diff options
author | Ingo Molnar <mingo@elte.hu> | 2008-10-13 05:05:51 -0400 |
---|---|---|
committer | Ingo Molnar <mingo@elte.hu> | 2008-10-13 05:05:51 -0400 |
commit | accba5f3965d6a9d1bf7c1e1a7995d17e9d521b6 (patch) | |
tree | 8fb40782e79472ed882ff2098d4dd295557278ee /fs/jbd2 | |
parent | 6852fd9b86d05063c6ef49d2e12e061cc7f6a105 (diff) | |
parent | 4480f15b3306f43bbb0310d461142b4e897ca45b (diff) |
Merge branch 'linus' into oprofile-v2
Conflicts:
arch/x86/kernel/apic_32.c
arch/x86/oprofile/nmi_int.c
include/linux/pci_ids.h
Diffstat (limited to 'fs/jbd2')
-rw-r--r-- | fs/jbd2/checkpoint.c | 71 | ||||
-rw-r--r-- | fs/jbd2/commit.c | 52 | ||||
-rw-r--r-- | fs/jbd2/journal.c | 104 | ||||
-rw-r--r-- | fs/jbd2/recovery.c | 7 | ||||
-rw-r--r-- | fs/jbd2/transaction.c | 4 |
5 files changed, 165 insertions, 73 deletions
diff --git a/fs/jbd2/checkpoint.c b/fs/jbd2/checkpoint.c index 91389c8aee8a..9203c3332f17 100644 --- a/fs/jbd2/checkpoint.c +++ b/fs/jbd2/checkpoint.c | |||
@@ -20,6 +20,7 @@ | |||
20 | #include <linux/time.h> | 20 | #include <linux/time.h> |
21 | #include <linux/fs.h> | 21 | #include <linux/fs.h> |
22 | #include <linux/jbd2.h> | 22 | #include <linux/jbd2.h> |
23 | #include <linux/marker.h> | ||
23 | #include <linux/errno.h> | 24 | #include <linux/errno.h> |
24 | #include <linux/slab.h> | 25 | #include <linux/slab.h> |
25 | 26 | ||
@@ -93,7 +94,8 @@ static int __try_to_free_cp_buf(struct journal_head *jh) | |||
93 | int ret = 0; | 94 | int ret = 0; |
94 | struct buffer_head *bh = jh2bh(jh); | 95 | struct buffer_head *bh = jh2bh(jh); |
95 | 96 | ||
96 | if (jh->b_jlist == BJ_None && !buffer_locked(bh) && !buffer_dirty(bh)) { | 97 | if (jh->b_jlist == BJ_None && !buffer_locked(bh) && |
98 | !buffer_dirty(bh) && !buffer_write_io_error(bh)) { | ||
97 | JBUFFER_TRACE(jh, "remove from checkpoint list"); | 99 | JBUFFER_TRACE(jh, "remove from checkpoint list"); |
98 | ret = __jbd2_journal_remove_checkpoint(jh) + 1; | 100 | ret = __jbd2_journal_remove_checkpoint(jh) + 1; |
99 | jbd_unlock_bh_state(bh); | 101 | jbd_unlock_bh_state(bh); |
@@ -126,14 +128,29 @@ void __jbd2_log_wait_for_space(journal_t *journal) | |||
126 | 128 | ||
127 | /* | 129 | /* |
128 | * Test again, another process may have checkpointed while we | 130 | * Test again, another process may have checkpointed while we |
129 | * were waiting for the checkpoint lock | 131 | * were waiting for the checkpoint lock. If there are no |
132 | * outstanding transactions there is nothing to checkpoint and | ||
133 | * we can't make progress. Abort the journal in this case. | ||
130 | */ | 134 | */ |
131 | spin_lock(&journal->j_state_lock); | 135 | spin_lock(&journal->j_state_lock); |
136 | spin_lock(&journal->j_list_lock); | ||
132 | nblocks = jbd_space_needed(journal); | 137 | nblocks = jbd_space_needed(journal); |
133 | if (__jbd2_log_space_left(journal) < nblocks) { | 138 | if (__jbd2_log_space_left(journal) < nblocks) { |
139 | int chkpt = journal->j_checkpoint_transactions != NULL; | ||
140 | |||
141 | spin_unlock(&journal->j_list_lock); | ||
134 | spin_unlock(&journal->j_state_lock); | 142 | spin_unlock(&journal->j_state_lock); |
135 | jbd2_log_do_checkpoint(journal); | 143 | if (chkpt) { |
144 | jbd2_log_do_checkpoint(journal); | ||
145 | } else { | ||
146 | printk(KERN_ERR "%s: no transactions\n", | ||
147 | __func__); | ||
148 | jbd2_journal_abort(journal, 0); | ||
149 | } | ||
150 | |||
136 | spin_lock(&journal->j_state_lock); | 151 | spin_lock(&journal->j_state_lock); |
152 | } else { | ||
153 | spin_unlock(&journal->j_list_lock); | ||
137 | } | 154 | } |
138 | mutex_unlock(&journal->j_checkpoint_mutex); | 155 | mutex_unlock(&journal->j_checkpoint_mutex); |
139 | } | 156 | } |
@@ -160,21 +177,25 @@ static void jbd_sync_bh(journal_t *journal, struct buffer_head *bh) | |||
160 | * buffers. Note that we take the buffers in the opposite ordering | 177 | * buffers. Note that we take the buffers in the opposite ordering |
161 | * from the one in which they were submitted for IO. | 178 | * from the one in which they were submitted for IO. |
162 | * | 179 | * |
180 | * Return 0 on success, and return <0 if some buffers have failed | ||
181 | * to be written out. | ||
182 | * | ||
163 | * Called with j_list_lock held. | 183 | * Called with j_list_lock held. |
164 | */ | 184 | */ |
165 | static void __wait_cp_io(journal_t *journal, transaction_t *transaction) | 185 | static int __wait_cp_io(journal_t *journal, transaction_t *transaction) |
166 | { | 186 | { |
167 | struct journal_head *jh; | 187 | struct journal_head *jh; |
168 | struct buffer_head *bh; | 188 | struct buffer_head *bh; |
169 | tid_t this_tid; | 189 | tid_t this_tid; |
170 | int released = 0; | 190 | int released = 0; |
191 | int ret = 0; | ||
171 | 192 | ||
172 | this_tid = transaction->t_tid; | 193 | this_tid = transaction->t_tid; |
173 | restart: | 194 | restart: |
174 | /* Did somebody clean up the transaction in the meanwhile? */ | 195 | /* Did somebody clean up the transaction in the meanwhile? */ |
175 | if (journal->j_checkpoint_transactions != transaction || | 196 | if (journal->j_checkpoint_transactions != transaction || |
176 | transaction->t_tid != this_tid) | 197 | transaction->t_tid != this_tid) |
177 | return; | 198 | return ret; |
178 | while (!released && transaction->t_checkpoint_io_list) { | 199 | while (!released && transaction->t_checkpoint_io_list) { |
179 | jh = transaction->t_checkpoint_io_list; | 200 | jh = transaction->t_checkpoint_io_list; |
180 | bh = jh2bh(jh); | 201 | bh = jh2bh(jh); |
@@ -194,6 +215,9 @@ restart: | |||
194 | spin_lock(&journal->j_list_lock); | 215 | spin_lock(&journal->j_list_lock); |
195 | goto restart; | 216 | goto restart; |
196 | } | 217 | } |
218 | if (unlikely(buffer_write_io_error(bh))) | ||
219 | ret = -EIO; | ||
220 | |||
197 | /* | 221 | /* |
198 | * Now in whatever state the buffer currently is, we know that | 222 | * Now in whatever state the buffer currently is, we know that |
199 | * it has been written out and so we can drop it from the list | 223 | * it has been written out and so we can drop it from the list |
@@ -203,6 +227,8 @@ restart: | |||
203 | jbd2_journal_remove_journal_head(bh); | 227 | jbd2_journal_remove_journal_head(bh); |
204 | __brelse(bh); | 228 | __brelse(bh); |
205 | } | 229 | } |
230 | |||
231 | return ret; | ||
206 | } | 232 | } |
207 | 233 | ||
208 | #define NR_BATCH 64 | 234 | #define NR_BATCH 64 |
@@ -226,7 +252,8 @@ __flush_batch(journal_t *journal, struct buffer_head **bhs, int *batch_count) | |||
226 | * Try to flush one buffer from the checkpoint list to disk. | 252 | * Try to flush one buffer from the checkpoint list to disk. |
227 | * | 253 | * |
228 | * Return 1 if something happened which requires us to abort the current | 254 | * Return 1 if something happened which requires us to abort the current |
229 | * scan of the checkpoint list. | 255 | * scan of the checkpoint list. Return <0 if the buffer has failed to |
256 | * be written out. | ||
230 | * | 257 | * |
231 | * Called with j_list_lock held and drops it if 1 is returned | 258 | * Called with j_list_lock held and drops it if 1 is returned |
232 | * Called under jbd_lock_bh_state(jh2bh(jh)), and drops it | 259 | * Called under jbd_lock_bh_state(jh2bh(jh)), and drops it |
@@ -258,6 +285,9 @@ static int __process_buffer(journal_t *journal, struct journal_head *jh, | |||
258 | jbd2_log_wait_commit(journal, tid); | 285 | jbd2_log_wait_commit(journal, tid); |
259 | ret = 1; | 286 | ret = 1; |
260 | } else if (!buffer_dirty(bh)) { | 287 | } else if (!buffer_dirty(bh)) { |
288 | ret = 1; | ||
289 | if (unlikely(buffer_write_io_error(bh))) | ||
290 | ret = -EIO; | ||
261 | J_ASSERT_JH(jh, !buffer_jbddirty(bh)); | 291 | J_ASSERT_JH(jh, !buffer_jbddirty(bh)); |
262 | BUFFER_TRACE(bh, "remove from checkpoint"); | 292 | BUFFER_TRACE(bh, "remove from checkpoint"); |
263 | __jbd2_journal_remove_checkpoint(jh); | 293 | __jbd2_journal_remove_checkpoint(jh); |
@@ -265,7 +295,6 @@ static int __process_buffer(journal_t *journal, struct journal_head *jh, | |||
265 | jbd_unlock_bh_state(bh); | 295 | jbd_unlock_bh_state(bh); |
266 | jbd2_journal_remove_journal_head(bh); | 296 | jbd2_journal_remove_journal_head(bh); |
267 | __brelse(bh); | 297 | __brelse(bh); |
268 | ret = 1; | ||
269 | } else { | 298 | } else { |
270 | /* | 299 | /* |
271 | * Important: we are about to write the buffer, and | 300 | * Important: we are about to write the buffer, and |
@@ -298,6 +327,7 @@ static int __process_buffer(journal_t *journal, struct journal_head *jh, | |||
298 | * to disk. We submit larger chunks of data at once. | 327 | * to disk. We submit larger chunks of data at once. |
299 | * | 328 | * |
300 | * The journal should be locked before calling this function. | 329 | * The journal should be locked before calling this function. |
330 | * Called with j_checkpoint_mutex held. | ||
301 | */ | 331 | */ |
302 | int jbd2_log_do_checkpoint(journal_t *journal) | 332 | int jbd2_log_do_checkpoint(journal_t *journal) |
303 | { | 333 | { |
@@ -313,6 +343,8 @@ int jbd2_log_do_checkpoint(journal_t *journal) | |||
313 | * journal straight away. | 343 | * journal straight away. |
314 | */ | 344 | */ |
315 | result = jbd2_cleanup_journal_tail(journal); | 345 | result = jbd2_cleanup_journal_tail(journal); |
346 | trace_mark(jbd2_checkpoint, "dev %s need_checkpoint %d", | ||
347 | journal->j_devname, result); | ||
316 | jbd_debug(1, "cleanup_journal_tail returned %d\n", result); | 348 | jbd_debug(1, "cleanup_journal_tail returned %d\n", result); |
317 | if (result <= 0) | 349 | if (result <= 0) |
318 | return result; | 350 | return result; |
@@ -321,6 +353,7 @@ int jbd2_log_do_checkpoint(journal_t *journal) | |||
321 | * OK, we need to start writing disk blocks. Take one transaction | 353 | * OK, we need to start writing disk blocks. Take one transaction |
322 | * and write it. | 354 | * and write it. |
323 | */ | 355 | */ |
356 | result = 0; | ||
324 | spin_lock(&journal->j_list_lock); | 357 | spin_lock(&journal->j_list_lock); |
325 | if (!journal->j_checkpoint_transactions) | 358 | if (!journal->j_checkpoint_transactions) |
326 | goto out; | 359 | goto out; |
@@ -339,7 +372,7 @@ restart: | |||
339 | int batch_count = 0; | 372 | int batch_count = 0; |
340 | struct buffer_head *bhs[NR_BATCH]; | 373 | struct buffer_head *bhs[NR_BATCH]; |
341 | struct journal_head *jh; | 374 | struct journal_head *jh; |
342 | int retry = 0; | 375 | int retry = 0, err; |
343 | 376 | ||
344 | while (!retry && transaction->t_checkpoint_list) { | 377 | while (!retry && transaction->t_checkpoint_list) { |
345 | struct buffer_head *bh; | 378 | struct buffer_head *bh; |
@@ -353,6 +386,8 @@ restart: | |||
353 | } | 386 | } |
354 | retry = __process_buffer(journal, jh, bhs, &batch_count, | 387 | retry = __process_buffer(journal, jh, bhs, &batch_count, |
355 | transaction); | 388 | transaction); |
389 | if (retry < 0 && !result) | ||
390 | result = retry; | ||
356 | if (!retry && (need_resched() || | 391 | if (!retry && (need_resched() || |
357 | spin_needbreak(&journal->j_list_lock))) { | 392 | spin_needbreak(&journal->j_list_lock))) { |
358 | spin_unlock(&journal->j_list_lock); | 393 | spin_unlock(&journal->j_list_lock); |
@@ -377,14 +412,18 @@ restart: | |||
377 | * Now we have cleaned up the first transaction's checkpoint | 412 | * Now we have cleaned up the first transaction's checkpoint |
378 | * list. Let's clean up the second one | 413 | * list. Let's clean up the second one |
379 | */ | 414 | */ |
380 | __wait_cp_io(journal, transaction); | 415 | err = __wait_cp_io(journal, transaction); |
416 | if (!result) | ||
417 | result = err; | ||
381 | } | 418 | } |
382 | out: | 419 | out: |
383 | spin_unlock(&journal->j_list_lock); | 420 | spin_unlock(&journal->j_list_lock); |
384 | result = jbd2_cleanup_journal_tail(journal); | ||
385 | if (result < 0) | 421 | if (result < 0) |
386 | return result; | 422 | jbd2_journal_abort(journal, result); |
387 | return 0; | 423 | else |
424 | result = jbd2_cleanup_journal_tail(journal); | ||
425 | |||
426 | return (result < 0) ? result : 0; | ||
388 | } | 427 | } |
389 | 428 | ||
390 | /* | 429 | /* |
@@ -400,8 +439,9 @@ out: | |||
400 | * This is the only part of the journaling code which really needs to be | 439 | * This is the only part of the journaling code which really needs to be |
401 | * aware of transaction aborts. Checkpointing involves writing to the | 440 | * aware of transaction aborts. Checkpointing involves writing to the |
402 | * main filesystem area rather than to the journal, so it can proceed | 441 | * main filesystem area rather than to the journal, so it can proceed |
403 | * even in abort state, but we must not update the journal superblock if | 442 | * even in abort state, but we must not update the super block if |
404 | * we have an abort error outstanding. | 443 | * checkpointing may have failed. Otherwise, we would lose some metadata |
444 | * buffers which should be written-back to the filesystem. | ||
405 | */ | 445 | */ |
406 | 446 | ||
407 | int jbd2_cleanup_journal_tail(journal_t *journal) | 447 | int jbd2_cleanup_journal_tail(journal_t *journal) |
@@ -410,6 +450,9 @@ int jbd2_cleanup_journal_tail(journal_t *journal) | |||
410 | tid_t first_tid; | 450 | tid_t first_tid; |
411 | unsigned long blocknr, freed; | 451 | unsigned long blocknr, freed; |
412 | 452 | ||
453 | if (is_journal_aborted(journal)) | ||
454 | return 1; | ||
455 | |||
413 | /* OK, work out the oldest transaction remaining in the log, and | 456 | /* OK, work out the oldest transaction remaining in the log, and |
414 | * the log block it starts at. | 457 | * the log block it starts at. |
415 | * | 458 | * |
diff --git a/fs/jbd2/commit.c b/fs/jbd2/commit.c index f8b3be873226..0abe02c4242a 100644 --- a/fs/jbd2/commit.c +++ b/fs/jbd2/commit.c | |||
@@ -16,6 +16,7 @@ | |||
16 | #include <linux/time.h> | 16 | #include <linux/time.h> |
17 | #include <linux/fs.h> | 17 | #include <linux/fs.h> |
18 | #include <linux/jbd2.h> | 18 | #include <linux/jbd2.h> |
19 | #include <linux/marker.h> | ||
19 | #include <linux/errno.h> | 20 | #include <linux/errno.h> |
20 | #include <linux/slab.h> | 21 | #include <linux/slab.h> |
21 | #include <linux/mm.h> | 22 | #include <linux/mm.h> |
@@ -67,7 +68,7 @@ static void release_buffer_page(struct buffer_head *bh) | |||
67 | goto nope; | 68 | goto nope; |
68 | 69 | ||
69 | /* OK, it's a truncated page */ | 70 | /* OK, it's a truncated page */ |
70 | if (TestSetPageLocked(page)) | 71 | if (!trylock_page(page)) |
71 | goto nope; | 72 | goto nope; |
72 | 73 | ||
73 | page_cache_get(page); | 74 | page_cache_get(page); |
@@ -126,8 +127,7 @@ static int journal_submit_commit_record(journal_t *journal, | |||
126 | 127 | ||
127 | JBUFFER_TRACE(descriptor, "submit commit block"); | 128 | JBUFFER_TRACE(descriptor, "submit commit block"); |
128 | lock_buffer(bh); | 129 | lock_buffer(bh); |
129 | get_bh(bh); | 130 | clear_buffer_dirty(bh); |
130 | set_buffer_dirty(bh); | ||
131 | set_buffer_uptodate(bh); | 131 | set_buffer_uptodate(bh); |
132 | bh->b_end_io = journal_end_buffer_io_sync; | 132 | bh->b_end_io = journal_end_buffer_io_sync; |
133 | 133 | ||
@@ -147,12 +147,9 @@ static int journal_submit_commit_record(journal_t *journal, | |||
147 | * to remember if we sent a barrier request | 147 | * to remember if we sent a barrier request |
148 | */ | 148 | */ |
149 | if (ret == -EOPNOTSUPP && barrier_done) { | 149 | if (ret == -EOPNOTSUPP && barrier_done) { |
150 | char b[BDEVNAME_SIZE]; | ||
151 | |||
152 | printk(KERN_WARNING | 150 | printk(KERN_WARNING |
153 | "JBD: barrier-based sync failed on %s - " | 151 | "JBD: barrier-based sync failed on %s - " |
154 | "disabling barriers\n", | 152 | "disabling barriers\n", journal->j_devname); |
155 | bdevname(journal->j_dev, b)); | ||
156 | spin_lock(&journal->j_state_lock); | 153 | spin_lock(&journal->j_state_lock); |
157 | journal->j_flags &= ~JBD2_BARRIER; | 154 | journal->j_flags &= ~JBD2_BARRIER; |
158 | spin_unlock(&journal->j_state_lock); | 155 | spin_unlock(&journal->j_state_lock); |
@@ -160,7 +157,7 @@ static int journal_submit_commit_record(journal_t *journal, | |||
160 | /* And try again, without the barrier */ | 157 | /* And try again, without the barrier */ |
161 | lock_buffer(bh); | 158 | lock_buffer(bh); |
162 | set_buffer_uptodate(bh); | 159 | set_buffer_uptodate(bh); |
163 | set_buffer_dirty(bh); | 160 | clear_buffer_dirty(bh); |
164 | ret = submit_bh(WRITE, bh); | 161 | ret = submit_bh(WRITE, bh); |
165 | } | 162 | } |
166 | *cbh = bh; | 163 | *cbh = bh; |
@@ -262,8 +259,18 @@ static int journal_finish_inode_data_buffers(journal_t *journal, | |||
262 | jinode->i_flags |= JI_COMMIT_RUNNING; | 259 | jinode->i_flags |= JI_COMMIT_RUNNING; |
263 | spin_unlock(&journal->j_list_lock); | 260 | spin_unlock(&journal->j_list_lock); |
264 | err = filemap_fdatawait(jinode->i_vfs_inode->i_mapping); | 261 | err = filemap_fdatawait(jinode->i_vfs_inode->i_mapping); |
265 | if (!ret) | 262 | if (err) { |
266 | ret = err; | 263 | /* |
264 | * Because AS_EIO is cleared by | ||
265 | * wait_on_page_writeback_range(), set it again so | ||
266 | * that user process can get -EIO from fsync(). | ||
267 | */ | ||
268 | set_bit(AS_EIO, | ||
269 | &jinode->i_vfs_inode->i_mapping->flags); | ||
270 | |||
271 | if (!ret) | ||
272 | ret = err; | ||
273 | } | ||
267 | spin_lock(&journal->j_list_lock); | 274 | spin_lock(&journal->j_list_lock); |
268 | jinode->i_flags &= ~JI_COMMIT_RUNNING; | 275 | jinode->i_flags &= ~JI_COMMIT_RUNNING; |
269 | wake_up_bit(&jinode->i_flags, __JI_COMMIT_RUNNING); | 276 | wake_up_bit(&jinode->i_flags, __JI_COMMIT_RUNNING); |
@@ -361,6 +368,8 @@ void jbd2_journal_commit_transaction(journal_t *journal) | |||
361 | commit_transaction = journal->j_running_transaction; | 368 | commit_transaction = journal->j_running_transaction; |
362 | J_ASSERT(commit_transaction->t_state == T_RUNNING); | 369 | J_ASSERT(commit_transaction->t_state == T_RUNNING); |
363 | 370 | ||
371 | trace_mark(jbd2_start_commit, "dev %s transaction %d", | ||
372 | journal->j_devname, commit_transaction->t_tid); | ||
364 | jbd_debug(1, "JBD: starting commit of transaction %d\n", | 373 | jbd_debug(1, "JBD: starting commit of transaction %d\n", |
365 | commit_transaction->t_tid); | 374 | commit_transaction->t_tid); |
366 | 375 | ||
@@ -495,9 +504,10 @@ void jbd2_journal_commit_transaction(journal_t *journal) | |||
495 | jh = commit_transaction->t_buffers; | 504 | jh = commit_transaction->t_buffers; |
496 | 505 | ||
497 | /* If we're in abort mode, we just un-journal the buffer and | 506 | /* If we're in abort mode, we just un-journal the buffer and |
498 | release it for background writing. */ | 507 | release it. */ |
499 | 508 | ||
500 | if (is_journal_aborted(journal)) { | 509 | if (is_journal_aborted(journal)) { |
510 | clear_buffer_jbddirty(jh2bh(jh)); | ||
501 | JBUFFER_TRACE(jh, "journal is aborting: refile"); | 511 | JBUFFER_TRACE(jh, "journal is aborting: refile"); |
502 | jbd2_journal_refile_buffer(journal, jh); | 512 | jbd2_journal_refile_buffer(journal, jh); |
503 | /* If that was the last one, we need to clean up | 513 | /* If that was the last one, we need to clean up |
@@ -670,8 +680,14 @@ start_journal_io: | |||
670 | * commit block, which happens below in such setting. | 680 | * commit block, which happens below in such setting. |
671 | */ | 681 | */ |
672 | err = journal_finish_inode_data_buffers(journal, commit_transaction); | 682 | err = journal_finish_inode_data_buffers(journal, commit_transaction); |
673 | if (err) | 683 | if (err) { |
674 | jbd2_journal_abort(journal, err); | 684 | printk(KERN_WARNING |
685 | "JBD2: Detected IO errors while flushing file data " | ||
686 | "on %s\n", journal->j_devname); | ||
687 | if (journal->j_flags & JBD2_ABORT_ON_SYNCDATA_ERR) | ||
688 | jbd2_journal_abort(journal, err); | ||
689 | err = 0; | ||
690 | } | ||
675 | 691 | ||
676 | /* Lo and behold: we have just managed to send a transaction to | 692 | /* Lo and behold: we have just managed to send a transaction to |
677 | the log. Before we can commit it, wait for the IO so far to | 693 | the log. Before we can commit it, wait for the IO so far to |
@@ -770,6 +786,9 @@ wait_for_iobuf: | |||
770 | /* AKPM: bforget here */ | 786 | /* AKPM: bforget here */ |
771 | } | 787 | } |
772 | 788 | ||
789 | if (err) | ||
790 | jbd2_journal_abort(journal, err); | ||
791 | |||
773 | jbd_debug(3, "JBD: commit phase 5\n"); | 792 | jbd_debug(3, "JBD: commit phase 5\n"); |
774 | 793 | ||
775 | if (!JBD2_HAS_INCOMPAT_FEATURE(journal, | 794 | if (!JBD2_HAS_INCOMPAT_FEATURE(journal, |
@@ -868,6 +887,8 @@ restart_loop: | |||
868 | if (buffer_jbddirty(bh)) { | 887 | if (buffer_jbddirty(bh)) { |
869 | JBUFFER_TRACE(jh, "add to new checkpointing trans"); | 888 | JBUFFER_TRACE(jh, "add to new checkpointing trans"); |
870 | __jbd2_journal_insert_checkpoint(jh, commit_transaction); | 889 | __jbd2_journal_insert_checkpoint(jh, commit_transaction); |
890 | if (is_journal_aborted(journal)) | ||
891 | clear_buffer_jbddirty(bh); | ||
871 | JBUFFER_TRACE(jh, "refile for checkpoint writeback"); | 892 | JBUFFER_TRACE(jh, "refile for checkpoint writeback"); |
872 | __jbd2_journal_refile_buffer(jh); | 893 | __jbd2_journal_refile_buffer(jh); |
873 | jbd_unlock_bh_state(bh); | 894 | jbd_unlock_bh_state(bh); |
@@ -974,6 +995,9 @@ restart_loop: | |||
974 | } | 995 | } |
975 | spin_unlock(&journal->j_list_lock); | 996 | spin_unlock(&journal->j_list_lock); |
976 | 997 | ||
998 | trace_mark(jbd2_end_commit, "dev %s transaction %d head %d", | ||
999 | journal->j_devname, commit_transaction->t_tid, | ||
1000 | journal->j_tail_sequence); | ||
977 | jbd_debug(1, "JBD: commit %d complete, head %d\n", | 1001 | jbd_debug(1, "JBD: commit %d complete, head %d\n", |
978 | journal->j_commit_sequence, journal->j_tail_sequence); | 1002 | journal->j_commit_sequence, journal->j_tail_sequence); |
979 | 1003 | ||
diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c index b26c6d9fe6ae..783de118de92 100644 --- a/fs/jbd2/journal.c +++ b/fs/jbd2/journal.c | |||
@@ -68,7 +68,6 @@ EXPORT_SYMBOL(jbd2_journal_set_features); | |||
68 | EXPORT_SYMBOL(jbd2_journal_create); | 68 | EXPORT_SYMBOL(jbd2_journal_create); |
69 | EXPORT_SYMBOL(jbd2_journal_load); | 69 | EXPORT_SYMBOL(jbd2_journal_load); |
70 | EXPORT_SYMBOL(jbd2_journal_destroy); | 70 | EXPORT_SYMBOL(jbd2_journal_destroy); |
71 | EXPORT_SYMBOL(jbd2_journal_update_superblock); | ||
72 | EXPORT_SYMBOL(jbd2_journal_abort); | 71 | EXPORT_SYMBOL(jbd2_journal_abort); |
73 | EXPORT_SYMBOL(jbd2_journal_errno); | 72 | EXPORT_SYMBOL(jbd2_journal_errno); |
74 | EXPORT_SYMBOL(jbd2_journal_ack_err); | 73 | EXPORT_SYMBOL(jbd2_journal_ack_err); |
@@ -598,13 +597,9 @@ int jbd2_journal_bmap(journal_t *journal, unsigned long blocknr, | |||
598 | if (ret) | 597 | if (ret) |
599 | *retp = ret; | 598 | *retp = ret; |
600 | else { | 599 | else { |
601 | char b[BDEVNAME_SIZE]; | ||
602 | |||
603 | printk(KERN_ALERT "%s: journal block not found " | 600 | printk(KERN_ALERT "%s: journal block not found " |
604 | "at offset %lu on %s\n", | 601 | "at offset %lu on %s\n", |
605 | __func__, | 602 | __func__, blocknr, journal->j_devname); |
606 | blocknr, | ||
607 | bdevname(journal->j_dev, b)); | ||
608 | err = -EIO; | 603 | err = -EIO; |
609 | __journal_abort_soft(journal, err); | 604 | __journal_abort_soft(journal, err); |
610 | } | 605 | } |
@@ -902,10 +897,7 @@ static struct proc_dir_entry *proc_jbd2_stats; | |||
902 | 897 | ||
903 | static void jbd2_stats_proc_init(journal_t *journal) | 898 | static void jbd2_stats_proc_init(journal_t *journal) |
904 | { | 899 | { |
905 | char name[BDEVNAME_SIZE]; | 900 | journal->j_proc_entry = proc_mkdir(journal->j_devname, proc_jbd2_stats); |
906 | |||
907 | bdevname(journal->j_dev, name); | ||
908 | journal->j_proc_entry = proc_mkdir(name, proc_jbd2_stats); | ||
909 | if (journal->j_proc_entry) { | 901 | if (journal->j_proc_entry) { |
910 | proc_create_data("history", S_IRUGO, journal->j_proc_entry, | 902 | proc_create_data("history", S_IRUGO, journal->j_proc_entry, |
911 | &jbd2_seq_history_fops, journal); | 903 | &jbd2_seq_history_fops, journal); |
@@ -916,12 +908,9 @@ static void jbd2_stats_proc_init(journal_t *journal) | |||
916 | 908 | ||
917 | static void jbd2_stats_proc_exit(journal_t *journal) | 909 | static void jbd2_stats_proc_exit(journal_t *journal) |
918 | { | 910 | { |
919 | char name[BDEVNAME_SIZE]; | ||
920 | |||
921 | bdevname(journal->j_dev, name); | ||
922 | remove_proc_entry("info", journal->j_proc_entry); | 911 | remove_proc_entry("info", journal->j_proc_entry); |
923 | remove_proc_entry("history", journal->j_proc_entry); | 912 | remove_proc_entry("history", journal->j_proc_entry); |
924 | remove_proc_entry(name, proc_jbd2_stats); | 913 | remove_proc_entry(journal->j_devname, proc_jbd2_stats); |
925 | } | 914 | } |
926 | 915 | ||
927 | static void journal_init_stats(journal_t *journal) | 916 | static void journal_init_stats(journal_t *journal) |
@@ -1019,6 +1008,7 @@ journal_t * jbd2_journal_init_dev(struct block_device *bdev, | |||
1019 | { | 1008 | { |
1020 | journal_t *journal = journal_init_common(); | 1009 | journal_t *journal = journal_init_common(); |
1021 | struct buffer_head *bh; | 1010 | struct buffer_head *bh; |
1011 | char *p; | ||
1022 | int n; | 1012 | int n; |
1023 | 1013 | ||
1024 | if (!journal) | 1014 | if (!journal) |
@@ -1040,6 +1030,10 @@ journal_t * jbd2_journal_init_dev(struct block_device *bdev, | |||
1040 | journal->j_fs_dev = fs_dev; | 1030 | journal->j_fs_dev = fs_dev; |
1041 | journal->j_blk_offset = start; | 1031 | journal->j_blk_offset = start; |
1042 | journal->j_maxlen = len; | 1032 | journal->j_maxlen = len; |
1033 | bdevname(journal->j_dev, journal->j_devname); | ||
1034 | p = journal->j_devname; | ||
1035 | while ((p = strchr(p, '/'))) | ||
1036 | *p = '!'; | ||
1043 | jbd2_stats_proc_init(journal); | 1037 | jbd2_stats_proc_init(journal); |
1044 | 1038 | ||
1045 | bh = __getblk(journal->j_dev, start, journal->j_blocksize); | 1039 | bh = __getblk(journal->j_dev, start, journal->j_blocksize); |
@@ -1062,6 +1056,7 @@ journal_t * jbd2_journal_init_inode (struct inode *inode) | |||
1062 | { | 1056 | { |
1063 | struct buffer_head *bh; | 1057 | struct buffer_head *bh; |
1064 | journal_t *journal = journal_init_common(); | 1058 | journal_t *journal = journal_init_common(); |
1059 | char *p; | ||
1065 | int err; | 1060 | int err; |
1066 | int n; | 1061 | int n; |
1067 | unsigned long long blocknr; | 1062 | unsigned long long blocknr; |
@@ -1071,6 +1066,12 @@ journal_t * jbd2_journal_init_inode (struct inode *inode) | |||
1071 | 1066 | ||
1072 | journal->j_dev = journal->j_fs_dev = inode->i_sb->s_bdev; | 1067 | journal->j_dev = journal->j_fs_dev = inode->i_sb->s_bdev; |
1073 | journal->j_inode = inode; | 1068 | journal->j_inode = inode; |
1069 | bdevname(journal->j_dev, journal->j_devname); | ||
1070 | p = journal->j_devname; | ||
1071 | while ((p = strchr(p, '/'))) | ||
1072 | *p = '!'; | ||
1073 | p = journal->j_devname + strlen(journal->j_devname); | ||
1074 | sprintf(p, ":%lu", journal->j_inode->i_ino); | ||
1074 | jbd_debug(1, | 1075 | jbd_debug(1, |
1075 | "journal %p: inode %s/%ld, size %Ld, bits %d, blksize %ld\n", | 1076 | "journal %p: inode %s/%ld, size %Ld, bits %d, blksize %ld\n", |
1076 | journal, inode->i_sb->s_id, inode->i_ino, | 1077 | journal, inode->i_sb->s_id, inode->i_ino, |
@@ -1254,6 +1255,22 @@ void jbd2_journal_update_superblock(journal_t *journal, int wait) | |||
1254 | goto out; | 1255 | goto out; |
1255 | } | 1256 | } |
1256 | 1257 | ||
1258 | if (buffer_write_io_error(bh)) { | ||
1259 | /* | ||
1260 | * Oh, dear. A previous attempt to write the journal | ||
1261 | * superblock failed. This could happen because the | ||
1262 | * USB device was yanked out. Or it could happen to | ||
1263 | * be a transient write error and maybe the block will | ||
1264 | * be remapped. Nothing we can do but to retry the | ||
1265 | * write and hope for the best. | ||
1266 | */ | ||
1267 | printk(KERN_ERR "JBD2: previous I/O error detected " | ||
1268 | "for journal superblock update for %s.\n", | ||
1269 | journal->j_devname); | ||
1270 | clear_buffer_write_io_error(bh); | ||
1271 | set_buffer_uptodate(bh); | ||
1272 | } | ||
1273 | |||
1257 | spin_lock(&journal->j_state_lock); | 1274 | spin_lock(&journal->j_state_lock); |
1258 | jbd_debug(1,"JBD: updating superblock (start %ld, seq %d, errno %d)\n", | 1275 | jbd_debug(1,"JBD: updating superblock (start %ld, seq %d, errno %d)\n", |
1259 | journal->j_tail, journal->j_tail_sequence, journal->j_errno); | 1276 | journal->j_tail, journal->j_tail_sequence, journal->j_errno); |
@@ -1265,9 +1282,16 @@ void jbd2_journal_update_superblock(journal_t *journal, int wait) | |||
1265 | 1282 | ||
1266 | BUFFER_TRACE(bh, "marking dirty"); | 1283 | BUFFER_TRACE(bh, "marking dirty"); |
1267 | mark_buffer_dirty(bh); | 1284 | mark_buffer_dirty(bh); |
1268 | if (wait) | 1285 | if (wait) { |
1269 | sync_dirty_buffer(bh); | 1286 | sync_dirty_buffer(bh); |
1270 | else | 1287 | if (buffer_write_io_error(bh)) { |
1288 | printk(KERN_ERR "JBD2: I/O error detected " | ||
1289 | "when updating journal superblock for %s.\n", | ||
1290 | journal->j_devname); | ||
1291 | clear_buffer_write_io_error(bh); | ||
1292 | set_buffer_uptodate(bh); | ||
1293 | } | ||
1294 | } else | ||
1271 | ll_rw_block(SWRITE, 1, &bh); | 1295 | ll_rw_block(SWRITE, 1, &bh); |
1272 | 1296 | ||
1273 | out: | 1297 | out: |
@@ -1427,9 +1451,12 @@ recovery_error: | |||
1427 | * | 1451 | * |
1428 | * Release a journal_t structure once it is no longer in use by the | 1452 | * Release a journal_t structure once it is no longer in use by the |
1429 | * journaled object. | 1453 | * journaled object. |
1454 | * Return <0 if we couldn't clean up the journal. | ||
1430 | */ | 1455 | */ |
1431 | void jbd2_journal_destroy(journal_t *journal) | 1456 | int jbd2_journal_destroy(journal_t *journal) |
1432 | { | 1457 | { |
1458 | int err = 0; | ||
1459 | |||
1433 | /* Wait for the commit thread to wake up and die. */ | 1460 | /* Wait for the commit thread to wake up and die. */ |
1434 | journal_kill_thread(journal); | 1461 | journal_kill_thread(journal); |
1435 | 1462 | ||
@@ -1452,11 +1479,16 @@ void jbd2_journal_destroy(journal_t *journal) | |||
1452 | J_ASSERT(journal->j_checkpoint_transactions == NULL); | 1479 | J_ASSERT(journal->j_checkpoint_transactions == NULL); |
1453 | spin_unlock(&journal->j_list_lock); | 1480 | spin_unlock(&journal->j_list_lock); |
1454 | 1481 | ||
1455 | /* We can now mark the journal as empty. */ | ||
1456 | journal->j_tail = 0; | ||
1457 | journal->j_tail_sequence = ++journal->j_transaction_sequence; | ||
1458 | if (journal->j_sb_buffer) { | 1482 | if (journal->j_sb_buffer) { |
1459 | jbd2_journal_update_superblock(journal, 1); | 1483 | if (!is_journal_aborted(journal)) { |
1484 | /* We can now mark the journal as empty. */ | ||
1485 | journal->j_tail = 0; | ||
1486 | journal->j_tail_sequence = | ||
1487 | ++journal->j_transaction_sequence; | ||
1488 | jbd2_journal_update_superblock(journal, 1); | ||
1489 | } else { | ||
1490 | err = -EIO; | ||
1491 | } | ||
1460 | brelse(journal->j_sb_buffer); | 1492 | brelse(journal->j_sb_buffer); |
1461 | } | 1493 | } |
1462 | 1494 | ||
@@ -1468,6 +1500,8 @@ void jbd2_journal_destroy(journal_t *journal) | |||
1468 | jbd2_journal_destroy_revoke(journal); | 1500 | jbd2_journal_destroy_revoke(journal); |
1469 | kfree(journal->j_wbuf); | 1501 | kfree(journal->j_wbuf); |
1470 | kfree(journal); | 1502 | kfree(journal); |
1503 | |||
1504 | return err; | ||
1471 | } | 1505 | } |
1472 | 1506 | ||
1473 | 1507 | ||
@@ -1693,10 +1727,16 @@ int jbd2_journal_flush(journal_t *journal) | |||
1693 | spin_lock(&journal->j_list_lock); | 1727 | spin_lock(&journal->j_list_lock); |
1694 | while (!err && journal->j_checkpoint_transactions != NULL) { | 1728 | while (!err && journal->j_checkpoint_transactions != NULL) { |
1695 | spin_unlock(&journal->j_list_lock); | 1729 | spin_unlock(&journal->j_list_lock); |
1730 | mutex_lock(&journal->j_checkpoint_mutex); | ||
1696 | err = jbd2_log_do_checkpoint(journal); | 1731 | err = jbd2_log_do_checkpoint(journal); |
1732 | mutex_unlock(&journal->j_checkpoint_mutex); | ||
1697 | spin_lock(&journal->j_list_lock); | 1733 | spin_lock(&journal->j_list_lock); |
1698 | } | 1734 | } |
1699 | spin_unlock(&journal->j_list_lock); | 1735 | spin_unlock(&journal->j_list_lock); |
1736 | |||
1737 | if (is_journal_aborted(journal)) | ||
1738 | return -EIO; | ||
1739 | |||
1700 | jbd2_cleanup_journal_tail(journal); | 1740 | jbd2_cleanup_journal_tail(journal); |
1701 | 1741 | ||
1702 | /* Finally, mark the journal as really needing no recovery. | 1742 | /* Finally, mark the journal as really needing no recovery. |
@@ -1718,7 +1758,7 @@ int jbd2_journal_flush(journal_t *journal) | |||
1718 | J_ASSERT(journal->j_head == journal->j_tail); | 1758 | J_ASSERT(journal->j_head == journal->j_tail); |
1719 | J_ASSERT(journal->j_tail_sequence == journal->j_transaction_sequence); | 1759 | J_ASSERT(journal->j_tail_sequence == journal->j_transaction_sequence); |
1720 | spin_unlock(&journal->j_state_lock); | 1760 | spin_unlock(&journal->j_state_lock); |
1721 | return err; | 1761 | return 0; |
1722 | } | 1762 | } |
1723 | 1763 | ||
1724 | /** | 1764 | /** |
@@ -1762,23 +1802,6 @@ int jbd2_journal_wipe(journal_t *journal, int write) | |||
1762 | } | 1802 | } |
1763 | 1803 | ||
1764 | /* | 1804 | /* |
1765 | * journal_dev_name: format a character string to describe on what | ||
1766 | * device this journal is present. | ||
1767 | */ | ||
1768 | |||
1769 | static const char *journal_dev_name(journal_t *journal, char *buffer) | ||
1770 | { | ||
1771 | struct block_device *bdev; | ||
1772 | |||
1773 | if (journal->j_inode) | ||
1774 | bdev = journal->j_inode->i_sb->s_bdev; | ||
1775 | else | ||
1776 | bdev = journal->j_dev; | ||
1777 | |||
1778 | return bdevname(bdev, buffer); | ||
1779 | } | ||
1780 | |||
1781 | /* | ||
1782 | * Journal abort has very specific semantics, which we describe | 1805 | * Journal abort has very specific semantics, which we describe |
1783 | * for journal abort. | 1806 | * for journal abort. |
1784 | * | 1807 | * |
@@ -1794,13 +1817,12 @@ static const char *journal_dev_name(journal_t *journal, char *buffer) | |||
1794 | void __jbd2_journal_abort_hard(journal_t *journal) | 1817 | void __jbd2_journal_abort_hard(journal_t *journal) |
1795 | { | 1818 | { |
1796 | transaction_t *transaction; | 1819 | transaction_t *transaction; |
1797 | char b[BDEVNAME_SIZE]; | ||
1798 | 1820 | ||
1799 | if (journal->j_flags & JBD2_ABORT) | 1821 | if (journal->j_flags & JBD2_ABORT) |
1800 | return; | 1822 | return; |
1801 | 1823 | ||
1802 | printk(KERN_ERR "Aborting journal on device %s.\n", | 1824 | printk(KERN_ERR "Aborting journal on device %s.\n", |
1803 | journal_dev_name(journal, b)); | 1825 | journal->j_devname); |
1804 | 1826 | ||
1805 | spin_lock(&journal->j_state_lock); | 1827 | spin_lock(&journal->j_state_lock); |
1806 | journal->j_flags |= JBD2_ABORT; | 1828 | journal->j_flags |= JBD2_ABORT; |
diff --git a/fs/jbd2/recovery.c b/fs/jbd2/recovery.c index 058f50f65b76..73063285b13f 100644 --- a/fs/jbd2/recovery.c +++ b/fs/jbd2/recovery.c | |||
@@ -225,7 +225,7 @@ do { \ | |||
225 | */ | 225 | */ |
226 | int jbd2_journal_recover(journal_t *journal) | 226 | int jbd2_journal_recover(journal_t *journal) |
227 | { | 227 | { |
228 | int err; | 228 | int err, err2; |
229 | journal_superblock_t * sb; | 229 | journal_superblock_t * sb; |
230 | 230 | ||
231 | struct recovery_info info; | 231 | struct recovery_info info; |
@@ -263,7 +263,10 @@ int jbd2_journal_recover(journal_t *journal) | |||
263 | journal->j_transaction_sequence = ++info.end_transaction; | 263 | journal->j_transaction_sequence = ++info.end_transaction; |
264 | 264 | ||
265 | jbd2_journal_clear_revoke(journal); | 265 | jbd2_journal_clear_revoke(journal); |
266 | sync_blockdev(journal->j_fs_dev); | 266 | err2 = sync_blockdev(journal->j_fs_dev); |
267 | if (!err) | ||
268 | err = err2; | ||
269 | |||
267 | return err; | 270 | return err; |
268 | } | 271 | } |
269 | 272 | ||
diff --git a/fs/jbd2/transaction.c b/fs/jbd2/transaction.c index 4f7cadbb19fa..e5d540588fa9 100644 --- a/fs/jbd2/transaction.c +++ b/fs/jbd2/transaction.c | |||
@@ -301,7 +301,7 @@ handle_t *jbd2_journal_start(journal_t *journal, int nblocks) | |||
301 | goto out; | 301 | goto out; |
302 | } | 302 | } |
303 | 303 | ||
304 | lock_acquire(&handle->h_lockdep_map, 0, 0, 0, 2, _THIS_IP_); | 304 | lock_map_acquire(&handle->h_lockdep_map); |
305 | out: | 305 | out: |
306 | return handle; | 306 | return handle; |
307 | } | 307 | } |
@@ -1279,7 +1279,7 @@ int jbd2_journal_stop(handle_t *handle) | |||
1279 | spin_unlock(&journal->j_state_lock); | 1279 | spin_unlock(&journal->j_state_lock); |
1280 | } | 1280 | } |
1281 | 1281 | ||
1282 | lock_release(&handle->h_lockdep_map, 1, _THIS_IP_); | 1282 | lock_map_release(&handle->h_lockdep_map); |
1283 | 1283 | ||
1284 | jbd2_free_handle(handle); | 1284 | jbd2_free_handle(handle); |
1285 | return err; | 1285 | return err; |