Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/include/linux/jbd.h | |
ae6ddcc5 | 3 | * |
1da177e4 LT |
4 | * Written by Stephen C. Tweedie <sct@redhat.com> |
5 | * | |
6 | * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved | |
7 | * | |
8 | * This file is part of the Linux kernel and is made available under | |
9 | * the terms of the GNU General Public License, version 2, or at your | |
10 | * option, any later version, incorporated herein by reference. | |
11 | * | |
12 | * Definitions for transaction data structures for the buffer cache | |
13 | * filesystem journaling support. | |
14 | */ | |
15 | ||
16 | #ifndef _LINUX_JBD_H | |
17 | #define _LINUX_JBD_H | |
18 | ||
1da177e4 LT |
19 | /* Allow this file to be included directly into e2fsprogs */ |
20 | #ifndef __KERNEL__ | |
21 | #include "jfs_compat.h" | |
22 | #define JFS_DEBUG | |
23 | #define jfs_debug jbd_debug | |
24 | #else | |
25 | ||
fe1dcbc4 | 26 | #include <linux/types.h> |
1da177e4 LT |
27 | #include <linux/buffer_head.h> |
28 | #include <linux/journal-head.h> | |
29 | #include <linux/stddef.h> | |
fb1c8f93 | 30 | #include <linux/bit_spinlock.h> |
2c68ee75 | 31 | #include <linux/mutex.h> |
e3df1898 | 32 | #include <linux/timer.h> |
34a3d1e8 | 33 | #include <linux/lockdep.h> |
5a0e3ad6 | 34 | #include <linux/slab.h> |
e3df1898 | 35 | |
1da177e4 LT |
36 | #define journal_oom_retry 1 |
37 | ||
38 | /* | |
0cbfdc86 | 39 | * Define JBD_PARANOID_IOFAIL to cause a kernel BUG() if ext3 finds |
1da177e4 LT |
40 | * certain classes of error which can occur due to failed IOs. Under |
41 | * normal use we want ext3 to continue after such errors, because | |
42 | * hardware _can_ fail, but for debugging purposes when running tests on | |
43 | * known-good hardware we may want to trap these errors. | |
44 | */ | |
45 | #undef JBD_PARANOID_IOFAIL | |
46 | ||
47 | /* | |
48 | * The default maximum commit age, in seconds. | |
49 | */ | |
50 | #define JBD_DEFAULT_MAX_COMMIT_AGE 5 | |
51 | ||
52 | #ifdef CONFIG_JBD_DEBUG | |
53 | /* | |
54 | * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal | |
55 | * consistency checks. By default we don't do this unless | |
56 | * CONFIG_JBD_DEBUG is on. | |
57 | */ | |
58 | #define JBD_EXPENSIVE_CHECKING | |
c2a9159c | 59 | extern u8 journal_enable_debug; |
1da177e4 LT |
60 | |
61 | #define jbd_debug(n, f, a...) \ | |
62 | do { \ | |
63 | if ((n) <= journal_enable_debug) { \ | |
64 | printk (KERN_DEBUG "(%s, %d): %s: ", \ | |
d5c003b4 | 65 | __FILE__, __LINE__, __func__); \ |
e9ad5620 | 66 | printk (f, ## a); \ |
1da177e4 LT |
67 | } \ |
68 | } while (0) | |
69 | #else | |
70 | #define jbd_debug(f, a...) /**/ | |
71 | #endif | |
72 | ||
c089d490 MC |
73 | static inline void *jbd_alloc(size_t size, gfp_t flags) |
74 | { | |
75 | return (void *)__get_free_pages(flags, get_order(size)); | |
76 | } | |
77 | ||
78 | static inline void jbd_free(void *ptr, size_t size) | |
79 | { | |
80 | free_pages((unsigned long)ptr, get_order(size)); | |
81 | }; | |
82 | ||
1da177e4 LT |
83 | #define JFS_MIN_JOURNAL_BLOCKS 1024 |
84 | ||
1da177e4 LT |
85 | |
86 | /** | |
87 | * typedef handle_t - The handle_t type represents a single atomic update being performed by some process. | |
88 | * | |
89 | * All filesystem modifications made by the process go | |
90 | * through this handle. Recursive operations (such as quota operations) | |
91 | * are gathered into a single update. | |
92 | * | |
93 | * The buffer credits field is used to account for journaled buffers | |
94 | * being modified by the running process. To ensure that there is | |
95 | * enough log space for all outstanding operations, we need to limit the | |
96 | * number of outstanding buffers possible at any time. When the | |
97 | * operation completes, any buffer credits not used are credited back to | |
98 | * the transaction, so that at all times we know how many buffers the | |
ae6ddcc5 MC |
99 | * outstanding updates on a transaction might possibly touch. |
100 | * | |
1da177e4 LT |
101 | * This is an opaque datatype. |
102 | **/ | |
103 | typedef struct handle_s handle_t; /* Atomic operation type */ | |
104 | ||
105 | ||
106 | /** | |
107 | * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem. | |
108 | * | |
109 | * journal_t is linked to from the fs superblock structure. | |
ae6ddcc5 | 110 | * |
1da177e4 LT |
111 | * We use the journal_t to keep track of all outstanding transaction |
112 | * activity on the filesystem, and to manage the state of the log | |
113 | * writing process. | |
114 | * | |
115 | * This is an opaque datatype. | |
116 | **/ | |
117 | typedef struct journal_s journal_t; /* Journal control structure */ | |
118 | #endif | |
119 | ||
120 | /* | |
121 | * Internal structures used by the logging mechanism: | |
122 | */ | |
123 | ||
124 | #define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */ | |
125 | ||
126 | /* | |
127 | * On-disk structures | |
128 | */ | |
129 | ||
ae6ddcc5 | 130 | /* |
1da177e4 LT |
131 | * Descriptor block types: |
132 | */ | |
133 | ||
134 | #define JFS_DESCRIPTOR_BLOCK 1 | |
135 | #define JFS_COMMIT_BLOCK 2 | |
136 | #define JFS_SUPERBLOCK_V1 3 | |
137 | #define JFS_SUPERBLOCK_V2 4 | |
138 | #define JFS_REVOKE_BLOCK 5 | |
139 | ||
140 | /* | |
141 | * Standard header for all descriptor blocks: | |
142 | */ | |
143 | typedef struct journal_header_s | |
144 | { | |
145 | __be32 h_magic; | |
146 | __be32 h_blocktype; | |
147 | __be32 h_sequence; | |
148 | } journal_header_t; | |
149 | ||
150 | ||
ae6ddcc5 MC |
151 | /* |
152 | * The block tag: used to describe a single buffer in the journal | |
1da177e4 LT |
153 | */ |
154 | typedef struct journal_block_tag_s | |
155 | { | |
156 | __be32 t_blocknr; /* The on-disk block number */ | |
157 | __be32 t_flags; /* See below */ | |
158 | } journal_block_tag_t; | |
159 | ||
ae6ddcc5 | 160 | /* |
1da177e4 | 161 | * The revoke descriptor: used on disk to describe a series of blocks to |
ae6ddcc5 | 162 | * be revoked from the log |
1da177e4 LT |
163 | */ |
164 | typedef struct journal_revoke_header_s | |
165 | { | |
166 | journal_header_t r_header; | |
167 | __be32 r_count; /* Count of bytes used in the block */ | |
168 | } journal_revoke_header_t; | |
169 | ||
170 | ||
171 | /* Definitions for the journal tag flags word: */ | |
172 | #define JFS_FLAG_ESCAPE 1 /* on-disk block is escaped */ | |
173 | #define JFS_FLAG_SAME_UUID 2 /* block has same uuid as previous */ | |
174 | #define JFS_FLAG_DELETED 4 /* block deleted by this transaction */ | |
175 | #define JFS_FLAG_LAST_TAG 8 /* last tag in this descriptor block */ | |
176 | ||
177 | ||
178 | /* | |
179 | * The journal superblock. All fields are in big-endian byte order. | |
180 | */ | |
181 | typedef struct journal_superblock_s | |
182 | { | |
183 | /* 0x0000 */ | |
184 | journal_header_t s_header; | |
185 | ||
186 | /* 0x000C */ | |
187 | /* Static information describing the journal */ | |
188 | __be32 s_blocksize; /* journal device blocksize */ | |
189 | __be32 s_maxlen; /* total blocks in journal file */ | |
190 | __be32 s_first; /* first block of log information */ | |
191 | ||
192 | /* 0x0018 */ | |
193 | /* Dynamic information describing the current state of the log */ | |
194 | __be32 s_sequence; /* first commit ID expected in log */ | |
195 | __be32 s_start; /* blocknr of start of log */ | |
196 | ||
197 | /* 0x0020 */ | |
198 | /* Error value, as set by journal_abort(). */ | |
199 | __be32 s_errno; | |
200 | ||
201 | /* 0x0024 */ | |
202 | /* Remaining fields are only valid in a version-2 superblock */ | |
e9ad5620 DK |
203 | __be32 s_feature_compat; /* compatible feature set */ |
204 | __be32 s_feature_incompat; /* incompatible feature set */ | |
205 | __be32 s_feature_ro_compat; /* readonly-compatible feature set */ | |
1da177e4 LT |
206 | /* 0x0030 */ |
207 | __u8 s_uuid[16]; /* 128-bit uuid for journal */ | |
208 | ||
209 | /* 0x0040 */ | |
210 | __be32 s_nr_users; /* Nr of filesystems sharing log */ | |
211 | ||
212 | __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/ | |
213 | ||
214 | /* 0x0048 */ | |
215 | __be32 s_max_transaction; /* Limit of journal blocks per trans.*/ | |
216 | __be32 s_max_trans_data; /* Limit of data blocks per trans. */ | |
217 | ||
218 | /* 0x0050 */ | |
219 | __u32 s_padding[44]; | |
220 | ||
221 | /* 0x0100 */ | |
222 | __u8 s_users[16*48]; /* ids of all fs'es sharing the log */ | |
223 | /* 0x0400 */ | |
224 | } journal_superblock_t; | |
225 | ||
226 | #define JFS_HAS_COMPAT_FEATURE(j,mask) \ | |
227 | ((j)->j_format_version >= 2 && \ | |
228 | ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask)))) | |
229 | #define JFS_HAS_RO_COMPAT_FEATURE(j,mask) \ | |
230 | ((j)->j_format_version >= 2 && \ | |
231 | ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask)))) | |
232 | #define JFS_HAS_INCOMPAT_FEATURE(j,mask) \ | |
233 | ((j)->j_format_version >= 2 && \ | |
234 | ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask)))) | |
235 | ||
236 | #define JFS_FEATURE_INCOMPAT_REVOKE 0x00000001 | |
237 | ||
238 | /* Features known to this kernel version: */ | |
239 | #define JFS_KNOWN_COMPAT_FEATURES 0 | |
240 | #define JFS_KNOWN_ROCOMPAT_FEATURES 0 | |
241 | #define JFS_KNOWN_INCOMPAT_FEATURES JFS_FEATURE_INCOMPAT_REVOKE | |
242 | ||
243 | #ifdef __KERNEL__ | |
244 | ||
245 | #include <linux/fs.h> | |
246 | #include <linux/sched.h> | |
44606672 | 247 | #include <linux/jbd_common.h> |
1da177e4 | 248 | |
1c099244 | 249 | #define J_ASSERT(assert) BUG_ON(!(assert)) |
1da177e4 | 250 | |
1da177e4 LT |
251 | #define J_ASSERT_BH(bh, expr) J_ASSERT(expr) |
252 | #define J_ASSERT_JH(jh, expr) J_ASSERT(expr) | |
1da177e4 | 253 | |
1da177e4 LT |
254 | #if defined(JBD_PARANOID_IOFAIL) |
255 | #define J_EXPECT(expr, why...) J_ASSERT(expr) | |
256 | #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr) | |
257 | #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr) | |
258 | #else | |
259 | #define __journal_expect(expr, why...) \ | |
260 | ({ \ | |
261 | int val = (expr); \ | |
262 | if (!val) { \ | |
263 | printk(KERN_ERR \ | |
264 | "EXT3-fs unexpected failure: %s;\n",# expr); \ | |
265 | printk(KERN_ERR why "\n"); \ | |
266 | } \ | |
267 | val; \ | |
268 | }) | |
269 | #define J_EXPECT(expr, why...) __journal_expect(expr, ## why) | |
270 | #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why) | |
271 | #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why) | |
272 | #endif | |
273 | ||
1da177e4 LT |
274 | struct jbd_revoke_table_s; |
275 | ||
276 | /** | |
a6b91919 | 277 | * struct handle_s - this is the concrete type associated with handle_t. |
1da177e4 LT |
278 | * @h_transaction: Which compound transaction is this update a part of? |
279 | * @h_buffer_credits: Number of remaining buffers we are allowed to dirty. | |
280 | * @h_ref: Reference count on this handle | |
281 | * @h_err: Field for caller's use to track errors through large fs operations | |
282 | * @h_sync: flag for sync-on-close | |
283 | * @h_jdata: flag to force data journaling | |
284 | * @h_aborted: flag indicating fatal error on handle | |
e6716b87 | 285 | * @h_lockdep_map: lockdep info for debugging lock problems |
1da177e4 | 286 | */ |
ae6ddcc5 | 287 | struct handle_s |
1da177e4 LT |
288 | { |
289 | /* Which compound transaction is this update a part of? */ | |
290 | transaction_t *h_transaction; | |
291 | ||
292 | /* Number of remaining buffers we are allowed to dirty: */ | |
293 | int h_buffer_credits; | |
294 | ||
295 | /* Reference count on this handle */ | |
296 | int h_ref; | |
297 | ||
298 | /* Field for caller's use to track errors through large fs */ | |
299 | /* operations */ | |
300 | int h_err; | |
301 | ||
302 | /* Flags [no locking] */ | |
303 | unsigned int h_sync: 1; /* sync-on-close */ | |
304 | unsigned int h_jdata: 1; /* force data journaling */ | |
305 | unsigned int h_aborted: 1; /* fatal error on handle */ | |
34a3d1e8 PZ |
306 | |
307 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
308 | struct lockdep_map h_lockdep_map; | |
309 | #endif | |
1da177e4 LT |
310 | }; |
311 | ||
312 | ||
313 | /* The transaction_t type is the guts of the journaling mechanism. It | |
314 | * tracks a compound transaction through its various states: | |
315 | * | |
316 | * RUNNING: accepting new updates | |
317 | * LOCKED: Updates still running but we don't accept new ones | |
318 | * RUNDOWN: Updates are tidying up but have finished requesting | |
319 | * new buffers to modify (state not used for now) | |
320 | * FLUSH: All updates complete, but we are still writing to disk | |
321 | * COMMIT: All data on disk, writing commit record | |
322 | * FINISHED: We still have to keep the transaction for checkpointing. | |
323 | * | |
324 | * The transaction keeps track of all of the buffers modified by a | |
325 | * running transaction, and all of the buffers committed but not yet | |
326 | * flushed to home for finished transactions. | |
327 | */ | |
328 | ||
329 | /* | |
330 | * Lock ranking: | |
331 | * | |
332 | * j_list_lock | |
333 | * ->jbd_lock_bh_journal_head() (This is "innermost") | |
334 | * | |
335 | * j_state_lock | |
336 | * ->jbd_lock_bh_state() | |
337 | * | |
338 | * jbd_lock_bh_state() | |
339 | * ->j_list_lock | |
340 | * | |
341 | * j_state_lock | |
342 | * ->t_handle_lock | |
343 | * | |
344 | * j_state_lock | |
345 | * ->j_list_lock (journal_unmap_buffer) | |
346 | * | |
347 | */ | |
348 | ||
ae6ddcc5 | 349 | struct transaction_s |
1da177e4 LT |
350 | { |
351 | /* Pointer to the journal for this transaction. [no locking] */ | |
352 | journal_t *t_journal; | |
353 | ||
354 | /* Sequence number for this transaction [no locking] */ | |
355 | tid_t t_tid; | |
356 | ||
357 | /* | |
358 | * Transaction's current state | |
359 | * [no locking - only kjournald alters this] | |
d4beaf4a JK |
360 | * [j_list_lock] guards transition of a transaction into T_FINISHED |
361 | * state and subsequent call of __journal_drop_transaction() | |
1da177e4 LT |
362 | * FIXME: needs barriers |
363 | * KLUDGE: [use j_state_lock] | |
364 | */ | |
365 | enum { | |
366 | T_RUNNING, | |
367 | T_LOCKED, | |
1da177e4 LT |
368 | T_FLUSH, |
369 | T_COMMIT, | |
03f4d804 | 370 | T_COMMIT_RECORD, |
ae6ddcc5 | 371 | T_FINISHED |
1da177e4 LT |
372 | } t_state; |
373 | ||
374 | /* | |
375 | * Where in the log does this transaction's commit start? [no locking] | |
376 | */ | |
9c28cbcc | 377 | unsigned int t_log_start; |
1da177e4 LT |
378 | |
379 | /* Number of buffers on the t_buffers list [j_list_lock] */ | |
380 | int t_nr_buffers; | |
381 | ||
382 | /* | |
383 | * Doubly-linked circular list of all buffers reserved but not yet | |
384 | * modified by this transaction [j_list_lock] | |
385 | */ | |
386 | struct journal_head *t_reserved_list; | |
387 | ||
388 | /* | |
389 | * Doubly-linked circular list of all buffers under writeout during | |
390 | * commit [j_list_lock] | |
391 | */ | |
392 | struct journal_head *t_locked_list; | |
393 | ||
394 | /* | |
395 | * Doubly-linked circular list of all metadata buffers owned by this | |
396 | * transaction [j_list_lock] | |
397 | */ | |
398 | struct journal_head *t_buffers; | |
399 | ||
400 | /* | |
401 | * Doubly-linked circular list of all data buffers still to be | |
402 | * flushed before this transaction can be committed [j_list_lock] | |
403 | */ | |
404 | struct journal_head *t_sync_datalist; | |
405 | ||
406 | /* | |
407 | * Doubly-linked circular list of all forget buffers (superseded | |
408 | * buffers which we can un-checkpoint once this transaction commits) | |
409 | * [j_list_lock] | |
410 | */ | |
411 | struct journal_head *t_forget; | |
412 | ||
413 | /* | |
414 | * Doubly-linked circular list of all buffers still to be flushed before | |
415 | * this transaction can be checkpointed. [j_list_lock] | |
416 | */ | |
417 | struct journal_head *t_checkpoint_list; | |
418 | ||
78ce89c9 JK |
419 | /* |
420 | * Doubly-linked circular list of all buffers submitted for IO while | |
421 | * checkpointing. [j_list_lock] | |
422 | */ | |
423 | struct journal_head *t_checkpoint_io_list; | |
424 | ||
1da177e4 LT |
425 | /* |
426 | * Doubly-linked circular list of temporary buffers currently undergoing | |
427 | * IO in the log [j_list_lock] | |
428 | */ | |
429 | struct journal_head *t_iobuf_list; | |
430 | ||
431 | /* | |
432 | * Doubly-linked circular list of metadata buffers being shadowed by log | |
433 | * IO. The IO buffers on the iobuf list and the shadow buffers on this | |
434 | * list match each other one for one at all times. [j_list_lock] | |
435 | */ | |
436 | struct journal_head *t_shadow_list; | |
437 | ||
438 | /* | |
439 | * Doubly-linked circular list of control buffers being written to the | |
440 | * log. [j_list_lock] | |
441 | */ | |
442 | struct journal_head *t_log_list; | |
443 | ||
444 | /* | |
445 | * Protects info related to handles | |
446 | */ | |
447 | spinlock_t t_handle_lock; | |
448 | ||
449 | /* | |
450 | * Number of outstanding updates running on this transaction | |
451 | * [t_handle_lock] | |
452 | */ | |
453 | int t_updates; | |
454 | ||
455 | /* | |
456 | * Number of buffers reserved for use by all handles in this transaction | |
457 | * handle but not yet modified. [t_handle_lock] | |
458 | */ | |
459 | int t_outstanding_credits; | |
460 | ||
461 | /* | |
462 | * Forward and backward links for the circular list of all transactions | |
463 | * awaiting checkpoint. [j_list_lock] | |
464 | */ | |
465 | transaction_t *t_cpnext, *t_cpprev; | |
466 | ||
467 | /* | |
468 | * When will the transaction expire (become due for commit), in jiffies? | |
469 | * [no locking] | |
470 | */ | |
471 | unsigned long t_expires; | |
472 | ||
f420d4dc JB |
473 | /* |
474 | * When this transaction started, in nanoseconds [no locking] | |
475 | */ | |
476 | ktime_t t_start_time; | |
477 | ||
1da177e4 LT |
478 | /* |
479 | * How many handles used this transaction? [t_handle_lock] | |
480 | */ | |
481 | int t_handle_count; | |
1da177e4 LT |
482 | }; |
483 | ||
484 | /** | |
a6b91919 | 485 | * struct journal_s - this is the concrete type associated with journal_t. |
1da177e4 LT |
486 | * @j_flags: General journaling state flags |
487 | * @j_errno: Is there an outstanding uncleared error on the journal (from a | |
ae6ddcc5 | 488 | * prior abort)? |
1da177e4 LT |
489 | * @j_sb_buffer: First part of superblock buffer |
490 | * @j_superblock: Second part of superblock buffer | |
491 | * @j_format_version: Version of the superblock format | |
492 | * @j_state_lock: Protect the various scalars in the journal | |
493 | * @j_barrier_count: Number of processes waiting to create a barrier lock | |
1da177e4 LT |
494 | * @j_running_transaction: The current running transaction.. |
495 | * @j_committing_transaction: the transaction we are pushing to disk | |
496 | * @j_checkpoint_transactions: a linked circular list of all transactions | |
497 | * waiting for checkpointing | |
498 | * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction | |
499 | * to start committing, or for a barrier lock to be released | |
500 | * @j_wait_logspace: Wait queue for waiting for checkpointing to complete | |
ae6ddcc5 | 501 | * @j_wait_done_commit: Wait queue for waiting for commit to complete |
1da177e4 LT |
502 | * @j_wait_checkpoint: Wait queue to trigger checkpointing |
503 | * @j_wait_commit: Wait queue to trigger commit | |
504 | * @j_wait_updates: Wait queue to wait for updates to complete | |
2c68ee75 | 505 | * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints |
1da177e4 LT |
506 | * @j_head: Journal head - identifies the first unused block in the journal |
507 | * @j_tail: Journal tail - identifies the oldest still-used block in the | |
508 | * journal. | |
509 | * @j_free: Journal free - how many free blocks are there in the journal? | |
ae6ddcc5 | 510 | * @j_first: The block number of the first usable block |
1da177e4 LT |
511 | * @j_last: The block number one beyond the last usable block |
512 | * @j_dev: Device where we store the journal | |
513 | * @j_blocksize: blocksize for the location where we store the journal. | |
514 | * @j_blk_offset: starting block offset for into the device where we store the | |
515 | * journal | |
516 | * @j_fs_dev: Device which holds the client fs. For internal journal this will | |
517 | * be equal to j_dev | |
518 | * @j_maxlen: Total maximum capacity of the journal region on disk. | |
519 | * @j_list_lock: Protects the buffer lists and internal buffer state. | |
520 | * @j_inode: Optional inode where we store the journal. If present, all journal | |
521 | * block numbers are mapped into this inode via bmap(). | |
ae6ddcc5 | 522 | * @j_tail_sequence: Sequence number of the oldest transaction in the log |
1da177e4 LT |
523 | * @j_transaction_sequence: Sequence number of the next transaction to grant |
524 | * @j_commit_sequence: Sequence number of the most recently committed | |
525 | * transaction | |
526 | * @j_commit_request: Sequence number of the most recent transaction wanting | |
ae6ddcc5 | 527 | * commit |
2db938be JK |
528 | * @j_commit_waited: Sequence number of the most recent transaction someone |
529 | * is waiting for to commit. | |
1da177e4 LT |
530 | * @j_uuid: Uuid of client object. |
531 | * @j_task: Pointer to the current commit thread for this journal | |
532 | * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a | |
533 | * single compound commit transaction | |
534 | * @j_commit_interval: What is the maximum transaction lifetime before we begin | |
535 | * a commit? | |
536 | * @j_commit_timer: The timer used to wakeup the commit thread | |
537 | * @j_revoke_lock: Protect the revoke table | |
538 | * @j_revoke: The revoke table - maintains the list of revoked blocks in the | |
539 | * current transaction. | |
540 | * @j_revoke_table: alternate revoke tables for j_revoke | |
6c8bec6d RD |
541 | * @j_wbuf: array of buffer_heads for journal_commit_transaction |
542 | * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the | |
543 | * number that will fit in j_blocksize | |
fe1dcbc4 | 544 | * @j_last_sync_writer: most recent pid which did a synchronous write |
3eabdb76 RD |
545 | * @j_average_commit_time: the average amount of time in nanoseconds it |
546 | * takes to commit a transaction to the disk. | |
1da177e4 LT |
547 | * @j_private: An opaque pointer to fs-private information. |
548 | */ | |
549 | ||
550 | struct journal_s | |
551 | { | |
552 | /* General journaling state flags [j_state_lock] */ | |
553 | unsigned long j_flags; | |
554 | ||
555 | /* | |
556 | * Is there an outstanding uncleared error on the journal (from a prior | |
557 | * abort)? [j_state_lock] | |
558 | */ | |
559 | int j_errno; | |
560 | ||
561 | /* The superblock buffer */ | |
562 | struct buffer_head *j_sb_buffer; | |
563 | journal_superblock_t *j_superblock; | |
564 | ||
565 | /* Version of the superblock format */ | |
566 | int j_format_version; | |
567 | ||
568 | /* | |
569 | * Protect the various scalars in the journal | |
570 | */ | |
571 | spinlock_t j_state_lock; | |
572 | ||
573 | /* | |
574 | * Number of processes waiting to create a barrier lock [j_state_lock] | |
575 | */ | |
576 | int j_barrier_count; | |
577 | ||
1da177e4 LT |
578 | /* |
579 | * Transactions: The current running transaction... | |
580 | * [j_state_lock] [caller holding open handle] | |
581 | */ | |
582 | transaction_t *j_running_transaction; | |
583 | ||
584 | /* | |
585 | * the transaction we are pushing to disk | |
586 | * [j_state_lock] [caller holding open handle] | |
587 | */ | |
588 | transaction_t *j_committing_transaction; | |
589 | ||
590 | /* | |
591 | * ... and a linked circular list of all transactions waiting for | |
592 | * checkpointing. [j_list_lock] | |
593 | */ | |
594 | transaction_t *j_checkpoint_transactions; | |
595 | ||
596 | /* | |
597 | * Wait queue for waiting for a locked transaction to start committing, | |
598 | * or for a barrier lock to be released | |
599 | */ | |
600 | wait_queue_head_t j_wait_transaction_locked; | |
601 | ||
602 | /* Wait queue for waiting for checkpointing to complete */ | |
603 | wait_queue_head_t j_wait_logspace; | |
604 | ||
605 | /* Wait queue for waiting for commit to complete */ | |
606 | wait_queue_head_t j_wait_done_commit; | |
607 | ||
608 | /* Wait queue to trigger checkpointing */ | |
609 | wait_queue_head_t j_wait_checkpoint; | |
610 | ||
611 | /* Wait queue to trigger commit */ | |
612 | wait_queue_head_t j_wait_commit; | |
613 | ||
614 | /* Wait queue to wait for updates to complete */ | |
615 | wait_queue_head_t j_wait_updates; | |
616 | ||
617 | /* Semaphore for locking against concurrent checkpoints */ | |
e9ad5620 | 618 | struct mutex j_checkpoint_mutex; |
1da177e4 LT |
619 | |
620 | /* | |
621 | * Journal head: identifies the first unused block in the journal. | |
622 | * [j_state_lock] | |
623 | */ | |
9c28cbcc | 624 | unsigned int j_head; |
1da177e4 LT |
625 | |
626 | /* | |
627 | * Journal tail: identifies the oldest still-used block in the journal. | |
628 | * [j_state_lock] | |
629 | */ | |
9c28cbcc | 630 | unsigned int j_tail; |
1da177e4 LT |
631 | |
632 | /* | |
633 | * Journal free: how many free blocks are there in the journal? | |
634 | * [j_state_lock] | |
635 | */ | |
9c28cbcc | 636 | unsigned int j_free; |
1da177e4 LT |
637 | |
638 | /* | |
639 | * Journal start and end: the block numbers of the first usable block | |
640 | * and one beyond the last usable block in the journal. [j_state_lock] | |
641 | */ | |
9c28cbcc JK |
642 | unsigned int j_first; |
643 | unsigned int j_last; | |
1da177e4 LT |
644 | |
645 | /* | |
646 | * Device, blocksize and starting block offset for the location where we | |
647 | * store the journal. | |
648 | */ | |
649 | struct block_device *j_dev; | |
650 | int j_blocksize; | |
9c28cbcc | 651 | unsigned int j_blk_offset; |
1da177e4 LT |
652 | |
653 | /* | |
654 | * Device which holds the client fs. For internal journal this will be | |
655 | * equal to j_dev. | |
656 | */ | |
657 | struct block_device *j_fs_dev; | |
658 | ||
659 | /* Total maximum capacity of the journal region on disk. */ | |
660 | unsigned int j_maxlen; | |
661 | ||
662 | /* | |
663 | * Protects the buffer lists and internal buffer state. | |
664 | */ | |
665 | spinlock_t j_list_lock; | |
666 | ||
667 | /* Optional inode where we store the journal. If present, all */ | |
668 | /* journal block numbers are mapped into this inode via */ | |
669 | /* bmap(). */ | |
670 | struct inode *j_inode; | |
671 | ||
672 | /* | |
673 | * Sequence number of the oldest transaction in the log [j_state_lock] | |
674 | */ | |
675 | tid_t j_tail_sequence; | |
676 | ||
677 | /* | |
678 | * Sequence number of the next transaction to grant [j_state_lock] | |
679 | */ | |
680 | tid_t j_transaction_sequence; | |
681 | ||
682 | /* | |
683 | * Sequence number of the most recently committed transaction | |
684 | * [j_state_lock]. | |
685 | */ | |
686 | tid_t j_commit_sequence; | |
687 | ||
688 | /* | |
689 | * Sequence number of the most recent transaction wanting commit | |
690 | * [j_state_lock] | |
691 | */ | |
692 | tid_t j_commit_request; | |
693 | ||
2db938be JK |
694 | /* |
695 | * Sequence number of the most recent transaction someone is waiting | |
696 | * for to commit. | |
697 | * [j_state_lock] | |
698 | */ | |
699 | tid_t j_commit_waited; | |
700 | ||
1da177e4 LT |
701 | /* |
702 | * Journal uuid: identifies the object (filesystem, LVM volume etc) | |
703 | * backed by this journal. This will eventually be replaced by an array | |
704 | * of uuids, allowing us to index multiple devices within a single | |
705 | * journal and to perform atomic updates across them. | |
706 | */ | |
707 | __u8 j_uuid[16]; | |
708 | ||
709 | /* Pointer to the current commit thread for this journal */ | |
710 | struct task_struct *j_task; | |
711 | ||
712 | /* | |
713 | * Maximum number of metadata buffers to allow in a single compound | |
714 | * commit transaction | |
715 | */ | |
716 | int j_max_transaction_buffers; | |
717 | ||
718 | /* | |
719 | * What is the maximum transaction lifetime before we begin a commit? | |
720 | */ | |
721 | unsigned long j_commit_interval; | |
722 | ||
723 | /* The timer used to wakeup the commit thread: */ | |
e3df1898 | 724 | struct timer_list j_commit_timer; |
1da177e4 LT |
725 | |
726 | /* | |
727 | * The revoke table: maintains the list of revoked blocks in the | |
728 | * current transaction. [j_revoke_lock] | |
729 | */ | |
730 | spinlock_t j_revoke_lock; | |
731 | struct jbd_revoke_table_s *j_revoke; | |
732 | struct jbd_revoke_table_s *j_revoke_table[2]; | |
733 | ||
734 | /* | |
735 | * array of bhs for journal_commit_transaction | |
736 | */ | |
737 | struct buffer_head **j_wbuf; | |
738 | int j_wbufsize; | |
739 | ||
f420d4dc JB |
740 | /* |
741 | * this is the pid of the last person to run a synchronous operation | |
742 | * through the journal. | |
743 | */ | |
fe1dcbc4 AM |
744 | pid_t j_last_sync_writer; |
745 | ||
f420d4dc JB |
746 | /* |
747 | * the average amount of time in nanoseconds it takes to commit a | |
748 | * transaction to the disk. [j_state_lock] | |
749 | */ | |
750 | u64 j_average_commit_time; | |
751 | ||
1da177e4 LT |
752 | /* |
753 | * An opaque pointer to fs-private information. ext3 puts its | |
754 | * superblock pointer here | |
755 | */ | |
756 | void *j_private; | |
757 | }; | |
758 | ||
ae6ddcc5 MC |
759 | /* |
760 | * Journal flag definitions | |
1da177e4 LT |
761 | */ |
762 | #define JFS_UNMOUNT 0x001 /* Journal thread is being destroyed */ | |
763 | #define JFS_ABORT 0x002 /* Journaling has been aborted for errors. */ | |
764 | #define JFS_ACK_ERR 0x004 /* The errno in the sb has been acked */ | |
765 | #define JFS_FLUSHED 0x008 /* The journal superblock has been flushed */ | |
766 | #define JFS_LOADED 0x010 /* The journal superblock has been loaded */ | |
767 | #define JFS_BARRIER 0x020 /* Use IDE barriers */ | |
0e4fb5e2 HK |
768 | #define JFS_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file |
769 | * data write error in ordered | |
770 | * mode */ | |
1da177e4 | 771 | |
ae6ddcc5 | 772 | /* |
1da177e4 LT |
773 | * Function declarations for the journaling transaction and buffer |
774 | * management | |
775 | */ | |
776 | ||
777 | /* Filing buffers */ | |
1da177e4 LT |
778 | extern void journal_unfile_buffer(journal_t *, struct journal_head *); |
779 | extern void __journal_unfile_buffer(struct journal_head *); | |
780 | extern void __journal_refile_buffer(struct journal_head *); | |
781 | extern void journal_refile_buffer(journal_t *, struct journal_head *); | |
782 | extern void __journal_file_buffer(struct journal_head *, transaction_t *, int); | |
783 | extern void __journal_free_buffer(struct journal_head *bh); | |
784 | extern void journal_file_buffer(struct journal_head *, transaction_t *, int); | |
785 | extern void __journal_clean_data_list(transaction_t *transaction); | |
786 | ||
787 | /* Log buffer allocation */ | |
788 | extern struct journal_head * journal_get_descriptor_buffer(journal_t *); | |
9c28cbcc | 789 | int journal_next_log_block(journal_t *, unsigned int *); |
1da177e4 LT |
790 | |
791 | /* Commit management */ | |
792 | extern void journal_commit_transaction(journal_t *); | |
793 | ||
794 | /* Checkpoint list management */ | |
795 | int __journal_clean_checkpoint_list(journal_t *journal); | |
78ce89c9 | 796 | int __journal_remove_checkpoint(struct journal_head *); |
1da177e4 LT |
797 | void __journal_insert_checkpoint(struct journal_head *, transaction_t *); |
798 | ||
799 | /* Buffer IO */ | |
ae6ddcc5 | 800 | extern int |
1da177e4 LT |
801 | journal_write_metadata_buffer(transaction_t *transaction, |
802 | struct journal_head *jh_in, | |
803 | struct journal_head **jh_out, | |
9c28cbcc | 804 | unsigned int blocknr); |
1da177e4 LT |
805 | |
806 | /* Transaction locking */ | |
807 | extern void __wait_on_journal (journal_t *); | |
808 | ||
809 | /* | |
810 | * Journal locking. | |
811 | * | |
812 | * We need to lock the journal during transaction state changes so that nobody | |
813 | * ever tries to take a handle on the running transaction while we are in the | |
814 | * middle of moving it to the commit phase. j_state_lock does this. | |
815 | * | |
816 | * Note that the locking is completely interrupt unsafe. We never touch | |
817 | * journal structures from interrupts. | |
818 | */ | |
819 | ||
820 | static inline handle_t *journal_current_handle(void) | |
821 | { | |
822 | return current->journal_info; | |
823 | } | |
824 | ||
825 | /* The journaling code user interface: | |
826 | * | |
827 | * Create and destroy handles | |
ae6ddcc5 | 828 | * Register buffer modifications against the current transaction. |
1da177e4 LT |
829 | */ |
830 | ||
831 | extern handle_t *journal_start(journal_t *, int nblocks); | |
832 | extern int journal_restart (handle_t *, int nblocks); | |
833 | extern int journal_extend (handle_t *, int nblocks); | |
834 | extern int journal_get_write_access(handle_t *, struct buffer_head *); | |
835 | extern int journal_get_create_access (handle_t *, struct buffer_head *); | |
836 | extern int journal_get_undo_access(handle_t *, struct buffer_head *); | |
837 | extern int journal_dirty_data (handle_t *, struct buffer_head *); | |
838 | extern int journal_dirty_metadata (handle_t *, struct buffer_head *); | |
839 | extern void journal_release_buffer (handle_t *, struct buffer_head *); | |
840 | extern int journal_forget (handle_t *, struct buffer_head *); | |
841 | extern void journal_sync_buffer (struct buffer_head *); | |
2ff28e22 | 842 | extern void journal_invalidatepage(journal_t *, |
1da177e4 | 843 | struct page *, unsigned long); |
27496a8c | 844 | extern int journal_try_to_free_buffers(journal_t *, struct page *, gfp_t); |
1da177e4 LT |
845 | extern int journal_stop(handle_t *); |
846 | extern int journal_flush (journal_t *); | |
847 | extern void journal_lock_updates (journal_t *); | |
848 | extern void journal_unlock_updates (journal_t *); | |
849 | ||
850 | extern journal_t * journal_init_dev(struct block_device *bdev, | |
851 | struct block_device *fs_dev, | |
852 | int start, int len, int bsize); | |
853 | extern journal_t * journal_init_inode (struct inode *); | |
854 | extern int journal_update_format (journal_t *); | |
ae6ddcc5 | 855 | extern int journal_check_used_features |
1da177e4 | 856 | (journal_t *, unsigned long, unsigned long, unsigned long); |
ae6ddcc5 | 857 | extern int journal_check_available_features |
1da177e4 | 858 | (journal_t *, unsigned long, unsigned long, unsigned long); |
ae6ddcc5 | 859 | extern int journal_set_features |
1da177e4 LT |
860 | (journal_t *, unsigned long, unsigned long, unsigned long); |
861 | extern int journal_create (journal_t *); | |
862 | extern int journal_load (journal_t *journal); | |
4afe9785 | 863 | extern int journal_destroy (journal_t *); |
1da177e4 LT |
864 | extern int journal_recover (journal_t *journal); |
865 | extern int journal_wipe (journal_t *, int); | |
866 | extern int journal_skip_recovery (journal_t *); | |
fd2cbd4d JK |
867 | extern void journal_update_sb_log_tail (journal_t *, tid_t, unsigned int, |
868 | int); | |
1da177e4 LT |
869 | extern void journal_abort (journal_t *, int); |
870 | extern int journal_errno (journal_t *); | |
871 | extern void journal_ack_err (journal_t *); | |
872 | extern int journal_clear_err (journal_t *); | |
9c28cbcc | 873 | extern int journal_bmap(journal_t *, unsigned int, unsigned int *); |
1da177e4 LT |
874 | extern int journal_force_commit(journal_t *); |
875 | ||
876 | /* | |
877 | * journal_head management | |
878 | */ | |
879 | struct journal_head *journal_add_journal_head(struct buffer_head *bh); | |
880 | struct journal_head *journal_grab_journal_head(struct buffer_head *bh); | |
1da177e4 LT |
881 | void journal_put_journal_head(struct journal_head *jh); |
882 | ||
883 | /* | |
884 | * handle management | |
885 | */ | |
e18b890b | 886 | extern struct kmem_cache *jbd_handle_cache; |
1da177e4 | 887 | |
dd0fc66f | 888 | static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags) |
1da177e4 | 889 | { |
e162b2f8 | 890 | return kmem_cache_zalloc(jbd_handle_cache, gfp_flags); |
1da177e4 LT |
891 | } |
892 | ||
893 | static inline void jbd_free_handle(handle_t *handle) | |
894 | { | |
895 | kmem_cache_free(jbd_handle_cache, handle); | |
896 | } | |
897 | ||
898 | /* Primary revoke support */ | |
899 | #define JOURNAL_REVOKE_DEFAULT_HASH 256 | |
900 | extern int journal_init_revoke(journal_t *, int); | |
901 | extern void journal_destroy_revoke_caches(void); | |
902 | extern int journal_init_revoke_caches(void); | |
903 | ||
904 | extern void journal_destroy_revoke(journal_t *); | |
905 | extern int journal_revoke (handle_t *, | |
9c28cbcc | 906 | unsigned int, struct buffer_head *); |
1da177e4 | 907 | extern int journal_cancel_revoke(handle_t *, struct journal_head *); |
38d726d1 TT |
908 | extern void journal_write_revoke_records(journal_t *, |
909 | transaction_t *, int); | |
1da177e4 LT |
910 | |
911 | /* Recovery revoke support */ | |
9c28cbcc JK |
912 | extern int journal_set_revoke(journal_t *, unsigned int, tid_t); |
913 | extern int journal_test_revoke(journal_t *, unsigned int, tid_t); | |
1da177e4 | 914 | extern void journal_clear_revoke(journal_t *); |
1da177e4 | 915 | extern void journal_switch_revoke_table(journal_t *journal); |
8c111b3f | 916 | extern void journal_clear_buffer_revoked_flags(journal_t *journal); |
1da177e4 LT |
917 | |
918 | /* | |
919 | * The log thread user interface: | |
920 | * | |
921 | * Request space in the current transaction, and force transaction commit | |
922 | * transitions on demand. | |
923 | */ | |
924 | ||
925 | int __log_space_left(journal_t *); /* Called with journal locked */ | |
926 | int log_start_commit(journal_t *journal, tid_t tid); | |
927 | int __log_start_commit(journal_t *journal, tid_t tid); | |
928 | int journal_start_commit(journal_t *journal, tid_t *tid); | |
929 | int journal_force_commit_nested(journal_t *journal); | |
930 | int log_wait_commit(journal_t *journal, tid_t tid); | |
931 | int log_do_checkpoint(journal_t *journal); | |
03f4d804 | 932 | int journal_trans_will_send_data_barrier(journal_t *journal, tid_t tid); |
1da177e4 LT |
933 | |
934 | void __log_wait_for_space(journal_t *journal); | |
935 | extern void __journal_drop_transaction(journal_t *, transaction_t *); | |
936 | extern int cleanup_journal_tail(journal_t *); | |
937 | ||
938 | /* Debugging code only: */ | |
939 | ||
940 | #define jbd_ENOSYS() \ | |
941 | do { \ | |
d5c003b4 | 942 | printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \ |
1da177e4 LT |
943 | current->state = TASK_UNINTERRUPTIBLE; \ |
944 | schedule(); \ | |
945 | } while (1) | |
946 | ||
947 | /* | |
948 | * is_journal_abort | |
949 | * | |
950 | * Simple test wrapper function to test the JFS_ABORT state flag. This | |
951 | * bit, when set, indicates that we have had a fatal error somewhere, | |
952 | * either inside the journaling layer or indicated to us by the client | |
953 | * (eg. ext3), and that we and should not commit any further | |
ae6ddcc5 | 954 | * transactions. |
1da177e4 LT |
955 | */ |
956 | ||
957 | static inline int is_journal_aborted(journal_t *journal) | |
958 | { | |
959 | return journal->j_flags & JFS_ABORT; | |
960 | } | |
961 | ||
962 | static inline int is_handle_aborted(handle_t *handle) | |
963 | { | |
964 | if (handle->h_aborted) | |
965 | return 1; | |
966 | return is_journal_aborted(handle->h_transaction->t_journal); | |
967 | } | |
968 | ||
969 | static inline void journal_abort_handle(handle_t *handle) | |
970 | { | |
971 | handle->h_aborted = 1; | |
972 | } | |
973 | ||
974 | #endif /* __KERNEL__ */ | |
975 | ||
976 | /* Comparison functions for transaction IDs: perform comparisons using | |
977 | * modulo arithmetic so that they work over sequence number wraps. */ | |
978 | ||
979 | static inline int tid_gt(tid_t x, tid_t y) | |
980 | { | |
981 | int difference = (x - y); | |
982 | return (difference > 0); | |
983 | } | |
984 | ||
985 | static inline int tid_geq(tid_t x, tid_t y) | |
986 | { | |
987 | int difference = (x - y); | |
988 | return (difference >= 0); | |
989 | } | |
990 | ||
991 | extern int journal_blocks_per_page(struct inode *inode); | |
992 | ||
993 | /* | |
994 | * Return the minimum number of blocks which must be free in the journal | |
995 | * before a new transaction may be started. Must be called under j_state_lock. | |
996 | */ | |
997 | static inline int jbd_space_needed(journal_t *journal) | |
998 | { | |
999 | int nblocks = journal->j_max_transaction_buffers; | |
1000 | if (journal->j_committing_transaction) | |
1001 | nblocks += journal->j_committing_transaction-> | |
1002 | t_outstanding_credits; | |
1003 | return nblocks; | |
1004 | } | |
1005 | ||
1006 | /* | |
1007 | * Definitions which augment the buffer_head layer | |
1008 | */ | |
1009 | ||
1010 | /* journaling buffer types */ | |
1011 | #define BJ_None 0 /* Not journaled */ | |
1012 | #define BJ_SyncData 1 /* Normal data: flush before commit */ | |
1013 | #define BJ_Metadata 2 /* Normal journaled metadata */ | |
1014 | #define BJ_Forget 3 /* Buffer superseded by this transaction */ | |
1015 | #define BJ_IO 4 /* Buffer is for temporary IO use */ | |
1016 | #define BJ_Shadow 5 /* Buffer contents being shadowed to the log */ | |
1017 | #define BJ_LogCtl 6 /* Buffer contains log descriptors */ | |
1018 | #define BJ_Reserved 7 /* Buffer is reserved for access by journal */ | |
1019 | #define BJ_Locked 8 /* Locked for I/O during commit */ | |
1020 | #define BJ_Types 9 | |
ae6ddcc5 | 1021 | |
1da177e4 LT |
1022 | extern int jbd_blocks_per_page(struct inode *inode); |
1023 | ||
1024 | #ifdef __KERNEL__ | |
1025 | ||
1026 | #define buffer_trace_init(bh) do {} while (0) | |
1027 | #define print_buffer_fields(bh) do {} while (0) | |
1028 | #define print_buffer_trace(bh) do {} while (0) | |
1029 | #define BUFFER_TRACE(bh, info) do {} while (0) | |
1030 | #define BUFFER_TRACE2(bh, bh2, info) do {} while (0) | |
1031 | #define JBUFFER_TRACE(jh, info) do {} while (0) | |
1032 | ||
1033 | #endif /* __KERNEL__ */ | |
1034 | ||
1da177e4 | 1035 | #endif /* _LINUX_JBD_H */ |