summaryrefslogtreecommitdiffstats
path: root/fs/gfs2/glock.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/gfs2/glock.c')
-rw-r--r--fs/gfs2/glock.c286
1 files changed, 198 insertions, 88 deletions
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 959a19ced4d5..11066d8647d2 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -15,6 +15,7 @@
#include <linux/buffer_head.h>
#include <linux/delay.h>
#include <linux/sort.h>
+#include <linux/hash.h>
#include <linux/jhash.h>
#include <linux/kallsyms.h>
#include <linux/gfs2_ondisk.h>
@@ -71,7 +72,7 @@ static DEFINE_SPINLOCK(lru_lock);
#define GFS2_GL_HASH_SHIFT 15
#define GFS2_GL_HASH_SIZE BIT(GFS2_GL_HASH_SHIFT)
-static struct rhashtable_params ht_parms = {
+static const struct rhashtable_params ht_parms = {
.nelem_hint = GFS2_GL_HASH_SIZE * 3 / 4,
.key_len = offsetofend(struct lm_lockname, ln_type),
.key_offset = offsetof(struct gfs2_glock, gl_name),
@@ -80,9 +81,52 @@ static struct rhashtable_params ht_parms = {
static struct rhashtable gl_hash_table;
-void gfs2_glock_free(struct gfs2_glock *gl)
+#define GLOCK_WAIT_TABLE_BITS 12
+#define GLOCK_WAIT_TABLE_SIZE (1 << GLOCK_WAIT_TABLE_BITS)
+static wait_queue_head_t glock_wait_table[GLOCK_WAIT_TABLE_SIZE] __cacheline_aligned;
+
+struct wait_glock_queue {
+ struct lm_lockname *name;
+ wait_queue_entry_t wait;
+};
+
+static int glock_wake_function(wait_queue_entry_t *wait, unsigned int mode,
+ int sync, void *key)
{
- struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
+ struct wait_glock_queue *wait_glock =
+ container_of(wait, struct wait_glock_queue, wait);
+ struct lm_lockname *wait_name = wait_glock->name;
+ struct lm_lockname *wake_name = key;
+
+ if (wake_name->ln_sbd != wait_name->ln_sbd ||
+ wake_name->ln_number != wait_name->ln_number ||
+ wake_name->ln_type != wait_name->ln_type)
+ return 0;
+ return autoremove_wake_function(wait, mode, sync, key);
+}
+
+static wait_queue_head_t *glock_waitqueue(struct lm_lockname *name)
+{
+ u32 hash = jhash2((u32 *)name, sizeof(*name) / 4, 0);
+
+ return glock_wait_table + hash_32(hash, GLOCK_WAIT_TABLE_BITS);
+}
+
+/**
+ * wake_up_glock - Wake up waiters on a glock
+ * @gl: the glock
+ */
+static void wake_up_glock(struct gfs2_glock *gl)
+{
+ wait_queue_head_t *wq = glock_waitqueue(&gl->gl_name);
+
+ if (waitqueue_active(wq))
+ __wake_up(wq, TASK_NORMAL, 1, &gl->gl_name);
+}
+
+static void gfs2_glock_dealloc(struct rcu_head *rcu)
+{
+ struct gfs2_glock *gl = container_of(rcu, struct gfs2_glock, gl_rcu);
if (gl->gl_ops->go_flags & GLOF_ASPACE) {
kmem_cache_free(gfs2_glock_aspace_cachep, gl);
@@ -90,6 +134,16 @@ void gfs2_glock_free(struct gfs2_glock *gl)
kfree(gl->gl_lksb.sb_lvbptr);
kmem_cache_free(gfs2_glock_cachep, gl);
}
+}
+
+void gfs2_glock_free(struct gfs2_glock *gl)
+{
+ struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
+
+ rhashtable_remove_fast(&gl_hash_table, &gl->gl_node, ht_parms);
+ smp_mb();
+ wake_up_glock(gl);
+ call_rcu(&gl->gl_rcu, gfs2_glock_dealloc);
if (atomic_dec_and_test(&sdp->sd_glock_disposal))
wake_up(&sdp->sd_glock_wait);
}
@@ -100,7 +154,7 @@ void gfs2_glock_free(struct gfs2_glock *gl)
*
*/
-static void gfs2_glock_hold(struct gfs2_glock *gl)
+void gfs2_glock_hold(struct gfs2_glock *gl)
{
GLOCK_BUG_ON(gl, __lockref_is_dead(&gl->gl_lockref));
lockref_get(&gl->gl_lockref);
@@ -143,6 +197,9 @@ void gfs2_glock_add_to_lru(struct gfs2_glock *gl)
static void gfs2_glock_remove_from_lru(struct gfs2_glock *gl)
{
+ if (!(gl->gl_ops->go_flags & GLOF_LRU))
+ return;
+
spin_lock(&lru_lock);
if (!list_empty(&gl->gl_lru)) {
list_del_init(&gl->gl_lru);
@@ -152,31 +209,66 @@ static void gfs2_glock_remove_from_lru(struct gfs2_glock *gl)
spin_unlock(&lru_lock);
}
-/**
- * gfs2_glock_put() - Decrement reference count on glock
- * @gl: The glock to put
- *
+/*
+ * Enqueue the glock on the work queue. Passes one glock reference on to the
+ * work queue.
*/
+static void __gfs2_glock_queue_work(struct gfs2_glock *gl, unsigned long delay) {
+ if (!queue_delayed_work(glock_workqueue, &gl->gl_work, delay)) {
+ /*
+ * We are holding the lockref spinlock, and the work was still
+ * queued above. The queued work (glock_work_func) takes that
+ * spinlock before dropping its glock reference(s), so it
+ * cannot have dropped them in the meantime.
+ */
+ GLOCK_BUG_ON(gl, gl->gl_lockref.count < 2);
+ gl->gl_lockref.count--;
+ }
+}
-void gfs2_glock_put(struct gfs2_glock *gl)
+static void gfs2_glock_queue_work(struct gfs2_glock *gl, unsigned long delay) {
+ spin_lock(&gl->gl_lockref.lock);
+ __gfs2_glock_queue_work(gl, delay);
+ spin_unlock(&gl->gl_lockref.lock);
+}
+
+static void __gfs2_glock_put(struct gfs2_glock *gl)
{
struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
struct address_space *mapping = gfs2_glock2aspace(gl);
- if (lockref_put_or_lock(&gl->gl_lockref))
- return;
-
lockref_mark_dead(&gl->gl_lockref);
gfs2_glock_remove_from_lru(gl);
spin_unlock(&gl->gl_lockref.lock);
- rhashtable_remove_fast(&gl_hash_table, &gl->gl_node, ht_parms);
GLOCK_BUG_ON(gl, !list_empty(&gl->gl_holders));
GLOCK_BUG_ON(gl, mapping && mapping->nrpages);
trace_gfs2_glock_put(gl);
sdp->sd_lockstruct.ls_ops->lm_put_lock(gl);
}
+/*
+ * Cause the glock to be put in work queue context.
+ */
+void gfs2_glock_queue_put(struct gfs2_glock *gl)
+{
+ gfs2_glock_queue_work(gl, 0);
+}
+
+/**
+ * gfs2_glock_put() - Decrement reference count on glock
+ * @gl: The glock to put
+ *
+ */
+
+void gfs2_glock_put(struct gfs2_glock *gl)
+{
+ if (lockref_put_or_lock(&gl->gl_lockref))
+ return;
+
+ __gfs2_glock_put(gl);
+}
+
/**
* may_grant - check if its ok to grant a new lock
* @gl: The glock
@@ -482,8 +574,7 @@ __acquires(&gl->gl_lockref.lock)
target == LM_ST_UNLOCKED &&
test_bit(SDF_SKIP_DLM_UNLOCK, &sdp->sd_flags)) {
finish_xmote(gl, target);
- if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
- gfs2_glock_put(gl);
+ gfs2_glock_queue_work(gl, 0);
}
else if (ret) {
pr_err("lm_lock ret %d\n", ret);
@@ -492,8 +583,7 @@ __acquires(&gl->gl_lockref.lock)
}
} else { /* lock_nolock */
finish_xmote(gl, target);
- if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
- gfs2_glock_put(gl);
+ gfs2_glock_queue_work(gl, 0);
}
spin_lock(&gl->gl_lockref.lock);
@@ -565,8 +655,7 @@ out_sched:
clear_bit(GLF_LOCK, &gl->gl_flags);
smp_mb__after_atomic();
gl->gl_lockref.count++;
- if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
- gl->gl_lockref.count--;
+ __gfs2_glock_queue_work(gl, 0);
return;
out_unlock:
@@ -601,11 +690,11 @@ static void glock_work_func(struct work_struct *work)
{
unsigned long delay = 0;
struct gfs2_glock *gl = container_of(work, struct gfs2_glock, gl_work.work);
- int drop_ref = 0;
+ unsigned int drop_refs = 1;
if (test_and_clear_bit(GLF_REPLY_PENDING, &gl->gl_flags)) {
finish_xmote(gl, gl->gl_reply);
- drop_ref = 1;
+ drop_refs++;
}
spin_lock(&gl->gl_lockref.lock);
if (test_bit(GLF_PENDING_DEMOTE, &gl->gl_flags) &&
@@ -623,17 +712,59 @@ static void glock_work_func(struct work_struct *work)
}
}
run_queue(gl, 0);
- spin_unlock(&gl->gl_lockref.lock);
- if (!delay)
- gfs2_glock_put(gl);
- else {
+ if (delay) {
+ /* Keep one glock reference for the work we requeue. */
+ drop_refs--;
if (gl->gl_name.ln_type != LM_TYPE_INODE)
delay = 0;
- if (queue_delayed_work(glock_workqueue, &gl->gl_work, delay) == 0)
- gfs2_glock_put(gl);
+ __gfs2_glock_queue_work(gl, delay);
}
- if (drop_ref)
- gfs2_glock_put(gl);
+
+ /*
+ * Drop the remaining glock references manually here. (Mind that
+ * __gfs2_glock_queue_work depends on the lockref spinlock begin held
+ * here as well.)
+ */
+ gl->gl_lockref.count -= drop_refs;
+ if (!gl->gl_lockref.count) {
+ __gfs2_glock_put(gl);
+ return;
+ }
+ spin_unlock(&gl->gl_lockref.lock);
+}
+
+static struct gfs2_glock *find_insert_glock(struct lm_lockname *name,
+ struct gfs2_glock *new)
+{
+ struct wait_glock_queue wait;
+ wait_queue_head_t *wq = glock_waitqueue(name);
+ struct gfs2_glock *gl;
+
+ wait.name = name;
+ init_wait(&wait.wait);
+ wait.wait.func = glock_wake_function;
+
+again:
+ prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
+ rcu_read_lock();
+ if (new) {
+ gl = rhashtable_lookup_get_insert_fast(&gl_hash_table,
+ &new->gl_node, ht_parms);
+ if (IS_ERR(gl))
+ goto out;
+ } else {
+ gl = rhashtable_lookup_fast(&gl_hash_table,
+ name, ht_parms);
+ }
+ if (gl && !lockref_get_not_dead(&gl->gl_lockref)) {
+ rcu_read_unlock();
+ schedule();
+ goto again;
+ }
+out:
+ rcu_read_unlock();
+ finish_wait(wq, &wait.wait);
+ return gl;
}
/**
@@ -662,15 +793,11 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
struct kmem_cache *cachep;
int ret = 0;
- rcu_read_lock();
- gl = rhashtable_lookup_fast(&gl_hash_table, &name, ht_parms);
- if (gl && !lockref_get_not_dead(&gl->gl_lockref))
- gl = NULL;
- rcu_read_unlock();
-
- *glp = gl;
- if (gl)
+ gl = find_insert_glock(&name, NULL);
+ if (gl) {
+ *glp = gl;
return 0;
+ }
if (!create)
return -ENOENT;
@@ -724,10 +851,7 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
mapping->writeback_index = 0;
}
-again:
- rcu_read_lock();
- tmp = rhashtable_lookup_get_insert_fast(&gl_hash_table, &gl->gl_node,
- ht_parms);
+ tmp = find_insert_glock(&name, gl);
if (!tmp) {
*glp = gl;
goto out;
@@ -736,13 +860,7 @@ again:
ret = PTR_ERR(tmp);
goto out_free;
}
- if (lockref_get_not_dead(&tmp->gl_lockref)) {
- *glp = tmp;
- goto out_free;
- }
- rcu_read_unlock();
- cond_resched();
- goto again;
+ *glp = tmp;
out_free:
kfree(gl->gl_lksb.sb_lvbptr);
@@ -750,7 +868,6 @@ out_free:
atomic_dec(&sdp->sd_glock_disposal);
out:
- rcu_read_unlock();
return ret;
}
@@ -986,8 +1103,7 @@ int gfs2_glock_nq(struct gfs2_holder *gh)
test_and_clear_bit(GLF_FROZEN, &gl->gl_flags))) {
set_bit(GLF_REPLY_PENDING, &gl->gl_flags);
gl->gl_lockref.count++;
- if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
- gl->gl_lockref.count--;
+ __gfs2_glock_queue_work(gl, 0);
}
run_queue(gl, 1);
spin_unlock(&gl->gl_lockref.lock);
@@ -1047,17 +1163,15 @@ void gfs2_glock_dq(struct gfs2_holder *gh)
gfs2_glock_add_to_lru(gl);
trace_gfs2_glock_queue(gh, 0);
+ if (unlikely(!fast_path)) {
+ gl->gl_lockref.count++;
+ if (test_bit(GLF_PENDING_DEMOTE, &gl->gl_flags) &&
+ !test_bit(GLF_DEMOTE, &gl->gl_flags) &&
+ gl->gl_name.ln_type == LM_TYPE_INODE)
+ delay = gl->gl_hold_time;
+ __gfs2_glock_queue_work(gl, delay);
+ }
spin_unlock(&gl->gl_lockref.lock);
- if (likely(fast_path))
- return;
-
- gfs2_glock_hold(gl);
- if (test_bit(GLF_PENDING_DEMOTE, &gl->gl_flags) &&
- !test_bit(GLF_DEMOTE, &gl->gl_flags) &&
- gl->gl_name.ln_type == LM_TYPE_INODE)
- delay = gl->gl_hold_time;
- if (queue_delayed_work(glock_workqueue, &gl->gl_work, delay) == 0)
- gfs2_glock_put(gl);
}
void gfs2_glock_dq_wait(struct gfs2_holder *gh)
@@ -1233,9 +1347,8 @@ void gfs2_glock_cb(struct gfs2_glock *gl, unsigned int state)
spin_lock(&gl->gl_lockref.lock);
handle_callback(gl, state, delay, true);
+ __gfs2_glock_queue_work(gl, delay);
spin_unlock(&gl->gl_lockref.lock);
- if (queue_delayed_work(glock_workqueue, &gl->gl_work, delay) == 0)
- gfs2_glock_put(gl);
}
/**
@@ -1294,10 +1407,8 @@ void gfs2_glock_complete(struct gfs2_glock *gl, int ret)
gl->gl_lockref.count++;
set_bit(GLF_REPLY_PENDING, &gl->gl_flags);
+ __gfs2_glock_queue_work(gl, 0);
spin_unlock(&gl->gl_lockref.lock);
-
- if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
- gfs2_glock_put(gl);
}
static int glock_cmp(void *priv, struct list_head *a, struct list_head *b)
@@ -1355,8 +1466,7 @@ add_back_to_lru:
if (demote_ok(gl))
handle_callback(gl, LM_ST_UNLOCKED, 0, false);
WARN_ON(!test_and_clear_bit(GLF_LOCK, &gl->gl_flags));
- if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
- gl->gl_lockref.count--;
+ __gfs2_glock_queue_work(gl, 0);
spin_unlock(&gl->gl_lockref.lock);
cond_resched_lock(&lru_lock);
}
@@ -1440,14 +1550,15 @@ static void glock_hash_walk(glock_examiner examiner, const struct gfs2_sbd *sdp)
do {
gl = ERR_PTR(rhashtable_walk_start(&iter));
- if (gl)
- continue;
+ if (IS_ERR(gl))
+ goto walk_stop;
while ((gl = rhashtable_walk_next(&iter)) && !IS_ERR(gl))
- if ((gl->gl_name.ln_sbd == sdp) &&
+ if (gl->gl_name.ln_sbd == sdp &&
lockref_get_not_dead(&gl->gl_lockref))
examiner(gl);
+walk_stop:
rhashtable_walk_stop(&iter);
} while (cond_resched(), gl == ERR_PTR(-EAGAIN));
@@ -1462,13 +1573,12 @@ static void glock_hash_walk(glock_examiner examiner, const struct gfs2_sbd *sdp)
static void thaw_glock(struct gfs2_glock *gl)
{
- if (!test_and_clear_bit(GLF_FROZEN, &gl->gl_flags))
- goto out;
- set_bit(GLF_REPLY_PENDING, &gl->gl_flags);
- if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0) {
-out:
+ if (!test_and_clear_bit(GLF_FROZEN, &gl->gl_flags)) {
gfs2_glock_put(gl);
+ return;
}
+ set_bit(GLF_REPLY_PENDING, &gl->gl_flags);
+ gfs2_glock_queue_work(gl, 0);
}
/**
@@ -1484,9 +1594,8 @@ static void clear_glock(struct gfs2_glock *gl)
spin_lock(&gl->gl_lockref.lock);
if (gl->gl_state != LM_ST_UNLOCKED)
handle_callback(gl, LM_ST_UNLOCKED, 0, false);
+ __gfs2_glock_queue_work(gl, 0);
spin_unlock(&gl->gl_lockref.lock);
- if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
- gfs2_glock_put(gl);
}
/**
@@ -1772,7 +1881,7 @@ static int gfs2_sbstats_seq_show(struct seq_file *seq, void *iter_ptr)
int __init gfs2_glock_init(void)
{
- int ret;
+ int i, ret;
ret = rhashtable_init(&gl_hash_table, &ht_parms);
if (ret < 0)
@@ -1801,6 +1910,9 @@ int __init gfs2_glock_init(void)
return ret;
}
+ for (i = 0; i < GLOCK_WAIT_TABLE_SIZE; i++)
+ init_waitqueue_head(glock_wait_table + i);
+
return 0;
}
@@ -1829,16 +1941,13 @@ static void gfs2_glock_iter_next(struct gfs2_glock_iter *gi)
}
static void *gfs2_glock_seq_start(struct seq_file *seq, loff_t *pos)
+ __acquires(RCU)
{
struct gfs2_glock_iter *gi = seq->private;
loff_t n = *pos;
- int ret;
- if (gi->last_pos <= *pos)
- n = (*pos - gi->last_pos);
-
- ret = rhashtable_walk_start(&gi->hti);
- if (ret)
+ rhashtable_walk_enter(&gl_hash_table, &gi->hti);
+ if (rhashtable_walk_start(&gi->hti) != 0)
return NULL;
do {
@@ -1846,6 +1955,7 @@ static void *gfs2_glock_seq_start(struct seq_file *seq, loff_t *pos)
} while (gi->gl && n--);
gi->last_pos = *pos;
+
return gi->gl;
}
@@ -1857,15 +1967,18 @@ static void *gfs2_glock_seq_next(struct seq_file *seq, void *iter_ptr,
(*pos)++;
gi->last_pos = *pos;
gfs2_glock_iter_next(gi);
+
return gi->gl;
}
static void gfs2_glock_seq_stop(struct seq_file *seq, void *iter_ptr)
+ __releases(RCU)
{
struct gfs2_glock_iter *gi = seq->private;
gi->gl = NULL;
rhashtable_walk_stop(&gi->hti);
+ rhashtable_walk_exit(&gi->hti);
}
static int gfs2_glock_seq_show(struct seq_file *seq, void *iter_ptr)
@@ -1928,12 +2041,10 @@ static int __gfs2_glocks_open(struct inode *inode, struct file *file,
struct gfs2_glock_iter *gi = seq->private;
gi->sdp = inode->i_private;
- gi->last_pos = 0;
seq->buf = kmalloc(GFS2_SEQ_GOODSIZE, GFP_KERNEL | __GFP_NOWARN);
if (seq->buf)
seq->size = GFS2_SEQ_GOODSIZE;
gi->gl = NULL;
- rhashtable_walk_enter(&gl_hash_table, &gi->hti);
}
return ret;
}
@@ -1949,7 +2060,6 @@ static int gfs2_glocks_release(struct inode *inode, struct file *file)
struct gfs2_glock_iter *gi = seq->private;
gi->gl = NULL;
- rhashtable_walk_exit(&gi->hti);
return seq_release_private(inode, file);
}
OpenPOWER on IntegriCloud