Re: [PATCHv3] QEMU(upstream): Disable xen's use of O_DIRECT by default as it results in crashes.

[Paolo Bonzini <pbonzini@redhat.com>]

Il 18/03/2013 13:18, Alex Bligh ha scritto:
> Due to what is almost certainly a kernel bug, writes with
> O_DIRECT may continue to reference the page after the write
> has been marked as completed, particularly in the case of
> TCP retransmit. In other scenarios, this "merely" risks
> data corruption on the write, but with Xen pages from domU
> are only transiently mapped into dom0's memory, resulting
> in kernel panics when they are subsequently accessed.
> 
> This brings PV devices in line with emulated devices. Removing
> O_DIRECT is safe as barrier operations are now correctly passed
> through.
> 
> See:
>   http://lists.xen.org/archives/html/xen-devel/2012-12/msg01154.html
> for more details.
> 
> This patch has already been applied to the xenbits.org
> qemu-upstream repository.
>   http://xenbits.xen.org/gitweb/?p=qemu-upstream-unstable.git;a=commit;h=f3903bbac78a81fcbce1350cdce860764a62783a
> Clearly it should go into qemu's own repository.
> 
> Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
> Signed-off-by: Alex Bligh <alex@alex.org.uk>
> ---
>  hw/xen_disk.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/hw/xen_disk.c b/hw/xen_disk.c
> index a402ac8..14f8723 100644
> --- a/hw/xen_disk.c
> +++ b/hw/xen_disk.c
> @@ -603,7 +603,7 @@ static int blk_init(struct XenDevice *xendev)
>      }
>  
>      /* read-only ? */
> -    qflags = BDRV_O_NOCACHE | BDRV_O_CACHE_WB | BDRV_O_NATIVE_AIO;
> +    qflags = BDRV_O_CACHE_WB | BDRV_O_NATIVE_AIO;
>      if (strcmp(blkdev->mode, "w") == 0) {
>          qflags |= BDRV_O_RDWR;
>      } else {
> 

How does migration work with this change?  The target may read stale
data from the page cache.  Fix the kernel bug please.

Paolo


Paolo