[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-block] [PATCH 4/4] virtio-blk: Clean up start/stop with mutex

From: Stefan Hajnoczi
Subject: Re: [Qemu-block] [PATCH 4/4] virtio-blk: Clean up start/stop with mutex and BH
Date: Thu, 17 Mar 2016 15:00:57 +0000
User-agent: Mutt/1.5.24 (2015-08-30)

On Wed, Mar 16, 2016 at 06:10:18PM +0800, Fam Zheng wrote:
> +    data = g_new(VirtIOBlockStartData, 1);
> +    data->vblk = vblk;
> +    data->bh = aio_bh_new(s->ctx, virtio_blk_data_plane_start_bh_cb, data);
> +    qemu_bh_schedule(data->bh);
> +    qemu_mutex_unlock(&s->start_stop_lock);
>      return;

This BH usage pattern is dangerous:

1. The BH is created and scheduled.
2. Before the BH executes the device is unrealized.
3. The data->bh pointer is inaccessible so we have a dangling BH that
   will access vblk after vblk has been freed.

In some cases it can be safe but I don't see why the pattern is safe in
this case.  Either the BH needs to hold some sort of reference to keep
vblk alive, or vblk needs to know about pending BHs so they can be

Attachment: signature.asc
Description: PGP signature

reply via email to

[Prev in Thread] Current Thread [Next in Thread]