Re: [PATCH 1/5] vhost-user-blk: Don't reconnect during initialisation

[Kevin Wolf]

Am 22.04.2021 um 19:02 hat Kevin Wolf geschrieben:
> This is a partial revert of commits 77542d43149 and bc79c87bcde.
> 
> Usually, an error during initialisation means that the configuration was
> wrong. Reconnecting won't make the error go away, but just turn the
> error condition into an endless loop. Avoid this and return errors
> again.
> 
> Additionally, calling vhost_user_blk_disconnect() from the chardev event
> handler could result in use-after-free because none of the
> initialisation code expects that the device could just go away in the
> middle. So removing the call fixes crashes in several places.
> 
> For example, using a num-queues setting that is incompatible with the
> backend would result in a crash like this (dereferencing dev->opaque,
> which is already NULL):
> 
>  #0  0x0000555555d0a4bd in vhost_user_read_cb (source=0x5555568f4690, 
> condition=(G_IO_IN | G_IO_HUP), opaque=0x7fffffffcbf0) at 
> ../hw/virtio/vhost-user.c:313
>  #1  0x0000555555d950d3 in qio_channel_fd_source_dispatch 
> (source=0x555557c3f750, callback=0x555555d0a478 <vhost_user_read_cb>, 
> user_data=0x7fffffffcbf0) at ../io/channel-watch.c:84
>  #2  0x00007ffff7b32a9f in g_main_context_dispatch () at 
> /lib64/libglib-2.0.so.0
>  #3  0x00007ffff7b84a98 in g_main_context_iterate.constprop () at 
> /lib64/libglib-2.0.so.0
>  #4  0x00007ffff7b32163 in g_main_loop_run () at /lib64/libglib-2.0.so.0
>  #5  0x0000555555d0a724 in vhost_user_read (dev=0x555557bc62f8, 
> msg=0x7fffffffcc50) at ../hw/virtio/vhost-user.c:402
>  #6  0x0000555555d0ee6b in vhost_user_get_config (dev=0x555557bc62f8, 
> config=0x555557bc62ac "", config_len=60) at ../hw/virtio/vhost-user.c:2133
>  #7  0x0000555555d56d46 in vhost_dev_get_config (hdev=0x555557bc62f8, 
> config=0x555557bc62ac "", config_len=60) at ../hw/virtio/vhost.c:1566
>  #8  0x0000555555cdd150 in vhost_user_blk_device_realize (dev=0x555557bc60b0, 
> errp=0x7fffffffcf90) at ../hw/block/vhost-user-blk.c:510
>  #9  0x0000555555d08f6d in virtio_device_realize (dev=0x555557bc60b0, 
> errp=0x7fffffffcff0) at ../hw/virtio/virtio.c:3660
> 
> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
> ---
>  hw/block/vhost-user-blk.c | 54 ++++++++++-----------------------------
>  1 file changed, 13 insertions(+), 41 deletions(-)
> 
> diff --git a/hw/block/vhost-user-blk.c b/hw/block/vhost-user-blk.c
> index f5e9682703..e824b0a759 100644
> --- a/hw/block/vhost-user-blk.c
> +++ b/hw/block/vhost-user-blk.c
> @@ -50,6 +50,8 @@ static const int user_feature_bits[] = {
>      VHOST_INVALID_FEATURE_BIT
>  };
>  
> +static void vhost_user_blk_event(void *opaque, QEMUChrEvent event);
> +
>  static void vhost_user_blk_update_config(VirtIODevice *vdev, uint8_t *config)
>  {
>      VHostUserBlk *s = VHOST_USER_BLK(vdev);
> @@ -362,19 +364,6 @@ static void vhost_user_blk_disconnect(DeviceState *dev)
>      vhost_dev_cleanup(&s->dev);
>  }
>  
> -static void vhost_user_blk_event(void *opaque, QEMUChrEvent event,
> -                                 bool realized);
> -
> -static void vhost_user_blk_event_realize(void *opaque, QEMUChrEvent event)
> -{
> -    vhost_user_blk_event(opaque, event, false);
> -}
> -
> -static void vhost_user_blk_event_oper(void *opaque, QEMUChrEvent event)
> -{
> -    vhost_user_blk_event(opaque, event, true);
> -}
> -
>  static void vhost_user_blk_chr_closed_bh(void *opaque)
>  {
>      DeviceState *dev = opaque;
> @@ -382,12 +371,11 @@ static void vhost_user_blk_chr_closed_bh(void *opaque)
>      VHostUserBlk *s = VHOST_USER_BLK(vdev);
>  
>      vhost_user_blk_disconnect(dev);
> -    qemu_chr_fe_set_handlers(&s->chardev, NULL, NULL,
> -            vhost_user_blk_event_oper, NULL, opaque, NULL, true);
> +    qemu_chr_fe_set_handlers(&s->chardev, NULL, NULL, vhost_user_blk_event,
> +                             NULL, opaque, NULL, true);
>  }
>  
> -static void vhost_user_blk_event(void *opaque, QEMUChrEvent event,
> -                                 bool realized)
> +static void vhost_user_blk_event(void *opaque, QEMUChrEvent event)
>  {
>      DeviceState *dev = opaque;
>      VirtIODevice *vdev = VIRTIO_DEVICE(dev);
> @@ -401,17 +389,7 @@ static void vhost_user_blk_event(void *opaque, 
> QEMUChrEvent event,
>          }
>          break;
>      case CHR_EVENT_CLOSED:
> -        /*
> -         * Closing the connection should happen differently on device
> -         * initialization and operation stages.
> -         * On initalization, we want to re-start vhost_dev initialization
> -         * from the very beginning right away when the connection is closed,
> -         * so we clean up vhost_dev on each connection closing.
> -         * On operation, we want to postpone vhost_dev cleanup to let the
> -         * other code perform its own cleanup sequence using vhost_dev data
> -         * (e.g. vhost_dev_set_log).
> -         */
> -        if (realized && !runstate_check(RUN_STATE_SHUTDOWN)) {
> +        if (!runstate_check(RUN_STATE_SHUTDOWN)) {
>              /*
>               * A close event may happen during a read/write, but vhost
>               * code assumes the vhost_dev remains setup, so delay the
> @@ -431,8 +409,6 @@ static void vhost_user_blk_event(void *opaque, 
> QEMUChrEvent event,
>               * knowing its type (in this case vhost-user).
>               */
>              s->dev.started = false;
> -        } else {
> -            vhost_user_blk_disconnect(dev);
>          }
>          break;
>      case CHR_EVENT_BREAK:
> @@ -490,31 +466,27 @@ static void vhost_user_blk_device_realize(DeviceState 
> *dev, Error **errp)
>      s->vhost_vqs = g_new0(struct vhost_virtqueue, s->num_queues);
>      s->connected = false;
>  
> -    qemu_chr_fe_set_handlers(&s->chardev,  NULL, NULL,
> -                             vhost_user_blk_event_realize, NULL, (void *)dev,
> -                             NULL, true);
> -
> -reconnect:
>      if (qemu_chr_fe_wait_connected(&s->chardev, &err) < 0) {
>          error_report_err(err);
>          goto virtio_err;
>      }

Preexisting bug: We have to set errp before jumping to virtio_err,
otherwise the caller (virtio_device_realize()) will take this as success
and crash when it later tries to access things that we've already freed
in the error path.

> -    /* check whether vhost_user_blk_connect() failed or not */
> -    if (!s->connected) {
> -        goto reconnect;
> +    if (vhost_user_blk_connect(dev) < 0) {
> +        qemu_chr_fe_disconnect(&s->chardev);
> +        goto virtio_err;

Here I'm newly introducing the same bug (fixed again by patch 2).

>      }
> +    assert(s->connected);
>  
>      ret = vhost_dev_get_config(&s->dev, (uint8_t *)&s->blkcfg,
>                                 sizeof(struct virtio_blk_config));
>      if (ret < 0) {
>          error_report("vhost-user-blk: get block config failed");
> -        goto reconnect;
> +        goto virtio_err;

And this one is wrong, too.

>      }
>  
> -    /* we're fully initialized, now we can operate, so change the handler */
> +    /* we're fully initialized, now we can operate, so add the handler */
>      qemu_chr_fe_set_handlers(&s->chardev,  NULL, NULL,
> -                             vhost_user_blk_event_oper, NULL, (void *)dev,
> +                             vhost_user_blk_event, NULL, (void *)dev,
>                               NULL, true);
>      return;

So maybe patch 2 should come first and also fix the preexisting bug, and
of course this patch needs a v2 that doesn't introduce the new instances
of the bug.

Kevin