jbd2: Remove __GFP_NOFAIL from jbd2 layer

__GFP_NOFAIL is going away, so add our own retry loop. Also add jbd2__journal_start() and jbd2__journal_restart() which take a gfp mask, so that file systems can optionally (re)start transaction handles using GFP_KERNEL. If they do this, then they need to be prepared to handle receiving an PTR_ERR(-ENOMEM) error, and be ready to reflect that error up to userspace. Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

jbd2: Remove __GFP_NOFAIL from jbd2 layer
__GFP_NOFAIL is going away, so add our own retry loop. Also add jbd2__journal_start() and jbd2__journal_restart() which take a gfp mask, so that file systems can optionally (re)start transaction handles using GFP_KERNEL. If they do this, then they need to be prepared to handle receiving an PTR_ERR(-ENOMEM) error, and be ready to reflect that error up to userspace. Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Theodore Ts'o
1 parent 4038968738
Showing 3 changed files with 57 additions and 23 deletions Side-by-side Diff
fs/jbd2/journal.c
fs/jbd2/transaction.c
include/linux/jbd2.h
@@ -41,6 +41,7 @@
 #include <linux/hash.h>
 #include <linux/log2.h>
 #include <linux/vmalloc.h>
+#include <linux/backing-dev.h>
  
 #define CREATE_TRACE_POINTS
 #include <trace/events/jbd2.h>
@@ -48,8 +49,6 @@
 #include <asm/uaccess.h>
 #include <asm/page.h>
  
-EXPORT_SYMBOL(jbd2_journal_start);
-EXPORT_SYMBOL(jbd2_journal_restart);
 EXPORT_SYMBOL(jbd2_journal_extend);
 EXPORT_SYMBOL(jbd2_journal_stop);
 EXPORT_SYMBOL(jbd2_journal_lock_updates);
@@ -311,7 +310,17 @@
 	 */
 	J_ASSERT_BH(bh_in, buffer_jbddirty(bh_in));
  
-	new_bh = alloc_buffer_head(GFP_NOFS|__GFP_NOFAIL);
+retry_alloc:
+	new_bh = alloc_buffer_head(GFP_NOFS);
+	if (!new_bh) {
+		/*
+		 * Failure is not an option, but __GFP_NOFAIL is going
+		 * away; so we retry ourselves here.
+		 */
+		congestion_wait(BLK_RW_ASYNC, HZ/50);
+		goto retry_alloc;
+	}
+
 	/* keep subsequent assertions sane */
 	new_bh->b_state = 0;
 	init_buffer(new_bh, NULL, NULL);
@@ -26,6 +26,8 @@
 #include <linux/mm.h>
 #include <linux/highmem.h>
 #include <linux/hrtimer.h>
+#include <linux/backing-dev.h>
+#include <linux/module.h>
  
 static void __jbd2_journal_temp_unlink_buffer(struct journal_head *jh);
  
  
  
  
  
@@ -83,30 +85,38 @@
  * transaction's buffer credits.
  */
  
-static int start_this_handle(journal_t *journal, handle_t *handle)
+static int start_this_handle(journal_t *journal, handle_t *handle,
+			     int gfp_mask)
 {
 	transaction_t *transaction;
 	int needed;
 	int nblocks = handle->h_buffer_credits;
 	transaction_t *new_transaction = NULL;
-	int ret = 0;
 	unsigned long ts = jiffies;
  
 	if (nblocks > journal->j_max_transaction_buffers) {
 		printk(KERN_ERR "JBD: %s wants too many credits (%d > %d)\n",
 		       current->comm, nblocks,
 		       journal->j_max_transaction_buffers);
-		ret = -ENOSPC;
-		goto out;
+		return -ENOSPC;
 	}
  
 alloc_transaction:
 	if (!journal->j_running_transaction) {
-		new_transaction = kzalloc(sizeof(*new_transaction),
-						GFP_NOFS|__GFP_NOFAIL);
+		new_transaction = kzalloc(sizeof(*new_transaction), gfp_mask);
 		if (!new_transaction) {
-			ret = -ENOMEM;
-			goto out;
+			/*
+			 * If __GFP_FS is not present, then we may be
+			 * being called from inside the fs writeback
+			 * layer, so we MUST NOT fail.  Since
+			 * __GFP_NOFAIL is going away, we will arrange
+			 * to retry the allocation ourselves.
+			 */
+			if ((gfp_mask & __GFP_FS) == 0) {
+				congestion_wait(BLK_RW_ASYNC, HZ/50);
+				goto alloc_transaction;
+			}
+			return -ENOMEM;
 		}
 	}
  
@@ -123,8 +133,8 @@
 	if (is_journal_aborted(journal) ||
 	    (journal->j_errno != 0 && !(journal->j_flags & JBD2_ACK_ERR))) {
 		spin_unlock(&journal->j_state_lock);
-		ret = -EROFS;
-		goto out;
+		kfree(new_transaction);
+		return -EROFS;
 	}
  
 	/* Wait on the journal's transaction barrier if necessary */
@@ -240,10 +250,8 @@
 	spin_unlock(&journal->j_state_lock);
  
 	lock_map_acquire(&handle->h_lockdep_map);
-out:
-	if (unlikely(new_transaction))		/* It's usually NULL */
-		kfree(new_transaction);
-	return ret;
+	kfree(new_transaction);
+	return 0;
 }
  
 static struct lock_class_key jbd2_handle_key;
@@ -278,7 +286,7 @@
  *
  * Return a pointer to a newly allocated handle, or NULL on failure
  */
-handle_t *jbd2_journal_start(journal_t *journal, int nblocks)
+handle_t *jbd2__journal_start(journal_t *journal, int nblocks, int gfp_mask)
 {
 	handle_t *handle = journal_current_handle();
 	int err;
@@ -298,7 +306,7 @@
  
 	current->journal_info = handle;
  
-	err = start_this_handle(journal, handle);
+	err = start_this_handle(journal, handle, gfp_mask);
 	if (err < 0) {
 		jbd2_free_handle(handle);
 		current->journal_info = NULL;
  
@@ -308,7 +316,16 @@
 out:
 	return handle;
 }
+EXPORT_SYMBOL(jbd2__journal_start);
  
+
+handle_t *jbd2_journal_start(journal_t *journal, int nblocks)
+{
+	return jbd2__journal_start(journal, nblocks, GFP_NOFS);
+}
+EXPORT_SYMBOL(jbd2_journal_start);
+
+
 /**
  * int jbd2_journal_extend() - extend buffer credits.
  * @handle:  handle to 'extend'
@@ -394,8 +411,7 @@
  * transaction capabable of guaranteeing the requested number of
  * credits.
  */
-
-int jbd2_journal_restart(handle_t *handle, int nblocks)
+int jbd2__journal_restart(handle_t *handle, int nblocks, int gfp_mask)
 {
 	transaction_t *transaction = handle->h_transaction;
 	journal_t *journal = transaction->t_journal;
  
  
@@ -428,10 +444,17 @@
  
 	lock_map_release(&handle->h_lockdep_map);
 	handle->h_buffer_credits = nblocks;
-	ret = start_this_handle(journal, handle);
+	ret = start_this_handle(journal, handle, gfp_mask);
 	return ret;
 }
+EXPORT_SYMBOL(jbd2__journal_restart);
  
+
+int jbd2_journal_restart(handle_t *handle, int nblocks)
+{
+	return jbd2__journal_restart(handle, nblocks, GFP_NOFS);
+}
+EXPORT_SYMBOL(jbd2_journal_restart);
  
 /**
  * void jbd2_journal_lock_updates () - establish a transaction barrier.
@@ -1081,7 +1081,9 @@
  */
  
 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
-extern int	 jbd2_journal_restart (handle_t *, int nblocks);
+extern handle_t *jbd2__journal_start(journal_t *, int nblocks, int gfp_mask);
+extern int	 jbd2_journal_restart(handle_t *, int nblocks);
+extern int	 jbd2__journal_restart(handle_t *, int nblocks, int gfp_mask);
 extern int	 jbd2_journal_extend (handle_t *, int nblocks);
 extern int	 jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
 extern int	 jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
...	...	@@ -41,6 +41,7 @@
41	41	#include <linux/hash.h>
42	42	#include <linux/log2.h>
43	43	#include <linux/vmalloc.h>
	44	+#include <linux/backing-dev.h>
44	45
45	46	#define CREATE_TRACE_POINTS
46	47	#include <trace/events/jbd2.h>
...	...	@@ -48,8 +49,6 @@
48	49	#include <asm/uaccess.h>
49	50	#include <asm/page.h>
50	51
51		-EXPORT_SYMBOL(jbd2_journal_start);
52		-EXPORT_SYMBOL(jbd2_journal_restart);
53	52	EXPORT_SYMBOL(jbd2_journal_extend);
54	53	EXPORT_SYMBOL(jbd2_journal_stop);
55	54	EXPORT_SYMBOL(jbd2_journal_lock_updates);
...	...	@@ -311,7 +310,17 @@
311	310	*/
312	311	J_ASSERT_BH(bh_in, buffer_jbddirty(bh_in));
313	312
314		- new_bh = alloc_buffer_head(GFP_NOFS\|__GFP_NOFAIL);
	313	+retry_alloc:
	314	+ new_bh = alloc_buffer_head(GFP_NOFS);
	315	+ if (!new_bh) {
	316	+ /*
	317	+ * Failure is not an option, but __GFP_NOFAIL is going
	318	+ * away; so we retry ourselves here.
	319	+ */
	320	+ congestion_wait(BLK_RW_ASYNC, HZ/50);
	321	+ goto retry_alloc;
	322	+ }
	323	+
315	324	/* keep subsequent assertions sane */
316	325	new_bh->b_state = 0;
317	326	init_buffer(new_bh, NULL, NULL);
...	...	@@ -26,6 +26,8 @@
26	26	#include <linux/mm.h>
27	27	#include <linux/highmem.h>
28	28	#include <linux/hrtimer.h>
	29	+#include <linux/backing-dev.h>
	30	+#include <linux/module.h>
29	31
30	32	static void __jbd2_journal_temp_unlink_buffer(struct journal_head *jh);
31	33
32	34
33	35
34	36
35	37
...	...	@@ -83,30 +85,38 @@
83	85	* transaction's buffer credits.
84	86	*/
85	87
86		-static int start_this_handle(journal_t journal, handle_t handle)
	88	+static int start_this_handle(journal_t journal, handle_t handle,
	89	+ int gfp_mask)
87	90	{
88	91	transaction_t *transaction;
89	92	int needed;
90	93	int nblocks = handle->h_buffer_credits;
91	94	transaction_t *new_transaction = NULL;
92		- int ret = 0;
93	95	unsigned long ts = jiffies;
94	96
95	97	if (nblocks > journal->j_max_transaction_buffers) {
96	98	printk(KERN_ERR "JBD: %s wants too many credits (%d > %d)\n",
97	99	current->comm, nblocks,
98	100	journal->j_max_transaction_buffers);
99		- ret = -ENOSPC;
100		- goto out;
	101	+ return -ENOSPC;
101	102	}
102	103
103	104	alloc_transaction:
104	105	if (!journal->j_running_transaction) {
105		- new_transaction = kzalloc(sizeof(*new_transaction),
106		- GFP_NOFS\|__GFP_NOFAIL);
	106	+ new_transaction = kzalloc(sizeof(*new_transaction), gfp_mask);
107	107	if (!new_transaction) {
108		- ret = -ENOMEM;
109		- goto out;
	108	+ /*
	109	+ * If __GFP_FS is not present, then we may be
	110	+ * being called from inside the fs writeback
	111	+ * layer, so we MUST NOT fail. Since
	112	+ * __GFP_NOFAIL is going away, we will arrange
	113	+ * to retry the allocation ourselves.
	114	+ */
	115	+ if ((gfp_mask & __GFP_FS) == 0) {
	116	+ congestion_wait(BLK_RW_ASYNC, HZ/50);
	117	+ goto alloc_transaction;
	118	+ }
	119	+ return -ENOMEM;
110	120	}
111	121	}
112	122
...	...	@@ -123,8 +133,8 @@
123	133	if (is_journal_aborted(journal) \|\|
124	134	(journal->j_errno != 0 && !(journal->j_flags & JBD2_ACK_ERR))) {
125	135	spin_unlock(&journal->j_state_lock);
126		- ret = -EROFS;
127		- goto out;
	136	+ kfree(new_transaction);
	137	+ return -EROFS;
128	138	}
129	139
130	140	/* Wait on the journal's transaction barrier if necessary */
...	...	@@ -240,10 +250,8 @@
240	250	spin_unlock(&journal->j_state_lock);
241	251
242	252	lock_map_acquire(&handle->h_lockdep_map);
243		-out:
244		- if (unlikely(new_transaction)) /* It's usually NULL */
245		- kfree(new_transaction);
246		- return ret;
	253	+ kfree(new_transaction);
	254	+ return 0;
247	255	}
248	256
249	257	static struct lock_class_key jbd2_handle_key;
...	...	@@ -278,7 +286,7 @@
278	286	*
279	287	* Return a pointer to a newly allocated handle, or NULL on failure
280	288	*/
281		-handle_t jbd2_journal_start(journal_t journal, int nblocks)
	289	+handle_t jbd2__journal_start(journal_t journal, int nblocks, int gfp_mask)
282	290	{
283	291	handle_t *handle = journal_current_handle();
284	292	int err;
...	...	@@ -298,7 +306,7 @@
298	306
299	307	current->journal_info = handle;
300	308
301		- err = start_this_handle(journal, handle);
	309	+ err = start_this_handle(journal, handle, gfp_mask);
302	310	if (err < 0) {
303	311	jbd2_free_handle(handle);
304	312	current->journal_info = NULL;
305	313
...	...	@@ -308,7 +316,16 @@
308	316	out:
309	317	return handle;
310	318	}
	319	+EXPORT_SYMBOL(jbd2__journal_start);
311	320
	321	+
	322	+handle_t jbd2_journal_start(journal_t journal, int nblocks)
	323	+{
	324	+ return jbd2__journal_start(journal, nblocks, GFP_NOFS);
	325	+}
	326	+EXPORT_SYMBOL(jbd2_journal_start);
	327	+
	328	+
312	329	/**
313	330	* int jbd2_journal_extend() - extend buffer credits.
314	331	* @handle: handle to 'extend'
...	...	@@ -394,8 +411,7 @@
394	411	* transaction capabable of guaranteeing the requested number of
395	412	* credits.
396	413	*/
397		-
398		-int jbd2_journal_restart(handle_t *handle, int nblocks)
	414	+int jbd2__journal_restart(handle_t *handle, int nblocks, int gfp_mask)
399	415	{
400	416	transaction_t *transaction = handle->h_transaction;
401	417	journal_t *journal = transaction->t_journal;
402	418
403	419
...	...	@@ -428,10 +444,17 @@
428	444
429	445	lock_map_release(&handle->h_lockdep_map);
430	446	handle->h_buffer_credits = nblocks;
431		- ret = start_this_handle(journal, handle);
	447	+ ret = start_this_handle(journal, handle, gfp_mask);
432	448	return ret;
433	449	}
	450	+EXPORT_SYMBOL(jbd2__journal_restart);
434	451
	452	+
	453	+int jbd2_journal_restart(handle_t *handle, int nblocks)
	454	+{
	455	+ return jbd2__journal_restart(handle, nblocks, GFP_NOFS);
	456	+}
	457	+EXPORT_SYMBOL(jbd2_journal_restart);
435	458
436	459	/**
437	460	* void jbd2_journal_lock_updates () - establish a transaction barrier.
...	...	@@ -1081,7 +1081,9 @@
1081	1081	*/
1082	1082
1083	1083	extern handle_t jbd2_journal_start(journal_t , int nblocks);
1084		-extern int jbd2_journal_restart (handle_t *, int nblocks);
	1084	+extern handle_t jbd2__journal_start(journal_t , int nblocks, int gfp_mask);
	1085	+extern int jbd2_journal_restart(handle_t *, int nblocks);
	1086	+extern int jbd2__journal_restart(handle_t *, int nblocks, int gfp_mask);
1085	1087	extern int jbd2_journal_extend (handle_t *, int nblocks);
1086	1088	extern int jbd2_journal_get_write_access(handle_t , struct buffer_head );
1087	1089	extern int jbd2_journal_get_create_access (handle_t , struct buffer_head );