jbd2.h 34 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106
  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_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 JBD2_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 jbd2_journal_enable_debug;
  53. #define jbd_debug(n, f, a...) \
  54. do { \
  55. if ((n) <= jbd2_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 * __jbd2_kmalloc (const char *where, size_t size, gfp_t flags, int retry);
  65. extern void * jbd2_slab_alloc(size_t size, gfp_t flags);
  66. extern void jbd2_slab_free(void *ptr, size_t size);
  67. #define jbd_kmalloc(size, flags) \
  68. __jbd2_kmalloc(__FUNCTION__, (size), (flags), journal_oom_retry)
  69. #define jbd_rep_kmalloc(size, flags) \
  70. __jbd2_kmalloc(__FUNCTION__, (size), (flags), 1)
  71. #define JBD2_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 JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
  108. /*
  109. * On-disk structures
  110. */
  111. /*
  112. * Descriptor block types:
  113. */
  114. #define JBD2_DESCRIPTOR_BLOCK 1
  115. #define JBD2_COMMIT_BLOCK 2
  116. #define JBD2_SUPERBLOCK_V1 3
  117. #define JBD2_SUPERBLOCK_V2 4
  118. #define JBD2_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. * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
  131. * raw struct shouldn't be used for pointer math or sizeof() - use
  132. * journal_tag_bytes(journal) instead to compute this.
  133. */
  134. typedef struct journal_block_tag_s
  135. {
  136. __be32 t_blocknr; /* The on-disk block number */
  137. __be32 t_flags; /* See below */
  138. __be32 t_blocknr_high; /* most-significant high 32bits. */
  139. } journal_block_tag_t;
  140. #define JBD_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
  141. #define JBD_TAG_SIZE64 (sizeof(journal_block_tag_t))
  142. /*
  143. * The revoke descriptor: used on disk to describe a series of blocks to
  144. * be revoked from the log
  145. */
  146. typedef struct jbd2_journal_revoke_header_s
  147. {
  148. journal_header_t r_header;
  149. __be32 r_count; /* Count of bytes used in the block */
  150. } jbd2_journal_revoke_header_t;
  151. /* Definitions for the journal tag flags word: */
  152. #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
  153. #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
  154. #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
  155. #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
  156. /*
  157. * The journal superblock. All fields are in big-endian byte order.
  158. */
  159. typedef struct journal_superblock_s
  160. {
  161. /* 0x0000 */
  162. journal_header_t s_header;
  163. /* 0x000C */
  164. /* Static information describing the journal */
  165. __be32 s_blocksize; /* journal device blocksize */
  166. __be32 s_maxlen; /* total blocks in journal file */
  167. __be32 s_first; /* first block of log information */
  168. /* 0x0018 */
  169. /* Dynamic information describing the current state of the log */
  170. __be32 s_sequence; /* first commit ID expected in log */
  171. __be32 s_start; /* blocknr of start of log */
  172. /* 0x0020 */
  173. /* Error value, as set by jbd2_journal_abort(). */
  174. __be32 s_errno;
  175. /* 0x0024 */
  176. /* Remaining fields are only valid in a version-2 superblock */
  177. __be32 s_feature_compat; /* compatible feature set */
  178. __be32 s_feature_incompat; /* incompatible feature set */
  179. __be32 s_feature_ro_compat; /* readonly-compatible feature set */
  180. /* 0x0030 */
  181. __u8 s_uuid[16]; /* 128-bit uuid for journal */
  182. /* 0x0040 */
  183. __be32 s_nr_users; /* Nr of filesystems sharing log */
  184. __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
  185. /* 0x0048 */
  186. __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
  187. __be32 s_max_trans_data; /* Limit of data blocks per trans. */
  188. /* 0x0050 */
  189. __u32 s_padding[44];
  190. /* 0x0100 */
  191. __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
  192. /* 0x0400 */
  193. } journal_superblock_t;
  194. #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
  195. ((j)->j_format_version >= 2 && \
  196. ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
  197. #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
  198. ((j)->j_format_version >= 2 && \
  199. ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
  200. #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
  201. ((j)->j_format_version >= 2 && \
  202. ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
  203. #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
  204. #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
  205. /* Features known to this kernel version: */
  206. #define JBD2_KNOWN_COMPAT_FEATURES 0
  207. #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
  208. #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
  209. JBD2_FEATURE_INCOMPAT_64BIT)
  210. #ifdef __KERNEL__
  211. #include <linux/fs.h>
  212. #include <linux/sched.h>
  213. #define JBD_ASSERTIONS
  214. #ifdef JBD_ASSERTIONS
  215. #define J_ASSERT(assert) \
  216. do { \
  217. if (!(assert)) { \
  218. printk (KERN_EMERG \
  219. "Assertion failure in %s() at %s:%d: \"%s\"\n", \
  220. __FUNCTION__, __FILE__, __LINE__, # assert); \
  221. BUG(); \
  222. } \
  223. } while (0)
  224. #if defined(CONFIG_BUFFER_DEBUG)
  225. void buffer_assertion_failure(struct buffer_head *bh);
  226. #define J_ASSERT_BH(bh, expr) \
  227. do { \
  228. if (!(expr)) \
  229. buffer_assertion_failure(bh); \
  230. J_ASSERT(expr); \
  231. } while (0)
  232. #define J_ASSERT_JH(jh, expr) J_ASSERT_BH(jh2bh(jh), expr)
  233. #else
  234. #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
  235. #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
  236. #endif
  237. #else
  238. #define J_ASSERT(assert) do { } while (0)
  239. #endif /* JBD_ASSERTIONS */
  240. #if defined(JBD_PARANOID_IOFAIL)
  241. #define J_EXPECT(expr, why...) J_ASSERT(expr)
  242. #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
  243. #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
  244. #else
  245. #define __journal_expect(expr, why...) \
  246. ({ \
  247. int val = (expr); \
  248. if (!val) { \
  249. printk(KERN_ERR \
  250. "EXT3-fs unexpected failure: %s;\n",# expr); \
  251. printk(KERN_ERR why "\n"); \
  252. } \
  253. val; \
  254. })
  255. #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
  256. #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
  257. #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
  258. #endif
  259. enum jbd_state_bits {
  260. BH_JBD /* Has an attached ext3 journal_head */
  261. = BH_PrivateStart,
  262. BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
  263. BH_Freed, /* Has been freed (truncated) */
  264. BH_Revoked, /* Has been revoked from the log */
  265. BH_RevokeValid, /* Revoked flag is valid */
  266. BH_JBDDirty, /* Is dirty but journaled */
  267. BH_State, /* Pins most journal_head state */
  268. BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
  269. BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
  270. };
  271. BUFFER_FNS(JBD, jbd)
  272. BUFFER_FNS(JWrite, jwrite)
  273. BUFFER_FNS(JBDDirty, jbddirty)
  274. TAS_BUFFER_FNS(JBDDirty, jbddirty)
  275. BUFFER_FNS(Revoked, revoked)
  276. TAS_BUFFER_FNS(Revoked, revoked)
  277. BUFFER_FNS(RevokeValid, revokevalid)
  278. TAS_BUFFER_FNS(RevokeValid, revokevalid)
  279. BUFFER_FNS(Freed, freed)
  280. static inline struct buffer_head *jh2bh(struct journal_head *jh)
  281. {
  282. return jh->b_bh;
  283. }
  284. static inline struct journal_head *bh2jh(struct buffer_head *bh)
  285. {
  286. return bh->b_private;
  287. }
  288. static inline void jbd_lock_bh_state(struct buffer_head *bh)
  289. {
  290. bit_spin_lock(BH_State, &bh->b_state);
  291. }
  292. static inline int jbd_trylock_bh_state(struct buffer_head *bh)
  293. {
  294. return bit_spin_trylock(BH_State, &bh->b_state);
  295. }
  296. static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
  297. {
  298. return bit_spin_is_locked(BH_State, &bh->b_state);
  299. }
  300. static inline void jbd_unlock_bh_state(struct buffer_head *bh)
  301. {
  302. bit_spin_unlock(BH_State, &bh->b_state);
  303. }
  304. static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
  305. {
  306. bit_spin_lock(BH_JournalHead, &bh->b_state);
  307. }
  308. static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
  309. {
  310. bit_spin_unlock(BH_JournalHead, &bh->b_state);
  311. }
  312. struct jbd2_revoke_table_s;
  313. /**
  314. * struct handle_s - The handle_s type is the concrete type associated with
  315. * handle_t.
  316. * @h_transaction: Which compound transaction is this update a part of?
  317. * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
  318. * @h_ref: Reference count on this handle
  319. * @h_err: Field for caller's use to track errors through large fs operations
  320. * @h_sync: flag for sync-on-close
  321. * @h_jdata: flag to force data journaling
  322. * @h_aborted: flag indicating fatal error on handle
  323. **/
  324. /* Docbook can't yet cope with the bit fields, but will leave the documentation
  325. * in so it can be fixed later.
  326. */
  327. struct handle_s
  328. {
  329. /* Which compound transaction is this update a part of? */
  330. transaction_t *h_transaction;
  331. /* Number of remaining buffers we are allowed to dirty: */
  332. int h_buffer_credits;
  333. /* Reference count on this handle */
  334. int h_ref;
  335. /* Field for caller's use to track errors through large fs */
  336. /* operations */
  337. int h_err;
  338. /* Flags [no locking] */
  339. unsigned int h_sync: 1; /* sync-on-close */
  340. unsigned int h_jdata: 1; /* force data journaling */
  341. unsigned int h_aborted: 1; /* fatal error on handle */
  342. };
  343. /* The transaction_t type is the guts of the journaling mechanism. It
  344. * tracks a compound transaction through its various states:
  345. *
  346. * RUNNING: accepting new updates
  347. * LOCKED: Updates still running but we don't accept new ones
  348. * RUNDOWN: Updates are tidying up but have finished requesting
  349. * new buffers to modify (state not used for now)
  350. * FLUSH: All updates complete, but we are still writing to disk
  351. * COMMIT: All data on disk, writing commit record
  352. * FINISHED: We still have to keep the transaction for checkpointing.
  353. *
  354. * The transaction keeps track of all of the buffers modified by a
  355. * running transaction, and all of the buffers committed but not yet
  356. * flushed to home for finished transactions.
  357. */
  358. /*
  359. * Lock ranking:
  360. *
  361. * j_list_lock
  362. * ->jbd_lock_bh_journal_head() (This is "innermost")
  363. *
  364. * j_state_lock
  365. * ->jbd_lock_bh_state()
  366. *
  367. * jbd_lock_bh_state()
  368. * ->j_list_lock
  369. *
  370. * j_state_lock
  371. * ->t_handle_lock
  372. *
  373. * j_state_lock
  374. * ->j_list_lock (journal_unmap_buffer)
  375. *
  376. */
  377. struct transaction_s
  378. {
  379. /* Pointer to the journal for this transaction. [no locking] */
  380. journal_t *t_journal;
  381. /* Sequence number for this transaction [no locking] */
  382. tid_t t_tid;
  383. /*
  384. * Transaction's current state
  385. * [no locking - only kjournald2 alters this]
  386. * FIXME: needs barriers
  387. * KLUDGE: [use j_state_lock]
  388. */
  389. enum {
  390. T_RUNNING,
  391. T_LOCKED,
  392. T_RUNDOWN,
  393. T_FLUSH,
  394. T_COMMIT,
  395. T_FINISHED
  396. } t_state;
  397. /*
  398. * Where in the log does this transaction's commit start? [no locking]
  399. */
  400. unsigned long t_log_start;
  401. /* Number of buffers on the t_buffers list [j_list_lock] */
  402. int t_nr_buffers;
  403. /*
  404. * Doubly-linked circular list of all buffers reserved but not yet
  405. * modified by this transaction [j_list_lock]
  406. */
  407. struct journal_head *t_reserved_list;
  408. /*
  409. * Doubly-linked circular list of all buffers under writeout during
  410. * commit [j_list_lock]
  411. */
  412. struct journal_head *t_locked_list;
  413. /*
  414. * Doubly-linked circular list of all metadata buffers owned by this
  415. * transaction [j_list_lock]
  416. */
  417. struct journal_head *t_buffers;
  418. /*
  419. * Doubly-linked circular list of all data buffers still to be
  420. * flushed before this transaction can be committed [j_list_lock]
  421. */
  422. struct journal_head *t_sync_datalist;
  423. /*
  424. * Doubly-linked circular list of all forget buffers (superseded
  425. * buffers which we can un-checkpoint once this transaction commits)
  426. * [j_list_lock]
  427. */
  428. struct journal_head *t_forget;
  429. /*
  430. * Doubly-linked circular list of all buffers still to be flushed before
  431. * this transaction can be checkpointed. [j_list_lock]
  432. */
  433. struct journal_head *t_checkpoint_list;
  434. /*
  435. * Doubly-linked circular list of all buffers submitted for IO while
  436. * checkpointing. [j_list_lock]
  437. */
  438. struct journal_head *t_checkpoint_io_list;
  439. /*
  440. * Doubly-linked circular list of temporary buffers currently undergoing
  441. * IO in the log [j_list_lock]
  442. */
  443. struct journal_head *t_iobuf_list;
  444. /*
  445. * Doubly-linked circular list of metadata buffers being shadowed by log
  446. * IO. The IO buffers on the iobuf list and the shadow buffers on this
  447. * list match each other one for one at all times. [j_list_lock]
  448. */
  449. struct journal_head *t_shadow_list;
  450. /*
  451. * Doubly-linked circular list of control buffers being written to the
  452. * log. [j_list_lock]
  453. */
  454. struct journal_head *t_log_list;
  455. /*
  456. * Protects info related to handles
  457. */
  458. spinlock_t t_handle_lock;
  459. /*
  460. * Number of outstanding updates running on this transaction
  461. * [t_handle_lock]
  462. */
  463. int t_updates;
  464. /*
  465. * Number of buffers reserved for use by all handles in this transaction
  466. * handle but not yet modified. [t_handle_lock]
  467. */
  468. int t_outstanding_credits;
  469. /*
  470. * Forward and backward links for the circular list of all transactions
  471. * awaiting checkpoint. [j_list_lock]
  472. */
  473. transaction_t *t_cpnext, *t_cpprev;
  474. /*
  475. * When will the transaction expire (become due for commit), in jiffies?
  476. * [no locking]
  477. */
  478. unsigned long t_expires;
  479. /*
  480. * How many handles used this transaction? [t_handle_lock]
  481. */
  482. int t_handle_count;
  483. };
  484. /**
  485. * struct journal_s - The journal_s type is the concrete type associated with
  486. * journal_t.
  487. * @j_flags: General journaling state flags
  488. * @j_errno: Is there an outstanding uncleared error on the journal (from a
  489. * prior abort)?
  490. * @j_sb_buffer: First part of superblock buffer
  491. * @j_superblock: Second part of superblock buffer
  492. * @j_format_version: Version of the superblock format
  493. * @j_state_lock: Protect the various scalars in the journal
  494. * @j_barrier_count: Number of processes waiting to create a barrier lock
  495. * @j_barrier: The barrier lock itself
  496. * @j_running_transaction: The current running transaction..
  497. * @j_committing_transaction: the transaction we are pushing to disk
  498. * @j_checkpoint_transactions: a linked circular list of all transactions
  499. * waiting for checkpointing
  500. * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
  501. * to start committing, or for a barrier lock to be released
  502. * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
  503. * @j_wait_done_commit: Wait queue for waiting for commit to complete
  504. * @j_wait_checkpoint: Wait queue to trigger checkpointing
  505. * @j_wait_commit: Wait queue to trigger commit
  506. * @j_wait_updates: Wait queue to wait for updates to complete
  507. * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
  508. * @j_head: Journal head - identifies the first unused block in the journal
  509. * @j_tail: Journal tail - identifies the oldest still-used block in the
  510. * journal.
  511. * @j_free: Journal free - how many free blocks are there in the journal?
  512. * @j_first: The block number of the first usable block
  513. * @j_last: The block number one beyond the last usable block
  514. * @j_dev: Device where we store the journal
  515. * @j_blocksize: blocksize for the location where we store the journal.
  516. * @j_blk_offset: starting block offset for into the device where we store the
  517. * journal
  518. * @j_fs_dev: Device which holds the client fs. For internal journal this will
  519. * be equal to j_dev
  520. * @j_maxlen: Total maximum capacity of the journal region on disk.
  521. * @j_list_lock: Protects the buffer lists and internal buffer state.
  522. * @j_inode: Optional inode where we store the journal. If present, all journal
  523. * block numbers are mapped into this inode via bmap().
  524. * @j_tail_sequence: Sequence number of the oldest transaction in the log
  525. * @j_transaction_sequence: Sequence number of the next transaction to grant
  526. * @j_commit_sequence: Sequence number of the most recently committed
  527. * transaction
  528. * @j_commit_request: Sequence number of the most recent transaction wanting
  529. * commit
  530. * @j_uuid: Uuid of client object.
  531. * @j_task: Pointer to the current commit thread for this journal
  532. * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
  533. * single compound commit transaction
  534. * @j_commit_interval: What is the maximum transaction lifetime before we begin
  535. * a commit?
  536. * @j_commit_timer: The timer used to wakeup the commit thread
  537. * @j_revoke_lock: Protect the revoke table
  538. * @j_revoke: The revoke table - maintains the list of revoked blocks in the
  539. * current transaction.
  540. * @j_revoke_table: alternate revoke tables for j_revoke
  541. * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
  542. * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
  543. * number that will fit in j_blocksize
  544. * @j_last_sync_writer: most recent pid which did a synchronous write
  545. * @j_private: An opaque pointer to fs-private information.
  546. */
  547. struct journal_s
  548. {
  549. /* General journaling state flags [j_state_lock] */
  550. unsigned long j_flags;
  551. /*
  552. * Is there an outstanding uncleared error on the journal (from a prior
  553. * abort)? [j_state_lock]
  554. */
  555. int j_errno;
  556. /* The superblock buffer */
  557. struct buffer_head *j_sb_buffer;
  558. journal_superblock_t *j_superblock;
  559. /* Version of the superblock format */
  560. int j_format_version;
  561. /*
  562. * Protect the various scalars in the journal
  563. */
  564. spinlock_t j_state_lock;
  565. /*
  566. * Number of processes waiting to create a barrier lock [j_state_lock]
  567. */
  568. int j_barrier_count;
  569. /* The barrier lock itself */
  570. struct mutex j_barrier;
  571. /*
  572. * Transactions: The current running transaction...
  573. * [j_state_lock] [caller holding open handle]
  574. */
  575. transaction_t *j_running_transaction;
  576. /*
  577. * the transaction we are pushing to disk
  578. * [j_state_lock] [caller holding open handle]
  579. */
  580. transaction_t *j_committing_transaction;
  581. /*
  582. * ... and a linked circular list of all transactions waiting for
  583. * checkpointing. [j_list_lock]
  584. */
  585. transaction_t *j_checkpoint_transactions;
  586. /*
  587. * Wait queue for waiting for a locked transaction to start committing,
  588. * or for a barrier lock to be released
  589. */
  590. wait_queue_head_t j_wait_transaction_locked;
  591. /* Wait queue for waiting for checkpointing to complete */
  592. wait_queue_head_t j_wait_logspace;
  593. /* Wait queue for waiting for commit to complete */
  594. wait_queue_head_t j_wait_done_commit;
  595. /* Wait queue to trigger checkpointing */
  596. wait_queue_head_t j_wait_checkpoint;
  597. /* Wait queue to trigger commit */
  598. wait_queue_head_t j_wait_commit;
  599. /* Wait queue to wait for updates to complete */
  600. wait_queue_head_t j_wait_updates;
  601. /* Semaphore for locking against concurrent checkpoints */
  602. struct mutex j_checkpoint_mutex;
  603. /*
  604. * Journal head: identifies the first unused block in the journal.
  605. * [j_state_lock]
  606. */
  607. unsigned long j_head;
  608. /*
  609. * Journal tail: identifies the oldest still-used block in the journal.
  610. * [j_state_lock]
  611. */
  612. unsigned long j_tail;
  613. /*
  614. * Journal free: how many free blocks are there in the journal?
  615. * [j_state_lock]
  616. */
  617. unsigned long j_free;
  618. /*
  619. * Journal start and end: the block numbers of the first usable block
  620. * and one beyond the last usable block in the journal. [j_state_lock]
  621. */
  622. unsigned long j_first;
  623. unsigned long j_last;
  624. /*
  625. * Device, blocksize and starting block offset for the location where we
  626. * store the journal.
  627. */
  628. struct block_device *j_dev;
  629. int j_blocksize;
  630. unsigned long long j_blk_offset;
  631. /*
  632. * Device which holds the client fs. For internal journal this will be
  633. * equal to j_dev.
  634. */
  635. struct block_device *j_fs_dev;
  636. /* Total maximum capacity of the journal region on disk. */
  637. unsigned int j_maxlen;
  638. /*
  639. * Protects the buffer lists and internal buffer state.
  640. */
  641. spinlock_t j_list_lock;
  642. /* Optional inode where we store the journal. If present, all */
  643. /* journal block numbers are mapped into this inode via */
  644. /* bmap(). */
  645. struct inode *j_inode;
  646. /*
  647. * Sequence number of the oldest transaction in the log [j_state_lock]
  648. */
  649. tid_t j_tail_sequence;
  650. /*
  651. * Sequence number of the next transaction to grant [j_state_lock]
  652. */
  653. tid_t j_transaction_sequence;
  654. /*
  655. * Sequence number of the most recently committed transaction
  656. * [j_state_lock].
  657. */
  658. tid_t j_commit_sequence;
  659. /*
  660. * Sequence number of the most recent transaction wanting commit
  661. * [j_state_lock]
  662. */
  663. tid_t j_commit_request;
  664. /*
  665. * Journal uuid: identifies the object (filesystem, LVM volume etc)
  666. * backed by this journal. This will eventually be replaced by an array
  667. * of uuids, allowing us to index multiple devices within a single
  668. * journal and to perform atomic updates across them.
  669. */
  670. __u8 j_uuid[16];
  671. /* Pointer to the current commit thread for this journal */
  672. struct task_struct *j_task;
  673. /*
  674. * Maximum number of metadata buffers to allow in a single compound
  675. * commit transaction
  676. */
  677. int j_max_transaction_buffers;
  678. /*
  679. * What is the maximum transaction lifetime before we begin a commit?
  680. */
  681. unsigned long j_commit_interval;
  682. /* The timer used to wakeup the commit thread: */
  683. struct timer_list j_commit_timer;
  684. /*
  685. * The revoke table: maintains the list of revoked blocks in the
  686. * current transaction. [j_revoke_lock]
  687. */
  688. spinlock_t j_revoke_lock;
  689. struct jbd2_revoke_table_s *j_revoke;
  690. struct jbd2_revoke_table_s *j_revoke_table[2];
  691. /*
  692. * array of bhs for jbd2_journal_commit_transaction
  693. */
  694. struct buffer_head **j_wbuf;
  695. int j_wbufsize;
  696. pid_t j_last_sync_writer;
  697. /*
  698. * An opaque pointer to fs-private information. ext3 puts its
  699. * superblock pointer here
  700. */
  701. void *j_private;
  702. };
  703. /*
  704. * Journal flag definitions
  705. */
  706. #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
  707. #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
  708. #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
  709. #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
  710. #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
  711. #define JBD2_BARRIER 0x020 /* Use IDE barriers */
  712. /*
  713. * Function declarations for the journaling transaction and buffer
  714. * management
  715. */
  716. /* Filing buffers */
  717. extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
  718. extern void __jbd2_journal_unfile_buffer(struct journal_head *);
  719. extern void __jbd2_journal_refile_buffer(struct journal_head *);
  720. extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
  721. extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
  722. extern void __journal_free_buffer(struct journal_head *bh);
  723. extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
  724. extern void __journal_clean_data_list(transaction_t *transaction);
  725. /* Log buffer allocation */
  726. extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
  727. int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
  728. /* Commit management */
  729. extern void jbd2_journal_commit_transaction(journal_t *);
  730. /* Checkpoint list management */
  731. int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
  732. int __jbd2_journal_remove_checkpoint(struct journal_head *);
  733. void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
  734. /* Buffer IO */
  735. extern int
  736. jbd2_journal_write_metadata_buffer(transaction_t *transaction,
  737. struct journal_head *jh_in,
  738. struct journal_head **jh_out,
  739. unsigned long long blocknr);
  740. /* Transaction locking */
  741. extern void __wait_on_journal (journal_t *);
  742. /*
  743. * Journal locking.
  744. *
  745. * We need to lock the journal during transaction state changes so that nobody
  746. * ever tries to take a handle on the running transaction while we are in the
  747. * middle of moving it to the commit phase. j_state_lock does this.
  748. *
  749. * Note that the locking is completely interrupt unsafe. We never touch
  750. * journal structures from interrupts.
  751. */
  752. static inline handle_t *journal_current_handle(void)
  753. {
  754. return current->journal_info;
  755. }
  756. /* The journaling code user interface:
  757. *
  758. * Create and destroy handles
  759. * Register buffer modifications against the current transaction.
  760. */
  761. extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
  762. extern int jbd2_journal_restart (handle_t *, int nblocks);
  763. extern int jbd2_journal_extend (handle_t *, int nblocks);
  764. extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
  765. extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
  766. extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
  767. extern int jbd2_journal_dirty_data (handle_t *, struct buffer_head *);
  768. extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
  769. extern void jbd2_journal_release_buffer (handle_t *, struct buffer_head *);
  770. extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
  771. extern void journal_sync_buffer (struct buffer_head *);
  772. extern void jbd2_journal_invalidatepage(journal_t *,
  773. struct page *, unsigned long);
  774. extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
  775. extern int jbd2_journal_stop(handle_t *);
  776. extern int jbd2_journal_flush (journal_t *);
  777. extern void jbd2_journal_lock_updates (journal_t *);
  778. extern void jbd2_journal_unlock_updates (journal_t *);
  779. extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
  780. struct block_device *fs_dev,
  781. unsigned long long start, int len, int bsize);
  782. extern journal_t * jbd2_journal_init_inode (struct inode *);
  783. extern int jbd2_journal_update_format (journal_t *);
  784. extern int jbd2_journal_check_used_features
  785. (journal_t *, unsigned long, unsigned long, unsigned long);
  786. extern int jbd2_journal_check_available_features
  787. (journal_t *, unsigned long, unsigned long, unsigned long);
  788. extern int jbd2_journal_set_features
  789. (journal_t *, unsigned long, unsigned long, unsigned long);
  790. extern int jbd2_journal_create (journal_t *);
  791. extern int jbd2_journal_load (journal_t *journal);
  792. extern void jbd2_journal_destroy (journal_t *);
  793. extern int jbd2_journal_recover (journal_t *journal);
  794. extern int jbd2_journal_wipe (journal_t *, int);
  795. extern int jbd2_journal_skip_recovery (journal_t *);
  796. extern void jbd2_journal_update_superblock (journal_t *, int);
  797. extern void __jbd2_journal_abort_hard (journal_t *);
  798. extern void jbd2_journal_abort (journal_t *, int);
  799. extern int jbd2_journal_errno (journal_t *);
  800. extern void jbd2_journal_ack_err (journal_t *);
  801. extern int jbd2_journal_clear_err (journal_t *);
  802. extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
  803. extern int jbd2_journal_force_commit(journal_t *);
  804. /*
  805. * journal_head management
  806. */
  807. struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
  808. struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
  809. void jbd2_journal_remove_journal_head(struct buffer_head *bh);
  810. void jbd2_journal_put_journal_head(struct journal_head *jh);
  811. /*
  812. * handle management
  813. */
  814. extern struct kmem_cache *jbd2_handle_cache;
  815. static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
  816. {
  817. return kmem_cache_alloc(jbd2_handle_cache, gfp_flags);
  818. }
  819. static inline void jbd_free_handle(handle_t *handle)
  820. {
  821. kmem_cache_free(jbd2_handle_cache, handle);
  822. }
  823. /* Primary revoke support */
  824. #define JOURNAL_REVOKE_DEFAULT_HASH 256
  825. extern int jbd2_journal_init_revoke(journal_t *, int);
  826. extern void jbd2_journal_destroy_revoke_caches(void);
  827. extern int jbd2_journal_init_revoke_caches(void);
  828. extern void jbd2_journal_destroy_revoke(journal_t *);
  829. extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
  830. extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
  831. extern void jbd2_journal_write_revoke_records(journal_t *, transaction_t *);
  832. /* Recovery revoke support */
  833. extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
  834. extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
  835. extern void jbd2_journal_clear_revoke(journal_t *);
  836. extern void jbd2_journal_switch_revoke_table(journal_t *journal);
  837. /*
  838. * The log thread user interface:
  839. *
  840. * Request space in the current transaction, and force transaction commit
  841. * transitions on demand.
  842. */
  843. int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
  844. int jbd2_log_start_commit(journal_t *journal, tid_t tid);
  845. int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
  846. int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
  847. int jbd2_journal_force_commit_nested(journal_t *journal);
  848. int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
  849. int jbd2_log_do_checkpoint(journal_t *journal);
  850. void __jbd2_log_wait_for_space(journal_t *journal);
  851. extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
  852. extern int jbd2_cleanup_journal_tail(journal_t *);
  853. /* Debugging code only: */
  854. #define jbd_ENOSYS() \
  855. do { \
  856. printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
  857. current->state = TASK_UNINTERRUPTIBLE; \
  858. schedule(); \
  859. } while (1)
  860. /*
  861. * is_journal_abort
  862. *
  863. * Simple test wrapper function to test the JBD2_ABORT state flag. This
  864. * bit, when set, indicates that we have had a fatal error somewhere,
  865. * either inside the journaling layer or indicated to us by the client
  866. * (eg. ext3), and that we and should not commit any further
  867. * transactions.
  868. */
  869. static inline int is_journal_aborted(journal_t *journal)
  870. {
  871. return journal->j_flags & JBD2_ABORT;
  872. }
  873. static inline int is_handle_aborted(handle_t *handle)
  874. {
  875. if (handle->h_aborted)
  876. return 1;
  877. return is_journal_aborted(handle->h_transaction->t_journal);
  878. }
  879. static inline void jbd2_journal_abort_handle(handle_t *handle)
  880. {
  881. handle->h_aborted = 1;
  882. }
  883. #endif /* __KERNEL__ */
  884. /* Comparison functions for transaction IDs: perform comparisons using
  885. * modulo arithmetic so that they work over sequence number wraps. */
  886. static inline int tid_gt(tid_t x, tid_t y)
  887. {
  888. int difference = (x - y);
  889. return (difference > 0);
  890. }
  891. static inline int tid_geq(tid_t x, tid_t y)
  892. {
  893. int difference = (x - y);
  894. return (difference >= 0);
  895. }
  896. extern int jbd2_journal_blocks_per_page(struct inode *inode);
  897. extern size_t journal_tag_bytes(journal_t *journal);
  898. /*
  899. * Return the minimum number of blocks which must be free in the journal
  900. * before a new transaction may be started. Must be called under j_state_lock.
  901. */
  902. static inline int jbd_space_needed(journal_t *journal)
  903. {
  904. int nblocks = journal->j_max_transaction_buffers;
  905. if (journal->j_committing_transaction)
  906. nblocks += journal->j_committing_transaction->
  907. t_outstanding_credits;
  908. return nblocks;
  909. }
  910. /*
  911. * Definitions which augment the buffer_head layer
  912. */
  913. /* journaling buffer types */
  914. #define BJ_None 0 /* Not journaled */
  915. #define BJ_SyncData 1 /* Normal data: flush before commit */
  916. #define BJ_Metadata 2 /* Normal journaled metadata */
  917. #define BJ_Forget 3 /* Buffer superseded by this transaction */
  918. #define BJ_IO 4 /* Buffer is for temporary IO use */
  919. #define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
  920. #define BJ_LogCtl 6 /* Buffer contains log descriptors */
  921. #define BJ_Reserved 7 /* Buffer is reserved for access by journal */
  922. #define BJ_Locked 8 /* Locked for I/O during commit */
  923. #define BJ_Types 9
  924. extern int jbd_blocks_per_page(struct inode *inode);
  925. #ifdef __KERNEL__
  926. #define buffer_trace_init(bh) do {} while (0)
  927. #define print_buffer_fields(bh) do {} while (0)
  928. #define print_buffer_trace(bh) do {} while (0)
  929. #define BUFFER_TRACE(bh, info) do {} while (0)
  930. #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
  931. #define JBUFFER_TRACE(jh, info) do {} while (0)
  932. #endif /* __KERNEL__ */
  933. #endif /* _LINUX_JBD_H */