Re: [Qemu-devel] [PATCH] qga-win: don't hang if vss hold writes timeout

[Chen Hanxiao]

在 2017-09-23 05:12:20，"Tomoki Sekiyama" <address@hidden> 写道：
 
> 
> Thanks for your patch.
> Which version of Windows have you found the hang?
> 
> 
  I tested on win08, win 08 r2 and win2012.

QGA versions:

  Use the latest qga-win from 
  
https://fedorapeople.org/groups/virt/virtio-win/direct-downloads/archive-qemu-ga/qemu-ga-win-7.4.5-1/
 
  and stable branch from
  
https://fedorapeople.org/groups/virt/virtio-win/direct-downloads/archive-virtio/virtio-win-0.1.126-2/virtio-win-0.1.126.iso

  All of them have the same behavior.

How to reproduce:

  Copy a big file in guest (from d: to c:), then do fsfreeze when IO is in 
progress.

  At this time, we had a very chance to see that qga hang for a long time.

  Libvirt will not recieve a reply.
  qemu-ga service even don't answer `net stop` command.


Regards,
- Chen
 
  

> Regards,
> Tomoki
> 
> 
> 2017-09-22 11:46 GMT+09:00 Chen Hanxiao <address@hidden>:
> 
> From: Chen Hanxiao <address@hidden>
> 
> 
> 
> When VM is in a heavy IO, if the command "guest-fsfreeze-freeze"
> 
> is executed, VSS may timeout when trying to hold writes.
> 
> 
> 
> Inside guest, Event ID 12298(VSS_ERROR_HOLD_WRITES_TIMEOUT)
> 
> is logged in the Event Viewer.
> 
> 
> 
> At that time, if we call AbortBackup, qga may hang forever.
> 
> 
> 
> This patch will solve this issues.
> 
> 
> 
> Signed-off-by: Chen Hanxiao <address@hidden>
> 
> ---
> 
>  qga/vss-win32/requester.cpp | 12 ++++++++++++
> 
>  1 file changed, 12 insertions(+)
> 
> 
> 
> diff --git a/qga/vss-win32/requester.cpp b/qga/vss-win32/requester.cpp
> 
> index 301762d..3d9c971 100644
> 
> --- a/qga/vss-win32/requester.cpp
> 
> +++ b/qga/vss-win32/requester.cpp
> 
> @@ -419,6 +419,16 @@ void requester_freeze(int *num_vols, ErrorSet *errset)
> 
>              break;
> 
>          }
> 
>      }
> 
> +
> 
> +    if (wait_status == WAIT_TIMEOUT) {
> 
> +        err_set(errset, E_FAIL,
> 
> +                "timeout when try to receive Frozen event from VSS 
> provider");
> 
> +        /* If we are here, VSS had timeout.
> 
> +         * Don't call AbortBackup, just return directly.
> 
> +         */
> 
> +        goto out1;
> 
> +    }
> 
> +
> 
>      if (wait_status != WAIT_OBJECT_0) {
> 
>          err_set(errset, E_FAIL,
> 
>                  "couldn't receive Frozen event from VSS provider");
> 
> @@ -432,6 +442,8 @@ out:
> 
>      if (vss_ctx.pVssbc) {
> 
>          vss_ctx.pVssbc->AbortBackup();
> 
>      }
> 
> +
> 
> +out1:
> 
>      requester_cleanup();
> 
>      CoUninitialize();
> 
>  }