[Qemu-devel] Fixing dma-helper qiov truncation

[Qemu-devel] Fixing dma-helper qiov truncation

[Mark Cave-Ayland]

Now that the 2.8 tree is open, I'd like to resurrect this earlier patch
to switch the macio controller over to the new byte-aligned DMA helpers:
https://lists.gnu.org/archive/html/qemu-devel/2016-05/msg04907.html.

The first patch in the series is a hack to work around the issue that
byte-aligned qiovs are being truncated with the existing code. git blame
shows that this code was introduced by this commit:


commit 58f423fbd5f7f435e8dc56dfa565b043d20d1e1b
Author: Kevin Wolf <kwolf@redhat.com>
Date:   Wed Jul 9 19:17:30 2014 +0200

    dma-helpers: Fix too long qiov

    If the size of the scatter/gather list isn't a multiple of 512, the
    number of sectors for the block layer request is rounded down, resulting
    in a qiov that doesn't match the request length. Truncate the qiov
to the
    new length of the request.

    This fixes the IDE qtest case /x86_64/ide/bmdma/short_prdt.

    Signed-off-by: Kevin Wolf <kwolf@redhat.com>
    Reviewed-by: Eric Blake <eblake@redhat.com>


Kevin/Paolo - what should be the correct behaviour here given that the
underlying APIs are now byte-aligned?


ATB,

Mark.

Re: [Qemu-devel] Fixing dma-helper qiov truncation

[Paolo Bonzini]

On 07/09/2016 09:31, Mark Cave-Ayland wrote:
> Now that the 2.8 tree is open, I'd like to resurrect this earlier patch
> to switch the macio controller over to the new byte-aligned DMA helpers:
> https://lists.gnu.org/archive/html/qemu-devel/2016-05/msg04907.html.
> 
> The first patch in the series is a hack to work around the issue that
> byte-aligned qiovs are being truncated with the existing code. git blame
> shows that this code was introduced by this commit:
> 
> 
> commit 58f423fbd5f7f435e8dc56dfa565b043d20d1e1b
> Author: Kevin Wolf <kwolf@redhat.com>
> Date:   Wed Jul 9 19:17:30 2014 +0200
> 
>     dma-helpers: Fix too long qiov
> 
>     If the size of the scatter/gather list isn't a multiple of 512, the
>     number of sectors for the block layer request is rounded down, resulting
>     in a qiov that doesn't match the request length. Truncate the qiov
> to the
>     new length of the request.
> 
>     This fixes the IDE qtest case /x86_64/ide/bmdma/short_prdt.
> 
>     Signed-off-by: Kevin Wolf <kwolf@redhat.com>
>     Reviewed-by: Eric Blake <eblake@redhat.com>
> 
> 
> Kevin/Paolo - what should be the correct behaviour here given that the
> underlying APIs are now byte-aligned?

I think the expected alignment should be passed down to dma-helpers, and
they should ignore the unaligned tail of the QEMUSGList.

Paolo