[PATCH v2 0/4] ArmPkg: more ArmDmaLib fixes

[PATCH v2 0/4] ArmPkg: more ArmDmaLib fixes

[Ard Biesheuvel]

Using patch #4 in anger (which was the only patch in the v1 submission of
this series) revealed some issues in the ArmDmaLib code related to abuse
of the device address where a CPU physical address (which usually, but
not always translates 1:1 to a CPU virtual address) was dereferenced by
the CPU.

Patch #1 is an unrelated fix that drops the direct dependency on ArmLib.

Patch #2 adds a missing virtual-to-physical conversion (which is required
when DebugUncachedMemoryAllocationLib is in use)

The device address of a DMA mapping is not to be dereferenced by the CPU,
especially not after patch #4 is applied, which allows the bus view and
the CPU view to be shifted by a fixed offset. So patch #3 corrects some
wrong uses of the device address.

Patch #4 is the original patch, with some in-code comments added to further
clarify how the offset should be specified, and why ConvertToPhysicalAddress()
needs to be called before taking this offset into account.  

Ard Biesheuvel (4):
  ArmPkg/ArmDmaLib: use DMA buffer alignment from CPU arch protocol
  ArmPkg/ArmDmaLib: fix incorrect device address of double buffer
  ArmPkg/ArmDmaLib: clean up abuse of device address
  ArmPkg/ArmDmaLib: add support for fixed host-to-device DMA offset

 ArmPkg/ArmPkg.dec                      |  8 +++
 ArmPkg/Library/ArmDmaLib/ArmDmaLib.c   | 51 +++++++++++++-------
 ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf |  3 +-
 3 files changed, 42 insertions(+), 20 deletions(-)

-- 
2.7.4

[PATCH v2 1/4] ArmPkg/ArmDmaLib: use DMA buffer alignment from CPU arch protocol

[Ard Biesheuvel]

Instead of depending on ArmLib to retrieve the CWG directly, use
the DMA buffer alignment exposed by the CPU arch protocol. This
removes our dependency on ArmLib, which makes the library a bit
more architecture independent.

While we're in there, rename gCpu to mCpu to better reflect its
local scope, and reflow some lines that we're modifying anyway.

Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
---
 ArmPkg/Library/ArmDmaLib/ArmDmaLib.c   | 18 ++++++++----------
 ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf |  2 --
 2 files changed, 8 insertions(+), 12 deletions(-)

diff --git a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
index d48d6ff6dbbb..03fd9f3278e6 100644
--- a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
+++ b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
@@ -22,7 +22,6 @@
 #include <Library/UncachedMemoryAllocationLib.h>
 #include <Library/IoLib.h>
 #include <Library/BaseMemoryLib.h>
-#include <Library/ArmLib.h>
 
 #include <Protocol/Cpu.h>
 
@@ -36,8 +35,7 @@ typedef struct {
 
 
 
-EFI_CPU_ARCH_PROTOCOL      *gCpu;
-UINTN                      gCacheAlignment = 0;
+STATIC EFI_CPU_ARCH_PROTOCOL      *mCpu;
 
 /**
   Provides the DMA controller-specific addresses needed to access system memory.
@@ -92,8 +90,8 @@ DmaMap (
 
   *Mapping = Map;
 
-  if ((((UINTN)HostAddress & (gCacheAlignment - 1)) != 0) ||
-      ((*NumberOfBytes & (gCacheAlignment - 1)) != 0)) {
+  if ((((UINTN)HostAddress & (mCpu->DmaBufferAlignment - 1)) != 0) ||
+      ((*NumberOfBytes & (mCpu->DmaBufferAlignment - 1)) != 0)) {
 
     // Get the cacheability of the region
     Status = gDS->GetMemorySpaceDescriptor (*DeviceAddress, &GcdDescriptor);
@@ -154,7 +152,8 @@ DmaMap (
     DEBUG_CODE_END ();
 
     // Flush the Data Cache (should not have any effect if the memory region is uncached)
-    gCpu->FlushDataCache (gCpu, *DeviceAddress, *NumberOfBytes, EfiCpuFlushTypeWriteBackInvalidate);
+    mCpu->FlushDataCache (mCpu, *DeviceAddress, *NumberOfBytes,
+            EfiCpuFlushTypeWriteBackInvalidate);
   }
 
   Map->HostAddress   = (UINTN)HostAddress;
@@ -211,7 +210,8 @@ DmaUnmap (
       //
       // Make sure we read buffer from uncached memory and not the cache
       //
-      gCpu->FlushDataCache (gCpu, Map->HostAddress, Map->NumberOfBytes, EfiCpuFlushTypeInvalidate);
+      mCpu->FlushDataCache (mCpu, Map->HostAddress, Map->NumberOfBytes,
+              EfiCpuFlushTypeInvalidate);
     }
   }
 
@@ -311,11 +311,9 @@ ArmDmaLibConstructor (
   EFI_STATUS              Status;
 
   // Get the Cpu protocol for later use
-  Status = gBS->LocateProtocol (&gEfiCpuArchProtocolGuid, NULL, (VOID **)&gCpu);
+  Status = gBS->LocateProtocol (&gEfiCpuArchProtocolGuid, NULL, (VOID **)&mCpu);
   ASSERT_EFI_ERROR(Status);
 
-  gCacheAlignment = ArmCacheWritebackGranule ();
-
   return Status;
 }
 
diff --git a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf
index 95c13006eaac..31de3cfd828c 100644
--- a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf
+++ b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf
@@ -37,8 +37,6 @@ [LibraryClasses]
   UncachedMemoryAllocationLib
   IoLib
   BaseMemoryLib
-  ArmLib
-
 
 [Protocols]
   gEfiCpuArchProtocolGuid
-- 
2.7.4

Re: [PATCH v2 1/4] ArmPkg/ArmDmaLib: use DMA buffer alignment from CPU arch protocol

[Leif Lindholm]

On Sat, Nov 12, 2016 at 02:02:25PM +0100, Ard Biesheuvel wrote:
> Instead of depending on ArmLib to retrieve the CWG directly, use
> the DMA buffer alignment exposed by the CPU arch protocol. This
> removes our dependency on ArmLib, which makes the library a bit
> more architecture independent.
> 
> While we're in there, rename gCpu to mCpu to better reflect its
> local scope, and reflow some lines that we're modifying anyway.
> 
> Contributed-under: TianoCore Contribution Agreement 1.0
> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>

Reviewed-by: Leif Lindholm <leif.lindholm@linaro.org>

> ---
>  ArmPkg/Library/ArmDmaLib/ArmDmaLib.c   | 18 ++++++++----------
>  ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf |  2 --
>  2 files changed, 8 insertions(+), 12 deletions(-)
> 
> diff --git a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
> index d48d6ff6dbbb..03fd9f3278e6 100644
> --- a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
> +++ b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
> @@ -22,7 +22,6 @@
>  #include <Library/UncachedMemoryAllocationLib.h>
>  #include <Library/IoLib.h>
>  #include <Library/BaseMemoryLib.h>
> -#include <Library/ArmLib.h>
>  
>  #include <Protocol/Cpu.h>
>  
> @@ -36,8 +35,7 @@ typedef struct {
>  
>  
>  
> -EFI_CPU_ARCH_PROTOCOL      *gCpu;
> -UINTN                      gCacheAlignment = 0;
> +STATIC EFI_CPU_ARCH_PROTOCOL      *mCpu;
>  
>  /**
>    Provides the DMA controller-specific addresses needed to access system memory.
> @@ -92,8 +90,8 @@ DmaMap (
>  
>    *Mapping = Map;
>  
> -  if ((((UINTN)HostAddress & (gCacheAlignment - 1)) != 0) ||
> -      ((*NumberOfBytes & (gCacheAlignment - 1)) != 0)) {
> +  if ((((UINTN)HostAddress & (mCpu->DmaBufferAlignment - 1)) != 0) ||
> +      ((*NumberOfBytes & (mCpu->DmaBufferAlignment - 1)) != 0)) {
>  
>      // Get the cacheability of the region
>      Status = gDS->GetMemorySpaceDescriptor (*DeviceAddress, &GcdDescriptor);
> @@ -154,7 +152,8 @@ DmaMap (
>      DEBUG_CODE_END ();
>  
>      // Flush the Data Cache (should not have any effect if the memory region is uncached)
> -    gCpu->FlushDataCache (gCpu, *DeviceAddress, *NumberOfBytes, EfiCpuFlushTypeWriteBackInvalidate);
> +    mCpu->FlushDataCache (mCpu, *DeviceAddress, *NumberOfBytes,
> +            EfiCpuFlushTypeWriteBackInvalidate);
>    }
>  
>    Map->HostAddress   = (UINTN)HostAddress;
> @@ -211,7 +210,8 @@ DmaUnmap (
>        //
>        // Make sure we read buffer from uncached memory and not the cache
>        //
> -      gCpu->FlushDataCache (gCpu, Map->HostAddress, Map->NumberOfBytes, EfiCpuFlushTypeInvalidate);
> +      mCpu->FlushDataCache (mCpu, Map->HostAddress, Map->NumberOfBytes,
> +              EfiCpuFlushTypeInvalidate);
>      }
>    }
>  
> @@ -311,11 +311,9 @@ ArmDmaLibConstructor (
>    EFI_STATUS              Status;
>  
>    // Get the Cpu protocol for later use
> -  Status = gBS->LocateProtocol (&gEfiCpuArchProtocolGuid, NULL, (VOID **)&gCpu);
> +  Status = gBS->LocateProtocol (&gEfiCpuArchProtocolGuid, NULL, (VOID **)&mCpu);
>    ASSERT_EFI_ERROR(Status);
>  
> -  gCacheAlignment = ArmCacheWritebackGranule ();
> -
>    return Status;
>  }
>  
> diff --git a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf
> index 95c13006eaac..31de3cfd828c 100644
> --- a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf
> +++ b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf
> @@ -37,8 +37,6 @@ [LibraryClasses]
>    UncachedMemoryAllocationLib
>    IoLib
>    BaseMemoryLib
> -  ArmLib
> -
>  
>  [Protocols]
>    gEfiCpuArchProtocolGuid
> -- 
> 2.7.4
> 

[PATCH v2 2/4] ArmPkg/ArmDmaLib: fix incorrect device address of double buffer

[Ard Biesheuvel]

If double buffering is not required in DmaMap(), the returned device
address is passed through ConvertToPhysicalAddress () to convert the
host address (which in case of DebugUncachedMemoryAllocationLib is not
1:1 mapped) to a physical address, which is what a device would expect
to be able to perform DMA.

By the same reasoning, a double buffer allocated using DmaAllocateBuffer ()
should be converted in the same way, considering that the buffer is allocated
using UncachedAllocatePages (), to which the above equally applies.

So add the missing ConvertToPhysicalAddress () invocation.

Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
---
 ArmPkg/Library/ArmDmaLib/ArmDmaLib.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
index 03fd9f3278e6..c2a44398d25a 100644
--- a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
+++ b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
@@ -127,7 +127,7 @@ DmaMap (
         CopyMem (Buffer, HostAddress, *NumberOfBytes);
       }
 
-      *DeviceAddress = (PHYSICAL_ADDRESS)(UINTN)Buffer;
+      *DeviceAddress = ConvertToPhysicalAddress ((UINTN)Buffer);
     } else {
       Map->DoubleBuffer  = FALSE;
     }
-- 
2.7.4

Re: [PATCH v2 2/4] ArmPkg/ArmDmaLib: fix incorrect device address of double buffer

[Leif Lindholm]

On Sat, Nov 12, 2016 at 02:02:26PM +0100, Ard Biesheuvel wrote:
> If double buffering is not required in DmaMap(), the returned device
> address is passed through ConvertToPhysicalAddress () to convert the
> host address (which in case of DebugUncachedMemoryAllocationLib is not
> 1:1 mapped) to a physical address, which is what a device would expect
> to be able to perform DMA.
> 
> By the same reasoning, a double buffer allocated using DmaAllocateBuffer ()
> should be converted in the same way, considering that the buffer is allocated
> using UncachedAllocatePages (), to which the above equally applies.
> 
> So add the missing ConvertToPhysicalAddress () invocation.
> 
> Contributed-under: TianoCore Contribution Agreement 1.0
> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>

Reviewed-by: Leif Lindholm <leif.lindholm@linaro.org>

> ---
>  ArmPkg/Library/ArmDmaLib/ArmDmaLib.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
> index 03fd9f3278e6..c2a44398d25a 100644
> --- a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
> +++ b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
> @@ -127,7 +127,7 @@ DmaMap (
>          CopyMem (Buffer, HostAddress, *NumberOfBytes);
>        }
>  
> -      *DeviceAddress = (PHYSICAL_ADDRESS)(UINTN)Buffer;
> +      *DeviceAddress = ConvertToPhysicalAddress ((UINTN)Buffer);
>      } else {
>        Map->DoubleBuffer  = FALSE;
>      }
> -- 
> 2.7.4
> 

[PATCH v2 3/4] ArmPkg/ArmDmaLib: clean up abuse of device address

[Ard Biesheuvel]

In preparation of adding support to ArmDmalib for DMA bus masters whose
view of memory is offset by a constant compared to the CPU's view, clean
up some abuse of the device address.

The device address is not defined in terms of the CPU's address space,
and so it should not be used in CopyMem () or cache maintenance operations
that require a valid mapping. This not only applies to the above use case,
but also to the DebugUncachedMemoryAllocationLib that unmaps the
primary, cached mapping of an allocation, and returns a host address
which is an uncached alias offset by a constant.

Since we should never access the device address from the CPU, there is
no need to record it in the MAPINFO struct. Instead, record the buffer
address in case of double buffering, since we do need to copy the contents
(in case of a bus master write) and free the buffer (in all cases) when
DmaUnmap() is called.

Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
---
 ArmPkg/Library/ArmDmaLib/ArmDmaLib.c | 15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)

diff --git a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
index c2a44398d25a..7321388de63e 100644
--- a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
+++ b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
@@ -27,7 +27,7 @@
 
 typedef struct {
   EFI_PHYSICAL_ADDRESS      HostAddress;
-  EFI_PHYSICAL_ADDRESS      DeviceAddress;
+  VOID                      *BufferAddress;
   UINTN                     NumberOfBytes;
   DMA_MAP_OPERATION         Operation;
   BOOLEAN                   DoubleBuffer;
@@ -94,7 +94,7 @@ DmaMap (
       ((*NumberOfBytes & (mCpu->DmaBufferAlignment - 1)) != 0)) {
 
     // Get the cacheability of the region
-    Status = gDS->GetMemorySpaceDescriptor (*DeviceAddress, &GcdDescriptor);
+    Status = gDS->GetMemorySpaceDescriptor ((UINTN)HostAddress, &GcdDescriptor);
     if (EFI_ERROR(Status)) {
       return Status;
     }
@@ -128,6 +128,7 @@ DmaMap (
       }
 
       *DeviceAddress = ConvertToPhysicalAddress ((UINTN)Buffer);
+      Map->BufferAddress = Buffer;
     } else {
       Map->DoubleBuffer  = FALSE;
     }
@@ -143,7 +144,7 @@ DmaMap (
     // So duplicate the check here when running in DEBUG mode, just to assert
     // that we are not trying to create a consistent mapping for cached memory.
     //
-    Status = gDS->GetMemorySpaceDescriptor (*DeviceAddress, &GcdDescriptor);
+    Status = gDS->GetMemorySpaceDescriptor ((UINTN)HostAddress, &GcdDescriptor);
     ASSERT_EFI_ERROR(Status);
 
     ASSERT (Operation != MapOperationBusMasterCommonBuffer ||
@@ -152,12 +153,11 @@ DmaMap (
     DEBUG_CODE_END ();
 
     // Flush the Data Cache (should not have any effect if the memory region is uncached)
-    mCpu->FlushDataCache (mCpu, *DeviceAddress, *NumberOfBytes,
+    mCpu->FlushDataCache (mCpu, (UINTN)HostAddress, *NumberOfBytes,
             EfiCpuFlushTypeWriteBackInvalidate);
   }
 
   Map->HostAddress   = (UINTN)HostAddress;
-  Map->DeviceAddress = *DeviceAddress;
   Map->NumberOfBytes = *NumberOfBytes;
   Map->Operation     = Operation;
 
@@ -200,10 +200,11 @@ DmaUnmap (
     if (Map->Operation == MapOperationBusMasterCommonBuffer) {
       Status = EFI_INVALID_PARAMETER;
     } else if (Map->Operation == MapOperationBusMasterWrite) {
-      CopyMem ((VOID *)(UINTN)Map->HostAddress, (VOID *)(UINTN)Map->DeviceAddress, Map->NumberOfBytes);
+      CopyMem ((VOID *)(UINTN)Map->HostAddress, Map->BufferAddress,
+        Map->NumberOfBytes);
     }
 
-    DmaFreeBuffer (EFI_SIZE_TO_PAGES (Map->NumberOfBytes), (VOID *)(UINTN)Map->DeviceAddress);
+    DmaFreeBuffer (EFI_SIZE_TO_PAGES (Map->NumberOfBytes), Map->BufferAddress);
 
   } else {
     if (Map->Operation == MapOperationBusMasterWrite) {
-- 
2.7.4

Re: [PATCH v2 3/4] ArmPkg/ArmDmaLib: clean up abuse of device address

[Leif Lindholm]

On Sat, Nov 12, 2016 at 02:02:27PM +0100, Ard Biesheuvel wrote:
> In preparation of adding support to ArmDmalib for DMA bus masters whose
> view of memory is offset by a constant compared to the CPU's view, clean
> up some abuse of the device address.
> 
> The device address is not defined in terms of the CPU's address space,
> and so it should not be used in CopyMem () or cache maintenance operations
> that require a valid mapping. This not only applies to the above use case,
> but also to the DebugUncachedMemoryAllocationLib that unmaps the
> primary, cached mapping of an allocation, and returns a host address
> which is an uncached alias offset by a constant.
> 
> Since we should never access the device address from the CPU, there is
> no need to record it in the MAPINFO struct. Instead, record the buffer
> address in case of double buffering, since we do need to copy the contents
> (in case of a bus master write) and free the buffer (in all cases) when
> DmaUnmap() is called.
> 
> Contributed-under: TianoCore Contribution Agreement 1.0
> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>

For the fix itself:
Reviewed-by: Leif Lindholm <leif.lindholm@linaro.org>

However, can we wait for a few Tested-by:s to ensure this fix does not
reveal any companion bugs?

> ---
>  ArmPkg/Library/ArmDmaLib/ArmDmaLib.c | 15 ++++++++-------
>  1 file changed, 8 insertions(+), 7 deletions(-)
> 
> diff --git a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
> index c2a44398d25a..7321388de63e 100644
> --- a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
> +++ b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
> @@ -27,7 +27,7 @@
>  
>  typedef struct {
>    EFI_PHYSICAL_ADDRESS      HostAddress;
> -  EFI_PHYSICAL_ADDRESS      DeviceAddress;
> +  VOID                      *BufferAddress;
>    UINTN                     NumberOfBytes;
>    DMA_MAP_OPERATION         Operation;
>    BOOLEAN                   DoubleBuffer;
> @@ -94,7 +94,7 @@ DmaMap (
>        ((*NumberOfBytes & (mCpu->DmaBufferAlignment - 1)) != 0)) {
>  
>      // Get the cacheability of the region
> -    Status = gDS->GetMemorySpaceDescriptor (*DeviceAddress, &GcdDescriptor);
> +    Status = gDS->GetMemorySpaceDescriptor ((UINTN)HostAddress, &GcdDescriptor);
>      if (EFI_ERROR(Status)) {
>        return Status;
>      }
> @@ -128,6 +128,7 @@ DmaMap (
>        }
>  
>        *DeviceAddress = ConvertToPhysicalAddress ((UINTN)Buffer);
> +      Map->BufferAddress = Buffer;
>      } else {
>        Map->DoubleBuffer  = FALSE;
>      }
> @@ -143,7 +144,7 @@ DmaMap (
>      // So duplicate the check here when running in DEBUG mode, just to assert
>      // that we are not trying to create a consistent mapping for cached memory.
>      //
> -    Status = gDS->GetMemorySpaceDescriptor (*DeviceAddress, &GcdDescriptor);
> +    Status = gDS->GetMemorySpaceDescriptor ((UINTN)HostAddress, &GcdDescriptor);
>      ASSERT_EFI_ERROR(Status);
>  
>      ASSERT (Operation != MapOperationBusMasterCommonBuffer ||
> @@ -152,12 +153,11 @@ DmaMap (
>      DEBUG_CODE_END ();
>  
>      // Flush the Data Cache (should not have any effect if the memory region is uncached)
> -    mCpu->FlushDataCache (mCpu, *DeviceAddress, *NumberOfBytes,
> +    mCpu->FlushDataCache (mCpu, (UINTN)HostAddress, *NumberOfBytes,
>              EfiCpuFlushTypeWriteBackInvalidate);
>    }
>  
>    Map->HostAddress   = (UINTN)HostAddress;
> -  Map->DeviceAddress = *DeviceAddress;
>    Map->NumberOfBytes = *NumberOfBytes;
>    Map->Operation     = Operation;
>  
> @@ -200,10 +200,11 @@ DmaUnmap (
>      if (Map->Operation == MapOperationBusMasterCommonBuffer) {
>        Status = EFI_INVALID_PARAMETER;
>      } else if (Map->Operation == MapOperationBusMasterWrite) {
> -      CopyMem ((VOID *)(UINTN)Map->HostAddress, (VOID *)(UINTN)Map->DeviceAddress, Map->NumberOfBytes);
> +      CopyMem ((VOID *)(UINTN)Map->HostAddress, Map->BufferAddress,
> +        Map->NumberOfBytes);
>      }
>  
> -    DmaFreeBuffer (EFI_SIZE_TO_PAGES (Map->NumberOfBytes), (VOID *)(UINTN)Map->DeviceAddress);
> +    DmaFreeBuffer (EFI_SIZE_TO_PAGES (Map->NumberOfBytes), Map->BufferAddress);
>  
>    } else {
>      if (Map->Operation == MapOperationBusMasterWrite) {
> -- 
> 2.7.4
> 

Re: [PATCH v2 3/4] ArmPkg/ArmDmaLib: clean up abuse of device address

[Ard Biesheuvel]

On 14 November 2016 at 16:16, Leif Lindholm <leif.lindholm@linaro.org> wrote:
> On Sat, Nov 12, 2016 at 02:02:27PM +0100, Ard Biesheuvel wrote:
>> In preparation of adding support to ArmDmalib for DMA bus masters whose
>> view of memory is offset by a constant compared to the CPU's view, clean
>> up some abuse of the device address.
>>
>> The device address is not defined in terms of the CPU's address space,
>> and so it should not be used in CopyMem () or cache maintenance operations
>> that require a valid mapping. This not only applies to the above use case,
>> but also to the DebugUncachedMemoryAllocationLib that unmaps the
>> primary, cached mapping of an allocation, and returns a host address
>> which is an uncached alias offset by a constant.
>>
>> Since we should never access the device address from the CPU, there is
>> no need to record it in the MAPINFO struct. Instead, record the buffer
>> address in case of double buffering, since we do need to copy the contents
>> (in case of a bus master write) and free the buffer (in all cases) when
>> DmaUnmap() is called.
>>
>> Contributed-under: TianoCore Contribution Agreement 1.0
>> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
>
> For the fix itself:
> Reviewed-by: Leif Lindholm <leif.lindholm@linaro.org>
>
> However, can we wait for a few Tested-by:s to ensure this fix does not
> reveal any companion bugs?
>

Perhaps, yes.

In case anyone is up to doing that, please find the branch here
https://git.linaro.org/people/ard.biesheuvel/uefi-next.git/log/?h=armdmalib-offset

However, given that the split CPU/bus master view is introduced in the
next patch, the only use case where the device address differs from
the host address is when using the DebugUncachedMemoryAllocationLib,
which is currently broken AFAICT (it attempts to unmap the linear
mapping of the allocation by setting the memory attributes to '0',
which triggers an assert in the ArmPkg MMU code)

Re: [PATCH v2 3/4] ArmPkg/ArmDmaLib: clean up abuse of device address

[Ryan Harkin]

On 15 November 2016 at 09:19, Ard Biesheuvel <ard.biesheuvel@linaro.org> wrote:
> On 14 November 2016 at 16:16, Leif Lindholm <leif.lindholm@linaro.org> wrote:
>> On Sat, Nov 12, 2016 at 02:02:27PM +0100, Ard Biesheuvel wrote:
>>> In preparation of adding support to ArmDmalib for DMA bus masters whose
>>> view of memory is offset by a constant compared to the CPU's view, clean
>>> up some abuse of the device address.
>>>
>>> The device address is not defined in terms of the CPU's address space,
>>> and so it should not be used in CopyMem () or cache maintenance operations
>>> that require a valid mapping. This not only applies to the above use case,
>>> but also to the DebugUncachedMemoryAllocationLib that unmaps the
>>> primary, cached mapping of an allocation, and returns a host address
>>> which is an uncached alias offset by a constant.
>>>
>>> Since we should never access the device address from the CPU, there is
>>> no need to record it in the MAPINFO struct. Instead, record the buffer
>>> address in case of double buffering, since we do need to copy the contents
>>> (in case of a bus master write) and free the buffer (in all cases) when
>>> DmaUnmap() is called.
>>>
>>> Contributed-under: TianoCore Contribution Agreement 1.0
>>> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
>>
>> For the fix itself:
>> Reviewed-by: Leif Lindholm <leif.lindholm@linaro.org>
>>
>> However, can we wait for a few Tested-by:s to ensure this fix does not
>> reveal any companion bugs?
>>
>
> Perhaps, yes.
>
> In case anyone is up to doing that, please find the branch here
> https://git.linaro.org/people/ard.biesheuvel/uefi-next.git/log/?h=armdmalib-offset
>

I tested your branch on the usual victims (R0/1/2, FVP Foundation &
AEMv8 and TC2) and they all work fine for me.

Tested-by: Ryan Harkin <ryan.harkin@linaro.org>


> However, given that the split CPU/bus master view is introduced in the
> next patch, the only use case where the device address differs from
> the host address is when using the DebugUncachedMemoryAllocationLib,
> which is currently broken AFAICT (it attempts to unmap the linear
> mapping of the allocation by setting the memory attributes to '0',
> which triggers an assert in the ArmPkg MMU code)

Re: [PATCH v2 3/4] ArmPkg/ArmDmaLib: clean up abuse of device address

[Ard Biesheuvel]

On 15 November 2016 at 11:34, Ryan Harkin <ryan.harkin@linaro.org> wrote:
> On 15 November 2016 at 09:19, Ard Biesheuvel <ard.biesheuvel@linaro.org> wrote:
>> On 14 November 2016 at 16:16, Leif Lindholm <leif.lindholm@linaro.org> wrote:
>>> On Sat, Nov 12, 2016 at 02:02:27PM +0100, Ard Biesheuvel wrote:
>>>> In preparation of adding support to ArmDmalib for DMA bus masters whose
>>>> view of memory is offset by a constant compared to the CPU's view, clean
>>>> up some abuse of the device address.
>>>>
>>>> The device address is not defined in terms of the CPU's address space,
>>>> and so it should not be used in CopyMem () or cache maintenance operations
>>>> that require a valid mapping. This not only applies to the above use case,
>>>> but also to the DebugUncachedMemoryAllocationLib that unmaps the
>>>> primary, cached mapping of an allocation, and returns a host address
>>>> which is an uncached alias offset by a constant.
>>>>
>>>> Since we should never access the device address from the CPU, there is
>>>> no need to record it in the MAPINFO struct. Instead, record the buffer
>>>> address in case of double buffering, since we do need to copy the contents
>>>> (in case of a bus master write) and free the buffer (in all cases) when
>>>> DmaUnmap() is called.
>>>>
>>>> Contributed-under: TianoCore Contribution Agreement 1.0
>>>> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
>>>
>>> For the fix itself:
>>> Reviewed-by: Leif Lindholm <leif.lindholm@linaro.org>
>>>
>>> However, can we wait for a few Tested-by:s to ensure this fix does not
>>> reveal any companion bugs?
>>>
>>
>> Perhaps, yes.
>>
>> In case anyone is up to doing that, please find the branch here
>> https://git.linaro.org/people/ard.biesheuvel/uefi-next.git/log/?h=armdmalib-offset
>>
>
> I tested your branch on the usual victims (R0/1/2, FVP Foundation &
> AEMv8 and TC2) and they all work fine for me.
>
> Tested-by: Ryan Harkin <ryan.harkin@linaro.org>
>

Thanks!

Re: [PATCH v2 3/4] ArmPkg/ArmDmaLib: clean up abuse of device address

[Leif Lindholm]

On Tue, Nov 15, 2016 at 01:07:37PM +0000, Ard Biesheuvel wrote:
> On 15 November 2016 at 11:34, Ryan Harkin <ryan.harkin@linaro.org> wrote:
> >>>> Since we should never access the device address from the CPU, there is
> >>>> no need to record it in the MAPINFO struct. Instead, record the buffer
> >>>> address in case of double buffering, since we do need to copy the contents
> >>>> (in case of a bus master write) and free the buffer (in all cases) when
> >>>> DmaUnmap() is called.
> >>>>
> >>>> Contributed-under: TianoCore Contribution Agreement 1.0
> >>>> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
> >>>
> >>> For the fix itself:
> >>> Reviewed-by: Leif Lindholm <leif.lindholm@linaro.org>
> >>>
> >>> However, can we wait for a few Tested-by:s to ensure this fix does not
> >>> reveal any companion bugs?
> >>>
> >>
> >> Perhaps, yes.
> >>
> >> In case anyone is up to doing that, please find the branch here
> >> https://git.linaro.org/people/ard.biesheuvel/uefi-next.git/log/?h=armdmalib-offset
> >>
> >
> > I tested your branch on the usual victims (R0/1/2, FVP Foundation &
> > AEMv8 and TC2) and they all work fine for me.
> >
> > Tested-by: Ryan Harkin <ryan.harkin@linaro.org>

OK, so on the off-hand chance that this does break some non-upstream
drivers, I'll wait for further comments until I'm back from holiday,
28 November, before pushing this.

Additional Tested-by:s would still be appreciated.

Regards,

Leif

Re: [PATCH v2 3/4] ArmPkg/ArmDmaLib: clean up abuse of device address

[Leif Lindholm]

On 15 November 2016 at 18:01, Leif Lindholm <leif.lindholm@linaro.org> wrote:
>> >> In case anyone is up to doing that, please find the branch here
>> >> https://git.linaro.org/people/ard.biesheuvel/uefi-next.git/log/?h=armdmalib-offset
>> >
>> > I tested your branch on the usual victims (R0/1/2, FVP Foundation &
>> > AEMv8 and TC2) and they all work fine for me.
>> >
>> > Tested-by: Ryan Harkin <ryan.harkin@linaro.org>
>
> OK, so on the off-hand chance that this does break some non-upstream
> drivers, I'll wait for further comments until I'm back from holiday,
> 28 November, before pushing this.

Series pushed as 3571884..bfe3427 - thanks!

/
    Leif

[PATCH v2 4/4] ArmPkg/ArmDmaLib: add support for fixed host-to-device DMA offset

[Ard Biesheuvel]

Some devices, such as the Raspberry Pi3, have a fixed offset between memory
addresses as seen by the host and as seen by the other bus masters. So add
a new PCD that allows this fixed offset to be recorded, and to be used when
returning device addresses from the DmaLib mapping routines.

Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
---
 ArmPkg/ArmPkg.dec                      |  8 ++++++++
 ArmPkg/Library/ArmDmaLib/ArmDmaLib.c   | 20 ++++++++++++++++++--
 ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf |  1 +
 3 files changed, 27 insertions(+), 2 deletions(-)

diff --git a/ArmPkg/ArmPkg.dec b/ArmPkg/ArmPkg.dec
index 3cdb5da3d4f3..090ed9951366 100644
--- a/ArmPkg/ArmPkg.dec
+++ b/ArmPkg/ArmPkg.dec
@@ -134,6 +134,14 @@ [PcdsFixedAtBuild.common]
   gArmTokenSpaceGuid.PcdFdSize|0|UINT32|0x0000002C
   gArmTokenSpaceGuid.PcdFvSize|0|UINT32|0x0000002E
 
+  #
+  # Value to add to a host address to obtain a device address, using
+  # unsigned 64-bit integer arithmetic on both ARM and AArch64. This
+  # means we can rely on truncation on overflow to specify negative
+  # offsets.
+  #
+  gArmTokenSpaceGuid.PcdArmDmaDeviceOffset|0x0|UINT64|0x0000044
+
 [PcdsFixedAtBuild.common, PcdsPatchableInModule.common]
   gArmTokenSpaceGuid.PcdFdBaseAddress|0|UINT64|0x0000002B
   gArmTokenSpaceGuid.PcdFvBaseAddress|0|UINT64|0x0000002D
diff --git a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
index 7321388de63e..1de4a6e5b04f 100644
--- a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
+++ b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
@@ -37,6 +37,15 @@ typedef struct {
 
 STATIC EFI_CPU_ARCH_PROTOCOL      *mCpu;
 
+STATIC
+PHYSICAL_ADDRESS
+HostToDeviceAddress (
+  IN  PHYSICAL_ADDRESS  HostAddress
+  )
+{
+  return HostAddress + PcdGet64 (PcdArmDmaDeviceOffset);
+}
+
 /**
   Provides the DMA controller-specific addresses needed to access system memory.
 
@@ -80,7 +89,14 @@ DmaMap (
     return EFI_INVALID_PARAMETER;
   }
 
-  *DeviceAddress = ConvertToPhysicalAddress (HostAddress);
+  //
+  // The debug implementation of UncachedMemoryAllocationLib in ArmPkg returns
+  // a virtual uncached alias, and unmaps the cached ID mapping of the buffer,
+  // in order to catch inadvertent references to the cached mapping.
+  // Since HostToDeviceAddress () expects ID mapped input addresses, convert
+  // the host address to an ID mapped address first.
+  //
+  *DeviceAddress = HostToDeviceAddress (ConvertToPhysicalAddress (HostAddress));
 
   // Remember range so we can flush on the other side
   Map = AllocatePool (sizeof (MAP_INFO_INSTANCE));
@@ -127,7 +143,7 @@ DmaMap (
         CopyMem (Buffer, HostAddress, *NumberOfBytes);
       }
 
-      *DeviceAddress = ConvertToPhysicalAddress ((UINTN)Buffer);
+      *DeviceAddress = HostToDeviceAddress (ConvertToPhysicalAddress ((UINTN)Buffer));
       Map->BufferAddress = Buffer;
     } else {
       Map->DoubleBuffer  = FALSE;
diff --git a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf
index 31de3cfd828c..9b7dad114b18 100644
--- a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf
+++ b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf
@@ -44,6 +44,7 @@ [Protocols]
 [Guids]
 
 [Pcd]
+  gArmTokenSpaceGuid.PcdArmDmaDeviceOffset
 
 [Depex]
   gEfiCpuArchProtocolGuid
-- 
2.7.4

Re: [PATCH v2 4/4] ArmPkg/ArmDmaLib: add support for fixed host-to-device DMA offset

[Leif Lindholm]

On Sat, Nov 12, 2016 at 02:02:28PM +0100, Ard Biesheuvel wrote:
> Some devices, such as the Raspberry Pi3, have a fixed offset between memory
> addresses as seen by the host and as seen by the other bus masters. So add
> a new PCD that allows this fixed offset to be recorded, and to be used when
> returning device addresses from the DmaLib mapping routines.
> 
> Contributed-under: TianoCore Contribution Agreement 1.0
> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>

Much more Leif-friendly, thanks.
Reviewed-by: Leif Lindholm <leif.lindholm@linaro.org>

> ---
>  ArmPkg/ArmPkg.dec                      |  8 ++++++++
>  ArmPkg/Library/ArmDmaLib/ArmDmaLib.c   | 20 ++++++++++++++++++--
>  ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf |  1 +
>  3 files changed, 27 insertions(+), 2 deletions(-)
> 
> diff --git a/ArmPkg/ArmPkg.dec b/ArmPkg/ArmPkg.dec
> index 3cdb5da3d4f3..090ed9951366 100644
> --- a/ArmPkg/ArmPkg.dec
> +++ b/ArmPkg/ArmPkg.dec
> @@ -134,6 +134,14 @@ [PcdsFixedAtBuild.common]
>    gArmTokenSpaceGuid.PcdFdSize|0|UINT32|0x0000002C
>    gArmTokenSpaceGuid.PcdFvSize|0|UINT32|0x0000002E
>  
> +  #
> +  # Value to add to a host address to obtain a device address, using
> +  # unsigned 64-bit integer arithmetic on both ARM and AArch64. This
> +  # means we can rely on truncation on overflow to specify negative
> +  # offsets.
> +  #
> +  gArmTokenSpaceGuid.PcdArmDmaDeviceOffset|0x0|UINT64|0x0000044
> +
>  [PcdsFixedAtBuild.common, PcdsPatchableInModule.common]
>    gArmTokenSpaceGuid.PcdFdBaseAddress|0|UINT64|0x0000002B
>    gArmTokenSpaceGuid.PcdFvBaseAddress|0|UINT64|0x0000002D
> diff --git a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
> index 7321388de63e..1de4a6e5b04f 100644
> --- a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
> +++ b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.c
> @@ -37,6 +37,15 @@ typedef struct {
>  
>  STATIC EFI_CPU_ARCH_PROTOCOL      *mCpu;
>  
> +STATIC
> +PHYSICAL_ADDRESS
> +HostToDeviceAddress (
> +  IN  PHYSICAL_ADDRESS  HostAddress
> +  )
> +{
> +  return HostAddress + PcdGet64 (PcdArmDmaDeviceOffset);
> +}
> +
>  /**
>    Provides the DMA controller-specific addresses needed to access system memory.
>  
> @@ -80,7 +89,14 @@ DmaMap (
>      return EFI_INVALID_PARAMETER;
>    }
>  
> -  *DeviceAddress = ConvertToPhysicalAddress (HostAddress);
> +  //
> +  // The debug implementation of UncachedMemoryAllocationLib in ArmPkg returns
> +  // a virtual uncached alias, and unmaps the cached ID mapping of the buffer,
> +  // in order to catch inadvertent references to the cached mapping.
> +  // Since HostToDeviceAddress () expects ID mapped input addresses, convert
> +  // the host address to an ID mapped address first.
> +  //
> +  *DeviceAddress = HostToDeviceAddress (ConvertToPhysicalAddress (HostAddress));
>  
>    // Remember range so we can flush on the other side
>    Map = AllocatePool (sizeof (MAP_INFO_INSTANCE));
> @@ -127,7 +143,7 @@ DmaMap (
>          CopyMem (Buffer, HostAddress, *NumberOfBytes);
>        }
>  
> -      *DeviceAddress = ConvertToPhysicalAddress ((UINTN)Buffer);
> +      *DeviceAddress = HostToDeviceAddress (ConvertToPhysicalAddress ((UINTN)Buffer));
>        Map->BufferAddress = Buffer;
>      } else {
>        Map->DoubleBuffer  = FALSE;
> diff --git a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf
> index 31de3cfd828c..9b7dad114b18 100644
> --- a/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf
> +++ b/ArmPkg/Library/ArmDmaLib/ArmDmaLib.inf
> @@ -44,6 +44,7 @@ [Protocols]
>  [Guids]
>  
>  [Pcd]
> +  gArmTokenSpaceGuid.PcdArmDmaDeviceOffset
>  
>  [Depex]
>    gEfiCpuArchProtocolGuid
> -- 
> 2.7.4
>