]>
Commit | Line | Data |
---|---|---|
2bd0ea18 | 1 | /* |
0d3e0b37 | 2 | * Copyright (c) 2000-2002 Silicon Graphics, Inc. All Rights Reserved. |
5000d01d | 3 | * |
2bd0ea18 NS |
4 | * This program is free software; you can redistribute it and/or modify it |
5 | * under the terms of version 2 of the GNU General Public License as | |
6 | * published by the Free Software Foundation. | |
5000d01d | 7 | * |
2bd0ea18 NS |
8 | * This program is distributed in the hope that it would be useful, but |
9 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. | |
5000d01d | 11 | * |
2bd0ea18 NS |
12 | * Further, this software is distributed without any warranty that it is |
13 | * free of the rightful claim of any third person regarding infringement | |
dfc130f3 | 14 | * or the like. Any license provided herein, whether implied or |
2bd0ea18 NS |
15 | * otherwise, applies only to this software file. Patent licenses, if |
16 | * any, provided herein do not apply to combinations of this program with | |
17 | * other software, or any other product whatsoever. | |
5000d01d | 18 | * |
2bd0ea18 NS |
19 | * You should have received a copy of the GNU General Public License along |
20 | * with this program; if not, write the Free Software Foundation, Inc., 59 | |
21 | * Temple Place - Suite 330, Boston MA 02111-1307, USA. | |
5000d01d | 22 | * |
2bd0ea18 NS |
23 | * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy, |
24 | * Mountain View, CA 94043, or: | |
5000d01d SL |
25 | * |
26 | * http://www.sgi.com | |
27 | * | |
28 | * For further information regarding this notice, see: | |
29 | * | |
2bd0ea18 NS |
30 | * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/ |
31 | */ | |
dfc130f3 | 32 | #ifndef __XFS_LOG_PRIV_H__ |
2bd0ea18 NS |
33 | #define __XFS_LOG_PRIV_H__ |
34 | ||
35 | #if defined(XFS_ALL_TRACE) | |
dfc130f3 | 36 | #define XFS_LOG_TRACE |
2bd0ea18 NS |
37 | #endif |
38 | ||
39 | #if !defined(DEBUG) | |
40 | #undef XFS_LOG_TRACE | |
41 | #endif | |
42 | ||
43 | struct xfs_buf; | |
44 | struct ktrace; | |
45 | struct log; | |
46 | struct xfs_buf_cancel; | |
47 | struct xfs_mount; | |
48 | ||
49 | /* | |
50 | * Macros, structures, prototypes for internal log manager use. | |
51 | */ | |
52 | ||
53 | #define XLOG_NUM_ICLOGS 2 | |
5ac5bb99 | 54 | #define XLOG_MAX_ICLOGS 8 |
2bd0ea18 NS |
55 | #define XLOG_CALLBACK_SIZE 10 |
56 | #define XLOG_HEADER_MAGIC_NUM 0xFEEDbabe /* Illegal cycle number */ | |
73bf5988 SL |
57 | #define XLOG_VERSION_1 1 |
58 | #define XLOG_VERSION_2 2 /* Large IClogs, Log sunit */ | |
59 | #define XLOG_VERSION_OKBITS (XLOG_VERSION_1 | XLOG_VERSION_2) | |
2bd0ea18 | 60 | #define XLOG_RECORD_BSIZE (16*1024) /* eventually 32k */ |
73bf5988 SL |
61 | #define XLOG_BIG_RECORD_BSIZE (32*1024) /* 32k buffers */ |
62 | #define XLOG_MAX_RECORD_BSIZE (256*1024) | |
63 | #define XLOG_HEADER_CYCLE_SIZE (32*1024) /* cycle data in header */ | |
2bd0ea18 | 64 | #define XLOG_RECORD_BSHIFT 14 /* 16384 == 1 << 14 */ |
73bf5988 SL |
65 | #define XLOG_BIG_RECORD_BSHIFT 15 /* 32k == 1 << 15 */ |
66 | #define XLOG_MAX_RECORD_BSHIFT 18 /* 256k == 1 << 18 */ | |
2bd0ea18 NS |
67 | #if XFS_WANT_FUNCS || (XFS_WANT_SPACE && XFSSO_XLOG_BTOLRBB) |
68 | int xlog_btolrbb(int b); | |
69 | #define XLOG_BTOLRBB(b) xlog_btolrbb(b) | |
70 | #else | |
71 | #define XLOG_BTOLRBB(b) (((b)+XLOG_RECORD_BSIZE-1) >> XLOG_RECORD_BSHIFT) | |
72 | #endif | |
73 | ||
74 | #define XLOG_HEADER_SIZE 512 | |
75 | ||
73bf5988 SL |
76 | #define XLOG_TOTAL_REC_SHIFT(log) \ |
77 | BTOBB(XLOG_MAX_ICLOGS << (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? \ | |
78 | XLOG_MAX_RECORD_BSHIFT : XLOG_BIG_RECORD_BSHIFT)) | |
5000d01d | 79 | |
2bd0ea18 NS |
80 | /* |
81 | * set lsns | |
82 | */ | |
83 | ||
84 | #define ASSIGN_LSN_CYCLE(lsn,cycle,arch) \ | |
85 | INT_SET(((uint *)&(lsn))[LSN_FIELD_CYCLE(arch)], arch, (cycle)); | |
86 | #define ASSIGN_LSN_BLOCK(lsn,block,arch) \ | |
87 | INT_SET(((uint *)&(lsn))[LSN_FIELD_BLOCK(arch)], arch, (block)); | |
88 | #define ASSIGN_ANY_LSN(lsn,cycle,block,arch) \ | |
89 | { \ | |
5000d01d SL |
90 | ASSIGN_LSN_CYCLE(lsn,cycle,arch); \ |
91 | ASSIGN_LSN_BLOCK(lsn,block,arch); \ | |
2bd0ea18 NS |
92 | } |
93 | #define ASSIGN_LSN(lsn,log,arch) \ | |
94 | ASSIGN_ANY_LSN(lsn,(log)->l_curr_cycle,(log)->l_curr_block,arch); | |
5000d01d | 95 | |
2bd0ea18 NS |
96 | #define XLOG_SET(f,b) (((f) & (b)) == (b)) |
97 | ||
98 | #define GET_CYCLE(ptr, arch) \ | |
99 | (INT_GET(*(uint *)(ptr), arch) == XLOG_HEADER_MAGIC_NUM ? \ | |
5000d01d SL |
100 | INT_GET(*((uint *)(ptr)+1), arch) : \ |
101 | INT_GET(*(uint *)(ptr), arch) \ | |
2bd0ea18 | 102 | ) |
5000d01d | 103 | |
2bd0ea18 NS |
104 | #define BLK_AVG(blk1, blk2) ((blk1+blk2) >> 1) |
105 | ||
106 | ||
107 | #ifdef __KERNEL__ | |
108 | /* | |
109 | * get client id from packed copy. | |
110 | * | |
111 | * this hack is here because the xlog_pack code copies four bytes | |
5000d01d | 112 | * of xlog_op_header containing the fields oh_clientid, oh_flags |
2bd0ea18 NS |
113 | * and oh_res2 into the packed copy. |
114 | * | |
5000d01d | 115 | * later on this four byte chunk is treated as an int and the |
2bd0ea18 NS |
116 | * client id is pulled out. |
117 | * | |
118 | * this has endian issues, of course. | |
5000d01d | 119 | */ |
2bd0ea18 NS |
120 | |
121 | #if __BYTE_ORDER == __LITTLE_ENDIAN | |
122 | #define GET_CLIENT_ID(i,arch) \ | |
123 | ((i) & 0xff) | |
124 | #else | |
125 | #define GET_CLIENT_ID(i,arch) \ | |
126 | ((i) >> 24) | |
127 | #endif | |
73bf5988 | 128 | |
2bd0ea18 NS |
129 | #if XFS_WANT_FUNCS || (XFS_WANT_SPACE && XFSSO_XLOG_GRANT_SUB_SPACE) |
130 | void xlog_grant_sub_space(struct log *log, int bytes, int type); | |
131 | #define XLOG_GRANT_SUB_SPACE(log,bytes,type) \ | |
132 | xlog_grant_sub_space(log,bytes,type) | |
133 | #else | |
134 | #define XLOG_GRANT_SUB_SPACE(log,bytes,type) \ | |
135 | { \ | |
136 | if (type == 'w') { \ | |
137 | (log)->l_grant_write_bytes -= (bytes); \ | |
138 | if ((log)->l_grant_write_bytes < 0) { \ | |
dfc130f3 | 139 | (log)->l_grant_write_bytes += (log)->l_logsize; \ |
2bd0ea18 NS |
140 | (log)->l_grant_write_cycle--; \ |
141 | } \ | |
142 | } else { \ | |
143 | (log)->l_grant_reserve_bytes -= (bytes); \ | |
144 | if ((log)->l_grant_reserve_bytes < 0) { \ | |
145 | (log)->l_grant_reserve_bytes += (log)->l_logsize;\ | |
146 | (log)->l_grant_reserve_cycle--; \ | |
147 | } \ | |
148 | } \ | |
149 | } | |
150 | #endif | |
151 | #if XFS_WANT_FUNCS || (XFS_WANT_SPACE && XFSSO_XLOG_GRANT_ADD_SPACE) | |
152 | void xlog_grant_add_space(struct log *log, int bytes, int type); | |
153 | #define XLOG_GRANT_ADD_SPACE(log,bytes,type) \ | |
154 | xlog_grant_add_space(log,bytes,type) | |
155 | #else | |
156 | #define XLOG_GRANT_ADD_SPACE(log,bytes,type) \ | |
157 | { \ | |
158 | if (type == 'w') { \ | |
159 | (log)->l_grant_write_bytes += (bytes); \ | |
160 | if ((log)->l_grant_write_bytes > (log)->l_logsize) { \ | |
dfc130f3 | 161 | (log)->l_grant_write_bytes -= (log)->l_logsize; \ |
2bd0ea18 NS |
162 | (log)->l_grant_write_cycle++; \ |
163 | } \ | |
164 | } else { \ | |
165 | (log)->l_grant_reserve_bytes += (bytes); \ | |
166 | if ((log)->l_grant_reserve_bytes > (log)->l_logsize) { \ | |
167 | (log)->l_grant_reserve_bytes -= (log)->l_logsize;\ | |
168 | (log)->l_grant_reserve_cycle++; \ | |
169 | } \ | |
170 | } \ | |
171 | } | |
172 | #endif | |
173 | #define XLOG_INS_TICKETQ(q,tic) \ | |
174 | { \ | |
175 | if (q) { \ | |
176 | (tic)->t_next = (q); \ | |
177 | (tic)->t_prev = (q)->t_prev; \ | |
178 | (q)->t_prev->t_next = (tic); \ | |
179 | (q)->t_prev = (tic); \ | |
180 | } else { \ | |
181 | (tic)->t_prev = (tic)->t_next = (tic); \ | |
182 | (q) = (tic); \ | |
183 | } \ | |
184 | (tic)->t_flags |= XLOG_TIC_IN_Q; \ | |
185 | } | |
186 | #define XLOG_DEL_TICKETQ(q,tic) \ | |
187 | { \ | |
188 | if ((tic) == (tic)->t_next) { \ | |
189 | (q) = NULL; \ | |
190 | } else { \ | |
191 | (q) = (tic)->t_next; \ | |
192 | (tic)->t_next->t_prev = (tic)->t_prev; \ | |
193 | (tic)->t_prev->t_next = (tic)->t_next; \ | |
194 | } \ | |
195 | (tic)->t_next = (tic)->t_prev = NULL; \ | |
196 | (tic)->t_flags &= ~XLOG_TIC_IN_Q; \ | |
197 | } | |
198 | ||
199 | ||
200 | #define GRANT_LOCK(log) mutex_spinlock(&(log)->l_grant_lock) | |
201 | #define GRANT_UNLOCK(log, s) mutex_spinunlock(&(log)->l_grant_lock, s) | |
202 | #define LOG_LOCK(log) mutex_spinlock(&(log)->l_icloglock) | |
203 | #define LOG_UNLOCK(log, s) mutex_spinunlock(&(log)->l_icloglock, s) | |
204 | ||
205 | #define xlog_panic(s) {cmn_err(CE_PANIC, s); } | |
206 | #define xlog_exit(s) {cmn_err(CE_PANIC, s); } | |
207 | #define xlog_warn(s) {cmn_err(CE_WARN, s); } | |
208 | ||
209 | /* | |
210 | * In core log state | |
211 | */ | |
212 | #define XLOG_STATE_ACTIVE 0x0001 /* Current IC log being written to */ | |
213 | #define XLOG_STATE_WANT_SYNC 0x0002 /* Want to sync this iclog; no more writes */ | |
214 | #define XLOG_STATE_SYNCING 0x0004 /* This IC log is syncing */ | |
215 | #define XLOG_STATE_DONE_SYNC 0x0008 /* Done syncing to disk */ | |
216 | #define XLOG_STATE_DO_CALLBACK \ | |
217 | 0x0010 /* Process callback functions */ | |
218 | #define XLOG_STATE_CALLBACK 0x0020 /* Callback functions now */ | |
219 | #define XLOG_STATE_DIRTY 0x0040 /* Dirty IC log, not ready for ACTIVE status*/ | |
220 | #define XLOG_STATE_IOERROR 0x0080 /* IO error happened in sync'ing log */ | |
221 | #define XLOG_STATE_ALL 0x7FFF /* All possible valid flags */ | |
222 | #define XLOG_STATE_NOTUSED 0x8000 /* This IC log not being used */ | |
223 | #endif /* __KERNEL__ */ | |
224 | ||
225 | /* | |
226 | * Flags to log operation header | |
227 | * | |
228 | * The first write of a new transaction will be preceded with a start | |
229 | * record, XLOG_START_TRANS. Once a transaction is committed, a commit | |
230 | * record is written, XLOG_COMMIT_TRANS. If a single region can not fit into | |
231 | * the remainder of the current active in-core log, it is split up into | |
232 | * multiple regions. Each partial region will be marked with a | |
233 | * XLOG_CONTINUE_TRANS until the last one, which gets marked with XLOG_END_TRANS. | |
234 | * | |
235 | */ | |
236 | #define XLOG_START_TRANS 0x01 /* Start a new transaction */ | |
237 | #define XLOG_COMMIT_TRANS 0x02 /* Commit this transaction */ | |
238 | #define XLOG_CONTINUE_TRANS 0x04 /* Cont this trans into new region */ | |
239 | #define XLOG_WAS_CONT_TRANS 0x08 /* Cont this trans into new region */ | |
240 | #define XLOG_END_TRANS 0x10 /* End a continued transaction */ | |
241 | #define XLOG_UNMOUNT_TRANS 0x20 /* Unmount a filesystem transaction */ | |
242 | #define XLOG_SKIP_TRANS (XLOG_COMMIT_TRANS | XLOG_CONTINUE_TRANS | \ | |
243 | XLOG_WAS_CONT_TRANS | XLOG_END_TRANS | \ | |
244 | XLOG_UNMOUNT_TRANS) | |
245 | ||
246 | #ifdef __KERNEL__ | |
247 | /* | |
248 | * Flags to log ticket | |
249 | */ | |
250 | #define XLOG_TIC_INITED 0x1 /* has been initialized */ | |
251 | #define XLOG_TIC_PERM_RESERV 0x2 /* permanent reservation */ | |
252 | #define XLOG_TIC_IN_Q 0x4 | |
253 | #endif /* __KERNEL__ */ | |
254 | ||
255 | #define XLOG_UNMOUNT_TYPE 0x556e /* Un for Unmount */ | |
256 | ||
257 | /* | |
258 | * Flags for log structure | |
259 | */ | |
260 | #define XLOG_CHKSUM_MISMATCH 0x1 /* used only during recovery */ | |
261 | #define XLOG_ACTIVE_RECOVERY 0x2 /* in the middle of recovery */ | |
dfc130f3 | 262 | #define XLOG_RECOVERY_NEEDED 0x4 /* log was recovered */ |
2bd0ea18 NS |
263 | #define XLOG_IO_ERROR 0x8 /* log hit an I/O error, and being |
264 | shutdown */ | |
265 | typedef __uint32_t xlog_tid_t; | |
266 | ||
267 | ||
268 | #ifdef __KERNEL__ | |
269 | /* | |
270 | * Below are states for covering allocation transactions. | |
271 | * By covering, we mean changing the h_tail_lsn in the last on-disk | |
272 | * log write such that no allocation transactions will be re-done during | |
273 | * recovery after a system crash. Recovery starts at the last on-disk | |
274 | * log write. | |
275 | * | |
276 | * These states are used to insert dummy log entries to cover | |
277 | * space allocation transactions which can undo non-transactional changes | |
278 | * after a crash. Writes to a file with space | |
279 | * already allocated do not result in any transactions. Allocations | |
280 | * might include space beyond the EOF. So if we just push the EOF a | |
281 | * little, the last transaction for the file could contain the wrong | |
282 | * size. If there is no file system activity, after an allocation | |
283 | * transaction, and the system crashes, the allocation transaction | |
284 | * will get replayed and the file will be truncated. This could | |
285 | * be hours/days/... after the allocation occurred. | |
286 | * | |
287 | * The fix for this is to do two dummy transactions when the | |
288 | * system is idle. We need two dummy transaction because the h_tail_lsn | |
289 | * in the log record header needs to point beyond the last possible | |
290 | * non-dummy transaction. The first dummy changes the h_tail_lsn to | |
291 | * the first transaction before the dummy. The second dummy causes | |
292 | * h_tail_lsn to point to the first dummy. Recovery starts at h_tail_lsn. | |
5000d01d | 293 | * |
2bd0ea18 NS |
294 | * These dummy transactions get committed when everything |
295 | * is idle (after there has been some activity). | |
296 | * | |
297 | * There are 5 states used to control this. | |
298 | * | |
299 | * IDLE -- no logging has been done on the file system or | |
300 | * we are done covering previous transactions. | |
301 | * NEED -- logging has occurred and we need a dummy transaction | |
302 | * when the log becomes idle. | |
303 | * DONE -- we were in the NEED state and have committed a dummy | |
304 | * transaction. | |
305 | * NEED2 -- we detected that a dummy transaction has gone to the | |
306 | * on disk log with no other transactions. | |
307 | * DONE2 -- we committed a dummy transaction when in the NEED2 state. | |
308 | * | |
309 | * There are two places where we switch states: | |
310 | * | |
311 | * 1.) In xfs_sync, when we detect an idle log and are in NEED or NEED2. | |
312 | * We commit the dummy transaction and switch to DONE or DONE2, | |
5000d01d | 313 | * respectively. In all other states, we don't do anything. |
2bd0ea18 NS |
314 | * |
315 | * 2.) When we finish writing the on-disk log (xlog_state_clean_log). | |
316 | * | |
317 | * No matter what state we are in, if this isn't the dummy | |
318 | * transaction going out, the next state is NEED. | |
319 | * So, if we aren't in the DONE or DONE2 states, the next state | |
320 | * is NEED. We can't be finishing a write of the dummy record | |
321 | * unless it was committed and the state switched to DONE or DONE2. | |
5000d01d | 322 | * |
2bd0ea18 NS |
323 | * If we are in the DONE state and this was a write of the |
324 | * dummy transaction, we move to NEED2. | |
325 | * | |
326 | * If we are in the DONE2 state and this was a write of the | |
327 | * dummy transaction, we move to IDLE. | |
328 | * | |
329 | * | |
330 | * Writing only one dummy transaction can get appended to | |
331 | * one file space allocation. When this happens, the log recovery | |
332 | * code replays the space allocation and a file could be truncated. | |
333 | * This is why we have the NEED2 and DONE2 states before going idle. | |
334 | */ | |
335 | ||
336 | #define XLOG_STATE_COVER_IDLE 0 | |
337 | #define XLOG_STATE_COVER_NEED 1 | |
338 | #define XLOG_STATE_COVER_DONE 2 | |
339 | #define XLOG_STATE_COVER_NEED2 3 | |
340 | #define XLOG_STATE_COVER_DONE2 4 | |
341 | ||
342 | #define XLOG_COVER_OPS 5 | |
343 | ||
344 | typedef struct xlog_ticket { | |
345 | sv_t t_sema; /* sleep on this semaphore :20 */ | |
dfc130f3 | 346 | struct xlog_ticket *t_next; /* : 4 */ |
2bd0ea18 NS |
347 | struct xlog_ticket *t_prev; /* : 4 */ |
348 | xlog_tid_t t_tid; /* transaction identifier : 4 */ | |
349 | int t_curr_res; /* current reservation in bytes : 4 */ | |
dfc130f3 | 350 | int t_unit_res; /* unit reservation in bytes : 4 */ |
a7565da0 NS |
351 | __uint8_t t_ocnt; /* original count : 1 */ |
352 | __uint8_t t_cnt; /* current count : 1 */ | |
353 | __uint8_t t_clientid; /* who does this belong to; : 1 */ | |
354 | __uint8_t t_flags; /* properties of reservation : 1 */ | |
2bd0ea18 NS |
355 | } xlog_ticket_t; |
356 | #endif | |
357 | ||
358 | ||
359 | typedef struct xlog_op_header { | |
360 | xlog_tid_t oh_tid; /* transaction id of operation : 4 b */ | |
73bf5988 | 361 | int oh_len; /* bytes in data region : 4 b */ |
dfc130f3 | 362 | __uint8_t oh_clientid; /* who sent me this : 1 b */ |
5000d01d | 363 | __uint8_t oh_flags; /* : 1 b */ |
2bd0ea18 NS |
364 | ushort oh_res2; /* 32 bit align : 2 b */ |
365 | } xlog_op_header_t; | |
366 | ||
367 | ||
368 | /* valid values for h_fmt */ | |
369 | #define XLOG_FMT_UNKNOWN 0 | |
370 | #define XLOG_FMT_LINUX_LE 1 | |
371 | #define XLOG_FMT_LINUX_BE 2 | |
372 | #define XLOG_FMT_IRIX_BE 3 | |
373 | ||
374 | /* our fmt */ | |
375 | #if __BYTE_ORDER == __LITTLE_ENDIAN | |
376 | #define XLOG_FMT XLOG_FMT_LINUX_LE | |
377 | #else | |
378 | #if __BYTE_ORDER == __BIG_ENDIAN | |
379 | #define XLOG_FMT XLOG_FMT_LINUX_BE | |
380 | #else | |
381 | #error unknown byte order | |
382 | #endif | |
383 | #endif | |
384 | ||
385 | typedef struct xlog_rec_header { | |
386 | uint h_magicno; /* log record (LR) identifier : 4 */ | |
387 | uint h_cycle; /* write cycle of log : 4 */ | |
388 | int h_version; /* LR version : 4 */ | |
389 | int h_len; /* len in bytes; should be 64-bit aligned: 4 */ | |
390 | xfs_lsn_t h_lsn; /* lsn of this LR : 8 */ | |
391 | xfs_lsn_t h_tail_lsn; /* lsn of 1st LR w/ buffers not committed: 8 */ | |
392 | uint h_chksum; /* may not be used; non-zero if used : 4 */ | |
393 | int h_prev_block; /* block number to previous LR : 4 */ | |
dfc130f3 | 394 | int h_num_logops; /* number of log operations in this LR : 4 */ |
5000d01d SL |
395 | uint h_cycle_data[XLOG_HEADER_CYCLE_SIZE / BBSIZE]; |
396 | /* new fields */ | |
dfc130f3 RC |
397 | int h_fmt; /* format of log record : 4 */ |
398 | uuid_t h_fs_uuid; /* uuid of FS : 16 */ | |
399 | int h_size; /* iclog size : 4 */ | |
2bd0ea18 NS |
400 | } xlog_rec_header_t; |
401 | ||
73bf5988 SL |
402 | typedef struct xlog_rec_ext_header { |
403 | uint xh_cycle; /* write cycle of log : 4 */ | |
404 | uint xh_cycle_data[XLOG_HEADER_CYCLE_SIZE / BBSIZE]; /* : 256 */ | |
405 | } xlog_rec_ext_header_t; | |
2bd0ea18 NS |
406 | #ifdef __KERNEL__ |
407 | /* | |
dfc130f3 | 408 | * - A log record header is 512 bytes. There is plenty of room to grow the |
2bd0ea18 NS |
409 | * xlog_rec_header_t into the reserved space. |
410 | * - ic_data follows, so a write to disk can start at the beginning of | |
411 | * the iclog. | |
412 | * - ic_forcesema is used to implement synchronous forcing of the iclog to disk. | |
413 | * - ic_next is the pointer to the next iclog in the ring. | |
414 | * - ic_bp is a pointer to the buffer used to write this incore log to disk. | |
415 | * - ic_log is a pointer back to the global log structure. | |
416 | * - ic_callback is a linked list of callback function/argument pairs to be | |
417 | * called after an iclog finishes writing. | |
418 | * - ic_size is the full size of the header plus data. | |
419 | * - ic_offset is the current number of bytes written to in this iclog. | |
420 | * - ic_refcnt is bumped when someone is writing to the log. | |
421 | * - ic_state is the state of the iclog. | |
422 | */ | |
423 | typedef struct xlog_iclog_fields { | |
424 | sv_t ic_forcesema; | |
f9e56f43 | 425 | sv_t ic_writesema; |
2bd0ea18 NS |
426 | struct xlog_in_core *ic_next; |
427 | struct xlog_in_core *ic_prev; | |
5000d01d | 428 | struct xfs_buf *ic_bp; |
2bd0ea18 NS |
429 | struct log *ic_log; |
430 | xfs_log_callback_t *ic_callback; | |
431 | xfs_log_callback_t **ic_callback_tail; | |
432 | #ifdef DEBUG | |
433 | struct ktrace *ic_trace; | |
434 | #endif | |
5000d01d SL |
435 | int ic_size; |
436 | int ic_offset; | |
437 | int ic_refcnt; | |
2bd0ea18 NS |
438 | int ic_roundoff; |
439 | int ic_bwritecnt; | |
440 | ushort_t ic_state; | |
73bf5988 | 441 | char *ic_datap; /* pointer to iclog data */ |
2bd0ea18 NS |
442 | } xlog_iclog_fields_t; |
443 | ||
b7d5131f | 444 | typedef struct xlog_in_core2 { |
2bd0ea18 NS |
445 | union { |
446 | xlog_rec_header_t hic_header; | |
73bf5988 | 447 | xlog_rec_ext_header_t hic_xheader; |
2bd0ea18 | 448 | char hic_sector[XLOG_HEADER_SIZE]; |
b7d5131f | 449 | } ic_h; |
b7d5131f SL |
450 | } xlog_in_core_2_t; |
451 | ||
452 | typedef struct xlog_in_core { | |
453 | xlog_iclog_fields_t hic_fields; | |
454 | xlog_in_core_2_t *hic_data; | |
2bd0ea18 NS |
455 | } xlog_in_core_t; |
456 | ||
457 | /* | |
458 | * Defines to save our code from this glop. | |
459 | */ | |
dfc130f3 | 460 | #define ic_forcesema hic_fields.ic_forcesema |
f9e56f43 | 461 | #define ic_writesema hic_fields.ic_writesema |
dfc130f3 RC |
462 | #define ic_next hic_fields.ic_next |
463 | #define ic_prev hic_fields.ic_prev | |
464 | #define ic_bp hic_fields.ic_bp | |
465 | #define ic_log hic_fields.ic_log | |
466 | #define ic_callback hic_fields.ic_callback | |
467 | #define ic_callback_tail hic_fields.ic_callback_tail | |
468 | #define ic_trace hic_fields.ic_trace | |
469 | #define ic_size hic_fields.ic_size | |
470 | #define ic_offset hic_fields.ic_offset | |
471 | #define ic_refcnt hic_fields.ic_refcnt | |
472 | #define ic_roundoff hic_fields.ic_roundoff | |
473 | #define ic_bwritecnt hic_fields.ic_bwritecnt | |
474 | #define ic_state hic_fields.ic_state | |
73bf5988 | 475 | #define ic_datap hic_fields.ic_datap |
b7d5131f | 476 | #define ic_header hic_data->ic_h.hic_header |
2bd0ea18 NS |
477 | |
478 | /* | |
479 | * The reservation head lsn is not made up of a cycle number and block number. | |
480 | * Instead, it uses a cycle number and byte number. Logs don't expect to | |
481 | * overflow 31 bits worth of byte offset, so using a byte number will mean | |
482 | * that round off problems won't occur when releasing partial reservations. | |
483 | */ | |
484 | typedef struct log { | |
485 | /* The following block of fields are changed while holding icloglock */ | |
dfc130f3 | 486 | sema_t l_flushsema; /* iclog flushing semaphore */ |
2bd0ea18 NS |
487 | int l_flushcnt; /* # of procs waiting on this sema */ |
488 | int l_ticket_cnt; /* free ticket count */ | |
489 | int l_ticket_tcnt; /* total ticket count */ | |
490 | int l_covered_state;/* state of "covering disk log entries" */ | |
dfc130f3 | 491 | xlog_ticket_t *l_freelist; /* free list of tickets */ |
2bd0ea18 | 492 | xlog_ticket_t *l_unmount_free;/* kmem_free these addresses */ |
dfc130f3 RC |
493 | xlog_ticket_t *l_tail; /* free list of tickets */ |
494 | xlog_in_core_t *l_iclog; /* head log queue */ | |
495 | lock_t l_icloglock; /* grab to change iclog state */ | |
496 | xfs_lsn_t l_tail_lsn; /* lsn of 1st LR w/ unflush buffers */ | |
2bd0ea18 | 497 | xfs_lsn_t l_last_sync_lsn;/* lsn of last LR on disk */ |
dfc130f3 RC |
498 | struct xfs_mount *l_mp; /* mount point */ |
499 | struct xfs_buf *l_xbuf; /* extra buffer for log wrapping */ | |
500 | dev_t l_dev; /* dev_t of log */ | |
501 | xfs_daddr_t l_logBBstart; /* start block of log */ | |
502 | int l_logsize; /* size of log in bytes */ | |
503 | int l_logBBsize; /* size of log in 512 byte chunks */ | |
2bd0ea18 | 504 | int l_roundoff; /* round off error of all iclogs */ |
dfc130f3 RC |
505 | int l_curr_cycle; /* Cycle number of log writes */ |
506 | int l_prev_cycle; /* Cycle # b4 last block increment */ | |
507 | int l_curr_block; /* current logical block of log */ | |
508 | int l_prev_block; /* previous logical block of log */ | |
2bd0ea18 NS |
509 | int l_iclog_size; /* size of log in bytes */ |
510 | int l_iclog_size_log;/* log power size of log */ | |
511 | int l_iclog_bufs; /* number of iclog buffers */ | |
512 | ||
513 | /* The following field are used for debugging; need to hold icloglock */ | |
514 | char *l_iclog_bak[XLOG_MAX_ICLOGS]; | |
515 | ||
516 | /* The following block of fields are changed while holding grant_lock */ | |
517 | lock_t l_grant_lock; /* protects below fields */ | |
518 | xlog_ticket_t *l_reserve_headq; /* */ | |
519 | xlog_ticket_t *l_write_headq; /* */ | |
520 | int l_grant_reserve_cycle; /* */ | |
521 | int l_grant_reserve_bytes; /* */ | |
522 | int l_grant_write_cycle; /* */ | |
523 | int l_grant_write_bytes; /* */ | |
524 | ||
525 | /* The following fields don't need locking */ | |
526 | #ifdef DEBUG | |
527 | struct ktrace *l_trace; | |
528 | struct ktrace *l_grant_trace; | |
529 | #endif | |
530 | uint l_flags; | |
531 | uint l_quotaoffs_flag;/* XFS_DQ_*, if QUOTAOFFs found */ | |
5000d01d | 532 | struct xfs_buf_cancel **l_buf_cancel_table; |
dfc130f3 RC |
533 | int l_iclog_hsize; /* size of iclog header */ |
534 | int l_iclog_heads; /* number of iclog header sectors */ | |
2bd0ea18 NS |
535 | } xlog_t; |
536 | ||
537 | ||
538 | /* common routines */ | |
4ca431fc | 539 | extern xfs_lsn_t xlog_assign_tail_lsn(struct xfs_mount *mp); |
2bd0ea18 NS |
540 | extern int xlog_find_head(xlog_t *log, xfs_daddr_t *head_blk); |
541 | extern int xlog_find_tail(xlog_t *log, | |
542 | xfs_daddr_t *head_blk, | |
543 | xfs_daddr_t *tail_blk, | |
544 | int readonly); | |
545 | extern int xlog_print_find_oldest(xlog_t *log, xfs_daddr_t *last_blk); | |
546 | extern int xlog_recover(xlog_t *log, int readonly); | |
547 | extern int xlog_recover_finish(xlog_t *log, int mfsi_flags); | |
548 | extern void xlog_pack_data(xlog_t *log, xlog_in_core_t *iclog); | |
549 | extern struct xfs_buf *xlog_get_bp(int,xfs_mount_t *); | |
550 | extern void xlog_put_bp(struct xfs_buf *); | |
551 | extern int xlog_bread(xlog_t *, xfs_daddr_t blkno, int bblks, struct xfs_buf *bp); | |
552 | extern void xlog_recover_process_iunlinks(xlog_t *log); | |
553 | ||
554 | #define XLOG_TRACE_GRAB_FLUSH 1 | |
555 | #define XLOG_TRACE_REL_FLUSH 2 | |
556 | #define XLOG_TRACE_SLEEP_FLUSH 3 | |
557 | #define XLOG_TRACE_WAKE_FLUSH 4 | |
558 | ||
559 | #endif /* __KERNEL__ */ | |
560 | ||
561 | #endif /* __XFS_LOG_PRIV_H__ */ |