Re: [PATCH 0/2] MdeModulePkg/AtaAtapiPassThru: Revert patch to disable PCI attributes

[Laszlo Ersek <lersek@redhat.com>]

Hello Ray,

On 11/27/17 02:19, Ruiyu Ni wrote:
> The patches caused Windows 10 S4 resume failure.
> Considering the similar changes are reverted from PciBus driver,
> revert the patches from AtaAtapiPassThru as well.
> 
> Ruiyu Ni (2):
>   MdeModulePkg/AtaAtapiPassThru: Revert patch to disable Bus Master
>   MdeModulePkg/AtaAtapiPassThru: Revert patch to disable PCI attributes
> 
>  .../Bus/Ata/AtaAtapiPassThru/AtaAtapiPassThru.c    | 58 +---------------------
>  .../Bus/Ata/AtaAtapiPassThru/AtaAtapiPassThru.h    |  5 --
>  2 files changed, 1 insertion(+), 62 deletions(-)
> 

it looks like these patches have not been committed yet, which is a good
thing, because apparently there's a better solution than a full revert.
Namely, in the other sub-thread at
<http://mid.mail-archive.com/0236afa2-e365-af7a-9374-7fd1ad742c36@redhat.com>
(alternative link:
<https://lists.01.org/pipermail/edk2-devel/2017-November/018046.html>),
Jiewen and Star seem to accept AhciReset() as a better way to abort
pending DMA.

This means that we need not revert the EBS-handler altogether, only
change what it does. Could you give that a try please?

(If the Windows regression is very urgent to fix, then I don't mind if
the Bus Master disabling is removed separately, before AhciReset() is
added; but in that case, a full revert looks unjustified, since the EBS
handler will have to be re-added for AhciReset() anyway.)

Thanks,
Laszlo