jbd.h 33 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097
  1. /*
  2. * linux/include/linux/jbd.h
  3. *
  4. * Written by Stephen C. Tweedie <sct@redhat.com>
  5. *
  6. * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
  7. *
  8. * This file is part of the Linux kernel and is made available under
  9. * the terms of the GNU General Public License, version 2, or at your
  10. * option, any later version, incorporated herein by reference.
  11. *
  12. * Definitions for transaction data structures for the buffer cache
  13. * filesystem journaling support.
  14. */
  15. #ifndef _LINUX_JBD_H
  16. #define _LINUX_JBD_H
  17. /* Allow this file to be included directly into e2fsprogs */
  18. #ifndef __KERNEL__
  19. #include "jfs_compat.h"
  20. #define JFS_DEBUG
  21. #define jfs_debug jbd_debug
  22. #else
  23. #include <linux/types.h>
  24. #include <linux/buffer_head.h>
  25. #include <linux/journal-head.h>
  26. #include <linux/stddef.h>
  27. #include <linux/bit_spinlock.h>
  28. #include <linux/mutex.h>
  29. #include <linux/timer.h>
  30. #include <asm/semaphore.h>
  31. #endif
  32. #define journal_oom_retry 1
  33. /*
  34. * Define JBD_PARANIOD_IOFAIL to cause a kernel BUG() if ext3 finds
  35. * certain classes of error which can occur due to failed IOs. Under
  36. * normal use we want ext3 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 JBD_PARANOID_IOFAIL
  41. /*
  42. * The default maximum commit age, in seconds.
  43. */
  44. #define JBD_DEFAULT_MAX_COMMIT_AGE 5
  45. #ifdef CONFIG_JBD_DEBUG
  46. /*
  47. * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal
  48. * consistency checks. By default we don't do this unless
  49. * CONFIG_JBD_DEBUG is on.
  50. */
  51. #define JBD_EXPENSIVE_CHECKING
  52. extern int journal_enable_debug;
  53. #define jbd_debug(n, f, a...) \
  54. do { \
  55. if ((n) <= journal_enable_debug) { \
  56. printk (KERN_DEBUG "(%s, %d): %s: ", \
  57. __FILE__, __LINE__, __FUNCTION__); \
  58. printk (f, ## a); \
  59. } \
  60. } while (0)
  61. #else
  62. #define jbd_debug(f, a...) /**/
  63. #endif
  64. extern void * __jbd_kmalloc (const char *where, size_t size, gfp_t flags, int retry);
  65. extern void * jbd_slab_alloc(size_t size, gfp_t flags);
  66. extern void jbd_slab_free(void *ptr, size_t size);
  67. #define jbd_kmalloc(size, flags) \
  68. __jbd_kmalloc(__FUNCTION__, (size), (flags), journal_oom_retry)
  69. #define jbd_rep_kmalloc(size, flags) \
  70. __jbd_kmalloc(__FUNCTION__, (size), (flags), 1)
  71. #define JFS_MIN_JOURNAL_BLOCKS 1024
  72. #ifdef __KERNEL__
  73. /**
  74. * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
  75. *
  76. * All filesystem modifications made by the process go
  77. * through this handle. Recursive operations (such as quota operations)
  78. * are gathered into a single update.
  79. *
  80. * The buffer credits field is used to account for journaled buffers
  81. * being modified by the running process. To ensure that there is
  82. * enough log space for all outstanding operations, we need to limit the
  83. * number of outstanding buffers possible at any time. When the
  84. * operation completes, any buffer credits not used are credited back to
  85. * the transaction, so that at all times we know how many buffers the
  86. * outstanding updates on a transaction might possibly touch.
  87. *
  88. * This is an opaque datatype.
  89. **/
  90. typedef struct handle_s handle_t; /* Atomic operation type */
  91. /**
  92. * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
  93. *
  94. * journal_t is linked to from the fs superblock structure.
  95. *
  96. * We use the journal_t to keep track of all outstanding transaction
  97. * activity on the filesystem, and to manage the state of the log
  98. * writing process.
  99. *
  100. * This is an opaque datatype.
  101. **/
  102. typedef struct journal_s journal_t; /* Journal control structure */
  103. #endif
  104. /*
  105. * Internal structures used by the logging mechanism:
  106. */
  107. #define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
  108. /*
  109. * On-disk structures
  110. */
  111. /*
  112. * Descriptor block types:
  113. */
  114. #define JFS_DESCRIPTOR_BLOCK 1
  115. #define JFS_COMMIT_BLOCK 2
  116. #define JFS_SUPERBLOCK_V1 3
  117. #define JFS_SUPERBLOCK_V2 4
  118. #define JFS_REVOKE_BLOCK 5
  119. /*
  120. * Standard header for all descriptor blocks:
  121. */
  122. typedef struct journal_header_s
  123. {
  124. __be32 h_magic;
  125. __be32 h_blocktype;
  126. __be32 h_sequence;
  127. } journal_header_t;
  128. /*
  129. * The block tag: used to describe a single buffer in the journal
  130. */
  131. typedef struct journal_block_tag_s
  132. {
  133. __be32 t_blocknr; /* The on-disk block number */
  134. __be32 t_flags; /* See below */
  135. } journal_block_tag_t;
  136. /*
  137. * The revoke descriptor: used on disk to describe a series of blocks to
  138. * be revoked from the log
  139. */
  140. typedef struct journal_revoke_header_s
  141. {
  142. journal_header_t r_header;
  143. __be32 r_count; /* Count of bytes used in the block */
  144. } journal_revoke_header_t;
  145. /* Definitions for the journal tag flags word: */
  146. #define JFS_FLAG_ESCAPE 1 /* on-disk block is escaped */
  147. #define JFS_FLAG_SAME_UUID 2 /* block has same uuid as previous */
  148. #define JFS_FLAG_DELETED 4 /* block deleted by this transaction */
  149. #define JFS_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
  150. /*
  151. * The journal superblock. All fields are in big-endian byte order.
  152. */
  153. typedef struct journal_superblock_s
  154. {
  155. /* 0x0000 */
  156. journal_header_t s_header;
  157. /* 0x000C */
  158. /* Static information describing the journal */
  159. __be32 s_blocksize; /* journal device blocksize */
  160. __be32 s_maxlen; /* total blocks in journal file */
  161. __be32 s_first; /* first block of log information */
  162. /* 0x0018 */
  163. /* Dynamic information describing the current state of the log */
  164. __be32 s_sequence; /* first commit ID expected in log */
  165. __be32 s_start; /* blocknr of start of log */
  166. /* 0x0020 */
  167. /* Error value, as set by journal_abort(). */
  168. __be32 s_errno;
  169. /* 0x0024 */
  170. /* Remaining fields are only valid in a version-2 superblock */
  171. __be32 s_feature_compat; /* compatible feature set */
  172. __be32 s_feature_incompat; /* incompatible feature set */
  173. __be32 s_feature_ro_compat; /* readonly-compatible feature set */
  174. /* 0x0030 */
  175. __u8 s_uuid[16]; /* 128-bit uuid for journal */
  176. /* 0x0040 */
  177. __be32 s_nr_users; /* Nr of filesystems sharing log */
  178. __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
  179. /* 0x0048 */
  180. __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
  181. __be32 s_max_trans_data; /* Limit of data blocks per trans. */
  182. /* 0x0050 */
  183. __u32 s_padding[44];
  184. /* 0x0100 */
  185. __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
  186. /* 0x0400 */
  187. } journal_superblock_t;
  188. #define JFS_HAS_COMPAT_FEATURE(j,mask) \
  189. ((j)->j_format_version >= 2 && \
  190. ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
  191. #define JFS_HAS_RO_COMPAT_FEATURE(j,mask) \
  192. ((j)->j_format_version >= 2 && \
  193. ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
  194. #define JFS_HAS_INCOMPAT_FEATURE(j,mask) \
  195. ((j)->j_format_version >= 2 && \
  196. ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
  197. #define JFS_FEATURE_INCOMPAT_REVOKE 0x00000001
  198. /* Features known to this kernel version: */
  199. #define JFS_KNOWN_COMPAT_FEATURES 0
  200. #define JFS_KNOWN_ROCOMPAT_FEATURES 0
  201. #define JFS_KNOWN_INCOMPAT_FEATURES JFS_FEATURE_INCOMPAT_REVOKE
  202. #ifdef __KERNEL__
  203. #include <linux/fs.h>
  204. #include <linux/sched.h>
  205. #define JBD_ASSERTIONS
  206. #ifdef JBD_ASSERTIONS
  207. #define J_ASSERT(assert) \
  208. do { \
  209. if (!(assert)) { \
  210. printk (KERN_EMERG \
  211. "Assertion failure in %s() at %s:%d: \"%s\"\n", \
  212. __FUNCTION__, __FILE__, __LINE__, # assert); \
  213. BUG(); \
  214. } \
  215. } while (0)
  216. #if defined(CONFIG_BUFFER_DEBUG)
  217. void buffer_assertion_failure(struct buffer_head *bh);
  218. #define J_ASSERT_BH(bh, expr) \
  219. do { \
  220. if (!(expr)) \
  221. buffer_assertion_failure(bh); \
  222. J_ASSERT(expr); \
  223. } while (0)
  224. #define J_ASSERT_JH(jh, expr) J_ASSERT_BH(jh2bh(jh), expr)
  225. #else
  226. #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
  227. #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
  228. #endif
  229. #else
  230. #define J_ASSERT(assert) do { } while (0)
  231. #endif /* JBD_ASSERTIONS */
  232. #if defined(JBD_PARANOID_IOFAIL)
  233. #define J_EXPECT(expr, why...) J_ASSERT(expr)
  234. #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
  235. #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
  236. #else
  237. #define __journal_expect(expr, why...) \
  238. ({ \
  239. int val = (expr); \
  240. if (!val) { \
  241. printk(KERN_ERR \
  242. "EXT3-fs unexpected failure: %s;\n",# expr); \
  243. printk(KERN_ERR why "\n"); \
  244. } \
  245. val; \
  246. })
  247. #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
  248. #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
  249. #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
  250. #endif
  251. enum jbd_state_bits {
  252. BH_JBD /* Has an attached ext3 journal_head */
  253. = BH_PrivateStart,
  254. BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
  255. BH_Freed, /* Has been freed (truncated) */
  256. BH_Revoked, /* Has been revoked from the log */
  257. BH_RevokeValid, /* Revoked flag is valid */
  258. BH_JBDDirty, /* Is dirty but journaled */
  259. BH_State, /* Pins most journal_head state */
  260. BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
  261. BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
  262. };
  263. BUFFER_FNS(JBD, jbd)
  264. BUFFER_FNS(JWrite, jwrite)
  265. BUFFER_FNS(JBDDirty, jbddirty)
  266. TAS_BUFFER_FNS(JBDDirty, jbddirty)
  267. BUFFER_FNS(Revoked, revoked)
  268. TAS_BUFFER_FNS(Revoked, revoked)
  269. BUFFER_FNS(RevokeValid, revokevalid)
  270. TAS_BUFFER_FNS(RevokeValid, revokevalid)
  271. BUFFER_FNS(Freed, freed)
  272. static inline struct buffer_head *jh2bh(struct journal_head *jh)
  273. {
  274. return jh->b_bh;
  275. }
  276. static inline struct journal_head *bh2jh(struct buffer_head *bh)
  277. {
  278. return bh->b_private;
  279. }
  280. static inline void jbd_lock_bh_state(struct buffer_head *bh)
  281. {
  282. bit_spin_lock(BH_State, &bh->b_state);
  283. }
  284. static inline int jbd_trylock_bh_state(struct buffer_head *bh)
  285. {
  286. return bit_spin_trylock(BH_State, &bh->b_state);
  287. }
  288. static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
  289. {
  290. return bit_spin_is_locked(BH_State, &bh->b_state);
  291. }
  292. static inline void jbd_unlock_bh_state(struct buffer_head *bh)
  293. {
  294. bit_spin_unlock(BH_State, &bh->b_state);
  295. }
  296. static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
  297. {
  298. bit_spin_lock(BH_JournalHead, &bh->b_state);
  299. }
  300. static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
  301. {
  302. bit_spin_unlock(BH_JournalHead, &bh->b_state);
  303. }
  304. struct jbd_revoke_table_s;
  305. /**
  306. * struct handle_s - The handle_s type is the concrete type associated with
  307. * handle_t.
  308. * @h_transaction: Which compound transaction is this update a part of?
  309. * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
  310. * @h_ref: Reference count on this handle
  311. * @h_err: Field for caller's use to track errors through large fs operations
  312. * @h_sync: flag for sync-on-close
  313. * @h_jdata: flag to force data journaling
  314. * @h_aborted: flag indicating fatal error on handle
  315. **/
  316. /* Docbook can't yet cope with the bit fields, but will leave the documentation
  317. * in so it can be fixed later.
  318. */
  319. struct handle_s
  320. {
  321. /* Which compound transaction is this update a part of? */
  322. transaction_t *h_transaction;
  323. /* Number of remaining buffers we are allowed to dirty: */
  324. int h_buffer_credits;
  325. /* Reference count on this handle */
  326. int h_ref;
  327. /* Field for caller's use to track errors through large fs */
  328. /* operations */
  329. int h_err;
  330. /* Flags [no locking] */
  331. unsigned int h_sync: 1; /* sync-on-close */
  332. unsigned int h_jdata: 1; /* force data journaling */
  333. unsigned int h_aborted: 1; /* fatal error on handle */
  334. };
  335. /* The transaction_t type is the guts of the journaling mechanism. It
  336. * tracks a compound transaction through its various states:
  337. *
  338. * RUNNING: accepting new updates
  339. * LOCKED: Updates still running but we don't accept new ones
  340. * RUNDOWN: Updates are tidying up but have finished requesting
  341. * new buffers to modify (state not used for now)
  342. * FLUSH: All updates complete, but we are still writing to disk
  343. * COMMIT: All data on disk, writing commit record
  344. * FINISHED: We still have to keep the transaction for checkpointing.
  345. *
  346. * The transaction keeps track of all of the buffers modified by a
  347. * running transaction, and all of the buffers committed but not yet
  348. * flushed to home for finished transactions.
  349. */
  350. /*
  351. * Lock ranking:
  352. *
  353. * j_list_lock
  354. * ->jbd_lock_bh_journal_head() (This is "innermost")
  355. *
  356. * j_state_lock
  357. * ->jbd_lock_bh_state()
  358. *
  359. * jbd_lock_bh_state()
  360. * ->j_list_lock
  361. *
  362. * j_state_lock
  363. * ->t_handle_lock
  364. *
  365. * j_state_lock
  366. * ->j_list_lock (journal_unmap_buffer)
  367. *
  368. */
  369. struct transaction_s
  370. {
  371. /* Pointer to the journal for this transaction. [no locking] */
  372. journal_t *t_journal;
  373. /* Sequence number for this transaction [no locking] */
  374. tid_t t_tid;
  375. /*
  376. * Transaction's current state
  377. * [no locking - only kjournald alters this]
  378. * FIXME: needs barriers
  379. * KLUDGE: [use j_state_lock]
  380. */
  381. enum {
  382. T_RUNNING,
  383. T_LOCKED,
  384. T_RUNDOWN,
  385. T_FLUSH,
  386. T_COMMIT,
  387. T_FINISHED
  388. } t_state;
  389. /*
  390. * Where in the log does this transaction's commit start? [no locking]
  391. */
  392. unsigned long t_log_start;
  393. /* Number of buffers on the t_buffers list [j_list_lock] */
  394. int t_nr_buffers;
  395. /*
  396. * Doubly-linked circular list of all buffers reserved but not yet
  397. * modified by this transaction [j_list_lock]
  398. */
  399. struct journal_head *t_reserved_list;
  400. /*
  401. * Doubly-linked circular list of all buffers under writeout during
  402. * commit [j_list_lock]
  403. */
  404. struct journal_head *t_locked_list;
  405. /*
  406. * Doubly-linked circular list of all metadata buffers owned by this
  407. * transaction [j_list_lock]
  408. */
  409. struct journal_head *t_buffers;
  410. /*
  411. * Doubly-linked circular list of all data buffers still to be
  412. * flushed before this transaction can be committed [j_list_lock]
  413. */
  414. struct journal_head *t_sync_datalist;
  415. /*
  416. * Doubly-linked circular list of all forget buffers (superseded
  417. * buffers which we can un-checkpoint once this transaction commits)
  418. * [j_list_lock]
  419. */
  420. struct journal_head *t_forget;
  421. /*
  422. * Doubly-linked circular list of all buffers still to be flushed before
  423. * this transaction can be checkpointed. [j_list_lock]
  424. */
  425. struct journal_head *t_checkpoint_list;
  426. /*
  427. * Doubly-linked circular list of all buffers submitted for IO while
  428. * checkpointing. [j_list_lock]
  429. */
  430. struct journal_head *t_checkpoint_io_list;
  431. /*
  432. * Doubly-linked circular list of temporary buffers currently undergoing
  433. * IO in the log [j_list_lock]
  434. */
  435. struct journal_head *t_iobuf_list;
  436. /*
  437. * Doubly-linked circular list of metadata buffers being shadowed by log
  438. * IO. The IO buffers on the iobuf list and the shadow buffers on this
  439. * list match each other one for one at all times. [j_list_lock]
  440. */
  441. struct journal_head *t_shadow_list;
  442. /*
  443. * Doubly-linked circular list of control buffers being written to the
  444. * log. [j_list_lock]
  445. */
  446. struct journal_head *t_log_list;
  447. /*
  448. * Protects info related to handles
  449. */
  450. spinlock_t t_handle_lock;
  451. /*
  452. * Number of outstanding updates running on this transaction
  453. * [t_handle_lock]
  454. */
  455. int t_updates;
  456. /*
  457. * Number of buffers reserved for use by all handles in this transaction
  458. * handle but not yet modified. [t_handle_lock]
  459. */
  460. int t_outstanding_credits;
  461. /*
  462. * Forward and backward links for the circular list of all transactions
  463. * awaiting checkpoint. [j_list_lock]
  464. */
  465. transaction_t *t_cpnext, *t_cpprev;
  466. /*
  467. * When will the transaction expire (become due for commit), in jiffies?
  468. * [no locking]
  469. */
  470. unsigned long t_expires;
  471. /*
  472. * How many handles used this transaction? [t_handle_lock]
  473. */
  474. int t_handle_count;
  475. };
  476. /**
  477. * struct journal_s - The journal_s type is the concrete type associated with
  478. * journal_t.
  479. * @j_flags: General journaling state flags
  480. * @j_errno: Is there an outstanding uncleared error on the journal (from a
  481. * prior abort)?
  482. * @j_sb_buffer: First part of superblock buffer
  483. * @j_superblock: Second part of superblock buffer
  484. * @j_format_version: Version of the superblock format
  485. * @j_state_lock: Protect the various scalars in the journal
  486. * @j_barrier_count: Number of processes waiting to create a barrier lock
  487. * @j_barrier: The barrier lock itself
  488. * @j_running_transaction: The current running transaction..
  489. * @j_committing_transaction: the transaction we are pushing to disk
  490. * @j_checkpoint_transactions: a linked circular list of all transactions
  491. * waiting for checkpointing
  492. * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
  493. * to start committing, or for a barrier lock to be released
  494. * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
  495. * @j_wait_done_commit: Wait queue for waiting for commit to complete
  496. * @j_wait_checkpoint: Wait queue to trigger checkpointing
  497. * @j_wait_commit: Wait queue to trigger commit
  498. * @j_wait_updates: Wait queue to wait for updates to complete
  499. * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
  500. * @j_head: Journal head - identifies the first unused block in the journal
  501. * @j_tail: Journal tail - identifies the oldest still-used block in the
  502. * journal.
  503. * @j_free: Journal free - how many free blocks are there in the journal?
  504. * @j_first: The block number of the first usable block
  505. * @j_last: The block number one beyond the last usable block
  506. * @j_dev: Device where we store the journal
  507. * @j_blocksize: blocksize for the location where we store the journal.
  508. * @j_blk_offset: starting block offset for into the device where we store the
  509. * journal
  510. * @j_fs_dev: Device which holds the client fs. For internal journal this will
  511. * be equal to j_dev
  512. * @j_maxlen: Total maximum capacity of the journal region on disk.
  513. * @j_list_lock: Protects the buffer lists and internal buffer state.
  514. * @j_inode: Optional inode where we store the journal. If present, all journal
  515. * block numbers are mapped into this inode via bmap().
  516. * @j_tail_sequence: Sequence number of the oldest transaction in the log
  517. * @j_transaction_sequence: Sequence number of the next transaction to grant
  518. * @j_commit_sequence: Sequence number of the most recently committed
  519. * transaction
  520. * @j_commit_request: Sequence number of the most recent transaction wanting
  521. * commit
  522. * @j_uuid: Uuid of client object.
  523. * @j_task: Pointer to the current commit thread for this journal
  524. * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
  525. * single compound commit transaction
  526. * @j_commit_interval: What is the maximum transaction lifetime before we begin
  527. * a commit?
  528. * @j_commit_timer: The timer used to wakeup the commit thread
  529. * @j_revoke_lock: Protect the revoke table
  530. * @j_revoke: The revoke table - maintains the list of revoked blocks in the
  531. * current transaction.
  532. * @j_revoke_table: alternate revoke tables for j_revoke
  533. * @j_wbuf: array of buffer_heads for journal_commit_transaction
  534. * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
  535. * number that will fit in j_blocksize
  536. * @j_last_sync_writer: most recent pid which did a synchronous write
  537. * @j_private: An opaque pointer to fs-private information.
  538. */
  539. struct journal_s
  540. {
  541. /* General journaling state flags [j_state_lock] */
  542. unsigned long j_flags;
  543. /*
  544. * Is there an outstanding uncleared error on the journal (from a prior
  545. * abort)? [j_state_lock]
  546. */
  547. int j_errno;
  548. /* The superblock buffer */
  549. struct buffer_head *j_sb_buffer;
  550. journal_superblock_t *j_superblock;
  551. /* Version of the superblock format */
  552. int j_format_version;
  553. /*
  554. * Protect the various scalars in the journal
  555. */
  556. spinlock_t j_state_lock;
  557. /*
  558. * Number of processes waiting to create a barrier lock [j_state_lock]
  559. */
  560. int j_barrier_count;
  561. /* The barrier lock itself */
  562. struct mutex j_barrier;
  563. /*
  564. * Transactions: The current running transaction...
  565. * [j_state_lock] [caller holding open handle]
  566. */
  567. transaction_t *j_running_transaction;
  568. /*
  569. * the transaction we are pushing to disk
  570. * [j_state_lock] [caller holding open handle]
  571. */
  572. transaction_t *j_committing_transaction;
  573. /*
  574. * ... and a linked circular list of all transactions waiting for
  575. * checkpointing. [j_list_lock]
  576. */
  577. transaction_t *j_checkpoint_transactions;
  578. /*
  579. * Wait queue for waiting for a locked transaction to start committing,
  580. * or for a barrier lock to be released
  581. */
  582. wait_queue_head_t j_wait_transaction_locked;
  583. /* Wait queue for waiting for checkpointing to complete */
  584. wait_queue_head_t j_wait_logspace;
  585. /* Wait queue for waiting for commit to complete */
  586. wait_queue_head_t j_wait_done_commit;
  587. /* Wait queue to trigger checkpointing */
  588. wait_queue_head_t j_wait_checkpoint;
  589. /* Wait queue to trigger commit */
  590. wait_queue_head_t j_wait_commit;
  591. /* Wait queue to wait for updates to complete */
  592. wait_queue_head_t j_wait_updates;
  593. /* Semaphore for locking against concurrent checkpoints */
  594. struct mutex j_checkpoint_mutex;
  595. /*
  596. * Journal head: identifies the first unused block in the journal.
  597. * [j_state_lock]
  598. */
  599. unsigned long j_head;
  600. /*
  601. * Journal tail: identifies the oldest still-used block in the journal.
  602. * [j_state_lock]
  603. */
  604. unsigned long j_tail;
  605. /*
  606. * Journal free: how many free blocks are there in the journal?
  607. * [j_state_lock]
  608. */
  609. unsigned long j_free;
  610. /*
  611. * Journal start and end: the block numbers of the first usable block
  612. * and one beyond the last usable block in the journal. [j_state_lock]
  613. */
  614. unsigned long j_first;
  615. unsigned long j_last;
  616. /*
  617. * Device, blocksize and starting block offset for the location where we
  618. * store the journal.
  619. */
  620. struct block_device *j_dev;
  621. int j_blocksize;
  622. unsigned long j_blk_offset;
  623. /*
  624. * Device which holds the client fs. For internal journal this will be
  625. * equal to j_dev.
  626. */
  627. struct block_device *j_fs_dev;
  628. /* Total maximum capacity of the journal region on disk. */
  629. unsigned int j_maxlen;
  630. /*
  631. * Protects the buffer lists and internal buffer state.
  632. */
  633. spinlock_t j_list_lock;
  634. /* Optional inode where we store the journal. If present, all */
  635. /* journal block numbers are mapped into this inode via */
  636. /* bmap(). */
  637. struct inode *j_inode;
  638. /*
  639. * Sequence number of the oldest transaction in the log [j_state_lock]
  640. */
  641. tid_t j_tail_sequence;
  642. /*
  643. * Sequence number of the next transaction to grant [j_state_lock]
  644. */
  645. tid_t j_transaction_sequence;
  646. /*
  647. * Sequence number of the most recently committed transaction
  648. * [j_state_lock].
  649. */
  650. tid_t j_commit_sequence;
  651. /*
  652. * Sequence number of the most recent transaction wanting commit
  653. * [j_state_lock]
  654. */
  655. tid_t j_commit_request;
  656. /*
  657. * Journal uuid: identifies the object (filesystem, LVM volume etc)
  658. * backed by this journal. This will eventually be replaced by an array
  659. * of uuids, allowing us to index multiple devices within a single
  660. * journal and to perform atomic updates across them.
  661. */
  662. __u8 j_uuid[16];
  663. /* Pointer to the current commit thread for this journal */
  664. struct task_struct *j_task;
  665. /*
  666. * Maximum number of metadata buffers to allow in a single compound
  667. * commit transaction
  668. */
  669. int j_max_transaction_buffers;
  670. /*
  671. * What is the maximum transaction lifetime before we begin a commit?
  672. */
  673. unsigned long j_commit_interval;
  674. /* The timer used to wakeup the commit thread: */
  675. struct timer_list j_commit_timer;
  676. /*
  677. * The revoke table: maintains the list of revoked blocks in the
  678. * current transaction. [j_revoke_lock]
  679. */
  680. spinlock_t j_revoke_lock;
  681. struct jbd_revoke_table_s *j_revoke;
  682. struct jbd_revoke_table_s *j_revoke_table[2];
  683. /*
  684. * array of bhs for journal_commit_transaction
  685. */
  686. struct buffer_head **j_wbuf;
  687. int j_wbufsize;
  688. pid_t j_last_sync_writer;
  689. /*
  690. * An opaque pointer to fs-private information. ext3 puts its
  691. * superblock pointer here
  692. */
  693. void *j_private;
  694. };
  695. /*
  696. * Journal flag definitions
  697. */
  698. #define JFS_UNMOUNT 0x001 /* Journal thread is being destroyed */
  699. #define JFS_ABORT 0x002 /* Journaling has been aborted for errors. */
  700. #define JFS_ACK_ERR 0x004 /* The errno in the sb has been acked */
  701. #define JFS_FLUSHED 0x008 /* The journal superblock has been flushed */
  702. #define JFS_LOADED 0x010 /* The journal superblock has been loaded */
  703. #define JFS_BARRIER 0x020 /* Use IDE barriers */
  704. /*
  705. * Function declarations for the journaling transaction and buffer
  706. * management
  707. */
  708. /* Filing buffers */
  709. extern void journal_unfile_buffer(journal_t *, struct journal_head *);
  710. extern void __journal_unfile_buffer(struct journal_head *);
  711. extern void __journal_refile_buffer(struct journal_head *);
  712. extern void journal_refile_buffer(journal_t *, struct journal_head *);
  713. extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
  714. extern void __journal_free_buffer(struct journal_head *bh);
  715. extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
  716. extern void __journal_clean_data_list(transaction_t *transaction);
  717. /* Log buffer allocation */
  718. extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
  719. int journal_next_log_block(journal_t *, unsigned long *);
  720. /* Commit management */
  721. extern void journal_commit_transaction(journal_t *);
  722. /* Checkpoint list management */
  723. int __journal_clean_checkpoint_list(journal_t *journal);
  724. int __journal_remove_checkpoint(struct journal_head *);
  725. void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
  726. /* Buffer IO */
  727. extern int
  728. journal_write_metadata_buffer(transaction_t *transaction,
  729. struct journal_head *jh_in,
  730. struct journal_head **jh_out,
  731. unsigned long blocknr);
  732. /* Transaction locking */
  733. extern void __wait_on_journal (journal_t *);
  734. /*
  735. * Journal locking.
  736. *
  737. * We need to lock the journal during transaction state changes so that nobody
  738. * ever tries to take a handle on the running transaction while we are in the
  739. * middle of moving it to the commit phase. j_state_lock does this.
  740. *
  741. * Note that the locking is completely interrupt unsafe. We never touch
  742. * journal structures from interrupts.
  743. */
  744. static inline handle_t *journal_current_handle(void)
  745. {
  746. return current->journal_info;
  747. }
  748. /* The journaling code user interface:
  749. *
  750. * Create and destroy handles
  751. * Register buffer modifications against the current transaction.
  752. */
  753. extern handle_t *journal_start(journal_t *, int nblocks);
  754. extern int journal_restart (handle_t *, int nblocks);
  755. extern int journal_extend (handle_t *, int nblocks);
  756. extern int journal_get_write_access(handle_t *, struct buffer_head *);
  757. extern int journal_get_create_access (handle_t *, struct buffer_head *);
  758. extern int journal_get_undo_access(handle_t *, struct buffer_head *);
  759. extern int journal_dirty_data (handle_t *, struct buffer_head *);
  760. extern int journal_dirty_metadata (handle_t *, struct buffer_head *);
  761. extern void journal_release_buffer (handle_t *, struct buffer_head *);
  762. extern int journal_forget (handle_t *, struct buffer_head *);
  763. extern void journal_sync_buffer (struct buffer_head *);
  764. extern void journal_invalidatepage(journal_t *,
  765. struct page *, unsigned long);
  766. extern int journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
  767. extern int journal_stop(handle_t *);
  768. extern int journal_flush (journal_t *);
  769. extern void journal_lock_updates (journal_t *);
  770. extern void journal_unlock_updates (journal_t *);
  771. extern journal_t * journal_init_dev(struct block_device *bdev,
  772. struct block_device *fs_dev,
  773. int start, int len, int bsize);
  774. extern journal_t * journal_init_inode (struct inode *);
  775. extern int journal_update_format (journal_t *);
  776. extern int journal_check_used_features
  777. (journal_t *, unsigned long, unsigned long, unsigned long);
  778. extern int journal_check_available_features
  779. (journal_t *, unsigned long, unsigned long, unsigned long);
  780. extern int journal_set_features
  781. (journal_t *, unsigned long, unsigned long, unsigned long);
  782. extern int journal_create (journal_t *);
  783. extern int journal_load (journal_t *journal);
  784. extern void journal_destroy (journal_t *);
  785. extern int journal_recover (journal_t *journal);
  786. extern int journal_wipe (journal_t *, int);
  787. extern int journal_skip_recovery (journal_t *);
  788. extern void journal_update_superblock (journal_t *, int);
  789. extern void __journal_abort_hard (journal_t *);
  790. extern void journal_abort (journal_t *, int);
  791. extern int journal_errno (journal_t *);
  792. extern void journal_ack_err (journal_t *);
  793. extern int journal_clear_err (journal_t *);
  794. extern int journal_bmap(journal_t *, unsigned long, unsigned long *);
  795. extern int journal_force_commit(journal_t *);
  796. /*
  797. * journal_head management
  798. */
  799. struct journal_head *journal_add_journal_head(struct buffer_head *bh);
  800. struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
  801. void journal_remove_journal_head(struct buffer_head *bh);
  802. void journal_put_journal_head(struct journal_head *jh);
  803. /*
  804. * handle management
  805. */
  806. extern struct kmem_cache *jbd_handle_cache;
  807. static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
  808. {
  809. return kmem_cache_alloc(jbd_handle_cache, gfp_flags);
  810. }
  811. static inline void jbd_free_handle(handle_t *handle)
  812. {
  813. kmem_cache_free(jbd_handle_cache, handle);
  814. }
  815. /* Primary revoke support */
  816. #define JOURNAL_REVOKE_DEFAULT_HASH 256
  817. extern int journal_init_revoke(journal_t *, int);
  818. extern void journal_destroy_revoke_caches(void);
  819. extern int journal_init_revoke_caches(void);
  820. extern void journal_destroy_revoke(journal_t *);
  821. extern int journal_revoke (handle_t *,
  822. unsigned long, struct buffer_head *);
  823. extern int journal_cancel_revoke(handle_t *, struct journal_head *);
  824. extern void journal_write_revoke_records(journal_t *, transaction_t *);
  825. /* Recovery revoke support */
  826. extern int journal_set_revoke(journal_t *, unsigned long, tid_t);
  827. extern int journal_test_revoke(journal_t *, unsigned long, tid_t);
  828. extern void journal_clear_revoke(journal_t *);
  829. extern void journal_switch_revoke_table(journal_t *journal);
  830. /*
  831. * The log thread user interface:
  832. *
  833. * Request space in the current transaction, and force transaction commit
  834. * transitions on demand.
  835. */
  836. int __log_space_left(journal_t *); /* Called with journal locked */
  837. int log_start_commit(journal_t *journal, tid_t tid);
  838. int __log_start_commit(journal_t *journal, tid_t tid);
  839. int journal_start_commit(journal_t *journal, tid_t *tid);
  840. int journal_force_commit_nested(journal_t *journal);
  841. int log_wait_commit(journal_t *journal, tid_t tid);
  842. int log_do_checkpoint(journal_t *journal);
  843. void __log_wait_for_space(journal_t *journal);
  844. extern void __journal_drop_transaction(journal_t *, transaction_t *);
  845. extern int cleanup_journal_tail(journal_t *);
  846. /* Debugging code only: */
  847. #define jbd_ENOSYS() \
  848. do { \
  849. printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
  850. current->state = TASK_UNINTERRUPTIBLE; \
  851. schedule(); \
  852. } while (1)
  853. /*
  854. * is_journal_abort
  855. *
  856. * Simple test wrapper function to test the JFS_ABORT state flag. This
  857. * bit, when set, indicates that we have had a fatal error somewhere,
  858. * either inside the journaling layer or indicated to us by the client
  859. * (eg. ext3), and that we and should not commit any further
  860. * transactions.
  861. */
  862. static inline int is_journal_aborted(journal_t *journal)
  863. {
  864. return journal->j_flags & JFS_ABORT;
  865. }
  866. static inline int is_handle_aborted(handle_t *handle)
  867. {
  868. if (handle->h_aborted)
  869. return 1;
  870. return is_journal_aborted(handle->h_transaction->t_journal);
  871. }
  872. static inline void journal_abort_handle(handle_t *handle)
  873. {
  874. handle->h_aborted = 1;
  875. }
  876. #endif /* __KERNEL__ */
  877. /* Comparison functions for transaction IDs: perform comparisons using
  878. * modulo arithmetic so that they work over sequence number wraps. */
  879. static inline int tid_gt(tid_t x, tid_t y)
  880. {
  881. int difference = (x - y);
  882. return (difference > 0);
  883. }
  884. static inline int tid_geq(tid_t x, tid_t y)
  885. {
  886. int difference = (x - y);
  887. return (difference >= 0);
  888. }
  889. extern int journal_blocks_per_page(struct inode *inode);
  890. /*
  891. * Return the minimum number of blocks which must be free in the journal
  892. * before a new transaction may be started. Must be called under j_state_lock.
  893. */
  894. static inline int jbd_space_needed(journal_t *journal)
  895. {
  896. int nblocks = journal->j_max_transaction_buffers;
  897. if (journal->j_committing_transaction)
  898. nblocks += journal->j_committing_transaction->
  899. t_outstanding_credits;
  900. return nblocks;
  901. }
  902. /*
  903. * Definitions which augment the buffer_head layer
  904. */
  905. /* journaling buffer types */
  906. #define BJ_None 0 /* Not journaled */
  907. #define BJ_SyncData 1 /* Normal data: flush before commit */
  908. #define BJ_Metadata 2 /* Normal journaled metadata */
  909. #define BJ_Forget 3 /* Buffer superseded by this transaction */
  910. #define BJ_IO 4 /* Buffer is for temporary IO use */
  911. #define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
  912. #define BJ_LogCtl 6 /* Buffer contains log descriptors */
  913. #define BJ_Reserved 7 /* Buffer is reserved for access by journal */
  914. #define BJ_Locked 8 /* Locked for I/O during commit */
  915. #define BJ_Types 9
  916. extern int jbd_blocks_per_page(struct inode *inode);
  917. #ifdef __KERNEL__
  918. #define buffer_trace_init(bh) do {} while (0)
  919. #define print_buffer_fields(bh) do {} while (0)
  920. #define print_buffer_trace(bh) do {} while (0)
  921. #define BUFFER_TRACE(bh, info) do {} while (0)
  922. #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
  923. #define JBUFFER_TRACE(jh, info) do {} while (0)
  924. #endif /* __KERNEL__ */
  925. #endif /* _LINUX_JBD_H */