Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/include/linux/jbd.h | |
3 | * | |
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 AM |
32 | #include <linux/timer.h> |
33 | ||
1da177e4 LT |
34 | #include <asm/semaphore.h> |
35 | #endif | |
36 | ||
37 | #define journal_oom_retry 1 | |
38 | ||
39 | /* | |
40 | * Define JBD_PARANIOD_IOFAIL to cause a kernel BUG() if ext3 finds | |
41 | * certain classes of error which can occur due to failed IOs. Under | |
42 | * normal use we want ext3 to continue after such errors, because | |
43 | * hardware _can_ fail, but for debugging purposes when running tests on | |
44 | * known-good hardware we may want to trap these errors. | |
45 | */ | |
46 | #undef JBD_PARANOID_IOFAIL | |
47 | ||
48 | /* | |
49 | * The default maximum commit age, in seconds. | |
50 | */ | |
51 | #define JBD_DEFAULT_MAX_COMMIT_AGE 5 | |
52 | ||
53 | #ifdef CONFIG_JBD_DEBUG | |
54 | /* | |
55 | * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal | |
56 | * consistency checks. By default we don't do this unless | |
57 | * CONFIG_JBD_DEBUG is on. | |
58 | */ | |
59 | #define JBD_EXPENSIVE_CHECKING | |
60 | extern int journal_enable_debug; | |
61 | ||
62 | #define jbd_debug(n, f, a...) \ | |
63 | do { \ | |
64 | if ((n) <= journal_enable_debug) { \ | |
65 | printk (KERN_DEBUG "(%s, %d): %s: ", \ | |
66 | __FILE__, __LINE__, __FUNCTION__); \ | |
67 | printk (f, ## a); \ | |
68 | } \ | |
69 | } while (0) | |
70 | #else | |
71 | #define jbd_debug(f, a...) /**/ | |
72 | #endif | |
73 | ||
27496a8c | 74 | extern void * __jbd_kmalloc (const char *where, size_t size, gfp_t flags, int retry); |
1da177e4 LT |
75 | #define jbd_kmalloc(size, flags) \ |
76 | __jbd_kmalloc(__FUNCTION__, (size), (flags), journal_oom_retry) | |
77 | #define jbd_rep_kmalloc(size, flags) \ | |
78 | __jbd_kmalloc(__FUNCTION__, (size), (flags), 1) | |
79 | ||
80 | #define JFS_MIN_JOURNAL_BLOCKS 1024 | |
81 | ||
82 | #ifdef __KERNEL__ | |
83 | ||
84 | /** | |
85 | * typedef handle_t - The handle_t type represents a single atomic update being performed by some process. | |
86 | * | |
87 | * All filesystem modifications made by the process go | |
88 | * through this handle. Recursive operations (such as quota operations) | |
89 | * are gathered into a single update. | |
90 | * | |
91 | * The buffer credits field is used to account for journaled buffers | |
92 | * being modified by the running process. To ensure that there is | |
93 | * enough log space for all outstanding operations, we need to limit the | |
94 | * number of outstanding buffers possible at any time. When the | |
95 | * operation completes, any buffer credits not used are credited back to | |
96 | * the transaction, so that at all times we know how many buffers the | |
97 | * outstanding updates on a transaction might possibly touch. | |
98 | * | |
99 | * This is an opaque datatype. | |
100 | **/ | |
101 | typedef struct handle_s handle_t; /* Atomic operation type */ | |
102 | ||
103 | ||
104 | /** | |
105 | * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem. | |
106 | * | |
107 | * journal_t is linked to from the fs superblock structure. | |
108 | * | |
109 | * We use the journal_t to keep track of all outstanding transaction | |
110 | * activity on the filesystem, and to manage the state of the log | |
111 | * writing process. | |
112 | * | |
113 | * This is an opaque datatype. | |
114 | **/ | |
115 | typedef struct journal_s journal_t; /* Journal control structure */ | |
116 | #endif | |
117 | ||
118 | /* | |
119 | * Internal structures used by the logging mechanism: | |
120 | */ | |
121 | ||
122 | #define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */ | |
123 | ||
124 | /* | |
125 | * On-disk structures | |
126 | */ | |
127 | ||
128 | /* | |
129 | * Descriptor block types: | |
130 | */ | |
131 | ||
132 | #define JFS_DESCRIPTOR_BLOCK 1 | |
133 | #define JFS_COMMIT_BLOCK 2 | |
134 | #define JFS_SUPERBLOCK_V1 3 | |
135 | #define JFS_SUPERBLOCK_V2 4 | |
136 | #define JFS_REVOKE_BLOCK 5 | |
137 | ||
138 | /* | |
139 | * Standard header for all descriptor blocks: | |
140 | */ | |
141 | typedef struct journal_header_s | |
142 | { | |
143 | __be32 h_magic; | |
144 | __be32 h_blocktype; | |
145 | __be32 h_sequence; | |
146 | } journal_header_t; | |
147 | ||
148 | ||
149 | /* | |
150 | * The block tag: used to describe a single buffer in the journal | |
151 | */ | |
152 | typedef struct journal_block_tag_s | |
153 | { | |
154 | __be32 t_blocknr; /* The on-disk block number */ | |
155 | __be32 t_flags; /* See below */ | |
156 | } journal_block_tag_t; | |
157 | ||
158 | /* | |
159 | * The revoke descriptor: used on disk to describe a series of blocks to | |
160 | * be revoked from the log | |
161 | */ | |
162 | typedef struct journal_revoke_header_s | |
163 | { | |
164 | journal_header_t r_header; | |
165 | __be32 r_count; /* Count of bytes used in the block */ | |
166 | } journal_revoke_header_t; | |
167 | ||
168 | ||
169 | /* Definitions for the journal tag flags word: */ | |
170 | #define JFS_FLAG_ESCAPE 1 /* on-disk block is escaped */ | |
171 | #define JFS_FLAG_SAME_UUID 2 /* block has same uuid as previous */ | |
172 | #define JFS_FLAG_DELETED 4 /* block deleted by this transaction */ | |
173 | #define JFS_FLAG_LAST_TAG 8 /* last tag in this descriptor block */ | |
174 | ||
175 | ||
176 | /* | |
177 | * The journal superblock. All fields are in big-endian byte order. | |
178 | */ | |
179 | typedef struct journal_superblock_s | |
180 | { | |
181 | /* 0x0000 */ | |
182 | journal_header_t s_header; | |
183 | ||
184 | /* 0x000C */ | |
185 | /* Static information describing the journal */ | |
186 | __be32 s_blocksize; /* journal device blocksize */ | |
187 | __be32 s_maxlen; /* total blocks in journal file */ | |
188 | __be32 s_first; /* first block of log information */ | |
189 | ||
190 | /* 0x0018 */ | |
191 | /* Dynamic information describing the current state of the log */ | |
192 | __be32 s_sequence; /* first commit ID expected in log */ | |
193 | __be32 s_start; /* blocknr of start of log */ | |
194 | ||
195 | /* 0x0020 */ | |
196 | /* Error value, as set by journal_abort(). */ | |
197 | __be32 s_errno; | |
198 | ||
199 | /* 0x0024 */ | |
200 | /* Remaining fields are only valid in a version-2 superblock */ | |
201 | __be32 s_feature_compat; /* compatible feature set */ | |
202 | __be32 s_feature_incompat; /* incompatible feature set */ | |
203 | __be32 s_feature_ro_compat; /* readonly-compatible feature set */ | |
204 | /* 0x0030 */ | |
205 | __u8 s_uuid[16]; /* 128-bit uuid for journal */ | |
206 | ||
207 | /* 0x0040 */ | |
208 | __be32 s_nr_users; /* Nr of filesystems sharing log */ | |
209 | ||
210 | __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/ | |
211 | ||
212 | /* 0x0048 */ | |
213 | __be32 s_max_transaction; /* Limit of journal blocks per trans.*/ | |
214 | __be32 s_max_trans_data; /* Limit of data blocks per trans. */ | |
215 | ||
216 | /* 0x0050 */ | |
217 | __u32 s_padding[44]; | |
218 | ||
219 | /* 0x0100 */ | |
220 | __u8 s_users[16*48]; /* ids of all fs'es sharing the log */ | |
221 | /* 0x0400 */ | |
222 | } journal_superblock_t; | |
223 | ||
224 | #define JFS_HAS_COMPAT_FEATURE(j,mask) \ | |
225 | ((j)->j_format_version >= 2 && \ | |
226 | ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask)))) | |
227 | #define JFS_HAS_RO_COMPAT_FEATURE(j,mask) \ | |
228 | ((j)->j_format_version >= 2 && \ | |
229 | ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask)))) | |
230 | #define JFS_HAS_INCOMPAT_FEATURE(j,mask) \ | |
231 | ((j)->j_format_version >= 2 && \ | |
232 | ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask)))) | |
233 | ||
234 | #define JFS_FEATURE_INCOMPAT_REVOKE 0x00000001 | |
235 | ||
236 | /* Features known to this kernel version: */ | |
237 | #define JFS_KNOWN_COMPAT_FEATURES 0 | |
238 | #define JFS_KNOWN_ROCOMPAT_FEATURES 0 | |
239 | #define JFS_KNOWN_INCOMPAT_FEATURES JFS_FEATURE_INCOMPAT_REVOKE | |
240 | ||
241 | #ifdef __KERNEL__ | |
242 | ||
243 | #include <linux/fs.h> | |
244 | #include <linux/sched.h> | |
1da177e4 LT |
245 | |
246 | #define JBD_ASSERTIONS | |
247 | #ifdef JBD_ASSERTIONS | |
248 | #define J_ASSERT(assert) \ | |
249 | do { \ | |
250 | if (!(assert)) { \ | |
251 | printk (KERN_EMERG \ | |
252 | "Assertion failure in %s() at %s:%d: \"%s\"\n", \ | |
253 | __FUNCTION__, __FILE__, __LINE__, # assert); \ | |
254 | BUG(); \ | |
255 | } \ | |
256 | } while (0) | |
257 | ||
258 | #if defined(CONFIG_BUFFER_DEBUG) | |
259 | void buffer_assertion_failure(struct buffer_head *bh); | |
260 | #define J_ASSERT_BH(bh, expr) \ | |
261 | do { \ | |
262 | if (!(expr)) \ | |
263 | buffer_assertion_failure(bh); \ | |
264 | J_ASSERT(expr); \ | |
265 | } while (0) | |
266 | #define J_ASSERT_JH(jh, expr) J_ASSERT_BH(jh2bh(jh), expr) | |
267 | #else | |
268 | #define J_ASSERT_BH(bh, expr) J_ASSERT(expr) | |
269 | #define J_ASSERT_JH(jh, expr) J_ASSERT(expr) | |
270 | #endif | |
271 | ||
272 | #else | |
273 | #define J_ASSERT(assert) do { } while (0) | |
274 | #endif /* JBD_ASSERTIONS */ | |
275 | ||
276 | #if defined(JBD_PARANOID_IOFAIL) | |
277 | #define J_EXPECT(expr, why...) J_ASSERT(expr) | |
278 | #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr) | |
279 | #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr) | |
280 | #else | |
281 | #define __journal_expect(expr, why...) \ | |
282 | ({ \ | |
283 | int val = (expr); \ | |
284 | if (!val) { \ | |
285 | printk(KERN_ERR \ | |
286 | "EXT3-fs unexpected failure: %s;\n",# expr); \ | |
287 | printk(KERN_ERR why "\n"); \ | |
288 | } \ | |
289 | val; \ | |
290 | }) | |
291 | #define J_EXPECT(expr, why...) __journal_expect(expr, ## why) | |
292 | #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why) | |
293 | #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why) | |
294 | #endif | |
295 | ||
296 | enum jbd_state_bits { | |
297 | BH_JBD /* Has an attached ext3 journal_head */ | |
298 | = BH_PrivateStart, | |
299 | BH_JWrite, /* Being written to log (@@@ DEBUGGING) */ | |
300 | BH_Freed, /* Has been freed (truncated) */ | |
301 | BH_Revoked, /* Has been revoked from the log */ | |
302 | BH_RevokeValid, /* Revoked flag is valid */ | |
303 | BH_JBDDirty, /* Is dirty but journaled */ | |
304 | BH_State, /* Pins most journal_head state */ | |
305 | BH_JournalHead, /* Pins bh->b_private and jh->b_bh */ | |
306 | BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */ | |
307 | }; | |
308 | ||
309 | BUFFER_FNS(JBD, jbd) | |
310 | BUFFER_FNS(JWrite, jwrite) | |
311 | BUFFER_FNS(JBDDirty, jbddirty) | |
312 | TAS_BUFFER_FNS(JBDDirty, jbddirty) | |
313 | BUFFER_FNS(Revoked, revoked) | |
314 | TAS_BUFFER_FNS(Revoked, revoked) | |
315 | BUFFER_FNS(RevokeValid, revokevalid) | |
316 | TAS_BUFFER_FNS(RevokeValid, revokevalid) | |
317 | BUFFER_FNS(Freed, freed) | |
318 | ||
319 | static inline struct buffer_head *jh2bh(struct journal_head *jh) | |
320 | { | |
321 | return jh->b_bh; | |
322 | } | |
323 | ||
324 | static inline struct journal_head *bh2jh(struct buffer_head *bh) | |
325 | { | |
326 | return bh->b_private; | |
327 | } | |
328 | ||
329 | static inline void jbd_lock_bh_state(struct buffer_head *bh) | |
330 | { | |
331 | bit_spin_lock(BH_State, &bh->b_state); | |
332 | } | |
333 | ||
334 | static inline int jbd_trylock_bh_state(struct buffer_head *bh) | |
335 | { | |
336 | return bit_spin_trylock(BH_State, &bh->b_state); | |
337 | } | |
338 | ||
339 | static inline int jbd_is_locked_bh_state(struct buffer_head *bh) | |
340 | { | |
341 | return bit_spin_is_locked(BH_State, &bh->b_state); | |
342 | } | |
343 | ||
344 | static inline void jbd_unlock_bh_state(struct buffer_head *bh) | |
345 | { | |
346 | bit_spin_unlock(BH_State, &bh->b_state); | |
347 | } | |
348 | ||
349 | static inline void jbd_lock_bh_journal_head(struct buffer_head *bh) | |
350 | { | |
351 | bit_spin_lock(BH_JournalHead, &bh->b_state); | |
352 | } | |
353 | ||
354 | static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh) | |
355 | { | |
356 | bit_spin_unlock(BH_JournalHead, &bh->b_state); | |
357 | } | |
358 | ||
359 | struct jbd_revoke_table_s; | |
360 | ||
361 | /** | |
362 | * struct handle_s - The handle_s type is the concrete type associated with | |
363 | * handle_t. | |
364 | * @h_transaction: Which compound transaction is this update a part of? | |
365 | * @h_buffer_credits: Number of remaining buffers we are allowed to dirty. | |
366 | * @h_ref: Reference count on this handle | |
367 | * @h_err: Field for caller's use to track errors through large fs operations | |
368 | * @h_sync: flag for sync-on-close | |
369 | * @h_jdata: flag to force data journaling | |
370 | * @h_aborted: flag indicating fatal error on handle | |
371 | **/ | |
372 | ||
373 | /* Docbook can't yet cope with the bit fields, but will leave the documentation | |
374 | * in so it can be fixed later. | |
375 | */ | |
376 | ||
377 | struct handle_s | |
378 | { | |
379 | /* Which compound transaction is this update a part of? */ | |
380 | transaction_t *h_transaction; | |
381 | ||
382 | /* Number of remaining buffers we are allowed to dirty: */ | |
383 | int h_buffer_credits; | |
384 | ||
385 | /* Reference count on this handle */ | |
386 | int h_ref; | |
387 | ||
388 | /* Field for caller's use to track errors through large fs */ | |
389 | /* operations */ | |
390 | int h_err; | |
391 | ||
392 | /* Flags [no locking] */ | |
393 | unsigned int h_sync: 1; /* sync-on-close */ | |
394 | unsigned int h_jdata: 1; /* force data journaling */ | |
395 | unsigned int h_aborted: 1; /* fatal error on handle */ | |
396 | }; | |
397 | ||
398 | ||
399 | /* The transaction_t type is the guts of the journaling mechanism. It | |
400 | * tracks a compound transaction through its various states: | |
401 | * | |
402 | * RUNNING: accepting new updates | |
403 | * LOCKED: Updates still running but we don't accept new ones | |
404 | * RUNDOWN: Updates are tidying up but have finished requesting | |
405 | * new buffers to modify (state not used for now) | |
406 | * FLUSH: All updates complete, but we are still writing to disk | |
407 | * COMMIT: All data on disk, writing commit record | |
408 | * FINISHED: We still have to keep the transaction for checkpointing. | |
409 | * | |
410 | * The transaction keeps track of all of the buffers modified by a | |
411 | * running transaction, and all of the buffers committed but not yet | |
412 | * flushed to home for finished transactions. | |
413 | */ | |
414 | ||
415 | /* | |
416 | * Lock ranking: | |
417 | * | |
418 | * j_list_lock | |
419 | * ->jbd_lock_bh_journal_head() (This is "innermost") | |
420 | * | |
421 | * j_state_lock | |
422 | * ->jbd_lock_bh_state() | |
423 | * | |
424 | * jbd_lock_bh_state() | |
425 | * ->j_list_lock | |
426 | * | |
427 | * j_state_lock | |
428 | * ->t_handle_lock | |
429 | * | |
430 | * j_state_lock | |
431 | * ->j_list_lock (journal_unmap_buffer) | |
432 | * | |
433 | */ | |
434 | ||
435 | struct transaction_s | |
436 | { | |
437 | /* Pointer to the journal for this transaction. [no locking] */ | |
438 | journal_t *t_journal; | |
439 | ||
440 | /* Sequence number for this transaction [no locking] */ | |
441 | tid_t t_tid; | |
442 | ||
443 | /* | |
444 | * Transaction's current state | |
445 | * [no locking - only kjournald alters this] | |
446 | * FIXME: needs barriers | |
447 | * KLUDGE: [use j_state_lock] | |
448 | */ | |
449 | enum { | |
450 | T_RUNNING, | |
451 | T_LOCKED, | |
452 | T_RUNDOWN, | |
453 | T_FLUSH, | |
454 | T_COMMIT, | |
455 | T_FINISHED | |
456 | } t_state; | |
457 | ||
458 | /* | |
459 | * Where in the log does this transaction's commit start? [no locking] | |
460 | */ | |
461 | unsigned long t_log_start; | |
462 | ||
463 | /* Number of buffers on the t_buffers list [j_list_lock] */ | |
464 | int t_nr_buffers; | |
465 | ||
466 | /* | |
467 | * Doubly-linked circular list of all buffers reserved but not yet | |
468 | * modified by this transaction [j_list_lock] | |
469 | */ | |
470 | struct journal_head *t_reserved_list; | |
471 | ||
472 | /* | |
473 | * Doubly-linked circular list of all buffers under writeout during | |
474 | * commit [j_list_lock] | |
475 | */ | |
476 | struct journal_head *t_locked_list; | |
477 | ||
478 | /* | |
479 | * Doubly-linked circular list of all metadata buffers owned by this | |
480 | * transaction [j_list_lock] | |
481 | */ | |
482 | struct journal_head *t_buffers; | |
483 | ||
484 | /* | |
485 | * Doubly-linked circular list of all data buffers still to be | |
486 | * flushed before this transaction can be committed [j_list_lock] | |
487 | */ | |
488 | struct journal_head *t_sync_datalist; | |
489 | ||
490 | /* | |
491 | * Doubly-linked circular list of all forget buffers (superseded | |
492 | * buffers which we can un-checkpoint once this transaction commits) | |
493 | * [j_list_lock] | |
494 | */ | |
495 | struct journal_head *t_forget; | |
496 | ||
497 | /* | |
498 | * Doubly-linked circular list of all buffers still to be flushed before | |
499 | * this transaction can be checkpointed. [j_list_lock] | |
500 | */ | |
501 | struct journal_head *t_checkpoint_list; | |
502 | ||
503 | /* | |
504 | * Doubly-linked circular list of temporary buffers currently undergoing | |
505 | * IO in the log [j_list_lock] | |
506 | */ | |
507 | struct journal_head *t_iobuf_list; | |
508 | ||
509 | /* | |
510 | * Doubly-linked circular list of metadata buffers being shadowed by log | |
511 | * IO. The IO buffers on the iobuf list and the shadow buffers on this | |
512 | * list match each other one for one at all times. [j_list_lock] | |
513 | */ | |
514 | struct journal_head *t_shadow_list; | |
515 | ||
516 | /* | |
517 | * Doubly-linked circular list of control buffers being written to the | |
518 | * log. [j_list_lock] | |
519 | */ | |
520 | struct journal_head *t_log_list; | |
521 | ||
522 | /* | |
523 | * Protects info related to handles | |
524 | */ | |
525 | spinlock_t t_handle_lock; | |
526 | ||
527 | /* | |
528 | * Number of outstanding updates running on this transaction | |
529 | * [t_handle_lock] | |
530 | */ | |
531 | int t_updates; | |
532 | ||
533 | /* | |
534 | * Number of buffers reserved for use by all handles in this transaction | |
535 | * handle but not yet modified. [t_handle_lock] | |
536 | */ | |
537 | int t_outstanding_credits; | |
538 | ||
539 | /* | |
540 | * Forward and backward links for the circular list of all transactions | |
541 | * awaiting checkpoint. [j_list_lock] | |
542 | */ | |
543 | transaction_t *t_cpnext, *t_cpprev; | |
544 | ||
545 | /* | |
546 | * When will the transaction expire (become due for commit), in jiffies? | |
547 | * [no locking] | |
548 | */ | |
549 | unsigned long t_expires; | |
550 | ||
551 | /* | |
552 | * How many handles used this transaction? [t_handle_lock] | |
553 | */ | |
554 | int t_handle_count; | |
555 | ||
556 | }; | |
557 | ||
558 | /** | |
559 | * struct journal_s - The journal_s type is the concrete type associated with | |
560 | * journal_t. | |
561 | * @j_flags: General journaling state flags | |
562 | * @j_errno: Is there an outstanding uncleared error on the journal (from a | |
563 | * prior abort)? | |
564 | * @j_sb_buffer: First part of superblock buffer | |
565 | * @j_superblock: Second part of superblock buffer | |
566 | * @j_format_version: Version of the superblock format | |
567 | * @j_state_lock: Protect the various scalars in the journal | |
568 | * @j_barrier_count: Number of processes waiting to create a barrier lock | |
569 | * @j_barrier: The barrier lock itself | |
570 | * @j_running_transaction: The current running transaction.. | |
571 | * @j_committing_transaction: the transaction we are pushing to disk | |
572 | * @j_checkpoint_transactions: a linked circular list of all transactions | |
573 | * waiting for checkpointing | |
574 | * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction | |
575 | * to start committing, or for a barrier lock to be released | |
576 | * @j_wait_logspace: Wait queue for waiting for checkpointing to complete | |
577 | * @j_wait_done_commit: Wait queue for waiting for commit to complete | |
578 | * @j_wait_checkpoint: Wait queue to trigger checkpointing | |
579 | * @j_wait_commit: Wait queue to trigger commit | |
580 | * @j_wait_updates: Wait queue to wait for updates to complete | |
2c68ee75 | 581 | * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints |
1da177e4 LT |
582 | * @j_head: Journal head - identifies the first unused block in the journal |
583 | * @j_tail: Journal tail - identifies the oldest still-used block in the | |
584 | * journal. | |
585 | * @j_free: Journal free - how many free blocks are there in the journal? | |
586 | * @j_first: The block number of the first usable block | |
587 | * @j_last: The block number one beyond the last usable block | |
588 | * @j_dev: Device where we store the journal | |
589 | * @j_blocksize: blocksize for the location where we store the journal. | |
590 | * @j_blk_offset: starting block offset for into the device where we store the | |
591 | * journal | |
592 | * @j_fs_dev: Device which holds the client fs. For internal journal this will | |
593 | * be equal to j_dev | |
594 | * @j_maxlen: Total maximum capacity of the journal region on disk. | |
595 | * @j_list_lock: Protects the buffer lists and internal buffer state. | |
596 | * @j_inode: Optional inode where we store the journal. If present, all journal | |
597 | * block numbers are mapped into this inode via bmap(). | |
598 | * @j_tail_sequence: Sequence number of the oldest transaction in the log | |
599 | * @j_transaction_sequence: Sequence number of the next transaction to grant | |
600 | * @j_commit_sequence: Sequence number of the most recently committed | |
601 | * transaction | |
602 | * @j_commit_request: Sequence number of the most recent transaction wanting | |
603 | * commit | |
604 | * @j_uuid: Uuid of client object. | |
605 | * @j_task: Pointer to the current commit thread for this journal | |
606 | * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a | |
607 | * single compound commit transaction | |
608 | * @j_commit_interval: What is the maximum transaction lifetime before we begin | |
609 | * a commit? | |
610 | * @j_commit_timer: The timer used to wakeup the commit thread | |
611 | * @j_revoke_lock: Protect the revoke table | |
612 | * @j_revoke: The revoke table - maintains the list of revoked blocks in the | |
613 | * current transaction. | |
614 | * @j_revoke_table: alternate revoke tables for j_revoke | |
6c8bec6d RD |
615 | * @j_wbuf: array of buffer_heads for journal_commit_transaction |
616 | * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the | |
617 | * number that will fit in j_blocksize | |
fe1dcbc4 | 618 | * @j_last_sync_writer: most recent pid which did a synchronous write |
1da177e4 LT |
619 | * @j_private: An opaque pointer to fs-private information. |
620 | */ | |
621 | ||
622 | struct journal_s | |
623 | { | |
624 | /* General journaling state flags [j_state_lock] */ | |
625 | unsigned long j_flags; | |
626 | ||
627 | /* | |
628 | * Is there an outstanding uncleared error on the journal (from a prior | |
629 | * abort)? [j_state_lock] | |
630 | */ | |
631 | int j_errno; | |
632 | ||
633 | /* The superblock buffer */ | |
634 | struct buffer_head *j_sb_buffer; | |
635 | journal_superblock_t *j_superblock; | |
636 | ||
637 | /* Version of the superblock format */ | |
638 | int j_format_version; | |
639 | ||
640 | /* | |
641 | * Protect the various scalars in the journal | |
642 | */ | |
643 | spinlock_t j_state_lock; | |
644 | ||
645 | /* | |
646 | * Number of processes waiting to create a barrier lock [j_state_lock] | |
647 | */ | |
648 | int j_barrier_count; | |
649 | ||
650 | /* The barrier lock itself */ | |
2c68ee75 | 651 | struct mutex j_barrier; |
1da177e4 LT |
652 | |
653 | /* | |
654 | * Transactions: The current running transaction... | |
655 | * [j_state_lock] [caller holding open handle] | |
656 | */ | |
657 | transaction_t *j_running_transaction; | |
658 | ||
659 | /* | |
660 | * the transaction we are pushing to disk | |
661 | * [j_state_lock] [caller holding open handle] | |
662 | */ | |
663 | transaction_t *j_committing_transaction; | |
664 | ||
665 | /* | |
666 | * ... and a linked circular list of all transactions waiting for | |
667 | * checkpointing. [j_list_lock] | |
668 | */ | |
669 | transaction_t *j_checkpoint_transactions; | |
670 | ||
671 | /* | |
672 | * Wait queue for waiting for a locked transaction to start committing, | |
673 | * or for a barrier lock to be released | |
674 | */ | |
675 | wait_queue_head_t j_wait_transaction_locked; | |
676 | ||
677 | /* Wait queue for waiting for checkpointing to complete */ | |
678 | wait_queue_head_t j_wait_logspace; | |
679 | ||
680 | /* Wait queue for waiting for commit to complete */ | |
681 | wait_queue_head_t j_wait_done_commit; | |
682 | ||
683 | /* Wait queue to trigger checkpointing */ | |
684 | wait_queue_head_t j_wait_checkpoint; | |
685 | ||
686 | /* Wait queue to trigger commit */ | |
687 | wait_queue_head_t j_wait_commit; | |
688 | ||
689 | /* Wait queue to wait for updates to complete */ | |
690 | wait_queue_head_t j_wait_updates; | |
691 | ||
692 | /* Semaphore for locking against concurrent checkpoints */ | |
2c68ee75 | 693 | struct mutex j_checkpoint_mutex; |
1da177e4 LT |
694 | |
695 | /* | |
696 | * Journal head: identifies the first unused block in the journal. | |
697 | * [j_state_lock] | |
698 | */ | |
699 | unsigned long j_head; | |
700 | ||
701 | /* | |
702 | * Journal tail: identifies the oldest still-used block in the journal. | |
703 | * [j_state_lock] | |
704 | */ | |
705 | unsigned long j_tail; | |
706 | ||
707 | /* | |
708 | * Journal free: how many free blocks are there in the journal? | |
709 | * [j_state_lock] | |
710 | */ | |
711 | unsigned long j_free; | |
712 | ||
713 | /* | |
714 | * Journal start and end: the block numbers of the first usable block | |
715 | * and one beyond the last usable block in the journal. [j_state_lock] | |
716 | */ | |
717 | unsigned long j_first; | |
718 | unsigned long j_last; | |
719 | ||
720 | /* | |
721 | * Device, blocksize and starting block offset for the location where we | |
722 | * store the journal. | |
723 | */ | |
724 | struct block_device *j_dev; | |
725 | int j_blocksize; | |
726 | unsigned int j_blk_offset; | |
727 | ||
728 | /* | |
729 | * Device which holds the client fs. For internal journal this will be | |
730 | * equal to j_dev. | |
731 | */ | |
732 | struct block_device *j_fs_dev; | |
733 | ||
734 | /* Total maximum capacity of the journal region on disk. */ | |
735 | unsigned int j_maxlen; | |
736 | ||
737 | /* | |
738 | * Protects the buffer lists and internal buffer state. | |
739 | */ | |
740 | spinlock_t j_list_lock; | |
741 | ||
742 | /* Optional inode where we store the journal. If present, all */ | |
743 | /* journal block numbers are mapped into this inode via */ | |
744 | /* bmap(). */ | |
745 | struct inode *j_inode; | |
746 | ||
747 | /* | |
748 | * Sequence number of the oldest transaction in the log [j_state_lock] | |
749 | */ | |
750 | tid_t j_tail_sequence; | |
751 | ||
752 | /* | |
753 | * Sequence number of the next transaction to grant [j_state_lock] | |
754 | */ | |
755 | tid_t j_transaction_sequence; | |
756 | ||
757 | /* | |
758 | * Sequence number of the most recently committed transaction | |
759 | * [j_state_lock]. | |
760 | */ | |
761 | tid_t j_commit_sequence; | |
762 | ||
763 | /* | |
764 | * Sequence number of the most recent transaction wanting commit | |
765 | * [j_state_lock] | |
766 | */ | |
767 | tid_t j_commit_request; | |
768 | ||
769 | /* | |
770 | * Journal uuid: identifies the object (filesystem, LVM volume etc) | |
771 | * backed by this journal. This will eventually be replaced by an array | |
772 | * of uuids, allowing us to index multiple devices within a single | |
773 | * journal and to perform atomic updates across them. | |
774 | */ | |
775 | __u8 j_uuid[16]; | |
776 | ||
777 | /* Pointer to the current commit thread for this journal */ | |
778 | struct task_struct *j_task; | |
779 | ||
780 | /* | |
781 | * Maximum number of metadata buffers to allow in a single compound | |
782 | * commit transaction | |
783 | */ | |
784 | int j_max_transaction_buffers; | |
785 | ||
786 | /* | |
787 | * What is the maximum transaction lifetime before we begin a commit? | |
788 | */ | |
789 | unsigned long j_commit_interval; | |
790 | ||
791 | /* The timer used to wakeup the commit thread: */ | |
e3df1898 | 792 | struct timer_list j_commit_timer; |
1da177e4 LT |
793 | |
794 | /* | |
795 | * The revoke table: maintains the list of revoked blocks in the | |
796 | * current transaction. [j_revoke_lock] | |
797 | */ | |
798 | spinlock_t j_revoke_lock; | |
799 | struct jbd_revoke_table_s *j_revoke; | |
800 | struct jbd_revoke_table_s *j_revoke_table[2]; | |
801 | ||
802 | /* | |
803 | * array of bhs for journal_commit_transaction | |
804 | */ | |
805 | struct buffer_head **j_wbuf; | |
806 | int j_wbufsize; | |
807 | ||
fe1dcbc4 AM |
808 | pid_t j_last_sync_writer; |
809 | ||
1da177e4 LT |
810 | /* |
811 | * An opaque pointer to fs-private information. ext3 puts its | |
812 | * superblock pointer here | |
813 | */ | |
814 | void *j_private; | |
815 | }; | |
816 | ||
817 | /* | |
818 | * Journal flag definitions | |
819 | */ | |
820 | #define JFS_UNMOUNT 0x001 /* Journal thread is being destroyed */ | |
821 | #define JFS_ABORT 0x002 /* Journaling has been aborted for errors. */ | |
822 | #define JFS_ACK_ERR 0x004 /* The errno in the sb has been acked */ | |
823 | #define JFS_FLUSHED 0x008 /* The journal superblock has been flushed */ | |
824 | #define JFS_LOADED 0x010 /* The journal superblock has been loaded */ | |
825 | #define JFS_BARRIER 0x020 /* Use IDE barriers */ | |
826 | ||
827 | /* | |
828 | * Function declarations for the journaling transaction and buffer | |
829 | * management | |
830 | */ | |
831 | ||
832 | /* Filing buffers */ | |
833 | extern void __journal_temp_unlink_buffer(struct journal_head *jh); | |
834 | extern void journal_unfile_buffer(journal_t *, struct journal_head *); | |
835 | extern void __journal_unfile_buffer(struct journal_head *); | |
836 | extern void __journal_refile_buffer(struct journal_head *); | |
837 | extern void journal_refile_buffer(journal_t *, struct journal_head *); | |
838 | extern void __journal_file_buffer(struct journal_head *, transaction_t *, int); | |
839 | extern void __journal_free_buffer(struct journal_head *bh); | |
840 | extern void journal_file_buffer(struct journal_head *, transaction_t *, int); | |
841 | extern void __journal_clean_data_list(transaction_t *transaction); | |
842 | ||
843 | /* Log buffer allocation */ | |
844 | extern struct journal_head * journal_get_descriptor_buffer(journal_t *); | |
845 | int journal_next_log_block(journal_t *, unsigned long *); | |
846 | ||
847 | /* Commit management */ | |
848 | extern void journal_commit_transaction(journal_t *); | |
849 | ||
850 | /* Checkpoint list management */ | |
851 | int __journal_clean_checkpoint_list(journal_t *journal); | |
7c8903f6 | 852 | void __journal_remove_checkpoint(struct journal_head *); |
1da177e4 LT |
853 | void __journal_insert_checkpoint(struct journal_head *, transaction_t *); |
854 | ||
855 | /* Buffer IO */ | |
856 | extern int | |
857 | journal_write_metadata_buffer(transaction_t *transaction, | |
858 | struct journal_head *jh_in, | |
859 | struct journal_head **jh_out, | |
860 | int blocknr); | |
861 | ||
862 | /* Transaction locking */ | |
863 | extern void __wait_on_journal (journal_t *); | |
864 | ||
865 | /* | |
866 | * Journal locking. | |
867 | * | |
868 | * We need to lock the journal during transaction state changes so that nobody | |
869 | * ever tries to take a handle on the running transaction while we are in the | |
870 | * middle of moving it to the commit phase. j_state_lock does this. | |
871 | * | |
872 | * Note that the locking is completely interrupt unsafe. We never touch | |
873 | * journal structures from interrupts. | |
874 | */ | |
875 | ||
876 | static inline handle_t *journal_current_handle(void) | |
877 | { | |
878 | return current->journal_info; | |
879 | } | |
880 | ||
881 | /* The journaling code user interface: | |
882 | * | |
883 | * Create and destroy handles | |
884 | * Register buffer modifications against the current transaction. | |
885 | */ | |
886 | ||
887 | extern handle_t *journal_start(journal_t *, int nblocks); | |
888 | extern int journal_restart (handle_t *, int nblocks); | |
889 | extern int journal_extend (handle_t *, int nblocks); | |
890 | extern int journal_get_write_access(handle_t *, struct buffer_head *); | |
891 | extern int journal_get_create_access (handle_t *, struct buffer_head *); | |
892 | extern int journal_get_undo_access(handle_t *, struct buffer_head *); | |
893 | extern int journal_dirty_data (handle_t *, struct buffer_head *); | |
894 | extern int journal_dirty_metadata (handle_t *, struct buffer_head *); | |
895 | extern void journal_release_buffer (handle_t *, struct buffer_head *); | |
896 | extern int journal_forget (handle_t *, struct buffer_head *); | |
897 | extern void journal_sync_buffer (struct buffer_head *); | |
898 | extern int journal_invalidatepage(journal_t *, | |
899 | struct page *, unsigned long); | |
27496a8c | 900 | extern int journal_try_to_free_buffers(journal_t *, struct page *, gfp_t); |
1da177e4 LT |
901 | extern int journal_stop(handle_t *); |
902 | extern int journal_flush (journal_t *); | |
903 | extern void journal_lock_updates (journal_t *); | |
904 | extern void journal_unlock_updates (journal_t *); | |
905 | ||
906 | extern journal_t * journal_init_dev(struct block_device *bdev, | |
907 | struct block_device *fs_dev, | |
908 | int start, int len, int bsize); | |
909 | extern journal_t * journal_init_inode (struct inode *); | |
910 | extern int journal_update_format (journal_t *); | |
911 | extern int journal_check_used_features | |
912 | (journal_t *, unsigned long, unsigned long, unsigned long); | |
913 | extern int journal_check_available_features | |
914 | (journal_t *, unsigned long, unsigned long, unsigned long); | |
915 | extern int journal_set_features | |
916 | (journal_t *, unsigned long, unsigned long, unsigned long); | |
917 | extern int journal_create (journal_t *); | |
918 | extern int journal_load (journal_t *journal); | |
919 | extern void journal_destroy (journal_t *); | |
920 | extern int journal_recover (journal_t *journal); | |
921 | extern int journal_wipe (journal_t *, int); | |
922 | extern int journal_skip_recovery (journal_t *); | |
923 | extern void journal_update_superblock (journal_t *, int); | |
924 | extern void __journal_abort_hard (journal_t *); | |
1da177e4 LT |
925 | extern void journal_abort (journal_t *, int); |
926 | extern int journal_errno (journal_t *); | |
927 | extern void journal_ack_err (journal_t *); | |
928 | extern int journal_clear_err (journal_t *); | |
929 | extern int journal_bmap(journal_t *, unsigned long, unsigned long *); | |
930 | extern int journal_force_commit(journal_t *); | |
931 | ||
932 | /* | |
933 | * journal_head management | |
934 | */ | |
935 | struct journal_head *journal_add_journal_head(struct buffer_head *bh); | |
936 | struct journal_head *journal_grab_journal_head(struct buffer_head *bh); | |
937 | void journal_remove_journal_head(struct buffer_head *bh); | |
938 | void journal_put_journal_head(struct journal_head *jh); | |
939 | ||
940 | /* | |
941 | * handle management | |
942 | */ | |
943 | extern kmem_cache_t *jbd_handle_cache; | |
944 | ||
dd0fc66f | 945 | static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags) |
1da177e4 LT |
946 | { |
947 | return kmem_cache_alloc(jbd_handle_cache, gfp_flags); | |
948 | } | |
949 | ||
950 | static inline void jbd_free_handle(handle_t *handle) | |
951 | { | |
952 | kmem_cache_free(jbd_handle_cache, handle); | |
953 | } | |
954 | ||
955 | /* Primary revoke support */ | |
956 | #define JOURNAL_REVOKE_DEFAULT_HASH 256 | |
957 | extern int journal_init_revoke(journal_t *, int); | |
958 | extern void journal_destroy_revoke_caches(void); | |
959 | extern int journal_init_revoke_caches(void); | |
960 | ||
961 | extern void journal_destroy_revoke(journal_t *); | |
962 | extern int journal_revoke (handle_t *, | |
963 | unsigned long, struct buffer_head *); | |
964 | extern int journal_cancel_revoke(handle_t *, struct journal_head *); | |
965 | extern void journal_write_revoke_records(journal_t *, transaction_t *); | |
966 | ||
967 | /* Recovery revoke support */ | |
968 | extern int journal_set_revoke(journal_t *, unsigned long, tid_t); | |
969 | extern int journal_test_revoke(journal_t *, unsigned long, tid_t); | |
970 | extern void journal_clear_revoke(journal_t *); | |
971 | extern void journal_brelse_array(struct buffer_head *b[], int n); | |
972 | extern void journal_switch_revoke_table(journal_t *journal); | |
973 | ||
974 | /* | |
975 | * The log thread user interface: | |
976 | * | |
977 | * Request space in the current transaction, and force transaction commit | |
978 | * transitions on demand. | |
979 | */ | |
980 | ||
981 | int __log_space_left(journal_t *); /* Called with journal locked */ | |
982 | int log_start_commit(journal_t *journal, tid_t tid); | |
983 | int __log_start_commit(journal_t *journal, tid_t tid); | |
984 | int journal_start_commit(journal_t *journal, tid_t *tid); | |
985 | int journal_force_commit_nested(journal_t *journal); | |
986 | int log_wait_commit(journal_t *journal, tid_t tid); | |
987 | int log_do_checkpoint(journal_t *journal); | |
988 | ||
989 | void __log_wait_for_space(journal_t *journal); | |
990 | extern void __journal_drop_transaction(journal_t *, transaction_t *); | |
991 | extern int cleanup_journal_tail(journal_t *); | |
992 | ||
993 | /* Debugging code only: */ | |
994 | ||
995 | #define jbd_ENOSYS() \ | |
996 | do { \ | |
997 | printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \ | |
998 | current->state = TASK_UNINTERRUPTIBLE; \ | |
999 | schedule(); \ | |
1000 | } while (1) | |
1001 | ||
1002 | /* | |
1003 | * is_journal_abort | |
1004 | * | |
1005 | * Simple test wrapper function to test the JFS_ABORT state flag. This | |
1006 | * bit, when set, indicates that we have had a fatal error somewhere, | |
1007 | * either inside the journaling layer or indicated to us by the client | |
1008 | * (eg. ext3), and that we and should not commit any further | |
1009 | * transactions. | |
1010 | */ | |
1011 | ||
1012 | static inline int is_journal_aborted(journal_t *journal) | |
1013 | { | |
1014 | return journal->j_flags & JFS_ABORT; | |
1015 | } | |
1016 | ||
1017 | static inline int is_handle_aborted(handle_t *handle) | |
1018 | { | |
1019 | if (handle->h_aborted) | |
1020 | return 1; | |
1021 | return is_journal_aborted(handle->h_transaction->t_journal); | |
1022 | } | |
1023 | ||
1024 | static inline void journal_abort_handle(handle_t *handle) | |
1025 | { | |
1026 | handle->h_aborted = 1; | |
1027 | } | |
1028 | ||
1029 | #endif /* __KERNEL__ */ | |
1030 | ||
1031 | /* Comparison functions for transaction IDs: perform comparisons using | |
1032 | * modulo arithmetic so that they work over sequence number wraps. */ | |
1033 | ||
1034 | static inline int tid_gt(tid_t x, tid_t y) | |
1035 | { | |
1036 | int difference = (x - y); | |
1037 | return (difference > 0); | |
1038 | } | |
1039 | ||
1040 | static inline int tid_geq(tid_t x, tid_t y) | |
1041 | { | |
1042 | int difference = (x - y); | |
1043 | return (difference >= 0); | |
1044 | } | |
1045 | ||
1046 | extern int journal_blocks_per_page(struct inode *inode); | |
1047 | ||
1048 | /* | |
1049 | * Return the minimum number of blocks which must be free in the journal | |
1050 | * before a new transaction may be started. Must be called under j_state_lock. | |
1051 | */ | |
1052 | static inline int jbd_space_needed(journal_t *journal) | |
1053 | { | |
1054 | int nblocks = journal->j_max_transaction_buffers; | |
1055 | if (journal->j_committing_transaction) | |
1056 | nblocks += journal->j_committing_transaction-> | |
1057 | t_outstanding_credits; | |
1058 | return nblocks; | |
1059 | } | |
1060 | ||
1061 | /* | |
1062 | * Definitions which augment the buffer_head layer | |
1063 | */ | |
1064 | ||
1065 | /* journaling buffer types */ | |
1066 | #define BJ_None 0 /* Not journaled */ | |
1067 | #define BJ_SyncData 1 /* Normal data: flush before commit */ | |
1068 | #define BJ_Metadata 2 /* Normal journaled metadata */ | |
1069 | #define BJ_Forget 3 /* Buffer superseded by this transaction */ | |
1070 | #define BJ_IO 4 /* Buffer is for temporary IO use */ | |
1071 | #define BJ_Shadow 5 /* Buffer contents being shadowed to the log */ | |
1072 | #define BJ_LogCtl 6 /* Buffer contains log descriptors */ | |
1073 | #define BJ_Reserved 7 /* Buffer is reserved for access by journal */ | |
1074 | #define BJ_Locked 8 /* Locked for I/O during commit */ | |
1075 | #define BJ_Types 9 | |
1076 | ||
1077 | extern int jbd_blocks_per_page(struct inode *inode); | |
1078 | ||
1079 | #ifdef __KERNEL__ | |
1080 | ||
1081 | #define buffer_trace_init(bh) do {} while (0) | |
1082 | #define print_buffer_fields(bh) do {} while (0) | |
1083 | #define print_buffer_trace(bh) do {} while (0) | |
1084 | #define BUFFER_TRACE(bh, info) do {} while (0) | |
1085 | #define BUFFER_TRACE2(bh, bh2, info) do {} while (0) | |
1086 | #define JBUFFER_TRACE(jh, info) do {} while (0) | |
1087 | ||
1088 | #endif /* __KERNEL__ */ | |
1089 | ||
1da177e4 | 1090 | #endif /* _LINUX_JBD_H */ |