]>
Commit | Line | Data |
---|---|---|
7c673cae FG |
1 | // -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*- |
2 | // vim: ts=8 sw=2 smarttab | |
3 | /* | |
4 | * Ceph - scalable distributed file system | |
5 | * | |
6 | * Copyright (C) 2004-2006 Sage Weil <sage@newdream.net> | |
7 | * Copyright (c) 2015 Hewlett-Packard Development Company, L.P. | |
8 | * | |
9 | * This is free software; you can redistribute it and/or | |
10 | * modify it under the terms of the GNU Lesser General Public | |
11 | * License version 2.1, as published by the Free Software | |
12 | * Foundation. See file COPYING. | |
13 | * | |
14 | */ | |
15 | #include "include/compat.h" | |
16 | #include "include/int_types.h" | |
17 | #include "boost/tuple/tuple.hpp" | |
18 | ||
19 | #include <unistd.h> | |
20 | #include <stdlib.h> | |
21 | #include <sys/types.h> | |
22 | #include <sys/stat.h> | |
23 | #include <fcntl.h> | |
24 | #include <sys/file.h> | |
25 | #include <errno.h> | |
26 | #include <dirent.h> | |
27 | #include <sys/ioctl.h> | |
28 | ||
29 | #if defined(__linux__) | |
30 | #include <linux/fs.h> | |
31 | #endif | |
32 | ||
33 | #include <iostream> | |
34 | #include <map> | |
35 | ||
36 | #include "include/linux_fiemap.h" | |
37 | ||
38 | #include "common/xattr.h" | |
39 | #include "chain_xattr.h" | |
40 | ||
41 | #if defined(DARWIN) || defined(__FreeBSD__) | |
42 | #include <sys/param.h> | |
43 | #include <sys/mount.h> | |
44 | #endif // DARWIN | |
45 | ||
46 | ||
47 | #include <fstream> | |
48 | #include <sstream> | |
49 | ||
50 | #include "FileStore.h" | |
51 | #include "GenericFileStoreBackend.h" | |
52 | #include "BtrfsFileStoreBackend.h" | |
53 | #include "XfsFileStoreBackend.h" | |
54 | #include "ZFSFileStoreBackend.h" | |
55 | #include "common/BackTrace.h" | |
56 | #include "include/types.h" | |
57 | #include "FileJournal.h" | |
58 | ||
59 | #include "osd/osd_types.h" | |
60 | #include "include/color.h" | |
61 | #include "include/buffer.h" | |
62 | ||
63 | #include "common/Timer.h" | |
64 | #include "common/debug.h" | |
65 | #include "common/errno.h" | |
66 | #include "common/run_cmd.h" | |
67 | #include "common/safe_io.h" | |
68 | #include "common/perf_counters.h" | |
69 | #include "common/sync_filesystem.h" | |
70 | #include "common/fd.h" | |
71 | #include "HashIndex.h" | |
72 | #include "DBObjectMap.h" | |
73 | #include "kv/KeyValueDB.h" | |
74 | ||
75 | #include "common/ceph_crypto.h" | |
76 | using ceph::crypto::SHA1; | |
77 | ||
78 | #include "include/assert.h" | |
79 | ||
80 | #include "common/config.h" | |
81 | #include "common/blkdev.h" | |
82 | ||
83 | #ifdef WITH_LTTNG | |
84 | #define TRACEPOINT_DEFINE | |
85 | #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE | |
86 | #include "tracing/objectstore.h" | |
87 | #undef TRACEPOINT_PROBE_DYNAMIC_LINKAGE | |
88 | #undef TRACEPOINT_DEFINE | |
89 | #else | |
90 | #define tracepoint(...) | |
91 | #endif | |
92 | ||
93 | #define dout_context cct | |
94 | #define dout_subsys ceph_subsys_filestore | |
95 | #undef dout_prefix | |
96 | #define dout_prefix *_dout << "filestore(" << basedir << ") " | |
97 | ||
98 | #define COMMIT_SNAP_ITEM "snap_%llu" | |
99 | #define CLUSTER_SNAP_ITEM "clustersnap_%s" | |
100 | ||
101 | #define REPLAY_GUARD_XATTR "user.cephos.seq" | |
102 | #define GLOBAL_REPLAY_GUARD_XATTR "user.cephos.gseq" | |
103 | ||
104 | // XATTR_SPILL_OUT_NAME as a xattr is used to maintain that indicates whether | |
105 | // xattrs spill over into DBObjectMap, if XATTR_SPILL_OUT_NAME exists in file | |
106 | // xattrs and the value is "no", it indicates no xattrs in DBObjectMap | |
107 | #define XATTR_SPILL_OUT_NAME "user.cephos.spill_out" | |
108 | #define XATTR_NO_SPILL_OUT "0" | |
109 | #define XATTR_SPILL_OUT "1" | |
31f18b77 | 110 | #define __FUNC__ __func__ << "(" << __LINE__ << ")" |
7c673cae FG |
111 | |
112 | //Initial features in new superblock. | |
113 | static CompatSet get_fs_initial_compat_set() { | |
114 | CompatSet::FeatureSet ceph_osd_feature_compat; | |
115 | CompatSet::FeatureSet ceph_osd_feature_ro_compat; | |
116 | CompatSet::FeatureSet ceph_osd_feature_incompat; | |
117 | return CompatSet(ceph_osd_feature_compat, ceph_osd_feature_ro_compat, | |
118 | ceph_osd_feature_incompat); | |
119 | } | |
120 | ||
121 | //Features are added here that this FileStore supports. | |
122 | static CompatSet get_fs_supported_compat_set() { | |
123 | CompatSet compat = get_fs_initial_compat_set(); | |
124 | //Any features here can be set in code, but not in initial superblock | |
125 | compat.incompat.insert(CEPH_FS_FEATURE_INCOMPAT_SHARDS); | |
126 | return compat; | |
127 | } | |
128 | ||
129 | int FileStore::validate_hobject_key(const hobject_t &obj) const | |
130 | { | |
131 | unsigned len = LFNIndex::get_max_escaped_name_len(obj); | |
132 | return len > m_filestore_max_xattr_value_size ? -ENAMETOOLONG : 0; | |
133 | } | |
134 | ||
135 | int FileStore::get_block_device_fsid(CephContext* cct, const string& path, | |
136 | uuid_d *fsid) | |
137 | { | |
138 | // make sure we don't try to use aio or direct_io (and get annoying | |
139 | // error messages from failing to do so); performance implications | |
140 | // should be irrelevant for this use | |
141 | FileJournal j(cct, *fsid, 0, 0, path.c_str(), false, false); | |
142 | return j.peek_fsid(*fsid); | |
143 | } | |
144 | ||
145 | void FileStore::FSPerfTracker::update_from_perfcounters( | |
146 | PerfCounters &logger) | |
147 | { | |
148 | os_commit_latency.consume_next( | |
149 | logger.get_tavg_ms( | |
150 | l_filestore_journal_latency)); | |
151 | os_apply_latency.consume_next( | |
152 | logger.get_tavg_ms( | |
153 | l_filestore_apply_latency)); | |
154 | } | |
155 | ||
156 | ||
157 | ostream& operator<<(ostream& out, const FileStore::OpSequencer& s) | |
158 | { | |
159 | return out << *s.parent; | |
160 | } | |
161 | ||
162 | int FileStore::get_cdir(const coll_t& cid, char *s, int len) | |
163 | { | |
164 | const string &cid_str(cid.to_str()); | |
165 | return snprintf(s, len, "%s/current/%s", basedir.c_str(), cid_str.c_str()); | |
166 | } | |
167 | ||
168 | int FileStore::get_index(const coll_t& cid, Index *index) | |
169 | { | |
170 | int r = index_manager.get_index(cid, basedir, index); | |
171 | assert(!m_filestore_fail_eio || r != -EIO); | |
172 | return r; | |
173 | } | |
174 | ||
175 | int FileStore::init_index(const coll_t& cid) | |
176 | { | |
177 | char path[PATH_MAX]; | |
178 | get_cdir(cid, path, sizeof(path)); | |
179 | int r = index_manager.init_index(cid, path, target_version); | |
180 | assert(!m_filestore_fail_eio || r != -EIO); | |
181 | return r; | |
182 | } | |
183 | ||
184 | int FileStore::lfn_find(const ghobject_t& oid, const Index& index, IndexedPath *path) | |
185 | { | |
186 | IndexedPath path2; | |
187 | if (!path) | |
188 | path = &path2; | |
189 | int r, exist; | |
190 | assert(NULL != index.index); | |
191 | r = (index.index)->lookup(oid, path, &exist); | |
192 | if (r < 0) { | |
193 | assert(!m_filestore_fail_eio || r != -EIO); | |
194 | return r; | |
195 | } | |
196 | if (!exist) | |
197 | return -ENOENT; | |
198 | return 0; | |
199 | } | |
200 | ||
201 | int FileStore::lfn_truncate(const coll_t& cid, const ghobject_t& oid, off_t length) | |
202 | { | |
203 | FDRef fd; | |
204 | int r = lfn_open(cid, oid, false, &fd); | |
205 | if (r < 0) | |
206 | return r; | |
207 | r = ::ftruncate(**fd, length); | |
208 | if (r < 0) | |
209 | r = -errno; | |
210 | if (r >= 0 && m_filestore_sloppy_crc) { | |
211 | int rc = backend->_crc_update_truncate(**fd, length); | |
212 | assert(rc >= 0); | |
213 | } | |
214 | lfn_close(fd); | |
215 | assert(!m_filestore_fail_eio || r != -EIO); | |
216 | return r; | |
217 | } | |
218 | ||
219 | int FileStore::lfn_stat(const coll_t& cid, const ghobject_t& oid, struct stat *buf) | |
220 | { | |
221 | IndexedPath path; | |
222 | Index index; | |
223 | int r = get_index(cid, &index); | |
224 | if (r < 0) | |
225 | return r; | |
226 | ||
227 | assert(NULL != index.index); | |
228 | RWLock::RLocker l((index.index)->access_lock); | |
229 | ||
230 | r = lfn_find(oid, index, &path); | |
231 | if (r < 0) | |
232 | return r; | |
233 | r = ::stat(path->path(), buf); | |
234 | if (r < 0) | |
235 | r = -errno; | |
236 | return r; | |
237 | } | |
238 | ||
239 | int FileStore::lfn_open(const coll_t& cid, | |
240 | const ghobject_t& oid, | |
241 | bool create, | |
242 | FDRef *outfd, | |
243 | Index *index) | |
244 | { | |
245 | assert(outfd); | |
246 | int r = 0; | |
247 | bool need_lock = true; | |
248 | int flags = O_RDWR; | |
249 | ||
250 | if (create) | |
251 | flags |= O_CREAT; | |
252 | if (cct->_conf->filestore_odsync_write) { | |
253 | flags |= O_DSYNC; | |
254 | } | |
255 | ||
256 | Index index2; | |
257 | if (!index) { | |
258 | index = &index2; | |
259 | } | |
260 | if (!((*index).index)) { | |
261 | r = get_index(cid, index); | |
262 | if (r < 0) { | |
31f18b77 | 263 | dout(10) << __FUNC__ << ": could not get index r = " << r << dendl; |
7c673cae FG |
264 | return r; |
265 | } | |
266 | } else { | |
267 | need_lock = false; | |
268 | } | |
269 | ||
270 | int fd, exist; | |
271 | assert(NULL != (*index).index); | |
272 | if (need_lock) { | |
273 | ((*index).index)->access_lock.get_write(); | |
274 | } | |
275 | if (!replaying) { | |
276 | *outfd = fdcache.lookup(oid); | |
277 | if (*outfd) { | |
278 | if (need_lock) { | |
279 | ((*index).index)->access_lock.put_write(); | |
280 | } | |
281 | return 0; | |
282 | } | |
283 | } | |
284 | ||
285 | ||
286 | IndexedPath path2; | |
287 | IndexedPath *path = &path2; | |
288 | ||
289 | r = (*index)->lookup(oid, path, &exist); | |
290 | if (r < 0) { | |
291 | derr << "could not find " << oid << " in index: " | |
292 | << cpp_strerror(-r) << dendl; | |
293 | goto fail; | |
294 | } | |
295 | ||
296 | r = ::open((*path)->path(), flags, 0644); | |
297 | if (r < 0) { | |
298 | r = -errno; | |
299 | dout(10) << "error opening file " << (*path)->path() << " with flags=" | |
300 | << flags << ": " << cpp_strerror(-r) << dendl; | |
301 | goto fail; | |
302 | } | |
303 | fd = r; | |
304 | if (create && (!exist)) { | |
305 | r = (*index)->created(oid, (*path)->path()); | |
306 | if (r < 0) { | |
307 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
308 | derr << "error creating " << oid << " (" << (*path)->path() | |
309 | << ") in index: " << cpp_strerror(-r) << dendl; | |
310 | goto fail; | |
311 | } | |
312 | r = chain_fsetxattr<true, true>( | |
313 | fd, XATTR_SPILL_OUT_NAME, | |
314 | XATTR_NO_SPILL_OUT, sizeof(XATTR_NO_SPILL_OUT)); | |
315 | if (r < 0) { | |
316 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
317 | derr << "error setting spillout xattr for oid " << oid << " (" << (*path)->path() | |
318 | << "):" << cpp_strerror(-r) << dendl; | |
319 | goto fail; | |
320 | } | |
321 | } | |
322 | ||
323 | if (!replaying) { | |
324 | bool existed; | |
325 | *outfd = fdcache.add(oid, fd, &existed); | |
326 | if (existed) { | |
327 | TEMP_FAILURE_RETRY(::close(fd)); | |
328 | } | |
329 | } else { | |
330 | *outfd = std::make_shared<FDCache::FD>(fd); | |
331 | } | |
332 | ||
333 | if (need_lock) { | |
334 | ((*index).index)->access_lock.put_write(); | |
335 | } | |
336 | ||
337 | return 0; | |
338 | ||
339 | fail: | |
340 | ||
341 | if (need_lock) { | |
342 | ((*index).index)->access_lock.put_write(); | |
343 | } | |
344 | ||
345 | assert(!m_filestore_fail_eio || r != -EIO); | |
346 | return r; | |
347 | } | |
348 | ||
349 | void FileStore::lfn_close(FDRef fd) | |
350 | { | |
351 | } | |
352 | ||
353 | int FileStore::lfn_link(const coll_t& c, const coll_t& newcid, const ghobject_t& o, const ghobject_t& newoid) | |
354 | { | |
355 | Index index_new, index_old; | |
356 | IndexedPath path_new, path_old; | |
357 | int exist; | |
358 | int r; | |
359 | bool index_same = false; | |
360 | if (c < newcid) { | |
361 | r = get_index(newcid, &index_new); | |
362 | if (r < 0) | |
363 | return r; | |
364 | r = get_index(c, &index_old); | |
365 | if (r < 0) | |
366 | return r; | |
367 | } else if (c == newcid) { | |
368 | r = get_index(c, &index_old); | |
369 | if (r < 0) | |
370 | return r; | |
371 | index_new = index_old; | |
372 | index_same = true; | |
373 | } else { | |
374 | r = get_index(c, &index_old); | |
375 | if (r < 0) | |
376 | return r; | |
377 | r = get_index(newcid, &index_new); | |
378 | if (r < 0) | |
379 | return r; | |
380 | } | |
381 | ||
382 | assert(NULL != index_old.index); | |
383 | assert(NULL != index_new.index); | |
384 | ||
385 | if (!index_same) { | |
386 | ||
387 | RWLock::RLocker l1((index_old.index)->access_lock); | |
388 | ||
389 | r = index_old->lookup(o, &path_old, &exist); | |
390 | if (r < 0) { | |
391 | assert(!m_filestore_fail_eio || r != -EIO); | |
392 | return r; | |
393 | } | |
394 | if (!exist) | |
395 | return -ENOENT; | |
396 | ||
397 | RWLock::WLocker l2((index_new.index)->access_lock); | |
398 | ||
399 | r = index_new->lookup(newoid, &path_new, &exist); | |
400 | if (r < 0) { | |
401 | assert(!m_filestore_fail_eio || r != -EIO); | |
402 | return r; | |
403 | } | |
404 | if (exist) | |
405 | return -EEXIST; | |
406 | ||
31f18b77 FG |
407 | dout(25) << __FUNC__ << ": path_old: " << path_old << dendl; |
408 | dout(25) << __FUNC__ << ": path_new: " << path_new << dendl; | |
7c673cae FG |
409 | r = ::link(path_old->path(), path_new->path()); |
410 | if (r < 0) | |
411 | return -errno; | |
412 | ||
413 | r = index_new->created(newoid, path_new->path()); | |
414 | if (r < 0) { | |
415 | assert(!m_filestore_fail_eio || r != -EIO); | |
416 | return r; | |
417 | } | |
418 | } else { | |
419 | RWLock::WLocker l1((index_old.index)->access_lock); | |
420 | ||
421 | r = index_old->lookup(o, &path_old, &exist); | |
422 | if (r < 0) { | |
423 | assert(!m_filestore_fail_eio || r != -EIO); | |
424 | return r; | |
425 | } | |
426 | if (!exist) | |
427 | return -ENOENT; | |
428 | ||
429 | r = index_new->lookup(newoid, &path_new, &exist); | |
430 | if (r < 0) { | |
431 | assert(!m_filestore_fail_eio || r != -EIO); | |
432 | return r; | |
433 | } | |
434 | if (exist) | |
435 | return -EEXIST; | |
436 | ||
31f18b77 FG |
437 | dout(25) << __FUNC__ << ": path_old: " << path_old << dendl; |
438 | dout(25) << __FUNC__ << ": path_new: " << path_new << dendl; | |
7c673cae FG |
439 | r = ::link(path_old->path(), path_new->path()); |
440 | if (r < 0) | |
441 | return -errno; | |
442 | ||
443 | // make sure old fd for unlinked/overwritten file is gone | |
444 | fdcache.clear(newoid); | |
445 | ||
446 | r = index_new->created(newoid, path_new->path()); | |
447 | if (r < 0) { | |
448 | assert(!m_filestore_fail_eio || r != -EIO); | |
449 | return r; | |
450 | } | |
451 | } | |
452 | return 0; | |
453 | } | |
454 | ||
455 | int FileStore::lfn_unlink(const coll_t& cid, const ghobject_t& o, | |
456 | const SequencerPosition &spos, | |
457 | bool force_clear_omap) | |
458 | { | |
459 | Index index; | |
460 | int r = get_index(cid, &index); | |
461 | if (r < 0) { | |
31f18b77 | 462 | dout(25) << __FUNC__ << ": get_index failed " << cpp_strerror(r) << dendl; |
7c673cae FG |
463 | return r; |
464 | } | |
465 | ||
466 | assert(NULL != index.index); | |
467 | RWLock::WLocker l((index.index)->access_lock); | |
468 | ||
469 | { | |
470 | IndexedPath path; | |
471 | int hardlink; | |
472 | r = index->lookup(o, &path, &hardlink); | |
473 | if (r < 0) { | |
474 | assert(!m_filestore_fail_eio || r != -EIO); | |
475 | return r; | |
476 | } | |
477 | ||
478 | if (!force_clear_omap) { | |
479 | if (hardlink == 0 || hardlink == 1) { | |
480 | force_clear_omap = true; | |
481 | } | |
482 | } | |
483 | if (force_clear_omap) { | |
31f18b77 | 484 | dout(20) << __FUNC__ << ": clearing omap on " << o |
7c673cae FG |
485 | << " in cid " << cid << dendl; |
486 | r = object_map->clear(o, &spos); | |
487 | if (r < 0 && r != -ENOENT) { | |
31f18b77 | 488 | dout(25) << __FUNC__ << ": omap clear failed " << cpp_strerror(r) << dendl; |
7c673cae FG |
489 | assert(!m_filestore_fail_eio || r != -EIO); |
490 | return r; | |
491 | } | |
492 | if (cct->_conf->filestore_debug_inject_read_err) { | |
493 | debug_obj_on_delete(o); | |
494 | } | |
495 | if (!m_disable_wbthrottle) { | |
496 | wbthrottle.clear_object(o); // should be only non-cache ref | |
497 | } | |
498 | fdcache.clear(o); | |
499 | } else { | |
500 | /* Ensure that replay of this op doesn't result in the object_map | |
501 | * going away. | |
502 | */ | |
503 | if (!backend->can_checkpoint()) | |
504 | object_map->sync(&o, &spos); | |
505 | } | |
506 | if (hardlink == 0) { | |
507 | if (!m_disable_wbthrottle) { | |
508 | wbthrottle.clear_object(o); // should be only non-cache ref | |
509 | } | |
510 | return 0; | |
511 | } | |
512 | } | |
513 | r = index->unlink(o); | |
514 | if (r < 0) { | |
31f18b77 | 515 | dout(25) << __FUNC__ << ": index unlink failed " << cpp_strerror(r) << dendl; |
7c673cae FG |
516 | return r; |
517 | } | |
518 | return 0; | |
519 | } | |
520 | ||
521 | FileStore::FileStore(CephContext* cct, const std::string &base, | |
522 | const std::string &jdev, osflagbits_t flags, | |
523 | const char *name, bool do_update) : | |
524 | JournalingObjectStore(cct, base), | |
525 | internal_name(name), | |
526 | basedir(base), journalpath(jdev), | |
527 | generic_flags(flags), | |
528 | blk_size(0), | |
529 | fsid_fd(-1), op_fd(-1), | |
530 | basedir_fd(-1), current_fd(-1), | |
531 | backend(NULL), | |
532 | index_manager(cct, do_update), | |
533 | lock("FileStore::lock"), | |
534 | force_sync(false), | |
535 | sync_entry_timeo_lock("FileStore::sync_entry_timeo_lock"), | |
536 | timer(cct, sync_entry_timeo_lock), | |
537 | stop(false), sync_thread(this), | |
538 | fdcache(cct), | |
539 | wbthrottle(cct), | |
540 | next_osr_id(0), | |
541 | m_disable_wbthrottle(cct->_conf->filestore_odsync_write || | |
542 | !cct->_conf->filestore_wbthrottle_enable), | |
543 | throttle_ops(cct, "filestore_ops", cct->_conf->filestore_caller_concurrency), | |
544 | throttle_bytes(cct, "filestore_bytes", cct->_conf->filestore_caller_concurrency), | |
545 | m_ondisk_finisher_num(cct->_conf->filestore_ondisk_finisher_threads), | |
546 | m_apply_finisher_num(cct->_conf->filestore_apply_finisher_threads), | |
547 | op_tp(cct, "FileStore::op_tp", "tp_fstore_op", cct->_conf->filestore_op_threads, "filestore_op_threads"), | |
548 | op_wq(this, cct->_conf->filestore_op_thread_timeout, | |
549 | cct->_conf->filestore_op_thread_suicide_timeout, &op_tp), | |
550 | logger(NULL), | |
551 | trace_endpoint("0.0.0.0", 0, "FileStore"), | |
552 | read_error_lock("FileStore::read_error_lock"), | |
553 | m_filestore_commit_timeout(cct->_conf->filestore_commit_timeout), | |
554 | m_filestore_journal_parallel(cct->_conf->filestore_journal_parallel ), | |
555 | m_filestore_journal_trailing(cct->_conf->filestore_journal_trailing), | |
556 | m_filestore_journal_writeahead(cct->_conf->filestore_journal_writeahead), | |
557 | m_filestore_fiemap_threshold(cct->_conf->filestore_fiemap_threshold), | |
558 | m_filestore_max_sync_interval(cct->_conf->filestore_max_sync_interval), | |
559 | m_filestore_min_sync_interval(cct->_conf->filestore_min_sync_interval), | |
560 | m_filestore_fail_eio(cct->_conf->filestore_fail_eio), | |
561 | m_filestore_fadvise(cct->_conf->filestore_fadvise), | |
562 | do_update(do_update), | |
563 | m_journal_dio(cct->_conf->journal_dio), | |
564 | m_journal_aio(cct->_conf->journal_aio), | |
565 | m_journal_force_aio(cct->_conf->journal_force_aio), | |
566 | m_osd_rollback_to_cluster_snap(cct->_conf->osd_rollback_to_cluster_snap), | |
567 | m_osd_use_stale_snap(cct->_conf->osd_use_stale_snap), | |
568 | m_filestore_do_dump(false), | |
569 | m_filestore_dump_fmt(true), | |
570 | m_filestore_sloppy_crc(cct->_conf->filestore_sloppy_crc), | |
571 | m_filestore_sloppy_crc_block_size(cct->_conf->filestore_sloppy_crc_block_size), | |
572 | m_filestore_max_alloc_hint_size(cct->_conf->filestore_max_alloc_hint_size), | |
573 | m_fs_type(0), | |
574 | m_filestore_max_inline_xattr_size(0), | |
575 | m_filestore_max_inline_xattrs(0), | |
576 | m_filestore_max_xattr_value_size(0) | |
577 | { | |
31f18b77 | 578 | m_filestore_kill_at = cct->_conf->filestore_kill_at; |
7c673cae FG |
579 | for (int i = 0; i < m_ondisk_finisher_num; ++i) { |
580 | ostringstream oss; | |
581 | oss << "filestore-ondisk-" << i; | |
582 | Finisher *f = new Finisher(cct, oss.str(), "fn_odsk_fstore"); | |
583 | ondisk_finishers.push_back(f); | |
584 | } | |
585 | for (int i = 0; i < m_apply_finisher_num; ++i) { | |
586 | ostringstream oss; | |
587 | oss << "filestore-apply-" << i; | |
588 | Finisher *f = new Finisher(cct, oss.str(), "fn_appl_fstore"); | |
589 | apply_finishers.push_back(f); | |
590 | } | |
591 | ||
592 | ostringstream oss; | |
593 | oss << basedir << "/current"; | |
594 | current_fn = oss.str(); | |
595 | ||
596 | ostringstream sss; | |
597 | sss << basedir << "/current/commit_op_seq"; | |
598 | current_op_seq_fn = sss.str(); | |
599 | ||
600 | ostringstream omss; | |
601 | if (cct->_conf->filestore_omap_backend_path != "") { | |
602 | omap_dir = cct->_conf->filestore_omap_backend_path; | |
603 | } else { | |
604 | omss << basedir << "/current/omap"; | |
605 | omap_dir = omss.str(); | |
606 | } | |
607 | ||
608 | // initialize logger | |
609 | PerfCountersBuilder plb(cct, internal_name, l_filestore_first, l_filestore_last); | |
610 | ||
611 | plb.add_u64(l_filestore_journal_queue_ops, "journal_queue_ops", "Operations in journal queue"); | |
612 | plb.add_u64(l_filestore_journal_ops, "journal_ops", "Active journal entries to be applied"); | |
613 | plb.add_u64(l_filestore_journal_queue_bytes, "journal_queue_bytes", "Size of journal queue"); | |
614 | plb.add_u64(l_filestore_journal_bytes, "journal_bytes", "Active journal operation size to be applied"); | |
28e407b8 AA |
615 | plb.add_time_avg(l_filestore_journal_latency, "journal_latency", "Average journal queue completing latency", |
616 | NULL, PerfCountersBuilder::PRIO_USEFUL); | |
7c673cae FG |
617 | plb.add_u64_counter(l_filestore_journal_wr, "journal_wr", "Journal write IOs"); |
618 | plb.add_u64_avg(l_filestore_journal_wr_bytes, "journal_wr_bytes", "Journal data written"); | |
619 | plb.add_u64(l_filestore_op_queue_max_ops, "op_queue_max_ops", "Max operations in writing to FS queue"); | |
620 | plb.add_u64(l_filestore_op_queue_ops, "op_queue_ops", "Operations in writing to FS queue"); | |
621 | plb.add_u64_counter(l_filestore_ops, "ops", "Operations written to store"); | |
622 | plb.add_u64(l_filestore_op_queue_max_bytes, "op_queue_max_bytes", "Max data in writing to FS queue"); | |
623 | plb.add_u64(l_filestore_op_queue_bytes, "op_queue_bytes", "Size of writing to FS queue"); | |
624 | plb.add_u64_counter(l_filestore_bytes, "bytes", "Data written to store"); | |
625 | plb.add_time_avg(l_filestore_apply_latency, "apply_latency", "Apply latency"); | |
626 | plb.add_u64(l_filestore_committing, "committing", "Is currently committing"); | |
627 | ||
628 | plb.add_u64_counter(l_filestore_commitcycle, "commitcycle", "Commit cycles"); | |
629 | plb.add_time_avg(l_filestore_commitcycle_interval, "commitcycle_interval", "Average interval between commits"); | |
630 | plb.add_time_avg(l_filestore_commitcycle_latency, "commitcycle_latency", "Average latency of commit"); | |
631 | plb.add_u64_counter(l_filestore_journal_full, "journal_full", "Journal writes while full"); | |
28e407b8 AA |
632 | plb.add_time_avg(l_filestore_queue_transaction_latency_avg, "queue_transaction_latency_avg", |
633 | "Store operation queue latency", NULL, PerfCountersBuilder::PRIO_USEFUL); | |
224ce89b | 634 | plb.add_time(l_filestore_sync_pause_max_lat, "sync_pause_max_latency", "Max latency of op_wq pause before syncfs"); |
7c673cae FG |
635 | |
636 | logger = plb.create_perf_counters(); | |
637 | ||
638 | cct->get_perfcounters_collection()->add(logger); | |
639 | cct->_conf->add_observer(this); | |
640 | ||
641 | superblock.compat_features = get_fs_initial_compat_set(); | |
642 | } | |
643 | ||
644 | FileStore::~FileStore() | |
645 | { | |
646 | for (vector<Finisher*>::iterator it = ondisk_finishers.begin(); it != ondisk_finishers.end(); ++it) { | |
647 | delete *it; | |
648 | *it = NULL; | |
649 | } | |
650 | for (vector<Finisher*>::iterator it = apply_finishers.begin(); it != apply_finishers.end(); ++it) { | |
651 | delete *it; | |
652 | *it = NULL; | |
653 | } | |
654 | cct->_conf->remove_observer(this); | |
655 | cct->get_perfcounters_collection()->remove(logger); | |
656 | ||
657 | if (journal) | |
658 | journal->logger = NULL; | |
659 | delete logger; | |
660 | ||
661 | if (m_filestore_do_dump) { | |
662 | dump_stop(); | |
663 | } | |
664 | } | |
665 | ||
666 | static void get_attrname(const char *name, char *buf, int len) | |
667 | { | |
668 | snprintf(buf, len, "user.ceph.%s", name); | |
669 | } | |
670 | ||
671 | bool parse_attrname(char **name) | |
672 | { | |
673 | if (strncmp(*name, "user.ceph.", 10) == 0) { | |
674 | *name += 10; | |
675 | return true; | |
676 | } | |
677 | return false; | |
678 | } | |
679 | ||
680 | void FileStore::collect_metadata(map<string,string> *pm) | |
681 | { | |
682 | char partition_path[PATH_MAX]; | |
683 | char dev_node[PATH_MAX]; | |
684 | int rc = 0; | |
685 | ||
686 | (*pm)["filestore_backend"] = backend->get_name(); | |
687 | ostringstream ss; | |
688 | ss << "0x" << std::hex << m_fs_type << std::dec; | |
689 | (*pm)["filestore_f_type"] = ss.str(); | |
690 | ||
691 | if (cct->_conf->filestore_collect_device_partition_information) { | |
3efd9988 | 692 | rc = get_device_by_fd(fsid_fd, partition_path, dev_node, PATH_MAX); |
7c673cae FG |
693 | } else { |
694 | rc = -EINVAL; | |
695 | } | |
696 | ||
697 | switch (rc) { | |
698 | case -EOPNOTSUPP: | |
699 | case -EINVAL: | |
700 | (*pm)["backend_filestore_partition_path"] = "unknown"; | |
701 | (*pm)["backend_filestore_dev_node"] = "unknown"; | |
702 | break; | |
703 | case -ENODEV: | |
704 | (*pm)["backend_filestore_partition_path"] = string(partition_path); | |
705 | (*pm)["backend_filestore_dev_node"] = "unknown"; | |
706 | break; | |
707 | default: | |
708 | (*pm)["backend_filestore_partition_path"] = string(partition_path); | |
709 | (*pm)["backend_filestore_dev_node"] = string(dev_node); | |
710 | } | |
711 | } | |
712 | ||
713 | int FileStore::statfs(struct store_statfs_t *buf0) | |
714 | { | |
715 | struct statfs buf; | |
716 | buf0->reset(); | |
717 | if (::statfs(basedir.c_str(), &buf) < 0) { | |
718 | int r = -errno; | |
719 | assert(!m_filestore_fail_eio || r != -EIO); | |
720 | assert(r != -ENOENT); | |
721 | return r; | |
722 | } | |
723 | buf0->total = buf.f_blocks * buf.f_bsize; | |
724 | buf0->available = buf.f_bavail * buf.f_bsize; | |
725 | // Adjust for writes pending in the journal | |
726 | if (journal) { | |
727 | uint64_t estimate = journal->get_journal_size_estimate(); | |
728 | if (buf0->available > estimate) | |
729 | buf0->available -= estimate; | |
730 | else | |
731 | buf0->available = 0; | |
732 | } | |
733 | return 0; | |
734 | } | |
735 | ||
736 | ||
737 | void FileStore::new_journal() | |
738 | { | |
739 | if (journalpath.length()) { | |
740 | dout(10) << "open_journal at " << journalpath << dendl; | |
741 | journal = new FileJournal(cct, fsid, &finisher, &sync_cond, | |
742 | journalpath.c_str(), | |
743 | m_journal_dio, m_journal_aio, | |
744 | m_journal_force_aio); | |
745 | if (journal) | |
746 | journal->logger = logger; | |
747 | } | |
748 | return; | |
749 | } | |
750 | ||
751 | int FileStore::dump_journal(ostream& out) | |
752 | { | |
753 | int r; | |
754 | ||
755 | if (!journalpath.length()) | |
756 | return -EINVAL; | |
757 | ||
758 | FileJournal *journal = new FileJournal(cct, fsid, &finisher, &sync_cond, journalpath.c_str(), m_journal_dio); | |
759 | r = journal->dump(out); | |
760 | delete journal; | |
761 | return r; | |
762 | } | |
763 | ||
764 | FileStoreBackend *FileStoreBackend::create(long f_type, FileStore *fs) | |
765 | { | |
766 | switch (f_type) { | |
767 | #if defined(__linux__) | |
768 | case BTRFS_SUPER_MAGIC: | |
769 | return new BtrfsFileStoreBackend(fs); | |
770 | # ifdef HAVE_LIBXFS | |
771 | case XFS_SUPER_MAGIC: | |
772 | return new XfsFileStoreBackend(fs); | |
773 | # endif | |
774 | #endif | |
775 | #ifdef HAVE_LIBZFS | |
776 | case ZFS_SUPER_MAGIC: | |
777 | return new ZFSFileStoreBackend(fs); | |
778 | #endif | |
779 | default: | |
780 | return new GenericFileStoreBackend(fs); | |
781 | } | |
782 | } | |
783 | ||
784 | void FileStore::create_backend(long f_type) | |
785 | { | |
786 | m_fs_type = f_type; | |
787 | ||
788 | assert(backend == NULL); | |
789 | backend = FileStoreBackend::create(f_type, this); | |
790 | ||
791 | dout(0) << "backend " << backend->get_name() | |
792 | << " (magic 0x" << std::hex << f_type << std::dec << ")" | |
793 | << dendl; | |
794 | ||
795 | switch (f_type) { | |
796 | #if defined(__linux__) | |
797 | case BTRFS_SUPER_MAGIC: | |
798 | if (!m_disable_wbthrottle){ | |
799 | wbthrottle.set_fs(WBThrottle::BTRFS); | |
800 | } | |
801 | break; | |
802 | ||
803 | case XFS_SUPER_MAGIC: | |
804 | // wbthrottle is constructed with fs(WBThrottle::XFS) | |
805 | break; | |
806 | #endif | |
807 | } | |
808 | ||
809 | set_xattr_limits_via_conf(); | |
810 | } | |
811 | ||
812 | int FileStore::mkfs() | |
813 | { | |
814 | int ret = 0; | |
815 | char fsid_fn[PATH_MAX]; | |
816 | char fsid_str[40]; | |
817 | uuid_d old_fsid; | |
818 | uuid_d old_omap_fsid; | |
819 | ||
820 | dout(1) << "mkfs in " << basedir << dendl; | |
821 | basedir_fd = ::open(basedir.c_str(), O_RDONLY); | |
822 | if (basedir_fd < 0) { | |
823 | ret = -errno; | |
224ce89b | 824 | derr << __FUNC__ << ": failed to open base dir " << basedir << ": " << cpp_strerror(ret) << dendl; |
7c673cae FG |
825 | return ret; |
826 | } | |
827 | ||
828 | // open+lock fsid | |
829 | snprintf(fsid_fn, sizeof(fsid_fn), "%s/fsid", basedir.c_str()); | |
830 | fsid_fd = ::open(fsid_fn, O_RDWR|O_CREAT, 0644); | |
831 | if (fsid_fd < 0) { | |
832 | ret = -errno; | |
224ce89b | 833 | derr << __FUNC__ << ": failed to open " << fsid_fn << ": " << cpp_strerror(ret) << dendl; |
7c673cae FG |
834 | goto close_basedir_fd; |
835 | } | |
836 | ||
837 | if (lock_fsid() < 0) { | |
838 | ret = -EBUSY; | |
839 | goto close_fsid_fd; | |
840 | } | |
841 | ||
842 | if (read_fsid(fsid_fd, &old_fsid) < 0 || old_fsid.is_zero()) { | |
843 | if (fsid.is_zero()) { | |
844 | fsid.generate_random(); | |
224ce89b | 845 | dout(1) << __FUNC__ << ": generated fsid " << fsid << dendl; |
7c673cae | 846 | } else { |
224ce89b | 847 | dout(1) << __FUNC__ << ": using provided fsid " << fsid << dendl; |
7c673cae FG |
848 | } |
849 | ||
850 | fsid.print(fsid_str); | |
851 | strcat(fsid_str, "\n"); | |
852 | ret = ::ftruncate(fsid_fd, 0); | |
853 | if (ret < 0) { | |
854 | ret = -errno; | |
31f18b77 | 855 | derr << __FUNC__ << ": failed to truncate fsid: " |
7c673cae FG |
856 | << cpp_strerror(ret) << dendl; |
857 | goto close_fsid_fd; | |
858 | } | |
859 | ret = safe_write(fsid_fd, fsid_str, strlen(fsid_str)); | |
860 | if (ret < 0) { | |
31f18b77 | 861 | derr << __FUNC__ << ": failed to write fsid: " |
7c673cae FG |
862 | << cpp_strerror(ret) << dendl; |
863 | goto close_fsid_fd; | |
864 | } | |
865 | if (::fsync(fsid_fd) < 0) { | |
866 | ret = -errno; | |
31f18b77 | 867 | derr << __FUNC__ << ": close failed: can't write fsid: " |
7c673cae FG |
868 | << cpp_strerror(ret) << dendl; |
869 | goto close_fsid_fd; | |
870 | } | |
224ce89b | 871 | dout(10) << __FUNC__ << ": fsid is " << fsid << dendl; |
7c673cae FG |
872 | } else { |
873 | if (!fsid.is_zero() && fsid != old_fsid) { | |
31f18b77 | 874 | derr << __FUNC__ << ": on-disk fsid " << old_fsid << " != provided " << fsid << dendl; |
7c673cae FG |
875 | ret = -EINVAL; |
876 | goto close_fsid_fd; | |
877 | } | |
878 | fsid = old_fsid; | |
31f18b77 | 879 | dout(1) << __FUNC__ << ": fsid is already set to " << fsid << dendl; |
7c673cae FG |
880 | } |
881 | ||
882 | // version stamp | |
883 | ret = write_version_stamp(); | |
884 | if (ret < 0) { | |
31f18b77 | 885 | derr << __FUNC__ << ": write_version_stamp() failed: " |
7c673cae FG |
886 | << cpp_strerror(ret) << dendl; |
887 | goto close_fsid_fd; | |
888 | } | |
889 | ||
890 | // superblock | |
891 | superblock.omap_backend = cct->_conf->filestore_omap_backend; | |
892 | ret = write_superblock(); | |
893 | if (ret < 0) { | |
31f18b77 | 894 | derr << __FUNC__ << ": write_superblock() failed: " |
7c673cae FG |
895 | << cpp_strerror(ret) << dendl; |
896 | goto close_fsid_fd; | |
897 | } | |
898 | ||
899 | struct statfs basefs; | |
900 | ret = ::fstatfs(basedir_fd, &basefs); | |
901 | if (ret < 0) { | |
902 | ret = -errno; | |
31f18b77 | 903 | derr << __FUNC__ << ": cannot fstatfs basedir " |
7c673cae FG |
904 | << cpp_strerror(ret) << dendl; |
905 | goto close_fsid_fd; | |
906 | } | |
907 | ||
224ce89b WB |
908 | #if defined(__linux__) |
909 | if (basefs.f_type == BTRFS_SUPER_MAGIC && | |
910 | !g_ceph_context->check_experimental_feature_enabled("btrfs")) { | |
911 | derr << __FUNC__ << ": deprecated btrfs support is not enabled" << dendl; | |
912 | goto close_fsid_fd; | |
913 | } | |
914 | #endif | |
915 | ||
7c673cae FG |
916 | create_backend(basefs.f_type); |
917 | ||
918 | ret = backend->create_current(); | |
919 | if (ret < 0) { | |
31f18b77 | 920 | derr << __FUNC__ << ": failed to create current/ " << cpp_strerror(ret) << dendl; |
7c673cae FG |
921 | goto close_fsid_fd; |
922 | } | |
923 | ||
924 | // write initial op_seq | |
925 | { | |
926 | uint64_t initial_seq = 0; | |
927 | int fd = read_op_seq(&initial_seq); | |
928 | if (fd < 0) { | |
929 | ret = fd; | |
31f18b77 | 930 | derr << __FUNC__ << ": failed to create " << current_op_seq_fn << ": " |
7c673cae FG |
931 | << cpp_strerror(ret) << dendl; |
932 | goto close_fsid_fd; | |
933 | } | |
934 | if (initial_seq == 0) { | |
935 | ret = write_op_seq(fd, 1); | |
936 | if (ret < 0) { | |
937 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
31f18b77 | 938 | derr << __FUNC__ << ": failed to write to " << current_op_seq_fn << ": " |
7c673cae FG |
939 | << cpp_strerror(ret) << dendl; |
940 | goto close_fsid_fd; | |
941 | } | |
942 | ||
943 | if (backend->can_checkpoint()) { | |
944 | // create snap_1 too | |
945 | current_fd = ::open(current_fn.c_str(), O_RDONLY); | |
946 | assert(current_fd >= 0); | |
947 | char s[NAME_MAX]; | |
948 | snprintf(s, sizeof(s), COMMIT_SNAP_ITEM, 1ull); | |
949 | ret = backend->create_checkpoint(s, NULL); | |
950 | VOID_TEMP_FAILURE_RETRY(::close(current_fd)); | |
951 | if (ret < 0 && ret != -EEXIST) { | |
952 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
31f18b77 | 953 | derr << __FUNC__ << ": failed to create snap_1: " << cpp_strerror(ret) << dendl; |
7c673cae FG |
954 | goto close_fsid_fd; |
955 | } | |
956 | } | |
957 | } | |
958 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
959 | } | |
960 | ret = KeyValueDB::test_init(superblock.omap_backend, omap_dir); | |
961 | if (ret < 0) { | |
31f18b77 | 962 | derr << __FUNC__ << ": failed to create " << cct->_conf->filestore_omap_backend << dendl; |
7c673cae FG |
963 | goto close_fsid_fd; |
964 | } | |
965 | // create fsid under omap | |
966 | // open+lock fsid | |
967 | int omap_fsid_fd; | |
968 | char omap_fsid_fn[PATH_MAX]; | |
969 | snprintf(omap_fsid_fn, sizeof(omap_fsid_fn), "%s/osd_uuid", omap_dir.c_str()); | |
970 | omap_fsid_fd = ::open(omap_fsid_fn, O_RDWR|O_CREAT, 0644); | |
971 | if (omap_fsid_fd < 0) { | |
972 | ret = -errno; | |
31f18b77 | 973 | derr << __FUNC__ << ": failed to open " << omap_fsid_fn << ": " << cpp_strerror(ret) << dendl; |
7c673cae FG |
974 | goto close_fsid_fd; |
975 | } | |
976 | ||
977 | if (read_fsid(omap_fsid_fd, &old_omap_fsid) < 0 || old_omap_fsid.is_zero()) { | |
978 | assert(!fsid.is_zero()); | |
979 | fsid.print(fsid_str); | |
980 | strcat(fsid_str, "\n"); | |
981 | ret = ::ftruncate(omap_fsid_fd, 0); | |
982 | if (ret < 0) { | |
983 | ret = -errno; | |
31f18b77 | 984 | derr << __FUNC__ << ": failed to truncate fsid: " |
7c673cae FG |
985 | << cpp_strerror(ret) << dendl; |
986 | goto close_omap_fsid_fd; | |
987 | } | |
988 | ret = safe_write(omap_fsid_fd, fsid_str, strlen(fsid_str)); | |
989 | if (ret < 0) { | |
31f18b77 | 990 | derr << __FUNC__ << ": failed to write fsid: " |
7c673cae FG |
991 | << cpp_strerror(ret) << dendl; |
992 | goto close_omap_fsid_fd; | |
993 | } | |
31f18b77 | 994 | dout(10) << __FUNC__ << ": write success, fsid:" << fsid_str << ", ret:" << ret << dendl; |
7c673cae FG |
995 | if (::fsync(omap_fsid_fd) < 0) { |
996 | ret = -errno; | |
31f18b77 | 997 | derr << __FUNC__ << ": close failed: can't write fsid: " |
7c673cae FG |
998 | << cpp_strerror(ret) << dendl; |
999 | goto close_omap_fsid_fd; | |
1000 | } | |
1001 | dout(10) << "mkfs omap fsid is " << fsid << dendl; | |
1002 | } else { | |
1003 | if (fsid != old_omap_fsid) { | |
31f18b77 | 1004 | derr << __FUNC__ << ": " << omap_fsid_fn |
7c673cae FG |
1005 | << " has existed omap fsid " << old_omap_fsid |
1006 | << " != expected osd fsid " << fsid | |
1007 | << dendl; | |
1008 | ret = -EINVAL; | |
1009 | goto close_omap_fsid_fd; | |
1010 | } | |
31f18b77 | 1011 | dout(1) << __FUNC__ << ": omap fsid is already set to " << fsid << dendl; |
7c673cae FG |
1012 | } |
1013 | ||
1014 | dout(1) << cct->_conf->filestore_omap_backend << " db exists/created" << dendl; | |
1015 | ||
1016 | // journal? | |
1017 | ret = mkjournal(); | |
1018 | if (ret) | |
1019 | goto close_omap_fsid_fd; | |
1020 | ||
1021 | ret = write_meta("type", "filestore"); | |
1022 | if (ret) | |
1023 | goto close_omap_fsid_fd; | |
1024 | ||
1025 | dout(1) << "mkfs done in " << basedir << dendl; | |
1026 | ret = 0; | |
1027 | ||
1028 | close_omap_fsid_fd: | |
1029 | VOID_TEMP_FAILURE_RETRY(::close(omap_fsid_fd)); | |
1030 | close_fsid_fd: | |
1031 | VOID_TEMP_FAILURE_RETRY(::close(fsid_fd)); | |
1032 | fsid_fd = -1; | |
1033 | close_basedir_fd: | |
1034 | VOID_TEMP_FAILURE_RETRY(::close(basedir_fd)); | |
1035 | delete backend; | |
1036 | backend = NULL; | |
1037 | return ret; | |
1038 | } | |
1039 | ||
1040 | int FileStore::mkjournal() | |
1041 | { | |
1042 | // read fsid | |
1043 | int ret; | |
1044 | char fn[PATH_MAX]; | |
1045 | snprintf(fn, sizeof(fn), "%s/fsid", basedir.c_str()); | |
1046 | int fd = ::open(fn, O_RDONLY, 0644); | |
1047 | if (fd < 0) { | |
1048 | int err = errno; | |
31f18b77 | 1049 | derr << __FUNC__ << ": open error: " << cpp_strerror(err) << dendl; |
7c673cae FG |
1050 | return -err; |
1051 | } | |
1052 | ret = read_fsid(fd, &fsid); | |
1053 | if (ret < 0) { | |
31f18b77 | 1054 | derr << __FUNC__ << ": read error: " << cpp_strerror(ret) << dendl; |
7c673cae FG |
1055 | VOID_TEMP_FAILURE_RETRY(::close(fd)); |
1056 | return ret; | |
1057 | } | |
1058 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
1059 | ||
1060 | ret = 0; | |
1061 | ||
1062 | new_journal(); | |
1063 | if (journal) { | |
1064 | ret = journal->check(); | |
1065 | if (ret < 0) { | |
1066 | ret = journal->create(); | |
1067 | if (ret) | |
31f18b77 | 1068 | derr << __FUNC__ << ": error creating journal on " << journalpath |
7c673cae FG |
1069 | << ": " << cpp_strerror(ret) << dendl; |
1070 | else | |
31f18b77 | 1071 | dout(0) << __FUNC__ << ": created journal on " << journalpath << dendl; |
7c673cae FG |
1072 | } |
1073 | delete journal; | |
1074 | journal = 0; | |
1075 | } | |
1076 | return ret; | |
1077 | } | |
1078 | ||
1079 | int FileStore::read_fsid(int fd, uuid_d *uuid) | |
1080 | { | |
1081 | char fsid_str[40]; | |
1082 | memset(fsid_str, 0, sizeof(fsid_str)); | |
1083 | int ret = safe_read(fd, fsid_str, sizeof(fsid_str)); | |
1084 | if (ret < 0) | |
1085 | return ret; | |
1086 | if (ret == 8) { | |
1087 | // old 64-bit fsid... mirror it. | |
1088 | *(uint64_t*)&uuid->bytes()[0] = *(uint64_t*)fsid_str; | |
1089 | *(uint64_t*)&uuid->bytes()[8] = *(uint64_t*)fsid_str; | |
1090 | return 0; | |
1091 | } | |
1092 | ||
1093 | if (ret > 36) | |
1094 | fsid_str[36] = 0; | |
1095 | else | |
1096 | fsid_str[ret] = 0; | |
1097 | if (!uuid->parse(fsid_str)) | |
1098 | return -EINVAL; | |
1099 | return 0; | |
1100 | } | |
1101 | ||
1102 | int FileStore::lock_fsid() | |
1103 | { | |
1104 | struct flock l; | |
1105 | memset(&l, 0, sizeof(l)); | |
1106 | l.l_type = F_WRLCK; | |
1107 | l.l_whence = SEEK_SET; | |
1108 | l.l_start = 0; | |
1109 | l.l_len = 0; | |
1110 | int r = ::fcntl(fsid_fd, F_SETLK, &l); | |
1111 | if (r < 0) { | |
1112 | int err = errno; | |
31f18b77 | 1113 | dout(0) << __FUNC__ << ": failed to lock " << basedir << "/fsid, is another ceph-osd still running? " |
7c673cae FG |
1114 | << cpp_strerror(err) << dendl; |
1115 | return -err; | |
1116 | } | |
1117 | return 0; | |
1118 | } | |
1119 | ||
1120 | bool FileStore::test_mount_in_use() | |
1121 | { | |
31f18b77 | 1122 | dout(5) << __FUNC__ << ": basedir " << basedir << " journal " << journalpath << dendl; |
7c673cae FG |
1123 | char fn[PATH_MAX]; |
1124 | snprintf(fn, sizeof(fn), "%s/fsid", basedir.c_str()); | |
1125 | ||
1126 | // verify fs isn't in use | |
1127 | ||
1128 | fsid_fd = ::open(fn, O_RDWR, 0644); | |
1129 | if (fsid_fd < 0) | |
1130 | return 0; // no fsid, ok. | |
1131 | bool inuse = lock_fsid() < 0; | |
1132 | VOID_TEMP_FAILURE_RETRY(::close(fsid_fd)); | |
1133 | fsid_fd = -1; | |
1134 | return inuse; | |
1135 | } | |
1136 | ||
31f18b77 FG |
1137 | bool FileStore::is_rotational() |
1138 | { | |
1139 | bool rotational; | |
1140 | if (backend) { | |
1141 | rotational = backend->is_rotational(); | |
1142 | } else { | |
1143 | int fd = ::open(basedir.c_str(), O_RDONLY); | |
1144 | if (fd < 0) | |
1145 | return true; | |
1146 | struct statfs st; | |
1147 | int r = ::fstatfs(fd, &st); | |
1148 | ::close(fd); | |
1149 | if (r < 0) { | |
1150 | return true; | |
1151 | } | |
1152 | create_backend(st.f_type); | |
1153 | rotational = backend->is_rotational(); | |
1154 | delete backend; | |
1155 | backend = NULL; | |
1156 | } | |
1157 | dout(10) << __func__ << " " << (int)rotational << dendl; | |
1158 | return rotational; | |
1159 | } | |
1160 | ||
d2e6a577 FG |
1161 | bool FileStore::is_journal_rotational() |
1162 | { | |
1163 | bool journal_rotational; | |
1164 | if (backend) { | |
1165 | journal_rotational = backend->is_journal_rotational(); | |
1166 | } else { | |
1167 | int fd = ::open(journalpath.c_str(), O_RDONLY); | |
1168 | if (fd < 0) | |
1169 | return true; | |
1170 | struct statfs st; | |
1171 | int r = ::fstatfs(fd, &st); | |
1172 | ::close(fd); | |
1173 | if (r < 0) { | |
1174 | return true; | |
1175 | } | |
1176 | create_backend(st.f_type); | |
1177 | journal_rotational = backend->is_journal_rotational(); | |
1178 | delete backend; | |
1179 | backend = NULL; | |
1180 | } | |
1181 | dout(10) << __func__ << " " << (int)journal_rotational << dendl; | |
1182 | return journal_rotational; | |
1183 | } | |
1184 | ||
7c673cae FG |
1185 | int FileStore::_detect_fs() |
1186 | { | |
1187 | struct statfs st; | |
1188 | int r = ::fstatfs(basedir_fd, &st); | |
1189 | if (r < 0) | |
1190 | return -errno; | |
1191 | ||
1192 | blk_size = st.f_bsize; | |
1193 | ||
224ce89b WB |
1194 | #if defined(__linux__) |
1195 | if (st.f_type == BTRFS_SUPER_MAGIC && | |
1196 | !g_ceph_context->check_experimental_feature_enabled("btrfs")) { | |
1197 | derr <<__FUNC__ << ": deprecated btrfs support is not enabled" << dendl; | |
1198 | return -EPERM; | |
1199 | } | |
1200 | #endif | |
1201 | ||
7c673cae FG |
1202 | create_backend(st.f_type); |
1203 | ||
1204 | r = backend->detect_features(); | |
1205 | if (r < 0) { | |
31f18b77 | 1206 | derr << __FUNC__ << ": detect_features error: " << cpp_strerror(r) << dendl; |
7c673cae FG |
1207 | return r; |
1208 | } | |
1209 | ||
1210 | // test xattrs | |
1211 | char fn[PATH_MAX]; | |
1212 | int x = rand(); | |
1213 | int y = x+1; | |
1214 | snprintf(fn, sizeof(fn), "%s/xattr_test", basedir.c_str()); | |
1215 | int tmpfd = ::open(fn, O_CREAT|O_WRONLY|O_TRUNC, 0700); | |
1216 | if (tmpfd < 0) { | |
1217 | int ret = -errno; | |
31f18b77 | 1218 | derr << __FUNC__ << ": unable to create " << fn << ": " << cpp_strerror(ret) << dendl; |
7c673cae FG |
1219 | return ret; |
1220 | } | |
1221 | ||
1222 | int ret = chain_fsetxattr(tmpfd, "user.test", &x, sizeof(x)); | |
1223 | if (ret >= 0) | |
1224 | ret = chain_fgetxattr(tmpfd, "user.test", &y, sizeof(y)); | |
1225 | if ((ret < 0) || (x != y)) { | |
1226 | derr << "Extended attributes don't appear to work. "; | |
1227 | if (ret) | |
1228 | *_dout << "Got error " + cpp_strerror(ret) + ". "; | |
1229 | *_dout << "If you are using ext3 or ext4, be sure to mount the underlying " | |
1230 | << "file system with the 'user_xattr' option." << dendl; | |
1231 | ::unlink(fn); | |
1232 | VOID_TEMP_FAILURE_RETRY(::close(tmpfd)); | |
1233 | return -ENOTSUP; | |
1234 | } | |
1235 | ||
1236 | char buf[1000]; | |
1237 | memset(buf, 0, sizeof(buf)); // shut up valgrind | |
1238 | chain_fsetxattr(tmpfd, "user.test", &buf, sizeof(buf)); | |
1239 | chain_fsetxattr(tmpfd, "user.test2", &buf, sizeof(buf)); | |
1240 | chain_fsetxattr(tmpfd, "user.test3", &buf, sizeof(buf)); | |
1241 | chain_fsetxattr(tmpfd, "user.test4", &buf, sizeof(buf)); | |
1242 | ret = chain_fsetxattr(tmpfd, "user.test5", &buf, sizeof(buf)); | |
1243 | if (ret == -ENOSPC) { | |
1244 | dout(0) << "limited size xattrs" << dendl; | |
1245 | } | |
1246 | chain_fremovexattr(tmpfd, "user.test"); | |
1247 | chain_fremovexattr(tmpfd, "user.test2"); | |
1248 | chain_fremovexattr(tmpfd, "user.test3"); | |
1249 | chain_fremovexattr(tmpfd, "user.test4"); | |
1250 | chain_fremovexattr(tmpfd, "user.test5"); | |
1251 | ||
1252 | ::unlink(fn); | |
1253 | VOID_TEMP_FAILURE_RETRY(::close(tmpfd)); | |
1254 | ||
1255 | return 0; | |
1256 | } | |
1257 | ||
1258 | int FileStore::_sanity_check_fs() | |
1259 | { | |
1260 | // sanity check(s) | |
1261 | ||
1262 | if (((int)m_filestore_journal_writeahead + | |
1263 | (int)m_filestore_journal_parallel + | |
1264 | (int)m_filestore_journal_trailing) > 1) { | |
1265 | dout(0) << "mount ERROR: more than one of filestore journal {writeahead,parallel,trailing} enabled" << dendl; | |
1266 | cerr << TEXT_RED | |
1267 | << " ** WARNING: more than one of 'filestore journal {writeahead,parallel,trailing}'\n" | |
1268 | << " is enabled in ceph.conf. You must choose a single journal mode." | |
1269 | << TEXT_NORMAL << std::endl; | |
1270 | return -EINVAL; | |
1271 | } | |
1272 | ||
1273 | if (!backend->can_checkpoint()) { | |
1274 | if (!journal || !m_filestore_journal_writeahead) { | |
1275 | dout(0) << "mount WARNING: no btrfs, and no journal in writeahead mode; data may be lost" << dendl; | |
1276 | cerr << TEXT_RED | |
1277 | << " ** WARNING: no btrfs AND (no journal OR journal not in writeahead mode)\n" | |
1278 | << " For non-btrfs volumes, a writeahead journal is required to\n" | |
1279 | << " maintain on-disk consistency in the event of a crash. Your conf\n" | |
1280 | << " should include something like:\n" | |
1281 | << " osd journal = /path/to/journal_device_or_file\n" | |
1282 | << " filestore journal writeahead = true\n" | |
1283 | << TEXT_NORMAL; | |
1284 | } | |
1285 | } | |
1286 | ||
1287 | if (!journal) { | |
1288 | dout(0) << "mount WARNING: no journal" << dendl; | |
1289 | cerr << TEXT_YELLOW | |
1290 | << " ** WARNING: No osd journal is configured: write latency may be high.\n" | |
1291 | << " If you will not be using an osd journal, write latency may be\n" | |
1292 | << " relatively high. It can be reduced somewhat by lowering\n" | |
1293 | << " filestore_max_sync_interval, but lower values mean lower write\n" | |
1294 | << " throughput, especially with spinning disks.\n" | |
1295 | << TEXT_NORMAL; | |
1296 | } | |
1297 | ||
1298 | return 0; | |
1299 | } | |
1300 | ||
1301 | int FileStore::write_superblock() | |
1302 | { | |
1303 | bufferlist bl; | |
1304 | ::encode(superblock, bl); | |
1305 | return safe_write_file(basedir.c_str(), "superblock", | |
1306 | bl.c_str(), bl.length()); | |
1307 | } | |
1308 | ||
1309 | int FileStore::read_superblock() | |
1310 | { | |
1311 | bufferptr bp(PATH_MAX); | |
1312 | int ret = safe_read_file(basedir.c_str(), "superblock", | |
1313 | bp.c_str(), bp.length()); | |
1314 | if (ret < 0) { | |
1315 | if (ret == -ENOENT) { | |
1316 | // If the file doesn't exist write initial CompatSet | |
1317 | return write_superblock(); | |
1318 | } | |
1319 | return ret; | |
1320 | } | |
1321 | ||
1322 | bufferlist bl; | |
1323 | bl.push_back(std::move(bp)); | |
1324 | bufferlist::iterator i = bl.begin(); | |
1325 | ::decode(superblock, i); | |
1326 | return 0; | |
1327 | } | |
1328 | ||
1329 | int FileStore::update_version_stamp() | |
1330 | { | |
1331 | return write_version_stamp(); | |
1332 | } | |
1333 | ||
1334 | int FileStore::version_stamp_is_valid(uint32_t *version) | |
1335 | { | |
1336 | bufferptr bp(PATH_MAX); | |
1337 | int ret = safe_read_file(basedir.c_str(), "store_version", | |
1338 | bp.c_str(), bp.length()); | |
1339 | if (ret < 0) { | |
1340 | return ret; | |
1341 | } | |
1342 | bufferlist bl; | |
1343 | bl.push_back(std::move(bp)); | |
1344 | bufferlist::iterator i = bl.begin(); | |
1345 | ::decode(*version, i); | |
31f18b77 | 1346 | dout(10) << __FUNC__ << ": was " << *version << " vs target " |
7c673cae FG |
1347 | << target_version << dendl; |
1348 | if (*version == target_version) | |
1349 | return 1; | |
1350 | else | |
1351 | return 0; | |
1352 | } | |
1353 | ||
1354 | int FileStore::write_version_stamp() | |
1355 | { | |
31f18b77 | 1356 | dout(1) << __FUNC__ << ": " << target_version << dendl; |
7c673cae FG |
1357 | bufferlist bl; |
1358 | ::encode(target_version, bl); | |
1359 | ||
1360 | return safe_write_file(basedir.c_str(), "store_version", | |
1361 | bl.c_str(), bl.length()); | |
1362 | } | |
1363 | ||
1364 | int FileStore::upgrade() | |
1365 | { | |
31f18b77 | 1366 | dout(1) << __FUNC__ << dendl; |
7c673cae FG |
1367 | uint32_t version; |
1368 | int r = version_stamp_is_valid(&version); | |
1369 | ||
1370 | if (r == -ENOENT) { | |
1371 | derr << "The store_version file doesn't exist." << dendl; | |
1372 | return -EINVAL; | |
1373 | } | |
1374 | if (r < 0) | |
1375 | return r; | |
1376 | if (r == 1) | |
1377 | return 0; | |
1378 | ||
1379 | if (version < 3) { | |
1380 | derr << "ObjectStore is old at version " << version << ". Please upgrade to firefly v0.80.x, convert your store, and then upgrade." << dendl; | |
1381 | return -EINVAL; | |
1382 | } | |
1383 | ||
1384 | // nothing necessary in FileStore for v3 -> v4 upgrade; we just need to | |
1385 | // open up DBObjectMap with the do_upgrade flag, which we already did. | |
1386 | update_version_stamp(); | |
1387 | return 0; | |
1388 | } | |
1389 | ||
1390 | int FileStore::read_op_seq(uint64_t *seq) | |
1391 | { | |
1392 | int op_fd = ::open(current_op_seq_fn.c_str(), O_CREAT|O_RDWR, 0644); | |
1393 | if (op_fd < 0) { | |
1394 | int r = -errno; | |
1395 | assert(!m_filestore_fail_eio || r != -EIO); | |
1396 | return r; | |
1397 | } | |
1398 | char s[40]; | |
1399 | memset(s, 0, sizeof(s)); | |
1400 | int ret = safe_read(op_fd, s, sizeof(s) - 1); | |
1401 | if (ret < 0) { | |
31f18b77 | 1402 | derr << __FUNC__ << ": error reading " << current_op_seq_fn << ": " << cpp_strerror(ret) << dendl; |
7c673cae FG |
1403 | VOID_TEMP_FAILURE_RETRY(::close(op_fd)); |
1404 | assert(!m_filestore_fail_eio || ret != -EIO); | |
1405 | return ret; | |
1406 | } | |
1407 | *seq = atoll(s); | |
1408 | return op_fd; | |
1409 | } | |
1410 | ||
1411 | int FileStore::write_op_seq(int fd, uint64_t seq) | |
1412 | { | |
1413 | char s[30]; | |
1414 | snprintf(s, sizeof(s), "%" PRId64 "\n", seq); | |
1415 | int ret = TEMP_FAILURE_RETRY(::pwrite(fd, s, strlen(s), 0)); | |
1416 | if (ret < 0) { | |
1417 | ret = -errno; | |
1418 | assert(!m_filestore_fail_eio || ret != -EIO); | |
1419 | } | |
1420 | return ret; | |
1421 | } | |
1422 | ||
1423 | int FileStore::mount() | |
1424 | { | |
1425 | int ret; | |
1426 | char buf[PATH_MAX]; | |
1427 | uint64_t initial_op_seq; | |
1428 | uuid_d omap_fsid; | |
1429 | set<string> cluster_snaps; | |
1430 | CompatSet supported_compat_set = get_fs_supported_compat_set(); | |
1431 | ||
1432 | dout(5) << "basedir " << basedir << " journal " << journalpath << dendl; | |
1433 | ||
1434 | ret = set_throttle_params(); | |
1435 | if (ret != 0) | |
1436 | goto done; | |
1437 | ||
1438 | // make sure global base dir exists | |
1439 | if (::access(basedir.c_str(), R_OK | W_OK)) { | |
1440 | ret = -errno; | |
31f18b77 | 1441 | derr << __FUNC__ << ": unable to access basedir '" << basedir << "': " |
7c673cae FG |
1442 | << cpp_strerror(ret) << dendl; |
1443 | goto done; | |
1444 | } | |
1445 | ||
1446 | // get fsid | |
1447 | snprintf(buf, sizeof(buf), "%s/fsid", basedir.c_str()); | |
1448 | fsid_fd = ::open(buf, O_RDWR, 0644); | |
1449 | if (fsid_fd < 0) { | |
1450 | ret = -errno; | |
31f18b77 | 1451 | derr << __FUNC__ << ": error opening '" << buf << "': " |
7c673cae FG |
1452 | << cpp_strerror(ret) << dendl; |
1453 | goto done; | |
1454 | } | |
1455 | ||
1456 | ret = read_fsid(fsid_fd, &fsid); | |
1457 | if (ret < 0) { | |
31f18b77 | 1458 | derr << __FUNC__ << ": error reading fsid_fd: " << cpp_strerror(ret) |
7c673cae FG |
1459 | << dendl; |
1460 | goto close_fsid_fd; | |
1461 | } | |
1462 | ||
1463 | if (lock_fsid() < 0) { | |
31f18b77 | 1464 | derr << __FUNC__ << ": lock_fsid failed" << dendl; |
7c673cae FG |
1465 | ret = -EBUSY; |
1466 | goto close_fsid_fd; | |
1467 | } | |
1468 | ||
1469 | dout(10) << "mount fsid is " << fsid << dendl; | |
1470 | ||
1471 | ||
1472 | uint32_t version_stamp; | |
1473 | ret = version_stamp_is_valid(&version_stamp); | |
1474 | if (ret < 0) { | |
31f18b77 | 1475 | derr << __FUNC__ << ": error in version_stamp_is_valid: " |
7c673cae FG |
1476 | << cpp_strerror(ret) << dendl; |
1477 | goto close_fsid_fd; | |
1478 | } else if (ret == 0) { | |
1479 | if (do_update || (int)version_stamp < cct->_conf->filestore_update_to) { | |
31f18b77 | 1480 | derr << __FUNC__ << ": stale version stamp detected: " |
7c673cae FG |
1481 | << version_stamp |
1482 | << ". Proceeding, do_update " | |
1483 | << "is set, performing disk format upgrade." | |
1484 | << dendl; | |
1485 | do_update = true; | |
1486 | } else { | |
1487 | ret = -EINVAL; | |
31f18b77 | 1488 | derr << __FUNC__ << ": stale version stamp " << version_stamp |
7c673cae FG |
1489 | << ". Please run the FileStore update script before starting the " |
1490 | << "OSD, or set filestore_update_to to " << target_version | |
1491 | << " (currently " << cct->_conf->filestore_update_to << ")" | |
1492 | << dendl; | |
1493 | goto close_fsid_fd; | |
1494 | } | |
1495 | } | |
1496 | ||
1497 | ret = read_superblock(); | |
1498 | if (ret < 0) { | |
1499 | goto close_fsid_fd; | |
1500 | } | |
1501 | ||
1502 | // Check if this FileStore supports all the necessary features to mount | |
1503 | if (supported_compat_set.compare(superblock.compat_features) == -1) { | |
31f18b77 | 1504 | derr << __FUNC__ << ": Incompatible features set " |
7c673cae FG |
1505 | << superblock.compat_features << dendl; |
1506 | ret = -EINVAL; | |
1507 | goto close_fsid_fd; | |
1508 | } | |
1509 | ||
1510 | // open some dir handles | |
1511 | basedir_fd = ::open(basedir.c_str(), O_RDONLY); | |
1512 | if (basedir_fd < 0) { | |
1513 | ret = -errno; | |
31f18b77 | 1514 | derr << __FUNC__ << ": failed to open " << basedir << ": " |
7c673cae FG |
1515 | << cpp_strerror(ret) << dendl; |
1516 | basedir_fd = -1; | |
1517 | goto close_fsid_fd; | |
1518 | } | |
1519 | ||
1520 | // test for btrfs, xattrs, etc. | |
1521 | ret = _detect_fs(); | |
1522 | if (ret < 0) { | |
31f18b77 | 1523 | derr << __FUNC__ << ": error in _detect_fs: " |
7c673cae FG |
1524 | << cpp_strerror(ret) << dendl; |
1525 | goto close_basedir_fd; | |
1526 | } | |
1527 | ||
1528 | { | |
1529 | list<string> ls; | |
1530 | ret = backend->list_checkpoints(ls); | |
1531 | if (ret < 0) { | |
31f18b77 | 1532 | derr << __FUNC__ << ": error in _list_snaps: "<< cpp_strerror(ret) << dendl; |
7c673cae FG |
1533 | goto close_basedir_fd; |
1534 | } | |
1535 | ||
1536 | long long unsigned c, prev = 0; | |
1537 | char clustersnap[NAME_MAX]; | |
1538 | for (list<string>::iterator it = ls.begin(); it != ls.end(); ++it) { | |
1539 | if (sscanf(it->c_str(), COMMIT_SNAP_ITEM, &c) == 1) { | |
1540 | assert(c > prev); | |
1541 | prev = c; | |
1542 | snaps.push_back(c); | |
1543 | } else if (sscanf(it->c_str(), CLUSTER_SNAP_ITEM, clustersnap) == 1) | |
1544 | cluster_snaps.insert(*it); | |
1545 | } | |
1546 | } | |
1547 | ||
1548 | if (m_osd_rollback_to_cluster_snap.length() && | |
1549 | cluster_snaps.count(m_osd_rollback_to_cluster_snap) == 0) { | |
1550 | derr << "rollback to cluster snapshot '" << m_osd_rollback_to_cluster_snap << "': not found" << dendl; | |
1551 | ret = -ENOENT; | |
1552 | goto close_basedir_fd; | |
1553 | } | |
1554 | ||
1555 | char nosnapfn[200]; | |
1556 | snprintf(nosnapfn, sizeof(nosnapfn), "%s/nosnap", current_fn.c_str()); | |
1557 | ||
1558 | if (backend->can_checkpoint()) { | |
1559 | if (snaps.empty()) { | |
31f18b77 | 1560 | dout(0) << __FUNC__ << ": WARNING: no consistent snaps found, store may be in inconsistent state" << dendl; |
7c673cae FG |
1561 | } else { |
1562 | char s[NAME_MAX]; | |
1563 | uint64_t curr_seq = 0; | |
1564 | ||
1565 | if (m_osd_rollback_to_cluster_snap.length()) { | |
1566 | derr << TEXT_RED | |
1567 | << " ** NOTE: rolling back to cluster snapshot " << m_osd_rollback_to_cluster_snap << " **" | |
1568 | << TEXT_NORMAL | |
1569 | << dendl; | |
1570 | assert(cluster_snaps.count(m_osd_rollback_to_cluster_snap)); | |
1571 | snprintf(s, sizeof(s), CLUSTER_SNAP_ITEM, m_osd_rollback_to_cluster_snap.c_str()); | |
1572 | } else { | |
1573 | { | |
1574 | int fd = read_op_seq(&curr_seq); | |
1575 | if (fd >= 0) { | |
1576 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
1577 | } | |
1578 | } | |
1579 | if (curr_seq) | |
1580 | dout(10) << " current/ seq was " << curr_seq << dendl; | |
1581 | else | |
1582 | dout(10) << " current/ missing entirely (unusual, but okay)" << dendl; | |
1583 | ||
1584 | uint64_t cp = snaps.back(); | |
1585 | dout(10) << " most recent snap from " << snaps << " is " << cp << dendl; | |
1586 | ||
1587 | // if current/ is marked as non-snapshotted, refuse to roll | |
1588 | // back (without clear direction) to avoid throwing out new | |
1589 | // data. | |
1590 | struct stat st; | |
1591 | if (::stat(nosnapfn, &st) == 0) { | |
1592 | if (!m_osd_use_stale_snap) { | |
1593 | derr << "ERROR: " << nosnapfn << " exists, not rolling back to avoid losing new data" << dendl; | |
1594 | derr << "Force rollback to old snapshotted version with 'osd use stale snap = true'" << dendl; | |
1595 | derr << "config option for --osd-use-stale-snap startup argument." << dendl; | |
1596 | ret = -ENOTSUP; | |
1597 | goto close_basedir_fd; | |
1598 | } | |
1599 | derr << "WARNING: user forced start with data sequence mismatch: current was " << curr_seq | |
1600 | << ", newest snap is " << cp << dendl; | |
1601 | cerr << TEXT_YELLOW | |
1602 | << " ** WARNING: forcing the use of stale snapshot data **" | |
1603 | << TEXT_NORMAL << std::endl; | |
1604 | } | |
1605 | ||
31f18b77 | 1606 | dout(10) << __FUNC__ << ": rolling back to consistent snap " << cp << dendl; |
7c673cae FG |
1607 | snprintf(s, sizeof(s), COMMIT_SNAP_ITEM, (long long unsigned)cp); |
1608 | } | |
1609 | ||
1610 | // drop current? | |
1611 | ret = backend->rollback_to(s); | |
1612 | if (ret) { | |
31f18b77 | 1613 | derr << __FUNC__ << ": error rolling back to " << s << ": " |
7c673cae FG |
1614 | << cpp_strerror(ret) << dendl; |
1615 | goto close_basedir_fd; | |
1616 | } | |
1617 | } | |
1618 | } | |
1619 | initial_op_seq = 0; | |
1620 | ||
1621 | current_fd = ::open(current_fn.c_str(), O_RDONLY); | |
1622 | if (current_fd < 0) { | |
1623 | ret = -errno; | |
31f18b77 | 1624 | derr << __FUNC__ << ": error opening: " << current_fn << ": " << cpp_strerror(ret) << dendl; |
7c673cae FG |
1625 | goto close_basedir_fd; |
1626 | } | |
1627 | ||
1628 | assert(current_fd >= 0); | |
1629 | ||
1630 | op_fd = read_op_seq(&initial_op_seq); | |
1631 | if (op_fd < 0) { | |
1632 | ret = op_fd; | |
31f18b77 | 1633 | derr << __FUNC__ << ": read_op_seq failed" << dendl; |
7c673cae FG |
1634 | goto close_current_fd; |
1635 | } | |
1636 | ||
1637 | dout(5) << "mount op_seq is " << initial_op_seq << dendl; | |
1638 | if (initial_op_seq == 0) { | |
1639 | derr << "mount initial op seq is 0; something is wrong" << dendl; | |
1640 | ret = -EINVAL; | |
1641 | goto close_current_fd; | |
1642 | } | |
1643 | ||
1644 | if (!backend->can_checkpoint()) { | |
1645 | // mark current/ as non-snapshotted so that we don't rollback away | |
1646 | // from it. | |
1647 | int r = ::creat(nosnapfn, 0644); | |
1648 | if (r < 0) { | |
1649 | ret = -errno; | |
31f18b77 | 1650 | derr << __FUNC__ << ": failed to create current/nosnap" << dendl; |
7c673cae FG |
1651 | goto close_current_fd; |
1652 | } | |
1653 | VOID_TEMP_FAILURE_RETRY(::close(r)); | |
1654 | } else { | |
1655 | // clear nosnap marker, if present. | |
1656 | ::unlink(nosnapfn); | |
1657 | } | |
1658 | ||
1659 | // check fsid with omap | |
1660 | // get omap fsid | |
1661 | int omap_fsid_fd; | |
1662 | char omap_fsid_buf[PATH_MAX]; | |
1663 | struct ::stat omap_fsid_stat; | |
1664 | snprintf(omap_fsid_buf, sizeof(omap_fsid_buf), "%s/osd_uuid", omap_dir.c_str()); | |
1665 | // if osd_uuid not exists, assume as this omap matchs corresponding osd | |
1666 | if (::stat(omap_fsid_buf, &omap_fsid_stat) != 0){ | |
31f18b77 | 1667 | dout(10) << __FUNC__ << ": osd_uuid not found under omap, " |
7c673cae FG |
1668 | << "assume as matched." |
1669 | << dendl; | |
1670 | }else{ | |
1671 | // if osd_uuid exists, compares osd_uuid with fsid | |
1672 | omap_fsid_fd = ::open(omap_fsid_buf, O_RDONLY, 0644); | |
1673 | if (omap_fsid_fd < 0) { | |
1674 | ret = -errno; | |
31f18b77 | 1675 | derr << __FUNC__ << ": error opening '" << omap_fsid_buf << "': " |
7c673cae FG |
1676 | << cpp_strerror(ret) |
1677 | << dendl; | |
1678 | goto close_current_fd; | |
1679 | } | |
1680 | ret = read_fsid(omap_fsid_fd, &omap_fsid); | |
1681 | VOID_TEMP_FAILURE_RETRY(::close(omap_fsid_fd)); | |
1682 | omap_fsid_fd = -1; // defensive | |
1683 | if (ret < 0) { | |
31f18b77 | 1684 | derr << __FUNC__ << ": error reading omap_fsid_fd" |
7c673cae FG |
1685 | << ", omap_fsid = " << omap_fsid |
1686 | << cpp_strerror(ret) | |
1687 | << dendl; | |
1688 | goto close_current_fd; | |
1689 | } | |
1690 | if (fsid != omap_fsid) { | |
31f18b77 | 1691 | derr << __FUNC__ << ": " << omap_fsid_buf |
7c673cae FG |
1692 | << " has existed omap fsid " << omap_fsid |
1693 | << " != expected osd fsid " << fsid | |
1694 | << dendl; | |
1695 | ret = -EINVAL; | |
1696 | goto close_current_fd; | |
1697 | } | |
1698 | } | |
1699 | ||
1700 | dout(0) << "start omap initiation" << dendl; | |
1701 | if (!(generic_flags & SKIP_MOUNT_OMAP)) { | |
1702 | KeyValueDB * omap_store = KeyValueDB::create(cct, | |
1703 | superblock.omap_backend, | |
1704 | omap_dir); | |
1705 | if (omap_store == NULL) | |
1706 | { | |
31f18b77 | 1707 | derr << __FUNC__ << ": Error creating " << superblock.omap_backend << dendl; |
7c673cae FG |
1708 | ret = -1; |
1709 | goto close_current_fd; | |
1710 | } | |
1711 | ||
1712 | if (superblock.omap_backend == "rocksdb") | |
1713 | ret = omap_store->init(cct->_conf->filestore_rocksdb_options); | |
1714 | else | |
1715 | ret = omap_store->init(); | |
1716 | ||
1717 | if (ret < 0) { | |
31f18b77 | 1718 | derr << __FUNC__ << ": Error initializing omap_store: " << cpp_strerror(ret) << dendl; |
7c673cae FG |
1719 | goto close_current_fd; |
1720 | } | |
1721 | ||
1722 | stringstream err; | |
1723 | if (omap_store->create_and_open(err)) { | |
1724 | delete omap_store; | |
31f18b77 | 1725 | derr << __FUNC__ << ": Error initializing " << superblock.omap_backend |
7c673cae FG |
1726 | << " : " << err.str() << dendl; |
1727 | ret = -1; | |
1728 | goto close_current_fd; | |
1729 | } | |
1730 | ||
1731 | DBObjectMap *dbomap = new DBObjectMap(cct, omap_store); | |
1732 | ret = dbomap->init(do_update); | |
1733 | if (ret < 0) { | |
1734 | delete dbomap; | |
31f18b77 | 1735 | derr << __FUNC__ << ": Error initializing DBObjectMap: " << ret << dendl; |
7c673cae FG |
1736 | goto close_current_fd; |
1737 | } | |
1738 | stringstream err2; | |
1739 | ||
1740 | if (cct->_conf->filestore_debug_omap_check && !dbomap->check(err2)) { | |
1741 | derr << err2.str() << dendl; | |
1742 | delete dbomap; | |
1743 | ret = -EINVAL; | |
1744 | goto close_current_fd; | |
1745 | } | |
1746 | object_map.reset(dbomap); | |
1747 | } | |
1748 | ||
1749 | // journal | |
1750 | new_journal(); | |
1751 | ||
1752 | // select journal mode? | |
1753 | if (journal) { | |
1754 | if (!m_filestore_journal_writeahead && | |
1755 | !m_filestore_journal_parallel && | |
1756 | !m_filestore_journal_trailing) { | |
1757 | if (!backend->can_checkpoint()) { | |
1758 | m_filestore_journal_writeahead = true; | |
31f18b77 | 1759 | dout(0) << __FUNC__ << ": enabling WRITEAHEAD journal mode: checkpoint is not enabled" << dendl; |
7c673cae FG |
1760 | } else { |
1761 | m_filestore_journal_parallel = true; | |
31f18b77 | 1762 | dout(0) << __FUNC__ << ": enabling PARALLEL journal mode: fs, checkpoint is enabled" << dendl; |
7c673cae FG |
1763 | } |
1764 | } else { | |
1765 | if (m_filestore_journal_writeahead) | |
31f18b77 | 1766 | dout(0) << __FUNC__ << ": WRITEAHEAD journal mode explicitly enabled in conf" << dendl; |
7c673cae | 1767 | if (m_filestore_journal_parallel) |
31f18b77 | 1768 | dout(0) << __FUNC__ << ": PARALLEL journal mode explicitly enabled in conf" << dendl; |
7c673cae | 1769 | if (m_filestore_journal_trailing) |
31f18b77 | 1770 | dout(0) << __FUNC__ << ": TRAILING journal mode explicitly enabled in conf" << dendl; |
7c673cae FG |
1771 | } |
1772 | if (m_filestore_journal_writeahead) | |
1773 | journal->set_wait_on_full(true); | |
1774 | } else { | |
31f18b77 | 1775 | dout(0) << __FUNC__ << ": no journal" << dendl; |
7c673cae FG |
1776 | } |
1777 | ||
1778 | ret = _sanity_check_fs(); | |
1779 | if (ret) { | |
31f18b77 | 1780 | derr << __FUNC__ << ": _sanity_check_fs failed with error " |
7c673cae FG |
1781 | << ret << dendl; |
1782 | goto close_current_fd; | |
1783 | } | |
1784 | ||
1785 | // Cleanup possibly invalid collections | |
1786 | { | |
1787 | vector<coll_t> collections; | |
1788 | ret = list_collections(collections, true); | |
1789 | if (ret < 0) { | |
1790 | derr << "Error " << ret << " while listing collections" << dendl; | |
1791 | goto close_current_fd; | |
1792 | } | |
1793 | for (vector<coll_t>::iterator i = collections.begin(); | |
1794 | i != collections.end(); | |
1795 | ++i) { | |
1796 | Index index; | |
1797 | ret = get_index(*i, &index); | |
1798 | if (ret < 0) { | |
1799 | derr << "Unable to mount index " << *i | |
1800 | << " with error: " << ret << dendl; | |
1801 | goto close_current_fd; | |
1802 | } | |
1803 | assert(NULL != index.index); | |
1804 | RWLock::WLocker l((index.index)->access_lock); | |
1805 | ||
1806 | index->cleanup(); | |
1807 | } | |
1808 | } | |
1809 | if (!m_disable_wbthrottle) { | |
1810 | wbthrottle.start(); | |
1811 | } else { | |
31f18b77 | 1812 | dout(0) << __FUNC__ << ": INFO: WbThrottle is disabled" << dendl; |
7c673cae | 1813 | if (cct->_conf->filestore_odsync_write) { |
31f18b77 | 1814 | dout(0) << __FUNC__ << ": INFO: O_DSYNC write is enabled" << dendl; |
7c673cae FG |
1815 | } |
1816 | } | |
1817 | sync_thread.create("filestore_sync"); | |
1818 | ||
1819 | if (!(generic_flags & SKIP_JOURNAL_REPLAY)) { | |
1820 | ret = journal_replay(initial_op_seq); | |
1821 | if (ret < 0) { | |
31f18b77 | 1822 | derr << __FUNC__ << ": failed to open journal " << journalpath << ": " << cpp_strerror(ret) << dendl; |
7c673cae FG |
1823 | if (ret == -ENOTTY) { |
1824 | derr << "maybe journal is not pointing to a block device and its size " | |
1825 | << "wasn't configured?" << dendl; | |
1826 | } | |
1827 | ||
1828 | goto stop_sync; | |
1829 | } | |
1830 | } | |
1831 | ||
1832 | { | |
1833 | stringstream err2; | |
1834 | if (cct->_conf->filestore_debug_omap_check && !object_map->check(err2)) { | |
1835 | derr << err2.str() << dendl; | |
1836 | ret = -EINVAL; | |
1837 | goto stop_sync; | |
1838 | } | |
1839 | } | |
1840 | ||
1841 | init_temp_collections(); | |
1842 | ||
1843 | journal_start(); | |
1844 | ||
1845 | op_tp.start(); | |
1846 | for (vector<Finisher*>::iterator it = ondisk_finishers.begin(); it != ondisk_finishers.end(); ++it) { | |
1847 | (*it)->start(); | |
1848 | } | |
1849 | for (vector<Finisher*>::iterator it = apply_finishers.begin(); it != apply_finishers.end(); ++it) { | |
1850 | (*it)->start(); | |
1851 | } | |
1852 | ||
1853 | timer.init(); | |
1854 | ||
1855 | // upgrade? | |
1856 | if (cct->_conf->filestore_update_to >= (int)get_target_version()) { | |
1857 | int err = upgrade(); | |
1858 | if (err < 0) { | |
1859 | derr << "error converting store" << dendl; | |
1860 | umount(); | |
1861 | return err; | |
1862 | } | |
1863 | } | |
1864 | ||
1865 | // all okay. | |
1866 | return 0; | |
1867 | ||
1868 | stop_sync: | |
1869 | // stop sync thread | |
1870 | lock.Lock(); | |
1871 | stop = true; | |
1872 | sync_cond.Signal(); | |
1873 | lock.Unlock(); | |
1874 | sync_thread.join(); | |
1875 | if (!m_disable_wbthrottle) { | |
1876 | wbthrottle.stop(); | |
1877 | } | |
1878 | close_current_fd: | |
1879 | VOID_TEMP_FAILURE_RETRY(::close(current_fd)); | |
1880 | current_fd = -1; | |
1881 | close_basedir_fd: | |
1882 | VOID_TEMP_FAILURE_RETRY(::close(basedir_fd)); | |
1883 | basedir_fd = -1; | |
1884 | close_fsid_fd: | |
1885 | VOID_TEMP_FAILURE_RETRY(::close(fsid_fd)); | |
1886 | fsid_fd = -1; | |
1887 | done: | |
1888 | assert(!m_filestore_fail_eio || ret != -EIO); | |
1889 | delete backend; | |
1890 | backend = NULL; | |
1891 | object_map.reset(); | |
1892 | return ret; | |
1893 | } | |
1894 | ||
1895 | void FileStore::init_temp_collections() | |
1896 | { | |
31f18b77 | 1897 | dout(10) << __FUNC__ << dendl; |
7c673cae FG |
1898 | vector<coll_t> ls; |
1899 | int r = list_collections(ls, true); | |
1900 | assert(r >= 0); | |
1901 | ||
1902 | dout(20) << " ls " << ls << dendl; | |
1903 | ||
1904 | SequencerPosition spos; | |
1905 | ||
1906 | set<coll_t> temps; | |
1907 | for (vector<coll_t>::iterator p = ls.begin(); p != ls.end(); ++p) | |
1908 | if (p->is_temp()) | |
1909 | temps.insert(*p); | |
1910 | dout(20) << " temps " << temps << dendl; | |
1911 | ||
1912 | for (vector<coll_t>::iterator p = ls.begin(); p != ls.end(); ++p) { | |
1913 | if (p->is_temp()) | |
1914 | continue; | |
1915 | if (p->is_meta()) | |
1916 | continue; | |
1917 | coll_t temp = p->get_temp(); | |
1918 | if (temps.count(temp)) { | |
1919 | temps.erase(temp); | |
1920 | } else { | |
31f18b77 | 1921 | dout(10) << __FUNC__ << ": creating " << temp << dendl; |
7c673cae FG |
1922 | r = _create_collection(temp, 0, spos); |
1923 | assert(r == 0); | |
1924 | } | |
1925 | } | |
1926 | ||
1927 | for (set<coll_t>::iterator p = temps.begin(); p != temps.end(); ++p) { | |
31f18b77 | 1928 | dout(10) << __FUNC__ << ": removing stray " << *p << dendl; |
7c673cae FG |
1929 | r = _collection_remove_recursive(*p, spos); |
1930 | assert(r == 0); | |
1931 | } | |
1932 | } | |
1933 | ||
1934 | int FileStore::umount() | |
1935 | { | |
31f18b77 | 1936 | dout(5) << __FUNC__ << ": " << basedir << dendl; |
7c673cae FG |
1937 | |
1938 | flush(); | |
1939 | sync(); | |
1940 | do_force_sync(); | |
1941 | ||
1942 | lock.Lock(); | |
1943 | stop = true; | |
1944 | sync_cond.Signal(); | |
1945 | lock.Unlock(); | |
1946 | sync_thread.join(); | |
1947 | if (!m_disable_wbthrottle){ | |
1948 | wbthrottle.stop(); | |
1949 | } | |
1950 | op_tp.stop(); | |
1951 | ||
1952 | journal_stop(); | |
1953 | if (!(generic_flags & SKIP_JOURNAL_REPLAY)) | |
1954 | journal_write_close(); | |
1955 | ||
1956 | for (vector<Finisher*>::iterator it = ondisk_finishers.begin(); it != ondisk_finishers.end(); ++it) { | |
1957 | (*it)->stop(); | |
1958 | } | |
1959 | for (vector<Finisher*>::iterator it = apply_finishers.begin(); it != apply_finishers.end(); ++it) { | |
1960 | (*it)->stop(); | |
1961 | } | |
1962 | ||
1963 | if (fsid_fd >= 0) { | |
1964 | VOID_TEMP_FAILURE_RETRY(::close(fsid_fd)); | |
1965 | fsid_fd = -1; | |
1966 | } | |
1967 | if (op_fd >= 0) { | |
1968 | VOID_TEMP_FAILURE_RETRY(::close(op_fd)); | |
1969 | op_fd = -1; | |
1970 | } | |
1971 | if (current_fd >= 0) { | |
1972 | VOID_TEMP_FAILURE_RETRY(::close(current_fd)); | |
1973 | current_fd = -1; | |
1974 | } | |
1975 | if (basedir_fd >= 0) { | |
1976 | VOID_TEMP_FAILURE_RETRY(::close(basedir_fd)); | |
1977 | basedir_fd = -1; | |
1978 | } | |
1979 | ||
1980 | force_sync = false; | |
1981 | ||
1982 | delete backend; | |
1983 | backend = NULL; | |
1984 | ||
1985 | object_map.reset(); | |
1986 | ||
1987 | { | |
1988 | Mutex::Locker l(sync_entry_timeo_lock); | |
1989 | timer.shutdown(); | |
1990 | } | |
1991 | ||
1992 | // nothing | |
1993 | return 0; | |
1994 | } | |
1995 | ||
1996 | ||
1997 | ||
1998 | ||
1999 | /// ----------------------------- | |
2000 | ||
2001 | FileStore::Op *FileStore::build_op(vector<Transaction>& tls, | |
2002 | Context *onreadable, | |
2003 | Context *onreadable_sync, | |
2004 | TrackedOpRef osd_op) | |
2005 | { | |
2006 | uint64_t bytes = 0, ops = 0; | |
2007 | for (vector<Transaction>::iterator p = tls.begin(); | |
2008 | p != tls.end(); | |
2009 | ++p) { | |
2010 | bytes += (*p).get_num_bytes(); | |
2011 | ops += (*p).get_num_ops(); | |
2012 | } | |
2013 | ||
2014 | Op *o = new Op; | |
2015 | o->start = ceph_clock_now(); | |
2016 | o->tls = std::move(tls); | |
2017 | o->onreadable = onreadable; | |
2018 | o->onreadable_sync = onreadable_sync; | |
2019 | o->ops = ops; | |
2020 | o->bytes = bytes; | |
2021 | o->osd_op = osd_op; | |
2022 | return o; | |
2023 | } | |
2024 | ||
2025 | ||
2026 | ||
2027 | void FileStore::queue_op(OpSequencer *osr, Op *o) | |
2028 | { | |
2029 | // queue op on sequencer, then queue sequencer for the threadpool, | |
2030 | // so that regardless of which order the threads pick up the | |
2031 | // sequencer, the op order will be preserved. | |
2032 | ||
2033 | osr->queue(o); | |
2034 | o->trace.event("queued"); | |
2035 | ||
2036 | logger->inc(l_filestore_ops); | |
2037 | logger->inc(l_filestore_bytes, o->bytes); | |
2038 | ||
31f18b77 | 2039 | dout(5) << __FUNC__ << ": " << o << " seq " << o->op |
7c673cae FG |
2040 | << " " << *osr |
2041 | << " " << o->bytes << " bytes" | |
2042 | << " (queue has " << throttle_ops.get_current() << " ops and " << throttle_bytes.get_current() << " bytes)" | |
2043 | << dendl; | |
2044 | op_wq.queue(osr); | |
2045 | } | |
2046 | ||
2047 | void FileStore::op_queue_reserve_throttle(Op *o) | |
2048 | { | |
2049 | throttle_ops.get(); | |
2050 | throttle_bytes.get(o->bytes); | |
2051 | ||
2052 | logger->set(l_filestore_op_queue_ops, throttle_ops.get_current()); | |
2053 | logger->set(l_filestore_op_queue_bytes, throttle_bytes.get_current()); | |
2054 | } | |
2055 | ||
2056 | void FileStore::op_queue_release_throttle(Op *o) | |
2057 | { | |
2058 | throttle_ops.put(); | |
2059 | throttle_bytes.put(o->bytes); | |
2060 | logger->set(l_filestore_op_queue_ops, throttle_ops.get_current()); | |
2061 | logger->set(l_filestore_op_queue_bytes, throttle_bytes.get_current()); | |
2062 | } | |
2063 | ||
2064 | void FileStore::_do_op(OpSequencer *osr, ThreadPool::TPHandle &handle) | |
2065 | { | |
2066 | if (!m_disable_wbthrottle) { | |
2067 | wbthrottle.throttle(); | |
2068 | } | |
2069 | // inject a stall? | |
2070 | if (cct->_conf->filestore_inject_stall) { | |
2071 | int orig = cct->_conf->filestore_inject_stall; | |
31f18b77 | 2072 | dout(5) << __FUNC__ << ": filestore_inject_stall " << orig << ", sleeping" << dendl; |
7c673cae FG |
2073 | sleep(orig); |
2074 | cct->_conf->set_val("filestore_inject_stall", "0"); | |
31f18b77 | 2075 | dout(5) << __FUNC__ << ": done stalling" << dendl; |
7c673cae FG |
2076 | } |
2077 | ||
2078 | osr->apply_lock.Lock(); | |
2079 | Op *o = osr->peek_queue(); | |
2080 | o->trace.event("op_apply_start"); | |
2081 | apply_manager.op_apply_start(o->op); | |
31f18b77 | 2082 | dout(5) << __FUNC__ << ": " << o << " seq " << o->op << " " << *osr << "/" << osr->parent << " start" << dendl; |
7c673cae FG |
2083 | o->trace.event("_do_transactions start"); |
2084 | int r = _do_transactions(o->tls, o->op, &handle); | |
2085 | o->trace.event("op_apply_finish"); | |
2086 | apply_manager.op_apply_finish(o->op); | |
31f18b77 | 2087 | dout(10) << __FUNC__ << ": " << o << " seq " << o->op << " r = " << r |
7c673cae FG |
2088 | << ", finisher " << o->onreadable << " " << o->onreadable_sync << dendl; |
2089 | ||
2090 | o->tls.clear(); | |
2091 | ||
2092 | } | |
2093 | ||
2094 | void FileStore::_finish_op(OpSequencer *osr) | |
2095 | { | |
2096 | list<Context*> to_queue; | |
2097 | Op *o = osr->dequeue(&to_queue); | |
2098 | ||
2099 | utime_t lat = ceph_clock_now(); | |
2100 | lat -= o->start; | |
2101 | ||
31f18b77 | 2102 | dout(10) << __FUNC__ << ": " << o << " seq " << o->op << " " << *osr << "/" << osr->parent << " lat " << lat << dendl; |
7c673cae FG |
2103 | osr->apply_lock.Unlock(); // locked in _do_op |
2104 | o->trace.event("_finish_op"); | |
2105 | ||
2106 | // called with tp lock held | |
2107 | op_queue_release_throttle(o); | |
2108 | ||
2109 | logger->tinc(l_filestore_apply_latency, lat); | |
2110 | ||
2111 | if (o->onreadable_sync) { | |
2112 | o->onreadable_sync->complete(0); | |
2113 | } | |
2114 | if (o->onreadable) { | |
2115 | apply_finishers[osr->id % m_apply_finisher_num]->queue(o->onreadable); | |
2116 | } | |
2117 | if (!to_queue.empty()) { | |
2118 | apply_finishers[osr->id % m_apply_finisher_num]->queue(to_queue); | |
2119 | } | |
2120 | delete o; | |
2121 | } | |
2122 | ||
2123 | ||
2124 | struct C_JournaledAhead : public Context { | |
2125 | FileStore *fs; | |
2126 | FileStore::OpSequencer *osr; | |
2127 | FileStore::Op *o; | |
2128 | Context *ondisk; | |
2129 | ||
2130 | C_JournaledAhead(FileStore *f, FileStore::OpSequencer *os, FileStore::Op *o, Context *ondisk): | |
2131 | fs(f), osr(os), o(o), ondisk(ondisk) { } | |
2132 | void finish(int r) override { | |
2133 | fs->_journaled_ahead(osr, o, ondisk); | |
2134 | } | |
2135 | }; | |
2136 | ||
2137 | int FileStore::queue_transactions(Sequencer *posr, vector<Transaction>& tls, | |
2138 | TrackedOpRef osd_op, | |
2139 | ThreadPool::TPHandle *handle) | |
2140 | { | |
2141 | Context *onreadable; | |
2142 | Context *ondisk; | |
2143 | Context *onreadable_sync; | |
2144 | ObjectStore::Transaction::collect_contexts( | |
2145 | tls, &onreadable, &ondisk, &onreadable_sync); | |
2146 | ||
2147 | if (cct->_conf->objectstore_blackhole) { | |
31f18b77 | 2148 | dout(0) << __FUNC__ << ": objectstore_blackhole = TRUE, dropping transaction" |
7c673cae FG |
2149 | << dendl; |
2150 | delete ondisk; | |
2151 | delete onreadable; | |
2152 | delete onreadable_sync; | |
2153 | return 0; | |
2154 | } | |
2155 | ||
2156 | utime_t start = ceph_clock_now(); | |
2157 | // set up the sequencer | |
2158 | OpSequencer *osr; | |
2159 | assert(posr); | |
2160 | if (posr->p) { | |
2161 | osr = static_cast<OpSequencer *>(posr->p.get()); | |
31f18b77 | 2162 | dout(5) << __FUNC__ << ": existing " << osr << " " << *osr << dendl; |
7c673cae | 2163 | } else { |
31f18b77 | 2164 | osr = new OpSequencer(cct, ++next_osr_id); |
7c673cae FG |
2165 | osr->set_cct(cct); |
2166 | osr->parent = posr; | |
2167 | posr->p = osr; | |
31f18b77 | 2168 | dout(5) << __FUNC__ << ": new " << osr << " " << *osr << dendl; |
7c673cae FG |
2169 | } |
2170 | ||
2171 | // used to include osr information in tracepoints during transaction apply | |
2172 | for (vector<Transaction>::iterator i = tls.begin(); i != tls.end(); ++i) { | |
2173 | (*i).set_osr(osr); | |
2174 | } | |
2175 | ||
2176 | ZTracer::Trace trace; | |
2177 | if (osd_op && osd_op->pg_trace) { | |
2178 | osd_op->store_trace.init("filestore op", &trace_endpoint, &osd_op->pg_trace); | |
2179 | trace = osd_op->store_trace; | |
2180 | } | |
2181 | ||
2182 | if (journal && journal->is_writeable() && !m_filestore_journal_trailing) { | |
2183 | Op *o = build_op(tls, onreadable, onreadable_sync, osd_op); | |
2184 | ||
2185 | //prepare and encode transactions data out of lock | |
2186 | bufferlist tbl; | |
2187 | int orig_len = journal->prepare_entry(o->tls, &tbl); | |
2188 | ||
2189 | if (handle) | |
2190 | handle->suspend_tp_timeout(); | |
2191 | ||
2192 | op_queue_reserve_throttle(o); | |
2193 | journal->reserve_throttle_and_backoff(tbl.length()); | |
2194 | ||
2195 | if (handle) | |
2196 | handle->reset_tp_timeout(); | |
2197 | ||
2198 | uint64_t op_num = submit_manager.op_submit_start(); | |
2199 | o->op = op_num; | |
2200 | trace.keyval("opnum", op_num); | |
2201 | ||
2202 | if (m_filestore_do_dump) | |
2203 | dump_transactions(o->tls, o->op, osr); | |
2204 | ||
2205 | if (m_filestore_journal_parallel) { | |
31f18b77 | 2206 | dout(5) << __FUNC__ << ": (parallel) " << o->op << " " << o->tls << dendl; |
7c673cae FG |
2207 | |
2208 | trace.keyval("journal mode", "parallel"); | |
2209 | trace.event("journal started"); | |
2210 | _op_journal_transactions(tbl, orig_len, o->op, ondisk, osd_op); | |
2211 | ||
2212 | // queue inside submit_manager op submission lock | |
2213 | queue_op(osr, o); | |
2214 | trace.event("op queued"); | |
2215 | } else if (m_filestore_journal_writeahead) { | |
31f18b77 | 2216 | dout(5) << __FUNC__ << ": (writeahead) " << o->op << " " << o->tls << dendl; |
7c673cae FG |
2217 | |
2218 | osr->queue_journal(o->op); | |
2219 | ||
2220 | trace.keyval("journal mode", "writeahead"); | |
2221 | trace.event("journal started"); | |
2222 | _op_journal_transactions(tbl, orig_len, o->op, | |
2223 | new C_JournaledAhead(this, osr, o, ondisk), | |
2224 | osd_op); | |
2225 | } else { | |
2226 | ceph_abort(); | |
2227 | } | |
2228 | submit_manager.op_submit_finish(op_num); | |
2229 | utime_t end = ceph_clock_now(); | |
2230 | logger->tinc(l_filestore_queue_transaction_latency_avg, end - start); | |
2231 | return 0; | |
2232 | } | |
2233 | ||
2234 | if (!journal) { | |
2235 | Op *o = build_op(tls, onreadable, onreadable_sync, osd_op); | |
31f18b77 | 2236 | dout(5) << __FUNC__ << ": (no journal) " << o << " " << tls << dendl; |
7c673cae FG |
2237 | |
2238 | if (handle) | |
2239 | handle->suspend_tp_timeout(); | |
2240 | ||
2241 | op_queue_reserve_throttle(o); | |
2242 | ||
2243 | if (handle) | |
2244 | handle->reset_tp_timeout(); | |
2245 | ||
2246 | uint64_t op_num = submit_manager.op_submit_start(); | |
2247 | o->op = op_num; | |
2248 | ||
2249 | if (m_filestore_do_dump) | |
2250 | dump_transactions(o->tls, o->op, osr); | |
2251 | ||
2252 | queue_op(osr, o); | |
2253 | trace.keyval("opnum", op_num); | |
2254 | trace.keyval("journal mode", "none"); | |
2255 | trace.event("op queued"); | |
2256 | ||
2257 | if (ondisk) | |
2258 | apply_manager.add_waiter(op_num, ondisk); | |
2259 | submit_manager.op_submit_finish(op_num); | |
2260 | utime_t end = ceph_clock_now(); | |
2261 | logger->tinc(l_filestore_queue_transaction_latency_avg, end - start); | |
2262 | return 0; | |
2263 | } | |
2264 | ||
2265 | assert(journal); | |
2266 | //prepare and encode transactions data out of lock | |
2267 | bufferlist tbl; | |
2268 | int orig_len = -1; | |
2269 | if (journal->is_writeable()) { | |
2270 | orig_len = journal->prepare_entry(tls, &tbl); | |
2271 | } | |
2272 | uint64_t op = submit_manager.op_submit_start(); | |
31f18b77 | 2273 | dout(5) << __FUNC__ << ": (trailing journal) " << op << " " << tls << dendl; |
7c673cae FG |
2274 | |
2275 | if (m_filestore_do_dump) | |
2276 | dump_transactions(tls, op, osr); | |
2277 | ||
2278 | trace.event("op_apply_start"); | |
2279 | trace.keyval("opnum", op); | |
2280 | trace.keyval("journal mode", "trailing"); | |
2281 | apply_manager.op_apply_start(op); | |
2282 | trace.event("do_transactions"); | |
2283 | int r = do_transactions(tls, op); | |
2284 | ||
2285 | if (r >= 0) { | |
2286 | trace.event("journal started"); | |
2287 | _op_journal_transactions(tbl, orig_len, op, ondisk, osd_op); | |
2288 | } else { | |
2289 | delete ondisk; | |
2290 | } | |
2291 | ||
2292 | // start on_readable finisher after we queue journal item, as on_readable callback | |
2293 | // is allowed to delete the Transaction | |
2294 | if (onreadable_sync) { | |
2295 | onreadable_sync->complete(r); | |
2296 | } | |
2297 | apply_finishers[osr->id % m_apply_finisher_num]->queue(onreadable, r); | |
2298 | ||
2299 | submit_manager.op_submit_finish(op); | |
2300 | trace.event("op_apply_finish"); | |
2301 | apply_manager.op_apply_finish(op); | |
2302 | ||
2303 | utime_t end = ceph_clock_now(); | |
2304 | logger->tinc(l_filestore_queue_transaction_latency_avg, end - start); | |
2305 | return r; | |
2306 | } | |
2307 | ||
2308 | void FileStore::_journaled_ahead(OpSequencer *osr, Op *o, Context *ondisk) | |
2309 | { | |
31f18b77 | 2310 | dout(5) << __FUNC__ << ": " << o << " seq " << o->op << " " << *osr << " " << o->tls << dendl; |
7c673cae FG |
2311 | |
2312 | o->trace.event("writeahead journal finished"); | |
2313 | ||
2314 | // this should queue in order because the journal does it's completions in order. | |
2315 | queue_op(osr, o); | |
2316 | ||
2317 | list<Context*> to_queue; | |
2318 | osr->dequeue_journal(&to_queue); | |
2319 | ||
2320 | // do ondisk completions async, to prevent any onreadable_sync completions | |
2321 | // getting blocked behind an ondisk completion. | |
2322 | if (ondisk) { | |
2323 | dout(10) << " queueing ondisk " << ondisk << dendl; | |
2324 | ondisk_finishers[osr->id % m_ondisk_finisher_num]->queue(ondisk); | |
2325 | } | |
2326 | if (!to_queue.empty()) { | |
2327 | ondisk_finishers[osr->id % m_ondisk_finisher_num]->queue(to_queue); | |
2328 | } | |
2329 | } | |
2330 | ||
2331 | int FileStore::_do_transactions( | |
2332 | vector<Transaction> &tls, | |
2333 | uint64_t op_seq, | |
2334 | ThreadPool::TPHandle *handle) | |
2335 | { | |
2336 | int trans_num = 0; | |
2337 | ||
2338 | for (vector<Transaction>::iterator p = tls.begin(); | |
2339 | p != tls.end(); | |
2340 | ++p, trans_num++) { | |
2341 | _do_transaction(*p, op_seq, trans_num, handle); | |
2342 | if (handle) | |
2343 | handle->reset_tp_timeout(); | |
2344 | } | |
2345 | ||
2346 | return 0; | |
2347 | } | |
2348 | ||
2349 | void FileStore::_set_global_replay_guard(const coll_t& cid, | |
2350 | const SequencerPosition &spos) | |
2351 | { | |
2352 | if (backend->can_checkpoint()) | |
2353 | return; | |
2354 | ||
2355 | // sync all previous operations on this sequencer | |
2356 | int ret = object_map->sync(); | |
2357 | if (ret < 0) { | |
31f18b77 | 2358 | derr << __FUNC__ << ": omap sync error " << cpp_strerror(ret) << dendl; |
7c673cae FG |
2359 | assert(0 == "_set_global_replay_guard failed"); |
2360 | } | |
2361 | ret = sync_filesystem(basedir_fd); | |
2362 | if (ret < 0) { | |
31f18b77 | 2363 | derr << __FUNC__ << ": sync_filesystem error " << cpp_strerror(ret) << dendl; |
7c673cae FG |
2364 | assert(0 == "_set_global_replay_guard failed"); |
2365 | } | |
2366 | ||
2367 | char fn[PATH_MAX]; | |
2368 | get_cdir(cid, fn, sizeof(fn)); | |
2369 | int fd = ::open(fn, O_RDONLY); | |
2370 | if (fd < 0) { | |
2371 | int err = errno; | |
31f18b77 | 2372 | derr << __FUNC__ << ": " << cid << " error " << cpp_strerror(err) << dendl; |
7c673cae FG |
2373 | assert(0 == "_set_global_replay_guard failed"); |
2374 | } | |
2375 | ||
2376 | _inject_failure(); | |
2377 | ||
2378 | // then record that we did it | |
2379 | bufferlist v; | |
2380 | ::encode(spos, v); | |
2381 | int r = chain_fsetxattr<true, true>( | |
2382 | fd, GLOBAL_REPLAY_GUARD_XATTR, v.c_str(), v.length()); | |
2383 | if (r < 0) { | |
31f18b77 | 2384 | derr << __FUNC__ << ": fsetxattr " << GLOBAL_REPLAY_GUARD_XATTR |
7c673cae FG |
2385 | << " got " << cpp_strerror(r) << dendl; |
2386 | assert(0 == "fsetxattr failed"); | |
2387 | } | |
2388 | ||
2389 | // and make sure our xattr is durable. | |
2390 | ::fsync(fd); | |
2391 | ||
2392 | _inject_failure(); | |
2393 | ||
2394 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
31f18b77 | 2395 | dout(10) << __FUNC__ << ": " << spos << " done" << dendl; |
7c673cae FG |
2396 | } |
2397 | ||
2398 | int FileStore::_check_global_replay_guard(const coll_t& cid, | |
2399 | const SequencerPosition& spos) | |
2400 | { | |
2401 | char fn[PATH_MAX]; | |
2402 | get_cdir(cid, fn, sizeof(fn)); | |
2403 | int fd = ::open(fn, O_RDONLY); | |
2404 | if (fd < 0) { | |
31f18b77 | 2405 | dout(10) << __FUNC__ << ": " << cid << " dne" << dendl; |
7c673cae FG |
2406 | return 1; // if collection does not exist, there is no guard, and we can replay. |
2407 | } | |
2408 | ||
2409 | char buf[100]; | |
2410 | int r = chain_fgetxattr(fd, GLOBAL_REPLAY_GUARD_XATTR, buf, sizeof(buf)); | |
2411 | if (r < 0) { | |
31f18b77 | 2412 | dout(20) << __FUNC__ << ": no xattr" << dendl; |
7c673cae FG |
2413 | assert(!m_filestore_fail_eio || r != -EIO); |
2414 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
2415 | return 1; // no xattr | |
2416 | } | |
2417 | bufferlist bl; | |
2418 | bl.append(buf, r); | |
2419 | ||
2420 | SequencerPosition opos; | |
2421 | bufferlist::iterator p = bl.begin(); | |
2422 | ::decode(opos, p); | |
2423 | ||
2424 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
2425 | return spos >= opos ? 1 : -1; | |
2426 | } | |
2427 | ||
2428 | ||
2429 | void FileStore::_set_replay_guard(const coll_t& cid, | |
2430 | const SequencerPosition &spos, | |
2431 | bool in_progress=false) | |
2432 | { | |
2433 | char fn[PATH_MAX]; | |
2434 | get_cdir(cid, fn, sizeof(fn)); | |
2435 | int fd = ::open(fn, O_RDONLY); | |
2436 | if (fd < 0) { | |
2437 | int err = errno; | |
31f18b77 | 2438 | derr << __FUNC__ << ": " << cid << " error " << cpp_strerror(err) << dendl; |
7c673cae FG |
2439 | assert(0 == "_set_replay_guard failed"); |
2440 | } | |
2441 | _set_replay_guard(fd, spos, 0, in_progress); | |
2442 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
2443 | } | |
2444 | ||
2445 | ||
2446 | void FileStore::_set_replay_guard(int fd, | |
2447 | const SequencerPosition& spos, | |
2448 | const ghobject_t *hoid, | |
2449 | bool in_progress) | |
2450 | { | |
2451 | if (backend->can_checkpoint()) | |
2452 | return; | |
2453 | ||
31f18b77 | 2454 | dout(10) << __FUNC__ << ": " << spos << (in_progress ? " START" : "") << dendl; |
7c673cae FG |
2455 | |
2456 | _inject_failure(); | |
2457 | ||
2458 | // first make sure the previous operation commits | |
2459 | ::fsync(fd); | |
2460 | ||
2461 | if (!in_progress) { | |
2462 | // sync object_map too. even if this object has a header or keys, | |
2463 | // it have had them in the past and then removed them, so always | |
2464 | // sync. | |
2465 | object_map->sync(hoid, &spos); | |
2466 | } | |
2467 | ||
2468 | _inject_failure(); | |
2469 | ||
2470 | // then record that we did it | |
2471 | bufferlist v(40); | |
2472 | ::encode(spos, v); | |
2473 | ::encode(in_progress, v); | |
2474 | int r = chain_fsetxattr<true, true>( | |
2475 | fd, REPLAY_GUARD_XATTR, v.c_str(), v.length()); | |
2476 | if (r < 0) { | |
2477 | derr << "fsetxattr " << REPLAY_GUARD_XATTR << " got " << cpp_strerror(r) << dendl; | |
2478 | assert(0 == "fsetxattr failed"); | |
2479 | } | |
2480 | ||
2481 | // and make sure our xattr is durable. | |
2482 | ::fsync(fd); | |
2483 | ||
2484 | _inject_failure(); | |
2485 | ||
31f18b77 | 2486 | dout(10) << __FUNC__ << ": " << spos << " done" << dendl; |
7c673cae FG |
2487 | } |
2488 | ||
2489 | void FileStore::_close_replay_guard(const coll_t& cid, | |
2490 | const SequencerPosition &spos) | |
2491 | { | |
2492 | char fn[PATH_MAX]; | |
2493 | get_cdir(cid, fn, sizeof(fn)); | |
2494 | int fd = ::open(fn, O_RDONLY); | |
2495 | if (fd < 0) { | |
2496 | int err = errno; | |
31f18b77 | 2497 | derr << __FUNC__ << ": " << cid << " error " << cpp_strerror(err) << dendl; |
7c673cae FG |
2498 | assert(0 == "_close_replay_guard failed"); |
2499 | } | |
2500 | _close_replay_guard(fd, spos); | |
2501 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
2502 | } | |
2503 | ||
2504 | void FileStore::_close_replay_guard(int fd, const SequencerPosition& spos, | |
2505 | const ghobject_t *hoid) | |
2506 | { | |
2507 | if (backend->can_checkpoint()) | |
2508 | return; | |
2509 | ||
31f18b77 | 2510 | dout(10) << __FUNC__ << ": " << spos << dendl; |
7c673cae FG |
2511 | |
2512 | _inject_failure(); | |
2513 | ||
2514 | // sync object_map too. even if this object has a header or keys, | |
2515 | // it have had them in the past and then removed them, so always | |
2516 | // sync. | |
2517 | object_map->sync(hoid, &spos); | |
2518 | ||
2519 | // then record that we are done with this operation | |
2520 | bufferlist v(40); | |
2521 | ::encode(spos, v); | |
2522 | bool in_progress = false; | |
2523 | ::encode(in_progress, v); | |
2524 | int r = chain_fsetxattr<true, true>( | |
2525 | fd, REPLAY_GUARD_XATTR, v.c_str(), v.length()); | |
2526 | if (r < 0) { | |
2527 | derr << "fsetxattr " << REPLAY_GUARD_XATTR << " got " << cpp_strerror(r) << dendl; | |
2528 | assert(0 == "fsetxattr failed"); | |
2529 | } | |
2530 | ||
2531 | // and make sure our xattr is durable. | |
2532 | ::fsync(fd); | |
2533 | ||
2534 | _inject_failure(); | |
2535 | ||
31f18b77 | 2536 | dout(10) << __FUNC__ << ": " << spos << " done" << dendl; |
7c673cae FG |
2537 | } |
2538 | ||
2539 | int FileStore::_check_replay_guard(const coll_t& cid, const ghobject_t &oid, | |
2540 | const SequencerPosition& spos) | |
2541 | { | |
2542 | if (!replaying || backend->can_checkpoint()) | |
2543 | return 1; | |
2544 | ||
2545 | int r = _check_global_replay_guard(cid, spos); | |
2546 | if (r < 0) | |
2547 | return r; | |
2548 | ||
2549 | FDRef fd; | |
2550 | r = lfn_open(cid, oid, false, &fd); | |
2551 | if (r < 0) { | |
31f18b77 | 2552 | dout(10) << __FUNC__ << ": " << cid << " " << oid << " dne" << dendl; |
7c673cae FG |
2553 | return 1; // if file does not exist, there is no guard, and we can replay. |
2554 | } | |
2555 | int ret = _check_replay_guard(**fd, spos); | |
2556 | lfn_close(fd); | |
2557 | return ret; | |
2558 | } | |
2559 | ||
2560 | int FileStore::_check_replay_guard(const coll_t& cid, const SequencerPosition& spos) | |
2561 | { | |
2562 | if (!replaying || backend->can_checkpoint()) | |
2563 | return 1; | |
2564 | ||
2565 | char fn[PATH_MAX]; | |
2566 | get_cdir(cid, fn, sizeof(fn)); | |
2567 | int fd = ::open(fn, O_RDONLY); | |
2568 | if (fd < 0) { | |
31f18b77 | 2569 | dout(10) << __FUNC__ << ": " << cid << " dne" << dendl; |
7c673cae FG |
2570 | return 1; // if collection does not exist, there is no guard, and we can replay. |
2571 | } | |
2572 | int ret = _check_replay_guard(fd, spos); | |
2573 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
2574 | return ret; | |
2575 | } | |
2576 | ||
2577 | int FileStore::_check_replay_guard(int fd, const SequencerPosition& spos) | |
2578 | { | |
2579 | if (!replaying || backend->can_checkpoint()) | |
2580 | return 1; | |
2581 | ||
2582 | char buf[100]; | |
2583 | int r = chain_fgetxattr(fd, REPLAY_GUARD_XATTR, buf, sizeof(buf)); | |
2584 | if (r < 0) { | |
31f18b77 | 2585 | dout(20) << __FUNC__ << ": no xattr" << dendl; |
7c673cae FG |
2586 | assert(!m_filestore_fail_eio || r != -EIO); |
2587 | return 1; // no xattr | |
2588 | } | |
2589 | bufferlist bl; | |
2590 | bl.append(buf, r); | |
2591 | ||
2592 | SequencerPosition opos; | |
2593 | bufferlist::iterator p = bl.begin(); | |
2594 | ::decode(opos, p); | |
2595 | bool in_progress = false; | |
2596 | if (!p.end()) // older journals don't have this | |
2597 | ::decode(in_progress, p); | |
2598 | if (opos > spos) { | |
31f18b77 | 2599 | dout(10) << __FUNC__ << ": object has " << opos << " > current pos " << spos |
7c673cae FG |
2600 | << ", now or in future, SKIPPING REPLAY" << dendl; |
2601 | return -1; | |
2602 | } else if (opos == spos) { | |
2603 | if (in_progress) { | |
31f18b77 | 2604 | dout(10) << __FUNC__ << ": object has " << opos << " == current pos " << spos |
7c673cae FG |
2605 | << ", in_progress=true, CONDITIONAL REPLAY" << dendl; |
2606 | return 0; | |
2607 | } else { | |
31f18b77 | 2608 | dout(10) << __FUNC__ << ": object has " << opos << " == current pos " << spos |
7c673cae FG |
2609 | << ", in_progress=false, SKIPPING REPLAY" << dendl; |
2610 | return -1; | |
2611 | } | |
2612 | } else { | |
31f18b77 | 2613 | dout(10) << __FUNC__ << ": object has " << opos << " < current pos " << spos |
7c673cae FG |
2614 | << ", in past, will replay" << dendl; |
2615 | return 1; | |
2616 | } | |
2617 | } | |
2618 | ||
2619 | void FileStore::_do_transaction( | |
2620 | Transaction& t, uint64_t op_seq, int trans_num, | |
2621 | ThreadPool::TPHandle *handle) | |
2622 | { | |
31f18b77 | 2623 | dout(10) << __FUNC__ << ": on " << &t << dendl; |
7c673cae FG |
2624 | |
2625 | #ifdef WITH_LTTNG | |
2626 | const char *osr_name = t.get_osr() ? static_cast<OpSequencer*>(t.get_osr())->get_name().c_str() : "<NULL>"; | |
2627 | #endif | |
2628 | ||
2629 | Transaction::iterator i = t.begin(); | |
2630 | ||
2631 | SequencerPosition spos(op_seq, trans_num, 0); | |
2632 | while (i.have_op()) { | |
2633 | if (handle) | |
2634 | handle->reset_tp_timeout(); | |
2635 | ||
2636 | Transaction::Op *op = i.decode_op(); | |
2637 | int r = 0; | |
2638 | ||
2639 | _inject_failure(); | |
2640 | ||
2641 | switch (op->op) { | |
2642 | case Transaction::OP_NOP: | |
2643 | break; | |
2644 | case Transaction::OP_TOUCH: | |
2645 | { | |
2646 | const coll_t &_cid = i.get_cid(op->cid); | |
2647 | const ghobject_t &oid = i.get_oid(op->oid); | |
2648 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2649 | _cid : _cid.get_temp(); | |
2650 | tracepoint(objectstore, touch_enter, osr_name); | |
2651 | if (_check_replay_guard(cid, oid, spos) > 0) | |
2652 | r = _touch(cid, oid); | |
2653 | tracepoint(objectstore, touch_exit, r); | |
2654 | } | |
2655 | break; | |
2656 | ||
2657 | case Transaction::OP_WRITE: | |
2658 | { | |
2659 | const coll_t &_cid = i.get_cid(op->cid); | |
2660 | const ghobject_t &oid = i.get_oid(op->oid); | |
2661 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2662 | _cid : _cid.get_temp(); | |
2663 | uint64_t off = op->off; | |
2664 | uint64_t len = op->len; | |
2665 | uint32_t fadvise_flags = i.get_fadvise_flags(); | |
2666 | bufferlist bl; | |
2667 | i.decode_bl(bl); | |
2668 | tracepoint(objectstore, write_enter, osr_name, off, len); | |
2669 | if (_check_replay_guard(cid, oid, spos) > 0) | |
2670 | r = _write(cid, oid, off, len, bl, fadvise_flags); | |
2671 | tracepoint(objectstore, write_exit, r); | |
2672 | } | |
2673 | break; | |
2674 | ||
2675 | case Transaction::OP_ZERO: | |
2676 | { | |
2677 | const coll_t &_cid = i.get_cid(op->cid); | |
2678 | const ghobject_t &oid = i.get_oid(op->oid); | |
2679 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2680 | _cid : _cid.get_temp(); | |
2681 | uint64_t off = op->off; | |
2682 | uint64_t len = op->len; | |
2683 | tracepoint(objectstore, zero_enter, osr_name, off, len); | |
2684 | if (_check_replay_guard(cid, oid, spos) > 0) | |
2685 | r = _zero(cid, oid, off, len); | |
2686 | tracepoint(objectstore, zero_exit, r); | |
2687 | } | |
2688 | break; | |
2689 | ||
2690 | case Transaction::OP_TRIMCACHE: | |
2691 | { | |
2692 | // deprecated, no-op | |
2693 | } | |
2694 | break; | |
2695 | ||
2696 | case Transaction::OP_TRUNCATE: | |
2697 | { | |
2698 | const coll_t &_cid = i.get_cid(op->cid); | |
2699 | const ghobject_t &oid = i.get_oid(op->oid); | |
2700 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2701 | _cid : _cid.get_temp(); | |
2702 | uint64_t off = op->off; | |
2703 | tracepoint(objectstore, truncate_enter, osr_name, off); | |
2704 | if (_check_replay_guard(cid, oid, spos) > 0) | |
2705 | r = _truncate(cid, oid, off); | |
2706 | tracepoint(objectstore, truncate_exit, r); | |
2707 | } | |
2708 | break; | |
2709 | ||
2710 | case Transaction::OP_REMOVE: | |
2711 | { | |
2712 | const coll_t &_cid = i.get_cid(op->cid); | |
2713 | const ghobject_t &oid = i.get_oid(op->oid); | |
2714 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2715 | _cid : _cid.get_temp(); | |
2716 | tracepoint(objectstore, remove_enter, osr_name); | |
2717 | if (_check_replay_guard(cid, oid, spos) > 0) | |
2718 | r = _remove(cid, oid, spos); | |
2719 | tracepoint(objectstore, remove_exit, r); | |
2720 | } | |
2721 | break; | |
2722 | ||
2723 | case Transaction::OP_SETATTR: | |
2724 | { | |
2725 | const coll_t &_cid = i.get_cid(op->cid); | |
2726 | const ghobject_t &oid = i.get_oid(op->oid); | |
2727 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2728 | _cid : _cid.get_temp(); | |
2729 | string name = i.decode_string(); | |
2730 | bufferlist bl; | |
2731 | i.decode_bl(bl); | |
2732 | tracepoint(objectstore, setattr_enter, osr_name); | |
2733 | if (_check_replay_guard(cid, oid, spos) > 0) { | |
2734 | map<string, bufferptr> to_set; | |
2735 | to_set[name] = bufferptr(bl.c_str(), bl.length()); | |
2736 | r = _setattrs(cid, oid, to_set, spos); | |
2737 | if (r == -ENOSPC) | |
2738 | dout(0) << " ENOSPC on setxattr on " << cid << "/" << oid | |
2739 | << " name " << name << " size " << bl.length() << dendl; | |
2740 | } | |
2741 | tracepoint(objectstore, setattr_exit, r); | |
2742 | } | |
2743 | break; | |
2744 | ||
2745 | case Transaction::OP_SETATTRS: | |
2746 | { | |
2747 | const coll_t &_cid = i.get_cid(op->cid); | |
2748 | const ghobject_t &oid = i.get_oid(op->oid); | |
2749 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2750 | _cid : _cid.get_temp(); | |
2751 | map<string, bufferptr> aset; | |
2752 | i.decode_attrset(aset); | |
2753 | tracepoint(objectstore, setattrs_enter, osr_name); | |
2754 | if (_check_replay_guard(cid, oid, spos) > 0) | |
2755 | r = _setattrs(cid, oid, aset, spos); | |
2756 | tracepoint(objectstore, setattrs_exit, r); | |
2757 | if (r == -ENOSPC) | |
2758 | dout(0) << " ENOSPC on setxattrs on " << cid << "/" << oid << dendl; | |
2759 | } | |
2760 | break; | |
2761 | ||
2762 | case Transaction::OP_RMATTR: | |
2763 | { | |
2764 | const coll_t &_cid = i.get_cid(op->cid); | |
2765 | const ghobject_t &oid = i.get_oid(op->oid); | |
2766 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2767 | _cid : _cid.get_temp(); | |
2768 | string name = i.decode_string(); | |
2769 | tracepoint(objectstore, rmattr_enter, osr_name); | |
2770 | if (_check_replay_guard(cid, oid, spos) > 0) | |
2771 | r = _rmattr(cid, oid, name.c_str(), spos); | |
2772 | tracepoint(objectstore, rmattr_exit, r); | |
2773 | } | |
2774 | break; | |
2775 | ||
2776 | case Transaction::OP_RMATTRS: | |
2777 | { | |
2778 | const coll_t &_cid = i.get_cid(op->cid); | |
2779 | const ghobject_t &oid = i.get_oid(op->oid); | |
2780 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2781 | _cid : _cid.get_temp(); | |
2782 | tracepoint(objectstore, rmattrs_enter, osr_name); | |
2783 | if (_check_replay_guard(cid, oid, spos) > 0) | |
2784 | r = _rmattrs(cid, oid, spos); | |
2785 | tracepoint(objectstore, rmattrs_exit, r); | |
2786 | } | |
2787 | break; | |
2788 | ||
2789 | case Transaction::OP_CLONE: | |
2790 | { | |
2791 | const coll_t &_cid = i.get_cid(op->cid); | |
2792 | const ghobject_t &oid = i.get_oid(op->oid); | |
2793 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2794 | _cid : _cid.get_temp(); | |
2795 | const ghobject_t &noid = i.get_oid(op->dest_oid); | |
2796 | tracepoint(objectstore, clone_enter, osr_name); | |
2797 | r = _clone(cid, oid, noid, spos); | |
2798 | tracepoint(objectstore, clone_exit, r); | |
2799 | } | |
2800 | break; | |
2801 | ||
2802 | case Transaction::OP_CLONERANGE: | |
2803 | { | |
2804 | const coll_t &_cid = i.get_cid(op->cid); | |
2805 | const ghobject_t &oid = i.get_oid(op->oid); | |
2806 | const ghobject_t &noid = i.get_oid(op->dest_oid); | |
2807 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2808 | _cid : _cid.get_temp(); | |
2809 | const coll_t &ncid = !_need_temp_object_collection(_cid, noid) ? | |
2810 | _cid : _cid.get_temp(); | |
2811 | uint64_t off = op->off; | |
2812 | uint64_t len = op->len; | |
2813 | tracepoint(objectstore, clone_range_enter, osr_name, len); | |
2814 | r = _clone_range(cid, oid, ncid, noid, off, len, off, spos); | |
2815 | tracepoint(objectstore, clone_range_exit, r); | |
2816 | } | |
2817 | break; | |
2818 | ||
2819 | case Transaction::OP_CLONERANGE2: | |
2820 | { | |
2821 | const coll_t &_cid = i.get_cid(op->cid); | |
2822 | const ghobject_t &oid = i.get_oid(op->oid); | |
2823 | const ghobject_t &noid = i.get_oid(op->dest_oid); | |
2824 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2825 | _cid : _cid.get_temp(); | |
2826 | const coll_t &ncid = !_need_temp_object_collection(_cid, noid) ? | |
2827 | _cid : _cid.get_temp(); | |
2828 | uint64_t srcoff = op->off; | |
2829 | uint64_t len = op->len; | |
2830 | uint64_t dstoff = op->dest_off; | |
2831 | tracepoint(objectstore, clone_range2_enter, osr_name, len); | |
2832 | r = _clone_range(cid, oid, ncid, noid, srcoff, len, dstoff, spos); | |
2833 | tracepoint(objectstore, clone_range2_exit, r); | |
2834 | } | |
2835 | break; | |
2836 | ||
2837 | case Transaction::OP_MKCOLL: | |
2838 | { | |
2839 | const coll_t &cid = i.get_cid(op->cid); | |
2840 | tracepoint(objectstore, mkcoll_enter, osr_name); | |
2841 | if (_check_replay_guard(cid, spos) > 0) | |
2842 | r = _create_collection(cid, op->split_bits, spos); | |
2843 | tracepoint(objectstore, mkcoll_exit, r); | |
2844 | } | |
2845 | break; | |
2846 | ||
2847 | case Transaction::OP_COLL_SET_BITS: | |
2848 | { | |
2849 | const coll_t &cid = i.get_cid(op->cid); | |
2850 | int bits = op->split_bits; | |
2851 | r = _collection_set_bits(cid, bits); | |
2852 | } | |
2853 | break; | |
2854 | ||
2855 | case Transaction::OP_COLL_HINT: | |
2856 | { | |
2857 | const coll_t &cid = i.get_cid(op->cid); | |
2858 | uint32_t type = op->hint_type; | |
2859 | bufferlist hint; | |
2860 | i.decode_bl(hint); | |
2861 | bufferlist::iterator hiter = hint.begin(); | |
2862 | if (type == Transaction::COLL_HINT_EXPECTED_NUM_OBJECTS) { | |
2863 | uint32_t pg_num; | |
2864 | uint64_t num_objs; | |
2865 | ::decode(pg_num, hiter); | |
2866 | ::decode(num_objs, hiter); | |
2867 | if (_check_replay_guard(cid, spos) > 0) { | |
2868 | r = _collection_hint_expected_num_objs(cid, pg_num, num_objs, spos); | |
2869 | } | |
2870 | } else { | |
2871 | // Ignore the hint | |
2872 | dout(10) << "Unrecognized collection hint type: " << type << dendl; | |
2873 | } | |
2874 | } | |
2875 | break; | |
2876 | ||
2877 | case Transaction::OP_RMCOLL: | |
2878 | { | |
2879 | const coll_t &cid = i.get_cid(op->cid); | |
2880 | tracepoint(objectstore, rmcoll_enter, osr_name); | |
2881 | if (_check_replay_guard(cid, spos) > 0) | |
2882 | r = _destroy_collection(cid); | |
2883 | tracepoint(objectstore, rmcoll_exit, r); | |
2884 | } | |
2885 | break; | |
2886 | ||
2887 | case Transaction::OP_COLL_ADD: | |
2888 | { | |
2889 | const coll_t &ocid = i.get_cid(op->cid); | |
2890 | const coll_t &ncid = i.get_cid(op->dest_cid); | |
2891 | const ghobject_t &oid = i.get_oid(op->oid); | |
2892 | ||
2893 | assert(oid.hobj.pool >= -1); | |
2894 | ||
2895 | // always followed by OP_COLL_REMOVE | |
2896 | Transaction::Op *op2 = i.decode_op(); | |
2897 | const coll_t &ocid2 = i.get_cid(op2->cid); | |
2898 | const ghobject_t &oid2 = i.get_oid(op2->oid); | |
2899 | assert(op2->op == Transaction::OP_COLL_REMOVE); | |
2900 | assert(ocid2 == ocid); | |
2901 | assert(oid2 == oid); | |
2902 | ||
2903 | tracepoint(objectstore, coll_add_enter); | |
2904 | r = _collection_add(ncid, ocid, oid, spos); | |
2905 | tracepoint(objectstore, coll_add_exit, r); | |
2906 | spos.op++; | |
2907 | if (r < 0) | |
2908 | break; | |
2909 | tracepoint(objectstore, coll_remove_enter, osr_name); | |
2910 | if (_check_replay_guard(ocid, oid, spos) > 0) | |
2911 | r = _remove(ocid, oid, spos); | |
2912 | tracepoint(objectstore, coll_remove_exit, r); | |
2913 | } | |
2914 | break; | |
2915 | ||
2916 | case Transaction::OP_COLL_MOVE: | |
2917 | { | |
2918 | // WARNING: this is deprecated and buggy; only here to replay old journals. | |
2919 | const coll_t &ocid = i.get_cid(op->cid); | |
2920 | const coll_t &ncid = i.get_cid(op->dest_cid); | |
2921 | const ghobject_t &oid = i.get_oid(op->oid); | |
2922 | tracepoint(objectstore, coll_move_enter); | |
2923 | r = _collection_add(ocid, ncid, oid, spos); | |
2924 | if (r == 0 && | |
2925 | (_check_replay_guard(ocid, oid, spos) > 0)) | |
2926 | r = _remove(ocid, oid, spos); | |
2927 | tracepoint(objectstore, coll_move_exit, r); | |
2928 | } | |
2929 | break; | |
2930 | ||
2931 | case Transaction::OP_COLL_MOVE_RENAME: | |
2932 | { | |
2933 | const coll_t &_oldcid = i.get_cid(op->cid); | |
2934 | const ghobject_t &oldoid = i.get_oid(op->oid); | |
2935 | const coll_t &_newcid = i.get_cid(op->dest_cid); | |
2936 | const ghobject_t &newoid = i.get_oid(op->dest_oid); | |
2937 | const coll_t &oldcid = !_need_temp_object_collection(_oldcid, oldoid) ? | |
2938 | _oldcid : _oldcid.get_temp(); | |
2939 | const coll_t &newcid = !_need_temp_object_collection(_newcid, newoid) ? | |
2940 | _oldcid : _newcid.get_temp(); | |
2941 | tracepoint(objectstore, coll_move_rename_enter); | |
2942 | r = _collection_move_rename(oldcid, oldoid, newcid, newoid, spos); | |
2943 | tracepoint(objectstore, coll_move_rename_exit, r); | |
2944 | } | |
2945 | break; | |
2946 | ||
2947 | case Transaction::OP_TRY_RENAME: | |
2948 | { | |
2949 | const coll_t &_cid = i.get_cid(op->cid); | |
2950 | const ghobject_t &oldoid = i.get_oid(op->oid); | |
2951 | const ghobject_t &newoid = i.get_oid(op->dest_oid); | |
2952 | const coll_t &oldcid = !_need_temp_object_collection(_cid, oldoid) ? | |
2953 | _cid : _cid.get_temp(); | |
2954 | const coll_t &newcid = !_need_temp_object_collection(_cid, newoid) ? | |
2955 | _cid : _cid.get_temp(); | |
2956 | tracepoint(objectstore, coll_try_rename_enter); | |
2957 | r = _collection_move_rename(oldcid, oldoid, newcid, newoid, spos, true); | |
2958 | tracepoint(objectstore, coll_try_rename_exit, r); | |
2959 | } | |
2960 | break; | |
2961 | ||
2962 | case Transaction::OP_COLL_SETATTR: | |
2963 | case Transaction::OP_COLL_RMATTR: | |
2964 | assert(0 == "collection attr methods no longer implemented"); | |
2965 | break; | |
2966 | ||
2967 | case Transaction::OP_STARTSYNC: | |
2968 | tracepoint(objectstore, startsync_enter, osr_name); | |
2969 | _start_sync(); | |
2970 | tracepoint(objectstore, startsync_exit); | |
2971 | break; | |
2972 | ||
2973 | case Transaction::OP_COLL_RENAME: | |
2974 | { | |
2975 | r = -EOPNOTSUPP; | |
2976 | } | |
2977 | break; | |
2978 | ||
2979 | case Transaction::OP_OMAP_CLEAR: | |
2980 | { | |
2981 | const coll_t &_cid = i.get_cid(op->cid); | |
2982 | const ghobject_t &oid = i.get_oid(op->oid); | |
2983 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2984 | _cid : _cid.get_temp(); | |
2985 | tracepoint(objectstore, omap_clear_enter, osr_name); | |
28e407b8 AA |
2986 | if (_check_replay_guard(cid, oid, spos) > 0) |
2987 | r = _omap_clear(cid, oid, spos); | |
7c673cae FG |
2988 | tracepoint(objectstore, omap_clear_exit, r); |
2989 | } | |
2990 | break; | |
2991 | case Transaction::OP_OMAP_SETKEYS: | |
2992 | { | |
2993 | const coll_t &_cid = i.get_cid(op->cid); | |
2994 | const ghobject_t &oid = i.get_oid(op->oid); | |
2995 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
2996 | _cid : _cid.get_temp(); | |
2997 | map<string, bufferlist> aset; | |
2998 | i.decode_attrset(aset); | |
2999 | tracepoint(objectstore, omap_setkeys_enter, osr_name); | |
28e407b8 AA |
3000 | if (_check_replay_guard(cid, oid, spos) > 0) |
3001 | r = _omap_setkeys(cid, oid, aset, spos); | |
7c673cae FG |
3002 | tracepoint(objectstore, omap_setkeys_exit, r); |
3003 | } | |
3004 | break; | |
3005 | case Transaction::OP_OMAP_RMKEYS: | |
3006 | { | |
3007 | const coll_t &_cid = i.get_cid(op->cid); | |
3008 | const ghobject_t &oid = i.get_oid(op->oid); | |
3009 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
3010 | _cid : _cid.get_temp(); | |
3011 | set<string> keys; | |
3012 | i.decode_keyset(keys); | |
3013 | tracepoint(objectstore, omap_rmkeys_enter, osr_name); | |
28e407b8 AA |
3014 | if (_check_replay_guard(cid, oid, spos) > 0) |
3015 | r = _omap_rmkeys(cid, oid, keys, spos); | |
7c673cae FG |
3016 | tracepoint(objectstore, omap_rmkeys_exit, r); |
3017 | } | |
3018 | break; | |
3019 | case Transaction::OP_OMAP_RMKEYRANGE: | |
3020 | { | |
3021 | const coll_t &_cid = i.get_cid(op->cid); | |
3022 | const ghobject_t &oid = i.get_oid(op->oid); | |
3023 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
3024 | _cid : _cid.get_temp(); | |
3025 | string first, last; | |
3026 | first = i.decode_string(); | |
3027 | last = i.decode_string(); | |
3028 | tracepoint(objectstore, omap_rmkeyrange_enter, osr_name); | |
28e407b8 AA |
3029 | if (_check_replay_guard(cid, oid, spos) > 0) |
3030 | r = _omap_rmkeyrange(cid, oid, first, last, spos); | |
7c673cae FG |
3031 | tracepoint(objectstore, omap_rmkeyrange_exit, r); |
3032 | } | |
3033 | break; | |
3034 | case Transaction::OP_OMAP_SETHEADER: | |
3035 | { | |
3036 | const coll_t &_cid = i.get_cid(op->cid); | |
3037 | const ghobject_t &oid = i.get_oid(op->oid); | |
3038 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
3039 | _cid : _cid.get_temp(); | |
3040 | bufferlist bl; | |
3041 | i.decode_bl(bl); | |
3042 | tracepoint(objectstore, omap_setheader_enter, osr_name); | |
28e407b8 AA |
3043 | if (_check_replay_guard(cid, oid, spos) > 0) |
3044 | r = _omap_setheader(cid, oid, bl, spos); | |
7c673cae FG |
3045 | tracepoint(objectstore, omap_setheader_exit, r); |
3046 | } | |
3047 | break; | |
3048 | case Transaction::OP_SPLIT_COLLECTION: | |
3049 | { | |
3050 | assert(0 == "not legacy journal; upgrade to firefly first"); | |
3051 | } | |
3052 | break; | |
3053 | case Transaction::OP_SPLIT_COLLECTION2: | |
3054 | { | |
3055 | coll_t cid = i.get_cid(op->cid); | |
3056 | uint32_t bits = op->split_bits; | |
3057 | uint32_t rem = op->split_rem; | |
3058 | coll_t dest = i.get_cid(op->dest_cid); | |
3059 | tracepoint(objectstore, split_coll2_enter, osr_name); | |
3060 | r = _split_collection(cid, bits, rem, dest, spos); | |
3061 | tracepoint(objectstore, split_coll2_exit, r); | |
3062 | } | |
3063 | break; | |
3064 | ||
3065 | case Transaction::OP_SETALLOCHINT: | |
3066 | { | |
3067 | const coll_t &_cid = i.get_cid(op->cid); | |
3068 | const ghobject_t &oid = i.get_oid(op->oid); | |
3069 | const coll_t &cid = !_need_temp_object_collection(_cid, oid) ? | |
3070 | _cid : _cid.get_temp(); | |
3071 | uint64_t expected_object_size = op->expected_object_size; | |
3072 | uint64_t expected_write_size = op->expected_write_size; | |
3073 | tracepoint(objectstore, setallochint_enter, osr_name); | |
3074 | if (_check_replay_guard(cid, oid, spos) > 0) | |
3075 | r = _set_alloc_hint(cid, oid, expected_object_size, | |
3076 | expected_write_size); | |
3077 | tracepoint(objectstore, setallochint_exit, r); | |
3078 | } | |
3079 | break; | |
3080 | ||
3081 | default: | |
3082 | derr << "bad op " << op->op << dendl; | |
3083 | ceph_abort(); | |
3084 | } | |
3085 | ||
3086 | if (r < 0) { | |
3087 | bool ok = false; | |
3088 | ||
3089 | if (r == -ENOENT && !(op->op == Transaction::OP_CLONERANGE || | |
3090 | op->op == Transaction::OP_CLONE || | |
3091 | op->op == Transaction::OP_CLONERANGE2 || | |
3092 | op->op == Transaction::OP_COLL_ADD || | |
3093 | op->op == Transaction::OP_SETATTR || | |
3094 | op->op == Transaction::OP_SETATTRS || | |
3095 | op->op == Transaction::OP_RMATTR || | |
3096 | op->op == Transaction::OP_OMAP_SETKEYS || | |
3097 | op->op == Transaction::OP_OMAP_RMKEYS || | |
3098 | op->op == Transaction::OP_OMAP_RMKEYRANGE || | |
3099 | op->op == Transaction::OP_OMAP_SETHEADER)) | |
3100 | // -ENOENT is normally okay | |
3101 | // ...including on a replayed OP_RMCOLL with checkpoint mode | |
3102 | ok = true; | |
3103 | if (r == -ENODATA) | |
3104 | ok = true; | |
3105 | ||
3106 | if (op->op == Transaction::OP_SETALLOCHINT) | |
3107 | // Either EOPNOTSUPP or EINVAL most probably. EINVAL in most | |
3108 | // cases means invalid hint size (e.g. too big, not a multiple | |
3109 | // of block size, etc) or, at least on xfs, an attempt to set | |
3110 | // or change it when the file is not empty. However, | |
3111 | // OP_SETALLOCHINT is advisory, so ignore all errors. | |
3112 | ok = true; | |
3113 | ||
3114 | if (replaying && !backend->can_checkpoint()) { | |
3115 | if (r == -EEXIST && op->op == Transaction::OP_MKCOLL) { | |
3116 | dout(10) << "tolerating EEXIST during journal replay since checkpoint is not enabled" << dendl; | |
3117 | ok = true; | |
3118 | } | |
3119 | if (r == -EEXIST && op->op == Transaction::OP_COLL_ADD) { | |
3120 | dout(10) << "tolerating EEXIST during journal replay since checkpoint is not enabled" << dendl; | |
3121 | ok = true; | |
3122 | } | |
3123 | if (r == -EEXIST && op->op == Transaction::OP_COLL_MOVE) { | |
3124 | dout(10) << "tolerating EEXIST during journal replay since checkpoint is not enabled" << dendl; | |
3125 | ok = true; | |
3126 | } | |
3127 | if (r == -ERANGE) { | |
3128 | dout(10) << "tolerating ERANGE on replay" << dendl; | |
3129 | ok = true; | |
3130 | } | |
3131 | if (r == -ENOENT) { | |
3132 | dout(10) << "tolerating ENOENT on replay" << dendl; | |
3133 | ok = true; | |
3134 | } | |
3135 | } | |
3136 | ||
3137 | if (!ok) { | |
3138 | const char *msg = "unexpected error code"; | |
3139 | ||
3140 | if (r == -ENOENT && (op->op == Transaction::OP_CLONERANGE || | |
3141 | op->op == Transaction::OP_CLONE || | |
3142 | op->op == Transaction::OP_CLONERANGE2)) { | |
3143 | msg = "ENOENT on clone suggests osd bug"; | |
3144 | } else if (r == -ENOSPC) { | |
3145 | // For now, if we hit _any_ ENOSPC, crash, before we do any damage | |
3146 | // by partially applying transactions. | |
3147 | msg = "ENOSPC from disk filesystem, misconfigured cluster"; | |
3148 | } else if (r == -ENOTEMPTY) { | |
3149 | msg = "ENOTEMPTY suggests garbage data in osd data dir"; | |
3150 | } else if (r == -EPERM) { | |
3151 | msg = "EPERM suggests file(s) in osd data dir not owned by ceph user, or leveldb corruption"; | |
3152 | } | |
3153 | ||
3154 | derr << " error " << cpp_strerror(r) << " not handled on operation " << op | |
3155 | << " (" << spos << ", or op " << spos.op << ", counting from 0)" << dendl; | |
3156 | dout(0) << msg << dendl; | |
3157 | dout(0) << " transaction dump:\n"; | |
3158 | JSONFormatter f(true); | |
3159 | f.open_object_section("transaction"); | |
3160 | t.dump(&f); | |
3161 | f.close_section(); | |
3162 | f.flush(*_dout); | |
3163 | *_dout << dendl; | |
3164 | ||
3165 | if (r == -EMFILE) { | |
3166 | dump_open_fds(cct); | |
3167 | } | |
3168 | ||
3169 | assert(0 == "unexpected error"); | |
3170 | } | |
3171 | } | |
3172 | ||
3173 | spos.op++; | |
3174 | } | |
3175 | ||
3176 | _inject_failure(); | |
3177 | } | |
3178 | ||
3179 | /*********************************************/ | |
3180 | ||
3181 | ||
3182 | ||
3183 | // -------------------- | |
3184 | // objects | |
3185 | ||
3186 | bool FileStore::exists(const coll_t& _cid, const ghobject_t& oid) | |
3187 | { | |
3188 | tracepoint(objectstore, exists_enter, _cid.c_str()); | |
3189 | const coll_t& cid = !_need_temp_object_collection(_cid, oid) ? _cid : _cid.get_temp(); | |
3190 | struct stat st; | |
3191 | bool retval = stat(cid, oid, &st) == 0; | |
3192 | tracepoint(objectstore, exists_exit, retval); | |
3193 | return retval; | |
3194 | } | |
3195 | ||
3196 | int FileStore::stat( | |
3197 | const coll_t& _cid, const ghobject_t& oid, struct stat *st, bool allow_eio) | |
3198 | { | |
3199 | tracepoint(objectstore, stat_enter, _cid.c_str()); | |
3200 | const coll_t& cid = !_need_temp_object_collection(_cid, oid) ? _cid : _cid.get_temp(); | |
3201 | int r = lfn_stat(cid, oid, st); | |
3202 | assert(allow_eio || !m_filestore_fail_eio || r != -EIO); | |
3203 | if (r < 0) { | |
31f18b77 | 3204 | dout(10) << __FUNC__ << ": " << cid << "/" << oid |
7c673cae FG |
3205 | << " = " << r << dendl; |
3206 | } else { | |
31f18b77 | 3207 | dout(10) << __FUNC__ << ": " << cid << "/" << oid |
7c673cae FG |
3208 | << " = " << r |
3209 | << " (size " << st->st_size << ")" << dendl; | |
3210 | } | |
3211 | if (cct->_conf->filestore_debug_inject_read_err && | |
3212 | debug_mdata_eio(oid)) { | |
3213 | return -EIO; | |
3214 | } else { | |
3215 | tracepoint(objectstore, stat_exit, r); | |
3216 | return r; | |
3217 | } | |
3218 | } | |
3219 | ||
3220 | int FileStore::set_collection_opts( | |
3221 | const coll_t& cid, | |
3222 | const pool_opts_t& opts) | |
3223 | { | |
3224 | return -EOPNOTSUPP; | |
3225 | } | |
3226 | ||
3227 | int FileStore::read( | |
3228 | const coll_t& _cid, | |
3229 | const ghobject_t& oid, | |
3230 | uint64_t offset, | |
3231 | size_t len, | |
3232 | bufferlist& bl, | |
224ce89b | 3233 | uint32_t op_flags) |
7c673cae FG |
3234 | { |
3235 | int got; | |
3236 | tracepoint(objectstore, read_enter, _cid.c_str(), offset, len); | |
3237 | const coll_t& cid = !_need_temp_object_collection(_cid, oid) ? _cid : _cid.get_temp(); | |
3238 | ||
31f18b77 | 3239 | dout(15) << __FUNC__ << ": " << cid << "/" << oid << " " << offset << "~" << len << dendl; |
7c673cae FG |
3240 | |
3241 | FDRef fd; | |
3242 | int r = lfn_open(cid, oid, false, &fd); | |
3243 | if (r < 0) { | |
31f18b77 | 3244 | dout(10) << __FUNC__ << ": (" << cid << "/" << oid << ") open error: " |
7c673cae FG |
3245 | << cpp_strerror(r) << dendl; |
3246 | return r; | |
3247 | } | |
3248 | ||
3249 | if (offset == 0 && len == 0) { | |
3250 | struct stat st; | |
3251 | memset(&st, 0, sizeof(struct stat)); | |
3252 | int r = ::fstat(**fd, &st); | |
3253 | assert(r == 0); | |
3254 | len = st.st_size; | |
3255 | } | |
3256 | ||
3257 | #ifdef HAVE_POSIX_FADVISE | |
3258 | if (op_flags & CEPH_OSD_OP_FLAG_FADVISE_RANDOM) | |
3259 | posix_fadvise(**fd, offset, len, POSIX_FADV_RANDOM); | |
3260 | if (op_flags & CEPH_OSD_OP_FLAG_FADVISE_SEQUENTIAL) | |
3261 | posix_fadvise(**fd, offset, len, POSIX_FADV_SEQUENTIAL); | |
3262 | #endif | |
3263 | ||
3264 | bufferptr bptr(len); // prealloc space for entire read | |
3265 | got = safe_pread(**fd, bptr.c_str(), len, offset); | |
3266 | if (got < 0) { | |
31f18b77 | 3267 | dout(10) << __FUNC__ << ": (" << cid << "/" << oid << ") pread error: " << cpp_strerror(got) << dendl; |
7c673cae | 3268 | lfn_close(fd); |
7c673cae FG |
3269 | return got; |
3270 | } | |
3271 | bptr.set_length(got); // properly size the buffer | |
3272 | bl.clear(); | |
3273 | bl.push_back(std::move(bptr)); // put it in the target bufferlist | |
3274 | ||
3275 | #ifdef HAVE_POSIX_FADVISE | |
3276 | if (op_flags & CEPH_OSD_OP_FLAG_FADVISE_DONTNEED) | |
3277 | posix_fadvise(**fd, offset, len, POSIX_FADV_DONTNEED); | |
3278 | if (op_flags & (CEPH_OSD_OP_FLAG_FADVISE_RANDOM | CEPH_OSD_OP_FLAG_FADVISE_SEQUENTIAL)) | |
3279 | posix_fadvise(**fd, offset, len, POSIX_FADV_NORMAL); | |
3280 | #endif | |
3281 | ||
3282 | if (m_filestore_sloppy_crc && (!replaying || backend->can_checkpoint())) { | |
3283 | ostringstream ss; | |
3284 | int errors = backend->_crc_verify_read(**fd, offset, got, bl, &ss); | |
3285 | if (errors != 0) { | |
31f18b77 | 3286 | dout(0) << __FUNC__ << ": " << cid << "/" << oid << " " << offset << "~" |
7c673cae FG |
3287 | << got << " ... BAD CRC:\n" << ss.str() << dendl; |
3288 | assert(0 == "bad crc on read"); | |
3289 | } | |
3290 | } | |
3291 | ||
3292 | lfn_close(fd); | |
3293 | ||
31f18b77 | 3294 | dout(10) << __FUNC__ << ": " << cid << "/" << oid << " " << offset << "~" |
7c673cae FG |
3295 | << got << "/" << len << dendl; |
3296 | if (cct->_conf->filestore_debug_inject_read_err && | |
3297 | debug_data_eio(oid)) { | |
3298 | return -EIO; | |
224ce89b WB |
3299 | } else if (cct->_conf->filestore_debug_random_read_err && |
3300 | (rand() % (int)(cct->_conf->filestore_debug_random_read_err * 100.0)) == 0) { | |
3301 | dout(0) << __func__ << ": inject random EIO" << dendl; | |
3302 | return -EIO; | |
7c673cae FG |
3303 | } else { |
3304 | tracepoint(objectstore, read_exit, got); | |
3305 | return got; | |
3306 | } | |
3307 | } | |
3308 | ||
3309 | int FileStore::_do_fiemap(int fd, uint64_t offset, size_t len, | |
3310 | map<uint64_t, uint64_t> *m) | |
3311 | { | |
3312 | uint64_t i; | |
3313 | struct fiemap_extent *extent = NULL; | |
3314 | struct fiemap *fiemap = NULL; | |
3315 | int r = 0; | |
3316 | ||
3317 | more: | |
3318 | r = backend->do_fiemap(fd, offset, len, &fiemap); | |
3319 | if (r < 0) | |
3320 | return r; | |
3321 | ||
3322 | if (fiemap->fm_mapped_extents == 0) { | |
3323 | free(fiemap); | |
3324 | return r; | |
3325 | } | |
3326 | ||
3327 | extent = &fiemap->fm_extents[0]; | |
3328 | ||
3329 | /* start where we were asked to start */ | |
3330 | if (extent->fe_logical < offset) { | |
3331 | extent->fe_length -= offset - extent->fe_logical; | |
3332 | extent->fe_logical = offset; | |
3333 | } | |
3334 | ||
3335 | i = 0; | |
3336 | ||
3337 | struct fiemap_extent *last = nullptr; | |
3338 | while (i < fiemap->fm_mapped_extents) { | |
3339 | struct fiemap_extent *next = extent + 1; | |
3340 | ||
31f18b77 | 3341 | dout(10) << __FUNC__ << ": fm_mapped_extents=" << fiemap->fm_mapped_extents |
7c673cae FG |
3342 | << " fe_logical=" << extent->fe_logical << " fe_length=" << extent->fe_length << dendl; |
3343 | ||
3344 | /* try to merge extents */ | |
3345 | while ((i < fiemap->fm_mapped_extents - 1) && | |
3346 | (extent->fe_logical + extent->fe_length == next->fe_logical)) { | |
3347 | next->fe_length += extent->fe_length; | |
3348 | next->fe_logical = extent->fe_logical; | |
3349 | extent = next; | |
3350 | next = extent + 1; | |
3351 | i++; | |
3352 | } | |
3353 | ||
3354 | if (extent->fe_logical + extent->fe_length > offset + len) | |
3355 | extent->fe_length = offset + len - extent->fe_logical; | |
3356 | (*m)[extent->fe_logical] = extent->fe_length; | |
3357 | i++; | |
3358 | last = extent++; | |
3359 | } | |
3360 | uint64_t xoffset = last->fe_logical + last->fe_length - offset; | |
3361 | offset = last->fe_logical + last->fe_length; | |
3362 | len -= xoffset; | |
3363 | const bool is_last = (last->fe_flags & FIEMAP_EXTENT_LAST) || (len == 0); | |
3364 | free(fiemap); | |
3365 | if (!is_last) { | |
3366 | goto more; | |
3367 | } | |
3368 | ||
3369 | return r; | |
3370 | } | |
3371 | ||
3372 | int FileStore::_do_seek_hole_data(int fd, uint64_t offset, size_t len, | |
3373 | map<uint64_t, uint64_t> *m) | |
3374 | { | |
3375 | #if defined(__linux__) && defined(SEEK_HOLE) && defined(SEEK_DATA) | |
3376 | off_t hole_pos, data_pos; | |
3377 | int r = 0; | |
3378 | ||
3379 | // If lseek fails with errno setting to be ENXIO, this means the current | |
3380 | // file offset is beyond the end of the file. | |
3381 | off_t start = offset; | |
3382 | while(start < (off_t)(offset + len)) { | |
3383 | data_pos = lseek(fd, start, SEEK_DATA); | |
3384 | if (data_pos < 0) { | |
3385 | if (errno == ENXIO) | |
3386 | break; | |
3387 | else { | |
3388 | r = -errno; | |
3389 | dout(10) << "failed to lseek: " << cpp_strerror(r) << dendl; | |
3390 | return r; | |
3391 | } | |
3392 | } else if (data_pos > (off_t)(offset + len)) { | |
3393 | break; | |
3394 | } | |
3395 | ||
3396 | hole_pos = lseek(fd, data_pos, SEEK_HOLE); | |
3397 | if (hole_pos < 0) { | |
3398 | if (errno == ENXIO) { | |
3399 | break; | |
3400 | } else { | |
3401 | r = -errno; | |
3402 | dout(10) << "failed to lseek: " << cpp_strerror(r) << dendl; | |
3403 | return r; | |
3404 | } | |
3405 | } | |
3406 | ||
3407 | if (hole_pos >= (off_t)(offset + len)) { | |
3408 | (*m)[data_pos] = offset + len - data_pos; | |
3409 | break; | |
3410 | } | |
3411 | (*m)[data_pos] = hole_pos - data_pos; | |
3412 | start = hole_pos; | |
3413 | } | |
3414 | ||
3415 | return r; | |
3416 | #else | |
3417 | (*m)[offset] = len; | |
3418 | return 0; | |
3419 | #endif | |
3420 | } | |
3421 | ||
3422 | int FileStore::fiemap(const coll_t& _cid, const ghobject_t& oid, | |
3423 | uint64_t offset, size_t len, | |
3424 | bufferlist& bl) | |
3425 | { | |
3426 | map<uint64_t, uint64_t> exomap; | |
3427 | int r = fiemap(_cid, oid, offset, len, exomap); | |
3428 | if (r >= 0) { | |
3429 | ::encode(exomap, bl); | |
3430 | } | |
3431 | return r; | |
3432 | } | |
3433 | ||
3434 | int FileStore::fiemap(const coll_t& _cid, const ghobject_t& oid, | |
3435 | uint64_t offset, size_t len, | |
3436 | map<uint64_t, uint64_t>& destmap) | |
3437 | { | |
3438 | tracepoint(objectstore, fiemap_enter, _cid.c_str(), offset, len); | |
3439 | const coll_t& cid = !_need_temp_object_collection(_cid, oid) ? _cid : _cid.get_temp(); | |
3440 | destmap.clear(); | |
3441 | ||
3442 | if ((!backend->has_seek_data_hole() && !backend->has_fiemap()) || | |
3443 | len <= (size_t)m_filestore_fiemap_threshold) { | |
3444 | destmap[offset] = len; | |
3445 | return 0; | |
3446 | } | |
3447 | ||
31f18b77 | 3448 | dout(15) << __FUNC__ << ": " << cid << "/" << oid << " " << offset << "~" << len << dendl; |
7c673cae FG |
3449 | |
3450 | FDRef fd; | |
3451 | ||
3452 | int r = lfn_open(cid, oid, false, &fd); | |
3453 | if (r < 0) { | |
3454 | dout(10) << "read couldn't open " << cid << "/" << oid << ": " << cpp_strerror(r) << dendl; | |
3455 | goto done; | |
3456 | } | |
3457 | ||
3458 | if (backend->has_seek_data_hole()) { | |
3459 | dout(15) << "seek_data/seek_hole " << cid << "/" << oid << " " << offset << "~" << len << dendl; | |
3460 | r = _do_seek_hole_data(**fd, offset, len, &destmap); | |
3461 | } else if (backend->has_fiemap()) { | |
3462 | dout(15) << "fiemap ioctl" << cid << "/" << oid << " " << offset << "~" << len << dendl; | |
3463 | r = _do_fiemap(**fd, offset, len, &destmap); | |
3464 | } | |
3465 | ||
3466 | lfn_close(fd); | |
3467 | ||
3468 | done: | |
3469 | ||
31f18b77 | 3470 | dout(10) << __FUNC__ << ": " << cid << "/" << oid << " " << offset << "~" << len << " = " << r << " num_extents=" << destmap.size() << " " << destmap << dendl; |
7c673cae FG |
3471 | assert(!m_filestore_fail_eio || r != -EIO); |
3472 | tracepoint(objectstore, fiemap_exit, r); | |
3473 | return r; | |
3474 | } | |
3475 | ||
3476 | int FileStore::_remove(const coll_t& cid, const ghobject_t& oid, | |
3477 | const SequencerPosition &spos) | |
3478 | { | |
31f18b77 | 3479 | dout(15) << __FUNC__ << ": " << cid << "/" << oid << dendl; |
7c673cae | 3480 | int r = lfn_unlink(cid, oid, spos); |
31f18b77 | 3481 | dout(10) << __FUNC__ << ": " << cid << "/" << oid << " = " << r << dendl; |
7c673cae FG |
3482 | return r; |
3483 | } | |
3484 | ||
3485 | int FileStore::_truncate(const coll_t& cid, const ghobject_t& oid, uint64_t size) | |
3486 | { | |
31f18b77 | 3487 | dout(15) << __FUNC__ << ": " << cid << "/" << oid << " size " << size << dendl; |
7c673cae | 3488 | int r = lfn_truncate(cid, oid, size); |
31f18b77 | 3489 | dout(10) << __FUNC__ << ": " << cid << "/" << oid << " size " << size << " = " << r << dendl; |
7c673cae FG |
3490 | return r; |
3491 | } | |
3492 | ||
3493 | ||
3494 | int FileStore::_touch(const coll_t& cid, const ghobject_t& oid) | |
3495 | { | |
31f18b77 | 3496 | dout(15) << __FUNC__ << ": " << cid << "/" << oid << dendl; |
7c673cae FG |
3497 | |
3498 | FDRef fd; | |
3499 | int r = lfn_open(cid, oid, true, &fd); | |
3500 | if (r < 0) { | |
3501 | return r; | |
3502 | } else { | |
3503 | lfn_close(fd); | |
3504 | } | |
31f18b77 | 3505 | dout(10) << __FUNC__ << ": " << cid << "/" << oid << " = " << r << dendl; |
7c673cae FG |
3506 | return r; |
3507 | } | |
3508 | ||
3509 | int FileStore::_write(const coll_t& cid, const ghobject_t& oid, | |
3510 | uint64_t offset, size_t len, | |
3511 | const bufferlist& bl, uint32_t fadvise_flags) | |
3512 | { | |
31f18b77 | 3513 | dout(15) << __FUNC__ << ": " << cid << "/" << oid << " " << offset << "~" << len << dendl; |
7c673cae FG |
3514 | int r; |
3515 | ||
3516 | FDRef fd; | |
3517 | r = lfn_open(cid, oid, true, &fd); | |
3518 | if (r < 0) { | |
31f18b77 | 3519 | dout(0) << __FUNC__ << ": couldn't open " << cid << "/" |
7c673cae FG |
3520 | << oid << ": " |
3521 | << cpp_strerror(r) << dendl; | |
3522 | goto out; | |
3523 | } | |
3524 | ||
3525 | // write | |
3526 | r = bl.write_fd(**fd, offset); | |
3527 | if (r < 0) { | |
31f18b77 | 3528 | derr << __FUNC__ << ": write_fd on " << cid << "/" << oid |
7c673cae FG |
3529 | << " error: " << cpp_strerror(r) << dendl; |
3530 | lfn_close(fd); | |
3531 | goto out; | |
3532 | } | |
3533 | r = bl.length(); | |
3534 | ||
3535 | if (r >= 0 && m_filestore_sloppy_crc) { | |
3536 | int rc = backend->_crc_update_write(**fd, offset, len, bl); | |
3537 | assert(rc >= 0); | |
3538 | } | |
3539 | ||
3540 | if (replaying || m_disable_wbthrottle) { | |
3541 | if (fadvise_flags & CEPH_OSD_OP_FLAG_FADVISE_DONTNEED) { | |
3542 | #ifdef HAVE_POSIX_FADVISE | |
3543 | posix_fadvise(**fd, 0, 0, POSIX_FADV_DONTNEED); | |
3544 | #endif | |
3545 | } | |
3546 | } else { | |
3547 | wbthrottle.queue_wb(fd, oid, offset, len, | |
3548 | fadvise_flags & CEPH_OSD_OP_FLAG_FADVISE_DONTNEED); | |
3549 | } | |
3550 | ||
3551 | lfn_close(fd); | |
3552 | ||
3553 | out: | |
31f18b77 | 3554 | dout(10) << __FUNC__ << ": " << cid << "/" << oid << " " << offset << "~" << len << " = " << r << dendl; |
7c673cae FG |
3555 | return r; |
3556 | } | |
3557 | ||
3558 | int FileStore::_zero(const coll_t& cid, const ghobject_t& oid, uint64_t offset, size_t len) | |
3559 | { | |
31f18b77 | 3560 | dout(15) << __FUNC__ << ": " << cid << "/" << oid << " " << offset << "~" << len << dendl; |
7c673cae FG |
3561 | int ret = 0; |
3562 | ||
3563 | if (cct->_conf->filestore_punch_hole) { | |
3564 | #ifdef CEPH_HAVE_FALLOCATE | |
3565 | # if !defined(DARWIN) && !defined(__FreeBSD__) | |
3566 | # ifdef FALLOC_FL_KEEP_SIZE | |
3567 | // first try to punch a hole. | |
3568 | FDRef fd; | |
3569 | ret = lfn_open(cid, oid, false, &fd); | |
3570 | if (ret < 0) { | |
3571 | goto out; | |
3572 | } | |
3573 | ||
3574 | struct stat st; | |
3575 | ret = ::fstat(**fd, &st); | |
3576 | if (ret < 0) { | |
3577 | ret = -errno; | |
3578 | lfn_close(fd); | |
3579 | goto out; | |
3580 | } | |
3581 | ||
3582 | // first try fallocate | |
3583 | ret = fallocate(**fd, FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE, | |
3584 | offset, len); | |
3585 | if (ret < 0) { | |
3586 | ret = -errno; | |
3587 | } else { | |
b32b8144 FG |
3588 | // ensure we extend file size, if needed |
3589 | if (len > 0 && offset + len > (uint64_t)st.st_size) { | |
7c673cae FG |
3590 | ret = ::ftruncate(**fd, offset + len); |
3591 | if (ret < 0) { | |
3592 | ret = -errno; | |
3593 | lfn_close(fd); | |
3594 | goto out; | |
3595 | } | |
3596 | } | |
3597 | } | |
3598 | lfn_close(fd); | |
3599 | ||
3600 | if (ret >= 0 && m_filestore_sloppy_crc) { | |
3601 | int rc = backend->_crc_update_zero(**fd, offset, len); | |
3602 | assert(rc >= 0); | |
3603 | } | |
3604 | ||
3605 | if (ret == 0) | |
3606 | goto out; // yay! | |
3607 | if (ret != -EOPNOTSUPP) | |
3608 | goto out; // some other error | |
3609 | # endif | |
3610 | # endif | |
3611 | #endif | |
3612 | } | |
3613 | ||
3614 | // lame, kernel is old and doesn't support it. | |
3615 | // write zeros.. yuck! | |
31f18b77 | 3616 | dout(20) << __FUNC__ << ": falling back to writing zeros" << dendl; |
7c673cae FG |
3617 | { |
3618 | bufferlist bl; | |
3619 | bl.append_zero(len); | |
3620 | ret = _write(cid, oid, offset, len, bl); | |
3621 | } | |
3622 | ||
3623 | #ifdef CEPH_HAVE_FALLOCATE | |
3624 | # if !defined(DARWIN) && !defined(__FreeBSD__) | |
3625 | # ifdef FALLOC_FL_KEEP_SIZE | |
3626 | out: | |
3627 | # endif | |
3628 | # endif | |
3629 | #endif | |
31f18b77 | 3630 | dout(20) << __FUNC__ << ": " << cid << "/" << oid << " " << offset << "~" << len << " = " << ret << dendl; |
7c673cae FG |
3631 | return ret; |
3632 | } | |
3633 | ||
3634 | int FileStore::_clone(const coll_t& cid, const ghobject_t& oldoid, const ghobject_t& newoid, | |
3635 | const SequencerPosition& spos) | |
3636 | { | |
31f18b77 | 3637 | dout(15) << __FUNC__ << ": " << cid << "/" << oldoid << " -> " << cid << "/" << newoid << dendl; |
7c673cae FG |
3638 | |
3639 | if (_check_replay_guard(cid, newoid, spos) < 0) | |
3640 | return 0; | |
3641 | ||
3642 | int r; | |
3643 | FDRef o, n; | |
3644 | { | |
3645 | Index index; | |
3646 | r = lfn_open(cid, oldoid, false, &o, &index); | |
3647 | if (r < 0) { | |
3648 | goto out2; | |
3649 | } | |
3650 | assert(NULL != (index.index)); | |
3651 | RWLock::WLocker l((index.index)->access_lock); | |
3652 | ||
3653 | r = lfn_open(cid, newoid, true, &n, &index); | |
3654 | if (r < 0) { | |
3655 | goto out; | |
3656 | } | |
3657 | r = ::ftruncate(**n, 0); | |
3658 | if (r < 0) { | |
3659 | r = -errno; | |
3660 | goto out3; | |
3661 | } | |
3662 | struct stat st; | |
3663 | r = ::fstat(**o, &st); | |
3664 | if (r < 0) { | |
3665 | r = -errno; | |
3666 | goto out3; | |
3667 | } | |
3668 | ||
3669 | r = _do_clone_range(**o, **n, 0, st.st_size, 0); | |
3670 | if (r < 0) { | |
3671 | goto out3; | |
3672 | } | |
3673 | ||
3674 | dout(20) << "objectmap clone" << dendl; | |
3675 | r = object_map->clone(oldoid, newoid, &spos); | |
3676 | if (r < 0 && r != -ENOENT) | |
3677 | goto out3; | |
3678 | } | |
3679 | ||
3680 | { | |
3681 | char buf[2]; | |
3682 | map<string, bufferptr> aset; | |
3683 | r = _fgetattrs(**o, aset); | |
3684 | if (r < 0) | |
3685 | goto out3; | |
3686 | ||
3687 | r = chain_fgetxattr(**o, XATTR_SPILL_OUT_NAME, buf, sizeof(buf)); | |
3688 | if (r >= 0 && !strncmp(buf, XATTR_NO_SPILL_OUT, sizeof(XATTR_NO_SPILL_OUT))) { | |
3689 | r = chain_fsetxattr<true, true>(**n, XATTR_SPILL_OUT_NAME, XATTR_NO_SPILL_OUT, | |
3690 | sizeof(XATTR_NO_SPILL_OUT)); | |
3691 | } else { | |
3692 | r = chain_fsetxattr<true, true>(**n, XATTR_SPILL_OUT_NAME, XATTR_SPILL_OUT, | |
3693 | sizeof(XATTR_SPILL_OUT)); | |
3694 | } | |
3695 | if (r < 0) | |
3696 | goto out3; | |
3697 | ||
3698 | r = _fsetattrs(**n, aset); | |
3699 | if (r < 0) | |
3700 | goto out3; | |
3701 | } | |
3702 | ||
3703 | // clone is non-idempotent; record our work. | |
3704 | _set_replay_guard(**n, spos, &newoid); | |
3705 | ||
3706 | out3: | |
3707 | lfn_close(n); | |
3708 | out: | |
3709 | lfn_close(o); | |
3710 | out2: | |
31f18b77 | 3711 | dout(10) << __FUNC__ << ": " << cid << "/" << oldoid << " -> " << cid << "/" << newoid << " = " << r << dendl; |
7c673cae FG |
3712 | assert(!m_filestore_fail_eio || r != -EIO); |
3713 | return r; | |
3714 | } | |
3715 | ||
3716 | int FileStore::_do_clone_range(int from, int to, uint64_t srcoff, uint64_t len, uint64_t dstoff) | |
3717 | { | |
31f18b77 | 3718 | dout(20) << __FUNC__ << ": copy " << srcoff << "~" << len << " to " << dstoff << dendl; |
7c673cae FG |
3719 | return backend->clone_range(from, to, srcoff, len, dstoff); |
3720 | } | |
3721 | ||
3722 | int FileStore::_do_sparse_copy_range(int from, int to, uint64_t srcoff, uint64_t len, uint64_t dstoff) | |
3723 | { | |
31f18b77 | 3724 | dout(20) << __FUNC__ << ": " << srcoff << "~" << len << " to " << dstoff << dendl; |
7c673cae FG |
3725 | int r = 0; |
3726 | map<uint64_t, uint64_t> exomap; | |
3727 | // fiemap doesn't allow zero length | |
3728 | if (len == 0) | |
3729 | return 0; | |
3730 | ||
3731 | if (backend->has_seek_data_hole()) { | |
3732 | dout(15) << "seek_data/seek_hole " << from << " " << srcoff << "~" << len << dendl; | |
3733 | r = _do_seek_hole_data(from, srcoff, len, &exomap); | |
3734 | } else if (backend->has_fiemap()) { | |
3735 | dout(15) << "fiemap ioctl" << from << " " << srcoff << "~" << len << dendl; | |
3736 | r = _do_fiemap(from, srcoff, len, &exomap); | |
3737 | } | |
3738 | ||
3739 | ||
3740 | int64_t written = 0; | |
3741 | if (r < 0) | |
3742 | goto out; | |
3743 | ||
3744 | for (map<uint64_t, uint64_t>::iterator miter = exomap.begin(); miter != exomap.end(); ++miter) { | |
3745 | uint64_t it_off = miter->first - srcoff + dstoff; | |
3746 | r = _do_copy_range(from, to, miter->first, miter->second, it_off, true); | |
3747 | if (r < 0) { | |
31f18b77 | 3748 | derr << __FUNC__ << ": copy error at " << miter->first << "~" << miter->second |
7c673cae FG |
3749 | << " to " << it_off << ", " << cpp_strerror(r) << dendl; |
3750 | break; | |
3751 | } | |
3752 | written += miter->second; | |
3753 | } | |
3754 | ||
3755 | if (r >= 0) { | |
3756 | if (m_filestore_sloppy_crc) { | |
3757 | int rc = backend->_crc_update_clone_range(from, to, srcoff, len, dstoff); | |
3758 | assert(rc >= 0); | |
3759 | } | |
3760 | struct stat st; | |
3761 | r = ::fstat(to, &st); | |
3762 | if (r < 0) { | |
3763 | r = -errno; | |
31f18b77 | 3764 | derr << __FUNC__ << ": fstat error at " << to << " " << cpp_strerror(r) << dendl; |
7c673cae FG |
3765 | goto out; |
3766 | } | |
3767 | if (st.st_size < (int)(dstoff + len)) { | |
3768 | r = ::ftruncate(to, dstoff + len); | |
3769 | if (r < 0) { | |
3770 | r = -errno; | |
31f18b77 | 3771 | derr << __FUNC__ << ": ftruncate error at " << dstoff+len << " " << cpp_strerror(r) << dendl; |
7c673cae FG |
3772 | goto out; |
3773 | } | |
3774 | } | |
3775 | r = written; | |
3776 | } | |
3777 | ||
3778 | out: | |
31f18b77 | 3779 | dout(20) << __FUNC__ << ": " << srcoff << "~" << len << " to " << dstoff << " = " << r << dendl; |
7c673cae FG |
3780 | return r; |
3781 | } | |
3782 | ||
3783 | int FileStore::_do_copy_range(int from, int to, uint64_t srcoff, uint64_t len, uint64_t dstoff, bool skip_sloppycrc) | |
3784 | { | |
31f18b77 | 3785 | dout(20) << __FUNC__ << ": " << srcoff << "~" << len << " to " << dstoff << dendl; |
7c673cae FG |
3786 | int r = 0; |
3787 | loff_t pos = srcoff; | |
3788 | loff_t end = srcoff + len; | |
3789 | int buflen = 4096 * 16; //limit by pipe max size.see fcntl | |
3790 | ||
3791 | #ifdef CEPH_HAVE_SPLICE | |
3792 | if (backend->has_splice()) { | |
3793 | int pipefd[2]; | |
3794 | if (pipe(pipefd) < 0) { | |
3795 | r = -errno; | |
3796 | derr << " pipe " << " got " << cpp_strerror(r) << dendl; | |
3797 | return r; | |
3798 | } | |
3799 | ||
3800 | loff_t dstpos = dstoff; | |
3801 | while (pos < end) { | |
3802 | int l = MIN(end-pos, buflen); | |
3803 | r = safe_splice(from, &pos, pipefd[1], NULL, l, SPLICE_F_NONBLOCK); | |
3804 | dout(10) << " safe_splice read from " << pos << "~" << l << " got " << r << dendl; | |
3805 | if (r < 0) { | |
31f18b77 | 3806 | derr << __FUNC__ << ": safe_splice read error at " << pos << "~" << len |
7c673cae FG |
3807 | << ", " << cpp_strerror(r) << dendl; |
3808 | break; | |
3809 | } | |
3810 | if (r == 0) { | |
3811 | // hrm, bad source range, wtf. | |
3812 | r = -ERANGE; | |
31f18b77 | 3813 | derr << __FUNC__ << ": got short read result at " << pos |
7c673cae FG |
3814 | << " of fd " << from << " len " << len << dendl; |
3815 | break; | |
3816 | } | |
3817 | ||
3818 | r = safe_splice(pipefd[0], NULL, to, &dstpos, r, 0); | |
3819 | dout(10) << " safe_splice write to " << to << " len " << r | |
3820 | << " got " << r << dendl; | |
3821 | if (r < 0) { | |
31f18b77 | 3822 | derr << __FUNC__ << ": write error at " << pos << "~" |
7c673cae FG |
3823 | << r << ", " << cpp_strerror(r) << dendl; |
3824 | break; | |
3825 | } | |
3826 | } | |
3827 | close(pipefd[0]); | |
3828 | close(pipefd[1]); | |
3829 | } else | |
3830 | #endif | |
3831 | { | |
3832 | int64_t actual; | |
3833 | ||
3834 | actual = ::lseek64(from, srcoff, SEEK_SET); | |
3835 | if (actual != (int64_t)srcoff) { | |
3836 | if (actual < 0) | |
3837 | r = -errno; | |
3838 | else | |
3839 | r = -EINVAL; | |
3840 | derr << "lseek64 to " << srcoff << " got " << cpp_strerror(r) << dendl; | |
3841 | return r; | |
3842 | } | |
3843 | actual = ::lseek64(to, dstoff, SEEK_SET); | |
3844 | if (actual != (int64_t)dstoff) { | |
3845 | if (actual < 0) | |
3846 | r = -errno; | |
3847 | else | |
3848 | r = -EINVAL; | |
3849 | derr << "lseek64 to " << dstoff << " got " << cpp_strerror(r) << dendl; | |
3850 | return r; | |
3851 | } | |
3852 | ||
3853 | char buf[buflen]; | |
3854 | while (pos < end) { | |
3855 | int l = MIN(end-pos, buflen); | |
3856 | r = ::read(from, buf, l); | |
3857 | dout(25) << " read from " << pos << "~" << l << " got " << r << dendl; | |
3858 | if (r < 0) { | |
3859 | if (errno == EINTR) { | |
3860 | continue; | |
3861 | } else { | |
3862 | r = -errno; | |
31f18b77 | 3863 | derr << __FUNC__ << ": read error at " << pos << "~" << len |
7c673cae FG |
3864 | << ", " << cpp_strerror(r) << dendl; |
3865 | break; | |
3866 | } | |
3867 | } | |
3868 | if (r == 0) { | |
3869 | // hrm, bad source range, wtf. | |
3870 | r = -ERANGE; | |
31f18b77 | 3871 | derr << __FUNC__ << ": got short read result at " << pos |
7c673cae FG |
3872 | << " of fd " << from << " len " << len << dendl; |
3873 | break; | |
3874 | } | |
3875 | int op = 0; | |
3876 | while (op < r) { | |
3877 | int r2 = safe_write(to, buf+op, r-op); | |
3878 | dout(25) << " write to " << to << " len " << (r-op) | |
3879 | << " got " << r2 << dendl; | |
3880 | if (r2 < 0) { | |
3881 | r = r2; | |
31f18b77 | 3882 | derr << __FUNC__ << ": write error at " << pos << "~" |
7c673cae FG |
3883 | << r-op << ", " << cpp_strerror(r) << dendl; |
3884 | ||
3885 | break; | |
3886 | } | |
3887 | op += (r-op); | |
3888 | } | |
3889 | if (r < 0) | |
3890 | break; | |
3891 | pos += r; | |
3892 | } | |
3893 | } | |
3894 | ||
3895 | if (r < 0 && replaying) { | |
3896 | assert(r == -ERANGE); | |
31f18b77 | 3897 | derr << __FUNC__ << ": short source tolerated because we are replaying" << dendl; |
94b18763 | 3898 | r = len; |
7c673cae FG |
3899 | } |
3900 | assert(replaying || pos == end); | |
3901 | if (r >= 0 && !skip_sloppycrc && m_filestore_sloppy_crc) { | |
3902 | int rc = backend->_crc_update_clone_range(from, to, srcoff, len, dstoff); | |
3903 | assert(rc >= 0); | |
3904 | } | |
31f18b77 | 3905 | dout(20) << __FUNC__ << ": " << srcoff << "~" << len << " to " << dstoff << " = " << r << dendl; |
7c673cae FG |
3906 | return r; |
3907 | } | |
3908 | ||
3909 | int FileStore::_clone_range(const coll_t& oldcid, const ghobject_t& oldoid, const coll_t& newcid, const ghobject_t& newoid, | |
3910 | uint64_t srcoff, uint64_t len, uint64_t dstoff, | |
3911 | const SequencerPosition& spos) | |
3912 | { | |
31f18b77 | 3913 | dout(15) << __FUNC__ << ": " << oldcid << "/" << oldoid << " -> " << newcid << "/" << newoid << " " << srcoff << "~" << len << " to " << dstoff << dendl; |
7c673cae FG |
3914 | |
3915 | if (_check_replay_guard(newcid, newoid, spos) < 0) | |
3916 | return 0; | |
3917 | ||
3918 | int r; | |
3919 | FDRef o, n; | |
3920 | r = lfn_open(oldcid, oldoid, false, &o); | |
3921 | if (r < 0) { | |
3922 | goto out2; | |
3923 | } | |
3924 | r = lfn_open(newcid, newoid, true, &n); | |
3925 | if (r < 0) { | |
3926 | goto out; | |
3927 | } | |
3928 | r = _do_clone_range(**o, **n, srcoff, len, dstoff); | |
3929 | if (r < 0) { | |
3930 | goto out3; | |
3931 | } | |
3932 | ||
3933 | // clone is non-idempotent; record our work. | |
3934 | _set_replay_guard(**n, spos, &newoid); | |
3935 | ||
3936 | out3: | |
3937 | lfn_close(n); | |
3938 | out: | |
3939 | lfn_close(o); | |
3940 | out2: | |
31f18b77 | 3941 | dout(10) << __FUNC__ << ": " << oldcid << "/" << oldoid << " -> " << newcid << "/" << newoid << " " |
7c673cae FG |
3942 | << srcoff << "~" << len << " to " << dstoff << " = " << r << dendl; |
3943 | return r; | |
3944 | } | |
3945 | ||
3946 | class SyncEntryTimeout : public Context { | |
3947 | public: | |
3948 | CephContext* cct; | |
3949 | explicit SyncEntryTimeout(CephContext* cct, int commit_timeo) | |
3950 | : cct(cct), m_commit_timeo(commit_timeo) | |
3951 | { | |
3952 | } | |
3953 | ||
3954 | void finish(int r) override { | |
3955 | BackTrace *bt = new BackTrace(1); | |
3956 | generic_dout(-1) << "FileStore: sync_entry timed out after " | |
3957 | << m_commit_timeo << " seconds.\n"; | |
3958 | bt->print(*_dout); | |
3959 | *_dout << dendl; | |
3960 | delete bt; | |
3961 | ceph_abort(); | |
3962 | } | |
3963 | private: | |
3964 | int m_commit_timeo; | |
3965 | }; | |
3966 | ||
3967 | void FileStore::sync_entry() | |
3968 | { | |
3969 | lock.Lock(); | |
3970 | while (!stop) { | |
3971 | utime_t max_interval; | |
3972 | max_interval.set_from_double(m_filestore_max_sync_interval); | |
3973 | utime_t min_interval; | |
3974 | min_interval.set_from_double(m_filestore_min_sync_interval); | |
3975 | ||
3976 | utime_t startwait = ceph_clock_now(); | |
3977 | if (!force_sync) { | |
31f18b77 | 3978 | dout(20) << __FUNC__ << ": waiting for max_interval " << max_interval << dendl; |
7c673cae FG |
3979 | sync_cond.WaitInterval(lock, max_interval); |
3980 | } else { | |
31f18b77 | 3981 | dout(20) << __FUNC__ << ": not waiting, force_sync set" << dendl; |
7c673cae FG |
3982 | } |
3983 | ||
3984 | if (force_sync) { | |
31f18b77 | 3985 | dout(20) << __FUNC__ << ": force_sync set" << dendl; |
7c673cae FG |
3986 | force_sync = false; |
3987 | } else if (stop) { | |
31f18b77 | 3988 | dout(20) << __FUNC__ << ": stop set" << dendl; |
7c673cae FG |
3989 | break; |
3990 | } else { | |
3991 | // wait for at least the min interval | |
3992 | utime_t woke = ceph_clock_now(); | |
3993 | woke -= startwait; | |
31f18b77 | 3994 | dout(20) << __FUNC__ << ": woke after " << woke << dendl; |
7c673cae FG |
3995 | if (woke < min_interval) { |
3996 | utime_t t = min_interval; | |
3997 | t -= woke; | |
31f18b77 | 3998 | dout(20) << __FUNC__ << ": waiting for another " << t |
7c673cae FG |
3999 | << " to reach min interval " << min_interval << dendl; |
4000 | sync_cond.WaitInterval(lock, t); | |
4001 | } | |
4002 | } | |
4003 | ||
4004 | list<Context*> fin; | |
4005 | again: | |
4006 | fin.swap(sync_waiters); | |
4007 | lock.Unlock(); | |
4008 | ||
4009 | op_tp.pause(); | |
4010 | if (apply_manager.commit_start()) { | |
4011 | utime_t start = ceph_clock_now(); | |
4012 | uint64_t cp = apply_manager.get_committing_seq(); | |
4013 | ||
4014 | sync_entry_timeo_lock.Lock(); | |
4015 | SyncEntryTimeout *sync_entry_timeo = | |
4016 | new SyncEntryTimeout(cct, m_filestore_commit_timeout); | |
224ce89b WB |
4017 | if (!timer.add_event_after(m_filestore_commit_timeout, |
4018 | sync_entry_timeo)) { | |
4019 | sync_entry_timeo = nullptr; | |
4020 | } | |
7c673cae FG |
4021 | sync_entry_timeo_lock.Unlock(); |
4022 | ||
4023 | logger->set(l_filestore_committing, 1); | |
4024 | ||
31f18b77 | 4025 | dout(15) << __FUNC__ << ": committing " << cp << dendl; |
7c673cae FG |
4026 | stringstream errstream; |
4027 | if (cct->_conf->filestore_debug_omap_check && !object_map->check(errstream)) { | |
4028 | derr << errstream.str() << dendl; | |
4029 | ceph_abort(); | |
4030 | } | |
4031 | ||
4032 | if (backend->can_checkpoint()) { | |
4033 | int err = write_op_seq(op_fd, cp); | |
4034 | if (err < 0) { | |
4035 | derr << "Error during write_op_seq: " << cpp_strerror(err) << dendl; | |
4036 | assert(0 == "error during write_op_seq"); | |
4037 | } | |
4038 | ||
4039 | char s[NAME_MAX]; | |
4040 | snprintf(s, sizeof(s), COMMIT_SNAP_ITEM, (long long unsigned)cp); | |
4041 | uint64_t cid = 0; | |
4042 | err = backend->create_checkpoint(s, &cid); | |
4043 | if (err < 0) { | |
4044 | int err = errno; | |
4045 | derr << "snap create '" << s << "' got error " << err << dendl; | |
4046 | assert(err == 0); | |
4047 | } | |
4048 | ||
4049 | snaps.push_back(cp); | |
4050 | apply_manager.commit_started(); | |
4051 | op_tp.unpause(); | |
4052 | ||
4053 | if (cid > 0) { | |
4054 | dout(20) << " waiting for checkpoint " << cid << " to complete" << dendl; | |
4055 | err = backend->sync_checkpoint(cid); | |
4056 | if (err < 0) { | |
4057 | derr << "ioctl WAIT_SYNC got " << cpp_strerror(err) << dendl; | |
4058 | assert(0 == "wait_sync got error"); | |
4059 | } | |
4060 | dout(20) << " done waiting for checkpoint " << cid << " to complete" << dendl; | |
4061 | } | |
224ce89b | 4062 | } else { |
7c673cae FG |
4063 | apply_manager.commit_started(); |
4064 | op_tp.unpause(); | |
4065 | ||
4066 | int err = object_map->sync(); | |
4067 | if (err < 0) { | |
4068 | derr << "object_map sync got " << cpp_strerror(err) << dendl; | |
4069 | assert(0 == "object_map sync returned error"); | |
4070 | } | |
4071 | ||
4072 | err = backend->syncfs(); | |
4073 | if (err < 0) { | |
4074 | derr << "syncfs got " << cpp_strerror(err) << dendl; | |
4075 | assert(0 == "syncfs returned error"); | |
4076 | } | |
4077 | ||
4078 | err = write_op_seq(op_fd, cp); | |
4079 | if (err < 0) { | |
4080 | derr << "Error during write_op_seq: " << cpp_strerror(err) << dendl; | |
4081 | assert(0 == "error during write_op_seq"); | |
4082 | } | |
4083 | err = ::fsync(op_fd); | |
4084 | if (err < 0) { | |
4085 | derr << "Error during fsync of op_seq: " << cpp_strerror(err) << dendl; | |
4086 | assert(0 == "error during fsync of op_seq"); | |
4087 | } | |
4088 | } | |
4089 | ||
4090 | utime_t done = ceph_clock_now(); | |
4091 | utime_t lat = done - start; | |
4092 | utime_t dur = done - startwait; | |
31f18b77 | 4093 | dout(10) << __FUNC__ << ": commit took " << lat << ", interval was " << dur << dendl; |
224ce89b WB |
4094 | utime_t max_pause_lat = logger->tget(l_filestore_sync_pause_max_lat); |
4095 | if (max_pause_lat < dur - lat) { | |
4096 | logger->tinc(l_filestore_sync_pause_max_lat, dur - lat); | |
4097 | } | |
7c673cae FG |
4098 | |
4099 | logger->inc(l_filestore_commitcycle); | |
4100 | logger->tinc(l_filestore_commitcycle_latency, lat); | |
4101 | logger->tinc(l_filestore_commitcycle_interval, dur); | |
4102 | ||
4103 | apply_manager.commit_finish(); | |
4104 | if (!m_disable_wbthrottle) { | |
4105 | wbthrottle.clear(); | |
4106 | } | |
4107 | ||
4108 | logger->set(l_filestore_committing, 0); | |
4109 | ||
4110 | // remove old snaps? | |
4111 | if (backend->can_checkpoint()) { | |
4112 | char s[NAME_MAX]; | |
4113 | while (snaps.size() > 2) { | |
4114 | snprintf(s, sizeof(s), COMMIT_SNAP_ITEM, (long long unsigned)snaps.front()); | |
4115 | snaps.pop_front(); | |
4116 | dout(10) << "removing snap '" << s << "'" << dendl; | |
4117 | int r = backend->destroy_checkpoint(s); | |
4118 | if (r) { | |
4119 | int err = errno; | |
4120 | derr << "unable to destroy snap '" << s << "' got " << cpp_strerror(err) << dendl; | |
4121 | } | |
4122 | } | |
4123 | } | |
4124 | ||
31f18b77 | 4125 | dout(15) << __FUNC__ << ": committed to op_seq " << cp << dendl; |
7c673cae | 4126 | |
224ce89b WB |
4127 | if (sync_entry_timeo) { |
4128 | Mutex::Locker lock(sync_entry_timeo_lock); | |
4129 | timer.cancel_event(sync_entry_timeo); | |
4130 | } | |
7c673cae FG |
4131 | } else { |
4132 | op_tp.unpause(); | |
4133 | } | |
4134 | ||
4135 | lock.Lock(); | |
4136 | finish_contexts(cct, fin, 0); | |
4137 | fin.clear(); | |
4138 | if (!sync_waiters.empty()) { | |
31f18b77 | 4139 | dout(10) << __FUNC__ << ": more waiters, committing again" << dendl; |
7c673cae FG |
4140 | goto again; |
4141 | } | |
4142 | if (!stop && journal && journal->should_commit_now()) { | |
31f18b77 | 4143 | dout(10) << __FUNC__ << ": journal says we should commit again (probably is/was full)" << dendl; |
7c673cae FG |
4144 | goto again; |
4145 | } | |
4146 | } | |
4147 | stop = false; | |
4148 | lock.Unlock(); | |
4149 | } | |
4150 | ||
4151 | void FileStore::_start_sync() | |
4152 | { | |
4153 | if (!journal) { // don't do a big sync if the journal is on | |
31f18b77 | 4154 | dout(10) << __FUNC__ << dendl; |
7c673cae FG |
4155 | sync_cond.Signal(); |
4156 | } else { | |
31f18b77 | 4157 | dout(10) << __FUNC__ << ": - NOOP (journal is on)" << dendl; |
7c673cae FG |
4158 | } |
4159 | } | |
4160 | ||
4161 | void FileStore::do_force_sync() | |
4162 | { | |
31f18b77 | 4163 | dout(10) << __FUNC__ << dendl; |
7c673cae FG |
4164 | Mutex::Locker l(lock); |
4165 | force_sync = true; | |
4166 | sync_cond.Signal(); | |
4167 | } | |
4168 | ||
4169 | void FileStore::start_sync(Context *onsafe) | |
4170 | { | |
4171 | Mutex::Locker l(lock); | |
4172 | sync_waiters.push_back(onsafe); | |
4173 | sync_cond.Signal(); | |
4174 | force_sync = true; | |
31f18b77 | 4175 | dout(10) << __FUNC__ << dendl; |
7c673cae FG |
4176 | } |
4177 | ||
4178 | void FileStore::sync() | |
4179 | { | |
4180 | Mutex l("FileStore::sync"); | |
4181 | Cond c; | |
4182 | bool done; | |
4183 | C_SafeCond *fin = new C_SafeCond(&l, &c, &done); | |
4184 | ||
4185 | start_sync(fin); | |
4186 | ||
4187 | l.Lock(); | |
4188 | while (!done) { | |
4189 | dout(10) << "sync waiting" << dendl; | |
4190 | c.Wait(l); | |
4191 | } | |
4192 | l.Unlock(); | |
4193 | dout(10) << "sync done" << dendl; | |
4194 | } | |
4195 | ||
4196 | void FileStore::_flush_op_queue() | |
4197 | { | |
31f18b77 | 4198 | dout(10) << __FUNC__ << ": draining op tp" << dendl; |
7c673cae | 4199 | op_wq.drain(); |
31f18b77 | 4200 | dout(10) << __FUNC__ << ": waiting for apply finisher" << dendl; |
7c673cae FG |
4201 | for (vector<Finisher*>::iterator it = apply_finishers.begin(); it != apply_finishers.end(); ++it) { |
4202 | (*it)->wait_for_empty(); | |
4203 | } | |
4204 | } | |
4205 | ||
4206 | /* | |
4207 | * flush - make every queued write readable | |
4208 | */ | |
4209 | void FileStore::flush() | |
4210 | { | |
31f18b77 | 4211 | dout(10) << __FUNC__ << dendl; |
7c673cae FG |
4212 | |
4213 | if (cct->_conf->filestore_blackhole) { | |
4214 | // wait forever | |
4215 | Mutex lock("FileStore::flush::lock"); | |
4216 | Cond cond; | |
4217 | lock.Lock(); | |
4218 | while (true) | |
4219 | cond.Wait(lock); | |
4220 | ceph_abort(); | |
4221 | } | |
4222 | ||
4223 | if (m_filestore_journal_writeahead) { | |
4224 | if (journal) | |
4225 | journal->flush(); | |
31f18b77 | 4226 | dout(10) << __FUNC__ << ": draining ondisk finisher" << dendl; |
7c673cae FG |
4227 | for (vector<Finisher*>::iterator it = ondisk_finishers.begin(); it != ondisk_finishers.end(); ++it) { |
4228 | (*it)->wait_for_empty(); | |
4229 | } | |
4230 | } | |
4231 | ||
4232 | _flush_op_queue(); | |
31f18b77 | 4233 | dout(10) << __FUNC__ << ": complete" << dendl; |
7c673cae FG |
4234 | } |
4235 | ||
4236 | /* | |
4237 | * sync_and_flush - make every queued write readable AND committed to disk | |
4238 | */ | |
4239 | void FileStore::sync_and_flush() | |
4240 | { | |
31f18b77 | 4241 | dout(10) << __FUNC__ << dendl; |
7c673cae FG |
4242 | |
4243 | if (m_filestore_journal_writeahead) { | |
4244 | if (journal) | |
4245 | journal->flush(); | |
4246 | _flush_op_queue(); | |
4247 | } else { | |
4248 | // includes m_filestore_journal_parallel | |
4249 | _flush_op_queue(); | |
4250 | sync(); | |
4251 | } | |
31f18b77 | 4252 | dout(10) << __FUNC__ << ": done" << dendl; |
7c673cae FG |
4253 | } |
4254 | ||
4255 | int FileStore::flush_journal() | |
4256 | { | |
31f18b77 | 4257 | dout(10) << __FUNC__ << dendl; |
7c673cae FG |
4258 | sync_and_flush(); |
4259 | sync(); | |
4260 | return 0; | |
4261 | } | |
4262 | ||
4263 | int FileStore::snapshot(const string& name) | |
4264 | { | |
31f18b77 | 4265 | dout(10) << __FUNC__ << ": " << name << dendl; |
7c673cae FG |
4266 | sync_and_flush(); |
4267 | ||
4268 | if (!backend->can_checkpoint()) { | |
31f18b77 | 4269 | dout(0) << __FUNC__ << ": " << name << " failed, not supported" << dendl; |
7c673cae FG |
4270 | return -EOPNOTSUPP; |
4271 | } | |
4272 | ||
4273 | char s[NAME_MAX]; | |
4274 | snprintf(s, sizeof(s), CLUSTER_SNAP_ITEM, name.c_str()); | |
4275 | ||
4276 | int r = backend->create_checkpoint(s, NULL); | |
4277 | if (r) { | |
31f18b77 | 4278 | derr << __FUNC__ << ": " << name << " failed: " << cpp_strerror(r) << dendl; |
7c673cae FG |
4279 | } |
4280 | ||
4281 | return r; | |
4282 | } | |
4283 | ||
4284 | // ------------------------------- | |
4285 | // attributes | |
4286 | ||
4287 | int FileStore::_fgetattr(int fd, const char *name, bufferptr& bp) | |
4288 | { | |
4289 | char val[CHAIN_XATTR_MAX_BLOCK_LEN]; | |
4290 | int l = chain_fgetxattr(fd, name, val, sizeof(val)); | |
4291 | if (l >= 0) { | |
4292 | bp = buffer::create(l); | |
4293 | memcpy(bp.c_str(), val, l); | |
4294 | } else if (l == -ERANGE) { | |
4295 | l = chain_fgetxattr(fd, name, 0, 0); | |
4296 | if (l > 0) { | |
4297 | bp = buffer::create(l); | |
4298 | l = chain_fgetxattr(fd, name, bp.c_str(), l); | |
4299 | } | |
4300 | } | |
4301 | assert(!m_filestore_fail_eio || l != -EIO); | |
4302 | return l; | |
4303 | } | |
4304 | ||
4305 | int FileStore::_fgetattrs(int fd, map<string,bufferptr>& aset) | |
4306 | { | |
4307 | // get attr list | |
4308 | char names1[100]; | |
4309 | int len = chain_flistxattr(fd, names1, sizeof(names1)-1); | |
4310 | char *names2 = 0; | |
4311 | char *name = 0; | |
4312 | if (len == -ERANGE) { | |
4313 | len = chain_flistxattr(fd, 0, 0); | |
4314 | if (len < 0) { | |
4315 | assert(!m_filestore_fail_eio || len != -EIO); | |
4316 | return len; | |
4317 | } | |
4318 | dout(10) << " -ERANGE, len is " << len << dendl; | |
4319 | names2 = new char[len+1]; | |
4320 | len = chain_flistxattr(fd, names2, len); | |
4321 | dout(10) << " -ERANGE, got " << len << dendl; | |
4322 | if (len < 0) { | |
4323 | assert(!m_filestore_fail_eio || len != -EIO); | |
4324 | delete[] names2; | |
4325 | return len; | |
4326 | } | |
4327 | name = names2; | |
4328 | } else if (len < 0) { | |
4329 | assert(!m_filestore_fail_eio || len != -EIO); | |
4330 | return len; | |
4331 | } else { | |
4332 | name = names1; | |
4333 | } | |
4334 | name[len] = 0; | |
4335 | ||
4336 | char *end = name + len; | |
4337 | while (name < end) { | |
4338 | char *attrname = name; | |
4339 | if (parse_attrname(&name)) { | |
4340 | if (*name) { | |
31f18b77 | 4341 | dout(20) << __FUNC__ << ": " << fd << " getting '" << name << "'" << dendl; |
7c673cae FG |
4342 | int r = _fgetattr(fd, attrname, aset[name]); |
4343 | if (r < 0) { | |
4344 | delete[] names2; | |
4345 | return r; | |
4346 | } | |
4347 | } | |
4348 | } | |
4349 | name += strlen(name) + 1; | |
4350 | } | |
4351 | ||
4352 | delete[] names2; | |
4353 | return 0; | |
4354 | } | |
4355 | ||
4356 | int FileStore::_fsetattrs(int fd, map<string, bufferptr> &aset) | |
4357 | { | |
4358 | for (map<string, bufferptr>::iterator p = aset.begin(); | |
4359 | p != aset.end(); | |
4360 | ++p) { | |
4361 | char n[CHAIN_XATTR_MAX_NAME_LEN]; | |
4362 | get_attrname(p->first.c_str(), n, CHAIN_XATTR_MAX_NAME_LEN); | |
4363 | const char *val; | |
4364 | if (p->second.length()) | |
4365 | val = p->second.c_str(); | |
4366 | else | |
4367 | val = ""; | |
4368 | // ??? Why do we skip setting all the other attrs if one fails? | |
4369 | int r = chain_fsetxattr(fd, n, val, p->second.length()); | |
4370 | if (r < 0) { | |
31f18b77 | 4371 | derr << __FUNC__ << ": chain_setxattr returned " << r << dendl; |
7c673cae FG |
4372 | return r; |
4373 | } | |
4374 | } | |
4375 | return 0; | |
4376 | } | |
4377 | ||
4378 | // debug EIO injection | |
4379 | void FileStore::inject_data_error(const ghobject_t &oid) { | |
4380 | Mutex::Locker l(read_error_lock); | |
31f18b77 | 4381 | dout(10) << __FUNC__ << ": init error on " << oid << dendl; |
7c673cae FG |
4382 | data_error_set.insert(oid); |
4383 | } | |
4384 | void FileStore::inject_mdata_error(const ghobject_t &oid) { | |
4385 | Mutex::Locker l(read_error_lock); | |
31f18b77 | 4386 | dout(10) << __FUNC__ << ": init error on " << oid << dendl; |
7c673cae FG |
4387 | mdata_error_set.insert(oid); |
4388 | } | |
224ce89b | 4389 | |
7c673cae FG |
4390 | void FileStore::debug_obj_on_delete(const ghobject_t &oid) { |
4391 | Mutex::Locker l(read_error_lock); | |
31f18b77 | 4392 | dout(10) << __FUNC__ << ": clear error on " << oid << dendl; |
7c673cae FG |
4393 | data_error_set.erase(oid); |
4394 | mdata_error_set.erase(oid); | |
4395 | } | |
4396 | bool FileStore::debug_data_eio(const ghobject_t &oid) { | |
4397 | Mutex::Locker l(read_error_lock); | |
4398 | if (data_error_set.count(oid)) { | |
31f18b77 | 4399 | dout(10) << __FUNC__ << ": inject error on " << oid << dendl; |
7c673cae FG |
4400 | return true; |
4401 | } else { | |
4402 | return false; | |
4403 | } | |
4404 | } | |
4405 | bool FileStore::debug_mdata_eio(const ghobject_t &oid) { | |
4406 | Mutex::Locker l(read_error_lock); | |
4407 | if (mdata_error_set.count(oid)) { | |
31f18b77 | 4408 | dout(10) << __FUNC__ << ": inject error on " << oid << dendl; |
7c673cae FG |
4409 | return true; |
4410 | } else { | |
4411 | return false; | |
4412 | } | |
4413 | } | |
4414 | ||
4415 | ||
4416 | // objects | |
4417 | ||
4418 | int FileStore::getattr(const coll_t& _cid, const ghobject_t& oid, const char *name, bufferptr &bp) | |
4419 | { | |
4420 | tracepoint(objectstore, getattr_enter, _cid.c_str()); | |
4421 | const coll_t& cid = !_need_temp_object_collection(_cid, oid) ? _cid : _cid.get_temp(); | |
31f18b77 | 4422 | dout(15) << __FUNC__ << ": " << cid << "/" << oid << " '" << name << "'" << dendl; |
7c673cae FG |
4423 | FDRef fd; |
4424 | int r = lfn_open(cid, oid, false, &fd); | |
4425 | if (r < 0) { | |
4426 | goto out; | |
4427 | } | |
4428 | char n[CHAIN_XATTR_MAX_NAME_LEN]; | |
4429 | get_attrname(name, n, CHAIN_XATTR_MAX_NAME_LEN); | |
4430 | r = _fgetattr(**fd, n, bp); | |
4431 | lfn_close(fd); | |
4432 | if (r == -ENODATA) { | |
4433 | map<string, bufferlist> got; | |
4434 | set<string> to_get; | |
4435 | to_get.insert(string(name)); | |
4436 | Index index; | |
4437 | r = get_index(cid, &index); | |
4438 | if (r < 0) { | |
31f18b77 | 4439 | dout(10) << __FUNC__ << ": could not get index r = " << r << dendl; |
7c673cae FG |
4440 | goto out; |
4441 | } | |
4442 | r = object_map->get_xattrs(oid, to_get, &got); | |
4443 | if (r < 0 && r != -ENOENT) { | |
31f18b77 | 4444 | dout(10) << __FUNC__ << ": get_xattrs err r =" << r << dendl; |
7c673cae FG |
4445 | goto out; |
4446 | } | |
4447 | if (got.empty()) { | |
31f18b77 | 4448 | dout(10) << __FUNC__ << ": got.size() is 0" << dendl; |
7c673cae FG |
4449 | return -ENODATA; |
4450 | } | |
4451 | bp = bufferptr(got.begin()->second.c_str(), | |
4452 | got.begin()->second.length()); | |
4453 | r = bp.length(); | |
4454 | } | |
4455 | out: | |
31f18b77 | 4456 | dout(10) << __FUNC__ << ": " << cid << "/" << oid << " '" << name << "' = " << r << dendl; |
7c673cae FG |
4457 | assert(!m_filestore_fail_eio || r != -EIO); |
4458 | if (cct->_conf->filestore_debug_inject_read_err && | |
4459 | debug_mdata_eio(oid)) { | |
4460 | return -EIO; | |
4461 | } else { | |
4462 | tracepoint(objectstore, getattr_exit, r); | |
4463 | return r < 0 ? r : 0; | |
4464 | } | |
4465 | } | |
4466 | ||
4467 | int FileStore::getattrs(const coll_t& _cid, const ghobject_t& oid, map<string,bufferptr>& aset) | |
4468 | { | |
4469 | tracepoint(objectstore, getattrs_enter, _cid.c_str()); | |
4470 | const coll_t& cid = !_need_temp_object_collection(_cid, oid) ? _cid : _cid.get_temp(); | |
4471 | set<string> omap_attrs; | |
4472 | map<string, bufferlist> omap_aset; | |
4473 | Index index; | |
31f18b77 | 4474 | dout(15) << __FUNC__ << ": " << cid << "/" << oid << dendl; |
7c673cae FG |
4475 | FDRef fd; |
4476 | bool spill_out = true; | |
4477 | char buf[2]; | |
4478 | ||
4479 | int r = lfn_open(cid, oid, false, &fd); | |
4480 | if (r < 0) { | |
4481 | goto out; | |
4482 | } | |
4483 | ||
4484 | r = chain_fgetxattr(**fd, XATTR_SPILL_OUT_NAME, buf, sizeof(buf)); | |
4485 | if (r >= 0 && !strncmp(buf, XATTR_NO_SPILL_OUT, sizeof(XATTR_NO_SPILL_OUT))) | |
4486 | spill_out = false; | |
4487 | ||
4488 | r = _fgetattrs(**fd, aset); | |
4489 | lfn_close(fd); | |
4490 | fd = FDRef(); // defensive | |
4491 | if (r < 0) { | |
4492 | goto out; | |
4493 | } | |
4494 | ||
4495 | if (!spill_out) { | |
31f18b77 | 4496 | dout(10) << __FUNC__ << ": no xattr exists in object_map r = " << r << dendl; |
7c673cae FG |
4497 | goto out; |
4498 | } | |
4499 | ||
4500 | r = get_index(cid, &index); | |
4501 | if (r < 0) { | |
31f18b77 | 4502 | dout(10) << __FUNC__ << ": could not get index r = " << r << dendl; |
7c673cae FG |
4503 | goto out; |
4504 | } | |
4505 | { | |
4506 | r = object_map->get_all_xattrs(oid, &omap_attrs); | |
4507 | if (r < 0 && r != -ENOENT) { | |
31f18b77 | 4508 | dout(10) << __FUNC__ << ": could not get omap_attrs r = " << r << dendl; |
7c673cae FG |
4509 | goto out; |
4510 | } | |
4511 | ||
4512 | r = object_map->get_xattrs(oid, omap_attrs, &omap_aset); | |
4513 | if (r < 0 && r != -ENOENT) { | |
31f18b77 | 4514 | dout(10) << __FUNC__ << ": could not get omap_attrs r = " << r << dendl; |
7c673cae FG |
4515 | goto out; |
4516 | } | |
4517 | if (r == -ENOENT) | |
4518 | r = 0; | |
4519 | } | |
4520 | assert(omap_attrs.size() == omap_aset.size()); | |
4521 | for (map<string, bufferlist>::iterator i = omap_aset.begin(); | |
4522 | i != omap_aset.end(); | |
4523 | ++i) { | |
4524 | string key(i->first); | |
4525 | aset.insert(make_pair(key, | |
4526 | bufferptr(i->second.c_str(), i->second.length()))); | |
4527 | } | |
4528 | out: | |
31f18b77 | 4529 | dout(10) << __FUNC__ << ": " << cid << "/" << oid << " = " << r << dendl; |
7c673cae FG |
4530 | assert(!m_filestore_fail_eio || r != -EIO); |
4531 | ||
4532 | if (cct->_conf->filestore_debug_inject_read_err && | |
4533 | debug_mdata_eio(oid)) { | |
4534 | return -EIO; | |
4535 | } else { | |
4536 | tracepoint(objectstore, getattrs_exit, r); | |
4537 | return r; | |
4538 | } | |
4539 | } | |
4540 | ||
4541 | int FileStore::_setattrs(const coll_t& cid, const ghobject_t& oid, map<string,bufferptr>& aset, | |
4542 | const SequencerPosition &spos) | |
4543 | { | |
4544 | map<string, bufferlist> omap_set; | |
4545 | set<string> omap_remove; | |
4546 | map<string, bufferptr> inline_set; | |
4547 | map<string, bufferptr> inline_to_set; | |
4548 | FDRef fd; | |
4549 | int spill_out = -1; | |
4550 | bool incomplete_inline = false; | |
4551 | ||
4552 | int r = lfn_open(cid, oid, false, &fd); | |
4553 | if (r < 0) { | |
4554 | goto out; | |
4555 | } | |
4556 | ||
4557 | char buf[2]; | |
4558 | r = chain_fgetxattr(**fd, XATTR_SPILL_OUT_NAME, buf, sizeof(buf)); | |
4559 | if (r >= 0 && !strncmp(buf, XATTR_NO_SPILL_OUT, sizeof(XATTR_NO_SPILL_OUT))) | |
4560 | spill_out = 0; | |
4561 | else | |
4562 | spill_out = 1; | |
4563 | ||
4564 | r = _fgetattrs(**fd, inline_set); | |
4565 | incomplete_inline = (r == -E2BIG); | |
4566 | assert(!m_filestore_fail_eio || r != -EIO); | |
31f18b77 | 4567 | dout(15) << __FUNC__ << ": " << cid << "/" << oid |
7c673cae FG |
4568 | << (incomplete_inline ? " (incomplete_inline, forcing omap)" : "") |
4569 | << dendl; | |
4570 | ||
4571 | for (map<string,bufferptr>::iterator p = aset.begin(); | |
4572 | p != aset.end(); | |
4573 | ++p) { | |
4574 | char n[CHAIN_XATTR_MAX_NAME_LEN]; | |
4575 | get_attrname(p->first.c_str(), n, CHAIN_XATTR_MAX_NAME_LEN); | |
4576 | ||
4577 | if (incomplete_inline) { | |
4578 | chain_fremovexattr(**fd, n); // ignore any error | |
4579 | omap_set[p->first].push_back(p->second); | |
4580 | continue; | |
4581 | } | |
4582 | ||
4583 | if (p->second.length() > m_filestore_max_inline_xattr_size) { | |
4584 | if (inline_set.count(p->first)) { | |
4585 | inline_set.erase(p->first); | |
4586 | r = chain_fremovexattr(**fd, n); | |
4587 | if (r < 0) | |
4588 | goto out_close; | |
4589 | } | |
4590 | omap_set[p->first].push_back(p->second); | |
4591 | continue; | |
4592 | } | |
4593 | ||
4594 | if (!inline_set.count(p->first) && | |
4595 | inline_set.size() >= m_filestore_max_inline_xattrs) { | |
4596 | omap_set[p->first].push_back(p->second); | |
4597 | continue; | |
4598 | } | |
4599 | omap_remove.insert(p->first); | |
4600 | inline_set.insert(*p); | |
4601 | ||
4602 | inline_to_set.insert(*p); | |
4603 | } | |
4604 | ||
4605 | if (spill_out != 1 && !omap_set.empty()) { | |
4606 | chain_fsetxattr(**fd, XATTR_SPILL_OUT_NAME, XATTR_SPILL_OUT, | |
4607 | sizeof(XATTR_SPILL_OUT)); | |
4608 | } | |
4609 | ||
4610 | r = _fsetattrs(**fd, inline_to_set); | |
4611 | if (r < 0) | |
4612 | goto out_close; | |
4613 | ||
4614 | if (spill_out && !omap_remove.empty()) { | |
4615 | r = object_map->remove_xattrs(oid, omap_remove, &spos); | |
4616 | if (r < 0 && r != -ENOENT) { | |
31f18b77 | 4617 | dout(10) << __FUNC__ << ": could not remove_xattrs r = " << r << dendl; |
7c673cae FG |
4618 | assert(!m_filestore_fail_eio || r != -EIO); |
4619 | goto out_close; | |
4620 | } else { | |
4621 | r = 0; // don't confuse the debug output | |
4622 | } | |
4623 | } | |
4624 | ||
4625 | if (!omap_set.empty()) { | |
4626 | r = object_map->set_xattrs(oid, omap_set, &spos); | |
4627 | if (r < 0) { | |
31f18b77 | 4628 | dout(10) << __FUNC__ << ": could not set_xattrs r = " << r << dendl; |
7c673cae FG |
4629 | assert(!m_filestore_fail_eio || r != -EIO); |
4630 | goto out_close; | |
4631 | } | |
4632 | } | |
4633 | out_close: | |
4634 | lfn_close(fd); | |
4635 | out: | |
31f18b77 | 4636 | dout(10) << __FUNC__ << ": " << cid << "/" << oid << " = " << r << dendl; |
7c673cae FG |
4637 | return r; |
4638 | } | |
4639 | ||
4640 | ||
4641 | int FileStore::_rmattr(const coll_t& cid, const ghobject_t& oid, const char *name, | |
4642 | const SequencerPosition &spos) | |
4643 | { | |
31f18b77 | 4644 | dout(15) << __FUNC__ << ": " << cid << "/" << oid << " '" << name << "'" << dendl; |
7c673cae FG |
4645 | FDRef fd; |
4646 | bool spill_out = true; | |
4647 | ||
4648 | int r = lfn_open(cid, oid, false, &fd); | |
4649 | if (r < 0) { | |
4650 | goto out; | |
4651 | } | |
4652 | ||
4653 | char buf[2]; | |
4654 | r = chain_fgetxattr(**fd, XATTR_SPILL_OUT_NAME, buf, sizeof(buf)); | |
4655 | if (r >= 0 && !strncmp(buf, XATTR_NO_SPILL_OUT, sizeof(XATTR_NO_SPILL_OUT))) { | |
4656 | spill_out = false; | |
4657 | } | |
4658 | ||
4659 | char n[CHAIN_XATTR_MAX_NAME_LEN]; | |
4660 | get_attrname(name, n, CHAIN_XATTR_MAX_NAME_LEN); | |
4661 | r = chain_fremovexattr(**fd, n); | |
4662 | if (r == -ENODATA && spill_out) { | |
4663 | Index index; | |
4664 | r = get_index(cid, &index); | |
4665 | if (r < 0) { | |
31f18b77 | 4666 | dout(10) << __FUNC__ << ": could not get index r = " << r << dendl; |
7c673cae FG |
4667 | goto out_close; |
4668 | } | |
4669 | set<string> to_remove; | |
4670 | to_remove.insert(string(name)); | |
4671 | r = object_map->remove_xattrs(oid, to_remove, &spos); | |
4672 | if (r < 0 && r != -ENOENT) { | |
31f18b77 | 4673 | dout(10) << __FUNC__ << ": could not remove_xattrs index r = " << r << dendl; |
7c673cae FG |
4674 | assert(!m_filestore_fail_eio || r != -EIO); |
4675 | goto out_close; | |
4676 | } | |
4677 | } | |
4678 | out_close: | |
4679 | lfn_close(fd); | |
4680 | out: | |
31f18b77 | 4681 | dout(10) << __FUNC__ << ": " << cid << "/" << oid << " '" << name << "' = " << r << dendl; |
7c673cae FG |
4682 | return r; |
4683 | } | |
4684 | ||
4685 | int FileStore::_rmattrs(const coll_t& cid, const ghobject_t& oid, | |
4686 | const SequencerPosition &spos) | |
4687 | { | |
31f18b77 | 4688 | dout(15) << __FUNC__ << ": " << cid << "/" << oid << dendl; |
7c673cae FG |
4689 | |
4690 | map<string,bufferptr> aset; | |
4691 | FDRef fd; | |
4692 | set<string> omap_attrs; | |
4693 | Index index; | |
4694 | bool spill_out = true; | |
4695 | ||
4696 | int r = lfn_open(cid, oid, false, &fd); | |
4697 | if (r < 0) { | |
4698 | goto out; | |
4699 | } | |
4700 | ||
4701 | char buf[2]; | |
4702 | r = chain_fgetxattr(**fd, XATTR_SPILL_OUT_NAME, buf, sizeof(buf)); | |
4703 | if (r >= 0 && !strncmp(buf, XATTR_NO_SPILL_OUT, sizeof(XATTR_NO_SPILL_OUT))) { | |
4704 | spill_out = false; | |
4705 | } | |
4706 | ||
4707 | r = _fgetattrs(**fd, aset); | |
4708 | if (r >= 0) { | |
4709 | for (map<string,bufferptr>::iterator p = aset.begin(); p != aset.end(); ++p) { | |
4710 | char n[CHAIN_XATTR_MAX_NAME_LEN]; | |
4711 | get_attrname(p->first.c_str(), n, CHAIN_XATTR_MAX_NAME_LEN); | |
4712 | r = chain_fremovexattr(**fd, n); | |
4713 | if (r < 0) { | |
31f18b77 | 4714 | dout(10) << __FUNC__ << ": could not remove xattr r = " << r << dendl; |
7c673cae FG |
4715 | goto out_close; |
4716 | } | |
4717 | } | |
4718 | } | |
4719 | ||
4720 | if (!spill_out) { | |
31f18b77 | 4721 | dout(10) << __FUNC__ << ": no xattr exists in object_map r = " << r << dendl; |
7c673cae FG |
4722 | goto out_close; |
4723 | } | |
4724 | ||
4725 | r = get_index(cid, &index); | |
4726 | if (r < 0) { | |
31f18b77 | 4727 | dout(10) << __FUNC__ << ": could not get index r = " << r << dendl; |
7c673cae FG |
4728 | goto out_close; |
4729 | } | |
4730 | { | |
4731 | r = object_map->get_all_xattrs(oid, &omap_attrs); | |
4732 | if (r < 0 && r != -ENOENT) { | |
31f18b77 | 4733 | dout(10) << __FUNC__ << ": could not get omap_attrs r = " << r << dendl; |
7c673cae FG |
4734 | assert(!m_filestore_fail_eio || r != -EIO); |
4735 | goto out_close; | |
4736 | } | |
4737 | r = object_map->remove_xattrs(oid, omap_attrs, &spos); | |
4738 | if (r < 0 && r != -ENOENT) { | |
31f18b77 | 4739 | dout(10) << __FUNC__ << ": could not remove omap_attrs r = " << r << dendl; |
7c673cae FG |
4740 | goto out_close; |
4741 | } | |
4742 | if (r == -ENOENT) | |
4743 | r = 0; | |
4744 | chain_fsetxattr(**fd, XATTR_SPILL_OUT_NAME, XATTR_NO_SPILL_OUT, | |
4745 | sizeof(XATTR_NO_SPILL_OUT)); | |
4746 | } | |
4747 | ||
4748 | out_close: | |
4749 | lfn_close(fd); | |
4750 | out: | |
31f18b77 | 4751 | dout(10) << __FUNC__ << ": " << cid << "/" << oid << " = " << r << dendl; |
7c673cae FG |
4752 | return r; |
4753 | } | |
4754 | ||
4755 | ||
4756 | ||
4757 | ||
4758 | int FileStore::_collection_remove_recursive(const coll_t &cid, | |
4759 | const SequencerPosition &spos) | |
4760 | { | |
4761 | struct stat st; | |
4762 | int r = collection_stat(cid, &st); | |
4763 | if (r < 0) { | |
4764 | if (r == -ENOENT) | |
4765 | return 0; | |
4766 | return r; | |
4767 | } | |
4768 | ||
4769 | vector<ghobject_t> objects; | |
4770 | ghobject_t max; | |
4771 | while (!max.is_max()) { | |
4772 | r = collection_list(cid, max, ghobject_t::get_max(), | |
4773 | 300, &objects, &max); | |
4774 | if (r < 0) | |
4775 | return r; | |
4776 | for (vector<ghobject_t>::iterator i = objects.begin(); | |
4777 | i != objects.end(); | |
4778 | ++i) { | |
4779 | assert(_check_replay_guard(cid, *i, spos)); | |
4780 | r = _remove(cid, *i, spos); | |
4781 | if (r < 0) | |
4782 | return r; | |
4783 | } | |
4784 | objects.clear(); | |
4785 | } | |
4786 | return _destroy_collection(cid); | |
4787 | } | |
4788 | ||
4789 | // -------------------------- | |
4790 | // collections | |
4791 | ||
4792 | int FileStore::list_collections(vector<coll_t>& ls) | |
4793 | { | |
4794 | return list_collections(ls, false); | |
4795 | } | |
4796 | ||
4797 | int FileStore::list_collections(vector<coll_t>& ls, bool include_temp) | |
4798 | { | |
4799 | tracepoint(objectstore, list_collections_enter); | |
31f18b77 | 4800 | dout(10) << __FUNC__ << dendl; |
7c673cae FG |
4801 | |
4802 | char fn[PATH_MAX]; | |
4803 | snprintf(fn, sizeof(fn), "%s/current", basedir.c_str()); | |
4804 | ||
4805 | int r = 0; | |
4806 | DIR *dir = ::opendir(fn); | |
4807 | if (!dir) { | |
4808 | r = -errno; | |
4809 | derr << "tried opening directory " << fn << ": " << cpp_strerror(-r) << dendl; | |
4810 | assert(!m_filestore_fail_eio || r != -EIO); | |
4811 | return r; | |
4812 | } | |
4813 | ||
4814 | struct dirent *de = nullptr; | |
4815 | while ((de = ::readdir(dir))) { | |
4816 | if (de->d_type == DT_UNKNOWN) { | |
4817 | // d_type not supported (non-ext[234], btrfs), must stat | |
4818 | struct stat sb; | |
4819 | char filename[PATH_MAX]; | |
4820 | snprintf(filename, sizeof(filename), "%s/%s", fn, de->d_name); | |
4821 | ||
4822 | r = ::stat(filename, &sb); | |
4823 | if (r < 0) { | |
4824 | r = -errno; | |
4825 | derr << "stat on " << filename << ": " << cpp_strerror(-r) << dendl; | |
4826 | assert(!m_filestore_fail_eio || r != -EIO); | |
4827 | break; | |
4828 | } | |
4829 | if (!S_ISDIR(sb.st_mode)) { | |
4830 | continue; | |
4831 | } | |
4832 | } else if (de->d_type != DT_DIR) { | |
4833 | continue; | |
4834 | } | |
4835 | if (strcmp(de->d_name, "omap") == 0) { | |
4836 | continue; | |
4837 | } | |
4838 | if (de->d_name[0] == '.' && | |
4839 | (de->d_name[1] == '\0' || | |
4840 | (de->d_name[1] == '.' && | |
4841 | de->d_name[2] == '\0'))) | |
4842 | continue; | |
4843 | coll_t cid; | |
4844 | if (!cid.parse(de->d_name)) { | |
4845 | derr << "ignoring invalid collection '" << de->d_name << "'" << dendl; | |
4846 | continue; | |
4847 | } | |
4848 | if (!cid.is_temp() || include_temp) | |
4849 | ls.push_back(cid); | |
4850 | } | |
4851 | ||
4852 | if (r > 0) { | |
4853 | derr << "trying readdir " << fn << ": " << cpp_strerror(r) << dendl; | |
4854 | r = -r; | |
4855 | } | |
4856 | ||
4857 | ::closedir(dir); | |
4858 | assert(!m_filestore_fail_eio || r != -EIO); | |
4859 | tracepoint(objectstore, list_collections_exit, r); | |
4860 | return r; | |
4861 | } | |
4862 | ||
4863 | int FileStore::collection_stat(const coll_t& c, struct stat *st) | |
4864 | { | |
4865 | tracepoint(objectstore, collection_stat_enter, c.c_str()); | |
4866 | char fn[PATH_MAX]; | |
4867 | get_cdir(c, fn, sizeof(fn)); | |
31f18b77 | 4868 | dout(15) << __FUNC__ << ": " << fn << dendl; |
7c673cae FG |
4869 | int r = ::stat(fn, st); |
4870 | if (r < 0) | |
4871 | r = -errno; | |
31f18b77 | 4872 | dout(10) << __FUNC__ << ": " << fn << " = " << r << dendl; |
7c673cae FG |
4873 | assert(!m_filestore_fail_eio || r != -EIO); |
4874 | tracepoint(objectstore, collection_stat_exit, r); | |
4875 | return r; | |
4876 | } | |
4877 | ||
4878 | bool FileStore::collection_exists(const coll_t& c) | |
4879 | { | |
4880 | tracepoint(objectstore, collection_exists_enter, c.c_str()); | |
4881 | struct stat st; | |
4882 | bool ret = collection_stat(c, &st) == 0; | |
4883 | tracepoint(objectstore, collection_exists_exit, ret); | |
4884 | return ret; | |
4885 | } | |
4886 | ||
4887 | int FileStore::collection_empty(const coll_t& c, bool *empty) | |
4888 | { | |
4889 | tracepoint(objectstore, collection_empty_enter, c.c_str()); | |
31f18b77 | 4890 | dout(15) << __FUNC__ << ": " << c << dendl; |
7c673cae FG |
4891 | Index index; |
4892 | int r = get_index(c, &index); | |
4893 | if (r < 0) { | |
31f18b77 | 4894 | derr << __FUNC__ << ": get_index returned: " << cpp_strerror(r) |
7c673cae FG |
4895 | << dendl; |
4896 | return r; | |
4897 | } | |
4898 | ||
4899 | assert(NULL != index.index); | |
4900 | RWLock::RLocker l((index.index)->access_lock); | |
4901 | ||
4902 | vector<ghobject_t> ls; | |
4903 | r = index->collection_list_partial(ghobject_t(), ghobject_t::get_max(), | |
4904 | 1, &ls, NULL); | |
4905 | if (r < 0) { | |
31f18b77 | 4906 | derr << __FUNC__ << ": collection_list_partial returned: " |
7c673cae FG |
4907 | << cpp_strerror(r) << dendl; |
4908 | assert(!m_filestore_fail_eio || r != -EIO); | |
4909 | return r; | |
4910 | } | |
4911 | *empty = ls.empty(); | |
4912 | tracepoint(objectstore, collection_empty_exit, *empty); | |
4913 | return 0; | |
4914 | } | |
4915 | ||
4916 | int FileStore::_collection_set_bits(const coll_t& c, int bits) | |
4917 | { | |
4918 | char fn[PATH_MAX]; | |
4919 | get_cdir(c, fn, sizeof(fn)); | |
31f18b77 | 4920 | dout(10) << __FUNC__ << ": " << fn << " " << bits << dendl; |
7c673cae FG |
4921 | char n[PATH_MAX]; |
4922 | int r; | |
4923 | int32_t v = bits; | |
4924 | int fd = ::open(fn, O_RDONLY); | |
4925 | if (fd < 0) { | |
4926 | r = -errno; | |
4927 | goto out; | |
4928 | } | |
4929 | get_attrname("bits", n, PATH_MAX); | |
4930 | r = chain_fsetxattr(fd, n, (char*)&v, sizeof(v)); | |
4931 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
4932 | out: | |
31f18b77 | 4933 | dout(10) << __FUNC__ << ": " << fn << " " << bits << " = " << r << dendl; |
7c673cae FG |
4934 | return r; |
4935 | } | |
4936 | ||
4937 | int FileStore::collection_bits(const coll_t& c) | |
4938 | { | |
4939 | char fn[PATH_MAX]; | |
4940 | get_cdir(c, fn, sizeof(fn)); | |
31f18b77 | 4941 | dout(15) << __FUNC__ << ": " << fn << dendl; |
7c673cae FG |
4942 | int r; |
4943 | char n[PATH_MAX]; | |
4944 | int32_t bits; | |
4945 | int fd = ::open(fn, O_RDONLY); | |
4946 | if (fd < 0) { | |
4947 | bits = r = -errno; | |
4948 | goto out; | |
4949 | } | |
4950 | get_attrname("bits", n, PATH_MAX); | |
4951 | r = chain_fgetxattr(fd, n, (char*)&bits, sizeof(bits)); | |
4952 | VOID_TEMP_FAILURE_RETRY(::close(fd)); | |
4953 | if (r < 0) { | |
4954 | bits = r; | |
4955 | goto out; | |
4956 | } | |
4957 | out: | |
31f18b77 | 4958 | dout(10) << __FUNC__ << ": " << fn << " = " << bits << dendl; |
7c673cae FG |
4959 | return bits; |
4960 | } | |
4961 | ||
4962 | int FileStore::collection_list(const coll_t& c, | |
4963 | const ghobject_t& orig_start, | |
4964 | const ghobject_t& end, | |
4965 | int max, | |
4966 | vector<ghobject_t> *ls, ghobject_t *next) | |
4967 | { | |
4968 | ghobject_t start = orig_start; | |
4969 | if (start.is_max()) | |
4970 | return 0; | |
4971 | ||
4972 | ghobject_t temp_next; | |
4973 | if (!next) | |
4974 | next = &temp_next; | |
4975 | // figure out the pool id. we need this in order to generate a | |
4976 | // meaningful 'next' value. | |
4977 | int64_t pool = -1; | |
4978 | shard_id_t shard; | |
4979 | { | |
4980 | spg_t pgid; | |
4981 | if (c.is_temp(&pgid)) { | |
4982 | pool = -2 - pgid.pool(); | |
4983 | shard = pgid.shard; | |
4984 | } else if (c.is_pg(&pgid)) { | |
4985 | pool = pgid.pool(); | |
4986 | shard = pgid.shard; | |
4987 | } else if (c.is_meta()) { | |
4988 | pool = -1; | |
4989 | shard = shard_id_t::NO_SHARD; | |
4990 | } else { | |
4991 | // hrm, the caller is test code! we should get kill it off. for now, | |
4992 | // tolerate it. | |
4993 | pool = 0; | |
4994 | shard = shard_id_t::NO_SHARD; | |
4995 | } | |
31f18b77 | 4996 | dout(20) << __FUNC__ << ": pool is " << pool << " shard is " << shard |
7c673cae FG |
4997 | << " pgid " << pgid << dendl; |
4998 | } | |
4999 | ghobject_t sep; | |
5000 | sep.hobj.pool = -1; | |
5001 | sep.set_shard(shard); | |
5002 | if (!c.is_temp() && !c.is_meta()) { | |
5003 | if (start < sep) { | |
31f18b77 | 5004 | dout(10) << __FUNC__ << ": first checking temp pool" << dendl; |
7c673cae FG |
5005 | coll_t temp = c.get_temp(); |
5006 | int r = collection_list(temp, start, end, max, ls, next); | |
5007 | if (r < 0) | |
5008 | return r; | |
5009 | if (*next != ghobject_t::get_max()) | |
5010 | return r; | |
5011 | start = sep; | |
31f18b77 | 5012 | dout(10) << __FUNC__ << ": fall through to non-temp collection, start " |
7c673cae FG |
5013 | << start << dendl; |
5014 | } else { | |
31f18b77 | 5015 | dout(10) << __FUNC__ << ": start " << start << " >= sep " << sep << dendl; |
7c673cae FG |
5016 | } |
5017 | } | |
5018 | ||
5019 | Index index; | |
5020 | int r = get_index(c, &index); | |
5021 | if (r < 0) | |
5022 | return r; | |
5023 | ||
5024 | assert(NULL != index.index); | |
5025 | RWLock::RLocker l((index.index)->access_lock); | |
5026 | ||
5027 | r = index->collection_list_partial(start, end, max, ls, next); | |
5028 | ||
5029 | if (r < 0) { | |
5030 | assert(!m_filestore_fail_eio || r != -EIO); | |
5031 | return r; | |
5032 | } | |
5033 | dout(20) << "objects: " << *ls << dendl; | |
5034 | ||
5035 | // HashIndex doesn't know the pool when constructing a 'next' value | |
5036 | if (next && !next->is_max()) { | |
5037 | next->hobj.pool = pool; | |
5038 | next->set_shard(shard); | |
5039 | dout(20) << " next " << *next << dendl; | |
5040 | } | |
5041 | ||
5042 | return 0; | |
5043 | } | |
5044 | ||
5045 | int FileStore::omap_get(const coll_t& _c, const ghobject_t &hoid, | |
5046 | bufferlist *header, | |
5047 | map<string, bufferlist> *out) | |
5048 | { | |
5049 | tracepoint(objectstore, omap_get_enter, _c.c_str()); | |
5050 | const coll_t& c = !_need_temp_object_collection(_c, hoid) ? _c : _c.get_temp(); | |
31f18b77 | 5051 | dout(15) << __FUNC__ << ": " << c << "/" << hoid << dendl; |
7c673cae FG |
5052 | Index index; |
5053 | int r = get_index(c, &index); | |
5054 | if (r < 0) | |
5055 | return r; | |
5056 | { | |
5057 | assert(NULL != index.index); | |
5058 | RWLock::RLocker l((index.index)->access_lock); | |
5059 | r = lfn_find(hoid, index); | |
5060 | if (r < 0) | |
5061 | return r; | |
5062 | } | |
5063 | r = object_map->get(hoid, header, out); | |
5064 | if (r < 0 && r != -ENOENT) { | |
5065 | assert(!m_filestore_fail_eio || r != -EIO); | |
5066 | return r; | |
5067 | } | |
5068 | tracepoint(objectstore, omap_get_exit, 0); | |
5069 | return 0; | |
5070 | } | |
5071 | ||
5072 | int FileStore::omap_get_header( | |
5073 | const coll_t& _c, | |
5074 | const ghobject_t &hoid, | |
5075 | bufferlist *bl, | |
5076 | bool allow_eio) | |
5077 | { | |
5078 | tracepoint(objectstore, omap_get_header_enter, _c.c_str()); | |
5079 | const coll_t& c = !_need_temp_object_collection(_c, hoid) ? _c : _c.get_temp(); | |
31f18b77 | 5080 | dout(15) << __FUNC__ << ": " << c << "/" << hoid << dendl; |
7c673cae FG |
5081 | Index index; |
5082 | int r = get_index(c, &index); | |
5083 | if (r < 0) | |
5084 | return r; | |
5085 | { | |
5086 | assert(NULL != index.index); | |
5087 | RWLock::RLocker l((index.index)->access_lock); | |
5088 | r = lfn_find(hoid, index); | |
5089 | if (r < 0) | |
5090 | return r; | |
5091 | } | |
5092 | r = object_map->get_header(hoid, bl); | |
5093 | if (r < 0 && r != -ENOENT) { | |
5094 | assert(allow_eio || !m_filestore_fail_eio || r != -EIO); | |
5095 | return r; | |
5096 | } | |
5097 | tracepoint(objectstore, omap_get_header_exit, 0); | |
5098 | return 0; | |
5099 | } | |
5100 | ||
5101 | int FileStore::omap_get_keys(const coll_t& _c, const ghobject_t &hoid, set<string> *keys) | |
5102 | { | |
5103 | tracepoint(objectstore, omap_get_keys_enter, _c.c_str()); | |
5104 | const coll_t& c = !_need_temp_object_collection(_c, hoid) ? _c : _c.get_temp(); | |
31f18b77 | 5105 | dout(15) << __FUNC__ << ": " << c << "/" << hoid << dendl; |
7c673cae FG |
5106 | Index index; |
5107 | int r = get_index(c, &index); | |
5108 | if (r < 0) | |
5109 | return r; | |
5110 | { | |
5111 | assert(NULL != index.index); | |
5112 | RWLock::RLocker l((index.index)->access_lock); | |
5113 | r = lfn_find(hoid, index); | |
5114 | if (r < 0) | |
5115 | return r; | |
5116 | } | |
5117 | r = object_map->get_keys(hoid, keys); | |
5118 | if (r < 0 && r != -ENOENT) { | |
5119 | assert(!m_filestore_fail_eio || r != -EIO); | |
5120 | return r; | |
5121 | } | |
5122 | tracepoint(objectstore, omap_get_keys_exit, 0); | |
5123 | return 0; | |
5124 | } | |
5125 | ||
5126 | int FileStore::omap_get_values(const coll_t& _c, const ghobject_t &hoid, | |
5127 | const set<string> &keys, | |
5128 | map<string, bufferlist> *out) | |
5129 | { | |
5130 | tracepoint(objectstore, omap_get_values_enter, _c.c_str()); | |
5131 | const coll_t& c = !_need_temp_object_collection(_c, hoid) ? _c : _c.get_temp(); | |
31f18b77 | 5132 | dout(15) << __FUNC__ << ": " << c << "/" << hoid << dendl; |
7c673cae FG |
5133 | Index index; |
5134 | const char *where = "()"; | |
5135 | int r = get_index(c, &index); | |
5136 | if (r < 0) { | |
5137 | where = " (get_index)"; | |
5138 | goto out; | |
5139 | } | |
5140 | { | |
5141 | assert(NULL != index.index); | |
5142 | RWLock::RLocker l((index.index)->access_lock); | |
5143 | r = lfn_find(hoid, index); | |
5144 | if (r < 0) { | |
5145 | where = " (lfn_find)"; | |
5146 | goto out; | |
5147 | } | |
5148 | } | |
5149 | r = object_map->get_values(hoid, keys, out); | |
5150 | if (r < 0 && r != -ENOENT) { | |
5151 | assert(!m_filestore_fail_eio || r != -EIO); | |
5152 | where = " (get_values)"; | |
5153 | goto out; | |
5154 | } | |
5155 | r = 0; | |
5156 | out: | |
5157 | tracepoint(objectstore, omap_get_values_exit, r); | |
31f18b77 | 5158 | dout(15) << __FUNC__ << ": " << c << "/" << hoid << " = " << r |
7c673cae FG |
5159 | << where << dendl; |
5160 | return r; | |
5161 | } | |
5162 | ||
5163 | int FileStore::omap_check_keys(const coll_t& _c, const ghobject_t &hoid, | |
5164 | const set<string> &keys, | |
5165 | set<string> *out) | |
5166 | { | |
5167 | tracepoint(objectstore, omap_check_keys_enter, _c.c_str()); | |
5168 | const coll_t& c = !_need_temp_object_collection(_c, hoid) ? _c : _c.get_temp(); | |
31f18b77 | 5169 | dout(15) << __FUNC__ << ": " << c << "/" << hoid << dendl; |
7c673cae FG |
5170 | |
5171 | Index index; | |
5172 | int r = get_index(c, &index); | |
5173 | if (r < 0) | |
5174 | return r; | |
5175 | { | |
5176 | assert(NULL != index.index); | |
5177 | RWLock::RLocker l((index.index)->access_lock); | |
5178 | r = lfn_find(hoid, index); | |
5179 | if (r < 0) | |
5180 | return r; | |
5181 | } | |
5182 | r = object_map->check_keys(hoid, keys, out); | |
5183 | if (r < 0 && r != -ENOENT) { | |
5184 | assert(!m_filestore_fail_eio || r != -EIO); | |
5185 | return r; | |
5186 | } | |
5187 | tracepoint(objectstore, omap_check_keys_exit, 0); | |
5188 | return 0; | |
5189 | } | |
5190 | ||
5191 | ObjectMap::ObjectMapIterator FileStore::get_omap_iterator(const coll_t& _c, | |
5192 | const ghobject_t &hoid) | |
5193 | { | |
5194 | tracepoint(objectstore, get_omap_iterator, _c.c_str()); | |
5195 | const coll_t& c = !_need_temp_object_collection(_c, hoid) ? _c : _c.get_temp(); | |
31f18b77 | 5196 | dout(15) << __FUNC__ << ": " << c << "/" << hoid << dendl; |
7c673cae FG |
5197 | Index index; |
5198 | int r = get_index(c, &index); | |
5199 | if (r < 0) { | |
31f18b77 | 5200 | dout(10) << __FUNC__ << ": " << c << "/" << hoid << " = 0 " |
7c673cae FG |
5201 | << "(get_index failed with " << cpp_strerror(r) << ")" << dendl; |
5202 | return ObjectMap::ObjectMapIterator(); | |
5203 | } | |
5204 | { | |
5205 | assert(NULL != index.index); | |
5206 | RWLock::RLocker l((index.index)->access_lock); | |
5207 | r = lfn_find(hoid, index); | |
5208 | if (r < 0) { | |
31f18b77 | 5209 | dout(10) << __FUNC__ << ": " << c << "/" << hoid << " = 0 " |
7c673cae FG |
5210 | << "(lfn_find failed with " << cpp_strerror(r) << ")" << dendl; |
5211 | return ObjectMap::ObjectMapIterator(); | |
5212 | } | |
5213 | } | |
5214 | return object_map->get_iterator(hoid); | |
5215 | } | |
5216 | ||
5217 | int FileStore::_collection_hint_expected_num_objs(const coll_t& c, uint32_t pg_num, | |
5218 | uint64_t expected_num_objs, | |
5219 | const SequencerPosition &spos) | |
5220 | { | |
31f18b77 | 5221 | dout(15) << __FUNC__ << ": collection: " << c << " pg number: " |
7c673cae FG |
5222 | << pg_num << " expected number of objects: " << expected_num_objs << dendl; |
5223 | ||
5224 | bool empty; | |
5225 | int ret = collection_empty(c, &empty); | |
5226 | if (ret < 0) | |
5227 | return ret; | |
5228 | if (!empty && !replaying) { | |
5229 | dout(0) << "Failed to give an expected number of objects hint to collection : " | |
5230 | << c << ", only empty collection can take such type of hint. " << dendl; | |
5231 | return 0; | |
5232 | } | |
5233 | ||
5234 | Index index; | |
5235 | ret = get_index(c, &index); | |
5236 | if (ret < 0) | |
5237 | return ret; | |
5238 | // Pre-hash the collection | |
5239 | ret = index->pre_hash_collection(pg_num, expected_num_objs); | |
5240 | dout(10) << "pre_hash_collection " << c << " = " << ret << dendl; | |
5241 | if (ret < 0) | |
5242 | return ret; | |
5243 | _set_replay_guard(c, spos); | |
5244 | ||
5245 | return 0; | |
5246 | } | |
5247 | ||
5248 | int FileStore::_create_collection( | |
5249 | const coll_t& c, | |
5250 | int bits, | |
5251 | const SequencerPosition &spos) | |
5252 | { | |
5253 | char fn[PATH_MAX]; | |
5254 | get_cdir(c, fn, sizeof(fn)); | |
31f18b77 | 5255 | dout(15) << __FUNC__ << ": " << fn << dendl; |
7c673cae FG |
5256 | int r = ::mkdir(fn, 0755); |
5257 | if (r < 0) | |
5258 | r = -errno; | |
5259 | if (r == -EEXIST && replaying) | |
5260 | r = 0; | |
31f18b77 | 5261 | dout(10) << __FUNC__ << ": " << fn << " = " << r << dendl; |
7c673cae FG |
5262 | |
5263 | if (r < 0) | |
5264 | return r; | |
5265 | r = init_index(c); | |
5266 | if (r < 0) | |
5267 | return r; | |
5268 | r = _collection_set_bits(c, bits); | |
5269 | if (r < 0) | |
5270 | return r; | |
5271 | // create parallel temp collection, too | |
5272 | if (!c.is_meta() && !c.is_temp()) { | |
5273 | coll_t temp = c.get_temp(); | |
5274 | r = _create_collection(temp, 0, spos); | |
5275 | if (r < 0) | |
5276 | return r; | |
5277 | } | |
5278 | ||
5279 | _set_replay_guard(c, spos); | |
5280 | return 0; | |
5281 | } | |
5282 | ||
5283 | int FileStore::_destroy_collection(const coll_t& c) | |
5284 | { | |
5285 | int r = 0; | |
5286 | char fn[PATH_MAX]; | |
5287 | get_cdir(c, fn, sizeof(fn)); | |
31f18b77 | 5288 | dout(15) << __FUNC__ << ": " << fn << dendl; |
7c673cae FG |
5289 | { |
5290 | Index from; | |
5291 | r = get_index(c, &from); | |
5292 | if (r < 0) | |
5293 | goto out; | |
5294 | assert(NULL != from.index); | |
5295 | RWLock::WLocker l((from.index)->access_lock); | |
5296 | ||
5297 | r = from->prep_delete(); | |
5298 | if (r < 0) | |
5299 | goto out; | |
5300 | } | |
5301 | r = ::rmdir(fn); | |
5302 | if (r < 0) { | |
5303 | r = -errno; | |
5304 | goto out; | |
5305 | } | |
5306 | ||
5307 | out: | |
5308 | // destroy parallel temp collection, too | |
5309 | if (!c.is_meta() && !c.is_temp()) { | |
5310 | coll_t temp = c.get_temp(); | |
5311 | int r2 = _destroy_collection(temp); | |
5312 | if (r2 < 0) { | |
5313 | r = r2; | |
5314 | goto out_final; | |
5315 | } | |
5316 | } | |
5317 | ||
5318 | out_final: | |
31f18b77 | 5319 | dout(10) << __FUNC__ << ": " << fn << " = " << r << dendl; |
7c673cae FG |
5320 | return r; |
5321 | } | |
5322 | ||
5323 | ||
5324 | int FileStore::_collection_add(const coll_t& c, const coll_t& oldcid, const ghobject_t& o, | |
5325 | const SequencerPosition& spos) | |
5326 | { | |
31f18b77 | 5327 | dout(15) << __FUNC__ << ": " << c << "/" << o << " from " << oldcid << "/" << o << dendl; |
7c673cae FG |
5328 | |
5329 | int dstcmp = _check_replay_guard(c, o, spos); | |
5330 | if (dstcmp < 0) | |
5331 | return 0; | |
5332 | ||
5333 | // check the src name too; it might have a newer guard, and we don't | |
5334 | // want to clobber it | |
5335 | int srccmp = _check_replay_guard(oldcid, o, spos); | |
5336 | if (srccmp < 0) | |
5337 | return 0; | |
5338 | ||
5339 | // open guard on object so we don't any previous operations on the | |
5340 | // new name that will modify the source inode. | |
5341 | FDRef fd; | |
5342 | int r = lfn_open(oldcid, o, 0, &fd); | |
5343 | if (r < 0) { | |
5344 | // the source collection/object does not exist. If we are replaying, we | |
5345 | // should be safe, so just return 0 and move on. | |
5346 | assert(replaying); | |
31f18b77 | 5347 | dout(10) << __FUNC__ << ": " << c << "/" << o << " from " |
7c673cae FG |
5348 | << oldcid << "/" << o << " (dne, continue replay) " << dendl; |
5349 | return 0; | |
5350 | } | |
5351 | if (dstcmp > 0) { // if dstcmp == 0 the guard already says "in-progress" | |
5352 | _set_replay_guard(**fd, spos, &o, true); | |
5353 | } | |
5354 | ||
5355 | r = lfn_link(oldcid, c, o, o); | |
5356 | if (replaying && !backend->can_checkpoint() && | |
5357 | r == -EEXIST) // crashed between link() and set_replay_guard() | |
5358 | r = 0; | |
5359 | ||
5360 | _inject_failure(); | |
5361 | ||
5362 | // close guard on object so we don't do this again | |
5363 | if (r == 0) { | |
5364 | _close_replay_guard(**fd, spos); | |
5365 | } | |
5366 | lfn_close(fd); | |
5367 | ||
31f18b77 | 5368 | dout(10) << __FUNC__ << ": " << c << "/" << o << " from " << oldcid << "/" << o << " = " << r << dendl; |
7c673cae FG |
5369 | return r; |
5370 | } | |
5371 | ||
5372 | int FileStore::_collection_move_rename(const coll_t& oldcid, const ghobject_t& oldoid, | |
5373 | coll_t c, const ghobject_t& o, | |
5374 | const SequencerPosition& spos, | |
5375 | bool allow_enoent) | |
5376 | { | |
31f18b77 | 5377 | dout(15) << __FUNC__ << ": " << c << "/" << o << " from " << oldcid << "/" << oldoid << dendl; |
7c673cae FG |
5378 | int r = 0; |
5379 | int dstcmp, srccmp; | |
5380 | ||
5381 | if (replaying) { | |
5382 | /* If the destination collection doesn't exist during replay, | |
5383 | * we need to delete the src object and continue on | |
5384 | */ | |
5385 | if (!collection_exists(c)) | |
5386 | goto out_rm_src; | |
5387 | } | |
5388 | ||
5389 | dstcmp = _check_replay_guard(c, o, spos); | |
5390 | if (dstcmp < 0) | |
5391 | goto out_rm_src; | |
5392 | ||
5393 | // check the src name too; it might have a newer guard, and we don't | |
5394 | // want to clobber it | |
5395 | srccmp = _check_replay_guard(oldcid, oldoid, spos); | |
5396 | if (srccmp < 0) | |
5397 | return 0; | |
5398 | ||
5399 | { | |
5400 | // open guard on object so we don't any previous operations on the | |
5401 | // new name that will modify the source inode. | |
5402 | FDRef fd; | |
5403 | r = lfn_open(oldcid, oldoid, 0, &fd); | |
5404 | if (r < 0) { | |
5405 | // the source collection/object does not exist. If we are replaying, we | |
5406 | // should be safe, so just return 0 and move on. | |
5407 | if (replaying) { | |
31f18b77 | 5408 | dout(10) << __FUNC__ << ": " << c << "/" << o << " from " |
7c673cae FG |
5409 | << oldcid << "/" << oldoid << " (dne, continue replay) " << dendl; |
5410 | } else if (allow_enoent) { | |
31f18b77 | 5411 | dout(10) << __FUNC__ << ": " << c << "/" << o << " from " |
7c673cae FG |
5412 | << oldcid << "/" << oldoid << " (dne, ignoring enoent)" |
5413 | << dendl; | |
5414 | } else { | |
5415 | assert(0 == "ERROR: source must exist"); | |
5416 | } | |
5417 | ||
5418 | if (!replaying) { | |
5419 | return 0; | |
5420 | } | |
5421 | if (allow_enoent && dstcmp > 0) { // if dstcmp == 0, try_rename was started. | |
5422 | return 0; | |
5423 | } | |
5424 | ||
5425 | r = 0; // don't know if object_map was cloned | |
5426 | } else { | |
5427 | if (dstcmp > 0) { // if dstcmp == 0 the guard already says "in-progress" | |
5428 | _set_replay_guard(**fd, spos, &o, true); | |
5429 | } | |
5430 | ||
5431 | r = lfn_link(oldcid, c, oldoid, o); | |
5432 | if (replaying && !backend->can_checkpoint() && | |
5433 | r == -EEXIST) // crashed between link() and set_replay_guard() | |
5434 | r = 0; | |
5435 | ||
5436 | lfn_close(fd); | |
5437 | fd = FDRef(); | |
5438 | ||
5439 | _inject_failure(); | |
5440 | } | |
5441 | ||
5442 | if (r == 0) { | |
5443 | // the name changed; link the omap content | |
5444 | r = object_map->rename(oldoid, o, &spos); | |
5445 | if (r == -ENOENT) | |
5446 | r = 0; | |
5447 | } | |
5448 | ||
5449 | _inject_failure(); | |
5450 | ||
5451 | if (r == 0) | |
5452 | r = lfn_unlink(oldcid, oldoid, spos, true); | |
5453 | ||
5454 | if (r == 0) | |
5455 | r = lfn_open(c, o, 0, &fd); | |
5456 | ||
5457 | // close guard on object so we don't do this again | |
5458 | if (r == 0) { | |
5459 | _close_replay_guard(**fd, spos, &o); | |
5460 | lfn_close(fd); | |
5461 | } | |
5462 | } | |
5463 | ||
31f18b77 | 5464 | dout(10) << __FUNC__ << ": " << c << "/" << o << " from " << oldcid << "/" << oldoid |
7c673cae FG |
5465 | << " = " << r << dendl; |
5466 | return r; | |
5467 | ||
5468 | out_rm_src: | |
5469 | // remove source | |
5470 | if (_check_replay_guard(oldcid, oldoid, spos) > 0) { | |
5471 | r = lfn_unlink(oldcid, oldoid, spos, true); | |
5472 | } | |
5473 | ||
31f18b77 | 5474 | dout(10) << __FUNC__ << ": " << c << "/" << o << " from " << oldcid << "/" << oldoid |
7c673cae FG |
5475 | << " = " << r << dendl; |
5476 | return r; | |
5477 | } | |
5478 | ||
5479 | void FileStore::_inject_failure() | |
5480 | { | |
31f18b77 FG |
5481 | if (m_filestore_kill_at) { |
5482 | int final = --m_filestore_kill_at; | |
5483 | dout(5) << __FUNC__ << ": " << (final+1) << " -> " << final << dendl; | |
7c673cae | 5484 | if (final == 0) { |
31f18b77 | 5485 | derr << __FUNC__ << ": KILLING" << dendl; |
7c673cae FG |
5486 | cct->_log->flush(); |
5487 | _exit(1); | |
5488 | } | |
5489 | } | |
5490 | } | |
5491 | ||
5492 | int FileStore::_omap_clear(const coll_t& cid, const ghobject_t &hoid, | |
5493 | const SequencerPosition &spos) { | |
31f18b77 | 5494 | dout(15) << __FUNC__ << ": " << cid << "/" << hoid << dendl; |
7c673cae FG |
5495 | Index index; |
5496 | int r = get_index(cid, &index); | |
5497 | if (r < 0) | |
5498 | return r; | |
5499 | { | |
5500 | assert(NULL != index.index); | |
5501 | RWLock::RLocker l((index.index)->access_lock); | |
5502 | r = lfn_find(hoid, index); | |
5503 | if (r < 0) | |
5504 | return r; | |
5505 | } | |
5506 | r = object_map->clear_keys_header(hoid, &spos); | |
5507 | if (r < 0 && r != -ENOENT) | |
5508 | return r; | |
5509 | return 0; | |
5510 | } | |
5511 | ||
5512 | int FileStore::_omap_setkeys(const coll_t& cid, const ghobject_t &hoid, | |
5513 | const map<string, bufferlist> &aset, | |
5514 | const SequencerPosition &spos) { | |
31f18b77 | 5515 | dout(15) << __FUNC__ << ": " << cid << "/" << hoid << dendl; |
7c673cae FG |
5516 | Index index; |
5517 | int r; | |
5518 | //treat pgmeta as a logical object, skip to check exist | |
5519 | if (hoid.is_pgmeta()) | |
5520 | goto skip; | |
5521 | ||
5522 | r = get_index(cid, &index); | |
5523 | if (r < 0) { | |
31f18b77 | 5524 | dout(20) << __FUNC__ << ": get_index got " << cpp_strerror(r) << dendl; |
7c673cae FG |
5525 | return r; |
5526 | } | |
5527 | { | |
5528 | assert(NULL != index.index); | |
5529 | RWLock::RLocker l((index.index)->access_lock); | |
5530 | r = lfn_find(hoid, index); | |
5531 | if (r < 0) { | |
31f18b77 | 5532 | dout(20) << __FUNC__ << ": lfn_find got " << cpp_strerror(r) << dendl; |
7c673cae FG |
5533 | return r; |
5534 | } | |
5535 | } | |
5536 | skip: | |
5537 | if (g_conf->subsys.should_gather(ceph_subsys_filestore, 20)) { | |
5538 | for (auto& p : aset) { | |
31f18b77 | 5539 | dout(20) << __FUNC__ << ": set " << p.first << dendl; |
7c673cae FG |
5540 | } |
5541 | } | |
5542 | r = object_map->set_keys(hoid, aset, &spos); | |
31f18b77 | 5543 | dout(20) << __FUNC__ << ": " << cid << "/" << hoid << " = " << r << dendl; |
7c673cae FG |
5544 | return r; |
5545 | } | |
5546 | ||
5547 | int FileStore::_omap_rmkeys(const coll_t& cid, const ghobject_t &hoid, | |
5548 | const set<string> &keys, | |
5549 | const SequencerPosition &spos) { | |
31f18b77 | 5550 | dout(15) << __FUNC__ << ": " << cid << "/" << hoid << dendl; |
7c673cae FG |
5551 | Index index; |
5552 | int r; | |
5553 | //treat pgmeta as a logical object, skip to check exist | |
5554 | if (hoid.is_pgmeta()) | |
5555 | goto skip; | |
5556 | ||
5557 | r = get_index(cid, &index); | |
5558 | if (r < 0) | |
5559 | return r; | |
5560 | { | |
5561 | assert(NULL != index.index); | |
5562 | RWLock::RLocker l((index.index)->access_lock); | |
5563 | r = lfn_find(hoid, index); | |
5564 | if (r < 0) | |
5565 | return r; | |
5566 | } | |
5567 | skip: | |
5568 | r = object_map->rm_keys(hoid, keys, &spos); | |
5569 | if (r < 0 && r != -ENOENT) | |
5570 | return r; | |
5571 | return 0; | |
5572 | } | |
5573 | ||
5574 | int FileStore::_omap_rmkeyrange(const coll_t& cid, const ghobject_t &hoid, | |
5575 | const string& first, const string& last, | |
5576 | const SequencerPosition &spos) { | |
31f18b77 | 5577 | dout(15) << __FUNC__ << ": " << cid << "/" << hoid << " [" << first << "," << last << "]" << dendl; |
7c673cae FG |
5578 | set<string> keys; |
5579 | { | |
5580 | ObjectMap::ObjectMapIterator iter = get_omap_iterator(cid, hoid); | |
5581 | if (!iter) | |
5582 | return -ENOENT; | |
5583 | for (iter->lower_bound(first); iter->valid() && iter->key() < last; | |
5584 | iter->next()) { | |
5585 | keys.insert(iter->key()); | |
5586 | } | |
5587 | } | |
5588 | return _omap_rmkeys(cid, hoid, keys, spos); | |
5589 | } | |
5590 | ||
5591 | int FileStore::_omap_setheader(const coll_t& cid, const ghobject_t &hoid, | |
5592 | const bufferlist &bl, | |
5593 | const SequencerPosition &spos) | |
5594 | { | |
31f18b77 | 5595 | dout(15) << __FUNC__ << ": " << cid << "/" << hoid << dendl; |
7c673cae FG |
5596 | Index index; |
5597 | int r = get_index(cid, &index); | |
5598 | if (r < 0) | |
5599 | return r; | |
5600 | { | |
5601 | assert(NULL != index.index); | |
5602 | RWLock::RLocker l((index.index)->access_lock); | |
5603 | r = lfn_find(hoid, index); | |
5604 | if (r < 0) | |
5605 | return r; | |
5606 | } | |
5607 | return object_map->set_header(hoid, bl, &spos); | |
5608 | } | |
5609 | ||
5610 | int FileStore::_split_collection(const coll_t& cid, | |
5611 | uint32_t bits, | |
5612 | uint32_t rem, | |
5613 | coll_t dest, | |
5614 | const SequencerPosition &spos) | |
5615 | { | |
5616 | int r; | |
5617 | { | |
31f18b77 | 5618 | dout(15) << __FUNC__ << ": " << cid << " bits: " << bits << dendl; |
7c673cae | 5619 | if (!collection_exists(cid)) { |
31f18b77 | 5620 | dout(2) << __FUNC__ << ": " << cid << " DNE" << dendl; |
7c673cae FG |
5621 | assert(replaying); |
5622 | return 0; | |
5623 | } | |
5624 | if (!collection_exists(dest)) { | |
31f18b77 | 5625 | dout(2) << __FUNC__ << ": " << dest << " DNE" << dendl; |
7c673cae FG |
5626 | assert(replaying); |
5627 | return 0; | |
5628 | } | |
5629 | ||
5630 | int dstcmp = _check_replay_guard(dest, spos); | |
5631 | if (dstcmp < 0) | |
5632 | return 0; | |
5633 | ||
5634 | int srccmp = _check_replay_guard(cid, spos); | |
5635 | if (srccmp < 0) | |
5636 | return 0; | |
5637 | ||
5638 | _set_global_replay_guard(cid, spos); | |
5639 | _set_replay_guard(cid, spos, true); | |
5640 | _set_replay_guard(dest, spos, true); | |
5641 | ||
5642 | Index from; | |
5643 | r = get_index(cid, &from); | |
5644 | ||
5645 | Index to; | |
5646 | if (!r) | |
5647 | r = get_index(dest, &to); | |
5648 | ||
5649 | if (!r) { | |
5650 | assert(NULL != from.index); | |
5651 | RWLock::WLocker l1((from.index)->access_lock); | |
5652 | ||
5653 | assert(NULL != to.index); | |
5654 | RWLock::WLocker l2((to.index)->access_lock); | |
5655 | ||
5656 | r = from->split(rem, bits, to.index); | |
5657 | } | |
5658 | ||
5659 | _close_replay_guard(cid, spos); | |
5660 | _close_replay_guard(dest, spos); | |
5661 | } | |
5662 | _collection_set_bits(cid, bits); | |
5663 | if (!r && cct->_conf->filestore_debug_verify_split) { | |
5664 | vector<ghobject_t> objects; | |
5665 | ghobject_t next; | |
5666 | while (1) { | |
5667 | collection_list( | |
5668 | cid, | |
5669 | next, ghobject_t::get_max(), | |
5670 | get_ideal_list_max(), | |
5671 | &objects, | |
5672 | &next); | |
5673 | if (objects.empty()) | |
5674 | break; | |
5675 | for (vector<ghobject_t>::iterator i = objects.begin(); | |
5676 | i != objects.end(); | |
5677 | ++i) { | |
31f18b77 | 5678 | dout(20) << __FUNC__ << ": " << *i << " still in source " |
7c673cae FG |
5679 | << cid << dendl; |
5680 | assert(!i->match(bits, rem)); | |
5681 | } | |
5682 | objects.clear(); | |
5683 | } | |
5684 | next = ghobject_t(); | |
5685 | while (1) { | |
5686 | collection_list( | |
5687 | dest, | |
5688 | next, ghobject_t::get_max(), | |
5689 | get_ideal_list_max(), | |
5690 | &objects, | |
5691 | &next); | |
5692 | if (objects.empty()) | |
5693 | break; | |
5694 | for (vector<ghobject_t>::iterator i = objects.begin(); | |
5695 | i != objects.end(); | |
5696 | ++i) { | |
31f18b77 | 5697 | dout(20) << __FUNC__ << ": " << *i << " now in dest " |
7c673cae FG |
5698 | << *i << dendl; |
5699 | assert(i->match(bits, rem)); | |
5700 | } | |
5701 | objects.clear(); | |
5702 | } | |
5703 | } | |
5704 | return r; | |
5705 | } | |
5706 | ||
5707 | int FileStore::_set_alloc_hint(const coll_t& cid, const ghobject_t& oid, | |
5708 | uint64_t expected_object_size, | |
5709 | uint64_t expected_write_size) | |
5710 | { | |
31f18b77 | 5711 | dout(15) << __FUNC__ << ": " << cid << "/" << oid << " object_size " << expected_object_size << " write_size " << expected_write_size << dendl; |
7c673cae FG |
5712 | |
5713 | FDRef fd; | |
5714 | int ret = 0; | |
5715 | ||
5716 | if (expected_object_size == 0 || expected_write_size == 0) | |
5717 | goto out; | |
5718 | ||
5719 | ret = lfn_open(cid, oid, false, &fd); | |
5720 | if (ret < 0) | |
5721 | goto out; | |
5722 | ||
5723 | { | |
5724 | // TODO: a more elaborate hint calculation | |
5725 | uint64_t hint = MIN(expected_write_size, m_filestore_max_alloc_hint_size); | |
5726 | ||
5727 | ret = backend->set_alloc_hint(**fd, hint); | |
31f18b77 | 5728 | dout(20) << __FUNC__ << ": hint " << hint << " ret " << ret << dendl; |
7c673cae FG |
5729 | } |
5730 | ||
5731 | lfn_close(fd); | |
5732 | out: | |
31f18b77 | 5733 | dout(10) << __FUNC__ << ": " << cid << "/" << oid << " object_size " << expected_object_size << " write_size " << expected_write_size << " = " << ret << dendl; |
7c673cae FG |
5734 | assert(!m_filestore_fail_eio || ret != -EIO); |
5735 | return ret; | |
5736 | } | |
5737 | ||
5738 | const char** FileStore::get_tracked_conf_keys() const | |
5739 | { | |
5740 | static const char* KEYS[] = { | |
5741 | "filestore_max_inline_xattr_size", | |
5742 | "filestore_max_inline_xattr_size_xfs", | |
5743 | "filestore_max_inline_xattr_size_btrfs", | |
5744 | "filestore_max_inline_xattr_size_other", | |
5745 | "filestore_max_inline_xattrs", | |
5746 | "filestore_max_inline_xattrs_xfs", | |
5747 | "filestore_max_inline_xattrs_btrfs", | |
5748 | "filestore_max_inline_xattrs_other", | |
5749 | "filestore_max_xattr_value_size", | |
5750 | "filestore_max_xattr_value_size_xfs", | |
5751 | "filestore_max_xattr_value_size_btrfs", | |
5752 | "filestore_max_xattr_value_size_other", | |
5753 | "filestore_min_sync_interval", | |
5754 | "filestore_max_sync_interval", | |
5755 | "filestore_queue_max_ops", | |
5756 | "filestore_queue_max_bytes", | |
5757 | "filestore_expected_throughput_bytes", | |
5758 | "filestore_expected_throughput_ops", | |
5759 | "filestore_queue_low_threshhold", | |
5760 | "filestore_queue_high_threshhold", | |
5761 | "filestore_queue_high_delay_multiple", | |
5762 | "filestore_queue_max_delay_multiple", | |
5763 | "filestore_commit_timeout", | |
5764 | "filestore_dump_file", | |
5765 | "filestore_kill_at", | |
5766 | "filestore_fail_eio", | |
5767 | "filestore_fadvise", | |
5768 | "filestore_sloppy_crc", | |
5769 | "filestore_sloppy_crc_block_size", | |
5770 | "filestore_max_alloc_hint_size", | |
5771 | NULL | |
5772 | }; | |
5773 | return KEYS; | |
5774 | } | |
5775 | ||
5776 | void FileStore::handle_conf_change(const struct md_config_t *conf, | |
5777 | const std::set <std::string> &changed) | |
5778 | { | |
5779 | if (changed.count("filestore_max_inline_xattr_size") || | |
5780 | changed.count("filestore_max_inline_xattr_size_xfs") || | |
5781 | changed.count("filestore_max_inline_xattr_size_btrfs") || | |
5782 | changed.count("filestore_max_inline_xattr_size_other") || | |
5783 | changed.count("filestore_max_inline_xattrs") || | |
5784 | changed.count("filestore_max_inline_xattrs_xfs") || | |
5785 | changed.count("filestore_max_inline_xattrs_btrfs") || | |
5786 | changed.count("filestore_max_inline_xattrs_other") || | |
5787 | changed.count("filestore_max_xattr_value_size") || | |
5788 | changed.count("filestore_max_xattr_value_size_xfs") || | |
5789 | changed.count("filestore_max_xattr_value_size_btrfs") || | |
5790 | changed.count("filestore_max_xattr_value_size_other")) { | |
5791 | if (backend) { | |
5792 | Mutex::Locker l(lock); | |
5793 | set_xattr_limits_via_conf(); | |
5794 | } | |
5795 | } | |
5796 | ||
5797 | if (changed.count("filestore_queue_max_bytes") || | |
5798 | changed.count("filestore_queue_max_ops") || | |
5799 | changed.count("filestore_expected_throughput_bytes") || | |
5800 | changed.count("filestore_expected_throughput_ops") || | |
5801 | changed.count("filestore_queue_low_threshhold") || | |
5802 | changed.count("filestore_queue_high_threshhold") || | |
5803 | changed.count("filestore_queue_high_delay_multiple") || | |
5804 | changed.count("filestore_queue_max_delay_multiple")) { | |
5805 | Mutex::Locker l(lock); | |
5806 | set_throttle_params(); | |
5807 | } | |
5808 | ||
5809 | if (changed.count("filestore_min_sync_interval") || | |
5810 | changed.count("filestore_max_sync_interval") || | |
5811 | changed.count("filestore_kill_at") || | |
5812 | changed.count("filestore_fail_eio") || | |
5813 | changed.count("filestore_sloppy_crc") || | |
5814 | changed.count("filestore_sloppy_crc_block_size") || | |
5815 | changed.count("filestore_max_alloc_hint_size") || | |
5816 | changed.count("filestore_fadvise")) { | |
5817 | Mutex::Locker l(lock); | |
5818 | m_filestore_min_sync_interval = conf->filestore_min_sync_interval; | |
5819 | m_filestore_max_sync_interval = conf->filestore_max_sync_interval; | |
31f18b77 | 5820 | m_filestore_kill_at = conf->filestore_kill_at; |
7c673cae FG |
5821 | m_filestore_fail_eio = conf->filestore_fail_eio; |
5822 | m_filestore_fadvise = conf->filestore_fadvise; | |
5823 | m_filestore_sloppy_crc = conf->filestore_sloppy_crc; | |
5824 | m_filestore_sloppy_crc_block_size = conf->filestore_sloppy_crc_block_size; | |
5825 | m_filestore_max_alloc_hint_size = conf->filestore_max_alloc_hint_size; | |
5826 | } | |
5827 | if (changed.count("filestore_commit_timeout")) { | |
5828 | Mutex::Locker l(sync_entry_timeo_lock); | |
5829 | m_filestore_commit_timeout = conf->filestore_commit_timeout; | |
5830 | } | |
5831 | if (changed.count("filestore_dump_file")) { | |
5832 | if (conf->filestore_dump_file.length() && | |
5833 | conf->filestore_dump_file != "-") { | |
5834 | dump_start(conf->filestore_dump_file); | |
5835 | } else { | |
5836 | dump_stop(); | |
5837 | } | |
5838 | } | |
5839 | } | |
5840 | ||
5841 | int FileStore::set_throttle_params() | |
5842 | { | |
5843 | stringstream ss; | |
5844 | bool valid = throttle_bytes.set_params( | |
5845 | cct->_conf->filestore_queue_low_threshhold, | |
5846 | cct->_conf->filestore_queue_high_threshhold, | |
5847 | cct->_conf->filestore_expected_throughput_bytes, | |
5848 | cct->_conf->filestore_queue_high_delay_multiple, | |
5849 | cct->_conf->filestore_queue_max_delay_multiple, | |
5850 | cct->_conf->filestore_queue_max_bytes, | |
5851 | &ss); | |
5852 | ||
5853 | valid &= throttle_ops.set_params( | |
5854 | cct->_conf->filestore_queue_low_threshhold, | |
5855 | cct->_conf->filestore_queue_high_threshhold, | |
5856 | cct->_conf->filestore_expected_throughput_ops, | |
5857 | cct->_conf->filestore_queue_high_delay_multiple, | |
5858 | cct->_conf->filestore_queue_max_delay_multiple, | |
5859 | cct->_conf->filestore_queue_max_ops, | |
5860 | &ss); | |
5861 | ||
5862 | logger->set(l_filestore_op_queue_max_ops, throttle_ops.get_max()); | |
5863 | logger->set(l_filestore_op_queue_max_bytes, throttle_bytes.get_max()); | |
5864 | ||
5865 | if (!valid) { | |
5866 | derr << "tried to set invalid params: " | |
5867 | << ss.str() | |
5868 | << dendl; | |
5869 | } | |
5870 | return valid ? 0 : -EINVAL; | |
5871 | } | |
5872 | ||
5873 | void FileStore::dump_start(const std::string& file) | |
5874 | { | |
31f18b77 | 5875 | dout(10) << __FUNC__ << ": " << file << dendl; |
7c673cae FG |
5876 | if (m_filestore_do_dump) { |
5877 | dump_stop(); | |
5878 | } | |
5879 | m_filestore_dump_fmt.reset(); | |
5880 | m_filestore_dump_fmt.open_array_section("dump"); | |
5881 | m_filestore_dump.open(file.c_str()); | |
5882 | m_filestore_do_dump = true; | |
5883 | } | |
5884 | ||
5885 | void FileStore::dump_stop() | |
5886 | { | |
31f18b77 | 5887 | dout(10) << __FUNC__ << dendl; |
7c673cae FG |
5888 | m_filestore_do_dump = false; |
5889 | if (m_filestore_dump.is_open()) { | |
5890 | m_filestore_dump_fmt.close_section(); | |
5891 | m_filestore_dump_fmt.flush(m_filestore_dump); | |
5892 | m_filestore_dump.flush(); | |
5893 | m_filestore_dump.close(); | |
5894 | } | |
5895 | } | |
5896 | ||
5897 | void FileStore::dump_transactions(vector<ObjectStore::Transaction>& ls, uint64_t seq, OpSequencer *osr) | |
5898 | { | |
5899 | m_filestore_dump_fmt.open_array_section("transactions"); | |
5900 | unsigned trans_num = 0; | |
5901 | for (vector<ObjectStore::Transaction>::iterator i = ls.begin(); i != ls.end(); ++i, ++trans_num) { | |
5902 | m_filestore_dump_fmt.open_object_section("transaction"); | |
5903 | m_filestore_dump_fmt.dump_string("osr", osr->get_name()); | |
5904 | m_filestore_dump_fmt.dump_unsigned("seq", seq); | |
5905 | m_filestore_dump_fmt.dump_unsigned("trans_num", trans_num); | |
5906 | (*i).dump(&m_filestore_dump_fmt); | |
5907 | m_filestore_dump_fmt.close_section(); | |
5908 | } | |
5909 | m_filestore_dump_fmt.close_section(); | |
5910 | m_filestore_dump_fmt.flush(m_filestore_dump); | |
5911 | m_filestore_dump.flush(); | |
5912 | } | |
5913 | ||
5914 | void FileStore::set_xattr_limits_via_conf() | |
5915 | { | |
5916 | uint32_t fs_xattr_size; | |
5917 | uint32_t fs_xattrs; | |
5918 | uint32_t fs_xattr_max_value_size; | |
5919 | ||
5920 | switch (m_fs_type) { | |
5921 | #if defined(__linux__) | |
5922 | case XFS_SUPER_MAGIC: | |
5923 | fs_xattr_size = cct->_conf->filestore_max_inline_xattr_size_xfs; | |
5924 | fs_xattrs = cct->_conf->filestore_max_inline_xattrs_xfs; | |
5925 | fs_xattr_max_value_size = cct->_conf->filestore_max_xattr_value_size_xfs; | |
5926 | break; | |
5927 | case BTRFS_SUPER_MAGIC: | |
5928 | fs_xattr_size = cct->_conf->filestore_max_inline_xattr_size_btrfs; | |
5929 | fs_xattrs = cct->_conf->filestore_max_inline_xattrs_btrfs; | |
5930 | fs_xattr_max_value_size = cct->_conf->filestore_max_xattr_value_size_btrfs; | |
5931 | break; | |
5932 | #endif | |
5933 | default: | |
5934 | fs_xattr_size = cct->_conf->filestore_max_inline_xattr_size_other; | |
5935 | fs_xattrs = cct->_conf->filestore_max_inline_xattrs_other; | |
5936 | fs_xattr_max_value_size = cct->_conf->filestore_max_xattr_value_size_other; | |
5937 | break; | |
5938 | } | |
5939 | ||
5940 | // Use override value if set | |
5941 | if (cct->_conf->filestore_max_inline_xattr_size) | |
5942 | m_filestore_max_inline_xattr_size = cct->_conf->filestore_max_inline_xattr_size; | |
5943 | else | |
5944 | m_filestore_max_inline_xattr_size = fs_xattr_size; | |
5945 | ||
5946 | // Use override value if set | |
5947 | if (cct->_conf->filestore_max_inline_xattrs) | |
5948 | m_filestore_max_inline_xattrs = cct->_conf->filestore_max_inline_xattrs; | |
5949 | else | |
5950 | m_filestore_max_inline_xattrs = fs_xattrs; | |
5951 | ||
5952 | // Use override value if set | |
5953 | if (cct->_conf->filestore_max_xattr_value_size) | |
5954 | m_filestore_max_xattr_value_size = cct->_conf->filestore_max_xattr_value_size; | |
5955 | else | |
5956 | m_filestore_max_xattr_value_size = fs_xattr_max_value_size; | |
5957 | ||
5958 | if (m_filestore_max_xattr_value_size < cct->_conf->osd_max_object_name_len) { | |
5959 | derr << "WARNING: max attr value size (" | |
5960 | << m_filestore_max_xattr_value_size | |
5961 | << ") is smaller than osd_max_object_name_len (" | |
5962 | << cct->_conf->osd_max_object_name_len | |
5963 | << "). Your backend filesystem appears to not support attrs large " | |
5964 | << "enough to handle the configured max rados name size. You may get " | |
5965 | << "unexpected ENAMETOOLONG errors on rados operations or buggy " | |
5966 | << "behavior" | |
5967 | << dendl; | |
5968 | } | |
5969 | } | |
5970 | ||
5971 | uint64_t FileStore::estimate_objects_overhead(uint64_t num_objects) | |
5972 | { | |
5973 | uint64_t res = num_objects * blk_size / 2; //assumes that each object uses ( in average ) additional 1/2 block due to FS allocation granularity. | |
5974 | return res; | |
5975 | } | |
5976 | ||
5977 | int FileStore::apply_layout_settings(const coll_t &cid) | |
5978 | { | |
31f18b77 | 5979 | dout(20) << __FUNC__ << ": " << cid << dendl; |
7c673cae FG |
5980 | Index index; |
5981 | int r = get_index(cid, &index); | |
5982 | if (r < 0) { | |
5983 | dout(10) << "Error getting index for " << cid << ": " << cpp_strerror(r) | |
5984 | << dendl; | |
5985 | return r; | |
5986 | } | |
5987 | ||
5988 | return index->apply_layout_settings(); | |
5989 | } | |
5990 | ||
5991 | ||
5992 | // -- FSSuperblock -- | |
5993 | ||
5994 | void FSSuperblock::encode(bufferlist &bl) const | |
5995 | { | |
5996 | ENCODE_START(2, 1, bl); | |
5997 | compat_features.encode(bl); | |
5998 | ::encode(omap_backend, bl); | |
5999 | ENCODE_FINISH(bl); | |
6000 | } | |
6001 | ||
6002 | void FSSuperblock::decode(bufferlist::iterator &bl) | |
6003 | { | |
6004 | DECODE_START(2, bl); | |
6005 | compat_features.decode(bl); | |
6006 | if (struct_v >= 2) | |
6007 | ::decode(omap_backend, bl); | |
6008 | else | |
6009 | omap_backend = "leveldb"; | |
6010 | DECODE_FINISH(bl); | |
6011 | } | |
6012 | ||
6013 | void FSSuperblock::dump(Formatter *f) const | |
6014 | { | |
6015 | f->open_object_section("compat"); | |
6016 | compat_features.dump(f); | |
6017 | f->dump_string("omap_backend", omap_backend); | |
6018 | f->close_section(); | |
6019 | } | |
6020 | ||
6021 | void FSSuperblock::generate_test_instances(list<FSSuperblock*>& o) | |
6022 | { | |
6023 | FSSuperblock z; | |
6024 | o.push_back(new FSSuperblock(z)); | |
6025 | CompatSet::FeatureSet feature_compat; | |
6026 | CompatSet::FeatureSet feature_ro_compat; | |
6027 | CompatSet::FeatureSet feature_incompat; | |
6028 | feature_incompat.insert(CEPH_FS_FEATURE_INCOMPAT_SHARDS); | |
6029 | z.compat_features = CompatSet(feature_compat, feature_ro_compat, | |
6030 | feature_incompat); | |
6031 | o.push_back(new FSSuperblock(z)); | |
6032 | z.omap_backend = "rocksdb"; | |
6033 | o.push_back(new FSSuperblock(z)); | |
6034 | } |