On 7 June 2018 at 07:41, Zeng, Star <[email protected]> wrote: > Thanks, got the point. > > It seems vague that who to ensure the cache coherency. MMU? Caller of > UpdateCapsule? UpdateCapsule? Consumer of capsule data? >
Unfortunately, since the spec does not mention it at all, we cannot rely on the caller of UpdateCapsule() to ensure this. That would require a spec change, and break backward compatibility with older revisions. The main issue here is that ARM does not provide the means to clean the entire cache, the only architectural method is clean cacheline by virtual address. So that leaves: - UpdateCapsule - problematic because it may be called at runtime and the firmware has no means of translating the physical addresses - ResetSystem - same as above: it is a runtime service, and so it cannot rely on a mapping to exist for those physical addresses - SEC - lacks the information about where the capsule resides - CapsulePei - already extracts the information about the capsule address in memory, and can perform the cache maintenance right before consuming the data. So unless anyone has any other suggestions, I think this approach is the only feasible one. If there are any concerns about adding this for architectures, I can look into refactoring CapsulePei and add it only for ARM/AARCH64. Thanks, Ard. > -----Original Message----- > From: edk2-devel [mailto:[email protected]] On Behalf Of Ard > Biesheuvel > Sent: Thursday, June 7, 2018 12:50 PM > To: Zeng, Star <[email protected]> > Cc: Ni, Ruiyu <[email protected]>; [email protected]; Leif Lindholm > <[email protected]>; Yao, Jiewen <[email protected]>; Gao, Liming > <[email protected]>; Kinney, Michael D <[email protected]> > Subject: Re: [edk2] [PATCH] MdeModulePkg/CapsulePei: clean Dcache before > consuming capsule data > > On 7 June 2018 at 03:37, Zeng, Star <[email protected]> wrote: >> Hi Ard, >> >> The input parameter CapsuleHeaderArray of UpdateCapsule has the virtual >> address. >> > > It has the virtual address of the capsules yes. But how about the data > structures passed as the ScatterGatherList? > > >> CapsuleHeaderArray >> Virtual pointer to an array of virtual pointers to the capsules being >> passed into update capsule. Each capsules is assumed to stored in >> contiguous virtual memory. The capsules in the CapsuleHeaderArray must >> be the same capsules as the ScatterGatherList. The CapsuleHeaderArray >> must have the >> >> >> Thanks, >> Star >> -----Original Message----- >> From: edk2-devel [mailto:[email protected]] On Behalf Of >> Ard Biesheuvel >> Sent: Wednesday, June 6, 2018 8:10 PM >> To: [email protected] >> Cc: Ni, Ruiyu <[email protected]>; Ard Biesheuvel >> <[email protected]>; Gao, Liming <[email protected]>; Yao, >> Jiewen <[email protected]>; Leif Lindholm >> <[email protected]>; Kinney, Michael D >> <[email protected]>; Zeng, Star <[email protected]> >> Subject: Re: [edk2] [PATCH] MdeModulePkg/CapsulePei: clean Dcache >> before consuming capsule data >> >> On 6 June 2018 at 11:52, Ard Biesheuvel <[email protected]> wrote: >>> When capsule updates are staged for processing after a warm reboot, >>> they are copied into memory with the MMU and caches enabled. When the >>> capsule PEI gets around to coalescing the capsule, the MMU and caches >>> may still be disabled, and so on architectures where uncached >>> accesses are incoherent with the caches (such as ARM and AARCH64), we >>> may read stale data if we don't clean the caches to memory first. >>> >>> Contributed-under: TianoCore Contribution Agreement 1.1 >>> Signed-off-by: Ard Biesheuvel <[email protected]> >> >> Leif asked me to include a note why this cannot be done when >> UpdateCapsule() is called. >> >> >> """ >> Note that this cache maintenance cannot be done during the invocation of >> UpdateCapsule(), since the ScatterGatherList structures are only identified >> by physical address, and at runtime, the firmware doesn't know whether and >> where this memory is mapped, and cache maintenance requires a virtual >> address. >> """ >> >>> --- >>> MdeModulePkg/Universal/CapsulePei/CapsulePei.inf | 1 + >>> MdeModulePkg/Universal/CapsulePei/Common/CapsuleCoalesce.c | 4 ++++ >>> 2 files changed, 5 insertions(+) >>> >>> diff --git a/MdeModulePkg/Universal/CapsulePei/CapsulePei.inf >>> b/MdeModulePkg/Universal/CapsulePei/CapsulePei.inf >>> index c54bc21a95a8..594e110d1f8a 100644 >>> --- a/MdeModulePkg/Universal/CapsulePei/CapsulePei.inf >>> +++ b/MdeModulePkg/Universal/CapsulePei/CapsulePei.inf >>> @@ -48,6 +48,7 @@ [Packages] >>> >>> [LibraryClasses] >>> BaseLib >>> + CacheMaintenanceLib >>> HobLib >>> BaseMemoryLib >>> PeiServicesLib >>> diff --git >>> a/MdeModulePkg/Universal/CapsulePei/Common/CapsuleCoalesce.c >>> b/MdeModulePkg/Universal/CapsulePei/Common/CapsuleCoalesce.c >>> index 3e7054cd38a9..1730f925adc5 100644 >>> --- a/MdeModulePkg/Universal/CapsulePei/Common/CapsuleCoalesce.c >>> +++ b/MdeModulePkg/Universal/CapsulePei/Common/CapsuleCoalesce.c >>> @@ -27,6 +27,7 @@ WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER >>> EXPRESS OR IMPLIED. >>> #include <Guid/CapsuleVendor.h> >>> >>> #include <Library/BaseMemoryLib.h> >>> +#include <Library/CacheMaintenanceLib.h> >>> #include <Library/DebugLib.h> >>> #include <Library/PrintLib.h> >>> #include <Library/BaseLib.h> >>> @@ -283,6 +284,9 @@ ValidateCapsuleByMemoryResource ( >>> DEBUG ((EFI_D_INFO, "Address(0x%lx) Size(0x%lx) in >>> MemoryResource[0x%x] - Start(0x%lx) Length(0x%lx)\n", >>> Address, Size, >>> Index, >>> MemoryResource[Index].PhysicalStart, >>> MemoryResource[Index].ResourceLength)); >>> + >>> + WriteBackDataCacheRange ((VOID *)(UINTN)Address, Size); >>> + >>> return TRUE; >>> } >>> } >>> -- >>> 2.17.0 >>> >> _______________________________________________ >> edk2-devel mailing list >> [email protected] >> https://lists.01.org/mailman/listinfo/edk2-devel > _______________________________________________ > edk2-devel mailing list > [email protected] > https://lists.01.org/mailman/listinfo/edk2-devel _______________________________________________ edk2-devel mailing list [email protected] https://lists.01.org/mailman/listinfo/edk2-devel
