From cc7e79b168a552152299bd8a8254dc099aacc993 Mon Sep 17 00:00:00 2001 From: Wendy Cheng Date: Fri, 5 Oct 2007 00:27:58 -0400 Subject: [GFS2] Handle multiple glock demote requests Fix a race condition where multiple glock demote requests are sent to a node back-to-back. This patch does a check inside handle_callback() to see whether a demote request is in progress. If true, it sets a flag to make sure run_queue() will loop again to handle the new request, instead of erronously setting gl_demote_state to a different state. Signed-off-by: S. Wendy Cheng Signed-off-by: Steven Whitehouse --- fs/gfs2/glock.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) (limited to 'fs/gfs2/glock.c') diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index a37efe4..104e83f 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -567,7 +567,10 @@ static int rq_demote(struct gfs2_glock *gl) gfs2_demote_wake(gl); return 0; } + set_bit(GLF_LOCK, &gl->gl_flags); + set_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags); + if (gl->gl_demote_state == LM_ST_UNLOCKED || gl->gl_state != LM_ST_EXCLUSIVE) { spin_unlock(&gl->gl_spin); @@ -576,7 +579,9 @@ static int rq_demote(struct gfs2_glock *gl) spin_unlock(&gl->gl_spin); gfs2_glock_xmote_th(gl, NULL); } + spin_lock(&gl->gl_spin); + clear_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags); return 0; } @@ -606,6 +611,11 @@ static void run_queue(struct gfs2_glock *gl) } else if (test_bit(GLF_DEMOTE, &gl->gl_flags)) { blocked = rq_demote(gl); + if (gl->gl_waiters2 && !blocked) { + set_bit(GLF_DEMOTE, &gl->gl_flags); + gl->gl_demote_state = LM_ST_UNLOCKED; + } + gl->gl_waiters2 = 0; } else if (!list_empty(&gl->gl_waiters3)) { gh = list_entry(gl->gl_waiters3.next, struct gfs2_holder, gh_list); @@ -722,7 +732,10 @@ static void handle_callback(struct gfs2_glock *gl, unsigned int state, } } else if (gl->gl_demote_state != LM_ST_UNLOCKED && gl->gl_demote_state != state) { - gl->gl_demote_state = LM_ST_UNLOCKED; + if (test_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags)) + gl->gl_waiters2 = 1; + else + gl->gl_demote_state = LM_ST_UNLOCKED; } spin_unlock(&gl->gl_spin); } -- cgit v1.1 From c2932e03dbcfe7ea9052953dbd5f3157183c1e9b Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Thu, 1 Nov 2007 09:26:54 +0000 Subject: [GFS2] Remove "reclaim limit" This call to reclaim glocks is not needed, and in particular we don't want it in the fast path for locking glocks. The limit was entirely arbitrary anyway and we can't expect users to adjust things like this, the remaining code will do the right thing on its own. Signed-off-by: Steven Whitehouse --- fs/gfs2/glock.c | 9 --------- 1 file changed, 9 deletions(-) (limited to 'fs/gfs2/glock.c') diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 104e83f..159a547 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -507,21 +507,12 @@ static int rq_mutex(struct gfs2_holder *gh) static int rq_promote(struct gfs2_holder *gh) { struct gfs2_glock *gl = gh->gh_gl; - struct gfs2_sbd *sdp = gl->gl_sbd; if (!relaxed_state_ok(gl->gl_state, gh->gh_state, gh->gh_flags)) { if (list_empty(&gl->gl_holders)) { gl->gl_req_gh = gh; set_bit(GLF_LOCK, &gl->gl_flags); spin_unlock(&gl->gl_spin); - - if (atomic_read(&sdp->sd_reclaim_count) > - gfs2_tune_get(sdp, gt_reclaim_limit) && - !(gh->gh_flags & LM_FLAG_PRIORITY)) { - gfs2_reclaim_glock(sdp); - gfs2_reclaim_glock(sdp); - } - gfs2_glock_xmote_th(gh->gh_gl, gh); spin_lock(&gl->gl_spin); } -- cgit v1.1 From 3042a2ccd68d2b609d283219e51cba363aa35c1d Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Fri, 2 Nov 2007 08:39:34 +0000 Subject: [GFS2] Reorder writeback for glock sync Previously we were doing (write data, wait for data, write metadata, wait for metadata). After this patch we so (write metadata, write data, wait for data, wait for metadata) which should be more efficient. Also I noticed that the drop_bh and xmote_bh functions were almost identical. In fact the only difference was a single test, and that test is such that in the drop_bh case, it would always evaluate to the correct result. As such we can use the xmote_bh functions in all the places where we were using the drop_bh function and remove the drop_bh functions. Signed-off-by: Steven Whitehouse --- fs/gfs2/glock.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) (limited to 'fs/gfs2/glock.c') diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 159a547..e668808b 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -947,8 +947,8 @@ static void gfs2_glock_drop_th(struct gfs2_glock *gl) const struct gfs2_glock_operations *glops = gl->gl_ops; unsigned int ret; - if (glops->go_drop_th) - glops->go_drop_th(gl); + if (glops->go_xmote_th) + glops->go_xmote_th(gl); gfs2_assert_warn(sdp, test_bit(GLF_LOCK, &gl->gl_flags)); gfs2_assert_warn(sdp, list_empty(&gl->gl_holders)); @@ -1252,12 +1252,11 @@ void gfs2_glock_dq(struct gfs2_holder *gh) list_del_init(&gh->gh_list); if (list_empty(&gl->gl_holders)) { - spin_unlock(&gl->gl_spin); - - if (glops->go_unlock) + if (glops->go_unlock) { + spin_unlock(&gl->gl_spin); glops->go_unlock(gh); - - spin_lock(&gl->gl_spin); + spin_lock(&gl->gl_spin); + } gl->gl_stamp = jiffies; } -- cgit v1.1 From e589665eb97b297412fb16b4c1737a01a91db903 Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Fri, 2 Nov 2007 09:14:31 +0000 Subject: [GFS2] Remove flags no longer required The HIF_MUTEX and HIF_PROMOTE flags were set on the glock holders depending upon which of the two waiters lists they were going to be queued upon. They were then tested when the holders were taken off the lists to ensure that the right type of holder was being dequeued. Since we are already using separate lists, there doesn't seem a lot of point having these flags as well, and since setting them and testing them is in the fast path for locking and unlocking glock, this patch removes them. Signed-off-by: Steven Whitehouse --- fs/gfs2/glock.c | 17 ++--------------- 1 file changed, 2 insertions(+), 15 deletions(-) (limited to 'fs/gfs2/glock.c') diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index e668808b..5fbd9d3 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -594,12 +594,7 @@ static void run_queue(struct gfs2_glock *gl) if (!list_empty(&gl->gl_waiters1)) { gh = list_entry(gl->gl_waiters1.next, struct gfs2_holder, gh_list); - - if (test_bit(HIF_MUTEX, &gh->gh_iflags)) - blocked = rq_mutex(gh); - else - gfs2_assert_warn(gl->gl_sbd, 0); - + blocked = rq_mutex(gh); } else if (test_bit(GLF_DEMOTE, &gl->gl_flags)) { blocked = rq_demote(gl); if (gl->gl_waiters2 && !blocked) { @@ -610,12 +605,7 @@ static void run_queue(struct gfs2_glock *gl) } else if (!list_empty(&gl->gl_waiters3)) { gh = list_entry(gl->gl_waiters3.next, struct gfs2_holder, gh_list); - - if (test_bit(HIF_PROMOTE, &gh->gh_iflags)) - blocked = rq_promote(gh); - else - gfs2_assert_warn(gl->gl_sbd, 0); - + blocked = rq_promote(gh); } else break; @@ -636,7 +626,6 @@ static void gfs2_glmutex_lock(struct gfs2_glock *gl) struct gfs2_holder gh; gfs2_holder_init(gl, 0, 0, &gh); - set_bit(HIF_MUTEX, &gh.gh_iflags); if (test_and_set_bit(HIF_WAIT, &gh.gh_iflags)) BUG(); @@ -1160,8 +1149,6 @@ restart: return -EIO; } - set_bit(HIF_PROMOTE, &gh->gh_iflags); - spin_lock(&gl->gl_spin); add_to_queue(gh); run_queue(gl); -- cgit v1.1 From 2bcd610d2fdea608a8fdac32788fc35a32a2327c Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Thu, 8 Nov 2007 14:25:12 +0000 Subject: [GFS2] Don't add glocks to the journal The only reason for adding glocks to the journal was to keep track of which locks required a log flush prior to release. We add a flag to the glock to allow this check to be made in a simpler way. This reduces the size of a glock (by 12 bytes on i386, 24 on x86_64) and means that we can avoid extra work during the journal flush. Signed-off-by: Steven Whitehouse --- fs/gfs2/glock.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'fs/gfs2/glock.c') diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 5fbd9d3..d83df68 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -346,7 +346,6 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number, gl->gl_object = NULL; gl->gl_sbd = sdp; gl->gl_aspace = NULL; - lops_init_le(&gl->gl_le, &gfs2_glock_lops); INIT_DELAYED_WORK(&gl->gl_work, glock_work_func); /* If this glock protects actual on-disk data or metadata blocks, @@ -1900,8 +1899,6 @@ static int dump_glock(struct glock_iter *gi, struct gfs2_glock *gl) print_dbg(gi, " req_bh = %s\n", (gl->gl_req_bh) ? "yes" : "no"); print_dbg(gi, " lvb_count = %d\n", atomic_read(&gl->gl_lvb_count)); print_dbg(gi, " object = %s\n", (gl->gl_object) ? "yes" : "no"); - print_dbg(gi, " le = %s\n", - (list_empty(&gl->gl_le.le_list)) ? "no" : "yes"); print_dbg(gi, " reclaim = %s\n", (list_empty(&gl->gl_reclaim)) ? "no" : "yes"); if (gl->gl_aspace) -- cgit v1.1 From 1a2781cfa5ed8eb82bb311d684f268c1822dae69 Mon Sep 17 00:00:00 2001 From: Fabio Massimo Di Nitto Date: Fri, 16 Nov 2007 09:50:40 +0000 Subject: [GFS2] Fix runtime issue with UP kernels The issue is indeed UP vs SMP and it is totally random. spin_is_locked() is a bad assertion because there is no correct answer on UP. on UP spin_is_locked() has to return either one value or another, always. This means that in my setup I am lucky enough to trigger the issue and your you are lucky enough not to. the patch in attachment removes the bogus calls to BUG_ON and according to David (in CC and thanks for the long explanation on the problem) we can rely upon things like lockdep to find problem that might be trying to catch. Signed-off-by: Fabio M. Di Nitto Cc: David S. Miller Signed-off-by: Steven Whitehouse --- fs/gfs2/glock.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'fs/gfs2/glock.c') diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index d83df68..a7f3c46 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -217,7 +217,6 @@ int gfs2_glock_put(struct gfs2_glock *gl) if (atomic_dec_and_test(&gl->gl_ref)) { hlist_del(&gl->gl_list); write_unlock(gl_lock_addr(gl->gl_hash)); - BUG_ON(spin_is_locked(&gl->gl_spin)); gfs2_assert(sdp, gl->gl_state == LM_ST_UNLOCKED); gfs2_assert(sdp, list_empty(&gl->gl_reclaim)); gfs2_assert(sdp, list_empty(&gl->gl_holders)); @@ -460,7 +459,6 @@ static void wait_on_holder(struct gfs2_holder *gh) static void gfs2_demote_wake(struct gfs2_glock *gl) { - BUG_ON(!spin_is_locked(&gl->gl_spin)); gl->gl_demote_state = LM_ST_EXCLUSIVE; clear_bit(GLF_DEMOTE, &gl->gl_flags); smp_mb__after_clear_bit(); @@ -680,7 +678,6 @@ static void gfs2_glmutex_unlock(struct gfs2_glock *gl) gl->gl_owner_pid = 0; gl->gl_ip = 0; run_queue(gl); - BUG_ON(!spin_is_locked(&gl->gl_spin)); spin_unlock(&gl->gl_spin); } -- cgit v1.1 From 398bbe68321947f6763fbc259a01eb548ce19408 Mon Sep 17 00:00:00 2001 From: Bob Peterson Date: Tue, 11 Dec 2007 19:13:54 -0600 Subject: [GFS2] Reorganize function gfs2_glmutex_lock This patch optimizes the function gfs2_glmutex_lock. The basic theory is: Why bother initializing a holder, setting up wait bits and then waiting on them, if you know the glock can be yours. So the holder stuff is placed inside the if checking if the glock is locked. This one needs careful scrutiny because changing anything to do with locking should strike terror into one's heart. Signed-off-by: Bob Peterson Signed-off-by: Steven Whitehouse --- fs/gfs2/glock.c | 23 +++++++++-------------- 1 file changed, 9 insertions(+), 14 deletions(-) (limited to 'fs/gfs2/glock.c') diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index a7f3c46..80e09c5 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -1,6 +1,6 @@ /* * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. - * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved. + * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved. * * This copyrighted material is made available to anyone wishing to use, * modify, copy, or redistribute it subject to the terms and conditions @@ -620,26 +620,21 @@ static void run_queue(struct gfs2_glock *gl) static void gfs2_glmutex_lock(struct gfs2_glock *gl) { - struct gfs2_holder gh; - - gfs2_holder_init(gl, 0, 0, &gh); - if (test_and_set_bit(HIF_WAIT, &gh.gh_iflags)) - BUG(); - spin_lock(&gl->gl_spin); if (test_and_set_bit(GLF_LOCK, &gl->gl_flags)) { + struct gfs2_holder gh; + + gfs2_holder_init(gl, 0, 0, &gh); + set_bit(HIF_WAIT, &gh.gh_iflags); list_add_tail(&gh.gh_list, &gl->gl_waiters1); + spin_unlock(&gl->gl_spin); + wait_on_holder(&gh); + gfs2_holder_uninit(&gh); } else { gl->gl_owner_pid = current->pid; gl->gl_ip = (unsigned long)__builtin_return_address(0); - clear_bit(HIF_WAIT, &gh.gh_iflags); - smp_mb(); - wake_up_bit(&gh.gh_iflags, HIF_WAIT); + spin_unlock(&gl->gl_spin); } - spin_unlock(&gl->gl_spin); - - wait_on_holder(&gh); - gfs2_holder_uninit(&gh); } /** -- cgit v1.1