]>
Commit | Line | Data |
---|---|---|
470decc6 | 1 | /* |
f7f4bccb | 2 | * linux/include/linux/jbd2.h |
470decc6 DK |
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 | ||
cd02ff0b MC |
16 | #ifndef _LINUX_JBD2_H |
17 | #define _LINUX_JBD2_H | |
470decc6 DK |
18 | |
19 | /* Allow this file to be included directly into e2fsprogs */ | |
20 | #ifndef __KERNEL__ | |
21 | #include "jfs_compat.h" | |
f7f4bccb | 22 | #define JBD2_DEBUG |
470decc6 DK |
23 | #else |
24 | ||
25 | #include <linux/types.h> | |
26 | #include <linux/buffer_head.h> | |
27 | #include <linux/journal-head.h> | |
28 | #include <linux/stddef.h> | |
470decc6 DK |
29 | #include <linux/mutex.h> |
30 | #include <linux/timer.h> | |
5a0e3ad6 | 31 | #include <linux/slab.h> |
01b5adce | 32 | #include <crypto/hash.h> |
470decc6 DK |
33 | #endif |
34 | ||
35 | #define journal_oom_retry 1 | |
36 | ||
37 | /* | |
cd02ff0b | 38 | * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds |
470decc6 | 39 | * certain classes of error which can occur due to failed IOs. Under |
cd02ff0b | 40 | * normal use we want ext4 to continue after such errors, because |
470decc6 DK |
41 | * hardware _can_ fail, but for debugging purposes when running tests on |
42 | * known-good hardware we may want to trap these errors. | |
43 | */ | |
cd02ff0b | 44 | #undef JBD2_PARANOID_IOFAIL |
470decc6 DK |
45 | |
46 | /* | |
47 | * The default maximum commit age, in seconds. | |
48 | */ | |
cd02ff0b | 49 | #define JBD2_DEFAULT_MAX_COMMIT_AGE 5 |
470decc6 | 50 | |
e23291b9 | 51 | #ifdef CONFIG_JBD2_DEBUG |
470decc6 | 52 | /* |
cd02ff0b | 53 | * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal |
470decc6 | 54 | * consistency checks. By default we don't do this unless |
e23291b9 | 55 | * CONFIG_JBD2_DEBUG is on. |
470decc6 | 56 | */ |
cd02ff0b | 57 | #define JBD2_EXPENSIVE_CHECKING |
b6e96d00 | 58 | extern ushort jbd2_journal_enable_debug; |
169f1a2a PG |
59 | void __jbd2_debug(int level, const char *file, const char *func, |
60 | unsigned int line, const char *fmt, ...); | |
470decc6 | 61 | |
169f1a2a PG |
62 | #define jbd_debug(n, fmt, a...) \ |
63 | __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a) | |
470decc6 | 64 | #else |
169f1a2a | 65 | #define jbd_debug(n, fmt, a...) /**/ |
470decc6 DK |
66 | #endif |
67 | ||
d2eecb03 TT |
68 | extern void *jbd2_alloc(size_t size, gfp_t flags); |
69 | extern void jbd2_free(void *ptr, size_t size); | |
af1e76d6 | 70 | |
f7f4bccb | 71 | #define JBD2_MIN_JOURNAL_BLOCKS 1024 |
470decc6 DK |
72 | |
73 | #ifdef __KERNEL__ | |
74 | ||
75 | /** | |
76 | * typedef handle_t - The handle_t type represents a single atomic update being performed by some process. | |
77 | * | |
78 | * All filesystem modifications made by the process go | |
79 | * through this handle. Recursive operations (such as quota operations) | |
80 | * are gathered into a single update. | |
81 | * | |
82 | * The buffer credits field is used to account for journaled buffers | |
83 | * being modified by the running process. To ensure that there is | |
84 | * enough log space for all outstanding operations, we need to limit the | |
85 | * number of outstanding buffers possible at any time. When the | |
86 | * operation completes, any buffer credits not used are credited back to | |
87 | * the transaction, so that at all times we know how many buffers the | |
88 | * outstanding updates on a transaction might possibly touch. | |
89 | * | |
90 | * This is an opaque datatype. | |
91 | **/ | |
8aefcd55 | 92 | typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */ |
470decc6 DK |
93 | |
94 | ||
95 | /** | |
96 | * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem. | |
97 | * | |
98 | * journal_t is linked to from the fs superblock structure. | |
99 | * | |
100 | * We use the journal_t to keep track of all outstanding transaction | |
101 | * activity on the filesystem, and to manage the state of the log | |
102 | * writing process. | |
103 | * | |
104 | * This is an opaque datatype. | |
105 | **/ | |
106 | typedef struct journal_s journal_t; /* Journal control structure */ | |
107 | #endif | |
108 | ||
109 | /* | |
110 | * Internal structures used by the logging mechanism: | |
111 | */ | |
112 | ||
f7f4bccb | 113 | #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */ |
470decc6 DK |
114 | |
115 | /* | |
116 | * On-disk structures | |
117 | */ | |
118 | ||
119 | /* | |
120 | * Descriptor block types: | |
121 | */ | |
122 | ||
f7f4bccb MC |
123 | #define JBD2_DESCRIPTOR_BLOCK 1 |
124 | #define JBD2_COMMIT_BLOCK 2 | |
125 | #define JBD2_SUPERBLOCK_V1 3 | |
126 | #define JBD2_SUPERBLOCK_V2 4 | |
127 | #define JBD2_REVOKE_BLOCK 5 | |
470decc6 DK |
128 | |
129 | /* | |
130 | * Standard header for all descriptor blocks: | |
131 | */ | |
132 | typedef struct journal_header_s | |
133 | { | |
134 | __be32 h_magic; | |
135 | __be32 h_blocktype; | |
136 | __be32 h_sequence; | |
137 | } journal_header_t; | |
138 | ||
818d276c GS |
139 | /* |
140 | * Checksum types. | |
141 | */ | |
142 | #define JBD2_CRC32_CHKSUM 1 | |
143 | #define JBD2_MD5_CHKSUM 2 | |
144 | #define JBD2_SHA1_CHKSUM 3 | |
8f888ef8 | 145 | #define JBD2_CRC32C_CHKSUM 4 |
818d276c GS |
146 | |
147 | #define JBD2_CRC32_CHKSUM_SIZE 4 | |
148 | ||
149 | #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32)) | |
150 | /* | |
151 | * Commit block header for storing transactional checksums: | |
8f888ef8 DW |
152 | * |
153 | * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum* | |
154 | * fields are used to store a checksum of the descriptor and data blocks. | |
155 | * | |
156 | * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum | |
157 | * field is used to store crc32c(uuid+commit_block). Each journal metadata | |
158 | * block gets its own checksum, and data block checksums are stored in | |
159 | * journal_block_tag (in the descriptor). The other h_chksum* fields are | |
160 | * not used. | |
161 | * | |
db9ee220 DW |
162 | * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses |
163 | * journal_block_tag3_t to store a full 32-bit checksum. Everything else | |
164 | * is the same as v2. | |
165 | * | |
166 | * Checksum v1, v2, and v3 are mutually exclusive features. | |
818d276c GS |
167 | */ |
168 | struct commit_header { | |
169 | __be32 h_magic; | |
170 | __be32 h_blocktype; | |
171 | __be32 h_sequence; | |
172 | unsigned char h_chksum_type; | |
173 | unsigned char h_chksum_size; | |
174 | unsigned char h_padding[2]; | |
175 | __be32 h_chksum[JBD2_CHECKSUM_BYTES]; | |
736603ab TT |
176 | __be64 h_commit_sec; |
177 | __be32 h_commit_nsec; | |
818d276c | 178 | }; |
470decc6 DK |
179 | |
180 | /* | |
b517bea1 ZB |
181 | * The block tag: used to describe a single buffer in the journal. |
182 | * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this | |
183 | * raw struct shouldn't be used for pointer math or sizeof() - use | |
184 | * journal_tag_bytes(journal) instead to compute this. | |
470decc6 | 185 | */ |
db9ee220 DW |
186 | typedef struct journal_block_tag3_s |
187 | { | |
188 | __be32 t_blocknr; /* The on-disk block number */ | |
189 | __be32 t_flags; /* See below */ | |
190 | __be32 t_blocknr_high; /* most-significant high 32bits. */ | |
191 | __be32 t_checksum; /* crc32c(uuid+seq+block) */ | |
192 | } journal_block_tag3_t; | |
193 | ||
470decc6 DK |
194 | typedef struct journal_block_tag_s |
195 | { | |
196 | __be32 t_blocknr; /* The on-disk block number */ | |
8f888ef8 DW |
197 | __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */ |
198 | __be16 t_flags; /* See below */ | |
b517bea1 | 199 | __be32 t_blocknr_high; /* most-significant high 32bits. */ |
470decc6 DK |
200 | } journal_block_tag_t; |
201 | ||
8f888ef8 DW |
202 | /* Tail of descriptor block, for checksumming */ |
203 | struct jbd2_journal_block_tail { | |
204 | __be32 t_checksum; /* crc32c(uuid+descr_block) */ | |
205 | }; | |
206 | ||
470decc6 DK |
207 | /* |
208 | * The revoke descriptor: used on disk to describe a series of blocks to | |
209 | * be revoked from the log | |
210 | */ | |
f7f4bccb | 211 | typedef struct jbd2_journal_revoke_header_s |
470decc6 DK |
212 | { |
213 | journal_header_t r_header; | |
214 | __be32 r_count; /* Count of bytes used in the block */ | |
f7f4bccb | 215 | } jbd2_journal_revoke_header_t; |
470decc6 | 216 | |
8f888ef8 DW |
217 | /* Tail of revoke block, for checksumming */ |
218 | struct jbd2_journal_revoke_tail { | |
219 | __be32 r_checksum; /* crc32c(uuid+revoke_block) */ | |
220 | }; | |
470decc6 DK |
221 | |
222 | /* Definitions for the journal tag flags word: */ | |
f7f4bccb MC |
223 | #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */ |
224 | #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */ | |
225 | #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */ | |
226 | #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */ | |
470decc6 DK |
227 | |
228 | ||
229 | /* | |
230 | * The journal superblock. All fields are in big-endian byte order. | |
231 | */ | |
232 | typedef struct journal_superblock_s | |
233 | { | |
234 | /* 0x0000 */ | |
235 | journal_header_t s_header; | |
236 | ||
237 | /* 0x000C */ | |
238 | /* Static information describing the journal */ | |
239 | __be32 s_blocksize; /* journal device blocksize */ | |
240 | __be32 s_maxlen; /* total blocks in journal file */ | |
241 | __be32 s_first; /* first block of log information */ | |
242 | ||
243 | /* 0x0018 */ | |
244 | /* Dynamic information describing the current state of the log */ | |
245 | __be32 s_sequence; /* first commit ID expected in log */ | |
246 | __be32 s_start; /* blocknr of start of log */ | |
247 | ||
248 | /* 0x0020 */ | |
f7f4bccb | 249 | /* Error value, as set by jbd2_journal_abort(). */ |
470decc6 DK |
250 | __be32 s_errno; |
251 | ||
252 | /* 0x0024 */ | |
253 | /* Remaining fields are only valid in a version-2 superblock */ | |
254 | __be32 s_feature_compat; /* compatible feature set */ | |
255 | __be32 s_feature_incompat; /* incompatible feature set */ | |
256 | __be32 s_feature_ro_compat; /* readonly-compatible feature set */ | |
257 | /* 0x0030 */ | |
258 | __u8 s_uuid[16]; /* 128-bit uuid for journal */ | |
259 | ||
260 | /* 0x0040 */ | |
261 | __be32 s_nr_users; /* Nr of filesystems sharing log */ | |
262 | ||
263 | __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/ | |
264 | ||
265 | /* 0x0048 */ | |
266 | __be32 s_max_transaction; /* Limit of journal blocks per trans.*/ | |
267 | __be32 s_max_trans_data; /* Limit of data blocks per trans. */ | |
268 | ||
269 | /* 0x0050 */ | |
8f888ef8 DW |
270 | __u8 s_checksum_type; /* checksum type */ |
271 | __u8 s_padding2[3]; | |
272 | __u32 s_padding[42]; | |
273 | __be32 s_checksum; /* crc32c(superblock) */ | |
470decc6 DK |
274 | |
275 | /* 0x0100 */ | |
276 | __u8 s_users[16*48]; /* ids of all fs'es sharing the log */ | |
277 | /* 0x0400 */ | |
278 | } journal_superblock_t; | |
279 | ||
f7f4bccb | 280 | #define JBD2_HAS_COMPAT_FEATURE(j,mask) \ |
470decc6 DK |
281 | ((j)->j_format_version >= 2 && \ |
282 | ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask)))) | |
f7f4bccb | 283 | #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \ |
470decc6 DK |
284 | ((j)->j_format_version >= 2 && \ |
285 | ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask)))) | |
f7f4bccb | 286 | #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \ |
470decc6 DK |
287 | ((j)->j_format_version >= 2 && \ |
288 | ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask)))) | |
289 | ||
818d276c GS |
290 | #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001 |
291 | ||
292 | #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001 | |
293 | #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002 | |
294 | #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004 | |
8f888ef8 | 295 | #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008 |
db9ee220 | 296 | #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010 |
470decc6 DK |
297 | |
298 | /* Features known to this kernel version: */ | |
818d276c | 299 | #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM |
f7f4bccb | 300 | #define JBD2_KNOWN_ROCOMPAT_FEATURES 0 |
b517bea1 | 301 | #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \ |
818d276c | 302 | JBD2_FEATURE_INCOMPAT_64BIT | \ |
e93376c2 | 303 | JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \ |
db9ee220 DW |
304 | JBD2_FEATURE_INCOMPAT_CSUM_V2 | \ |
305 | JBD2_FEATURE_INCOMPAT_CSUM_V3) | |
470decc6 DK |
306 | |
307 | #ifdef __KERNEL__ | |
308 | ||
309 | #include <linux/fs.h> | |
310 | #include <linux/sched.h> | |
b34090e5 JK |
311 | |
312 | enum jbd_state_bits { | |
313 | BH_JBD /* Has an attached ext3 journal_head */ | |
314 | = BH_PrivateStart, | |
315 | BH_JWrite, /* Being written to log (@@@ DEBUGGING) */ | |
316 | BH_Freed, /* Has been freed (truncated) */ | |
317 | BH_Revoked, /* Has been revoked from the log */ | |
318 | BH_RevokeValid, /* Revoked flag is valid */ | |
319 | BH_JBDDirty, /* Is dirty but journaled */ | |
320 | BH_State, /* Pins most journal_head state */ | |
321 | BH_JournalHead, /* Pins bh->b_private and jh->b_bh */ | |
322 | BH_Shadow, /* IO on shadow buffer is running */ | |
323 | BH_Verified, /* Metadata block has been verified ok */ | |
324 | BH_JBDPrivateStart, /* First bit available for private use by FS */ | |
325 | }; | |
326 | ||
327 | BUFFER_FNS(JBD, jbd) | |
328 | BUFFER_FNS(JWrite, jwrite) | |
329 | BUFFER_FNS(JBDDirty, jbddirty) | |
330 | TAS_BUFFER_FNS(JBDDirty, jbddirty) | |
331 | BUFFER_FNS(Revoked, revoked) | |
332 | TAS_BUFFER_FNS(Revoked, revoked) | |
333 | BUFFER_FNS(RevokeValid, revokevalid) | |
334 | TAS_BUFFER_FNS(RevokeValid, revokevalid) | |
335 | BUFFER_FNS(Freed, freed) | |
336 | BUFFER_FNS(Shadow, shadow) | |
337 | BUFFER_FNS(Verified, verified) | |
338 | ||
44606672 | 339 | #include <linux/jbd_common.h> |
470decc6 | 340 | |
36df53f4 | 341 | #define J_ASSERT(assert) BUG_ON(!(assert)) |
470decc6 | 342 | |
470decc6 DK |
343 | #define J_ASSERT_BH(bh, expr) J_ASSERT(expr) |
344 | #define J_ASSERT_JH(jh, expr) J_ASSERT(expr) | |
470decc6 | 345 | |
cd02ff0b | 346 | #if defined(JBD2_PARANOID_IOFAIL) |
470decc6 DK |
347 | #define J_EXPECT(expr, why...) J_ASSERT(expr) |
348 | #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr) | |
349 | #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr) | |
350 | #else | |
351 | #define __journal_expect(expr, why...) \ | |
352 | ({ \ | |
353 | int val = (expr); \ | |
354 | if (!val) { \ | |
355 | printk(KERN_ERR \ | |
08ec8c38 TT |
356 | "JBD2 unexpected failure: %s: %s;\n", \ |
357 | __func__, #expr); \ | |
470decc6 DK |
358 | printk(KERN_ERR why "\n"); \ |
359 | } \ | |
360 | val; \ | |
361 | }) | |
362 | #define J_EXPECT(expr, why...) __journal_expect(expr, ## why) | |
363 | #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why) | |
364 | #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why) | |
365 | #endif | |
366 | ||
c851ed54 JK |
367 | /* Flags in jbd_inode->i_flags */ |
368 | #define __JI_COMMIT_RUNNING 0 | |
369 | /* Commit of the inode data in progress. We use this flag to protect us from | |
370 | * concurrent deletion of inode. We cannot use reference to inode for this | |
371 | * since we cannot afford doing last iput() on behalf of kjournald | |
372 | */ | |
373 | #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING) | |
374 | ||
375 | /** | |
376 | * struct jbd_inode is the structure linking inodes in ordered mode | |
377 | * present in a transaction so that we can sync them during commit. | |
378 | */ | |
379 | struct jbd2_inode { | |
380 | /* Which transaction does this inode belong to? Either the running | |
381 | * transaction or the committing one. [j_list_lock] */ | |
382 | transaction_t *i_transaction; | |
383 | ||
384 | /* Pointer to the running transaction modifying inode's data in case | |
385 | * there is already a committing transaction touching it. [j_list_lock] */ | |
386 | transaction_t *i_next_transaction; | |
387 | ||
388 | /* List of inodes in the i_transaction [j_list_lock] */ | |
389 | struct list_head i_list; | |
390 | ||
391 | /* VFS inode this inode belongs to [constant during the lifetime | |
392 | * of the structure] */ | |
393 | struct inode *i_vfs_inode; | |
394 | ||
395 | /* Flags of inode [j_list_lock] */ | |
39e3ac25 | 396 | unsigned long i_flags; |
c851ed54 JK |
397 | }; |
398 | ||
f7f4bccb | 399 | struct jbd2_revoke_table_s; |
470decc6 DK |
400 | |
401 | /** | |
402 | * struct handle_s - The handle_s type is the concrete type associated with | |
403 | * handle_t. | |
404 | * @h_transaction: Which compound transaction is this update a part of? | |
405 | * @h_buffer_credits: Number of remaining buffers we are allowed to dirty. | |
406 | * @h_ref: Reference count on this handle | |
407 | * @h_err: Field for caller's use to track errors through large fs operations | |
408 | * @h_sync: flag for sync-on-close | |
409 | * @h_jdata: flag to force data journaling | |
410 | * @h_aborted: flag indicating fatal error on handle | |
411 | **/ | |
412 | ||
413 | /* Docbook can't yet cope with the bit fields, but will leave the documentation | |
414 | * in so it can be fixed later. | |
415 | */ | |
416 | ||
8aefcd55 | 417 | struct jbd2_journal_handle |
470decc6 | 418 | { |
8f7d89f3 JK |
419 | union { |
420 | /* Which compound transaction is this update a part of? */ | |
421 | transaction_t *h_transaction; | |
422 | /* Which journal handle belongs to - used iff h_reserved set */ | |
423 | journal_t *h_journal; | |
424 | }; | |
425 | ||
426 | /* Handle reserved for finishing the logical operation */ | |
427 | handle_t *h_rsv_handle; | |
470decc6 DK |
428 | |
429 | /* Number of remaining buffers we are allowed to dirty: */ | |
430 | int h_buffer_credits; | |
431 | ||
432 | /* Reference count on this handle */ | |
433 | int h_ref; | |
434 | ||
435 | /* Field for caller's use to track errors through large fs */ | |
436 | /* operations */ | |
437 | int h_err; | |
438 | ||
439 | /* Flags [no locking] */ | |
078d5039 TT |
440 | unsigned int h_sync: 1; /* sync-on-close */ |
441 | unsigned int h_jdata: 1; /* force data journaling */ | |
8f7d89f3 | 442 | unsigned int h_reserved: 1; /* handle with reserved credits */ |
078d5039 | 443 | unsigned int h_aborted: 1; /* fatal error on handle */ |
343d9c28 TT |
444 | unsigned int h_type: 8; /* for handle statistics */ |
445 | unsigned int h_line_no: 16; /* for handle statistics */ | |
446 | ||
447 | unsigned long h_start_jiffies; | |
448 | unsigned int h_requested_credits; | |
7b751066 MC |
449 | |
450 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
451 | struct lockdep_map h_lockdep_map; | |
452 | #endif | |
470decc6 DK |
453 | }; |
454 | ||
455 | ||
8e85fb3f JL |
456 | /* |
457 | * Some stats for checkpoint phase | |
458 | */ | |
459 | struct transaction_chp_stats_s { | |
460 | unsigned long cs_chp_time; | |
bf699327 TT |
461 | __u32 cs_forced_to_close; |
462 | __u32 cs_written; | |
463 | __u32 cs_dropped; | |
8e85fb3f JL |
464 | }; |
465 | ||
470decc6 DK |
466 | /* The transaction_t type is the guts of the journaling mechanism. It |
467 | * tracks a compound transaction through its various states: | |
468 | * | |
469 | * RUNNING: accepting new updates | |
470 | * LOCKED: Updates still running but we don't accept new ones | |
471 | * RUNDOWN: Updates are tidying up but have finished requesting | |
472 | * new buffers to modify (state not used for now) | |
473 | * FLUSH: All updates complete, but we are still writing to disk | |
474 | * COMMIT: All data on disk, writing commit record | |
475 | * FINISHED: We still have to keep the transaction for checkpointing. | |
476 | * | |
477 | * The transaction keeps track of all of the buffers modified by a | |
478 | * running transaction, and all of the buffers committed but not yet | |
479 | * flushed to home for finished transactions. | |
480 | */ | |
481 | ||
482 | /* | |
483 | * Lock ranking: | |
484 | * | |
485 | * j_list_lock | |
486 | * ->jbd_lock_bh_journal_head() (This is "innermost") | |
487 | * | |
488 | * j_state_lock | |
489 | * ->jbd_lock_bh_state() | |
490 | * | |
491 | * jbd_lock_bh_state() | |
492 | * ->j_list_lock | |
493 | * | |
494 | * j_state_lock | |
495 | * ->t_handle_lock | |
496 | * | |
497 | * j_state_lock | |
498 | * ->j_list_lock (journal_unmap_buffer) | |
499 | * | |
500 | */ | |
501 | ||
502 | struct transaction_s | |
503 | { | |
504 | /* Pointer to the journal for this transaction. [no locking] */ | |
505 | journal_t *t_journal; | |
506 | ||
507 | /* Sequence number for this transaction [no locking] */ | |
508 | tid_t t_tid; | |
509 | ||
510 | /* | |
511 | * Transaction's current state | |
f7f4bccb | 512 | * [no locking - only kjournald2 alters this] |
f5a7a6b0 JK |
513 | * [j_list_lock] guards transition of a transaction into T_FINISHED |
514 | * state and subsequent call of __jbd2_journal_drop_transaction() | |
470decc6 DK |
515 | * FIXME: needs barriers |
516 | * KLUDGE: [use j_state_lock] | |
517 | */ | |
518 | enum { | |
519 | T_RUNNING, | |
520 | T_LOCKED, | |
470decc6 DK |
521 | T_FLUSH, |
522 | T_COMMIT, | |
bbd2be36 JK |
523 | T_COMMIT_DFLUSH, |
524 | T_COMMIT_JFLUSH, | |
794446c6 | 525 | T_COMMIT_CALLBACK, |
470decc6 DK |
526 | T_FINISHED |
527 | } t_state; | |
528 | ||
529 | /* | |
530 | * Where in the log does this transaction's commit start? [no locking] | |
531 | */ | |
532 | unsigned long t_log_start; | |
533 | ||
534 | /* Number of buffers on the t_buffers list [j_list_lock] */ | |
535 | int t_nr_buffers; | |
536 | ||
537 | /* | |
538 | * Doubly-linked circular list of all buffers reserved but not yet | |
539 | * modified by this transaction [j_list_lock] | |
540 | */ | |
541 | struct journal_head *t_reserved_list; | |
542 | ||
470decc6 DK |
543 | /* |
544 | * Doubly-linked circular list of all metadata buffers owned by this | |
545 | * transaction [j_list_lock] | |
546 | */ | |
547 | struct journal_head *t_buffers; | |
548 | ||
470decc6 DK |
549 | /* |
550 | * Doubly-linked circular list of all forget buffers (superseded | |
551 | * buffers which we can un-checkpoint once this transaction commits) | |
552 | * [j_list_lock] | |
553 | */ | |
554 | struct journal_head *t_forget; | |
555 | ||
556 | /* | |
557 | * Doubly-linked circular list of all buffers still to be flushed before | |
558 | * this transaction can be checkpointed. [j_list_lock] | |
559 | */ | |
560 | struct journal_head *t_checkpoint_list; | |
561 | ||
562 | /* | |
563 | * Doubly-linked circular list of all buffers submitted for IO while | |
564 | * checkpointing. [j_list_lock] | |
565 | */ | |
566 | struct journal_head *t_checkpoint_io_list; | |
567 | ||
470decc6 DK |
568 | /* |
569 | * Doubly-linked circular list of metadata buffers being shadowed by log | |
570 | * IO. The IO buffers on the iobuf list and the shadow buffers on this | |
571 | * list match each other one for one at all times. [j_list_lock] | |
572 | */ | |
573 | struct journal_head *t_shadow_list; | |
574 | ||
c851ed54 JK |
575 | /* |
576 | * List of inodes whose data we've modified in data=ordered mode. | |
577 | * [j_list_lock] | |
578 | */ | |
579 | struct list_head t_inode_list; | |
580 | ||
470decc6 DK |
581 | /* |
582 | * Protects info related to handles | |
583 | */ | |
584 | spinlock_t t_handle_lock; | |
585 | ||
8e85fb3f JL |
586 | /* |
587 | * Longest time some handle had to wait for running transaction | |
588 | */ | |
589 | unsigned long t_max_wait; | |
590 | ||
591 | /* | |
592 | * When transaction started | |
593 | */ | |
594 | unsigned long t_start; | |
595 | ||
9fff24aa TT |
596 | /* |
597 | * When commit was requested | |
598 | */ | |
599 | unsigned long t_requested; | |
600 | ||
8e85fb3f JL |
601 | /* |
602 | * Checkpointing stats [j_checkpoint_sem] | |
603 | */ | |
604 | struct transaction_chp_stats_s t_chp_stats; | |
605 | ||
470decc6 DK |
606 | /* |
607 | * Number of outstanding updates running on this transaction | |
608 | * [t_handle_lock] | |
609 | */ | |
a51dca9c | 610 | atomic_t t_updates; |
470decc6 DK |
611 | |
612 | /* | |
613 | * Number of buffers reserved for use by all handles in this transaction | |
614 | * handle but not yet modified. [t_handle_lock] | |
615 | */ | |
a51dca9c | 616 | atomic_t t_outstanding_credits; |
470decc6 DK |
617 | |
618 | /* | |
619 | * Forward and backward links for the circular list of all transactions | |
620 | * awaiting checkpoint. [j_list_lock] | |
621 | */ | |
622 | transaction_t *t_cpnext, *t_cpprev; | |
623 | ||
624 | /* | |
625 | * When will the transaction expire (become due for commit), in jiffies? | |
626 | * [no locking] | |
627 | */ | |
628 | unsigned long t_expires; | |
629 | ||
e07f7183 JB |
630 | /* |
631 | * When this transaction started, in nanoseconds [no locking] | |
632 | */ | |
633 | ktime_t t_start_time; | |
634 | ||
470decc6 DK |
635 | /* |
636 | * How many handles used this transaction? [t_handle_lock] | |
637 | */ | |
8dd42046 | 638 | atomic_t t_handle_count; |
470decc6 | 639 | |
7058548c TT |
640 | /* |
641 | * This transaction is being forced and some process is | |
642 | * waiting for it to finish. | |
643 | */ | |
0ccff1a4 | 644 | unsigned int t_synchronous_commit:1; |
81be12c8 JK |
645 | |
646 | /* Disk flush needs to be sent to fs partition [no locking] */ | |
647 | int t_need_data_flush; | |
7058548c | 648 | |
3e624fc7 TT |
649 | /* |
650 | * For use by the filesystem to store fs-specific data | |
651 | * structures associated with the transaction | |
652 | */ | |
653 | struct list_head t_private_list; | |
470decc6 DK |
654 | }; |
655 | ||
8e85fb3f JL |
656 | struct transaction_run_stats_s { |
657 | unsigned long rs_wait; | |
9fff24aa | 658 | unsigned long rs_request_delay; |
8e85fb3f JL |
659 | unsigned long rs_running; |
660 | unsigned long rs_locked; | |
661 | unsigned long rs_flushing; | |
662 | unsigned long rs_logging; | |
663 | ||
bf699327 TT |
664 | __u32 rs_handle_count; |
665 | __u32 rs_blocks; | |
666 | __u32 rs_blocks_logged; | |
8e85fb3f JL |
667 | }; |
668 | ||
669 | struct transaction_stats_s { | |
8e85fb3f | 670 | unsigned long ts_tid; |
9fff24aa | 671 | unsigned long ts_requested; |
bf699327 | 672 | struct transaction_run_stats_s run; |
8e85fb3f JL |
673 | }; |
674 | ||
8e85fb3f JL |
675 | static inline unsigned long |
676 | jbd2_time_diff(unsigned long start, unsigned long end) | |
677 | { | |
678 | if (end >= start) | |
679 | return end - start; | |
680 | ||
681 | return end + (MAX_JIFFY_OFFSET - start); | |
682 | } | |
683 | ||
1a0d3786 TT |
684 | #define JBD2_NR_BATCH 64 |
685 | ||
470decc6 DK |
686 | /** |
687 | * struct journal_s - The journal_s type is the concrete type associated with | |
688 | * journal_t. | |
689 | * @j_flags: General journaling state flags | |
690 | * @j_errno: Is there an outstanding uncleared error on the journal (from a | |
691 | * prior abort)? | |
692 | * @j_sb_buffer: First part of superblock buffer | |
693 | * @j_superblock: Second part of superblock buffer | |
694 | * @j_format_version: Version of the superblock format | |
695 | * @j_state_lock: Protect the various scalars in the journal | |
696 | * @j_barrier_count: Number of processes waiting to create a barrier lock | |
697 | * @j_barrier: The barrier lock itself | |
698 | * @j_running_transaction: The current running transaction.. | |
699 | * @j_committing_transaction: the transaction we are pushing to disk | |
700 | * @j_checkpoint_transactions: a linked circular list of all transactions | |
701 | * waiting for checkpointing | |
702 | * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction | |
703 | * to start committing, or for a barrier lock to be released | |
470decc6 | 704 | * @j_wait_done_commit: Wait queue for waiting for commit to complete |
470decc6 DK |
705 | * @j_wait_commit: Wait queue to trigger commit |
706 | * @j_wait_updates: Wait queue to wait for updates to complete | |
8f7d89f3 | 707 | * @j_wait_reserved: Wait queue to wait for reserved buffer credits to drop |
470decc6 DK |
708 | * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints |
709 | * @j_head: Journal head - identifies the first unused block in the journal | |
710 | * @j_tail: Journal tail - identifies the oldest still-used block in the | |
711 | * journal. | |
712 | * @j_free: Journal free - how many free blocks are there in the journal? | |
713 | * @j_first: The block number of the first usable block | |
714 | * @j_last: The block number one beyond the last usable block | |
715 | * @j_dev: Device where we store the journal | |
716 | * @j_blocksize: blocksize for the location where we store the journal. | |
717 | * @j_blk_offset: starting block offset for into the device where we store the | |
718 | * journal | |
719 | * @j_fs_dev: Device which holds the client fs. For internal journal this will | |
720 | * be equal to j_dev | |
8f7d89f3 | 721 | * @j_reserved_credits: Number of buffers reserved from the running transaction |
470decc6 DK |
722 | * @j_maxlen: Total maximum capacity of the journal region on disk. |
723 | * @j_list_lock: Protects the buffer lists and internal buffer state. | |
724 | * @j_inode: Optional inode where we store the journal. If present, all journal | |
725 | * block numbers are mapped into this inode via bmap(). | |
726 | * @j_tail_sequence: Sequence number of the oldest transaction in the log | |
727 | * @j_transaction_sequence: Sequence number of the next transaction to grant | |
728 | * @j_commit_sequence: Sequence number of the most recently committed | |
729 | * transaction | |
730 | * @j_commit_request: Sequence number of the most recent transaction wanting | |
731 | * commit | |
732 | * @j_uuid: Uuid of client object. | |
733 | * @j_task: Pointer to the current commit thread for this journal | |
734 | * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a | |
735 | * single compound commit transaction | |
736 | * @j_commit_interval: What is the maximum transaction lifetime before we begin | |
737 | * a commit? | |
738 | * @j_commit_timer: The timer used to wakeup the commit thread | |
739 | * @j_revoke_lock: Protect the revoke table | |
740 | * @j_revoke: The revoke table - maintains the list of revoked blocks in the | |
741 | * current transaction. | |
742 | * @j_revoke_table: alternate revoke tables for j_revoke | |
f7f4bccb | 743 | * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction |
470decc6 DK |
744 | * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the |
745 | * number that will fit in j_blocksize | |
746 | * @j_last_sync_writer: most recent pid which did a synchronous write | |
8e85fb3f JL |
747 | * @j_history: Buffer storing the transactions statistics history |
748 | * @j_history_max: Maximum number of transactions in the statistics history | |
749 | * @j_history_cur: Current number of transactions in the statistics history | |
750 | * @j_history_lock: Protect the transactions statistics history | |
751 | * @j_proc_entry: procfs entry for the jbd statistics directory | |
752 | * @j_stats: Overall statistics | |
470decc6 DK |
753 | * @j_private: An opaque pointer to fs-private information. |
754 | */ | |
755 | ||
756 | struct journal_s | |
757 | { | |
758 | /* General journaling state flags [j_state_lock] */ | |
759 | unsigned long j_flags; | |
760 | ||
761 | /* | |
762 | * Is there an outstanding uncleared error on the journal (from a prior | |
763 | * abort)? [j_state_lock] | |
764 | */ | |
765 | int j_errno; | |
766 | ||
767 | /* The superblock buffer */ | |
768 | struct buffer_head *j_sb_buffer; | |
769 | journal_superblock_t *j_superblock; | |
770 | ||
771 | /* Version of the superblock format */ | |
772 | int j_format_version; | |
773 | ||
774 | /* | |
775 | * Protect the various scalars in the journal | |
776 | */ | |
a931da6a | 777 | rwlock_t j_state_lock; |
470decc6 DK |
778 | |
779 | /* | |
780 | * Number of processes waiting to create a barrier lock [j_state_lock] | |
781 | */ | |
782 | int j_barrier_count; | |
783 | ||
784 | /* The barrier lock itself */ | |
785 | struct mutex j_barrier; | |
786 | ||
787 | /* | |
788 | * Transactions: The current running transaction... | |
789 | * [j_state_lock] [caller holding open handle] | |
790 | */ | |
791 | transaction_t *j_running_transaction; | |
792 | ||
793 | /* | |
794 | * the transaction we are pushing to disk | |
795 | * [j_state_lock] [caller holding open handle] | |
796 | */ | |
797 | transaction_t *j_committing_transaction; | |
798 | ||
799 | /* | |
800 | * ... and a linked circular list of all transactions waiting for | |
801 | * checkpointing. [j_list_lock] | |
802 | */ | |
803 | transaction_t *j_checkpoint_transactions; | |
804 | ||
805 | /* | |
806 | * Wait queue for waiting for a locked transaction to start committing, | |
807 | * or for a barrier lock to be released | |
808 | */ | |
809 | wait_queue_head_t j_wait_transaction_locked; | |
810 | ||
470decc6 DK |
811 | /* Wait queue for waiting for commit to complete */ |
812 | wait_queue_head_t j_wait_done_commit; | |
813 | ||
470decc6 DK |
814 | /* Wait queue to trigger commit */ |
815 | wait_queue_head_t j_wait_commit; | |
816 | ||
817 | /* Wait queue to wait for updates to complete */ | |
818 | wait_queue_head_t j_wait_updates; | |
819 | ||
8f7d89f3 JK |
820 | /* Wait queue to wait for reserved buffer credits to drop */ |
821 | wait_queue_head_t j_wait_reserved; | |
822 | ||
470decc6 DK |
823 | /* Semaphore for locking against concurrent checkpoints */ |
824 | struct mutex j_checkpoint_mutex; | |
825 | ||
1a0d3786 TT |
826 | /* |
827 | * List of buffer heads used by the checkpoint routine. This | |
828 | * was moved from jbd2_log_do_checkpoint() to reduce stack | |
829 | * usage. Access to this array is controlled by the | |
830 | * j_checkpoint_mutex. [j_checkpoint_mutex] | |
831 | */ | |
832 | struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH]; | |
833 | ||
470decc6 DK |
834 | /* |
835 | * Journal head: identifies the first unused block in the journal. | |
836 | * [j_state_lock] | |
837 | */ | |
838 | unsigned long j_head; | |
839 | ||
840 | /* | |
841 | * Journal tail: identifies the oldest still-used block in the journal. | |
842 | * [j_state_lock] | |
843 | */ | |
844 | unsigned long j_tail; | |
845 | ||
846 | /* | |
847 | * Journal free: how many free blocks are there in the journal? | |
848 | * [j_state_lock] | |
849 | */ | |
850 | unsigned long j_free; | |
851 | ||
852 | /* | |
853 | * Journal start and end: the block numbers of the first usable block | |
854 | * and one beyond the last usable block in the journal. [j_state_lock] | |
855 | */ | |
856 | unsigned long j_first; | |
857 | unsigned long j_last; | |
858 | ||
859 | /* | |
860 | * Device, blocksize and starting block offset for the location where we | |
861 | * store the journal. | |
862 | */ | |
863 | struct block_device *j_dev; | |
864 | int j_blocksize; | |
05496769 TT |
865 | unsigned long long j_blk_offset; |
866 | char j_devname[BDEVNAME_SIZE+24]; | |
470decc6 DK |
867 | |
868 | /* | |
869 | * Device which holds the client fs. For internal journal this will be | |
870 | * equal to j_dev. | |
871 | */ | |
872 | struct block_device *j_fs_dev; | |
873 | ||
874 | /* Total maximum capacity of the journal region on disk. */ | |
875 | unsigned int j_maxlen; | |
876 | ||
8f7d89f3 JK |
877 | /* Number of buffers reserved from the running transaction */ |
878 | atomic_t j_reserved_credits; | |
879 | ||
470decc6 DK |
880 | /* |
881 | * Protects the buffer lists and internal buffer state. | |
882 | */ | |
883 | spinlock_t j_list_lock; | |
884 | ||
885 | /* Optional inode where we store the journal. If present, all */ | |
886 | /* journal block numbers are mapped into this inode via */ | |
887 | /* bmap(). */ | |
888 | struct inode *j_inode; | |
889 | ||
890 | /* | |
891 | * Sequence number of the oldest transaction in the log [j_state_lock] | |
892 | */ | |
893 | tid_t j_tail_sequence; | |
894 | ||
895 | /* | |
896 | * Sequence number of the next transaction to grant [j_state_lock] | |
897 | */ | |
898 | tid_t j_transaction_sequence; | |
899 | ||
900 | /* | |
901 | * Sequence number of the most recently committed transaction | |
902 | * [j_state_lock]. | |
903 | */ | |
904 | tid_t j_commit_sequence; | |
905 | ||
906 | /* | |
907 | * Sequence number of the most recent transaction wanting commit | |
908 | * [j_state_lock] | |
909 | */ | |
910 | tid_t j_commit_request; | |
911 | ||
912 | /* | |
913 | * Journal uuid: identifies the object (filesystem, LVM volume etc) | |
914 | * backed by this journal. This will eventually be replaced by an array | |
915 | * of uuids, allowing us to index multiple devices within a single | |
916 | * journal and to perform atomic updates across them. | |
917 | */ | |
918 | __u8 j_uuid[16]; | |
919 | ||
920 | /* Pointer to the current commit thread for this journal */ | |
921 | struct task_struct *j_task; | |
922 | ||
923 | /* | |
924 | * Maximum number of metadata buffers to allow in a single compound | |
925 | * commit transaction | |
926 | */ | |
927 | int j_max_transaction_buffers; | |
928 | ||
929 | /* | |
930 | * What is the maximum transaction lifetime before we begin a commit? | |
931 | */ | |
932 | unsigned long j_commit_interval; | |
933 | ||
934 | /* The timer used to wakeup the commit thread: */ | |
935 | struct timer_list j_commit_timer; | |
936 | ||
937 | /* | |
938 | * The revoke table: maintains the list of revoked blocks in the | |
939 | * current transaction. [j_revoke_lock] | |
940 | */ | |
941 | spinlock_t j_revoke_lock; | |
f7f4bccb MC |
942 | struct jbd2_revoke_table_s *j_revoke; |
943 | struct jbd2_revoke_table_s *j_revoke_table[2]; | |
470decc6 DK |
944 | |
945 | /* | |
f7f4bccb | 946 | * array of bhs for jbd2_journal_commit_transaction |
470decc6 DK |
947 | */ |
948 | struct buffer_head **j_wbuf; | |
949 | int j_wbufsize; | |
950 | ||
e07f7183 JB |
951 | /* |
952 | * this is the pid of hte last person to run a synchronous operation | |
953 | * through the journal | |
954 | */ | |
470decc6 DK |
955 | pid_t j_last_sync_writer; |
956 | ||
e07f7183 JB |
957 | /* |
958 | * the average amount of time in nanoseconds it takes to commit a | |
959 | * transaction to disk. [j_state_lock] | |
960 | */ | |
961 | u64 j_average_commit_time; | |
962 | ||
30773840 TT |
963 | /* |
964 | * minimum and maximum times that we should wait for | |
965 | * additional filesystem operations to get batched into a | |
966 | * synchronous handle in microseconds | |
967 | */ | |
968 | u32 j_min_batch_time; | |
969 | u32 j_max_batch_time; | |
970 | ||
3e624fc7 TT |
971 | /* This function is called when a transaction is closed */ |
972 | void (*j_commit_callback)(journal_t *, | |
973 | transaction_t *); | |
974 | ||
470decc6 | 975 | /* |
8e85fb3f JL |
976 | * Journal statistics |
977 | */ | |
8e85fb3f JL |
978 | spinlock_t j_history_lock; |
979 | struct proc_dir_entry *j_proc_entry; | |
980 | struct transaction_stats_s j_stats; | |
981 | ||
624080ed TT |
982 | /* Failed journal commit ID */ |
983 | unsigned int j_failed_commit; | |
984 | ||
8e85fb3f | 985 | /* |
470decc6 DK |
986 | * An opaque pointer to fs-private information. ext3 puts its |
987 | * superblock pointer here | |
988 | */ | |
989 | void *j_private; | |
01b5adce DW |
990 | |
991 | /* Reference to checksum algorithm driver via cryptoapi */ | |
992 | struct crypto_shash *j_chksum_driver; | |
4fd5ea43 DW |
993 | |
994 | /* Precomputed journal UUID checksum for seeding other checksums */ | |
995 | __u32 j_csum_seed; | |
470decc6 DK |
996 | }; |
997 | ||
998 | /* | |
999 | * Journal flag definitions | |
1000 | */ | |
f7f4bccb MC |
1001 | #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */ |
1002 | #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */ | |
1003 | #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */ | |
1004 | #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */ | |
1005 | #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */ | |
1006 | #define JBD2_BARRIER 0x020 /* Use IDE barriers */ | |
5bf5683a HK |
1007 | #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file |
1008 | * data write error in ordered | |
1009 | * mode */ | |
470decc6 DK |
1010 | |
1011 | /* | |
1012 | * Function declarations for the journaling transaction and buffer | |
1013 | * management | |
1014 | */ | |
1015 | ||
1016 | /* Filing buffers */ | |
f7f4bccb | 1017 | extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *); |
f7f4bccb MC |
1018 | extern void __jbd2_journal_refile_buffer(struct journal_head *); |
1019 | extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *); | |
1020 | extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); | |
470decc6 | 1021 | extern void __journal_free_buffer(struct journal_head *bh); |
f7f4bccb | 1022 | extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int); |
470decc6 | 1023 | extern void __journal_clean_data_list(transaction_t *transaction); |
f5113eff JK |
1024 | static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh) |
1025 | { | |
1026 | list_add_tail(&bh->b_assoc_buffers, head); | |
1027 | } | |
1028 | static inline void jbd2_unfile_log_bh(struct buffer_head *bh) | |
1029 | { | |
1030 | list_del_init(&bh->b_assoc_buffers); | |
1031 | } | |
470decc6 DK |
1032 | |
1033 | /* Log buffer allocation */ | |
e5a120ae | 1034 | struct buffer_head *jbd2_journal_get_descriptor_buffer(journal_t *journal); |
18eba7aa | 1035 | int jbd2_journal_next_log_block(journal_t *, unsigned long long *); |
79feb521 JK |
1036 | int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid, |
1037 | unsigned long *block); | |
1038 | void __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); | |
3339578f | 1039 | void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block); |
470decc6 DK |
1040 | |
1041 | /* Commit management */ | |
f7f4bccb | 1042 | extern void jbd2_journal_commit_transaction(journal_t *); |
470decc6 DK |
1043 | |
1044 | /* Checkpoint list management */ | |
50849db3 | 1045 | void __jbd2_journal_clean_checkpoint_list(journal_t *journal); |
f7f4bccb MC |
1046 | int __jbd2_journal_remove_checkpoint(struct journal_head *); |
1047 | void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *); | |
470decc6 | 1048 | |
e06c8227 JB |
1049 | |
1050 | /* | |
1051 | * Triggers | |
1052 | */ | |
1053 | ||
1054 | struct jbd2_buffer_trigger_type { | |
1055 | /* | |
13ceef09 JK |
1056 | * Fired a the moment data to write to the journal are known to be |
1057 | * stable - so either at the moment b_frozen_data is created or just | |
1058 | * before a buffer is written to the journal. mapped_data is a mapped | |
1059 | * buffer that is the frozen data for commit. | |
e06c8227 | 1060 | */ |
13ceef09 | 1061 | void (*t_frozen)(struct jbd2_buffer_trigger_type *type, |
e06c8227 JB |
1062 | struct buffer_head *bh, void *mapped_data, |
1063 | size_t size); | |
1064 | ||
1065 | /* | |
1066 | * Fired during journal abort for dirty buffers that will not be | |
1067 | * committed. | |
1068 | */ | |
1069 | void (*t_abort)(struct jbd2_buffer_trigger_type *type, | |
1070 | struct buffer_head *bh); | |
1071 | }; | |
1072 | ||
13ceef09 | 1073 | extern void jbd2_buffer_frozen_trigger(struct journal_head *jh, |
e06c8227 JB |
1074 | void *mapped_data, |
1075 | struct jbd2_buffer_trigger_type *triggers); | |
1076 | extern void jbd2_buffer_abort_trigger(struct journal_head *jh, | |
1077 | struct jbd2_buffer_trigger_type *triggers); | |
1078 | ||
470decc6 | 1079 | /* Buffer IO */ |
f5113eff JK |
1080 | extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction, |
1081 | struct journal_head *jh_in, | |
1082 | struct buffer_head **bh_out, | |
1083 | sector_t blocknr); | |
470decc6 DK |
1084 | |
1085 | /* Transaction locking */ | |
1086 | extern void __wait_on_journal (journal_t *); | |
1087 | ||
0c2022ec YY |
1088 | /* Transaction cache support */ |
1089 | extern void jbd2_journal_destroy_transaction_cache(void); | |
1090 | extern int jbd2_journal_init_transaction_cache(void); | |
1091 | extern void jbd2_journal_free_transaction(transaction_t *); | |
1092 | ||
470decc6 DK |
1093 | /* |
1094 | * Journal locking. | |
1095 | * | |
1096 | * We need to lock the journal during transaction state changes so that nobody | |
1097 | * ever tries to take a handle on the running transaction while we are in the | |
1098 | * middle of moving it to the commit phase. j_state_lock does this. | |
1099 | * | |
1100 | * Note that the locking is completely interrupt unsafe. We never touch | |
1101 | * journal structures from interrupts. | |
1102 | */ | |
1103 | ||
1104 | static inline handle_t *journal_current_handle(void) | |
1105 | { | |
1106 | return current->journal_info; | |
1107 | } | |
1108 | ||
1109 | /* The journaling code user interface: | |
1110 | * | |
1111 | * Create and destroy handles | |
1112 | * Register buffer modifications against the current transaction. | |
1113 | */ | |
1114 | ||
f7f4bccb | 1115 | extern handle_t *jbd2_journal_start(journal_t *, int nblocks); |
8f7d89f3 JK |
1116 | extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks, |
1117 | gfp_t gfp_mask, unsigned int type, | |
1118 | unsigned int line_no); | |
47def826 | 1119 | extern int jbd2_journal_restart(handle_t *, int nblocks); |
d2159fb7 | 1120 | extern int jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask); |
8f7d89f3 JK |
1121 | extern int jbd2_journal_start_reserved(handle_t *handle, |
1122 | unsigned int type, unsigned int line_no); | |
1123 | extern void jbd2_journal_free_reserved(handle_t *handle); | |
f7f4bccb MC |
1124 | extern int jbd2_journal_extend (handle_t *, int nblocks); |
1125 | extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *); | |
1126 | extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *); | |
1127 | extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *); | |
e06c8227 JB |
1128 | void jbd2_journal_set_triggers(struct buffer_head *, |
1129 | struct jbd2_buffer_trigger_type *type); | |
f7f4bccb | 1130 | extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *); |
f7f4bccb | 1131 | extern int jbd2_journal_forget (handle_t *, struct buffer_head *); |
470decc6 | 1132 | extern void journal_sync_buffer (struct buffer_head *); |
53e87268 | 1133 | extern int jbd2_journal_invalidatepage(journal_t *, |
259709b0 | 1134 | struct page *, unsigned int, unsigned int); |
f7f4bccb MC |
1135 | extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t); |
1136 | extern int jbd2_journal_stop(handle_t *); | |
1137 | extern int jbd2_journal_flush (journal_t *); | |
1138 | extern void jbd2_journal_lock_updates (journal_t *); | |
1139 | extern void jbd2_journal_unlock_updates (journal_t *); | |
470decc6 | 1140 | |
f7f4bccb | 1141 | extern journal_t * jbd2_journal_init_dev(struct block_device *bdev, |
470decc6 | 1142 | struct block_device *fs_dev, |
18eba7aa | 1143 | unsigned long long start, int len, int bsize); |
f7f4bccb MC |
1144 | extern journal_t * jbd2_journal_init_inode (struct inode *); |
1145 | extern int jbd2_journal_update_format (journal_t *); | |
1146 | extern int jbd2_journal_check_used_features | |
470decc6 | 1147 | (journal_t *, unsigned long, unsigned long, unsigned long); |
f7f4bccb | 1148 | extern int jbd2_journal_check_available_features |
470decc6 | 1149 | (journal_t *, unsigned long, unsigned long, unsigned long); |
f7f4bccb | 1150 | extern int jbd2_journal_set_features |
470decc6 | 1151 | (journal_t *, unsigned long, unsigned long, unsigned long); |
818d276c GS |
1152 | extern void jbd2_journal_clear_features |
1153 | (journal_t *, unsigned long, unsigned long, unsigned long); | |
f7f4bccb | 1154 | extern int jbd2_journal_load (journal_t *journal); |
44519faf | 1155 | extern int jbd2_journal_destroy (journal_t *); |
f7f4bccb MC |
1156 | extern int jbd2_journal_recover (journal_t *journal); |
1157 | extern int jbd2_journal_wipe (journal_t *, int); | |
1158 | extern int jbd2_journal_skip_recovery (journal_t *); | |
d796c52e | 1159 | extern void jbd2_journal_update_sb_errno(journal_t *); |
79feb521 JK |
1160 | extern void jbd2_journal_update_sb_log_tail (journal_t *, tid_t, |
1161 | unsigned long, int); | |
f7f4bccb MC |
1162 | extern void __jbd2_journal_abort_hard (journal_t *); |
1163 | extern void jbd2_journal_abort (journal_t *, int); | |
1164 | extern int jbd2_journal_errno (journal_t *); | |
1165 | extern void jbd2_journal_ack_err (journal_t *); | |
1166 | extern int jbd2_journal_clear_err (journal_t *); | |
18eba7aa | 1167 | extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *); |
f7f4bccb | 1168 | extern int jbd2_journal_force_commit(journal_t *); |
9ff86446 | 1169 | extern int jbd2_journal_force_commit_nested(journal_t *); |
c851ed54 | 1170 | extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode); |
7f5aa215 JK |
1171 | extern int jbd2_journal_begin_ordered_truncate(journal_t *journal, |
1172 | struct jbd2_inode *inode, loff_t new_size); | |
c851ed54 JK |
1173 | extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode); |
1174 | extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode); | |
470decc6 DK |
1175 | |
1176 | /* | |
1177 | * journal_head management | |
1178 | */ | |
f7f4bccb MC |
1179 | struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh); |
1180 | struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh); | |
f7f4bccb | 1181 | void jbd2_journal_put_journal_head(struct journal_head *jh); |
470decc6 DK |
1182 | |
1183 | /* | |
1184 | * handle management | |
1185 | */ | |
e18b890b | 1186 | extern struct kmem_cache *jbd2_handle_cache; |
470decc6 | 1187 | |
af1e76d6 | 1188 | static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags) |
470decc6 | 1189 | { |
28daf4fa | 1190 | return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags); |
470decc6 DK |
1191 | } |
1192 | ||
af1e76d6 | 1193 | static inline void jbd2_free_handle(handle_t *handle) |
470decc6 | 1194 | { |
f7f4bccb | 1195 | kmem_cache_free(jbd2_handle_cache, handle); |
470decc6 DK |
1196 | } |
1197 | ||
8aefcd55 TT |
1198 | /* |
1199 | * jbd2_inode management (optional, for those file systems that want to use | |
1200 | * dynamically allocated jbd2_inode structures) | |
1201 | */ | |
1202 | extern struct kmem_cache *jbd2_inode_cache; | |
1203 | ||
1204 | static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags) | |
1205 | { | |
1206 | return kmem_cache_alloc(jbd2_inode_cache, gfp_flags); | |
1207 | } | |
1208 | ||
1209 | static inline void jbd2_free_inode(struct jbd2_inode *jinode) | |
1210 | { | |
1211 | kmem_cache_free(jbd2_inode_cache, jinode); | |
1212 | } | |
1213 | ||
470decc6 DK |
1214 | /* Primary revoke support */ |
1215 | #define JOURNAL_REVOKE_DEFAULT_HASH 256 | |
f7f4bccb MC |
1216 | extern int jbd2_journal_init_revoke(journal_t *, int); |
1217 | extern void jbd2_journal_destroy_revoke_caches(void); | |
1218 | extern int jbd2_journal_init_revoke_caches(void); | |
470decc6 | 1219 | |
f7f4bccb | 1220 | extern void jbd2_journal_destroy_revoke(journal_t *); |
18eba7aa | 1221 | extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *); |
f7f4bccb | 1222 | extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *); |
e5a120ae JK |
1223 | extern void jbd2_journal_write_revoke_records(journal_t *journal, |
1224 | transaction_t *transaction, | |
1225 | struct list_head *log_bufs, | |
1226 | int write_op); | |
470decc6 DK |
1227 | |
1228 | /* Recovery revoke support */ | |
18eba7aa MC |
1229 | extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t); |
1230 | extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t); | |
f7f4bccb MC |
1231 | extern void jbd2_journal_clear_revoke(journal_t *); |
1232 | extern void jbd2_journal_switch_revoke_table(journal_t *journal); | |
1ba37268 | 1233 | extern void jbd2_clear_buffer_revoked_flags(journal_t *journal); |
470decc6 DK |
1234 | |
1235 | /* | |
1236 | * The log thread user interface: | |
1237 | * | |
1238 | * Request space in the current transaction, and force transaction commit | |
1239 | * transitions on demand. | |
1240 | */ | |
1241 | ||
f7f4bccb MC |
1242 | int jbd2_log_start_commit(journal_t *journal, tid_t tid); |
1243 | int __jbd2_log_start_commit(journal_t *journal, tid_t tid); | |
1244 | int jbd2_journal_start_commit(journal_t *journal, tid_t *tid); | |
f7f4bccb | 1245 | int jbd2_log_wait_commit(journal_t *journal, tid_t tid); |
d76a3a77 | 1246 | int jbd2_complete_transaction(journal_t *journal, tid_t tid); |
f7f4bccb | 1247 | int jbd2_log_do_checkpoint(journal_t *journal); |
bbd2be36 | 1248 | int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid); |
470decc6 | 1249 | |
f7f4bccb | 1250 | void __jbd2_log_wait_for_space(journal_t *journal); |
fb68407b AK |
1251 | extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *); |
1252 | extern int jbd2_cleanup_journal_tail(journal_t *); | |
470decc6 DK |
1253 | |
1254 | /* Debugging code only: */ | |
1255 | ||
1256 | #define jbd_ENOSYS() \ | |
1257 | do { \ | |
d5c003b4 | 1258 | printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \ |
470decc6 DK |
1259 | current->state = TASK_UNINTERRUPTIBLE; \ |
1260 | schedule(); \ | |
1261 | } while (1) | |
1262 | ||
1263 | /* | |
1264 | * is_journal_abort | |
1265 | * | |
f7f4bccb | 1266 | * Simple test wrapper function to test the JBD2_ABORT state flag. This |
470decc6 DK |
1267 | * bit, when set, indicates that we have had a fatal error somewhere, |
1268 | * either inside the journaling layer or indicated to us by the client | |
1269 | * (eg. ext3), and that we and should not commit any further | |
1270 | * transactions. | |
1271 | */ | |
1272 | ||
1273 | static inline int is_journal_aborted(journal_t *journal) | |
1274 | { | |
f7f4bccb | 1275 | return journal->j_flags & JBD2_ABORT; |
470decc6 DK |
1276 | } |
1277 | ||
1278 | static inline int is_handle_aborted(handle_t *handle) | |
1279 | { | |
41a5b913 | 1280 | if (handle->h_aborted || !handle->h_transaction) |
470decc6 DK |
1281 | return 1; |
1282 | return is_journal_aborted(handle->h_transaction->t_journal); | |
1283 | } | |
1284 | ||
f7f4bccb | 1285 | static inline void jbd2_journal_abort_handle(handle_t *handle) |
470decc6 DK |
1286 | { |
1287 | handle->h_aborted = 1; | |
1288 | } | |
1289 | ||
1290 | #endif /* __KERNEL__ */ | |
1291 | ||
1292 | /* Comparison functions for transaction IDs: perform comparisons using | |
1293 | * modulo arithmetic so that they work over sequence number wraps. */ | |
1294 | ||
1295 | static inline int tid_gt(tid_t x, tid_t y) | |
1296 | { | |
1297 | int difference = (x - y); | |
1298 | return (difference > 0); | |
1299 | } | |
1300 | ||
1301 | static inline int tid_geq(tid_t x, tid_t y) | |
1302 | { | |
1303 | int difference = (x - y); | |
1304 | return (difference >= 0); | |
1305 | } | |
1306 | ||
f7f4bccb | 1307 | extern int jbd2_journal_blocks_per_page(struct inode *inode); |
b517bea1 | 1308 | extern size_t journal_tag_bytes(journal_t *journal); |
470decc6 | 1309 | |
db9ee220 DW |
1310 | static inline int jbd2_journal_has_csum_v2or3(journal_t *journal) |
1311 | { | |
1312 | if (JBD2_HAS_INCOMPAT_FEATURE(journal, JBD2_FEATURE_INCOMPAT_CSUM_V2) || | |
1313 | JBD2_HAS_INCOMPAT_FEATURE(journal, JBD2_FEATURE_INCOMPAT_CSUM_V3)) | |
1314 | return 1; | |
1315 | ||
1316 | return 0; | |
1317 | } | |
1318 | ||
76c39904 JK |
1319 | /* |
1320 | * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for | |
1321 | * transaction control blocks. | |
1322 | */ | |
1323 | #define JBD2_CONTROL_BLOCKS_SHIFT 5 | |
1324 | ||
470decc6 DK |
1325 | /* |
1326 | * Return the minimum number of blocks which must be free in the journal | |
1327 | * before a new transaction may be started. Must be called under j_state_lock. | |
1328 | */ | |
76c39904 | 1329 | static inline int jbd2_space_needed(journal_t *journal) |
470decc6 DK |
1330 | { |
1331 | int nblocks = journal->j_max_transaction_buffers; | |
76c39904 JK |
1332 | return nblocks + (nblocks >> JBD2_CONTROL_BLOCKS_SHIFT); |
1333 | } | |
1334 | ||
1335 | /* | |
1336 | * Return number of free blocks in the log. Must be called under j_state_lock. | |
1337 | */ | |
1338 | static inline unsigned long jbd2_log_space_left(journal_t *journal) | |
1339 | { | |
1340 | /* Allow for rounding errors */ | |
1341 | unsigned long free = journal->j_free - 32; | |
1342 | ||
1343 | if (journal->j_committing_transaction) { | |
1344 | unsigned long committing = atomic_read(&journal-> | |
1345 | j_committing_transaction->t_outstanding_credits); | |
1346 | ||
1347 | /* Transaction + control blocks */ | |
1348 | free -= committing + (committing >> JBD2_CONTROL_BLOCKS_SHIFT); | |
1349 | } | |
1350 | return free; | |
470decc6 DK |
1351 | } |
1352 | ||
1353 | /* | |
1354 | * Definitions which augment the buffer_head layer | |
1355 | */ | |
1356 | ||
1357 | /* journaling buffer types */ | |
1358 | #define BJ_None 0 /* Not journaled */ | |
87c89c23 JK |
1359 | #define BJ_Metadata 1 /* Normal journaled metadata */ |
1360 | #define BJ_Forget 2 /* Buffer superseded by this transaction */ | |
f5113eff | 1361 | #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */ |
e5a120ae JK |
1362 | #define BJ_Reserved 4 /* Buffer is reserved for access by journal */ |
1363 | #define BJ_Types 5 | |
470decc6 DK |
1364 | |
1365 | extern int jbd_blocks_per_page(struct inode *inode); | |
1366 | ||
fd47d3e1 BW |
1367 | /* JBD uses a CRC32 checksum */ |
1368 | #define JBD_MAX_CHECKSUM_SIZE 4 | |
1369 | ||
01b5adce DW |
1370 | static inline u32 jbd2_chksum(journal_t *journal, u32 crc, |
1371 | const void *address, unsigned int length) | |
1372 | { | |
1373 | struct { | |
1374 | struct shash_desc shash; | |
fd47d3e1 | 1375 | char ctx[JBD_MAX_CHECKSUM_SIZE]; |
01b5adce DW |
1376 | } desc; |
1377 | int err; | |
1378 | ||
fd47d3e1 BW |
1379 | BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) > |
1380 | JBD_MAX_CHECKSUM_SIZE); | |
1381 | ||
01b5adce DW |
1382 | desc.shash.tfm = journal->j_chksum_driver; |
1383 | desc.shash.flags = 0; | |
1384 | *(u32 *)desc.ctx = crc; | |
1385 | ||
1386 | err = crypto_shash_update(&desc.shash, address, length); | |
1387 | BUG_ON(err); | |
1388 | ||
1389 | return *(u32 *)desc.ctx; | |
1390 | } | |
1391 | ||
06a407f1 DM |
1392 | /* Return most recent uncommitted transaction */ |
1393 | static inline tid_t jbd2_get_latest_transaction(journal_t *journal) | |
1394 | { | |
1395 | tid_t tid; | |
1396 | ||
1397 | read_lock(&journal->j_state_lock); | |
1398 | tid = journal->j_commit_request; | |
1399 | if (journal->j_running_transaction) | |
1400 | tid = journal->j_running_transaction->t_tid; | |
1401 | read_unlock(&journal->j_state_lock); | |
1402 | return tid; | |
1403 | } | |
1404 | ||
470decc6 DK |
1405 | #ifdef __KERNEL__ |
1406 | ||
1407 | #define buffer_trace_init(bh) do {} while (0) | |
1408 | #define print_buffer_fields(bh) do {} while (0) | |
1409 | #define print_buffer_trace(bh) do {} while (0) | |
1410 | #define BUFFER_TRACE(bh, info) do {} while (0) | |
1411 | #define BUFFER_TRACE2(bh, bh2, info) do {} while (0) | |
1412 | #define JBUFFER_TRACE(jh, info) do {} while (0) | |
1413 | ||
1414 | #endif /* __KERNEL__ */ | |
1415 | ||
cd02ff0b | 1416 | #endif /* _LINUX_JBD2_H */ |