[linux-2.6-block.git] / fs / btrfs / locking.c

/*
 * Copyright (C) 2008 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */
#include <linux/sched.h>
#include <linux/pagemap.h>
#include <linux/spinlock.h>
#include <linux/page-flags.h>
#include <asm/bug.h>
#include "ctree.h"
#include "extent_io.h"
#include "locking.h"

static void btrfs_assert_tree_read_locked(struct extent_buffer *eb);

/*
 * if we currently have a spinning reader or writer lock
 * (indicated by the rw flag) this will bump the count
 * of blocking holders and drop the spinlock.
 */
void btrfs_set_lock_blocking_rw(struct extent_buffer *eb, int rw)
{
	/*
	 * no lock is required.  The lock owner may change if
	 * we have a read lock, but it won't change to or away
	 * from us.  If we have the write lock, we are the owner
	 * and it'll never change.
	 */
	if (eb->lock_nested && current->pid == eb->lock_owner)
		return;
	if (rw == BTRFS_WRITE_LOCK) {
		if (atomic_read(&eb->blocking_writers) == 0) {
			WARN_ON(atomic_read(&eb->spinning_writers) != 1);
			atomic_dec(&eb->spinning_writers);
			btrfs_assert_tree_locked(eb);
			atomic_inc(&eb->blocking_writers);
			write_unlock(&eb->lock);
		}
	} else if (rw == BTRFS_READ_LOCK) {
		btrfs_assert_tree_read_locked(eb);
		atomic_inc(&eb->blocking_readers);
		WARN_ON(atomic_read(&eb->spinning_readers) == 0);
		atomic_dec(&eb->spinning_readers);
		read_unlock(&eb->lock);
	}
	return;
}

/*
 * if we currently have a blocking lock, take the spinlock
 * and drop our blocking count
 */
void btrfs_clear_lock_blocking_rw(struct extent_buffer *eb, int rw)
{
	/*
	 * no lock is required.  The lock owner may change if
	 * we have a read lock, but it won't change to or away
	 * from us.  If we have the write lock, we are the owner
	 * and it'll never change.
	 */
	if (eb->lock_nested && current->pid == eb->lock_owner)
		return;

	if (rw == BTRFS_WRITE_LOCK_BLOCKING) {
		BUG_ON(atomic_read(&eb->blocking_writers) != 1);
		write_lock(&eb->lock);
		WARN_ON(atomic_read(&eb->spinning_writers));
		atomic_inc(&eb->spinning_writers);
		if (atomic_dec_and_test(&eb->blocking_writers) &&
		    waitqueue_active(&eb->write_lock_wq))
			wake_up(&eb->write_lock_wq);
	} else if (rw == BTRFS_READ_LOCK_BLOCKING) {
		BUG_ON(atomic_read(&eb->blocking_readers) == 0);
		read_lock(&eb->lock);
		atomic_inc(&eb->spinning_readers);
		if (atomic_dec_and_test(&eb->blocking_readers) &&
		    waitqueue_active(&eb->read_lock_wq))
			wake_up(&eb->read_lock_wq);
	}
	return;
}

/*
 * take a spinning read lock.  This will wait for any blocking
 * writers
 */
void btrfs_tree_read_lock(struct extent_buffer *eb)
{
again:
	BUG_ON(!atomic_read(&eb->blocking_writers) &&
	       current->pid == eb->lock_owner);

	read_lock(&eb->lock);
	if (atomic_read(&eb->blocking_writers) &&
	    current->pid == eb->lock_owner) {
		/*
		 * This extent is already write-locked by our thread. We allow
		 * an additional read lock to be added because it's for the same
		 * thread. btrfs_find_all_roots() depends on this as it may be
		 * called on a partly (write-)locked tree.
		 */
		BUG_ON(eb->lock_nested);
		eb->lock_nested = 1;
		read_unlock(&eb->lock);
		return;
	}
	if (atomic_read(&eb->blocking_writers)) {
		read_unlock(&eb->lock);
		wait_event(eb->write_lock_wq,
			   atomic_read(&eb->blocking_writers) == 0);
		goto again;
	}
	atomic_inc(&eb->read_locks);
	atomic_inc(&eb->spinning_readers);
}

/*
 * take a spinning read lock.
 * returns 1 if we get the read lock and 0 if we don't
 * this won't wait for blocking writers
 */
int btrfs_tree_read_lock_atomic(struct extent_buffer *eb)
{
	if (atomic_read(&eb->blocking_writers))
		return 0;

	read_lock(&eb->lock);
	if (atomic_read(&eb->blocking_writers)) {
		read_unlock(&eb->lock);
		return 0;
	}
	atomic_inc(&eb->read_locks);
	atomic_inc(&eb->spinning_readers);
	return 1;
}

/*
 * returns 1 if we get the read lock and 0 if we don't
 * this won't wait for blocking writers
 */
int btrfs_try_tree_read_lock(struct extent_buffer *eb)
{
	if (atomic_read(&eb->blocking_writers))
		return 0;

	if (!read_trylock(&eb->lock))
		return 0;

	if (atomic_read(&eb->blocking_writers)) {
		read_unlock(&eb->lock);
		return 0;
	}
	atomic_inc(&eb->read_locks);
	atomic_inc(&eb->spinning_readers);
	return 1;
}

/*
 * returns 1 if we get the read lock and 0 if we don't
 * this won't wait for blocking writers or readers
 */
int btrfs_try_tree_write_lock(struct extent_buffer *eb)
{
	if (atomic_read(&eb->blocking_writers) ||
	    atomic_read(&eb->blocking_readers))
		return 0;

	write_lock(&eb->lock);
	if (atomic_read(&eb->blocking_writers) ||
	    atomic_read(&eb->blocking_readers)) {
		write_unlock(&eb->lock);
		return 0;
	}
	atomic_inc(&eb->write_locks);
	atomic_inc(&eb->spinning_writers);
	eb->lock_owner = current->pid;
	return 1;
}

/*
 * drop a spinning read lock
 */
void btrfs_tree_read_unlock(struct extent_buffer *eb)
{
	/*
	 * if we're nested, we have the write lock.  No new locking
	 * is needed as long as we are the lock owner.
	 * The write unlock will do a barrier for us, and the lock_nested
	 * field only matters to the lock owner.
	 */
	if (eb->lock_nested && current->pid == eb->lock_owner) {
		eb->lock_nested = 0;
		return;
	}
	btrfs_assert_tree_read_locked(eb);
	WARN_ON(atomic_read(&eb->spinning_readers) == 0);
	atomic_dec(&eb->spinning_readers);
	atomic_dec(&eb->read_locks);
	read_unlock(&eb->lock);
}

/*
 * drop a blocking read lock
 */
void btrfs_tree_read_unlock_blocking(struct extent_buffer *eb)
{
	/*
	 * if we're nested, we have the write lock.  No new locking
	 * is needed as long as we are the lock owner.
	 * The write unlock will do a barrier for us, and the lock_nested
	 * field only matters to the lock owner.
	 */
	if (eb->lock_nested && current->pid == eb->lock_owner) {
		eb->lock_nested = 0;
		return;
	}
	btrfs_assert_tree_read_locked(eb);
	WARN_ON(atomic_read(&eb->blocking_readers) == 0);
	if (atomic_dec_and_test(&eb->blocking_readers) &&
	    waitqueue_active(&eb->read_lock_wq))
		wake_up(&eb->read_lock_wq);
	atomic_dec(&eb->read_locks);
}

/*
 * take a spinning write lock.  This will wait for both
 * blocking readers or writers
 */
void btrfs_tree_lock(struct extent_buffer *eb)
{
again:
	wait_event(eb->read_lock_wq, atomic_read(&eb->blocking_readers) == 0);
	wait_event(eb->write_lock_wq, atomic_read(&eb->blocking_writers) == 0);
	write_lock(&eb->lock);
	if (atomic_read(&eb->blocking_readers)) {
		write_unlock(&eb->lock);
		wait_event(eb->read_lock_wq,
			   atomic_read(&eb->blocking_readers) == 0);
		goto again;
	}
	if (atomic_read(&eb->blocking_writers)) {
		write_unlock(&eb->lock);
		wait_event(eb->write_lock_wq,
			   atomic_read(&eb->blocking_writers) == 0);
		goto again;
	}
	WARN_ON(atomic_read(&eb->spinning_writers));
	atomic_inc(&eb->spinning_writers);
	atomic_inc(&eb->write_locks);
	eb->lock_owner = current->pid;
}

/*
 * drop a spinning or a blocking write lock.
 */
void btrfs_tree_unlock(struct extent_buffer *eb)
{
	int blockers = atomic_read(&eb->blocking_writers);

	BUG_ON(blockers > 1);

	btrfs_assert_tree_locked(eb);
	eb->lock_owner = 0;
	atomic_dec(&eb->write_locks);

	if (blockers) {
		WARN_ON(atomic_read(&eb->spinning_writers));
		atomic_dec(&eb->blocking_writers);
		smp_mb();
		if (waitqueue_active(&eb->write_lock_wq))
			wake_up(&eb->write_lock_wq);
	} else {
		WARN_ON(atomic_read(&eb->spinning_writers) != 1);
		atomic_dec(&eb->spinning_writers);
		write_unlock(&eb->lock);
	}
}

void btrfs_assert_tree_locked(struct extent_buffer *eb)
{
	BUG_ON(!atomic_read(&eb->write_locks));
}

static void btrfs_assert_tree_read_locked(struct extent_buffer *eb)
{
	BUG_ON(!atomic_read(&eb->read_locks));
}
Commit	Line	Data
925baedd CM	1	/*
	2	* Copyright (C) 2008 Oracle. All rights reserved.
	3	*
	4	* This program is free software; you can redistribute it and/or
	5	* modify it under the terms of the GNU General Public
	6	* License v2 as published by the Free Software Foundation.
	7	*
	8	* This program is distributed in the hope that it will be useful,
	9	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	10	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	11	* General Public License for more details.
	12	*
	13	* You should have received a copy of the GNU General Public
	14	* License along with this program; if not, write to the
	15	* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
	16	* Boston, MA 021110-1307, USA.
	17	*/
	18	#include <linux/sched.h>
925baedd CM	19	#include <linux/pagemap.h>
	20	#include <linux/spinlock.h>
	21	#include <linux/page-flags.h>
4881ee5a	22	#include <asm/bug.h>
925baedd CM	23	#include "ctree.h"
	24	#include "extent_io.h"
	25	#include "locking.h"
	26
48a3b636	27	static void btrfs_assert_tree_read_locked(struct extent_buffer *eb);
d397712b	28
b4ce94de	29	/*
bd681513 CM	30	* if we currently have a spinning reader or writer lock
	31	* (indicated by the rw flag) this will bump the count
	32	* of blocking holders and drop the spinlock.
b4ce94de	33	*/
bd681513	34	void btrfs_set_lock_blocking_rw(struct extent_buffer *eb, int rw)
925baedd	35	{
ea4ebde0 CM	36	/*
	37	* no lock is required. The lock owner may change if
	38	* we have a read lock, but it won't change to or away
	39	* from us. If we have the write lock, we are the owner
	40	* and it'll never change.
	41	*/
	42	if (eb->lock_nested && current->pid == eb->lock_owner)
	43	return;
bd681513 CM	44	if (rw == BTRFS_WRITE_LOCK) {
	45	if (atomic_read(&eb->blocking_writers) == 0) {
	46	WARN_ON(atomic_read(&eb->spinning_writers) != 1);
	47	atomic_dec(&eb->spinning_writers);
	48	btrfs_assert_tree_locked(eb);
	49	atomic_inc(&eb->blocking_writers);
	50	write_unlock(&eb->lock);
	51	}
	52	} else if (rw == BTRFS_READ_LOCK) {
	53	btrfs_assert_tree_read_locked(eb);
	54	atomic_inc(&eb->blocking_readers);
	55	WARN_ON(atomic_read(&eb->spinning_readers) == 0);
	56	atomic_dec(&eb->spinning_readers);
	57	read_unlock(&eb->lock);
b4ce94de	58	}
bd681513	59	return;
b4ce94de	60	}
f9efa9c7	61
b4ce94de	62	/*
bd681513 CM	63	* if we currently have a blocking lock, take the spinlock
bd681513 CM	64	* and drop our blocking count
b4ce94de	65	*/
bd681513	66	void btrfs_clear_lock_blocking_rw(struct extent_buffer *eb, int rw)
b4ce94de	67	{
ea4ebde0 CM	68	/*
	69	* no lock is required. The lock owner may change if
	70	* we have a read lock, but it won't change to or away
	71	* from us. If we have the write lock, we are the owner
	72	* and it'll never change.
	73	*/
	74	if (eb->lock_nested && current->pid == eb->lock_owner)
	75	return;
	76
bd681513 CM	77	if (rw == BTRFS_WRITE_LOCK_BLOCKING) {
	78	BUG_ON(atomic_read(&eb->blocking_writers) != 1);
	79	write_lock(&eb->lock);
	80	WARN_ON(atomic_read(&eb->spinning_writers));
	81	atomic_inc(&eb->spinning_writers);
cbea5ac1 CM	82	if (atomic_dec_and_test(&eb->blocking_writers) &&
cbea5ac1 CM	83	waitqueue_active(&eb->write_lock_wq))
bd681513 CM	84	wake_up(&eb->write_lock_wq);
	85	} else if (rw == BTRFS_READ_LOCK_BLOCKING) {
	86	BUG_ON(atomic_read(&eb->blocking_readers) == 0);
	87	read_lock(&eb->lock);
	88	atomic_inc(&eb->spinning_readers);
cbea5ac1 CM	89	if (atomic_dec_and_test(&eb->blocking_readers) &&
cbea5ac1 CM	90	waitqueue_active(&eb->read_lock_wq))
bd681513	91	wake_up(&eb->read_lock_wq);
b4ce94de	92	}
bd681513	93	return;
b4ce94de CM	94	}
	95
	96	/*
bd681513 CM	97	* take a spinning read lock. This will wait for any blocking
bd681513 CM	98	* writers
b4ce94de	99	*/
bd681513	100	void btrfs_tree_read_lock(struct extent_buffer *eb)
b4ce94de	101	{
bd681513	102	again:
ea4ebde0 CM	103	BUG_ON(!atomic_read(&eb->blocking_writers) &&
	104	current->pid == eb->lock_owner);
	105
5b25f70f AJ	106	read_lock(&eb->lock);
	107	if (atomic_read(&eb->blocking_writers) &&
	108	current->pid == eb->lock_owner) {
	109	/*
	110	* This extent is already write-locked by our thread. We allow
	111	* an additional read lock to be added because it's for the same
	112	* thread. btrfs_find_all_roots() depends on this as it may be
	113	* called on a partly (write-)locked tree.
	114	*/
	115	BUG_ON(eb->lock_nested);
	116	eb->lock_nested = 1;
	117	read_unlock(&eb->lock);
	118	return;
	119	}
bd681513 CM	120	if (atomic_read(&eb->blocking_writers)) {
bd681513 CM	121	read_unlock(&eb->lock);
39f9d028 LB	122	wait_event(eb->write_lock_wq,
39f9d028 LB	123	atomic_read(&eb->blocking_writers) == 0);
bd681513	124	goto again;
b4ce94de	125	}
bd681513 CM	126	atomic_inc(&eb->read_locks);
bd681513 CM	127	atomic_inc(&eb->spinning_readers);
b4ce94de CM	128	}
b4ce94de CM	129
f82c458a CM	130	/*
	131	* take a spinning read lock.
	132	* returns 1 if we get the read lock and 0 if we don't
	133	* this won't wait for blocking writers
	134	*/
	135	int btrfs_tree_read_lock_atomic(struct extent_buffer *eb)
	136	{
	137	if (atomic_read(&eb->blocking_writers))
	138	return 0;
	139
	140	read_lock(&eb->lock);
	141	if (atomic_read(&eb->blocking_writers)) {
	142	read_unlock(&eb->lock);
	143	return 0;
	144	}
	145	atomic_inc(&eb->read_locks);
	146	atomic_inc(&eb->spinning_readers);
	147	return 1;
	148	}
	149
b4ce94de	150	/*
bd681513 CM	151	* returns 1 if we get the read lock and 0 if we don't
bd681513 CM	152	* this won't wait for blocking writers
b4ce94de	153	*/
bd681513	154	int btrfs_try_tree_read_lock(struct extent_buffer *eb)
b4ce94de	155	{
bd681513 CM	156	if (atomic_read(&eb->blocking_writers))
bd681513 CM	157	return 0;
b4ce94de	158
ea4ebde0 CM	159	if (!read_trylock(&eb->lock))
	160	return 0;
	161
bd681513 CM	162	if (atomic_read(&eb->blocking_writers)) {
	163	read_unlock(&eb->lock);
	164	return 0;
b9473439	165	}
bd681513 CM	166	atomic_inc(&eb->read_locks);
	167	atomic_inc(&eb->spinning_readers);
	168	return 1;
b4ce94de CM	169	}
	170
	171	/*
bd681513 CM	172	* returns 1 if we get the read lock and 0 if we don't
bd681513 CM	173	* this won't wait for blocking writers or readers
b4ce94de	174	*/
bd681513	175	int btrfs_try_tree_write_lock(struct extent_buffer *eb)
b4ce94de	176	{
bd681513 CM	177	if (atomic_read(&eb->blocking_writers) \|\|
	178	atomic_read(&eb->blocking_readers))
	179	return 0;
ea4ebde0	180
f82c458a	181	write_lock(&eb->lock);
bd681513 CM	182	if (atomic_read(&eb->blocking_writers) \|\|
	183	atomic_read(&eb->blocking_readers)) {
	184	write_unlock(&eb->lock);
	185	return 0;
	186	}
	187	atomic_inc(&eb->write_locks);
	188	atomic_inc(&eb->spinning_writers);
5b25f70f	189	eb->lock_owner = current->pid;
b4ce94de CM	190	return 1;
	191	}
	192
	193	/*
bd681513 CM	194	* drop a spinning read lock
	195	*/
	196	void btrfs_tree_read_unlock(struct extent_buffer *eb)
	197	{
ea4ebde0 CM	198	/*
	199	* if we're nested, we have the write lock. No new locking
	200	* is needed as long as we are the lock owner.
	201	* The write unlock will do a barrier for us, and the lock_nested
	202	* field only matters to the lock owner.
	203	*/
	204	if (eb->lock_nested && current->pid == eb->lock_owner) {
	205	eb->lock_nested = 0;
	206	return;
5b25f70f	207	}
bd681513 CM	208	btrfs_assert_tree_read_locked(eb);
	209	WARN_ON(atomic_read(&eb->spinning_readers) == 0);
	210	atomic_dec(&eb->spinning_readers);
	211	atomic_dec(&eb->read_locks);
	212	read_unlock(&eb->lock);
	213	}
	214
	215	/*
	216	* drop a blocking read lock
	217	*/
	218	void btrfs_tree_read_unlock_blocking(struct extent_buffer *eb)
	219	{
ea4ebde0 CM	220	/*
	221	* if we're nested, we have the write lock. No new locking
	222	* is needed as long as we are the lock owner.
	223	* The write unlock will do a barrier for us, and the lock_nested
	224	* field only matters to the lock owner.
	225	*/
	226	if (eb->lock_nested && current->pid == eb->lock_owner) {
	227	eb->lock_nested = 0;
	228	return;
5b25f70f	229	}
bd681513 CM	230	btrfs_assert_tree_read_locked(eb);
bd681513 CM	231	WARN_ON(atomic_read(&eb->blocking_readers) == 0);
cbea5ac1 CM	232	if (atomic_dec_and_test(&eb->blocking_readers) &&
cbea5ac1 CM	233	waitqueue_active(&eb->read_lock_wq))
bd681513 CM	234	wake_up(&eb->read_lock_wq);
	235	atomic_dec(&eb->read_locks);
	236	}
	237
	238	/*
	239	* take a spinning write lock. This will wait for both
	240	* blocking readers or writers
b4ce94de	241	*/
143bede5	242	void btrfs_tree_lock(struct extent_buffer *eb)
b4ce94de	243	{
bd681513 CM	244	again:
	245	wait_event(eb->read_lock_wq, atomic_read(&eb->blocking_readers) == 0);
	246	wait_event(eb->write_lock_wq, atomic_read(&eb->blocking_writers) == 0);
	247	write_lock(&eb->lock);
	248	if (atomic_read(&eb->blocking_readers)) {
	249	write_unlock(&eb->lock);
	250	wait_event(eb->read_lock_wq,
	251	atomic_read(&eb->blocking_readers) == 0);
	252	goto again;
f9efa9c7	253	}
bd681513 CM	254	if (atomic_read(&eb->blocking_writers)) {
	255	write_unlock(&eb->lock);
	256	wait_event(eb->write_lock_wq,
	257	atomic_read(&eb->blocking_writers) == 0);
	258	goto again;
	259	}
	260	WARN_ON(atomic_read(&eb->spinning_writers));
	261	atomic_inc(&eb->spinning_writers);
	262	atomic_inc(&eb->write_locks);
5b25f70f	263	eb->lock_owner = current->pid;
925baedd CM	264	}
925baedd CM	265
bd681513 CM	266	/*
	267	* drop a spinning or a blocking write lock.
	268	*/
143bede5	269	void btrfs_tree_unlock(struct extent_buffer *eb)
925baedd	270	{
bd681513 CM	271	int blockers = atomic_read(&eb->blocking_writers);
	272
	273	BUG_ON(blockers > 1);
	274
	275	btrfs_assert_tree_locked(eb);
ea4ebde0	276	eb->lock_owner = 0;
bd681513 CM	277	atomic_dec(&eb->write_locks);
	278
	279	if (blockers) {
	280	WARN_ON(atomic_read(&eb->spinning_writers));
	281	atomic_dec(&eb->blocking_writers);
cbea5ac1 CM	282	smp_mb();
	283	if (waitqueue_active(&eb->write_lock_wq))
	284	wake_up(&eb->write_lock_wq);
bd681513 CM	285	} else {
	286	WARN_ON(atomic_read(&eb->spinning_writers) != 1);
	287	atomic_dec(&eb->spinning_writers);
	288	write_unlock(&eb->lock);
	289	}
925baedd CM	290	}
925baedd CM	291
b9447ef8	292	void btrfs_assert_tree_locked(struct extent_buffer *eb)
925baedd	293	{
bd681513 CM	294	BUG_ON(!atomic_read(&eb->write_locks));
	295	}
	296
48a3b636	297	static void btrfs_assert_tree_read_locked(struct extent_buffer *eb)
bd681513 CM	298	{
bd681513 CM	299	BUG_ON(!atomic_read(&eb->read_locks));
925baedd	300	}