]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Copyright (c) 2000-2003 Silicon Graphics, Inc. All Rights Reserved. | |
3 | * | |
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. | |
7 | * | |
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. | |
11 | * | |
12 | * Further, this software is distributed without any warranty that it is | |
13 | * free of the rightful claim of any third person regarding infringement | |
14 | * or the like. Any license provided herein, whether implied or | |
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. | |
18 | * | |
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. | |
22 | * | |
23 | * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy, | |
24 | * Mountain View, CA 94043, or: | |
25 | * | |
26 | * http://www.sgi.com | |
27 | * | |
28 | * For further information regarding this notice, see: | |
29 | * | |
30 | * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/ | |
31 | */ | |
32 | #ifndef __XFS_LOG_PRIV_H__ | |
33 | #define __XFS_LOG_PRIV_H__ | |
34 | ||
35 | struct xfs_buf; | |
36 | struct ktrace; | |
37 | struct log; | |
38 | struct xfs_buf_cancel; | |
39 | struct xfs_mount; | |
40 | ||
41 | /* | |
42 | * Macros, structures, prototypes for internal log manager use. | |
43 | */ | |
44 | ||
45 | #define XLOG_MIN_ICLOGS 2 | |
46 | #define XLOG_MED_ICLOGS 4 | |
47 | #define XLOG_MAX_ICLOGS 8 | |
48 | #define XLOG_CALLBACK_SIZE 10 | |
49 | #define XLOG_HEADER_MAGIC_NUM 0xFEEDbabe /* Invalid cycle number */ | |
50 | #define XLOG_VERSION_1 1 | |
51 | #define XLOG_VERSION_2 2 /* Large IClogs, Log sunit */ | |
52 | #define XLOG_VERSION_OKBITS (XLOG_VERSION_1 | XLOG_VERSION_2) | |
53 | #define XLOG_RECORD_BSIZE (16*1024) /* eventually 32k */ | |
54 | #define XLOG_BIG_RECORD_BSIZE (32*1024) /* 32k buffers */ | |
55 | #define XLOG_MAX_RECORD_BSIZE (256*1024) | |
56 | #define XLOG_HEADER_CYCLE_SIZE (32*1024) /* cycle data in header */ | |
57 | #define XLOG_RECORD_BSHIFT 14 /* 16384 == 1 << 14 */ | |
58 | #define XLOG_BIG_RECORD_BSHIFT 15 /* 32k == 1 << 15 */ | |
59 | #define XLOG_MAX_RECORD_BSHIFT 18 /* 256k == 1 << 18 */ | |
60 | #define XLOG_BTOLSUNIT(log, b) (((b)+(log)->l_mp->m_sb.sb_logsunit-1) / \ | |
61 | (log)->l_mp->m_sb.sb_logsunit) | |
62 | #define XLOG_LSUNITTOB(log, su) ((su) * (log)->l_mp->m_sb.sb_logsunit) | |
63 | ||
64 | #define XLOG_HEADER_SIZE 512 | |
65 | ||
66 | #define XLOG_REC_SHIFT(log) \ | |
67 | BTOBB(1 << (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? \ | |
68 | XLOG_MAX_RECORD_BSHIFT : XLOG_BIG_RECORD_BSHIFT)) | |
69 | #define XLOG_TOTAL_REC_SHIFT(log) \ | |
70 | BTOBB(XLOG_MAX_ICLOGS << (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? \ | |
71 | XLOG_MAX_RECORD_BSHIFT : XLOG_BIG_RECORD_BSHIFT)) | |
72 | ||
73 | /* | |
74 | * set lsns | |
75 | */ | |
76 | ||
77 | #define ASSIGN_ANY_LSN_HOST(lsn,cycle,block) \ | |
78 | { \ | |
79 | (lsn) = ((xfs_lsn_t)(cycle)<<32)|(block); \ | |
80 | } | |
81 | #define ASSIGN_ANY_LSN_DISK(lsn,cycle,block) \ | |
82 | { \ | |
83 | INT_SET(((uint *)&(lsn))[0], ARCH_CONVERT, (cycle)); \ | |
84 | INT_SET(((uint *)&(lsn))[1], ARCH_CONVERT, (block)); \ | |
85 | } | |
86 | #define ASSIGN_LSN(lsn,log) \ | |
87 | ASSIGN_ANY_LSN_DISK(lsn,(log)->l_curr_cycle,(log)->l_curr_block); | |
88 | ||
89 | #define XLOG_SET(f,b) (((f) & (b)) == (b)) | |
90 | ||
91 | #define GET_CYCLE(ptr, arch) \ | |
92 | (INT_GET(*(uint *)(ptr), arch) == XLOG_HEADER_MAGIC_NUM ? \ | |
93 | INT_GET(*((uint *)(ptr)+1), arch) : \ | |
94 | INT_GET(*(uint *)(ptr), arch) \ | |
95 | ) | |
96 | ||
97 | #define BLK_AVG(blk1, blk2) ((blk1+blk2) >> 1) | |
98 | ||
99 | ||
100 | #ifdef __KERNEL__ | |
101 | ||
102 | /* | |
103 | * get client id from packed copy. | |
104 | * | |
105 | * this hack is here because the xlog_pack code copies four bytes | |
106 | * of xlog_op_header containing the fields oh_clientid, oh_flags | |
107 | * and oh_res2 into the packed copy. | |
108 | * | |
109 | * later on this four byte chunk is treated as an int and the | |
110 | * client id is pulled out. | |
111 | * | |
112 | * this has endian issues, of course. | |
113 | */ | |
114 | ||
115 | #if __BYTE_ORDER == __LITTLE_ENDIAN | |
116 | #define GET_CLIENT_ID(i,arch) \ | |
117 | ((i) & 0xff) | |
118 | #else | |
119 | #define GET_CLIENT_ID(i,arch) \ | |
120 | ((i) >> 24) | |
121 | #endif | |
122 | ||
123 | #if XFS_WANT_FUNCS || (XFS_WANT_SPACE && XFSSO_XLOG_GRANT_SUB_SPACE) | |
124 | void xlog_grant_sub_space(struct log *log, int bytes, int type); | |
125 | #define XLOG_GRANT_SUB_SPACE(log,bytes,type) \ | |
126 | xlog_grant_sub_space(log,bytes,type) | |
127 | #else | |
128 | #define XLOG_GRANT_SUB_SPACE(log,bytes,type) \ | |
129 | { \ | |
130 | if (type == 'w') { \ | |
131 | (log)->l_grant_write_bytes -= (bytes); \ | |
132 | if ((log)->l_grant_write_bytes < 0) { \ | |
133 | (log)->l_grant_write_bytes += (log)->l_logsize; \ | |
134 | (log)->l_grant_write_cycle--; \ | |
135 | } \ | |
136 | } else { \ | |
137 | (log)->l_grant_reserve_bytes -= (bytes); \ | |
138 | if ((log)->l_grant_reserve_bytes < 0) { \ | |
139 | (log)->l_grant_reserve_bytes += (log)->l_logsize;\ | |
140 | (log)->l_grant_reserve_cycle--; \ | |
141 | } \ | |
142 | } \ | |
143 | } | |
144 | #endif | |
145 | #if XFS_WANT_FUNCS || (XFS_WANT_SPACE && XFSSO_XLOG_GRANT_ADD_SPACE) | |
146 | void xlog_grant_add_space(struct log *log, int bytes, int type); | |
147 | #define XLOG_GRANT_ADD_SPACE(log,bytes,type) \ | |
148 | xlog_grant_add_space(log,bytes,type) | |
149 | #else | |
150 | #define XLOG_GRANT_ADD_SPACE(log,bytes,type) \ | |
151 | { \ | |
152 | if (type == 'w') { \ | |
153 | (log)->l_grant_write_bytes += (bytes); \ | |
154 | if ((log)->l_grant_write_bytes > (log)->l_logsize) { \ | |
155 | (log)->l_grant_write_bytes -= (log)->l_logsize; \ | |
156 | (log)->l_grant_write_cycle++; \ | |
157 | } \ | |
158 | } else { \ | |
159 | (log)->l_grant_reserve_bytes += (bytes); \ | |
160 | if ((log)->l_grant_reserve_bytes > (log)->l_logsize) { \ | |
161 | (log)->l_grant_reserve_bytes -= (log)->l_logsize;\ | |
162 | (log)->l_grant_reserve_cycle++; \ | |
163 | } \ | |
164 | } \ | |
165 | } | |
166 | #endif | |
167 | #define XLOG_INS_TICKETQ(q,tic) \ | |
168 | { \ | |
169 | if (q) { \ | |
170 | (tic)->t_next = (q); \ | |
171 | (tic)->t_prev = (q)->t_prev; \ | |
172 | (q)->t_prev->t_next = (tic); \ | |
173 | (q)->t_prev = (tic); \ | |
174 | } else { \ | |
175 | (tic)->t_prev = (tic)->t_next = (tic); \ | |
176 | (q) = (tic); \ | |
177 | } \ | |
178 | (tic)->t_flags |= XLOG_TIC_IN_Q; \ | |
179 | } | |
180 | #define XLOG_DEL_TICKETQ(q,tic) \ | |
181 | { \ | |
182 | if ((tic) == (tic)->t_next) { \ | |
183 | (q) = NULL; \ | |
184 | } else { \ | |
185 | (q) = (tic)->t_next; \ | |
186 | (tic)->t_next->t_prev = (tic)->t_prev; \ | |
187 | (tic)->t_prev->t_next = (tic)->t_next; \ | |
188 | } \ | |
189 | (tic)->t_next = (tic)->t_prev = NULL; \ | |
190 | (tic)->t_flags &= ~XLOG_TIC_IN_Q; \ | |
191 | } | |
192 | ||
193 | ||
194 | #define GRANT_LOCK(log) mutex_spinlock(&(log)->l_grant_lock) | |
195 | #define GRANT_UNLOCK(log, s) mutex_spinunlock(&(log)->l_grant_lock, s) | |
196 | #define LOG_LOCK(log) mutex_spinlock(&(log)->l_icloglock) | |
197 | #define LOG_UNLOCK(log, s) mutex_spinunlock(&(log)->l_icloglock, s) | |
198 | ||
199 | #define xlog_panic(args...) cmn_err(CE_PANIC, ## args) | |
200 | #define xlog_exit(args...) cmn_err(CE_PANIC, ## args) | |
201 | #define xlog_warn(args...) cmn_err(CE_WARN, ## args) | |
202 | ||
203 | /* | |
204 | * In core log state | |
205 | */ | |
206 | #define XLOG_STATE_ACTIVE 0x0001 /* Current IC log being written to */ | |
207 | #define XLOG_STATE_WANT_SYNC 0x0002 /* Want to sync this iclog; no more writes */ | |
208 | #define XLOG_STATE_SYNCING 0x0004 /* This IC log is syncing */ | |
209 | #define XLOG_STATE_DONE_SYNC 0x0008 /* Done syncing to disk */ | |
210 | #define XLOG_STATE_DO_CALLBACK \ | |
211 | 0x0010 /* Process callback functions */ | |
212 | #define XLOG_STATE_CALLBACK 0x0020 /* Callback functions now */ | |
213 | #define XLOG_STATE_DIRTY 0x0040 /* Dirty IC log, not ready for ACTIVE status*/ | |
214 | #define XLOG_STATE_IOERROR 0x0080 /* IO error happened in sync'ing log */ | |
215 | #define XLOG_STATE_ALL 0x7FFF /* All possible valid flags */ | |
216 | #define XLOG_STATE_NOTUSED 0x8000 /* This IC log not being used */ | |
217 | #endif /* __KERNEL__ */ | |
218 | ||
219 | /* | |
220 | * Flags to log operation header | |
221 | * | |
222 | * The first write of a new transaction will be preceded with a start | |
223 | * record, XLOG_START_TRANS. Once a transaction is committed, a commit | |
224 | * record is written, XLOG_COMMIT_TRANS. If a single region can not fit into | |
225 | * the remainder of the current active in-core log, it is split up into | |
226 | * multiple regions. Each partial region will be marked with a | |
227 | * XLOG_CONTINUE_TRANS until the last one, which gets marked with XLOG_END_TRANS. | |
228 | * | |
229 | */ | |
230 | #define XLOG_START_TRANS 0x01 /* Start a new transaction */ | |
231 | #define XLOG_COMMIT_TRANS 0x02 /* Commit this transaction */ | |
232 | #define XLOG_CONTINUE_TRANS 0x04 /* Cont this trans into new region */ | |
233 | #define XLOG_WAS_CONT_TRANS 0x08 /* Cont this trans into new region */ | |
234 | #define XLOG_END_TRANS 0x10 /* End a continued transaction */ | |
235 | #define XLOG_UNMOUNT_TRANS 0x20 /* Unmount a filesystem transaction */ | |
236 | #define XLOG_SKIP_TRANS (XLOG_COMMIT_TRANS | XLOG_CONTINUE_TRANS | \ | |
237 | XLOG_WAS_CONT_TRANS | XLOG_END_TRANS | \ | |
238 | XLOG_UNMOUNT_TRANS) | |
239 | ||
240 | #ifdef __KERNEL__ | |
241 | /* | |
242 | * Flags to log ticket | |
243 | */ | |
244 | #define XLOG_TIC_INITED 0x1 /* has been initialized */ | |
245 | #define XLOG_TIC_PERM_RESERV 0x2 /* permanent reservation */ | |
246 | #define XLOG_TIC_IN_Q 0x4 | |
247 | #endif /* __KERNEL__ */ | |
248 | ||
249 | #define XLOG_UNMOUNT_TYPE 0x556e /* Un for Unmount */ | |
250 | ||
251 | /* | |
252 | * Flags for log structure | |
253 | */ | |
254 | #define XLOG_CHKSUM_MISMATCH 0x1 /* used only during recovery */ | |
255 | #define XLOG_ACTIVE_RECOVERY 0x2 /* in the middle of recovery */ | |
256 | #define XLOG_RECOVERY_NEEDED 0x4 /* log was recovered */ | |
257 | #define XLOG_IO_ERROR 0x8 /* log hit an I/O error, and being | |
258 | shutdown */ | |
259 | typedef __uint32_t xlog_tid_t; | |
260 | ||
261 | ||
262 | #ifdef __KERNEL__ | |
263 | /* | |
264 | * Below are states for covering allocation transactions. | |
265 | * By covering, we mean changing the h_tail_lsn in the last on-disk | |
266 | * log write such that no allocation transactions will be re-done during | |
267 | * recovery after a system crash. Recovery starts at the last on-disk | |
268 | * log write. | |
269 | * | |
270 | * These states are used to insert dummy log entries to cover | |
271 | * space allocation transactions which can undo non-transactional changes | |
272 | * after a crash. Writes to a file with space | |
273 | * already allocated do not result in any transactions. Allocations | |
274 | * might include space beyond the EOF. So if we just push the EOF a | |
275 | * little, the last transaction for the file could contain the wrong | |
276 | * size. If there is no file system activity, after an allocation | |
277 | * transaction, and the system crashes, the allocation transaction | |
278 | * will get replayed and the file will be truncated. This could | |
279 | * be hours/days/... after the allocation occurred. | |
280 | * | |
281 | * The fix for this is to do two dummy transactions when the | |
282 | * system is idle. We need two dummy transaction because the h_tail_lsn | |
283 | * in the log record header needs to point beyond the last possible | |
284 | * non-dummy transaction. The first dummy changes the h_tail_lsn to | |
285 | * the first transaction before the dummy. The second dummy causes | |
286 | * h_tail_lsn to point to the first dummy. Recovery starts at h_tail_lsn. | |
287 | * | |
288 | * These dummy transactions get committed when everything | |
289 | * is idle (after there has been some activity). | |
290 | * | |
291 | * There are 5 states used to control this. | |
292 | * | |
293 | * IDLE -- no logging has been done on the file system or | |
294 | * we are done covering previous transactions. | |
295 | * NEED -- logging has occurred and we need a dummy transaction | |
296 | * when the log becomes idle. | |
297 | * DONE -- we were in the NEED state and have committed a dummy | |
298 | * transaction. | |
299 | * NEED2 -- we detected that a dummy transaction has gone to the | |
300 | * on disk log with no other transactions. | |
301 | * DONE2 -- we committed a dummy transaction when in the NEED2 state. | |
302 | * | |
303 | * There are two places where we switch states: | |
304 | * | |
305 | * 1.) In xfs_sync, when we detect an idle log and are in NEED or NEED2. | |
306 | * We commit the dummy transaction and switch to DONE or DONE2, | |
307 | * respectively. In all other states, we don't do anything. | |
308 | * | |
309 | * 2.) When we finish writing the on-disk log (xlog_state_clean_log). | |
310 | * | |
311 | * No matter what state we are in, if this isn't the dummy | |
312 | * transaction going out, the next state is NEED. | |
313 | * So, if we aren't in the DONE or DONE2 states, the next state | |
314 | * is NEED. We can't be finishing a write of the dummy record | |
315 | * unless it was committed and the state switched to DONE or DONE2. | |
316 | * | |
317 | * If we are in the DONE state and this was a write of the | |
318 | * dummy transaction, we move to NEED2. | |
319 | * | |
320 | * If we are in the DONE2 state and this was a write of the | |
321 | * dummy transaction, we move to IDLE. | |
322 | * | |
323 | * | |
324 | * Writing only one dummy transaction can get appended to | |
325 | * one file space allocation. When this happens, the log recovery | |
326 | * code replays the space allocation and a file could be truncated. | |
327 | * This is why we have the NEED2 and DONE2 states before going idle. | |
328 | */ | |
329 | ||
330 | #define XLOG_STATE_COVER_IDLE 0 | |
331 | #define XLOG_STATE_COVER_NEED 1 | |
332 | #define XLOG_STATE_COVER_DONE 2 | |
333 | #define XLOG_STATE_COVER_NEED2 3 | |
334 | #define XLOG_STATE_COVER_DONE2 4 | |
335 | ||
336 | #define XLOG_COVER_OPS 5 | |
337 | ||
7e9c6396 TS |
338 | |
339 | /* Ticket reservation region accounting */ | |
340 | #if defined(XFS_LOG_RES_DEBUG) | |
341 | #define XLOG_TIC_LEN_MAX 15 | |
342 | #define XLOG_TIC_RESET_RES(t) ((t)->t_res_num = \ | |
343 | (t)->t_res_arr_sum = (t)->t_res_num_ophdrs = 0) | |
344 | #define XLOG_TIC_ADD_OPHDR(t) ((t)->t_res_num_ophdrs++) | |
345 | #define XLOG_TIC_ADD_REGION(t, len, type) \ | |
346 | do { \ | |
347 | if ((t)->t_res_num == XLOG_TIC_LEN_MAX) { \ | |
348 | /* add to overflow and start again */ \ | |
349 | (t)->t_res_o_flow += (t)->t_res_arr_sum; \ | |
350 | (t)->t_res_num = 0; \ | |
351 | (t)->t_res_arr_sum = 0; \ | |
352 | } \ | |
353 | (t)->t_res_arr[(t)->t_res_num].r_len = (len); \ | |
354 | (t)->t_res_arr[(t)->t_res_num].r_type = (type); \ | |
355 | (t)->t_res_arr_sum += (len); \ | |
356 | (t)->t_res_num++; \ | |
357 | } while (0) | |
358 | ||
359 | /* | |
360 | * Reservation region | |
361 | * As would be stored in xfs_log_iovec but without the i_addr which | |
362 | * we don't care about. | |
363 | */ | |
364 | typedef struct xlog_res { | |
365 | uint r_len; | |
366 | uint r_type; | |
367 | } xlog_res_t; | |
368 | #else | |
369 | #define XLOG_TIC_RESET_RES(t) | |
370 | #define XLOG_TIC_ADD_OPHDR(t) | |
371 | #define XLOG_TIC_ADD_REGION(t, len, type) | |
372 | #endif | |
373 | ||
374 | ||
1da177e4 | 375 | typedef struct xlog_ticket { |
7e9c6396 TS |
376 | sv_t t_sema; /* sleep on this semaphore : 20 */ |
377 | struct xlog_ticket *t_next; /* :4|8 */ | |
378 | struct xlog_ticket *t_prev; /* :4|8 */ | |
379 | xlog_tid_t t_tid; /* transaction identifier : 4 */ | |
380 | int t_curr_res; /* current reservation in bytes : 4 */ | |
381 | int t_unit_res; /* unit reservation in bytes : 4 */ | |
382 | char t_ocnt; /* original count : 1 */ | |
383 | char t_cnt; /* current count : 1 */ | |
384 | char t_clientid; /* who does this belong to; : 1 */ | |
385 | char t_flags; /* properties of reservation : 1 */ | |
386 | uint t_trans_type; /* transaction type : 4 */ | |
387 | ||
388 | #if defined (XFS_LOG_RES_DEBUG) | |
389 | /* reservation array fields */ | |
390 | uint t_res_num; /* num in array : 4 */ | |
391 | xlog_res_t t_res_arr[XLOG_TIC_LEN_MAX]; /* array of res : X */ | |
392 | uint t_res_num_ophdrs; /* num op hdrs : 4 */ | |
393 | uint t_res_arr_sum; /* array sum : 4 */ | |
394 | uint t_res_o_flow; /* sum overflow : 4 */ | |
395 | #endif | |
1da177e4 | 396 | } xlog_ticket_t; |
7e9c6396 | 397 | |
1da177e4 LT |
398 | #endif |
399 | ||
400 | ||
401 | typedef struct xlog_op_header { | |
402 | xlog_tid_t oh_tid; /* transaction id of operation : 4 b */ | |
403 | int oh_len; /* bytes in data region : 4 b */ | |
404 | __uint8_t oh_clientid; /* who sent me this : 1 b */ | |
405 | __uint8_t oh_flags; /* : 1 b */ | |
406 | ushort oh_res2; /* 32 bit align : 2 b */ | |
407 | } xlog_op_header_t; | |
408 | ||
409 | ||
410 | /* valid values for h_fmt */ | |
411 | #define XLOG_FMT_UNKNOWN 0 | |
412 | #define XLOG_FMT_LINUX_LE 1 | |
413 | #define XLOG_FMT_LINUX_BE 2 | |
414 | #define XLOG_FMT_IRIX_BE 3 | |
415 | ||
416 | /* our fmt */ | |
417 | #if __BYTE_ORDER == __LITTLE_ENDIAN | |
418 | #define XLOG_FMT XLOG_FMT_LINUX_LE | |
419 | #else | |
420 | #if __BYTE_ORDER == __BIG_ENDIAN | |
421 | #define XLOG_FMT XLOG_FMT_LINUX_BE | |
422 | #else | |
423 | #error unknown byte order | |
424 | #endif | |
425 | #endif | |
426 | ||
427 | typedef struct xlog_rec_header { | |
428 | uint h_magicno; /* log record (LR) identifier : 4 */ | |
429 | uint h_cycle; /* write cycle of log : 4 */ | |
430 | int h_version; /* LR version : 4 */ | |
431 | int h_len; /* len in bytes; should be 64-bit aligned: 4 */ | |
432 | xfs_lsn_t h_lsn; /* lsn of this LR : 8 */ | |
433 | xfs_lsn_t h_tail_lsn; /* lsn of 1st LR w/ buffers not committed: 8 */ | |
434 | uint h_chksum; /* may not be used; non-zero if used : 4 */ | |
435 | int h_prev_block; /* block number to previous LR : 4 */ | |
436 | int h_num_logops; /* number of log operations in this LR : 4 */ | |
437 | uint h_cycle_data[XLOG_HEADER_CYCLE_SIZE / BBSIZE]; | |
438 | /* new fields */ | |
439 | int h_fmt; /* format of log record : 4 */ | |
440 | uuid_t h_fs_uuid; /* uuid of FS : 16 */ | |
441 | int h_size; /* iclog size : 4 */ | |
442 | } xlog_rec_header_t; | |
443 | ||
444 | typedef struct xlog_rec_ext_header { | |
445 | uint xh_cycle; /* write cycle of log : 4 */ | |
446 | uint xh_cycle_data[XLOG_HEADER_CYCLE_SIZE / BBSIZE]; /* : 256 */ | |
447 | } xlog_rec_ext_header_t; | |
448 | ||
449 | #ifdef __KERNEL__ | |
450 | /* | |
451 | * - A log record header is 512 bytes. There is plenty of room to grow the | |
452 | * xlog_rec_header_t into the reserved space. | |
453 | * - ic_data follows, so a write to disk can start at the beginning of | |
454 | * the iclog. | |
455 | * - ic_forcesema is used to implement synchronous forcing of the iclog to disk. | |
456 | * - ic_next is the pointer to the next iclog in the ring. | |
457 | * - ic_bp is a pointer to the buffer used to write this incore log to disk. | |
458 | * - ic_log is a pointer back to the global log structure. | |
459 | * - ic_callback is a linked list of callback function/argument pairs to be | |
460 | * called after an iclog finishes writing. | |
461 | * - ic_size is the full size of the header plus data. | |
462 | * - ic_offset is the current number of bytes written to in this iclog. | |
463 | * - ic_refcnt is bumped when someone is writing to the log. | |
464 | * - ic_state is the state of the iclog. | |
465 | */ | |
466 | typedef struct xlog_iclog_fields { | |
467 | sv_t ic_forcesema; | |
468 | sv_t ic_writesema; | |
469 | struct xlog_in_core *ic_next; | |
470 | struct xlog_in_core *ic_prev; | |
471 | struct xfs_buf *ic_bp; | |
472 | struct log *ic_log; | |
473 | xfs_log_callback_t *ic_callback; | |
474 | xfs_log_callback_t **ic_callback_tail; | |
475 | #ifdef XFS_LOG_TRACE | |
476 | struct ktrace *ic_trace; | |
477 | #endif | |
478 | int ic_size; | |
479 | int ic_offset; | |
480 | int ic_refcnt; | |
481 | int ic_bwritecnt; | |
482 | ushort_t ic_state; | |
483 | char *ic_datap; /* pointer to iclog data */ | |
484 | } xlog_iclog_fields_t; | |
485 | ||
486 | typedef union xlog_in_core2 { | |
487 | xlog_rec_header_t hic_header; | |
488 | xlog_rec_ext_header_t hic_xheader; | |
489 | char hic_sector[XLOG_HEADER_SIZE]; | |
490 | } xlog_in_core_2_t; | |
491 | ||
492 | typedef struct xlog_in_core { | |
493 | xlog_iclog_fields_t hic_fields; | |
494 | xlog_in_core_2_t *hic_data; | |
495 | } xlog_in_core_t; | |
496 | ||
497 | /* | |
498 | * Defines to save our code from this glop. | |
499 | */ | |
500 | #define ic_forcesema hic_fields.ic_forcesema | |
501 | #define ic_writesema hic_fields.ic_writesema | |
502 | #define ic_next hic_fields.ic_next | |
503 | #define ic_prev hic_fields.ic_prev | |
504 | #define ic_bp hic_fields.ic_bp | |
505 | #define ic_log hic_fields.ic_log | |
506 | #define ic_callback hic_fields.ic_callback | |
507 | #define ic_callback_tail hic_fields.ic_callback_tail | |
508 | #define ic_trace hic_fields.ic_trace | |
509 | #define ic_size hic_fields.ic_size | |
510 | #define ic_offset hic_fields.ic_offset | |
511 | #define ic_refcnt hic_fields.ic_refcnt | |
512 | #define ic_bwritecnt hic_fields.ic_bwritecnt | |
513 | #define ic_state hic_fields.ic_state | |
514 | #define ic_datap hic_fields.ic_datap | |
515 | #define ic_header hic_data->hic_header | |
516 | ||
517 | /* | |
518 | * The reservation head lsn is not made up of a cycle number and block number. | |
519 | * Instead, it uses a cycle number and byte number. Logs don't expect to | |
520 | * overflow 31 bits worth of byte offset, so using a byte number will mean | |
521 | * that round off problems won't occur when releasing partial reservations. | |
522 | */ | |
523 | typedef struct log { | |
524 | /* The following block of fields are changed while holding icloglock */ | |
525 | sema_t l_flushsema; /* iclog flushing semaphore */ | |
526 | int l_flushcnt; /* # of procs waiting on this | |
527 | * sema */ | |
528 | int l_ticket_cnt; /* free ticket count */ | |
529 | int l_ticket_tcnt; /* total ticket count */ | |
530 | int l_covered_state;/* state of "covering disk | |
531 | * log entries" */ | |
532 | xlog_ticket_t *l_freelist; /* free list of tickets */ | |
533 | xlog_ticket_t *l_unmount_free;/* kmem_free these addresses */ | |
534 | xlog_ticket_t *l_tail; /* free list of tickets */ | |
535 | xlog_in_core_t *l_iclog; /* head log queue */ | |
536 | lock_t l_icloglock; /* grab to change iclog state */ | |
537 | xfs_lsn_t l_tail_lsn; /* lsn of 1st LR with unflushed | |
538 | * buffers */ | |
539 | xfs_lsn_t l_last_sync_lsn;/* lsn of last LR on disk */ | |
540 | struct xfs_mount *l_mp; /* mount point */ | |
541 | struct xfs_buf *l_xbuf; /* extra buffer for log | |
542 | * wrapping */ | |
543 | struct xfs_buftarg *l_targ; /* buftarg of log */ | |
544 | xfs_daddr_t l_logBBstart; /* start block of log */ | |
545 | int l_logsize; /* size of log in bytes */ | |
546 | int l_logBBsize; /* size of log in BB chunks */ | |
547 | int l_curr_cycle; /* Cycle number of log writes */ | |
548 | int l_prev_cycle; /* Cycle number before last | |
549 | * block increment */ | |
550 | int l_curr_block; /* current logical log block */ | |
551 | int l_prev_block; /* previous logical log block */ | |
552 | int l_iclog_size; /* size of log in bytes */ | |
553 | int l_iclog_size_log; /* log power size of log */ | |
554 | int l_iclog_bufs; /* number of iclog buffers */ | |
555 | ||
556 | /* The following field are used for debugging; need to hold icloglock */ | |
557 | char *l_iclog_bak[XLOG_MAX_ICLOGS]; | |
558 | ||
559 | /* The following block of fields are changed while holding grant_lock */ | |
560 | lock_t l_grant_lock; | |
561 | xlog_ticket_t *l_reserve_headq; | |
562 | xlog_ticket_t *l_write_headq; | |
563 | int l_grant_reserve_cycle; | |
564 | int l_grant_reserve_bytes; | |
565 | int l_grant_write_cycle; | |
566 | int l_grant_write_bytes; | |
567 | ||
568 | /* The following fields don't need locking */ | |
569 | #ifdef XFS_LOG_TRACE | |
570 | struct ktrace *l_trace; | |
571 | struct ktrace *l_grant_trace; | |
572 | #endif | |
573 | uint l_flags; | |
574 | uint l_quotaoffs_flag; /* XFS_DQ_*, for QUOTAOFFs */ | |
575 | struct xfs_buf_cancel **l_buf_cancel_table; | |
576 | int l_iclog_hsize; /* size of iclog header */ | |
577 | int l_iclog_heads; /* # of iclog header sectors */ | |
578 | uint l_sectbb_log; /* log2 of sector size in BBs */ | |
579 | uint l_sectbb_mask; /* sector size (in BBs) | |
580 | * alignment mask */ | |
581 | } xlog_t; | |
582 | ||
583 | ||
584 | /* common routines */ | |
585 | extern xfs_lsn_t xlog_assign_tail_lsn(struct xfs_mount *mp); | |
1da177e4 LT |
586 | extern int xlog_find_tail(xlog_t *log, |
587 | xfs_daddr_t *head_blk, | |
588 | xfs_daddr_t *tail_blk, | |
589 | int readonly); | |
590 | extern int xlog_recover(xlog_t *log, int readonly); | |
591 | extern int xlog_recover_finish(xlog_t *log, int mfsi_flags); | |
592 | extern void xlog_pack_data(xlog_t *log, xlog_in_core_t *iclog, int); | |
593 | extern void xlog_recover_process_iunlinks(xlog_t *log); | |
594 | ||
595 | extern struct xfs_buf *xlog_get_bp(xlog_t *, int); | |
596 | extern void xlog_put_bp(struct xfs_buf *); | |
597 | extern int xlog_bread(xlog_t *, xfs_daddr_t, int, struct xfs_buf *); | |
1da177e4 LT |
598 | |
599 | /* iclog tracing */ | |
600 | #define XLOG_TRACE_GRAB_FLUSH 1 | |
601 | #define XLOG_TRACE_REL_FLUSH 2 | |
602 | #define XLOG_TRACE_SLEEP_FLUSH 3 | |
603 | #define XLOG_TRACE_WAKE_FLUSH 4 | |
604 | ||
605 | #endif /* __KERNEL__ */ | |
606 | ||
607 | #endif /* __XFS_LOG_PRIV_H__ */ |