[Linaro-uefi,MdeModulePkg/Library,v1,1/1] MdeModulePkg/UefiBootManangerLib: Fix exception issue

Message ID 20190225091052.25461-2-ming.huang@linaro.org
State New
Headers show
Series
  • Fix exception issue while UsbMass block io uninstalled
Related show

Commit Message

Ming Huang Feb. 25, 2019, 9:10 a.m.
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(-)

Comments

Wu, Hao A March 19, 2019, 5:56 a.m. | #1
> -----Original Message-----

> From: Ming Huang [mailto:ming.huang@linaro.org]

> Sent: Tuesday, March 19, 2019 12:14 PM

> To: Wu, Hao A; Leif Lindholm

> Cc: linaro-uefi@lists.linaro.org; edk2-devel@lists.01.org; Zeng, Star; Dong,

> Eric; Ni, Ray; dann.frazier@canonical.com; ard.biesheuvel@linaro.org; Kinney,

> Michael D; Gao, Liming; wanghuiqiang@huawei.com;

> huangming23@huawei.com; zhangjinsong2@huawei.com;

> huangdaode@hisilicon.com; waip23@126.com; Wang, Jian J

> Subject: Re: [MdeModulePkg/Library v1 1/1]

> MdeModulePkg/UefiBootManangerLib: Fix exception issue

> 

> 

> 

> On 3/19/2019 10:25 AM, Wu, Hao A wrote:

> >> -----Original Message-----

> >> From: Leif Lindholm [mailto:leif.lindholm@linaro.org]

> >> Sent: Monday, March 18, 2019 8:43 PM

> >> To: Ming Huang

> >> Cc: linaro-uefi@lists.linaro.org; edk2-devel@lists.01.org; Zeng, Star; Dong,

> >> Eric; Ni, Ray; dann.frazier@canonical.com; ard.biesheuvel@linaro.org;

> Kinney,

> >> Michael D; Gao, Liming; wanghuiqiang@huawei.com;

> >> huangming23@huawei.com; zhangjinsong2@huawei.com;

> >> huangdaode@hisilicon.com; waip23@126.com; Wang, Jian J; Wu, Hao A;

> Ni,

> >> Ray

> >> Subject: Re: [MdeModulePkg/Library v1 1/1]

> >> MdeModulePkg/UefiBootManangerLib: Fix exception issue

> >>

> >> +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?

> >>

> >> 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

> >

> > Do you mean 'EFI_INVALID_PARAMETER' is returned from:

> >   Status = gBS->HandleProtocol (Handle, &gEfiBlockIoProtocolGuid, (VOID

> **) &BlockIo);

> 

> Yes.

> 

> >

> > If so, my guess is that 'Handle' is NULL at this point. An improvement can

> > be adding a previous check for 'Status' after the ASSERT at:

> >

> >   Status = gBS->LocateDevicePath (&gEfiBlockIoProtocolGuid,

> &TempDevicePath, &Handle);

> >   ASSERT_EFI_ERROR (Status);

> 

> As my debug output, this 'Status' is seccuss and Handle is not NULL, but

> gBS->ConnectController return:Not Found

> 

> Debug output:

> [BmExpandMediaDevicePath]:[1056L] Handle=3E3F3D18 BlockIo=3B2757B6

> Media=AFAF6C617470AFAF Status=Success

> EhcExecTransfer: transfer failed with 40

> EhcBulkTransfer: error - Device Error, transfer - 40

> .........

> [UsbOnHubInterrupt]:[632L] SignalEvent (HubIf->HubNotify)

> UsbBotExecCommand: UsbBotSendCommand (Device Error)

> UsbBootExecCmd: Device Error to Exec 0x0 Cmd (Result = 1)

> EhcExecTransfer: transfer failed with 40

> ...............

> [USBMassDriverBindingStop]:[1010L] Uninstall USB block io, free:

> 3E44F218(F0)

> [BmExpandMediaDevicePath]:[1064L] Connect Not Found

> [BmExpandMediaDevicePath]:[1076L] Handle=3E3F3D18 BlockIo=3B2757B6

> Media=AFAF6C617470AFAF Status=Invalid Parameter


Thanks for the debug information, I got it now.

The call to the gBS->ConnectController() leads to protocols being
uninstalled from 'Handle' and removing 'Handle' from the database. Then
within the call to gBS->HandleProtocol(), CoreValidateHandle() returns
EFI_INVALID_PARAMETER since the handle cannot be found.

I am good with this patch, please help to address Leif's previous comment
to keep the ASSERT.

Also, I have filed a Bugzilla tracker for this:
https://bugzilla.tianocore.org/show_bug.cgi?id=1631

Could you help to add the reference to the above BZ in the commit log
message? Thanks.


Best Regards,
Hao Wu

> 

> Thanks

> 

> >

> > And leave:

> >

> >   Status = gBS->HandleProtocol (Handle, &gEfiBlockIoProtocolGuid, (VOID

> **) &BlockIo);

> >   ASSERT_EFI_ERROR (Status);

> >

> > unchanged.

> 

> 

> 

> >

> > Best Regards,

> > Hao Wu

> >

> >>> 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

> >>>

Patch

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)) {
+    return NULL;
+  }
   Buffer = AllocatePool (BlockIo->Media->BlockSize);
   if (Buffer != NULL) {
     BlockIo->ReadBlocks (