[Cluster-devel] [GFS2 PATCH] GFS2: Allow glocks to be unlocked after withdraw

Andreas Gruenbacher agruenba at redhat.com
Fri May 5 15:38:14 UTC 2017


Bob,

On Fri, May 5, 2017 at 4:59 PM, Bob Peterson <rpeterso at redhat.com> wrote:
> Hi,
>
> Before this patch, glocks could not be unlocked after a withdraw,
> but that resulted in hangs during unmount. This patch allows GFS2
> function do_xmote to demote locks to an unlocked state, but not
> promoting any upward.

the commit message should point out that this relates to commit 0d1c7ae9d8 which
currently is on for-next, so it fixes a regression and should probably end up in
this cycle as well.

> Signed-off-by: Bob Peterson <rpeterso at redhat.com>
> ---
> diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
> index 9e81219..959a19c 100644
> --- a/fs/gfs2/glock.c
> +++ b/fs/gfs2/glock.c
> @@ -449,7 +449,8 @@ __acquires(&gl->gl_lockref.lock)
>         unsigned int lck_flags = (unsigned int)(gh ? gh->gh_flags : 0);
>         int ret;
>
> -       if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
> +       if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
> +           target != LM_ST_UNLOCKED)
>                 return;
>         lck_flags &= (LM_FLAG_TRY | LM_FLAG_TRY_1CB | LM_FLAG_NOEXP |
>                       LM_FLAG_PRIORITY);
> @@ -486,7 +487,8 @@ __acquires(&gl->gl_lockref.lock)
>                 }
>                 else if (ret) {
>                         pr_err("lm_lock ret %d\n", ret);
> -                       GLOCK_BUG_ON(gl, 1);
> +                       GLOCK_BUG_ON(gl, !test_bit(SDF_SHUTDOWN,
> +                                                  &sdp->sd_flags));
>                 }
>         } else { /* lock_nolock */
>                 finish_xmote(gl, target);

Thanks,
Andreas




More information about the Cluster-devel mailing list