Re: [MdeModulePkg/Library v1 1/1] MdeModulePkg/UefiBootManangerLib: Fix exception issue

[Ming Huang <ming.huang@linaro.org>]

On 3/18/2019 8:42 PM, Leif Lindholm wrote:
> +MdeModulePkg maintainers (you added MdePkg maintainers to cc)
> 
> This looks like an improvement to me.
> 
> Am I correct in guessing this behaviour refers to some specific corner
> case of a USB CDROM emulated from a BMC?

Yes, I found this issue with a USB CDROM emulated from a BMC.
I guess have the same symptom with physical USB CDROM.

Thanks

> 
> On Mon, Feb 25, 2019 at 05:10:52PM +0800, Ming Huang wrote:
>> The system environment: virtual-CDROM(USB interface) via BMC, insert a
>> iso file to CDROM, like ubuntu-18.04.1-server-arm64.iso, change CDROM
>> to first boot option.
>> With release version bios, disconnecting CDROM when boot to
>> "1 seconds left, Press Esc or F2 to enter Setup"
>> then system will get a exception.
>>
>> The root cause is the EFI_BLOCK_IO_PROTOCOL for UsbMass will be uninstalled
>> in this situation after print some transfer error. The status will be
>> invalid parameter. This line will get a exception for BlockIo not point
>> to right address:
>> AllocatePool (BlockIo->Media->BlockSize)
>> So, here need to judge the status not using ASSERT_EFI_ERROR.
>>
>> Contributed-under: TianoCore Contribution Agreement 1.1
>> Signed-off-by: Ming Huang <ming.huang@linaro.org>
>> ---
>>  MdeModulePkg/Library/UefiBootManagerLib/BmBoot.c | 4 +++-
>>  1 file changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/MdeModulePkg/Library/UefiBootManagerLib/BmBoot.c b/MdeModulePkg/Library/UefiBootManagerLib/BmBoot.c
>> index d5957db610d9..c2f1c651b02f 100644
>> --- a/MdeModulePkg/Library/UefiBootManagerLib/BmBoot.c
>> +++ b/MdeModulePkg/Library/UefiBootManagerLib/BmBoot.c
>> @@ -1068,7 +1068,9 @@ BmExpandMediaDevicePath (
>>    // Block IO read/write will success.
>>    //
>>    Status = gBS->HandleProtocol (Handle, &gEfiBlockIoProtocolGuid, (VOID **) &BlockIo);
>> -  ASSERT_EFI_ERROR (Status);
>> +  if (EFI_ERROR (Status)) {
> 
> It would still be worth including an ASSERT here, to let DEBUG builds
> report on point of failure rather than several steps up the chain.
> 
> /
>     Leif
> 
>> +    return NULL;
>> +  }
>>    Buffer = AllocatePool (BlockIo->Media->BlockSize);
>>    if (Buffer != NULL) {
>>      BlockIo->ReadBlocks (
>> -- 
>> 2.9.5
>>