[v1] EmbeddedPkg/MmcDxe: Align the ExtCSD buffer

[edk2] [PATCH] EmbeddedPkg/MmcDxe: Align the ExtCSD buffer

Posted by Jun Nie 6 years, 10 months ago

ExtCSD structure may be read via DMA. So align it to
page to avoid data corruption.

Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Jun Nie <jun.nie@linaro.org>
---
 EmbeddedPkg/Universal/MmcDxe/Mmc.h               |  2 +-
 EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c | 11 ++++++++---
 2 files changed, 9 insertions(+), 4 deletions(-)

diff --git a/EmbeddedPkg/Universal/MmcDxe/Mmc.h b/EmbeddedPkg/Universal/MmcDxe/Mmc.h
index 8a7d5a3..f3e56ff 100644
--- a/EmbeddedPkg/Universal/MmcDxe/Mmc.h
+++ b/EmbeddedPkg/Universal/MmcDxe/Mmc.h
@@ -319,7 +319,7 @@ typedef struct  {
   OCR       OCRData;
   CID       CIDData;
   CSD       CSDData;
-  ECSD      ECSDData;                         // MMC V4 extended card specific
+  ECSD      *ECSDData;                         // MMC V4 extended card specific
 } CARD_INFO;
 
 typedef struct _MMC_HOST_INSTANCE {
diff --git a/EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c b/EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c
index c28207e..6bb65c3 100644
--- a/EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c
+++ b/EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c
@@ -13,6 +13,7 @@
 **/
 
 #include <Library/BaseMemoryLib.h>
+#include <Library/MemoryAllocationLib.h>
 #include <Library/TimerLib.h>
 
 #include "Mmc.h"
@@ -210,12 +211,16 @@ EmmcIdentificationMode (
   }
 
   // Fetch ECSD
+  MmcHostInstance->CardInfo.ECSDData = AllocatePages (1);
+  if (MmcHostInstance->CardInfo.ECSDData == NULL) {
+    return EFI_BUFFER_TOO_SMALL;
+  }
   Status = Host->SendCommand (Host, MMC_CMD8, 0);
   if (EFI_ERROR (Status)) {
     DEBUG ((EFI_D_ERROR, "EmmcIdentificationMode(): ECSD fetch error, Status=%r.\n", Status));
   }
 
-  Status = Host->ReadBlockData (Host, 0, 512, (UINT32 *)&(MmcHostInstance->CardInfo.ECSDData));
+  Status = Host->ReadBlockData (Host, 0, 512, (UINT32 *)MmcHostInstance->CardInfo.ECSDData);
   if (EFI_ERROR (Status)) {
     DEBUG ((EFI_D_ERROR, "EmmcIdentificationMode(): ECSD read error, Status=%r.\n", Status));
     return Status;
@@ -237,7 +242,7 @@ EmmcIdentificationMode (
   Media->LogicalBlocksPerPhysicalBlock = 1;
   Media->IoAlign = 4;
   // Compute last block using bits [215:212] of the ECSD
-  Media->LastBlock = MmcHostInstance->CardInfo.ECSDData.SECTOR_COUNT - 1; // eMMC isn't supposed to report this for
+  Media->LastBlock = MmcHostInstance->CardInfo.ECSDData->SECTOR_COUNT - 1; // eMMC isn't supposed to report this for
   // Cards <2GB in size, but the model does.
 
   // Setup card type
@@ -258,7 +263,7 @@ InitializeEmmcDevice (
   UINT32     TimingMode[4] = {EMMCHS52DDR1V2, EMMCHS52DDR1V8, EMMCHS52, EMMCHS26};
 
   Host  = MmcHostInstance->MmcHost;
-  ECSDData = &MmcHostInstance->CardInfo.ECSDData;
+  ECSDData = MmcHostInstance->CardInfo.ECSDData;
   if (ECSDData->DEVICE_TYPE == EMMCBACKWARD)
     return EFI_SUCCESS;
 
-- 
1.9.1

_______________________________________________
edk2-devel mailing list
edk2-devel@lists.01.org
https://lists.01.org/mailman/listinfo/edk2-devel

Re: [edk2] [PATCH] EmbeddedPkg/MmcDxe: Align the ExtCSD buffer

Posted by Leif Lindholm 6 years, 10 months ago

On Thu, Jun 29, 2017 at 05:03:50PM +0800, Jun Nie wrote:
> ExtCSD structure may be read via DMA. So align it to
> page to avoid data corruption.

So, this is possibly a valid thing to do, but this is not what you
said in the last version. The last version said you needed 64-bit
alignment, and this is already guaranteed by all of the allocation
functions.

> Contributed-under: TianoCore Contribution Agreement 1.0
> Signed-off-by: Jun Nie <jun.nie@linaro.org>
> ---
>  EmbeddedPkg/Universal/MmcDxe/Mmc.h               |  2 +-
>  EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c | 11 ++++++++---
>  2 files changed, 9 insertions(+), 4 deletions(-)
> 
> diff --git a/EmbeddedPkg/Universal/MmcDxe/Mmc.h b/EmbeddedPkg/Universal/MmcDxe/Mmc.h
> index 8a7d5a3..f3e56ff 100644
> --- a/EmbeddedPkg/Universal/MmcDxe/Mmc.h
> +++ b/EmbeddedPkg/Universal/MmcDxe/Mmc.h
> @@ -319,7 +319,7 @@ typedef struct  {
>    OCR       OCRData;
>    CID       CIDData;
>    CSD       CSDData;
> -  ECSD      ECSDData;                         // MMC V4 extended card specific
> +  ECSD      *ECSDData;                         // MMC V4 extended card specific
>  } CARD_INFO;
>  
>  typedef struct _MMC_HOST_INSTANCE {
> diff --git a/EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c b/EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c
> index c28207e..6bb65c3 100644
> --- a/EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c
> +++ b/EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c
> @@ -13,6 +13,7 @@
>  **/
>  
>  #include <Library/BaseMemoryLib.h>
> +#include <Library/MemoryAllocationLib.h>
>  #include <Library/TimerLib.h>
>  
>  #include "Mmc.h"
> @@ -210,12 +211,16 @@ EmmcIdentificationMode (
>    }
>  
>    // Fetch ECSD
> +  MmcHostInstance->CardInfo.ECSDData = AllocatePages (1);

No magic numbers please.
Suggest EFI_SIZE_TO_PAGES (sizeof (ECSD));
Or AllocatePool (EFI_SIZE_TO_PAGES (sizeof (ECSD)));

I'm not seeing a balancing FreePages call in this patch.
My guess would be that DestroyMmcHostInstance() would be the
correct location for it.

> +  if (MmcHostInstance->CardInfo.ECSDData == NULL) {
> +    return EFI_BUFFER_TOO_SMALL;

I think EFI_OUT_OF_RESOURCES would be more appropriate.

/
    Leif

> +  }
>    Status = Host->SendCommand (Host, MMC_CMD8, 0);
>    if (EFI_ERROR (Status)) {
>      DEBUG ((EFI_D_ERROR, "EmmcIdentificationMode(): ECSD fetch error, Status=%r.\n", Status));
>    }
>  
> -  Status = Host->ReadBlockData (Host, 0, 512, (UINT32 *)&(MmcHostInstance->CardInfo.ECSDData));
> +  Status = Host->ReadBlockData (Host, 0, 512, (UINT32 *)MmcHostInstance->CardInfo.ECSDData);
>    if (EFI_ERROR (Status)) {
>      DEBUG ((EFI_D_ERROR, "EmmcIdentificationMode(): ECSD read error, Status=%r.\n", Status));
>      return Status;
> @@ -237,7 +242,7 @@ EmmcIdentificationMode (
>    Media->LogicalBlocksPerPhysicalBlock = 1;
>    Media->IoAlign = 4;
>    // Compute last block using bits [215:212] of the ECSD
> -  Media->LastBlock = MmcHostInstance->CardInfo.ECSDData.SECTOR_COUNT - 1; // eMMC isn't supposed to report this for
> +  Media->LastBlock = MmcHostInstance->CardInfo.ECSDData->SECTOR_COUNT - 1; // eMMC isn't supposed to report this for
>    // Cards <2GB in size, but the model does.
>  
>    // Setup card type
> @@ -258,7 +263,7 @@ InitializeEmmcDevice (
>    UINT32     TimingMode[4] = {EMMCHS52DDR1V2, EMMCHS52DDR1V8, EMMCHS52, EMMCHS26};
>  
>    Host  = MmcHostInstance->MmcHost;
> -  ECSDData = &MmcHostInstance->CardInfo.ECSDData;
> +  ECSDData = MmcHostInstance->CardInfo.ECSDData;
>    if (ECSDData->DEVICE_TYPE == EMMCBACKWARD)
>      return EFI_SUCCESS;
>  
> -- 
> 1.9.1
> 
_______________________________________________
edk2-devel mailing list
edk2-devel@lists.01.org
https://lists.01.org/mailman/listinfo/edk2-devel

Re: [edk2] [PATCH] EmbeddedPkg/MmcDxe: Align the ExtCSD buffer

Posted by Jun Nie 6 years, 10 months ago

2017-06-29 19:11 GMT+08:00 Leif Lindholm <leif.lindholm@linaro.org>:
> On Thu, Jun 29, 2017 at 05:03:50PM +0800, Jun Nie wrote:
>> ExtCSD structure may be read via DMA. So align it to
>> page to avoid data corruption.
>
> So, this is possibly a valid thing to do, but this is not what you
> said in the last version. The last version said you needed 64-bit
> alignment, and this is already guaranteed by all of the allocation
> functions.

Yes, I need 64 bytes alignment. page alignment is better, right? :-)

>
>> Contributed-under: TianoCore Contribution Agreement 1.0
>> Signed-off-by: Jun Nie <jun.nie@linaro.org>
>> ---
>>  EmbeddedPkg/Universal/MmcDxe/Mmc.h               |  2 +-
>>  EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c | 11 ++++++++---
>>  2 files changed, 9 insertions(+), 4 deletions(-)
>>
>> diff --git a/EmbeddedPkg/Universal/MmcDxe/Mmc.h b/EmbeddedPkg/Universal/MmcDxe/Mmc.h
>> index 8a7d5a3..f3e56ff 100644
>> --- a/EmbeddedPkg/Universal/MmcDxe/Mmc.h
>> +++ b/EmbeddedPkg/Universal/MmcDxe/Mmc.h
>> @@ -319,7 +319,7 @@ typedef struct  {
>>    OCR       OCRData;
>>    CID       CIDData;
>>    CSD       CSDData;
>> -  ECSD      ECSDData;                         // MMC V4 extended card specific
>> +  ECSD      *ECSDData;                         // MMC V4 extended card specific
>>  } CARD_INFO;
>>
>>  typedef struct _MMC_HOST_INSTANCE {
>> diff --git a/EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c b/EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c
>> index c28207e..6bb65c3 100644
>> --- a/EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c
>> +++ b/EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c
>> @@ -13,6 +13,7 @@
>>  **/
>>
>>  #include <Library/BaseMemoryLib.h>
>> +#include <Library/MemoryAllocationLib.h>
>>  #include <Library/TimerLib.h>
>>
>>  #include "Mmc.h"
>> @@ -210,12 +211,16 @@ EmmcIdentificationMode (
>>    }
>>
>>    // Fetch ECSD
>> +  MmcHostInstance->CardInfo.ECSDData = AllocatePages (1);
>
> No magic numbers please.
> Suggest EFI_SIZE_TO_PAGES (sizeof (ECSD));
> Or AllocatePool (EFI_SIZE_TO_PAGES (sizeof (ECSD)));

You are right.

>
> I'm not seeing a balancing FreePages call in this patch.
> My guess would be that DestroyMmcHostInstance() would be the
> correct location for it.

Add FreePages shall be more clear. Will do in next version.

>
>> +  if (MmcHostInstance->CardInfo.ECSDData == NULL) {
>> +    return EFI_BUFFER_TOO_SMALL;
>
> I think EFI_OUT_OF_RESOURCES would be more appropriate.

Will do.

>
> /
>     Leif
>
>> +  }
>>    Status = Host->SendCommand (Host, MMC_CMD8, 0);
>>    if (EFI_ERROR (Status)) {
>>      DEBUG ((EFI_D_ERROR, "EmmcIdentificationMode(): ECSD fetch error, Status=%r.\n", Status));
>>    }
>>
>> -  Status = Host->ReadBlockData (Host, 0, 512, (UINT32 *)&(MmcHostInstance->CardInfo.ECSDData));
>> +  Status = Host->ReadBlockData (Host, 0, 512, (UINT32 *)MmcHostInstance->CardInfo.ECSDData);
>>    if (EFI_ERROR (Status)) {
>>      DEBUG ((EFI_D_ERROR, "EmmcIdentificationMode(): ECSD read error, Status=%r.\n", Status));
>>      return Status;
>> @@ -237,7 +242,7 @@ EmmcIdentificationMode (
>>    Media->LogicalBlocksPerPhysicalBlock = 1;
>>    Media->IoAlign = 4;
>>    // Compute last block using bits [215:212] of the ECSD
>> -  Media->LastBlock = MmcHostInstance->CardInfo.ECSDData.SECTOR_COUNT - 1; // eMMC isn't supposed to report this for
>> +  Media->LastBlock = MmcHostInstance->CardInfo.ECSDData->SECTOR_COUNT - 1; // eMMC isn't supposed to report this for
>>    // Cards <2GB in size, but the model does.
>>
>>    // Setup card type
>> @@ -258,7 +263,7 @@ InitializeEmmcDevice (
>>    UINT32     TimingMode[4] = {EMMCHS52DDR1V2, EMMCHS52DDR1V8, EMMCHS52, EMMCHS26};
>>
>>    Host  = MmcHostInstance->MmcHost;
>> -  ECSDData = &MmcHostInstance->CardInfo.ECSDData;
>> +  ECSDData = MmcHostInstance->CardInfo.ECSDData;
>>    if (ECSDData->DEVICE_TYPE == EMMCBACKWARD)
>>      return EFI_SUCCESS;
>>
>> --
>> 1.9.1
>>
_______________________________________________
edk2-devel mailing list
edk2-devel@lists.01.org
https://lists.01.org/mailman/listinfo/edk2-devel