Hi,
Here are the current set of small fixes relating to GFS2.
This patch set contains two minor docs/spelling fixes, some fixes for
flock, a change to use GFP_NOFS to avoid recursion on a rarely used code
path and a fix for a race relating to the glock lru,
Steve.
We must not leave items on the LRU list with GLF_LOCK set, since
they can be removed if the glock is brought back into use, which
may then potentially result in a hang, waiting for GLF_LOCK to
clear.
It doesn't happen very often, since it requires a glock that has
not been used for a long time to be brought back into use at the
same moment that the shrinker is part way through disposing of
glocks.
The fix is to set GLF_LOCK at a later time, when we already know
that the other locks can be obtained. Also, we now only release
the lru_lock in case a resched is needed, rather than on every
iteration.
Signed-off-by: Steven Whitehouse <[email protected]>
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 278fae5..c1e5b12 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -1406,12 +1406,16 @@ __acquires(&lru_lock)
gl = list_entry(list->next, struct gfs2_glock, gl_lru);
list_del_init(&gl->gl_lru);
if (!spin_trylock(&gl->gl_spin)) {
+add_back_to_lru:
list_add(&gl->gl_lru, &lru_list);
atomic_inc(&lru_count);
continue;
}
+ if (test_and_set_bit(GLF_LOCK, &gl->gl_flags)) {
+ spin_unlock(&gl->gl_spin);
+ goto add_back_to_lru;
+ }
clear_bit(GLF_LRU, &gl->gl_flags);
- spin_unlock(&lru_lock);
gl->gl_lockref.count++;
if (demote_ok(gl))
handle_callback(gl, LM_ST_UNLOCKED, 0, false);
@@ -1419,7 +1423,7 @@ __acquires(&lru_lock)
if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
gl->gl_lockref.count--;
spin_unlock(&gl->gl_spin);
- spin_lock(&lru_lock);
+ cond_resched_lock(&lru_lock);
}
}
@@ -1444,7 +1448,7 @@ static long gfs2_scan_glock_lru(int nr)
gl = list_entry(lru_list.next, struct gfs2_glock, gl_lru);
/* Test for being demotable */
- if (!test_and_set_bit(GLF_LOCK, &gl->gl_flags)) {
+ if (!test_bit(GLF_LOCK, &gl->gl_flags)) {
list_move(&gl->gl_lru, &dispose);
atomic_dec(&lru_count);
freed++;
--
1.8.3.1
Normally GFP_KERNEL is ok here, but there is now a rarely used code path
relating to deallocation of unlinked inodes (in certain corner cases)
which if hit at times of memory shortage can cause recursion while
trying to free memory.
One solution would be to try and move the gfs2_glock_get() call so
that it is no longer called while another glock is held, but that
doesn't look at all easy, so GFP_NOFS is the best solution for the
time being.
Signed-off-by: Steven Whitehouse <[email protected]>
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index c1e5b12..b703dcc 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -731,14 +731,14 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
cachep = gfs2_glock_aspace_cachep;
else
cachep = gfs2_glock_cachep;
- gl = kmem_cache_alloc(cachep, GFP_KERNEL);
+ gl = kmem_cache_alloc(cachep, GFP_NOFS);
if (!gl)
return -ENOMEM;
memset(&gl->gl_lksb, 0, sizeof(struct dlm_lksb));
if (glops->go_flags & GLOF_LVB) {
- gl->gl_lksb.sb_lvbptr = kzalloc(GFS2_MIN_LVB_SIZE, GFP_KERNEL);
+ gl->gl_lksb.sb_lvbptr = kzalloc(GFS2_MIN_LVB_SIZE, GFP_NOFS);
if (!gl->gl_lksb.sb_lvbptr) {
kmem_cache_free(cachep, gl);
return -ENOMEM;
--
1.8.3.1
From: Bob Peterson <[email protected]>
This patch allows flock glocks to use a non-blocking dequeue rather
than dq_wait. It also reverts the previous patch I had posted regarding
dq_wait. The reverted patch isn't necessarily a bad idea, but I decided
this might avoid unforeseen side effects, and was therefore safer.
Signed-off-by: Bob Peterson <[email protected]>
Signed-off-by: Steven Whitehouse <[email protected]>
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index 4fc3a30..491e8e0 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -991,7 +991,7 @@ static int do_flock(struct file *file, int cmd, struct file_lock *fl)
goto out;
flock_lock_file_wait(file,
&(struct file_lock){.fl_type = F_UNLCK});
- gfs2_glock_dq_wait(fl_gh);
+ gfs2_glock_dq(fl_gh);
gfs2_holder_reinit(state, flags, fl_gh);
} else {
error = gfs2_glock_get(GFS2_SB(&ip->i_inode), ip->i_no_addr,
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index b703dcc..ee4e04f 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -1128,9 +1128,7 @@ void gfs2_glock_dq_wait(struct gfs2_holder *gh)
struct gfs2_glock *gl = gh->gh_gl;
gfs2_glock_dq(gh);
might_sleep();
- if (!find_first_holder(gl))
- wait_on_bit(&gl->gl_flags, GLF_DEMOTE, gfs2_glock_demote_wait,
- TASK_UNINTERRUPTIBLE);
+ wait_on_bit(&gl->gl_flags, GLF_DEMOTE, gfs2_glock_demote_wait, TASK_UNINTERRUPTIBLE);
}
/**
--
1.8.3.1
From: Geert Uytterhoeven <[email protected]>
Signed-off-by: Geert Uytterhoeven <[email protected]>
Cc: [email protected]
Signed-off-by: Steven Whitehouse <[email protected]>
diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c
index fc11007..2ffc67d 100644
--- a/fs/gfs2/glops.c
+++ b/fs/gfs2/glops.c
@@ -234,8 +234,8 @@ static void inode_go_sync(struct gfs2_glock *gl)
* inode_go_inval - prepare a inode glock to be released
* @gl: the glock
* @flags:
- *
- * Normally we invlidate everything, but if we are moving into
+ *
+ * Normally we invalidate everything, but if we are moving into
* LM_ST_DEFERRED from LM_ST_SHARED or LM_ST_EXCLUSIVE then we
* can keep hold of the metadata, since it won't have changed.
*
--
1.8.3.1
From: Fabian Frederick <[email protected]>
Cc: [email protected]
Signed-off-by: Fabian Frederick <[email protected]>
Signed-off-by: Steven Whitehouse <[email protected]>
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index db629d1..f4cb9c0 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -337,7 +337,7 @@ static bool gfs2_unaligned_extlen(struct gfs2_rbm *rbm, u32 n_unaligned, u32 *le
/**
* gfs2_free_extlen - Return extent length of free blocks
- * @rbm: Starting position
+ * @rrbm: Starting position
* @len: Max length to check
*
* Starting at the block specified by the rbm, see how many free blocks
@@ -2522,7 +2522,7 @@ void gfs2_rlist_alloc(struct gfs2_rgrp_list *rlist, unsigned int state)
/**
* gfs2_rlist_free - free a resource group list
- * @list: the list of resource groups
+ * @rlist: the list of resource groups
*
*/
--
1.8.3.1
From: Bob Peterson <[email protected]>
This patch removes the GLF_NOCACHE flag from the glocks associated with
flocks. There should be no good reason not to cache glocks for flocks:
they only force the glock to be demoted before they can be reacquired,
which can slow down performance and even cause glock hangs, especially
in cases where the flocks are held in Shared (SH) mode.
Signed-off-by: Bob Peterson <[email protected]>
Signed-off-by: Steven Whitehouse <[email protected]>
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index 491e8e0..26b3f95 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -981,7 +981,7 @@ static int do_flock(struct file *file, int cmd, struct file_lock *fl)
int error = 0;
state = (fl->fl_type == F_WRLCK) ? LM_ST_EXCLUSIVE : LM_ST_SHARED;
- flags = (IS_SETLKW(cmd) ? 0 : LM_FLAG_TRY) | GL_EXACT | GL_NOCACHE;
+ flags = (IS_SETLKW(cmd) ? 0 : LM_FLAG_TRY) | GL_EXACT;
mutex_lock(&fp->f_fl_mutex);
--
1.8.3.1
From: Fabian Frederick <[email protected]>
kcalloc manages count*sizeof overflow.
Cc: [email protected]
Signed-off-by: Fabian Frederick <[email protected]>
Signed-off-by: Steven Whitehouse <[email protected]>
diff --git a/fs/gfs2/lock_dlm.c b/fs/gfs2/lock_dlm.c
index 91f274d..4fafea1 100644
--- a/fs/gfs2/lock_dlm.c
+++ b/fs/gfs2/lock_dlm.c
@@ -1036,8 +1036,8 @@ static int set_recover_size(struct gfs2_sbd *sdp, struct dlm_slot *slots,
new_size = old_size + RECOVER_SIZE_INC;
- submit = kzalloc(new_size * sizeof(uint32_t), GFP_NOFS);
- result = kzalloc(new_size * sizeof(uint32_t), GFP_NOFS);
+ submit = kcalloc(new_size, sizeof(uint32_t), GFP_NOFS);
+ result = kcalloc(new_size, sizeof(uint32_t), GFP_NOFS);
if (!submit || !result) {
kfree(submit);
kfree(result);
--
1.8.3.1
From: Bob Peterson <[email protected]>
Function gfs2_glock_dq_wait is supposed to dequeue a glock and then
wait for the lock to be demoted. The problem is, if this is a shared
lock, its demote will depend on the other holders, which means you
might end up waiting forever because the other process is blocked.
This problem is especially apparent when dealing with nested flocks.
Signed-off-by: Bob Peterson <[email protected]>
Signed-off-by: Steven Whitehouse <[email protected]>
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index c355f73..278fae5 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -1128,7 +1128,9 @@ void gfs2_glock_dq_wait(struct gfs2_holder *gh)
struct gfs2_glock *gl = gh->gh_gl;
gfs2_glock_dq(gh);
might_sleep();
- wait_on_bit(&gl->gl_flags, GLF_DEMOTE, gfs2_glock_demote_wait, TASK_UNINTERRUPTIBLE);
+ if (!find_first_holder(gl))
+ wait_on_bit(&gl->gl_flags, GLF_DEMOTE, gfs2_glock_demote_wait,
+ TASK_UNINTERRUPTIBLE);
}
/**
--
1.8.3.1