On Sun, May 22, 2022 at 9:59 PM Christophe JAILLET
<christophe.jaillet@xxxxxxxxxx> wrote:
>
> In the error paths introduced by the commit in the Fixes tag, a mutex may
> be left locked.
> Add the correct goto instead of a direct return.
>
> Fixes: a1468175bb17 ("vhost-vdpa: support ASID based IOTLB API")
> Signed-off-by: Christophe JAILLET <christophe.jaillet@xxxxxxxxxx>
> ---
> WARNING: This patch only fixes the goto vs return mix-up in this function.
> However, the 2nd hunk looks really spurious to me. I think that the:
> - return -EINVAL;
> + r = -EINVAL;
> + goto unlock;
> should be done only in the 'if (!iotlb)' block.
It should be fine, the error happen if
1) the batched ASID based request is not equal (the first if)
2) there's no IOTLB for this ASID (the second if)
But I agree the code could be tweaked to use two different if instead
of using a or condition here.
Acked-by: Jason Wang <jasowang@xxxxxxxxxx>
>
> As I don't know this code, I just leave it as-is but draw your attention
> in case this is another bug lurking.
> ---
> drivers/vhost/vdpa.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/vhost/vdpa.c b/drivers/vhost/vdpa.c
> index 1f1d1c425573..3e86080041fc 100644
> --- a/drivers/vhost/vdpa.c
> +++ b/drivers/vhost/vdpa.c
> @@ -1000,7 +1000,8 @@ static int vhost_vdpa_process_iotlb_msg(struct vhost_dev *dev, u32 asid,
> if (!as) {
> dev_err(&v->dev, "can't find and alloc asid %d\n",
> asid);
> - return -EINVAL;
> + r = -EINVAL;
> + goto unlock;
> }
> iotlb = &as->iotlb;
> } else
> @@ -1013,7 +1014,8 @@ static int vhost_vdpa_process_iotlb_msg(struct vhost_dev *dev, u32 asid,
> }
> if (!iotlb)
> dev_err(&v->dev, "no iotlb for asid %d\n", asid);
> - return -EINVAL;
> + r = -EINVAL;
> + goto unlock;
> }
>
> switch (msg->type) {
> --
> 2.34.1
>
