabled? > > Perhaps just don't release the EFI_LOADER_DATA for other use? (if > this option is enabled only) > > > > > Do you have a full stack trace? How early does this run? > > For the place where it first occurs, ie the trace above, the acpi_reallocate_root_table() is > the call from acpi_early_init() from start_kernel(). > > We hit the table a lot during later calls though and hence would run into the > same problem. > Could you try the patch below? Since the memory holding the tables is already memblock_reserve()d, we can just mark it NOMAP, and permit r/o remapping of NOMAP regions. diff --git a/arch/arm64/kernel/acpi.c b/arch/arm64/kernel/acpi.c index a85174d05473..84da869c5ac4 100644 --- a/arch/arm64/kernel/acpi.c +++ b/arch/arm64/kernel/acpi.c @@ -298,8 +298,11 @@ void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size) case EFI_BOOT_SERVICES_DATA: case EFI_CONVENTIONAL_MEMORY: case EFI_PERSISTENT_MEMORY: - pr_warn(FW_BUG "requested region covers kernel memory @ %pa\n", &phys); - return NULL; + if (memblock_is_map_memory(phys)) { + pr_warn(FW_BUG "requested region covers kernel memory @ %pa\n", &phys); + return NULL; + } + fallthrough; case EFI_RUNTIME_SERVICES_CODE: /* @@ -388,3 +391,8 @@ int apei_claim_sea(struct pt_regs *regs) return err; } + +void arch_reserve_mem_area(acpi_physical_address addr, size_t size) +{ + memblock_mark_nomap(addr, size); +} diff --git a/include/linux/acpi.h b/include/linux/acpi.h index 1e4cdc6c7ae2..64ae25c59d55 100644 --- a/include/linux/acpi.h +++ b/include/linux/acpi.h @@ -958,7 +958,7 @@ void acpi_os_set_prepare_extended_sleep(int (*func)(u8 sleep_state, acpi_status acpi_os_prepare_extended_sleep(u8 sleep_state, u32 val_a, u32 val_b); -#ifdef CONFIG_X86 +#ifndef CONFIG_IA64 void arch_reserve_mem_area(acpi_physical_address addr, size_t size); #else static inline void arch_reserve_mem_area(acpi_physical_address addr,
On Tue, 29 Sep 2020 11:29:48 +0200 Ard Biesheuvel <ardb@kernel.org> wrote: > On Mon, 28 Sep 2020 at 19:18, Jonathan Cameron > <Jonathan.Cameron@huawei.com> wrote: > > > > On Mon, 28 Sep 2020 18:49:35 +0200 > > Ard Biesheuvel <ardb@kernel.org> wrote: > > > > > On Mon, 28 Sep 2020 at 18:02, Jonathan Cameron > > > <Jonathan.Cameron@huawei.com> wrote: > > > > > > > > On Fri, 26 Jun 2020 17:58:31 +0200 > > > > Ard Biesheuvel <ardb@kernel.org> wrote: > > > > > > > > > AML uses SystemMemory opregions to allow AML handlers to access MMIO > > > > > registers of, e.g., GPIO controllers, or access reserved regions of > > > > > memory that are owned by the firmware. > > > > > > > > > > Currently, we also allow AML access to memory that is owned by the > > > > > kernel and mapped via the linear region, which does not seem to be > > > > > supported by a valid use case, and exposes the kernel's internal > > > > > state to AML methods that may be buggy and exploitable. > > > > > > > > > > On arm64, ACPI support requires booting in EFI mode, and so we can cross > > > > > reference the requested region against the EFI memory map, rather than > > > > > just do a minimal check on the first page. So let's only permit regions > > > > > to be remapped by the ACPI core if > > > > > - they don't appear in the EFI memory map at all (which is the case for > > > > > most MMIO), or > > > > > - they are covered by a single region in the EFI memory map, which is not > > > > > of a type that describes memory that is given to the kernel at boot. > > > > > > > > > > Reported-by: Jason A. Donenfeld <Jason@zx2c4.com> > > > > > Signed-off-by: Ard Biesheuvel <ardb@kernel.org> > > > > > > > > Hi Ard, > > > > > > > > Ran into a problem with this one. See below > > > > > > > > > --- > > > > > arch/arm64/include/asm/acpi.h | 15 +---- > > > > > arch/arm64/kernel/acpi.c | 66 ++++++++++++++++++++ > > > > > 2 files changed, 67 insertions(+), 14 deletions(-) > > > > > > > > > > diff --git a/arch/arm64/include/asm/acpi.h b/arch/arm64/include/asm/acpi.h > > > > > index a45366c3909b..bd68e1b7f29f 100644 > > > > > --- a/arch/arm64/include/asm/acpi.h > > > > > +++ b/arch/arm64/include/asm/acpi.h > > > > > @@ -47,20 +47,7 @@ > > > > > pgprot_t __acpi_get_mem_attribute(phys_addr_t addr); > > > > > > > > > > /* ACPI table mapping after acpi_permanent_mmap is set */ > > > > > -static inline void __iomem *acpi_os_ioremap(acpi_physical_address phys, > > > > > - acpi_size size) > > > > > -{ > > > > > - /* For normal memory we already have a cacheable mapping. */ > > > > > - if (memblock_is_map_memory(phys)) > > > > > - return (void __iomem *)__phys_to_virt(phys); > > > > > - > > > > > - /* > > > > > - * We should still honor the memory's attribute here because > > > > > - * crash dump kernel possibly excludes some ACPI (reclaim) > > > > > - * regions from memblock list. > > > > > - */ > > > > > - return __ioremap(phys, size, __acpi_get_mem_attribute(phys)); > > > > > -} > > > > > +void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size); > > > > > #define acpi_os_ioremap acpi_os_ioremap > > > > > > > > > > typedef u64 phys_cpuid_t; > > > > > diff --git a/arch/arm64/kernel/acpi.c b/arch/arm64/kernel/acpi.c > > > > > index a7586a4db142..01b861e225b0 100644 > > > > > --- a/arch/arm64/kernel/acpi.c > > > > > +++ b/arch/arm64/kernel/acpi.c > > > > > @@ -261,6 +261,72 @@ pgprot_t __acpi_get_mem_attribute(phys_addr_t addr) > > > > > return __pgprot(PROT_DEVICE_nGnRnE); > > > > > } > > > > > > > > > > +void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size) > > > > > +{ > > > > > + efi_memory_desc_t *md, *region = NULL; > > > > > + pgprot_t prot; > > > > > + > > > > > + if (WARN_ON_ONCE(!efi_enabled(EFI_MEMMAP))) > > > > > + return NULL; > > > > > + > > > > > + for_each_efi_memory_desc(md) { > > > > > + u64 end = md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT); > > > > > + > > > > > + if (phys < md->phys_addr || phys >= end) > > > > > + continue; > > > > > + > > > > > + if (phys + size > end) { > > > > > + pr_warn(FW_BUG "requested region covers multiple EFI memory regions\n"); > > > > > + return NULL; > > > > > + } > > > > > + region = md; > > > > > + break; > > > > > + } > > > > > + > > > > > + /* > > > > > + * It is fine for AML to remap regions that are not represented in the > > > > > + * EFI memory map at all, as it only describes normal memory, and MMIO > > > > > + * regions that require a virtual mapping to make them accessible to > > > > > + * the EFI runtime services. > > > > > + */ > > > > > + prot = __pgprot(PROT_DEVICE_nGnRnE); > > > > > + if (region) { > > > > > + switch (region->type) { > > > > > + case EFI_LOADER_CODE: > > > > > + case EFI_LOADER_DATA: > > > > > > > > Unfortunately this seems to have broken overriding of ACPI tables from an initrd. > > > > My particular test environment is qemu + EDK2. > > > > > > > > It only has obvious visible affect on tables that are used late in the boot such as PPTT > > > > as they get dropped before they are used. > > > > > > > > These are read after ACPICA is initialized and acpi_reallocate_root_table() > > > > has been called. The back trace is: > > > > > > > > acpi_os_ioremap+0xfc/0x288 > > > > acpi_os_map_iomem+0xc4/0x188 > > > > acpi_os_map_memory+0x18/0x28 > > > > acpi_tb_acquire_table+0x54/0x8c > > > > acpi_tb_validate_table+0x34/0x5c > > > > acpi_tb_validate_temp_table+0x34/0x40 > > > > acpi_tb_verify_temp_table+0x48/0x250 > > > > acpi_reallocate_root_table+0x12c/0x160 > > > > > > > > Seems that the table is in a region of type EFI_LOADER_DATA. > > > > > > > > I don't really know enough about this area to be sure what the right fix is or > > > > even whether this is a kernel issue, or one that should be fixed elsewhere in > > > > the stack. > > > > > > > > For now I'm just carry a hack that treats EFI_LOADER_DATA in the same fashion as > > > > EFI_ACPI_RECLAIM_MEMORY below. > > > > > > > > What's the right way to fix this? > > > > > > > > > > Hi Jonathan, > > > > > > That is an excellent question. > > > > > > The purpose of this change is to ensure that firmware cannot > > > manipulate the internal state of the kernel. So as long as we can > > > ensure that this memory is not claimed by the kernel's memory > > > subsystem, we should be fine. > > > > > > Since this is an obvious debug feature, what we could do is reserve > > > this memory permanently in some way, and make the test take this into > > > account. > > > > Whilst it is a debug feature, I wonder if it gets shipped in production > > hardware. If not, could be we cynical and just drop the check if the > > relevant config option is enabled? > > > > Perhaps just don't release the EFI_LOADER_DATA for other use? (if > > this option is enabled only) > > > > > > > > Do you have a full stack trace? How early does this run? > > > > For the place where it first occurs, ie the trace above, the acpi_reallocate_root_table() is > > the call from acpi_early_init() from start_kernel(). > > > > We hit the table a lot during later calls though and hence would run into the > > same problem. > > > > Could you try the patch below? Since the memory holding the tables is > already memblock_reserve()d, we can just mark it NOMAP, and permit r/o > remapping of NOMAP regions. Looks good. Thanks. Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> > > diff --git a/arch/arm64/kernel/acpi.c b/arch/arm64/kernel/acpi.c > index a85174d05473..84da869c5ac4 100644 > --- a/arch/arm64/kernel/acpi.c > +++ b/arch/arm64/kernel/acpi.c > @@ -298,8 +298,11 @@ void __iomem > *acpi_os_ioremap(acpi_physical_address phys, acpi_size size) > case EFI_BOOT_SERVICES_DATA: > case EFI_CONVENTIONAL_MEMORY: > case EFI_PERSISTENT_MEMORY: > - pr_warn(FW_BUG "requested region covers kernel > memory @ %pa\n", &phys); > - return NULL; > + if (memblock_is_map_memory(phys)) { > + pr_warn(FW_BUG "requested region > covers kernel memory @ %pa\n", &phys); > + return NULL; > + } > + fallthrough; > > case EFI_RUNTIME_SERVICES_CODE: > /* > @@ -388,3 +391,8 @@ int apei_claim_sea(struct pt_regs *regs) > > return err; > } > + > +void arch_reserve_mem_area(acpi_physical_address addr, size_t size) > +{ > + memblock_mark_nomap(addr, size); > +} > diff --git a/include/linux/acpi.h b/include/linux/acpi.h > index 1e4cdc6c7ae2..64ae25c59d55 100644 > --- a/include/linux/acpi.h > +++ b/include/linux/acpi.h > @@ -958,7 +958,7 @@ void acpi_os_set_prepare_extended_sleep(int > (*func)(u8 sleep_state, > acpi_status acpi_os_prepare_extended_sleep(u8 sleep_state, > u32 val_a, u32 val_b); > > -#ifdef CONFIG_X86 > +#ifndef CONFIG_IA64 > void arch_reserve_mem_area(acpi_physical_address addr, size_t size); > #else > static inline void arch_reserve_mem_area(acpi_physical_address addr,
Hi Jonathan, On Wed, Sep 30, 2020 at 10:27:22AM +0100, Jonathan Cameron wrote: > On Tue, 29 Sep 2020 11:29:48 +0200 Ard Biesheuvel <ardb@kernel.org> wrote: > > Could you try the patch below? Since the memory holding the tables is > > already memblock_reserve()d, we can just mark it NOMAP, and permit r/o > > remapping of NOMAP regions. > > Looks good. Thanks. > > Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> > Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Could you please try the updated patch that Ard posted. There are a few minor differences: https://lore.kernel.org/linux-acpi/20200929132522.18067-1-ardb@kernel.org/ Thanks. -- Catalin
diff --git a/arch/arm64/include/asm/acpi.h b/arch/arm64/include/asm/acpi.h index a45366c3909b..bd68e1b7f29f 100644 --- a/arch/arm64/include/asm/acpi.h +++ b/arch/arm64/include/asm/acpi.h @@ -47,20 +47,7 @@ pgprot_t __acpi_get_mem_attribute(phys_addr_t addr); /* ACPI table mapping after acpi_permanent_mmap is set */ -static inline void __iomem *acpi_os_ioremap(acpi_physical_address phys, - acpi_size size) -{ - /* For normal memory we already have a cacheable mapping. */ - if (memblock_is_map_memory(phys)) - return (void __iomem *)__phys_to_virt(phys); - - /* - * We should still honor the memory's attribute here because - * crash dump kernel possibly excludes some ACPI (reclaim) - * regions from memblock list. - */ - return __ioremap(phys, size, __acpi_get_mem_attribute(phys)); -} +void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size); #define acpi_os_ioremap acpi_os_ioremap typedef u64 phys_cpuid_t; diff --git a/arch/arm64/kernel/acpi.c b/arch/arm64/kernel/acpi.c index a7586a4db142..01b861e225b0 100644 --- a/arch/arm64/kernel/acpi.c +++ b/arch/arm64/kernel/acpi.c @@ -261,6 +261,72 @@ pgprot_t __acpi_get_mem_attribute(phys_addr_t addr) return __pgprot(PROT_DEVICE_nGnRnE); } +void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size) +{ + efi_memory_desc_t *md, *region = NULL; + pgprot_t prot; + + if (WARN_ON_ONCE(!efi_enabled(EFI_MEMMAP))) + return NULL; + + for_each_efi_memory_desc(md) { + u64 end = md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT); + + if (phys < md->phys_addr || phys >= end) + continue; + + if (phys + size > end) { + pr_warn(FW_BUG "requested region covers multiple EFI memory regions\n"); + return NULL; + } + region = md; + break; + } + + /* + * It is fine for AML to remap regions that are not represented in the + * EFI memory map at all, as it only describes normal memory, and MMIO + * regions that require a virtual mapping to make them accessible to + * the EFI runtime services. + */ + prot = __pgprot(PROT_DEVICE_nGnRnE); + if (region) { + switch (region->type) { + case EFI_LOADER_CODE: + case EFI_LOADER_DATA: + case EFI_BOOT_SERVICES_CODE: + case EFI_BOOT_SERVICES_DATA: + case EFI_CONVENTIONAL_MEMORY: + case EFI_PERSISTENT_MEMORY: + pr_warn(FW_BUG "requested region covers kernel memory @ %pa\n", &phys); + return NULL; + + case EFI_ACPI_RECLAIM_MEMORY: + /* + * ACPI reclaim memory is used to pass firmware tables + * and other data that is intended for consumption by + * the OS only, which may decide it wants to reclaim + * that memory and use it for something else. We never + * do that, but we usually add it to the linear map + * anyway, in which case we should use the existing + * mapping. + */ + if (memblock_is_map_memory(phys)) + return (void __iomem *)__phys_to_virt(phys); + /* fall through */ + + default: + if (region->attribute & EFI_MEMORY_WB) + prot = PAGE_KERNEL; + else if (region->attribute & EFI_MEMORY_WT) + prot = __pgprot(PROT_NORMAL_WT); + else if (region->attribute & EFI_MEMORY_WC) + prot = __pgprot(PROT_NORMAL_NC); + } + } + return __ioremap(phys, size, prot); +} + /* * Claim Synchronous External Aborts as a firmware first notification. *
AML uses SystemMemory opregions to allow AML handlers to access MMIO registers of, e.g., GPIO controllers, or access reserved regions of memory that are owned by the firmware. Currently, we also allow AML access to memory that is owned by the kernel and mapped via the linear region, which does not seem to be supported by a valid use case, and exposes the kernel's internal state to AML methods that may be buggy and exploitable. On arm64, ACPI support requires booting in EFI mode, and so we can cross reference the requested region against the EFI memory map, rather than just do a minimal check on the first page. So let's only permit regions to be remapped by the ACPI core if - they don't appear in the EFI memory map at all (which is the case for most MMIO), or - they are covered by a single region in the EFI memory map, which is not of a type that describes memory that is given to the kernel at boot. Reported-by: Jason A. Donenfeld <Jason@zx2c4.com> Signed-off-by: Ard Biesheuvel <ardb@kernel.org> --- arch/arm64/include/asm/acpi.h | 15 +---- arch/arm64/kernel/acpi.c | 66 ++++++++++++++++++++ 2 files changed, 67 insertions(+), 14 deletions(-)