jbd2.h 46 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642
  1. /*
  2. * linux/include/linux/jbd2.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. #ifndef _LINUX_JBD2_H
  16. #define _LINUX_JBD2_H
  17. /* Allow this file to be included directly into e2fsprogs */
  18. #ifndef __KERNEL__
  19. #include "jfs_compat.h"
  20. #define JBD2_DEBUG
  21. #else
  22. #include <linux/types.h>
  23. #include <linux/buffer_head.h>
  24. #include <linux/journal-head.h>
  25. #include <linux/stddef.h>
  26. #include <linux/mutex.h>
  27. #include <linux/timer.h>
  28. #include <linux/slab.h>
  29. #include <linux/bit_spinlock.h>
  30. #include <crypto/hash.h>
  31. #endif
  32. #define journal_oom_retry 1
  33. /*
  34. * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
  35. * certain classes of error which can occur due to failed IOs. Under
  36. * normal use we want ext4 to continue after such errors, because
  37. * hardware _can_ fail, but for debugging purposes when running tests on
  38. * known-good hardware we may want to trap these errors.
  39. */
  40. #undef JBD2_PARANOID_IOFAIL
  41. /*
  42. * The default maximum commit age, in seconds.
  43. */
  44. #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
  45. #ifdef CONFIG_JBD2_DEBUG
  46. /*
  47. * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
  48. * consistency checks. By default we don't do this unless
  49. * CONFIG_JBD2_DEBUG is on.
  50. */
  51. #define JBD2_EXPENSIVE_CHECKING
  52. extern ushort jbd2_journal_enable_debug;
  53. void __jbd2_debug(int level, const char *file, const char *func,
  54. unsigned int line, const char *fmt, ...);
  55. #define jbd_debug(n, fmt, a...) \
  56. __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
  57. #else
  58. #define jbd_debug(n, fmt, a...) /**/
  59. #endif
  60. extern void *jbd2_alloc(size_t size, gfp_t flags);
  61. extern void jbd2_free(void *ptr, size_t size);
  62. #define JBD2_MIN_JOURNAL_BLOCKS 1024
  63. #ifdef __KERNEL__
  64. /**
  65. * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
  66. *
  67. * All filesystem modifications made by the process go
  68. * through this handle. Recursive operations (such as quota operations)
  69. * are gathered into a single update.
  70. *
  71. * The buffer credits field is used to account for journaled buffers
  72. * being modified by the running process. To ensure that there is
  73. * enough log space for all outstanding operations, we need to limit the
  74. * number of outstanding buffers possible at any time. When the
  75. * operation completes, any buffer credits not used are credited back to
  76. * the transaction, so that at all times we know how many buffers the
  77. * outstanding updates on a transaction might possibly touch.
  78. *
  79. * This is an opaque datatype.
  80. **/
  81. typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */
  82. /**
  83. * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
  84. *
  85. * journal_t is linked to from the fs superblock structure.
  86. *
  87. * We use the journal_t to keep track of all outstanding transaction
  88. * activity on the filesystem, and to manage the state of the log
  89. * writing process.
  90. *
  91. * This is an opaque datatype.
  92. **/
  93. typedef struct journal_s journal_t; /* Journal control structure */
  94. #endif
  95. /*
  96. * Internal structures used by the logging mechanism:
  97. */
  98. #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
  99. /*
  100. * On-disk structures
  101. */
  102. /*
  103. * Descriptor block types:
  104. */
  105. #define JBD2_DESCRIPTOR_BLOCK 1
  106. #define JBD2_COMMIT_BLOCK 2
  107. #define JBD2_SUPERBLOCK_V1 3
  108. #define JBD2_SUPERBLOCK_V2 4
  109. #define JBD2_REVOKE_BLOCK 5
  110. /*
  111. * Standard header for all descriptor blocks:
  112. */
  113. typedef struct journal_header_s
  114. {
  115. __be32 h_magic;
  116. __be32 h_blocktype;
  117. __be32 h_sequence;
  118. } journal_header_t;
  119. /*
  120. * Checksum types.
  121. */
  122. #define JBD2_CRC32_CHKSUM 1
  123. #define JBD2_MD5_CHKSUM 2
  124. #define JBD2_SHA1_CHKSUM 3
  125. #define JBD2_CRC32C_CHKSUM 4
  126. #define JBD2_CRC32_CHKSUM_SIZE 4
  127. #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
  128. /*
  129. * Commit block header for storing transactional checksums:
  130. *
  131. * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
  132. * fields are used to store a checksum of the descriptor and data blocks.
  133. *
  134. * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
  135. * field is used to store crc32c(uuid+commit_block). Each journal metadata
  136. * block gets its own checksum, and data block checksums are stored in
  137. * journal_block_tag (in the descriptor). The other h_chksum* fields are
  138. * not used.
  139. *
  140. * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses
  141. * journal_block_tag3_t to store a full 32-bit checksum. Everything else
  142. * is the same as v2.
  143. *
  144. * Checksum v1, v2, and v3 are mutually exclusive features.
  145. */
  146. struct commit_header {
  147. __be32 h_magic;
  148. __be32 h_blocktype;
  149. __be32 h_sequence;
  150. unsigned char h_chksum_type;
  151. unsigned char h_chksum_size;
  152. unsigned char h_padding[2];
  153. __be32 h_chksum[JBD2_CHECKSUM_BYTES];
  154. __be64 h_commit_sec;
  155. __be32 h_commit_nsec;
  156. };
  157. /*
  158. * The block tag: used to describe a single buffer in the journal.
  159. * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
  160. * raw struct shouldn't be used for pointer math or sizeof() - use
  161. * journal_tag_bytes(journal) instead to compute this.
  162. */
  163. typedef struct journal_block_tag3_s
  164. {
  165. __be32 t_blocknr; /* The on-disk block number */
  166. __be32 t_flags; /* See below */
  167. __be32 t_blocknr_high; /* most-significant high 32bits. */
  168. __be32 t_checksum; /* crc32c(uuid+seq+block) */
  169. } journal_block_tag3_t;
  170. typedef struct journal_block_tag_s
  171. {
  172. __be32 t_blocknr; /* The on-disk block number */
  173. __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */
  174. __be16 t_flags; /* See below */
  175. __be32 t_blocknr_high; /* most-significant high 32bits. */
  176. } journal_block_tag_t;
  177. /* Tail of descriptor or revoke block, for checksumming */
  178. struct jbd2_journal_block_tail {
  179. __be32 t_checksum; /* crc32c(uuid+descr_block) */
  180. };
  181. /*
  182. * The revoke descriptor: used on disk to describe a series of blocks to
  183. * be revoked from the log
  184. */
  185. typedef struct jbd2_journal_revoke_header_s
  186. {
  187. journal_header_t r_header;
  188. __be32 r_count; /* Count of bytes used in the block */
  189. } jbd2_journal_revoke_header_t;
  190. /* Definitions for the journal tag flags word: */
  191. #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
  192. #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
  193. #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
  194. #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
  195. /*
  196. * The journal superblock. All fields are in big-endian byte order.
  197. */
  198. typedef struct journal_superblock_s
  199. {
  200. /* 0x0000 */
  201. journal_header_t s_header;
  202. /* 0x000C */
  203. /* Static information describing the journal */
  204. __be32 s_blocksize; /* journal device blocksize */
  205. __be32 s_maxlen; /* total blocks in journal file */
  206. __be32 s_first; /* first block of log information */
  207. /* 0x0018 */
  208. /* Dynamic information describing the current state of the log */
  209. __be32 s_sequence; /* first commit ID expected in log */
  210. __be32 s_start; /* blocknr of start of log */
  211. /* 0x0020 */
  212. /* Error value, as set by jbd2_journal_abort(). */
  213. __be32 s_errno;
  214. /* 0x0024 */
  215. /* Remaining fields are only valid in a version-2 superblock */
  216. __be32 s_feature_compat; /* compatible feature set */
  217. __be32 s_feature_incompat; /* incompatible feature set */
  218. __be32 s_feature_ro_compat; /* readonly-compatible feature set */
  219. /* 0x0030 */
  220. __u8 s_uuid[16]; /* 128-bit uuid for journal */
  221. /* 0x0040 */
  222. __be32 s_nr_users; /* Nr of filesystems sharing log */
  223. __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
  224. /* 0x0048 */
  225. __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
  226. __be32 s_max_trans_data; /* Limit of data blocks per trans. */
  227. /* 0x0050 */
  228. __u8 s_checksum_type; /* checksum type */
  229. __u8 s_padding2[3];
  230. __u32 s_padding[42];
  231. __be32 s_checksum; /* crc32c(superblock) */
  232. /* 0x0100 */
  233. __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
  234. /* 0x0400 */
  235. } journal_superblock_t;
  236. /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */
  237. #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
  238. ((j)->j_format_version >= 2 && \
  239. ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
  240. #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
  241. ((j)->j_format_version >= 2 && \
  242. ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
  243. #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
  244. ((j)->j_format_version >= 2 && \
  245. ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
  246. #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
  247. #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
  248. #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
  249. #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
  250. #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008
  251. #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010
  252. /* See "journal feature predicate functions" below */
  253. /* Features known to this kernel version: */
  254. #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
  255. #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
  256. #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
  257. JBD2_FEATURE_INCOMPAT_64BIT | \
  258. JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
  259. JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
  260. JBD2_FEATURE_INCOMPAT_CSUM_V3)
  261. #ifdef __KERNEL__
  262. #include <linux/fs.h>
  263. #include <linux/sched.h>
  264. enum jbd_state_bits {
  265. BH_JBD /* Has an attached ext3 journal_head */
  266. = BH_PrivateStart,
  267. BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
  268. BH_Freed, /* Has been freed (truncated) */
  269. BH_Revoked, /* Has been revoked from the log */
  270. BH_RevokeValid, /* Revoked flag is valid */
  271. BH_JBDDirty, /* Is dirty but journaled */
  272. BH_State, /* Pins most journal_head state */
  273. BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
  274. BH_Shadow, /* IO on shadow buffer is running */
  275. BH_Verified, /* Metadata block has been verified ok */
  276. BH_JBDPrivateStart, /* First bit available for private use by FS */
  277. };
  278. BUFFER_FNS(JBD, jbd)
  279. BUFFER_FNS(JWrite, jwrite)
  280. BUFFER_FNS(JBDDirty, jbddirty)
  281. TAS_BUFFER_FNS(JBDDirty, jbddirty)
  282. BUFFER_FNS(Revoked, revoked)
  283. TAS_BUFFER_FNS(Revoked, revoked)
  284. BUFFER_FNS(RevokeValid, revokevalid)
  285. TAS_BUFFER_FNS(RevokeValid, revokevalid)
  286. BUFFER_FNS(Freed, freed)
  287. BUFFER_FNS(Shadow, shadow)
  288. BUFFER_FNS(Verified, verified)
  289. static inline struct buffer_head *jh2bh(struct journal_head *jh)
  290. {
  291. return jh->b_bh;
  292. }
  293. static inline struct journal_head *bh2jh(struct buffer_head *bh)
  294. {
  295. return bh->b_private;
  296. }
  297. static inline void jbd_lock_bh_state(struct buffer_head *bh)
  298. {
  299. bit_spin_lock(BH_State, &bh->b_state);
  300. }
  301. static inline int jbd_trylock_bh_state(struct buffer_head *bh)
  302. {
  303. return bit_spin_trylock(BH_State, &bh->b_state);
  304. }
  305. static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
  306. {
  307. return bit_spin_is_locked(BH_State, &bh->b_state);
  308. }
  309. static inline void jbd_unlock_bh_state(struct buffer_head *bh)
  310. {
  311. bit_spin_unlock(BH_State, &bh->b_state);
  312. }
  313. static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
  314. {
  315. bit_spin_lock(BH_JournalHead, &bh->b_state);
  316. }
  317. static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
  318. {
  319. bit_spin_unlock(BH_JournalHead, &bh->b_state);
  320. }
  321. #define J_ASSERT(assert) BUG_ON(!(assert))
  322. #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
  323. #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
  324. #if defined(JBD2_PARANOID_IOFAIL)
  325. #define J_EXPECT(expr, why...) J_ASSERT(expr)
  326. #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
  327. #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
  328. #else
  329. #define __journal_expect(expr, why...) \
  330. ({ \
  331. int val = (expr); \
  332. if (!val) { \
  333. printk(KERN_ERR \
  334. "JBD2 unexpected failure: %s: %s;\n", \
  335. __func__, #expr); \
  336. printk(KERN_ERR why "\n"); \
  337. } \
  338. val; \
  339. })
  340. #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
  341. #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
  342. #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
  343. #endif
  344. /* Flags in jbd_inode->i_flags */
  345. #define __JI_COMMIT_RUNNING 0
  346. #define __JI_WRITE_DATA 1
  347. #define __JI_WAIT_DATA 2
  348. /*
  349. * Commit of the inode data in progress. We use this flag to protect us from
  350. * concurrent deletion of inode. We cannot use reference to inode for this
  351. * since we cannot afford doing last iput() on behalf of kjournald
  352. */
  353. #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
  354. /* Write allocated dirty buffers in this inode before commit */
  355. #define JI_WRITE_DATA (1 << __JI_WRITE_DATA)
  356. /* Wait for outstanding data writes for this inode before commit */
  357. #define JI_WAIT_DATA (1 << __JI_WAIT_DATA)
  358. /**
  359. * struct jbd_inode - The jbd_inode type is the structure linking inodes in
  360. * ordered mode present in a transaction so that we can sync them during commit.
  361. */
  362. struct jbd2_inode {
  363. /**
  364. * @i_transaction:
  365. *
  366. * Which transaction does this inode belong to? Either the running
  367. * transaction or the committing one. [j_list_lock]
  368. */
  369. transaction_t *i_transaction;
  370. /**
  371. * @i_next_transaction:
  372. *
  373. * Pointer to the running transaction modifying inode's data in case
  374. * there is already a committing transaction touching it. [j_list_lock]
  375. */
  376. transaction_t *i_next_transaction;
  377. /**
  378. * @i_list: List of inodes in the i_transaction [j_list_lock]
  379. */
  380. struct list_head i_list;
  381. /**
  382. * @i_vfs_inode:
  383. *
  384. * VFS inode this inode belongs to [constant for lifetime of structure]
  385. */
  386. struct inode *i_vfs_inode;
  387. /**
  388. * @i_flags: Flags of inode [j_list_lock]
  389. */
  390. unsigned long i_flags;
  391. };
  392. struct jbd2_revoke_table_s;
  393. /**
  394. * struct handle_s - The handle_s type is the concrete type associated with
  395. * handle_t.
  396. * @h_transaction: Which compound transaction is this update a part of?
  397. * @h_journal: Which journal handle belongs to - used iff h_reserved set.
  398. * @h_rsv_handle: Handle reserved for finishing the logical operation.
  399. * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
  400. * @h_ref: Reference count on this handle.
  401. * @h_err: Field for caller's use to track errors through large fs operations.
  402. * @h_sync: Flag for sync-on-close.
  403. * @h_jdata: Flag to force data journaling.
  404. * @h_reserved: Flag for handle for reserved credits.
  405. * @h_aborted: Flag indicating fatal error on handle.
  406. * @h_type: For handle statistics.
  407. * @h_line_no: For handle statistics.
  408. * @h_start_jiffies: Handle Start time.
  409. * @h_requested_credits: Holds @h_buffer_credits after handle is started.
  410. * @saved_alloc_context: Saved context while transaction is open.
  411. **/
  412. /* Docbook can't yet cope with the bit fields, but will leave the documentation
  413. * in so it can be fixed later.
  414. */
  415. struct jbd2_journal_handle
  416. {
  417. union {
  418. transaction_t *h_transaction;
  419. /* Which journal handle belongs to - used iff h_reserved set */
  420. journal_t *h_journal;
  421. };
  422. handle_t *h_rsv_handle;
  423. int h_buffer_credits;
  424. int h_ref;
  425. int h_err;
  426. /* Flags [no locking] */
  427. unsigned int h_sync: 1;
  428. unsigned int h_jdata: 1;
  429. unsigned int h_reserved: 1;
  430. unsigned int h_aborted: 1;
  431. unsigned int h_type: 8;
  432. unsigned int h_line_no: 16;
  433. unsigned long h_start_jiffies;
  434. unsigned int h_requested_credits;
  435. };
  436. /*
  437. * Some stats for checkpoint phase
  438. */
  439. struct transaction_chp_stats_s {
  440. unsigned long cs_chp_time;
  441. __u32 cs_forced_to_close;
  442. __u32 cs_written;
  443. __u32 cs_dropped;
  444. };
  445. /* The transaction_t type is the guts of the journaling mechanism. It
  446. * tracks a compound transaction through its various states:
  447. *
  448. * RUNNING: accepting new updates
  449. * LOCKED: Updates still running but we don't accept new ones
  450. * RUNDOWN: Updates are tidying up but have finished requesting
  451. * new buffers to modify (state not used for now)
  452. * FLUSH: All updates complete, but we are still writing to disk
  453. * COMMIT: All data on disk, writing commit record
  454. * FINISHED: We still have to keep the transaction for checkpointing.
  455. *
  456. * The transaction keeps track of all of the buffers modified by a
  457. * running transaction, and all of the buffers committed but not yet
  458. * flushed to home for finished transactions.
  459. */
  460. /*
  461. * Lock ranking:
  462. *
  463. * j_list_lock
  464. * ->jbd_lock_bh_journal_head() (This is "innermost")
  465. *
  466. * j_state_lock
  467. * ->jbd_lock_bh_state()
  468. *
  469. * jbd_lock_bh_state()
  470. * ->j_list_lock
  471. *
  472. * j_state_lock
  473. * ->t_handle_lock
  474. *
  475. * j_state_lock
  476. * ->j_list_lock (journal_unmap_buffer)
  477. *
  478. */
  479. struct transaction_s
  480. {
  481. /* Pointer to the journal for this transaction. [no locking] */
  482. journal_t *t_journal;
  483. /* Sequence number for this transaction [no locking] */
  484. tid_t t_tid;
  485. /*
  486. * Transaction's current state
  487. * [no locking - only kjournald2 alters this]
  488. * [j_list_lock] guards transition of a transaction into T_FINISHED
  489. * state and subsequent call of __jbd2_journal_drop_transaction()
  490. * FIXME: needs barriers
  491. * KLUDGE: [use j_state_lock]
  492. */
  493. enum {
  494. T_RUNNING,
  495. T_LOCKED,
  496. T_FLUSH,
  497. T_COMMIT,
  498. T_COMMIT_DFLUSH,
  499. T_COMMIT_JFLUSH,
  500. T_COMMIT_CALLBACK,
  501. T_FINISHED
  502. } t_state;
  503. /*
  504. * Where in the log does this transaction's commit start? [no locking]
  505. */
  506. unsigned long t_log_start;
  507. /* Number of buffers on the t_buffers list [j_list_lock] */
  508. int t_nr_buffers;
  509. /*
  510. * Doubly-linked circular list of all buffers reserved but not yet
  511. * modified by this transaction [j_list_lock]
  512. */
  513. struct journal_head *t_reserved_list;
  514. /*
  515. * Doubly-linked circular list of all metadata buffers owned by this
  516. * transaction [j_list_lock]
  517. */
  518. struct journal_head *t_buffers;
  519. /*
  520. * Doubly-linked circular list of all forget buffers (superseded
  521. * buffers which we can un-checkpoint once this transaction commits)
  522. * [j_list_lock]
  523. */
  524. struct journal_head *t_forget;
  525. /*
  526. * Doubly-linked circular list of all buffers still to be flushed before
  527. * this transaction can be checkpointed. [j_list_lock]
  528. */
  529. struct journal_head *t_checkpoint_list;
  530. /*
  531. * Doubly-linked circular list of all buffers submitted for IO while
  532. * checkpointing. [j_list_lock]
  533. */
  534. struct journal_head *t_checkpoint_io_list;
  535. /*
  536. * Doubly-linked circular list of metadata buffers being shadowed by log
  537. * IO. The IO buffers on the iobuf list and the shadow buffers on this
  538. * list match each other one for one at all times. [j_list_lock]
  539. */
  540. struct journal_head *t_shadow_list;
  541. /*
  542. * List of inodes whose data we've modified in data=ordered mode.
  543. * [j_list_lock]
  544. */
  545. struct list_head t_inode_list;
  546. /*
  547. * Protects info related to handles
  548. */
  549. spinlock_t t_handle_lock;
  550. /*
  551. * Longest time some handle had to wait for running transaction
  552. */
  553. unsigned long t_max_wait;
  554. /*
  555. * When transaction started
  556. */
  557. unsigned long t_start;
  558. /*
  559. * When commit was requested
  560. */
  561. unsigned long t_requested;
  562. /*
  563. * Checkpointing stats [j_checkpoint_sem]
  564. */
  565. struct transaction_chp_stats_s t_chp_stats;
  566. /*
  567. * Number of outstanding updates running on this transaction
  568. * [t_handle_lock]
  569. */
  570. atomic_t t_updates;
  571. /*
  572. * Number of buffers reserved for use by all handles in this transaction
  573. * handle but not yet modified. [t_handle_lock]
  574. */
  575. atomic_t t_outstanding_credits;
  576. /*
  577. * Forward and backward links for the circular list of all transactions
  578. * awaiting checkpoint. [j_list_lock]
  579. */
  580. transaction_t *t_cpnext, *t_cpprev;
  581. /*
  582. * When will the transaction expire (become due for commit), in jiffies?
  583. * [no locking]
  584. */
  585. unsigned long t_expires;
  586. /*
  587. * When this transaction started, in nanoseconds [no locking]
  588. */
  589. ktime_t t_start_time;
  590. /*
  591. * How many handles used this transaction? [t_handle_lock]
  592. */
  593. atomic_t t_handle_count;
  594. /*
  595. * This transaction is being forced and some process is
  596. * waiting for it to finish.
  597. */
  598. unsigned int t_synchronous_commit:1;
  599. /* Disk flush needs to be sent to fs partition [no locking] */
  600. int t_need_data_flush;
  601. /*
  602. * For use by the filesystem to store fs-specific data
  603. * structures associated with the transaction
  604. */
  605. struct list_head t_private_list;
  606. };
  607. struct transaction_run_stats_s {
  608. unsigned long rs_wait;
  609. unsigned long rs_request_delay;
  610. unsigned long rs_running;
  611. unsigned long rs_locked;
  612. unsigned long rs_flushing;
  613. unsigned long rs_logging;
  614. __u32 rs_handle_count;
  615. __u32 rs_blocks;
  616. __u32 rs_blocks_logged;
  617. };
  618. struct transaction_stats_s {
  619. unsigned long ts_tid;
  620. unsigned long ts_requested;
  621. struct transaction_run_stats_s run;
  622. };
  623. static inline unsigned long
  624. jbd2_time_diff(unsigned long start, unsigned long end)
  625. {
  626. if (end >= start)
  627. return end - start;
  628. return end + (MAX_JIFFY_OFFSET - start);
  629. }
  630. #define JBD2_NR_BATCH 64
  631. /**
  632. * struct journal_s - The journal_s type is the concrete type associated with
  633. * journal_t.
  634. */
  635. struct journal_s
  636. {
  637. /**
  638. * @j_flags: General journaling state flags [j_state_lock]
  639. */
  640. unsigned long j_flags;
  641. /**
  642. * @j_errno:
  643. *
  644. * Is there an outstanding uncleared error on the journal (from a prior
  645. * abort)? [j_state_lock]
  646. */
  647. int j_errno;
  648. /**
  649. * @j_sb_buffer: The first part of the superblock buffer.
  650. */
  651. struct buffer_head *j_sb_buffer;
  652. /**
  653. * @j_superblock: The second part of the superblock buffer.
  654. */
  655. journal_superblock_t *j_superblock;
  656. /**
  657. * @j_format_version: Version of the superblock format.
  658. */
  659. int j_format_version;
  660. /**
  661. * @j_state_lock: Protect the various scalars in the journal.
  662. */
  663. rwlock_t j_state_lock;
  664. /**
  665. * @j_barrier_count:
  666. *
  667. * Number of processes waiting to create a barrier lock [j_state_lock]
  668. */
  669. int j_barrier_count;
  670. /**
  671. * @j_barrier: The barrier lock itself.
  672. */
  673. struct mutex j_barrier;
  674. /**
  675. * @j_running_transaction:
  676. *
  677. * Transactions: The current running transaction...
  678. * [j_state_lock] [caller holding open handle]
  679. */
  680. transaction_t *j_running_transaction;
  681. /**
  682. * @j_committing_transaction:
  683. *
  684. * the transaction we are pushing to disk
  685. * [j_state_lock] [caller holding open handle]
  686. */
  687. transaction_t *j_committing_transaction;
  688. /**
  689. * @j_checkpoint_transactions:
  690. *
  691. * ... and a linked circular list of all transactions waiting for
  692. * checkpointing. [j_list_lock]
  693. */
  694. transaction_t *j_checkpoint_transactions;
  695. /**
  696. * @j_wait_transaction_locked:
  697. *
  698. * Wait queue for waiting for a locked transaction to start committing,
  699. * or for a barrier lock to be released.
  700. */
  701. wait_queue_head_t j_wait_transaction_locked;
  702. /**
  703. * @j_wait_done_commit: Wait queue for waiting for commit to complete.
  704. */
  705. wait_queue_head_t j_wait_done_commit;
  706. /**
  707. * @j_wait_commit: Wait queue to trigger commit.
  708. */
  709. wait_queue_head_t j_wait_commit;
  710. /**
  711. * @j_wait_updates: Wait queue to wait for updates to complete.
  712. */
  713. wait_queue_head_t j_wait_updates;
  714. /**
  715. * @j_wait_reserved:
  716. *
  717. * Wait queue to wait for reserved buffer credits to drop.
  718. */
  719. wait_queue_head_t j_wait_reserved;
  720. /**
  721. * @j_checkpoint_mutex:
  722. *
  723. * Semaphore for locking against concurrent checkpoints.
  724. */
  725. struct mutex j_checkpoint_mutex;
  726. /**
  727. * @j_chkpt_bhs:
  728. *
  729. * List of buffer heads used by the checkpoint routine. This
  730. * was moved from jbd2_log_do_checkpoint() to reduce stack
  731. * usage. Access to this array is controlled by the
  732. * @j_checkpoint_mutex. [j_checkpoint_mutex]
  733. */
  734. struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
  735. /**
  736. * @j_head:
  737. *
  738. * Journal head: identifies the first unused block in the journal.
  739. * [j_state_lock]
  740. */
  741. unsigned long j_head;
  742. /**
  743. * @j_tail:
  744. *
  745. * Journal tail: identifies the oldest still-used block in the journal.
  746. * [j_state_lock]
  747. */
  748. unsigned long j_tail;
  749. /**
  750. * @j_free:
  751. *
  752. * Journal free: how many free blocks are there in the journal?
  753. * [j_state_lock]
  754. */
  755. unsigned long j_free;
  756. /**
  757. * @j_first:
  758. *
  759. * The block number of the first usable block in the journal
  760. * [j_state_lock].
  761. */
  762. unsigned long j_first;
  763. /**
  764. * @j_last:
  765. *
  766. * The block number one beyond the last usable block in the journal
  767. * [j_state_lock].
  768. */
  769. unsigned long j_last;
  770. /**
  771. * @j_dev: Device where we store the journal.
  772. */
  773. struct block_device *j_dev;
  774. /**
  775. * @j_blocksize: Block size for the location where we store the journal.
  776. */
  777. int j_blocksize;
  778. /**
  779. * @j_blk_offset:
  780. *
  781. * Starting block offset into the device where we store the journal.
  782. */
  783. unsigned long long j_blk_offset;
  784. /**
  785. * @j_devname: Journal device name.
  786. */
  787. char j_devname[BDEVNAME_SIZE+24];
  788. /**
  789. * @j_fs_dev:
  790. *
  791. * Device which holds the client fs. For internal journal this will be
  792. * equal to j_dev.
  793. */
  794. struct block_device *j_fs_dev;
  795. /**
  796. * @j_maxlen: Total maximum capacity of the journal region on disk.
  797. */
  798. unsigned int j_maxlen;
  799. /**
  800. * @j_reserved_credits:
  801. *
  802. * Number of buffers reserved from the running transaction.
  803. */
  804. atomic_t j_reserved_credits;
  805. /**
  806. * @j_list_lock: Protects the buffer lists and internal buffer state.
  807. */
  808. spinlock_t j_list_lock;
  809. /**
  810. * @j_inode:
  811. *
  812. * Optional inode where we store the journal. If present, all
  813. * journal block numbers are mapped into this inode via bmap().
  814. */
  815. struct inode *j_inode;
  816. /**
  817. * @j_tail_sequence:
  818. *
  819. * Sequence number of the oldest transaction in the log [j_state_lock]
  820. */
  821. tid_t j_tail_sequence;
  822. /**
  823. * @j_transaction_sequence:
  824. *
  825. * Sequence number of the next transaction to grant [j_state_lock]
  826. */
  827. tid_t j_transaction_sequence;
  828. /**
  829. * @j_commit_sequence:
  830. *
  831. * Sequence number of the most recently committed transaction
  832. * [j_state_lock].
  833. */
  834. tid_t j_commit_sequence;
  835. /**
  836. * @j_commit_request:
  837. *
  838. * Sequence number of the most recent transaction wanting commit
  839. * [j_state_lock]
  840. */
  841. tid_t j_commit_request;
  842. /**
  843. * @j_uuid:
  844. *
  845. * Journal uuid: identifies the object (filesystem, LVM volume etc)
  846. * backed by this journal. This will eventually be replaced by an array
  847. * of uuids, allowing us to index multiple devices within a single
  848. * journal and to perform atomic updates across them.
  849. */
  850. __u8 j_uuid[16];
  851. /**
  852. * @j_task: Pointer to the current commit thread for this journal.
  853. */
  854. struct task_struct *j_task;
  855. /**
  856. * @j_max_transaction_buffers:
  857. *
  858. * Maximum number of metadata buffers to allow in a single compound
  859. * commit transaction.
  860. */
  861. int j_max_transaction_buffers;
  862. /**
  863. * @j_commit_interval:
  864. *
  865. * What is the maximum transaction lifetime before we begin a commit?
  866. */
  867. unsigned long j_commit_interval;
  868. /**
  869. * @j_commit_timer: The timer used to wakeup the commit thread.
  870. */
  871. struct timer_list j_commit_timer;
  872. /**
  873. * @j_revoke_lock: Protect the revoke table.
  874. */
  875. spinlock_t j_revoke_lock;
  876. /**
  877. * @j_revoke:
  878. *
  879. * The revoke table - maintains the list of revoked blocks in the
  880. * current transaction.
  881. */
  882. struct jbd2_revoke_table_s *j_revoke;
  883. /**
  884. * @j_revoke_table: Alternate revoke tables for j_revoke.
  885. */
  886. struct jbd2_revoke_table_s *j_revoke_table[2];
  887. /**
  888. * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction.
  889. */
  890. struct buffer_head **j_wbuf;
  891. /**
  892. * @j_wbufsize:
  893. *
  894. * Size of @j_wbuf array.
  895. */
  896. int j_wbufsize;
  897. /**
  898. * @j_last_sync_writer:
  899. *
  900. * The pid of the last person to run a synchronous operation
  901. * through the journal.
  902. */
  903. pid_t j_last_sync_writer;
  904. /**
  905. * @j_average_commit_time:
  906. *
  907. * The average amount of time in nanoseconds it takes to commit a
  908. * transaction to disk. [j_state_lock]
  909. */
  910. u64 j_average_commit_time;
  911. /**
  912. * @j_min_batch_time:
  913. *
  914. * Minimum time that we should wait for additional filesystem operations
  915. * to get batched into a synchronous handle in microseconds.
  916. */
  917. u32 j_min_batch_time;
  918. /**
  919. * @j_max_batch_time:
  920. *
  921. * Maximum time that we should wait for additional filesystem operations
  922. * to get batched into a synchronous handle in microseconds.
  923. */
  924. u32 j_max_batch_time;
  925. /**
  926. * @j_commit_callback:
  927. *
  928. * This function is called when a transaction is closed.
  929. */
  930. void (*j_commit_callback)(journal_t *,
  931. transaction_t *);
  932. /*
  933. * Journal statistics
  934. */
  935. /**
  936. * @j_history_lock: Protect the transactions statistics history.
  937. */
  938. spinlock_t j_history_lock;
  939. /**
  940. * @j_proc_entry: procfs entry for the jbd statistics directory.
  941. */
  942. struct proc_dir_entry *j_proc_entry;
  943. /**
  944. * @j_stats: Overall statistics.
  945. */
  946. struct transaction_stats_s j_stats;
  947. /**
  948. * @j_failed_commit: Failed journal commit ID.
  949. */
  950. unsigned int j_failed_commit;
  951. /**
  952. * @j_private:
  953. *
  954. * An opaque pointer to fs-private information. ext3 puts its
  955. * superblock pointer here.
  956. */
  957. void *j_private;
  958. /**
  959. * @j_chksum_driver:
  960. *
  961. * Reference to checksum algorithm driver via cryptoapi.
  962. */
  963. struct crypto_shash *j_chksum_driver;
  964. /**
  965. * @j_csum_seed:
  966. *
  967. * Precomputed journal UUID checksum for seeding other checksums.
  968. */
  969. __u32 j_csum_seed;
  970. #ifdef CONFIG_DEBUG_LOCK_ALLOC
  971. /**
  972. * @j_trans_commit_map:
  973. *
  974. * Lockdep entity to track transaction commit dependencies. Handles
  975. * hold this "lock" for read, when we wait for commit, we acquire the
  976. * "lock" for writing. This matches the properties of jbd2 journalling
  977. * where the running transaction has to wait for all handles to be
  978. * dropped to commit that transaction and also acquiring a handle may
  979. * require transaction commit to finish.
  980. */
  981. struct lockdep_map j_trans_commit_map;
  982. #endif
  983. };
  984. #define jbd2_might_wait_for_commit(j) \
  985. do { \
  986. rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \
  987. rwsem_release(&j->j_trans_commit_map, 1, _THIS_IP_); \
  988. } while (0)
  989. /* journal feature predicate functions */
  990. #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \
  991. static inline bool jbd2_has_feature_##name(journal_t *j) \
  992. { \
  993. return ((j)->j_format_version >= 2 && \
  994. ((j)->j_superblock->s_feature_compat & \
  995. cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \
  996. } \
  997. static inline void jbd2_set_feature_##name(journal_t *j) \
  998. { \
  999. (j)->j_superblock->s_feature_compat |= \
  1000. cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
  1001. } \
  1002. static inline void jbd2_clear_feature_##name(journal_t *j) \
  1003. { \
  1004. (j)->j_superblock->s_feature_compat &= \
  1005. ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
  1006. }
  1007. #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
  1008. static inline bool jbd2_has_feature_##name(journal_t *j) \
  1009. { \
  1010. return ((j)->j_format_version >= 2 && \
  1011. ((j)->j_superblock->s_feature_ro_compat & \
  1012. cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \
  1013. } \
  1014. static inline void jbd2_set_feature_##name(journal_t *j) \
  1015. { \
  1016. (j)->j_superblock->s_feature_ro_compat |= \
  1017. cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
  1018. } \
  1019. static inline void jbd2_clear_feature_##name(journal_t *j) \
  1020. { \
  1021. (j)->j_superblock->s_feature_ro_compat &= \
  1022. ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
  1023. }
  1024. #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \
  1025. static inline bool jbd2_has_feature_##name(journal_t *j) \
  1026. { \
  1027. return ((j)->j_format_version >= 2 && \
  1028. ((j)->j_superblock->s_feature_incompat & \
  1029. cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \
  1030. } \
  1031. static inline void jbd2_set_feature_##name(journal_t *j) \
  1032. { \
  1033. (j)->j_superblock->s_feature_incompat |= \
  1034. cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
  1035. } \
  1036. static inline void jbd2_clear_feature_##name(journal_t *j) \
  1037. { \
  1038. (j)->j_superblock->s_feature_incompat &= \
  1039. ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
  1040. }
  1041. JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM)
  1042. JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE)
  1043. JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT)
  1044. JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT)
  1045. JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2)
  1046. JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3)
  1047. /*
  1048. * Journal flag definitions
  1049. */
  1050. #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
  1051. #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
  1052. #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
  1053. #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
  1054. #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
  1055. #define JBD2_BARRIER 0x020 /* Use IDE barriers */
  1056. #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
  1057. * data write error in ordered
  1058. * mode */
  1059. #define JBD2_REC_ERR 0x080 /* The errno in the sb has been recorded */
  1060. /*
  1061. * Function declarations for the journaling transaction and buffer
  1062. * management
  1063. */
  1064. /* Filing buffers */
  1065. extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
  1066. extern void __jbd2_journal_refile_buffer(struct journal_head *);
  1067. extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
  1068. extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
  1069. extern void __journal_free_buffer(struct journal_head *bh);
  1070. extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
  1071. extern void __journal_clean_data_list(transaction_t *transaction);
  1072. static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
  1073. {
  1074. list_add_tail(&bh->b_assoc_buffers, head);
  1075. }
  1076. static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
  1077. {
  1078. list_del_init(&bh->b_assoc_buffers);
  1079. }
  1080. /* Log buffer allocation */
  1081. struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
  1082. void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
  1083. int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
  1084. int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
  1085. unsigned long *block);
  1086. int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
  1087. void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
  1088. /* Commit management */
  1089. extern void jbd2_journal_commit_transaction(journal_t *);
  1090. /* Checkpoint list management */
  1091. void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy);
  1092. int __jbd2_journal_remove_checkpoint(struct journal_head *);
  1093. void jbd2_journal_destroy_checkpoint(journal_t *journal);
  1094. void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
  1095. /*
  1096. * Triggers
  1097. */
  1098. struct jbd2_buffer_trigger_type {
  1099. /*
  1100. * Fired a the moment data to write to the journal are known to be
  1101. * stable - so either at the moment b_frozen_data is created or just
  1102. * before a buffer is written to the journal. mapped_data is a mapped
  1103. * buffer that is the frozen data for commit.
  1104. */
  1105. void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
  1106. struct buffer_head *bh, void *mapped_data,
  1107. size_t size);
  1108. /*
  1109. * Fired during journal abort for dirty buffers that will not be
  1110. * committed.
  1111. */
  1112. void (*t_abort)(struct jbd2_buffer_trigger_type *type,
  1113. struct buffer_head *bh);
  1114. };
  1115. extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
  1116. void *mapped_data,
  1117. struct jbd2_buffer_trigger_type *triggers);
  1118. extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
  1119. struct jbd2_buffer_trigger_type *triggers);
  1120. /* Buffer IO */
  1121. extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
  1122. struct journal_head *jh_in,
  1123. struct buffer_head **bh_out,
  1124. sector_t blocknr);
  1125. /* Transaction locking */
  1126. extern void __wait_on_journal (journal_t *);
  1127. /* Transaction cache support */
  1128. extern void jbd2_journal_destroy_transaction_cache(void);
  1129. extern int jbd2_journal_init_transaction_cache(void);
  1130. extern void jbd2_journal_free_transaction(transaction_t *);
  1131. /*
  1132. * Journal locking.
  1133. *
  1134. * We need to lock the journal during transaction state changes so that nobody
  1135. * ever tries to take a handle on the running transaction while we are in the
  1136. * middle of moving it to the commit phase. j_state_lock does this.
  1137. *
  1138. * Note that the locking is completely interrupt unsafe. We never touch
  1139. * journal structures from interrupts.
  1140. */
  1141. static inline handle_t *journal_current_handle(void)
  1142. {
  1143. return current->journal_info;
  1144. }
  1145. /* The journaling code user interface:
  1146. *
  1147. * Create and destroy handles
  1148. * Register buffer modifications against the current transaction.
  1149. */
  1150. extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
  1151. extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
  1152. gfp_t gfp_mask, unsigned int type,
  1153. unsigned int line_no);
  1154. extern int jbd2_journal_restart(handle_t *, int nblocks);
  1155. extern int jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask);
  1156. extern int jbd2_journal_start_reserved(handle_t *handle,
  1157. unsigned int type, unsigned int line_no);
  1158. extern void jbd2_journal_free_reserved(handle_t *handle);
  1159. extern int jbd2_journal_extend (handle_t *, int nblocks);
  1160. extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
  1161. extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
  1162. extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
  1163. void jbd2_journal_set_triggers(struct buffer_head *,
  1164. struct jbd2_buffer_trigger_type *type);
  1165. extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
  1166. extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
  1167. extern void journal_sync_buffer (struct buffer_head *);
  1168. extern int jbd2_journal_invalidatepage(journal_t *,
  1169. struct page *, unsigned int, unsigned int);
  1170. extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
  1171. extern int jbd2_journal_stop(handle_t *);
  1172. extern int jbd2_journal_flush (journal_t *);
  1173. extern void jbd2_journal_lock_updates (journal_t *);
  1174. extern void jbd2_journal_unlock_updates (journal_t *);
  1175. extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
  1176. struct block_device *fs_dev,
  1177. unsigned long long start, int len, int bsize);
  1178. extern journal_t * jbd2_journal_init_inode (struct inode *);
  1179. extern int jbd2_journal_update_format (journal_t *);
  1180. extern int jbd2_journal_check_used_features
  1181. (journal_t *, unsigned long, unsigned long, unsigned long);
  1182. extern int jbd2_journal_check_available_features
  1183. (journal_t *, unsigned long, unsigned long, unsigned long);
  1184. extern int jbd2_journal_set_features
  1185. (journal_t *, unsigned long, unsigned long, unsigned long);
  1186. extern void jbd2_journal_clear_features
  1187. (journal_t *, unsigned long, unsigned long, unsigned long);
  1188. extern int jbd2_journal_load (journal_t *journal);
  1189. extern int jbd2_journal_destroy (journal_t *);
  1190. extern int jbd2_journal_recover (journal_t *journal);
  1191. extern int jbd2_journal_wipe (journal_t *, int);
  1192. extern int jbd2_journal_skip_recovery (journal_t *);
  1193. extern void jbd2_journal_update_sb_errno(journal_t *);
  1194. extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t,
  1195. unsigned long, int);
  1196. extern void __jbd2_journal_abort_hard (journal_t *);
  1197. extern void jbd2_journal_abort (journal_t *, int);
  1198. extern int jbd2_journal_errno (journal_t *);
  1199. extern void jbd2_journal_ack_err (journal_t *);
  1200. extern int jbd2_journal_clear_err (journal_t *);
  1201. extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
  1202. extern int jbd2_journal_force_commit(journal_t *);
  1203. extern int jbd2_journal_force_commit_nested(journal_t *);
  1204. extern int jbd2_journal_inode_add_write(handle_t *handle, struct jbd2_inode *inode);
  1205. extern int jbd2_journal_inode_add_wait(handle_t *handle, struct jbd2_inode *inode);
  1206. extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
  1207. struct jbd2_inode *inode, loff_t new_size);
  1208. extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
  1209. extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
  1210. /*
  1211. * journal_head management
  1212. */
  1213. struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
  1214. struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
  1215. void jbd2_journal_put_journal_head(struct journal_head *jh);
  1216. /*
  1217. * handle management
  1218. */
  1219. extern struct kmem_cache *jbd2_handle_cache;
  1220. static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
  1221. {
  1222. return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
  1223. }
  1224. static inline void jbd2_free_handle(handle_t *handle)
  1225. {
  1226. kmem_cache_free(jbd2_handle_cache, handle);
  1227. }
  1228. /*
  1229. * jbd2_inode management (optional, for those file systems that want to use
  1230. * dynamically allocated jbd2_inode structures)
  1231. */
  1232. extern struct kmem_cache *jbd2_inode_cache;
  1233. static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
  1234. {
  1235. return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
  1236. }
  1237. static inline void jbd2_free_inode(struct jbd2_inode *jinode)
  1238. {
  1239. kmem_cache_free(jbd2_inode_cache, jinode);
  1240. }
  1241. /* Primary revoke support */
  1242. #define JOURNAL_REVOKE_DEFAULT_HASH 256
  1243. extern int jbd2_journal_init_revoke(journal_t *, int);
  1244. extern void jbd2_journal_destroy_revoke_caches(void);
  1245. extern int jbd2_journal_init_revoke_caches(void);
  1246. extern void jbd2_journal_destroy_revoke(journal_t *);
  1247. extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
  1248. extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
  1249. extern void jbd2_journal_write_revoke_records(transaction_t *transaction,
  1250. struct list_head *log_bufs);
  1251. /* Recovery revoke support */
  1252. extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
  1253. extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
  1254. extern void jbd2_journal_clear_revoke(journal_t *);
  1255. extern void jbd2_journal_switch_revoke_table(journal_t *journal);
  1256. extern void jbd2_clear_buffer_revoked_flags(journal_t *journal);
  1257. /*
  1258. * The log thread user interface:
  1259. *
  1260. * Request space in the current transaction, and force transaction commit
  1261. * transitions on demand.
  1262. */
  1263. int jbd2_log_start_commit(journal_t *journal, tid_t tid);
  1264. int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
  1265. int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
  1266. int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
  1267. int jbd2_complete_transaction(journal_t *journal, tid_t tid);
  1268. int jbd2_log_do_checkpoint(journal_t *journal);
  1269. int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
  1270. void __jbd2_log_wait_for_space(journal_t *journal);
  1271. extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
  1272. extern int jbd2_cleanup_journal_tail(journal_t *);
  1273. /*
  1274. * is_journal_abort
  1275. *
  1276. * Simple test wrapper function to test the JBD2_ABORT state flag. This
  1277. * bit, when set, indicates that we have had a fatal error somewhere,
  1278. * either inside the journaling layer or indicated to us by the client
  1279. * (eg. ext3), and that we and should not commit any further
  1280. * transactions.
  1281. */
  1282. static inline int is_journal_aborted(journal_t *journal)
  1283. {
  1284. return journal->j_flags & JBD2_ABORT;
  1285. }
  1286. static inline int is_handle_aborted(handle_t *handle)
  1287. {
  1288. if (handle->h_aborted || !handle->h_transaction)
  1289. return 1;
  1290. return is_journal_aborted(handle->h_transaction->t_journal);
  1291. }
  1292. static inline void jbd2_journal_abort_handle(handle_t *handle)
  1293. {
  1294. handle->h_aborted = 1;
  1295. }
  1296. #endif /* __KERNEL__ */
  1297. /* Comparison functions for transaction IDs: perform comparisons using
  1298. * modulo arithmetic so that they work over sequence number wraps. */
  1299. static inline int tid_gt(tid_t x, tid_t y)
  1300. {
  1301. int difference = (x - y);
  1302. return (difference > 0);
  1303. }
  1304. static inline int tid_geq(tid_t x, tid_t y)
  1305. {
  1306. int difference = (x - y);
  1307. return (difference >= 0);
  1308. }
  1309. extern int jbd2_journal_blocks_per_page(struct inode *inode);
  1310. extern size_t journal_tag_bytes(journal_t *journal);
  1311. static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j)
  1312. {
  1313. return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j);
  1314. }
  1315. static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
  1316. {
  1317. WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) &&
  1318. journal->j_chksum_driver == NULL);
  1319. return journal->j_chksum_driver != NULL;
  1320. }
  1321. /*
  1322. * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for
  1323. * transaction control blocks.
  1324. */
  1325. #define JBD2_CONTROL_BLOCKS_SHIFT 5
  1326. /*
  1327. * Return the minimum number of blocks which must be free in the journal
  1328. * before a new transaction may be started. Must be called under j_state_lock.
  1329. */
  1330. static inline int jbd2_space_needed(journal_t *journal)
  1331. {
  1332. int nblocks = journal->j_max_transaction_buffers;
  1333. return nblocks + (nblocks >> JBD2_CONTROL_BLOCKS_SHIFT);
  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. if (journal->j_committing_transaction) {
  1343. unsigned long committing = atomic_read(&journal->
  1344. j_committing_transaction->t_outstanding_credits);
  1345. /* Transaction + control blocks */
  1346. free -= committing + (committing >> JBD2_CONTROL_BLOCKS_SHIFT);
  1347. }
  1348. return free;
  1349. }
  1350. /*
  1351. * Definitions which augment the buffer_head layer
  1352. */
  1353. /* journaling buffer types */
  1354. #define BJ_None 0 /* Not journaled */
  1355. #define BJ_Metadata 1 /* Normal journaled metadata */
  1356. #define BJ_Forget 2 /* Buffer superseded by this transaction */
  1357. #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */
  1358. #define BJ_Reserved 4 /* Buffer is reserved for access by journal */
  1359. #define BJ_Types 5
  1360. extern int jbd_blocks_per_page(struct inode *inode);
  1361. /* JBD uses a CRC32 checksum */
  1362. #define JBD_MAX_CHECKSUM_SIZE 4
  1363. static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
  1364. const void *address, unsigned int length)
  1365. {
  1366. struct {
  1367. struct shash_desc shash;
  1368. char ctx[JBD_MAX_CHECKSUM_SIZE];
  1369. } desc;
  1370. int err;
  1371. BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
  1372. JBD_MAX_CHECKSUM_SIZE);
  1373. desc.shash.tfm = journal->j_chksum_driver;
  1374. desc.shash.flags = 0;
  1375. *(u32 *)desc.ctx = crc;
  1376. err = crypto_shash_update(&desc.shash, address, length);
  1377. BUG_ON(err);
  1378. return *(u32 *)desc.ctx;
  1379. }
  1380. /* Return most recent uncommitted transaction */
  1381. static inline tid_t jbd2_get_latest_transaction(journal_t *journal)
  1382. {
  1383. tid_t tid;
  1384. read_lock(&journal->j_state_lock);
  1385. tid = journal->j_commit_request;
  1386. if (journal->j_running_transaction)
  1387. tid = journal->j_running_transaction->t_tid;
  1388. read_unlock(&journal->j_state_lock);
  1389. return tid;
  1390. }
  1391. #ifdef __KERNEL__
  1392. #define buffer_trace_init(bh) do {} while (0)
  1393. #define print_buffer_fields(bh) do {} while (0)
  1394. #define print_buffer_trace(bh) do {} while (0)
  1395. #define BUFFER_TRACE(bh, info) do {} while (0)
  1396. #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
  1397. #define JBUFFER_TRACE(jh, info) do {} while (0)
  1398. #endif /* __KERNEL__ */
  1399. #define EFSBADCRC EBADMSG /* Bad CRC detected */
  1400. #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
  1401. #endif /* _LINUX_JBD2_H */