~hc/RK356X_SDK_RELEASE.git

..	..	@@ -10,6 +10,7 @@
10	10	#include <linux/pagemap.h>
11	11	#include <linux/blkdev.h>
12	12	#include <linux/uuid.h>
	13	+#include "misc.h"
13	14	#include "ctree.h"
14	15	#include "disk-io.h"
15	16	#include "transaction.h"
..	..	@@ -19,12 +20,84 @@
19	20	#include "volumes.h"
20	21	#include "dev-replace.h"
21	22	#include "qgroup.h"
	23	+#include "block-group.h"
	24	+#include "space-info.h"
22	25
23	26	#define BTRFS_ROOT_TRANS_TAG 0
24	27
	28	+/*
	29	+ * Transaction states and transitions
	30	+ *
	31	+ * No running transaction (fs tree blocks are not modified)
	32	+ * \|
	33	+ * \| To next stage:
	34	+ * \| Call start_transaction() variants. Except btrfs_join_transaction_nostart().
	35	+ * V
	36	+ * Transaction N [[TRANS_STATE_RUNNING]]
	37	+ * \|
	38	+ * \| New trans handles can be attached to transaction N by calling all
	39	+ * \| start_transaction() variants.
	40	+ * \|
	41	+ * \| To next stage:
	42	+ * \| Call btrfs_commit_transaction() on any trans handle attached to
	43	+ * \| transaction N
	44	+ * V
	45	+ * Transaction N [[TRANS_STATE_COMMIT_START]]
	46	+ * \|
	47	+ * \| Will wait for previous running transaction to completely finish if there
	48	+ * \| is one
	49	+ * \|
	50	+ * \| Then one of the following happes:
	51	+ * \| - Wait for all other trans handle holders to release.
	52	+ * \| The btrfs_commit_transaction() caller will do the commit work.
	53	+ * \| - Wait for current transaction to be committed by others.
	54	+ * \| Other btrfs_commit_transaction() caller will do the commit work.
	55	+ * \|
	56	+ * \| At this stage, only btrfs_join_transaction*() variants can attach
	57	+ * \| to this running transaction.
	58	+ * \| All other variants will wait for current one to finish and attach to
	59	+ * \| transaction N+1.
	60	+ * \|
	61	+ * \| To next stage:
	62	+ * \| Caller is chosen to commit transaction N, and all other trans handle
	63	+ * \| haven been released.
	64	+ * V
	65	+ * Transaction N [[TRANS_STATE_COMMIT_DOING]]
	66	+ * \|
	67	+ * \| The heavy lifting transaction work is started.
	68	+ * \| From running delayed refs (modifying extent tree) to creating pending
	69	+ * \| snapshots, running qgroups.
	70	+ * \| In short, modify supporting trees to reflect modifications of subvolume
	71	+ * \| trees.
	72	+ * \|
	73	+ * \| At this stage, all start_transaction() calls will wait for this
	74	+ * \| transaction to finish and attach to transaction N+1.
	75	+ * \|
	76	+ * \| To next stage:
	77	+ * \| Until all supporting trees are updated.
	78	+ * V
	79	+ * Transaction N [[TRANS_STATE_UNBLOCKED]]
	80	+ * \| Transaction N+1
	81	+ * \| All needed trees are modified, thus we only [[TRANS_STATE_RUNNING]]
	82	+ * \| need to write them back to disk and update \|
	83	+ * \| super blocks. \|
	84	+ * \| \|
	85	+ * \| At this stage, new transaction is allowed to \|
	86	+ * \| start. \|
	87	+ * \| All new start_transaction() calls will be \|
	88	+ * \| attached to transid N+1. \|
	89	+ * \| \|
	90	+ * \| To next stage: \|
	91	+ * \| Until all tree blocks are super blocks are \|
	92	+ * \| written to block devices \|
	93	+ * V \|
	94	+ * Transaction N [[TRANS_STATE_COMPLETED]] V
	95	+ * All tree blocks and super blocks are written. Transaction N+1
	96	+ * This transaction is finished and all its [[TRANS_STATE_COMMIT_START]]
	97	+ * data structures will be cleaned up. \| Life goes on
	98	+ */
25	99	static const unsigned int btrfs_blocked_trans_types[TRANS_STATE_MAX] = {
26	100	[TRANS_STATE_RUNNING] = 0U,
27		- [TRANS_STATE_BLOCKED] = __TRANS_START,
28	101	[TRANS_STATE_COMMIT_START] = (__TRANS_START \| __TRANS_ATTACH),
29	102	[TRANS_STATE_COMMIT_DOING] = (__TRANS_START \|
30	103	__TRANS_ATTACH \|
..	..	@@ -47,19 +120,14 @@
47	120	WARN_ON(refcount_read(&transaction->use_count) == 0);
48	121	if (refcount_dec_and_test(&transaction->use_count)) {
49	122	BUG_ON(!list_empty(&transaction->list));
50		- WARN_ON(!RB_EMPTY_ROOT(&transaction->delayed_refs.href_root));
	123	+ WARN_ON(!RB_EMPTY_ROOT(
	124	+ &transaction->delayed_refs.href_root.rb_root));
	125	+ WARN_ON(!RB_EMPTY_ROOT(
	126	+ &transaction->delayed_refs.dirty_extent_root));
51	127	if (transaction->delayed_refs.pending_csums)
52	128	btrfs_err(transaction->fs_info,
53	129	"pending csums is %llu",
54	130	transaction->delayed_refs.pending_csums);
55		- while (!list_empty(&transaction->pending_chunks)) {
56		- struct extent_map *em;
57		-
58		- em = list_first_entry(&transaction->pending_chunks,
59		- struct extent_map, list);
60		- list_del_init(&em->list);
61		- free_extent_map(em);
62		- }
63	131	/*
64	132	* If any block groups are found in ->deleted_bgs then it's
65	133	* because the transaction was aborted and a commit did not
..	..	@@ -68,75 +136,90 @@
68	136	* discard the physical locations of the block groups.
69	137	*/
70	138	while (!list_empty(&transaction->deleted_bgs)) {
71		- struct btrfs_block_group_cache *cache;
	139	+ struct btrfs_block_group *cache;
72	140
73	141	cache = list_first_entry(&transaction->deleted_bgs,
74		- struct btrfs_block_group_cache,
	142	+ struct btrfs_block_group,
75	143	bg_list);
76	144	list_del_init(&cache->bg_list);
77		- btrfs_put_block_group_trimming(cache);
	145	+ btrfs_unfreeze_block_group(cache);
78	146	btrfs_put_block_group(cache);
79	147	}
	148	+ WARN_ON(!list_empty(&transaction->dev_update_list));
80	149	kfree(transaction);
81	150	}
82	151	}
83	152
84		-static void clear_btree_io_tree(struct extent_io_tree *tree)
	153	+static noinline void switch_commit_roots(struct btrfs_trans_handle *trans)
85	154	{
86		- spin_lock(&tree->lock);
87		- /*
88		- * Do a single barrier for the waitqueue_active check here, the state
89		- * of the waitqueue should not change once clear_btree_io_tree is
90		- * called.
91		- */
92		- smp_mb();
93		- while (!RB_EMPTY_ROOT(&tree->state)) {
94		- struct rb_node *node;
95		- struct extent_state *state;
96		-
97		- node = rb_first(&tree->state);
98		- state = rb_entry(node, struct extent_state, rb_node);
99		- rb_erase(&state->rb_node, &tree->state);
100		- RB_CLEAR_NODE(&state->rb_node);
101		- /*
102		- * btree io trees aren't supposed to have tasks waiting for
103		- * changes in the flags of extent states ever.
104		- */
105		- ASSERT(!waitqueue_active(&state->wq));
106		- free_extent_state(state);
107		-
108		- cond_resched_lock(&tree->lock);
109		- }
110		- spin_unlock(&tree->lock);
111		-}
112		-
113		-static noinline void switch_commit_roots(struct btrfs_transaction *trans)
114		-{
	155	+ struct btrfs_transaction *cur_trans = trans->transaction;
115	156	struct btrfs_fs_info *fs_info = trans->fs_info;
116	157	struct btrfs_root root, tmp;
	158	+ struct btrfs_caching_control caching_ctl, next;
117	159
118	160	down_write(&fs_info->commit_root_sem);
119		- list_for_each_entry_safe(root, tmp, &trans->switch_commits,
	161	+ list_for_each_entry_safe(root, tmp, &cur_trans->switch_commits,
120	162	dirty_list) {
121	163	list_del_init(&root->dirty_list);
122	164	free_extent_buffer(root->commit_root);
123	165	root->commit_root = btrfs_root_node(root);
124		- if (is_fstree(root->objectid))
	166	+ if (is_fstree(root->root_key.objectid))
125	167	btrfs_unpin_free_ino(root);
126		- clear_btree_io_tree(&root->dirty_log_pages);
	168	+ extent_io_tree_release(&root->dirty_log_pages);
	169	+ btrfs_qgroup_clean_swapped_blocks(root);
127	170	}
128	171
129	172	/* We can free old roots now. */
130		- spin_lock(&trans->dropped_roots_lock);
131		- while (!list_empty(&trans->dropped_roots)) {
132		- root = list_first_entry(&trans->dropped_roots,
	173	+ spin_lock(&cur_trans->dropped_roots_lock);
	174	+ while (!list_empty(&cur_trans->dropped_roots)) {
	175	+ root = list_first_entry(&cur_trans->dropped_roots,
133	176	struct btrfs_root, root_list);
134	177	list_del_init(&root->root_list);
135		- spin_unlock(&trans->dropped_roots_lock);
	178	+ spin_unlock(&cur_trans->dropped_roots_lock);
	179	+ btrfs_free_log(trans, root);
136	180	btrfs_drop_and_free_fs_root(fs_info, root);
137		- spin_lock(&trans->dropped_roots_lock);
	181	+ spin_lock(&cur_trans->dropped_roots_lock);
138	182	}
139		- spin_unlock(&trans->dropped_roots_lock);
	183	+ spin_unlock(&cur_trans->dropped_roots_lock);
	184	+
	185	+ /*
	186	+ * We have to update the last_byte_to_unpin under the commit_root_sem,
	187	+ * at the same time we swap out the commit roots.
	188	+ *
	189	+ * This is because we must have a real view of the last spot the caching
	190	+ * kthreads were while caching. Consider the following views of the
	191	+ * extent tree for a block group
	192	+ *
	193	+ * commit root
	194	+ * +----+----+----+----+----+----+----+
	195	+ * \|\\\\\| \|\\\\\|\\\\\| \|\\\\\|\\\\\|
	196	+ * +----+----+----+----+----+----+----+
	197	+ * 0 1 2 3 4 5 6 7
	198	+ *
	199	+ * new commit root
	200	+ * +----+----+----+----+----+----+----+
	201	+ * \| \| \| \|\\\\\| \| \|\\\\\|
	202	+ * +----+----+----+----+----+----+----+
	203	+ * 0 1 2 3 4 5 6 7
	204	+ *
	205	+ * If the cache_ctl->progress was at 3, then we are only allowed to
	206	+ * unpin [0,1) and [2,3], because the caching thread has already
	207	+ * processed those extents. We are not allowed to unpin [5,6), because
	208	+ * the caching thread will re-start it's search from 3, and thus find
	209	+ * the hole from [4,6) to add to the free space cache.
	210	+ */
	211	+ list_for_each_entry_safe(caching_ctl, next,
	212	+ &fs_info->caching_block_groups, list) {
	213	+ struct btrfs_block_group *cache = caching_ctl->block_group;
	214	+
	215	+ if (btrfs_block_group_done(cache)) {
	216	+ cache->last_byte_to_unpin = (u64)-1;
	217	+ list_del_init(&caching_ctl->list);
	218	+ btrfs_put_caching_control(caching_ctl);
	219	+ } else {
	220	+ cache->last_byte_to_unpin = caching_ctl->progress;
	221	+ }
	222	+ }
140	223	up_write(&fs_info->commit_root_sem);
141	224	}
142	225
..	..	@@ -166,6 +249,24 @@
166	249	}
167	250
168	251	/*
	252	+ * To be called after all the new block groups attached to the transaction
	253	+ * handle have been created (btrfs_create_pending_block_groups()).
	254	+ */
	255	+void btrfs_trans_release_chunk_metadata(struct btrfs_trans_handle *trans)
	256	+{
	257	+ struct btrfs_fs_info *fs_info = trans->fs_info;
	258	+
	259	+ if (!trans->chunk_bytes_reserved)
	260	+ return;
	261	+
	262	+ WARN_ON_ONCE(!list_empty(&trans->new_bgs));
	263	+
	264	+ btrfs_block_rsv_release(fs_info, &fs_info->chunk_block_rsv,
	265	+ trans->chunk_bytes_reserved, NULL);
	266	+ trans->chunk_bytes_reserved = 0;
	267	+}
	268	+
	269	+/*
169	270	* either allocate a new transaction or hop into the existing one
170	271	*/
171	272	static noinline int join_transaction(struct btrfs_fs_info *fs_info,
..	..	@@ -183,7 +284,7 @@
183	284
184	285	cur_trans = fs_info->running_transaction;
185	286	if (cur_trans) {
186		- if (cur_trans->aborted) {
	287	+ if (TRANS_ABORTED(cur_trans)) {
187	288	spin_unlock(&fs_info->trans_lock);
188	289	return cur_trans->aborted;
189	290	}
..	..	@@ -200,10 +301,11 @@
200	301	spin_unlock(&fs_info->trans_lock);
201	302
202	303	/*
203		- * If we are ATTACH, we just want to catch the current transaction,
204		- * and commit it. If there is no transaction, just return ENOENT.
	304	+ * If we are ATTACH or TRANS_JOIN_NOSTART, we just want to catch the
	305	+ * current transaction, and commit it. If there is no transaction, just
	306	+ * return ENOENT.
205	307	*/
206		- if (type == TRANS_ATTACH)
	308	+ if (type == TRANS_ATTACH \|\| type == TRANS_JOIN_NOSTART)
207	309	return -ENOENT;
208	310
209	311	/*
..	..	@@ -231,24 +333,24 @@
231	333	}
232	334
233	335	cur_trans->fs_info = fs_info;
	336	+ atomic_set(&cur_trans->pending_ordered, 0);
	337	+ init_waitqueue_head(&cur_trans->pending_wait);
234	338	atomic_set(&cur_trans->num_writers, 1);
235	339	extwriter_counter_init(cur_trans, type);
236	340	init_waitqueue_head(&cur_trans->writer_wait);
237	341	init_waitqueue_head(&cur_trans->commit_wait);
238		- init_waitqueue_head(&cur_trans->pending_wait);
239	342	cur_trans->state = TRANS_STATE_RUNNING;
240	343	/*
241	344	* One for this trans handle, one so it will live on until we
242	345	* commit the transaction.
243	346	*/
244	347	refcount_set(&cur_trans->use_count, 2);
245		- atomic_set(&cur_trans->pending_ordered, 0);
246	348	cur_trans->flags = 0;
247	349	cur_trans->start_time = ktime_get_seconds();
248	350
249	351	memset(&cur_trans->delayed_refs, 0, sizeof(cur_trans->delayed_refs));
250	352
251		- cur_trans->delayed_refs.href_root = RB_ROOT;
	353	+ cur_trans->delayed_refs.href_root = RB_ROOT_CACHED;
252	354	cur_trans->delayed_refs.dirty_extent_root = RB_ROOT;
253	355	atomic_set(&cur_trans->delayed_refs.num_entries, 0);
254	356
..	..	@@ -266,19 +368,20 @@
266	368	spin_lock_init(&cur_trans->delayed_refs.lock);
267	369
268	370	INIT_LIST_HEAD(&cur_trans->pending_snapshots);
269		- INIT_LIST_HEAD(&cur_trans->pending_chunks);
	371	+ INIT_LIST_HEAD(&cur_trans->dev_update_list);
270	372	INIT_LIST_HEAD(&cur_trans->switch_commits);
271	373	INIT_LIST_HEAD(&cur_trans->dirty_bgs);
272	374	INIT_LIST_HEAD(&cur_trans->io_bgs);
273	375	INIT_LIST_HEAD(&cur_trans->dropped_roots);
274	376	mutex_init(&cur_trans->cache_write_mutex);
275		- cur_trans->num_dirty_bgs = 0;
276	377	spin_lock_init(&cur_trans->dirty_bgs_lock);
277	378	INIT_LIST_HEAD(&cur_trans->deleted_bgs);
278	379	spin_lock_init(&cur_trans->dropped_roots_lock);
279	380	list_add_tail(&cur_trans->list, &fs_info->trans_list);
280		- extent_io_tree_init(&cur_trans->dirty_pages,
281		- fs_info->btree_inode);
	381	+ extent_io_tree_init(fs_info, &cur_trans->dirty_pages,
	382	+ IO_TREE_TRANS_DIRTY_PAGES, fs_info->btree_inode);
	383	+ extent_io_tree_init(fs_info, &cur_trans->pinned_extents,
	384	+ IO_TREE_FS_PINNED_EXTENTS, NULL);
282	385	fs_info->generation++;
283	386	cur_trans->transid = fs_info->generation;
284	387	fs_info->running_transaction = cur_trans;
..	..	@@ -289,10 +392,10 @@
289	392	}
290	393
291	394	/*
292		- * this does all the record keeping required to make sure that a reference
293		- * counted root is properly recorded in a given transaction. This is required
294		- * to make sure the old root from before we joined the transaction is deleted
295		- * when the transaction commits
	395	+ * This does all the record keeping required to make sure that a shareable root
	396	+ * is properly recorded in a given transaction. This is required to make sure
	397	+ * the old root from before we joined the transaction is deleted when the
	398	+ * transaction commits.
296	399	*/
297	400	static int record_root_in_trans(struct btrfs_trans_handle *trans,
298	401	struct btrfs_root *root,
..	..	@@ -300,7 +403,7 @@
300	403	{
301	404	struct btrfs_fs_info *fs_info = root->fs_info;
302	405
303		- if ((test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
	406	+ if ((test_bit(BTRFS_ROOT_SHAREABLE, &root->state) &&
304	407	root->last_trans < trans->transid) \|\| force) {
305	408	WARN_ON(root == fs_info->extent_root);
306	409	WARN_ON(!force && root->commit_root != root->node);
..	..	@@ -379,7 +482,7 @@
379	482	{
380	483	struct btrfs_fs_info *fs_info = root->fs_info;
381	484
382		- if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
	485	+ if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
383	486	return 0;
384	487
385	488	/*
..	..	@@ -400,9 +503,9 @@
400	503
401	504	static inline int is_transaction_blocked(struct btrfs_transaction *trans)
402	505	{
403		- return (trans->state >= TRANS_STATE_BLOCKED &&
	506	+ return (trans->state >= TRANS_STATE_COMMIT_START &&
404	507	trans->state < TRANS_STATE_UNBLOCKED &&
405		- !trans->aborted);
	508	+ !TRANS_ABORTED(trans));
406	509	}
407	510
408	511	/* wait for commit against the current transaction to become unblocked
..	..	@@ -421,7 +524,7 @@
421	524
422	525	wait_event(fs_info->transaction_wait,
423	526	cur_trans->state >= TRANS_STATE_UNBLOCKED \|\|
424		- cur_trans->aborted);
	527	+ TRANS_ABORTED(cur_trans));
425	528	btrfs_put_transaction(cur_trans);
426	529	} else {
427	530	spin_unlock(&fs_info->trans_lock);
..	..	@@ -444,7 +547,7 @@
444	547	struct btrfs_fs_info *fs_info = root->fs_info;
445	548
446	549	if (!fs_info->reloc_ctl \|\|
447		- !test_bit(BTRFS_ROOT_REF_COWS, &root->state) \|\|
	550	+ !test_bit(BTRFS_ROOT_SHAREABLE, &root->state) \|\|
448	551	root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID \|\|
449	552	root->reloc_root)
450	553	return false;
..	..	@@ -458,12 +561,13 @@
458	561	bool enforce_qgroups)
459	562	{
460	563	struct btrfs_fs_info *fs_info = root->fs_info;
461		-
	564	+ struct btrfs_block_rsv *delayed_refs_rsv = &fs_info->delayed_refs_rsv;
462	565	struct btrfs_trans_handle *h;
463	566	struct btrfs_transaction *cur_trans;
464	567	u64 num_bytes = 0;
465	568	u64 qgroup_reserved = 0;
466	569	bool reloc_reserved = false;
	570	+ bool do_chunk_alloc = false;
467	571	int ret;
468	572
469	573	/* Send isn't supposed to start transactions. */
..	..	@@ -487,13 +591,29 @@
487	591	* the appropriate flushing if need be.
488	592	*/
489	593	if (num_items && root != fs_info->chunk_root) {
	594	+ struct btrfs_block_rsv *rsv = &fs_info->trans_block_rsv;
	595	+ u64 delayed_refs_bytes = 0;
	596	+
490	597	qgroup_reserved = num_items * fs_info->nodesize;
491	598	ret = btrfs_qgroup_reserve_meta_pertrans(root, qgroup_reserved,
492	599	enforce_qgroups);
493	600	if (ret)
494	601	return ERR_PTR(ret);
495	602
496		- num_bytes = btrfs_calc_trans_metadata_size(fs_info, num_items);
	603	+ /*
	604	+ * We want to reserve all the bytes we may need all at once, so
	605	+ * we only do 1 enospc flushing cycle per transaction start. We
	606	+ * accomplish this by simply assuming we'll do 2 x num_items
	607	+ * worth of delayed refs updates in this trans handle, and
	608	+ * refill that amount for whatever is missing in the reserve.
	609	+ */
	610	+ num_bytes = btrfs_calc_insert_metadata_size(fs_info, num_items);
	611	+ if (flush == BTRFS_RESERVE_FLUSH_ALL &&
	612	+ delayed_refs_rsv->full == 0) {
	613	+ delayed_refs_bytes = num_bytes;
	614	+ num_bytes <<= 1;
	615	+ }
	616	+
497	617	/*
498	618	* Do the reservation for the relocation root creation
499	619	*/
..	..	@@ -502,8 +622,27 @@
502	622	reloc_reserved = true;
503	623	}
504	624
505		- ret = btrfs_block_rsv_add(root, &fs_info->trans_block_rsv,
506		- num_bytes, flush);
	625	+ ret = btrfs_block_rsv_add(root, rsv, num_bytes, flush);
	626	+ if (ret)
	627	+ goto reserve_fail;
	628	+ if (delayed_refs_bytes) {
	629	+ btrfs_migrate_to_delayed_refs_rsv(fs_info, rsv,
	630	+ delayed_refs_bytes);
	631	+ num_bytes -= delayed_refs_bytes;
	632	+ }
	633	+
	634	+ if (rsv->space_info->force_alloc)
	635	+ do_chunk_alloc = true;
	636	+ } else if (num_items == 0 && flush == BTRFS_RESERVE_FLUSH_ALL &&
	637	+ !delayed_refs_rsv->full) {
	638	+ /*
	639	+ * Some people call with btrfs_start_transaction(root, 0)
	640	+ * because they can be throttled, but have some other mechanism
	641	+ * for reserving space. We still want these guys to refill the
	642	+ * delayed block_rsv so just add 1 items worth of reservation
	643	+ * here.
	644	+ */
	645	+ ret = btrfs_delayed_refs_rsv_refill(fs_info, flush);
507	646	if (ret)
508	647	goto reserve_fail;
509	648	}
..	..	@@ -556,7 +695,7 @@
556	695	INIT_LIST_HEAD(&h->new_bgs);
557	696
558	697	smp_mb();
559		- if (cur_trans->state >= TRANS_STATE_BLOCKED &&
	698	+ if (cur_trans->state >= TRANS_STATE_COMMIT_START &&
560	699	may_wait_transaction(fs_info, type)) {
561	700	current->journal_info = h;
562	701	btrfs_commit_transaction(h);
..	..	@@ -574,6 +713,19 @@
574	713	got_it:
575	714	if (!current->journal_info)
576	715	current->journal_info = h;
	716	+
	717	+ /*
	718	+ * If the space_info is marked ALLOC_FORCE then we'll get upgraded to
	719	+ * ALLOC_FORCE the first run through, and then we won't allocate for
	720	+ * anybody else who races in later. We don't care about the return
	721	+ * value here.
	722	+ */
	723	+ if (do_chunk_alloc && num_bytes) {
	724	+ u64 flags = h->block_rsv->space_info->flags;
	725	+
	726	+ btrfs_chunk_alloc(h, btrfs_get_alloc_profile(fs_info, flags),
	727	+ CHUNK_ALLOC_NO_FORCE);
	728	+ }
577	729
578	730	/*
579	731	* btrfs_record_root_in_trans() needs to alloc new extents, and may
..	..	@@ -594,7 +746,7 @@
594	746	alloc_fail:
595	747	if (num_bytes)
596	748	btrfs_block_rsv_release(fs_info, &fs_info->trans_block_rsv,
597		- num_bytes);
	749	+ num_bytes, NULL);
598	750	reserve_fail:
599	751	btrfs_qgroup_free_meta_pertrans(root, qgroup_reserved);
600	752	return ERR_PTR(ret);
..	..	@@ -609,43 +761,10 @@
609	761
610	762	struct btrfs_trans_handle *btrfs_start_transaction_fallback_global_rsv(
611	763	struct btrfs_root *root,
612		- unsigned int num_items,
613		- int min_factor)
	764	+ unsigned int num_items)
614	765	{
615		- struct btrfs_fs_info *fs_info = root->fs_info;
616		- struct btrfs_trans_handle *trans;
617		- u64 num_bytes;
618		- int ret;
619		-
620		- /*
621		- * We have two callers: unlink and block group removal. The
622		- * former should succeed even if we will temporarily exceed
623		- * quota and the latter operates on the extent root so
624		- * qgroup enforcement is ignored anyway.
625		- */
626		- trans = start_transaction(root, num_items, TRANS_START,
627		- BTRFS_RESERVE_FLUSH_ALL, false);
628		- if (!IS_ERR(trans) \|\| PTR_ERR(trans) != -ENOSPC)
629		- return trans;
630		-
631		- trans = btrfs_start_transaction(root, 0);
632		- if (IS_ERR(trans))
633		- return trans;
634		-
635		- num_bytes = btrfs_calc_trans_metadata_size(fs_info, num_items);
636		- ret = btrfs_cond_migrate_bytes(fs_info, &fs_info->trans_block_rsv,
637		- num_bytes, min_factor);
638		- if (ret) {
639		- btrfs_end_transaction(trans);
640		- return ERR_PTR(ret);
641		- }
642		-
643		- trans->block_rsv = &fs_info->trans_block_rsv;
644		- trans->bytes_reserved = num_bytes;
645		- trace_btrfs_space_reservation(fs_info, "transaction",
646		- trans->transid, num_bytes, 1);
647		-
648		- return trans;
	766	+ return start_transaction(root, num_items, TRANS_START,
	767	+ BTRFS_RESERVE_FLUSH_ALL_STEAL, false);
649	768	}
650	769
651	770	struct btrfs_trans_handle btrfs_join_transaction(struct btrfs_root root)
..	..	@@ -654,7 +773,7 @@
654	773	true);
655	774	}
656	775
657		-struct btrfs_trans_handle btrfs_join_transaction_nolock(struct btrfs_root root)
	776	+struct btrfs_trans_handle btrfs_join_transaction_spacecache(struct btrfs_root root)
658	777	{
659	778	return start_transaction(root, 0, TRANS_JOIN_NOLOCK,
660	779	BTRFS_RESERVE_NO_FLUSH, true);
..	..	@@ -692,7 +811,7 @@
692	811	/*
693	812	* btrfs_attach_transaction_barrier() - catch the running transaction
694	813	*
695		- * It is similar to the above function, the differentia is this one
	814	+ * It is similar to the above function, the difference is this one
696	815	* will wait for all the inactive transactions until they fully
697	816	* complete.
698	817	*/
..	..	@@ -703,8 +822,13 @@
703	822
704	823	trans = start_transaction(root, 0, TRANS_ATTACH,
705	824	BTRFS_RESERVE_NO_FLUSH, true);
706		- if (trans == ERR_PTR(-ENOENT))
707		- btrfs_wait_for_commit(root->fs_info, 0);
	825	+ if (trans == ERR_PTR(-ENOENT)) {
	826	+ int ret;
	827	+
	828	+ ret = btrfs_wait_for_commit(root->fs_info, 0);
	829	+ if (ret)
	830	+ return ERR_PTR(ret);
	831	+ }
708	832
709	833	return trans;
710	834	}
..	..	@@ -768,6 +892,7 @@
768	892	}
769	893
770	894	wait_for_commit(cur_trans);
	895	+ ret = cur_trans->aborted;
771	896	btrfs_put_transaction(cur_trans);
772	897	out:
773	898	return ret;
..	..	@@ -782,7 +907,7 @@
782	907	{
783	908	struct btrfs_fs_info *fs_info = trans->fs_info;
784	909
785		- if (btrfs_check_space_for_delayed_refs(trans, fs_info))
	910	+ if (btrfs_check_space_for_delayed_refs(fs_info))
786	911	return 1;
787	912
788	913	return !!btrfs_block_rsv_check(&fs_info->global_block_rsv, 5);
..	..	@@ -791,21 +916,11 @@
791	916	int btrfs_should_end_transaction(struct btrfs_trans_handle *trans)
792	917	{
793	918	struct btrfs_transaction *cur_trans = trans->transaction;
794		- int updates;
795		- int err;
796	919
797	920	smp_mb();
798		- if (cur_trans->state >= TRANS_STATE_BLOCKED \|\|
	921	+ if (cur_trans->state >= TRANS_STATE_COMMIT_START \|\|
799	922	cur_trans->delayed_refs.flushing)
800	923	return 1;
801		-
802		- updates = trans->delayed_ref_updates;
803		- trans->delayed_ref_updates = 0;
804		- if (updates) {
805		- err = btrfs_run_delayed_refs(trans, updates * 2);
806		- if (err) /* Error code will also eval true */
807		- return err;
808		- }
809	924
810	925	return should_end_transaction(trans);
811	926	}
..	..	@@ -827,7 +942,7 @@
827	942	trace_btrfs_space_reservation(fs_info, "transaction",
828	943	trans->transid, trans->bytes_reserved, 0);
829	944	btrfs_block_rsv_release(fs_info, trans->block_rsv,
830		- trans->bytes_reserved);
	945	+ trans->bytes_reserved, NULL);
831	946	trans->bytes_reserved = 0;
832	947	}
833	948
..	..	@@ -836,11 +951,7 @@
836	951	{
837	952	struct btrfs_fs_info *info = trans->fs_info;
838	953	struct btrfs_transaction *cur_trans = trans->transaction;
839		- u64 transid = trans->transid;
840		- unsigned long cur = trans->delayed_ref_updates;
841		- int lock = (trans->type != TRANS_JOIN_NOLOCK);
842	954	int err = 0;
843		- int must_run_delayed_refs = 0;
844	955
845	956	if (refcount_read(&trans->use_count) > 1) {
846	957	refcount_dec(&trans->use_count);
..	..	@@ -851,46 +962,9 @@
851	962	btrfs_trans_release_metadata(trans);
852	963	trans->block_rsv = NULL;
853	964
854		- if (!list_empty(&trans->new_bgs))
855		- btrfs_create_pending_block_groups(trans);
856		-
857		- trans->delayed_ref_updates = 0;
858		- if (!trans->sync) {
859		- must_run_delayed_refs =
860		- btrfs_should_throttle_delayed_refs(trans, info);
861		- cur = max_t(unsigned long, cur, 32);
862		-
863		- /*
864		- * don't make the caller wait if they are from a NOLOCK
865		- * or ATTACH transaction, it will deadlock with commit
866		- */
867		- if (must_run_delayed_refs == 1 &&
868		- (trans->type & (__TRANS_JOIN_NOLOCK \| __TRANS_ATTACH)))
869		- must_run_delayed_refs = 2;
870		- }
871		-
872		- btrfs_trans_release_metadata(trans);
873		- trans->block_rsv = NULL;
874		-
875		- if (!list_empty(&trans->new_bgs))
876		- btrfs_create_pending_block_groups(trans);
	965	+ btrfs_create_pending_block_groups(trans);
877	966
878	967	btrfs_trans_release_chunk_metadata(trans);
879		-
880		- if (lock && should_end_transaction(trans) &&
881		- READ_ONCE(cur_trans->state) == TRANS_STATE_RUNNING) {
882		- spin_lock(&info->trans_lock);
883		- if (cur_trans->state == TRANS_STATE_RUNNING)
884		- cur_trans->state = TRANS_STATE_BLOCKED;
885		- spin_unlock(&info->trans_lock);
886		- }
887		-
888		- if (lock && READ_ONCE(cur_trans->state) == TRANS_STATE_BLOCKED) {
889		- if (throttle)
890		- return btrfs_commit_transaction(trans);
891		- else
892		- wake_up_process(info->transaction_kthread);
893		- }
894	968
895	969	if (trans->type & __TRANS_FREEZABLE)
896	970	sb_end_intwrite(info->sb);
..	..	@@ -909,17 +983,16 @@
909	983	if (throttle)
910	984	btrfs_run_delayed_iputs(info);
911	985
912		- if (trans->aborted \|\|
	986	+ if (TRANS_ABORTED(trans) \|\|
913	987	test_bit(BTRFS_FS_STATE_ERROR, &info->fs_state)) {
914	988	wake_up_process(info->transaction_kthread);
915		- err = -EIO;
	989	+ if (TRANS_ABORTED(trans))
	990	+ err = trans->aborted;
	991	+ else
	992	+ err = -EROFS;
916	993	}
917	994
918	995	kmem_cache_free(btrfs_trans_handle_cachep, trans);
919		- if (must_run_delayed_refs) {
920		- btrfs_async_run_delayed_refs(info, cur, transid,
921		- must_run_delayed_refs == 1);
922		- }
923	996	return err;
924	997	}
925	998
..	..	@@ -967,7 +1040,7 @@
967	1040	* superblock that points to btree nodes/leafs for which
968	1041	* writeback hasn't finished yet (and without errors).
969	1042	* We cleanup any entries left in the io tree when committing
970		- * the transaction (through clear_btree_io_tree()).
	1043	+ * the transaction (through extent_io_tree_release()).
971	1044	*/
972	1045	if (err == -ENOMEM) {
973	1046	err = 0;
..	..	@@ -1012,7 +1085,7 @@
1012	1085	* left in the io tree. For a log commit, we don't remove them
1013	1086	* after committing the log because the tree can be accessed
1014	1087	* concurrently - we do it only at transaction commit time when
1015		- * it's safe to do it (through clear_btree_io_tree()).
	1088	+ * it's safe to do it (through extent_io_tree_release()).
1016	1089	*/
1017	1090	err = clear_extent_bit(dirty_pages, start, end,
1018	1091	EXTENT_NEED_WAIT, 0, 0, &cached_state);
..	..	@@ -1032,7 +1105,7 @@
1032	1105	return werr;
1033	1106	}
1034	1107
1035		-int btrfs_wait_extents(struct btrfs_fs_info *fs_info,
	1108	+static int btrfs_wait_extents(struct btrfs_fs_info *fs_info,
1036	1109	struct extent_io_tree *dirty_pages)
1037	1110	{
1038	1111	bool errors = false;
..	..	@@ -1090,7 +1163,7 @@
1090	1163	blk_finish_plug(&plug);
1091	1164	ret2 = btrfs_wait_extents(fs_info, dirty_pages);
1092	1165
1093		- clear_btree_io_tree(&trans->transaction->dirty_pages);
	1166	+ extent_io_tree_release(&trans->transaction->dirty_pages);
1094	1167
1095	1168	if (ret)
1096	1169	return ret;
..	..	@@ -1158,7 +1231,7 @@
1158	1231
1159	1232	eb = btrfs_lock_root_node(fs_info->tree_root);
1160	1233	ret = btrfs_cow_block(trans, fs_info->tree_root, eb, NULL,
1161		- 0, &eb);
	1234	+ 0, &eb, BTRFS_NESTING_COW);
1162	1235	btrfs_tree_unlock(eb);
1163	1236	free_extent_buffer(eb);
1164	1237
..	..	@@ -1169,17 +1242,17 @@
1169	1242	if (ret)
1170	1243	return ret;
1171	1244
1172		- ret = btrfs_run_dev_stats(trans, fs_info);
	1245	+ ret = btrfs_run_dev_stats(trans);
1173	1246	if (ret)
1174	1247	return ret;
1175		- ret = btrfs_run_dev_replace(trans, fs_info);
	1248	+ ret = btrfs_run_dev_replace(trans);
1176	1249	if (ret)
1177	1250	return ret;
1178	1251	ret = btrfs_run_qgroups(trans);
1179	1252	if (ret)
1180	1253	return ret;
1181	1254
1182		- ret = btrfs_setup_space_cache(trans, fs_info);
	1255	+ ret = btrfs_setup_space_cache(trans);
1183	1256	if (ret)
1184	1257	return ret;
1185	1258
..	..	@@ -1207,7 +1280,7 @@
1207	1280	}
1208	1281
1209	1282	while (!list_empty(dirty_bgs) \|\| !list_empty(io_bgs)) {
1210		- ret = btrfs_write_dirty_block_groups(trans, fs_info);
	1283	+ ret = btrfs_write_dirty_block_groups(trans);
1211	1284	if (ret)
1212	1285	return ret;
1213	1286	ret = btrfs_run_delayed_refs(trans, (unsigned long)-1);
..	..	@@ -1220,7 +1293,10 @@
1220	1293
1221	1294	list_add_tail(&fs_info->extent_root->dirty_list,
1222	1295	&trans->transaction->switch_commits);
1223		- btrfs_after_dev_replace_commit(fs_info);
	1296	+
	1297	+ /* Update dev-replace pointer once everything is committed */
	1298	+ fs_info->dev_replace.committed_cursor_left =
	1299	+ fs_info->dev_replace.cursor_left_last_write_of_item;
1224	1300
1225	1301	return 0;
1226	1302	}
..	..	@@ -1235,8 +1311,10 @@
1235	1311	struct btrfs_fs_info *fs_info = root->fs_info;
1236	1312
1237	1313	spin_lock(&fs_info->trans_lock);
1238		- if (list_empty(&root->root_list))
	1314	+ if (list_empty(&root->root_list)) {
	1315	+ btrfs_grab_root(root);
1239	1316	list_add_tail(&root->root_list, &fs_info->dead_roots);
	1317	+ }
1240	1318	spin_unlock(&fs_info->trans_lock);
1241	1319	}
1242	1320
..	..	@@ -1360,7 +1438,7 @@
1360	1438	return 0;
1361	1439
1362	1440	/*
1363		- * Ensure dirty @src will be commited. Or, after comming
	1441	+ * Ensure dirty @src will be committed. Or, after coming
1364	1442	* commit_fs_roots() and switch_commit_roots(), any dirty but not
1365	1443	* recorded root will never be updated again, causing an outdated root
1366	1444	* item.
..	..	@@ -1401,7 +1479,7 @@
1401	1479	ret = commit_cowonly_roots(trans);
1402	1480	if (ret)
1403	1481	goto out;
1404		- switch_commit_roots(trans->transaction);
	1482	+ switch_commit_roots(trans);
1405	1483	ret = btrfs_write_and_wait_transaction(trans);
1406	1484	if (ret)
1407	1485	btrfs_handle_fs_error(fs_info, ret,
..	..	@@ -1453,7 +1531,6 @@
1453	1531	u64 index = 0;
1454	1532	u64 objectid;
1455	1533	u64 root_flags;
1456		- uuid_le new_uuid;
1457	1534
1458	1535	ASSERT(pending->path);
1459	1536	path = pending->path;
..	..	@@ -1546,8 +1623,7 @@
1546	1623
1547	1624	btrfs_set_root_generation_v2(new_root_item,
1548	1625	trans->transid);
1549		- uuid_le_gen(&new_uuid);
1550		- memcpy(new_root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
	1626	+ generate_random_guid(new_root_item->uuid);
1551	1627	memcpy(new_root_item->parent_uuid, root->root_item.uuid,
1552	1628	BTRFS_UUID_SIZE);
1553	1629	if (!(root_flags & BTRFS_ROOT_SUBVOL_RDONLY)) {
..	..	@@ -1563,7 +1639,8 @@
1563	1639	btrfs_set_root_otransid(new_root_item, trans->transid);
1564	1640
1565	1641	old = btrfs_lock_root_node(root);
1566		- ret = btrfs_cow_block(trans, root, old, NULL, 0, &old);
	1642	+ ret = btrfs_cow_block(trans, root, old, NULL, 0, &old,
	1643	+ BTRFS_NESTING_COW);
1567	1644	if (ret) {
1568	1645	btrfs_tree_unlock(old);
1569	1646	free_extent_buffer(old);
..	..	@@ -1571,7 +1648,7 @@
1571	1648	goto fail;
1572	1649	}
1573	1650
1574		- btrfs_set_lock_blocking(old);
	1651	+ btrfs_set_lock_blocking_write(old);
1575	1652
1576	1653	ret = btrfs_copy_root(trans, root, old, &tmp, objectid);
1577	1654	/* clean up in any case */
..	..	@@ -1609,9 +1686,10 @@
1609	1686	}
1610	1687
1611	1688	key.offset = (u64)-1;
1612		- pending->snap = btrfs_read_fs_root_no_name(fs_info, &key);
	1689	+ pending->snap = btrfs_get_new_fs_root(fs_info, objectid, pending->anon_dev);
1613	1690	if (IS_ERR(pending->snap)) {
1614	1691	ret = PTR_ERR(pending->snap);
	1692	+ pending->snap = NULL;
1615	1693	btrfs_abort_transaction(trans, ret);
1616	1694	goto fail;
1617	1695	}
..	..	@@ -1639,10 +1717,9 @@
1639	1717	if (ret < 0)
1640	1718	goto fail;
1641	1719
1642		- ret = btrfs_insert_dir_item(trans, parent_root,
1643		- dentry->d_name.name, dentry->d_name.len,
1644		- BTRFS_I(parent_inode), &key,
1645		- BTRFS_FT_DIR, index);
	1720	+ ret = btrfs_insert_dir_item(trans, dentry->d_name.name,
	1721	+ dentry->d_name.len, BTRFS_I(parent_inode),
	1722	+ &key, BTRFS_FT_DIR, index);
1646	1723	/* We have check then name at the beginning, so it is impossible. */
1647	1724	BUG_ON(ret == -EEXIST \|\| ret == -EOVERFLOW);
1648	1725	if (ret) {
..	..	@@ -1659,7 +1736,8 @@
1659	1736	btrfs_abort_transaction(trans, ret);
1660	1737	goto fail;
1661	1738	}
1662		- ret = btrfs_uuid_tree_add(trans, new_uuid.b, BTRFS_UUID_KEY_SUBVOL,
	1739	+ ret = btrfs_uuid_tree_add(trans, new_root_item->uuid,
	1740	+ BTRFS_UUID_KEY_SUBVOL,
1663	1741	objectid);
1664	1742	if (ret) {
1665	1743	btrfs_abort_transaction(trans, ret);
..	..	@@ -1771,7 +1849,8 @@
1771	1849	struct btrfs_transaction *trans)
1772	1850	{
1773	1851	wait_event(fs_info->transaction_blocked_wait,
1774		- trans->state >= TRANS_STATE_COMMIT_START \|\| trans->aborted);
	1852	+ trans->state >= TRANS_STATE_COMMIT_START \|\|
	1853	+ TRANS_ABORTED(trans));
1775	1854	}
1776	1855
1777	1856	/*
..	..	@@ -1783,7 +1862,8 @@
1783	1862	struct btrfs_transaction *trans)
1784	1863	{
1785	1864	wait_event(fs_info->transaction_wait,
1786		- trans->state >= TRANS_STATE_UNBLOCKED \|\| trans->aborted);
	1865	+ trans->state >= TRANS_STATE_UNBLOCKED \|\|
	1866	+ TRANS_ABORTED(trans));
1787	1867	}
1788	1868
1789	1869	/*
..	..	@@ -1865,7 +1945,6 @@
1865	1945	{
1866	1946	struct btrfs_fs_info *fs_info = trans->fs_info;
1867	1947	struct btrfs_transaction *cur_trans = trans->transaction;
1868		- DEFINE_WAIT(wait);
1869	1948
1870	1949	WARN_ON(refcount_read(&trans->use_count) > 1);
1871	1950
..	..	@@ -1880,7 +1959,6 @@
1880	1959	*/
1881	1960	BUG_ON(list_empty(&cur_trans->list));
1882	1961
1883		- list_del_init(&cur_trans->list);
1884	1962	if (cur_trans == fs_info->running_transaction) {
1885	1963	cur_trans->state = TRANS_STATE_COMMIT_DOING;
1886	1964	spin_unlock(&fs_info->trans_lock);
..	..	@@ -1889,6 +1967,17 @@
1889	1967
1890	1968	spin_lock(&fs_info->trans_lock);
1891	1969	}
	1970	+
	1971	+ /*
	1972	+ * Now that we know no one else is still using the transaction we can
	1973	+ * remove the transaction from the list of transactions. This avoids
	1974	+ * the transaction kthread from cleaning up the transaction while some
	1975	+ * other task is still using it, which could result in a use-after-free
	1976	+ * on things like log trees, as it forces the transaction kthread to
	1977	+ * wait for this transaction to be cleaned up by us.
	1978	+ */
	1979	+ list_del_init(&cur_trans->list);
	1980	+
1892	1981	spin_unlock(&fs_info->trans_lock);
1893	1982
1894	1983	btrfs_cleanup_one_transaction(trans->transaction, fs_info);
..	..	@@ -1912,8 +2001,25 @@
1912	2001	kmem_cache_free(btrfs_trans_handle_cachep, trans);
1913	2002	}
1914	2003
1915		-static inline int btrfs_start_delalloc_flush(struct btrfs_fs_info *fs_info)
	2004	+/*
	2005	+ * Release reserved delayed ref space of all pending block groups of the
	2006	+ * transaction and remove them from the list
	2007	+ */
	2008	+static void btrfs_cleanup_pending_block_groups(struct btrfs_trans_handle *trans)
1916	2009	{
	2010	+ struct btrfs_fs_info *fs_info = trans->fs_info;
	2011	+ struct btrfs_block_group block_group, tmp;
	2012	+
	2013	+ list_for_each_entry_safe(block_group, tmp, &trans->new_bgs, bg_list) {
	2014	+ btrfs_delayed_refs_rsv_release(fs_info, 1);
	2015	+ list_del_init(&block_group->bg_list);
	2016	+ }
	2017	+}
	2018	+
	2019	+static inline int btrfs_start_delalloc_flush(struct btrfs_trans_handle *trans)
	2020	+{
	2021	+ struct btrfs_fs_info *fs_info = trans->fs_info;
	2022	+
1917	2023	/*
1918	2024	* We use writeback_inodes_sb here because if we used
1919	2025	* btrfs_start_delalloc_roots we would deadlock with fs freeze.
..	..	@@ -1923,22 +2029,50 @@
1923	2029	* from already being in a transaction and our join_transaction doesn't
1924	2030	* have to re-take the fs freeze lock.
1925	2031	*/
1926		- if (btrfs_test_opt(fs_info, FLUSHONCOMMIT))
	2032	+ if (btrfs_test_opt(fs_info, FLUSHONCOMMIT)) {
1927	2033	writeback_inodes_sb(fs_info->sb, WB_REASON_SYNC);
	2034	+ } else {
	2035	+ struct btrfs_pending_snapshot *pending;
	2036	+ struct list_head *head = &trans->transaction->pending_snapshots;
	2037	+
	2038	+ /*
	2039	+ * Flush dellaloc for any root that is going to be snapshotted.
	2040	+ * This is done to avoid a corrupted version of files, in the
	2041	+ * snapshots, that had both buffered and direct IO writes (even
	2042	+ * if they were done sequentially) due to an unordered update of
	2043	+ * the inode's size on disk.
	2044	+ */
	2045	+ list_for_each_entry(pending, head, list) {
	2046	+ int ret;
	2047	+
	2048	+ ret = btrfs_start_delalloc_snapshot(pending->root);
	2049	+ if (ret)
	2050	+ return ret;
	2051	+ }
	2052	+ }
1928	2053	return 0;
1929	2054	}
1930	2055
1931		-static inline void btrfs_wait_delalloc_flush(struct btrfs_fs_info *fs_info)
	2056	+static inline void btrfs_wait_delalloc_flush(struct btrfs_trans_handle *trans)
1932	2057	{
1933		- if (btrfs_test_opt(fs_info, FLUSHONCOMMIT))
1934		- btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1935		-}
	2058	+ struct btrfs_fs_info *fs_info = trans->fs_info;
1936	2059
1937		-static inline void
1938		-btrfs_wait_pending_ordered(struct btrfs_transaction *cur_trans)
1939		-{
1940		- wait_event(cur_trans->pending_wait,
1941		- atomic_read(&cur_trans->pending_ordered) == 0);
	2060	+ if (btrfs_test_opt(fs_info, FLUSHONCOMMIT)) {
	2061	+ btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
	2062	+ } else {
	2063	+ struct btrfs_pending_snapshot *pending;
	2064	+ struct list_head *head = &trans->transaction->pending_snapshots;
	2065	+
	2066	+ /*
	2067	+ * Wait for any dellaloc that we started previously for the roots
	2068	+ * that are going to be snapshotted. This is to avoid a corrupted
	2069	+ * version of files in the snapshots that had both buffered and
	2070	+ * direct IO writes (even if they were done sequentially).
	2071	+ */
	2072	+ list_for_each_entry(pending, head, list)
	2073	+ btrfs_wait_ordered_extents(pending->root,
	2074	+ U64_MAX, 0, U64_MAX);
	2075	+ }
1942	2076	}
1943	2077
1944	2078	int btrfs_commit_transaction(struct btrfs_trans_handle *trans)
..	..	@@ -1947,6 +2081,8 @@
1947	2081	struct btrfs_transaction *cur_trans = trans->transaction;
1948	2082	struct btrfs_transaction *prev_trans = NULL;
1949	2083	int ret;
	2084	+
	2085	+ ASSERT(refcount_read(&trans->use_count) == 1);
1950	2086
1951	2087	/*
1952	2088	* Some places just start a transaction to commit it. We need to make
..	..	@@ -1957,7 +2093,7 @@
1957	2093	trans->dirty = true;
1958	2094
1959	2095	/* Stop the commit early if ->aborted is set */
1960		- if (unlikely(READ_ONCE(cur_trans->aborted))) {
	2096	+ if (TRANS_ABORTED(cur_trans)) {
1961	2097	ret = cur_trans->aborted;
1962	2098	btrfs_end_transaction(trans);
1963	2099	return ret;
..	..	@@ -1984,8 +2120,7 @@
1984	2120	cur_trans->delayed_refs.flushing = 1;
1985	2121	smp_wmb();
1986	2122
1987		- if (!list_empty(&trans->new_bgs))
1988		- btrfs_create_pending_block_groups(trans);
	2123	+ btrfs_create_pending_block_groups(trans);
1989	2124
1990	2125	ret = btrfs_run_delayed_refs(trans, 0);
1991	2126	if (ret) {
..	..	@@ -2032,7 +2167,7 @@
2032	2167
2033	2168	wait_for_commit(cur_trans);
2034	2169
2035		- if (unlikely(cur_trans->aborted))
	2170	+ if (TRANS_ABORTED(cur_trans))
2036	2171	ret = cur_trans->aborted;
2037	2172
2038	2173	btrfs_put_transaction(cur_trans);
..	..	@@ -2051,7 +2186,7 @@
2051	2186	spin_unlock(&fs_info->trans_lock);
2052	2187
2053	2188	wait_for_commit(prev_trans);
2054		- ret = prev_trans->aborted;
	2189	+ ret = READ_ONCE(prev_trans->aborted);
2055	2190
2056	2191	btrfs_put_transaction(prev_trans);
2057	2192	if (ret)
..	..	@@ -2075,7 +2210,7 @@
2075	2210
2076	2211	extwriter_counter_dec(cur_trans, trans->type);
2077	2212
2078		- ret = btrfs_start_delalloc_flush(fs_info);
	2213	+ ret = btrfs_start_delalloc_flush(trans);
2079	2214	if (ret)
2080	2215	goto cleanup_transaction;
2081	2216
..	..	@@ -2091,9 +2226,15 @@
2091	2226	if (ret)
2092	2227	goto cleanup_transaction;
2093	2228
2094		- btrfs_wait_delalloc_flush(fs_info);
	2229	+ btrfs_wait_delalloc_flush(trans);
2095	2230
2096		- btrfs_wait_pending_ordered(cur_trans);
	2231	+ /*
	2232	+ * Wait for all ordered extents started by a fast fsync that joined this
	2233	+ * transaction. Otherwise if this transaction commits before the ordered
	2234	+ * extents complete we lose logged data after a power failure.
	2235	+ */
	2236	+ wait_event(cur_trans->pending_wait,
	2237	+ atomic_read(&cur_trans->pending_ordered) == 0);
2097	2238
2098	2239	btrfs_scrub_pause(fs_info);
2099	2240	/*
..	..	@@ -2107,8 +2248,7 @@
2107	2248	wait_event(cur_trans->writer_wait,
2108	2249	atomic_read(&cur_trans->num_writers) == 1);
2109	2250
2110		- /* ->aborted might be set after the previous check, so check it */
2111		- if (unlikely(READ_ONCE(cur_trans->aborted))) {
	2251	+ if (TRANS_ABORTED(cur_trans)) {
2112	2252	ret = cur_trans->aborted;
2113	2253	goto scrub_continue;
2114	2254	}
..	..	@@ -2125,10 +2265,8 @@
2125	2265	* core function of the snapshot creation.
2126	2266	*/
2127	2267	ret = create_pending_snapshots(trans);
2128		- if (ret) {
2129		- mutex_unlock(&fs_info->reloc_mutex);
2130		- goto scrub_continue;
2131		- }
	2268	+ if (ret)
	2269	+ goto unlock_reloc;
2132	2270
2133	2271	/*
2134	2272	* We insert the dir indexes of the snapshots and update the inode
..	..	@@ -2141,16 +2279,12 @@
2141	2279	* the nodes and leaves.
2142	2280	*/
2143	2281	ret = btrfs_run_delayed_items(trans);
2144		- if (ret) {
2145		- mutex_unlock(&fs_info->reloc_mutex);
2146		- goto scrub_continue;
2147		- }
	2282	+ if (ret)
	2283	+ goto unlock_reloc;
2148	2284
2149	2285	ret = btrfs_run_delayed_refs(trans, (unsigned long)-1);
2150		- if (ret) {
2151		- mutex_unlock(&fs_info->reloc_mutex);
2152		- goto scrub_continue;
2153		- }
	2286	+ if (ret)
	2287	+ goto unlock_reloc;
2154	2288
2155	2289	/*
2156	2290	* make sure none of the code above managed to slip in a
..	..	@@ -2176,11 +2310,8 @@
2176	2310	mutex_lock(&fs_info->tree_log_mutex);
2177	2311
2178	2312	ret = commit_fs_roots(trans);
2179		- if (ret) {
2180		- mutex_unlock(&fs_info->tree_log_mutex);
2181		- mutex_unlock(&fs_info->reloc_mutex);
2182		- goto scrub_continue;
2183		- }
	2313	+ if (ret)
	2314	+ goto unlock_tree_log;
2184	2315
2185	2316	/*
2186	2317	* Since the transaction is done, we can apply the pending changes
..	..	@@ -2198,42 +2329,29 @@
2198	2329	* new delayed refs. Must handle them or qgroup can be wrong.
2199	2330	*/
2200	2331	ret = btrfs_run_delayed_refs(trans, (unsigned long)-1);
2201		- if (ret) {
2202		- mutex_unlock(&fs_info->tree_log_mutex);
2203		- mutex_unlock(&fs_info->reloc_mutex);
2204		- goto scrub_continue;
2205		- }
	2332	+ if (ret)
	2333	+ goto unlock_tree_log;
2206	2334
2207	2335	/*
2208	2336	* Since fs roots are all committed, we can get a quite accurate
2209	2337	* new_roots. So let's do quota accounting.
2210	2338	*/
2211	2339	ret = btrfs_qgroup_account_extents(trans);
2212		- if (ret < 0) {
2213		- mutex_unlock(&fs_info->tree_log_mutex);
2214		- mutex_unlock(&fs_info->reloc_mutex);
2215		- goto scrub_continue;
2216		- }
	2340	+ if (ret < 0)
	2341	+ goto unlock_tree_log;
2217	2342
2218	2343	ret = commit_cowonly_roots(trans);
2219		- if (ret) {
2220		- mutex_unlock(&fs_info->tree_log_mutex);
2221		- mutex_unlock(&fs_info->reloc_mutex);
2222		- goto scrub_continue;
2223		- }
	2344	+ if (ret)
	2345	+ goto unlock_tree_log;
2224	2346
2225	2347	/*
2226	2348	* The tasks which save the space cache and inode cache may also
2227	2349	* update ->aborted, check it.
2228	2350	*/
2229		- if (unlikely(READ_ONCE(cur_trans->aborted))) {
	2351	+ if (TRANS_ABORTED(cur_trans)) {
2230	2352	ret = cur_trans->aborted;
2231		- mutex_unlock(&fs_info->tree_log_mutex);
2232		- mutex_unlock(&fs_info->reloc_mutex);
2233		- goto scrub_continue;
	2353	+ goto unlock_tree_log;
2234	2354	}
2235		-
2236		- btrfs_prepare_extent_commit(fs_info);
2237	2355
2238	2356	cur_trans = fs_info->running_transaction;
2239	2357
..	..	@@ -2247,7 +2365,7 @@
2247	2365	list_add_tail(&fs_info->chunk_root->dirty_list,
2248	2366	&cur_trans->switch_commits);
2249	2367
2250		- switch_commit_roots(cur_trans);
	2368	+ switch_commit_roots(trans);
2251	2369
2252	2370	ASSERT(list_empty(&cur_trans->dirty_bgs));
2253	2371	ASSERT(list_empty(&cur_trans->io_bgs));
..	..	@@ -2258,8 +2376,7 @@
2258	2376	memcpy(fs_info->super_for_commit, fs_info->super_copy,
2259	2377	sizeof(*fs_info->super_copy));
2260	2378
2261		- btrfs_update_commit_device_size(fs_info);
2262		- btrfs_update_commit_device_bytes_used(cur_trans);
	2379	+ btrfs_commit_device_sizes(cur_trans);
2263	2380
2264	2381	clear_bit(BTRFS_FS_LOG1_ERR, &fs_info->flags);
2265	2382	clear_bit(BTRFS_FS_LOG2_ERR, &fs_info->flags);
..	..	@@ -2278,6 +2395,10 @@
2278	2395	if (ret) {
2279	2396	btrfs_handle_fs_error(fs_info, ret,
2280	2397	"Error while writing out transaction");
	2398	+ /*
	2399	+ * reloc_mutex has been unlocked, tree_log_mutex is still held
	2400	+ * but we can't jump to unlock_tree_log causing double unlock
	2401	+ */
2281	2402	mutex_unlock(&fs_info->tree_log_mutex);
2282	2403	goto scrub_continue;
2283	2404	}
..	..	@@ -2303,7 +2424,6 @@
2303	2424	*/
2304	2425	cur_trans->state = TRANS_STATE_COMPLETED;
2305	2426	wake_up(&cur_trans->commit_wait);
2306		- clear_bit(BTRFS_FS_NEED_ASYNC_COMMIT, &fs_info->flags);
2307	2427
2308	2428	spin_lock(&fs_info->trans_lock);
2309	2429	list_del_init(&cur_trans->list);
..	..	@@ -2326,10 +2446,15 @@
2326	2446
2327	2447	return ret;
2328	2448
	2449	+unlock_tree_log:
	2450	+ mutex_unlock(&fs_info->tree_log_mutex);
	2451	+unlock_reloc:
	2452	+ mutex_unlock(&fs_info->reloc_mutex);
2329	2453	scrub_continue:
2330	2454	btrfs_scrub_continue(fs_info);
2331	2455	cleanup_transaction:
2332	2456	btrfs_trans_release_metadata(trans);
	2457	+ btrfs_cleanup_pending_block_groups(trans);
2333	2458	btrfs_trans_release_chunk_metadata(trans);
2334	2459	trans->block_rsv = NULL;
2335	2460	btrfs_warn(fs_info, "Skipping commit of aborted transaction.");
..	..	@@ -2365,16 +2490,21 @@
2365	2490	list_del_init(&root->root_list);
2366	2491	spin_unlock(&fs_info->trans_lock);
2367	2492
2368		- btrfs_debug(fs_info, "cleaner removing %llu", root->objectid);
	2493	+ btrfs_debug(fs_info, "cleaner removing %llu", root->root_key.objectid);
2369	2494
2370	2495	btrfs_kill_all_delayed_nodes(root);
	2496	+ if (root->ino_cache_inode) {
	2497	+ iput(root->ino_cache_inode);
	2498	+ root->ino_cache_inode = NULL;
	2499	+ }
2371	2500
2372	2501	if (btrfs_header_backref_rev(root->node) <
2373	2502	BTRFS_MIXED_BACKREF_REV)
2374		- ret = btrfs_drop_snapshot(root, NULL, 0, 0);
	2503	+ ret = btrfs_drop_snapshot(root, 0, 0);
2375	2504	else
2376		- ret = btrfs_drop_snapshot(root, NULL, 1, 0);
	2505	+ ret = btrfs_drop_snapshot(root, 1, 0);
2377	2506
	2507	+ btrfs_put_root(root);
2378	2508	return (ret < 0) ? 0 : 1;
2379	2509	}
2380	2510