1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * Copyright (C) 2001-2003 Cluster File Systems, Inc.
5 * Author: Andreas Dilger <adilger@clusterfs.com>
7 * This file is part of Lustre, http://www.lustre.org.
9 * Lustre is free software; you can redistribute it and/or
10 * modify it under the terms of version 2 of the GNU General Public
11 * License as published by the Free Software Foundation.
13 * Lustre is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License
19 * along with Lustre; if not, write to the Free Software
20 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
22 * OST<->MDS recovery logging infrastructure.
24 * Invariants in implementation:
25 * - we do not share logs among different OST<->MDS connections, so that
26 * if an OST or MDS fails it need only look at log(s) relevant to itself
29 #define DEBUG_SUBSYSTEM S_LOG
38 #include <liblustre.h>
41 #include <linux/lustre_log.h>
42 #include <portals/list.h>
44 /* Create a new log handle and add it to the open list.
45 * This log handle will be closed when all of the records in it are removed.
47 * Assumes caller has already pushed us into the kernel context and is locking.
49 static struct llog_handle *llog_cat_new_log(struct llog_handle *cathandle)
51 struct llog_handle *loghandle;
52 struct llog_log_hdr *llh;
53 struct llog_logid_rec rec;
54 int rc, index, bitmap_size;
57 llh = cathandle->lgh_hdr;
58 bitmap_size = sizeof(llh->llh_bitmap) * 8;
60 index = (cathandle->lgh_last_idx + 1) % bitmap_size;
62 /* maximum number of available slots in catalog is bitmap_size - 2 */
63 if (llh->llh_cat_idx == cpu_to_le32(index)) {
64 CERROR("no free catalog slots for log...\n");
65 RETURN(ERR_PTR(-ENOSPC));
69 if (ext2_set_bit(index, llh->llh_bitmap)) {
70 CERROR("argh, index %u already set in log bitmap?\n",
72 LBUG(); /* should never happen */
74 cathandle->lgh_last_idx = index;
75 llh->llh_count = cpu_to_le32(le32_to_cpu(llh->llh_count) + 1);
76 llh->llh_tail.lrt_index = cpu_to_le32(index);
79 rc = llog_open(cathandle->lgh_ctxt, &loghandle, NULL, NULL,
82 CERROR("cannot create new log, error = %d\n", rc);
86 rc = llog_init_handle(loghandle,
87 LLOG_F_IS_PLAIN | LLOG_F_ZAP_WHEN_EMPTY,
88 &cathandle->lgh_hdr->llh_tgtuuid);
90 GOTO(out_destroy, rc);
92 CDEBUG(D_HA, "new recovery log "LPX64":%x for index %u of catalog "
93 LPX64"\n", loghandle->lgh_id.lgl_oid, loghandle->lgh_id.lgl_ogen,
94 index, cathandle->lgh_id.lgl_oid);
95 /* build the record for this log in the catalog */
96 rec.lid_hdr.lrh_len = cpu_to_le32(sizeof(rec));
97 rec.lid_hdr.lrh_index = cpu_to_le32(index);
98 rec.lid_hdr.lrh_type = cpu_to_le32(LLOG_LOGID_MAGIC);
99 rec.lid_id = loghandle->lgh_id;
100 rec.lid_tail.lrt_len = cpu_to_le32(sizeof(rec));
101 rec.lid_tail.lrt_index = cpu_to_le32(index);
103 /* update the catalog: header and record */
104 rc = llog_write_rec(cathandle, &rec.lid_hdr,
105 &loghandle->u.phd.phd_cookie, 1, NULL, index);
107 GOTO(out_destroy, rc);
109 loghandle->lgh_hdr->llh_cat_idx = cpu_to_le32(index);
110 cathandle->u.chd.chd_current_log = loghandle;
111 LASSERT(list_empty(&loghandle->u.phd.phd_entry));
112 list_add_tail(&loghandle->u.phd.phd_entry, &cathandle->u.chd.chd_head);
116 llog_destroy(loghandle);
121 /* Open an existent log handle and add it to the open list.
122 * This log handle will be closed when all of the records in it are removed.
124 * Assumes caller has already pushed us into the kernel context and is locking.
125 * We return a lock on the handle to ensure nobody yanks it from us.
127 int llog_cat_id2handle(struct llog_handle *cathandle, struct llog_handle **res,
128 struct llog_logid *logid)
130 struct llog_handle *loghandle;
134 if (cathandle == NULL)
137 list_for_each_entry(loghandle, &cathandle->u.chd.chd_head,
139 struct llog_logid *cgl = &loghandle->lgh_id;
140 if (cgl->lgl_oid == logid->lgl_oid) {
141 if (cgl->lgl_ogen != logid->lgl_ogen) {
142 CERROR("log "LPX64" generation %x != %x\n",
143 logid->lgl_oid, cgl->lgl_ogen,
147 loghandle->u.phd.phd_cat_handle = cathandle;
152 rc = llog_open(cathandle->lgh_ctxt, &loghandle, logid, NULL, 0);
154 CERROR("error opening log id "LPX64":%x: rc %d\n",
155 logid->lgl_oid, logid->lgl_ogen, rc);
157 rc = llog_init_handle(loghandle, LLOG_F_IS_PLAIN, NULL);
159 list_add(&loghandle->u.phd.phd_entry,
160 &cathandle->u.chd.chd_head);
164 loghandle->u.phd.phd_cat_handle = cathandle;
165 loghandle->u.phd.phd_cookie.lgc_lgl = cathandle->lgh_id;
166 loghandle->u.phd.phd_cookie.lgc_index =
167 le32_to_cpu(loghandle->lgh_hdr->llh_cat_idx);
174 EXPORT_SYMBOL(llog_cat_id2handle);
176 int llog_cat_put(struct llog_handle *cathandle)
178 struct llog_handle *loghandle, *n;
182 list_for_each_entry_safe(loghandle, n, &cathandle->u.chd.chd_head,
184 int err = llog_close(loghandle);
186 CERROR("error closing loghandle\n");
188 rc = llog_close(cathandle);
191 EXPORT_SYMBOL(llog_cat_put);
193 /* Return the currently active log handle. If the current log handle doesn't
194 * have enough space left for the current record, start a new one.
196 * If reclen is 0, we only want to know what the currently active log is,
197 * otherwise we get a lock on this log so nobody can steal our space.
199 * Assumes caller has already pushed us into the kernel context and is locking.
201 * NOTE: loghandle is write-locked upon successful return
203 static struct llog_handle *llog_cat_current_log(struct llog_handle *cathandle,
206 struct llog_handle *loghandle = NULL;
209 down_read(&cathandle->lgh_lock);
210 loghandle = cathandle->u.chd.chd_current_log;
212 struct llog_log_hdr *llh = loghandle->lgh_hdr;
213 if (loghandle->lgh_last_idx < (sizeof(llh->llh_bitmap)*8) - 1) {
214 down_write(&loghandle->lgh_lock);
215 up_read(&cathandle->lgh_lock);
221 down_write(&loghandle->lgh_lock);
222 up_read(&cathandle->lgh_lock);
225 up_read(&cathandle->lgh_lock);
227 /* time to create new log */
229 /* first, we have to make sure the state hasn't changed */
230 down_write(&cathandle->lgh_lock);
231 loghandle = cathandle->u.chd.chd_current_log;
233 struct llog_log_hdr *llh = loghandle->lgh_hdr;
234 if (loghandle->lgh_last_idx < (sizeof(llh->llh_bitmap)*8) - 1) {
235 down_write(&loghandle->lgh_lock);
236 up_write(&cathandle->lgh_lock);
241 CDEBUG(D_INODE, "creating new log\n");
242 loghandle = llog_cat_new_log(cathandle);
243 if (!IS_ERR(loghandle))
244 down_write(&loghandle->lgh_lock);
245 up_write(&cathandle->lgh_lock);
249 /* Add a single record to the recovery log(s) using a catalog
250 * Returns as llog_write_record
252 * Assumes caller has already pushed us into the kernel context.
254 int llog_cat_add_rec(struct llog_handle *cathandle, struct llog_rec_hdr *rec,
255 struct llog_cookie *reccookie, void *buf)
257 struct llog_handle *loghandle;
261 LASSERT(le32_to_cpu(rec->lrh_len) <= LLOG_CHUNK_SIZE);
262 loghandle = llog_cat_current_log(cathandle, 1);
263 if (IS_ERR(loghandle))
264 RETURN(PTR_ERR(loghandle));
265 /* loghandle is already locked by llog_cat_current_log() for us */
266 rc = llog_write_rec(loghandle, rec, reccookie, 1, buf, -1);
267 up_write(&loghandle->lgh_lock);
271 EXPORT_SYMBOL(llog_cat_add_rec);
273 /* For each cookie in the cookie array, we clear the log in-use bit and either:
274 * - the log is empty, so mark it free in the catalog header and delete it
275 * - the log is not empty, just write out the log header
277 * The cookies may be in different log files, so we need to get new logs
280 * Assumes caller has already pushed us into the kernel context.
282 int llog_cat_cancel_records(struct llog_handle *cathandle, int count,
283 struct llog_cookie *cookies)
285 int i, index, rc = 0;
288 down_write(&cathandle->lgh_lock);
289 for (i = 0; i < count; i++, cookies++) {
290 struct llog_handle *loghandle;
291 struct llog_logid *lgl = &cookies->lgc_lgl;
293 rc = llog_cat_id2handle(cathandle, &loghandle, lgl);
295 CERROR("Cannot find log "LPX64"\n", lgl->lgl_oid);
299 down_write(&loghandle->lgh_lock);
300 rc = llog_cancel_rec(loghandle, cookies->lgc_index);
301 up_write(&loghandle->lgh_lock);
303 if (rc == 1) { /* log has been destroyed */
304 index = loghandle->u.phd.phd_cookie.lgc_index;
305 if (cathandle->u.chd.chd_current_log == loghandle)
306 cathandle->u.chd.chd_current_log = NULL;
307 llog_free_handle(loghandle);
310 llog_cat_set_first_idx(cathandle, index);
311 rc = llog_cancel_rec(cathandle, index);
313 CDEBUG(D_HA, "cancel plain log at index %u "
314 "of catalog "LPX64"\n",
315 index, cathandle->lgh_id.lgl_oid);
318 up_write(&cathandle->lgh_lock);
322 EXPORT_SYMBOL(llog_cat_cancel_records);
324 static int llog_cat_process_cb(struct llog_handle *cat_llh,
325 struct llog_rec_hdr *rec, void *data)
327 struct llog_process_data *d = data;
328 struct llog_logid_rec *lir = (struct llog_logid_rec *)rec;
329 struct llog_handle *llh;
332 if (le32_to_cpu(rec->lrh_type) != LLOG_LOGID_MAGIC) {
333 CERROR("invalid record in catalog\n");
336 CWARN("processing log "LPX64":%x at index %u of catalog "LPX64"\n",
337 lir->lid_id.lgl_oid, lir->lid_id.lgl_ogen,
338 le32_to_cpu(rec->lrh_index), cat_llh->lgh_id.lgl_oid);
340 rc = llog_cat_id2handle(cat_llh, &llh, &lir->lid_id);
342 CERROR("Cannot find handle for log "LPX64"\n",
343 lir->lid_id.lgl_oid);
347 rc = llog_process(llh, d->lpd_cb, d->lpd_data, NULL);
351 int llog_cat_process(struct llog_handle *cat_llh, llog_cb_t cb, void *data)
353 struct llog_process_data d;
354 struct llog_process_cat_data cd;
355 struct llog_log_hdr *llh = cat_llh->lgh_hdr;
359 LASSERT(llh->llh_flags &cpu_to_le32(LLOG_F_IS_CAT));
363 if (llh->llh_cat_idx > cat_llh->lgh_last_idx) {
364 CWARN("catalog "LPX64" crosses index zero\n",
365 cat_llh->lgh_id.lgl_oid);
367 cd.first_idx = le32_to_cpu(llh->llh_cat_idx);
369 rc = llog_process(cat_llh, llog_cat_process_cb, &d, &cd);
374 cd.last_idx = cat_llh->lgh_last_idx;
375 rc = llog_process(cat_llh, llog_cat_process_cb, &d, &cd);
377 rc = llog_process(cat_llh, llog_cat_process_cb, &d, NULL);
382 EXPORT_SYMBOL(llog_cat_process);
384 static int llog_cat_reverse_process_cb(struct llog_handle *cat_llh,
385 struct llog_rec_hdr *rec, void *data)
387 struct llog_process_data *d = data;
388 struct llog_logid_rec *lir = (struct llog_logid_rec *)rec;
389 struct llog_handle *llh;
392 if (le32_to_cpu(rec->lrh_type) != LLOG_LOGID_MAGIC) {
393 CERROR("invalid record in catalog\n");
396 CWARN("processing log "LPX64":%x at index %u of catalog "LPX64"\n",
397 lir->lid_id.lgl_oid, lir->lid_id.lgl_ogen,
398 le32_to_cpu(rec->lrh_index), cat_llh->lgh_id.lgl_oid);
400 rc = llog_cat_id2handle(cat_llh, &llh, &lir->lid_id);
402 CERROR("Cannot find handle for log "LPX64"\n",
403 lir->lid_id.lgl_oid);
407 rc = llog_reverse_process(llh, d->lpd_cb, d->lpd_data, NULL);
411 int llog_cat_reverse_process(struct llog_handle *cat_llh,
412 llog_cb_t cb, void *data)
414 struct llog_process_data d;
415 struct llog_process_cat_data cd;
416 struct llog_log_hdr *llh = cat_llh->lgh_hdr;
420 LASSERT(llh->llh_flags &cpu_to_le32(LLOG_F_IS_CAT));
424 if (llh->llh_cat_idx > cat_llh->lgh_last_idx) {
425 CWARN("catalog "LPX64" crosses index zero\n",
426 cat_llh->lgh_id.lgl_oid);
429 cd.last_idx = cat_llh->lgh_last_idx;
430 rc = llog_reverse_process(cat_llh, llog_cat_reverse_process_cb,
435 cd.first_idx = le32_to_cpu(llh->llh_cat_idx);
437 rc = llog_reverse_process(cat_llh, llog_cat_reverse_process_cb,
440 rc = llog_reverse_process(cat_llh, llog_cat_reverse_process_cb,
446 EXPORT_SYMBOL(llog_cat_reverse_process);
448 int llog_cat_set_first_idx(struct llog_handle *cathandle, int index)
450 struct llog_log_hdr *llh = cathandle->lgh_hdr;
451 int i, bitmap_size, idx;
454 bitmap_size = sizeof(llh->llh_bitmap) * 8;
455 if (llh->llh_cat_idx == cpu_to_le32(index - 1)) {
456 idx = le32_to_cpu(llh->llh_cat_idx) + 1;
457 llh->llh_cat_idx = cpu_to_le32(idx);
458 if (idx == cathandle->lgh_last_idx)
460 for (i = (index + 1) % bitmap_size;
461 i != cathandle->lgh_last_idx;
462 i = (i + 1) % bitmap_size) {
463 if (!ext2_test_bit(i, llh->llh_bitmap)) {
464 idx = le32_to_cpu(llh->llh_cat_idx) + 1;
465 llh->llh_cat_idx = cpu_to_le32(idx);
467 llh->llh_cat_idx = 0;
473 CDEBUG(D_HA, "set catalog "LPX64" first idx %u\n",
474 cathandle->lgh_id.lgl_oid,le32_to_cpu(llh->llh_cat_idx));
479 EXPORT_SYMBOL(llog_cat_set_first_idx);
481 int llog_catalog_add(struct llog_ctxt *ctxt, struct llog_rec_hdr *rec,
482 void *buf, struct llog_cookie *logcookies,
483 int numcookies, void *data)
485 struct llog_handle *cathandle;
489 cathandle = ctxt->loc_handle;
490 LASSERT(cathandle != NULL);
492 rc = llog_cat_add_rec(cathandle, rec, logcookies, buf);
494 CERROR("write one catalog record failed: %d\n", rc);
497 EXPORT_SYMBOL(llog_catalog_add);
499 int llog_catalog_cancel(struct llog_ctxt *ctxt, int count,
500 struct llog_cookie *cookies, int flags, void *data)
502 struct llog_handle *cathandle;
506 if (cookies == NULL || count == 0)
508 cathandle = ctxt->loc_handle;
509 LASSERT(cathandle != NULL);
510 rc = llog_cat_cancel_records(cathandle, count, cookies);
513 EXPORT_SYMBOL(llog_catalog_cancel);
515 int llog_catalog_setup(struct llog_ctxt **res, char *name,
516 struct obd_export *exp,
517 struct lvfs_run_ctxt *lvfs_ctxt,
518 struct fsfilt_operations *fsops,
519 struct dentry *logs_de,
520 struct dentry *objects_de)
522 struct llog_ctxt *ctxt;
523 struct llog_catid catid;
524 struct llog_handle *handle;
529 OBD_ALLOC(ctxt, sizeof(*ctxt));
535 /* marking this ctxt alone. */
537 ctxt->loc_fsops = fsops;
538 ctxt->loc_lvfs_ctxt = lvfs_ctxt;
540 ctxt->loc_logs_dir = logs_de;
541 ctxt->loc_objects_dir = objects_de;
542 ctxt->loc_logops = &llog_lvfs_ops;
543 ctxt->loc_logops->lop_add = llog_catalog_add;
544 ctxt->loc_logops->lop_cancel = llog_catalog_cancel;
546 memset(&catid, 0, sizeof(struct llog_catid));
547 rc = llog_get_cat_list(lvfs_ctxt, fsops, name, 1, &catid);
549 CERROR("error llog_get_cat_list rc: %d\n", rc);
552 if (catid.lci_logid.lgl_oid)
553 rc = llog_open(ctxt, &handle, &catid.lci_logid, NULL,
556 rc = llog_open(ctxt, &handle, NULL, NULL, OBD_LLOG_FL_CREATE);
558 catid.lci_logid = handle->lgh_id;
563 ctxt->loc_handle = handle;
564 rc = llog_init_handle(handle, LLOG_F_IS_CAT, NULL);
568 rc = llog_put_cat_list(lvfs_ctxt, fsops, name, 1, &catid);
570 CERROR("error llog_get_cat_list rc: %d\n", rc);
573 OBD_FREE(ctxt, sizeof(*ctxt));
576 EXPORT_SYMBOL(llog_catalog_setup);
578 int llog_catalog_cleanup(struct llog_ctxt *ctxt)
580 struct llog_handle *cathandle, *n, *loghandle;
581 struct llog_log_hdr *llh;
588 cathandle = ctxt->loc_handle;
590 list_for_each_entry_safe(loghandle, n,
591 &cathandle->u.chd.chd_head,
593 llh = loghandle->lgh_hdr;
594 if ((le32_to_cpu(llh->llh_flags) &
595 LLOG_F_ZAP_WHEN_EMPTY) &&
596 (le32_to_cpu(llh->llh_count) == 1)) {
597 rc = llog_destroy(loghandle);
599 CERROR("failure destroying log during "
600 "cleanup: %d\n", rc);
602 index = loghandle->u.phd.phd_cookie.lgc_index;
603 llog_free_handle(loghandle);
606 llog_cat_set_first_idx(cathandle, index);
607 rc = llog_cancel_rec(cathandle, index);
609 CDEBUG(D_HA, "cancel plain log at index"
610 " %u of catalog "LPX64"\n",
611 index,cathandle->lgh_id.lgl_oid);
614 llog_cat_put(ctxt->loc_handle);
618 EXPORT_SYMBOL(llog_catalog_cleanup);
620 int llog_cat_half_bottom(struct llog_cookie *cookie, struct llog_handle *handle)
622 struct llog_handle *loghandle;
623 struct llog_logid *lgl = &cookie->lgc_lgl;
626 down_read(&handle->lgh_lock);
627 rc = llog_cat_id2handle(handle, &loghandle, lgl);
630 if (2 * loghandle->lgh_hdr->llh_cat_idx <=
631 handle->lgh_last_idx + handle->lgh_hdr->llh_cat_idx + 1)
636 up_read(&handle->lgh_lock);
639 EXPORT_SYMBOL(llog_cat_half_bottom);