From: Andreas Gruenbacher Date: Wed, 17 Dec 2025 22:22:49 +0000 (+0000) Subject: gfs2: Do not cancel internal demote requests X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=4928c36536d218ae70a1b79f1c0d3302ffe31f69;p=thirdparty%2Flinux.git gfs2: Do not cancel internal demote requests Trying to change the state of a glock may result in a "conversion deadlock" error, indicating that the requested state transition would cause a deadlock. In this case, we unlock and retry the state change. It makes no sense to try canceling those unlock requests, but the current code is not aware of that. In addition, if a locking request is canceled shortly after it is made, the cancelation request can currently overtake the locking request. This may result in deadlocks. Fix both of these bugs by repurposing the GLF_PENDING_REPLY flag into a GLF_MAY_CANCEL flag which is set only when the current locking request can be canceled. When trying to cancel a locking request in gfs2_glock_dq(), wait for this flag to be set. Signed-off-by: Andreas Gruenbacher --- diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 7761799333155..7c3d488327eef 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -60,7 +60,8 @@ struct gfs2_glock_iter { typedef void (*glock_examiner) (struct gfs2_glock * gl); -static void do_xmote(struct gfs2_glock *gl, struct gfs2_holder *gh, unsigned int target); +static void do_xmote(struct gfs2_glock *gl, struct gfs2_holder *gh, + unsigned int target, bool may_cancel); static void request_demote(struct gfs2_glock *gl, unsigned int state, unsigned long delay, bool remote); @@ -600,12 +601,14 @@ static void finish_xmote(struct gfs2_glock *gl, unsigned int ret) switch(gl->gl_state) { /* Unlocked due to conversion deadlock, try again */ case LM_ST_UNLOCKED: - do_xmote(gl, gh, gl->gl_target); + do_xmote(gl, gh, gl->gl_target, + !test_bit(GLF_DEMOTE_IN_PROGRESS, + &gl->gl_flags)); break; /* Conversion fails, unlock and try again */ case LM_ST_SHARED: case LM_ST_DEFERRED: - do_xmote(gl, gh, LM_ST_UNLOCKED); + do_xmote(gl, gh, LM_ST_UNLOCKED, false); break; default: /* Everything else */ fs_err(gl->gl_name.ln_sbd, @@ -638,7 +641,7 @@ static void finish_xmote(struct gfs2_glock *gl, unsigned int ret) } out: if (!test_bit(GLF_CANCELING, &gl->gl_flags)) - clear_bit(GLF_LOCK, &gl->gl_flags); + clear_and_wake_up_bit(GLF_LOCK, &gl->gl_flags); } /** @@ -646,11 +649,12 @@ out: * @gl: The lock state * @gh: The holder (only for promotes) * @target: The target lock state + * @may_cancel: Operation may be canceled * */ static void do_xmote(struct gfs2_glock *gl, struct gfs2_holder *gh, - unsigned int target) + unsigned int target, bool may_cancel) __releases(&gl->gl_lockref.lock) __acquires(&gl->gl_lockref.lock) { @@ -703,17 +707,20 @@ skip_inval: } if (ls->ls_ops->lm_lock) { - set_bit(GLF_PENDING_REPLY, &gl->gl_flags); spin_unlock(&gl->gl_lockref.lock); ret = ls->ls_ops->lm_lock(gl, target, gh ? gh->gh_flags : 0); spin_lock(&gl->gl_lockref.lock); if (!ret) { + if (may_cancel) { + set_bit(GLF_MAY_CANCEL, &gl->gl_flags); + smp_mb__after_atomic(); + wake_up_bit(&gl->gl_flags, GLF_LOCK); + } /* The operation will be completed asynchronously. */ gl->gl_lockref.count++; return; } - clear_bit(GLF_PENDING_REPLY, &gl->gl_flags); if (ret == -ENODEV) { /* @@ -774,7 +781,7 @@ __acquires(&gl->gl_lockref.lock) GLOCK_BUG_ON(gl, gl->gl_demote_state == LM_ST_EXCLUSIVE); gl->gl_target = gl->gl_demote_state; set_bit(GLF_LOCK, &gl->gl_flags); - do_xmote(gl, NULL, gl->gl_target); + do_xmote(gl, NULL, gl->gl_target, false); return; } @@ -791,7 +798,7 @@ promote: if (!(gh->gh_flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB))) do_error(gl, 0); /* Fail queued try locks */ set_bit(GLF_LOCK, &gl->gl_flags); - do_xmote(gl, gh, gl->gl_target); + do_xmote(gl, gh, gl->gl_target, true); return; out_sched: @@ -1588,6 +1595,7 @@ void gfs2_glock_dq(struct gfs2_holder *gh) { struct gfs2_glock *gl = gh->gh_gl; +again: spin_lock(&gl->gl_lockref.lock); if (!gfs2_holder_queued(gh)) { /* @@ -1602,13 +1610,25 @@ void gfs2_glock_dq(struct gfs2_holder *gh) test_bit(GLF_LOCK, &gl->gl_flags) && !test_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags) && !test_bit(GLF_CANCELING, &gl->gl_flags)) { + if (!test_bit(GLF_MAY_CANCEL, &gl->gl_flags)) { + struct wait_queue_head *wq; + DEFINE_WAIT(wait); + + wq = bit_waitqueue(&gl->gl_flags, GLF_LOCK); + prepare_to_wait(wq, &wait, TASK_UNINTERRUPTIBLE); + spin_unlock(&gl->gl_lockref.lock); + schedule(); + finish_wait(wq, &wait); + goto again; + } + set_bit(GLF_CANCELING, &gl->gl_flags); spin_unlock(&gl->gl_lockref.lock); 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); clear_bit(GLF_CANCELING, &gl->gl_flags); - clear_bit(GLF_LOCK, &gl->gl_flags); + clear_and_wake_up_bit(GLF_LOCK, &gl->gl_flags); if (!gfs2_holder_queued(gh)) goto out; } @@ -1838,7 +1858,7 @@ void gfs2_glock_complete(struct gfs2_glock *gl, int ret) struct lm_lockstruct *ls = &gl->gl_name.ln_sbd->sd_lockstruct; spin_lock(&gl->gl_lockref.lock); - clear_bit(GLF_PENDING_REPLY, &gl->gl_flags); + clear_bit(GLF_MAY_CANCEL, &gl->gl_flags); gl->gl_reply = ret; if (unlikely(test_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags))) { @@ -2245,8 +2265,8 @@ static const char *gflags2str(char *buf, const struct gfs2_glock *gl) *p++ = 'y'; if (test_bit(GLF_LFLUSH, gflags)) *p++ = 'f'; - if (test_bit(GLF_PENDING_REPLY, gflags)) - *p++ = 'R'; + if (test_bit(GLF_MAY_CANCEL, gflags)) + *p++ = 'c'; if (test_bit(GLF_HAVE_REPLY, gflags)) *p++ = 'r'; if (test_bit(GLF_INITIAL, gflags)) diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h index d05d8fe4e456b..f7909607936af 100644 --- a/fs/gfs2/incore.h +++ b/fs/gfs2/incore.h @@ -326,7 +326,7 @@ enum { GLF_BLOCKING = 15, GLF_TRY_TO_EVICT = 17, /* iopen glocks only */ GLF_VERIFY_DELETE = 18, /* iopen glocks only */ - GLF_PENDING_REPLY = 19, + GLF_MAY_CANCEL = 19, GLF_DEFER_DELETE = 20, /* iopen glocks only */ GLF_CANCELING = 21, }; diff --git a/fs/gfs2/trace_gfs2.h b/fs/gfs2/trace_gfs2.h index fcfbf68ec725c..a308228d5c2d8 100644 --- a/fs/gfs2/trace_gfs2.h +++ b/fs/gfs2/trace_gfs2.h @@ -52,7 +52,7 @@ {(1UL << GLF_DEMOTE_IN_PROGRESS), "p" }, \ {(1UL << GLF_DIRTY), "y" }, \ {(1UL << GLF_LFLUSH), "f" }, \ - {(1UL << GLF_PENDING_REPLY), "R" }, \ + {(1UL << GLF_MAY_CANCEL), "c" }, \ {(1UL << GLF_HAVE_REPLY), "r" }, \ {(1UL << GLF_INITIAL), "a" }, \ {(1UL << GLF_HAVE_FROZEN_REPLY), "F" }, \