gfs2: Fix request cancelation bug

In finish_xmote(), when a locking request is canceled, the corresponding
holder is moved to the tail of the holders list instead of being
dequeued immediately.  When there is only a single holder, the canceled
locking request is then immediately repeated.  This makes no sense; it
looks like another remnant of LM_FLAG_PRIORITY support.

Instead, dequeue canceled holders and proceed with the next holder in
finish_xmote().  We can then easily detect in gfs2_glock_dq() when a
holder has been canceled.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
This commit is contained in:
Andreas Gruenbacher 2025-02-05 17:16:38 +01:00
parent d838605fea
commit a431d49243

View file

@ -607,14 +607,19 @@ static void finish_xmote(struct gfs2_glock *gl, unsigned int ret)
if (gh && (ret & LM_OUT_CANCELED))
gfs2_holder_wake(gh);
if (gh && !test_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags)) {
/* move to back of queue and try next entry */
if (ret & LM_OUT_CANCELED) {
list_move_tail(&gh->gh_list, &gl->gl_holders);
list_del_init(&gh->gh_list);
trace_gfs2_glock_queue(gh, 0);
gl->gl_target = gl->gl_state;
gh = find_first_waiter(gl);
gl->gl_target = gh->gh_state;
if (do_promote(gl))
goto out;
goto retry;
if (gh) {
gl->gl_target = gh->gh_state;
if (do_promote(gl))
goto out;
do_xmote(gl, gh, gl->gl_target);
return;
}
goto out;
}
/* Some error or failed "try lock" - report it */
if ((ret & LM_OUT_ERROR) ||
@ -627,7 +632,6 @@ static void finish_xmote(struct gfs2_glock *gl, unsigned int ret)
switch(state) {
/* Unlocked due to conversion deadlock, try again */
case LM_ST_UNLOCKED:
retry:
do_xmote(gl, gh, gl->gl_target);
break;
/* Conversion fails, unlock and try again */
@ -1672,6 +1676,8 @@ void gfs2_glock_dq(struct gfs2_holder *gh)
gl->gl_name.ln_sbd->sd_lockstruct.ls_ops->lm_cancel(gl);
wait_on_bit(&gh->gh_iflags, HIF_WAIT, TASK_UNINTERRUPTIBLE);
spin_lock(&gl->gl_lockref.lock);
if (!gfs2_holder_queued(gh))
goto out;
}
/*