]>
Commit | Line | Data |
---|---|---|
959ef981 | 1 | // SPDX-License-Identifier: GPL-2.0 |
2bd0ea18 | 2 | /* |
da23017d NS |
3 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. |
4 | * All Rights Reserved. | |
2bd0ea18 NS |
5 | */ |
6 | ||
2bd0ea18 | 7 | #include <sys/stat.h> |
9440d84d | 8 | #include "init.h" |
29e62271 | 9 | |
9c799827 | 10 | #include "libxfs_priv.h" |
b626fb59 DC |
11 | #include "xfs_fs.h" |
12 | #include "xfs_shared.h" | |
13 | #include "xfs_format.h" | |
14 | #include "xfs_log_format.h" | |
15 | #include "xfs_trans_resv.h" | |
16 | #include "xfs_mount.h" | |
794a5604 | 17 | #include "xfs_defer.h" |
b626fb59 DC |
18 | #include "xfs_inode_buf.h" |
19 | #include "xfs_inode_fork.h" | |
20 | #include "xfs_inode.h" | |
21 | #include "xfs_trans.h" | |
b3a96b46 | 22 | #include "xfs_rmap_btree.h" |
e7be6330 | 23 | #include "xfs_refcount_btree.h" |
b658de93 | 24 | #include "libfrog/platform.h" |
b626fb59 | 25 | |
d57269fa CH |
26 | #include "xfs_format.h" |
27 | #include "xfs_da_format.h" | |
28 | #include "xfs_log_format.h" | |
29 | #include "xfs_ondisk.h" | |
30 | ||
6b803e5a | 31 | #include "libxfs.h" /* for now */ |
b626fb59 | 32 | |
7448af58 DW |
33 | #ifndef HAVE_LIBURCU_ATOMIC64 |
34 | pthread_mutex_t atomic64_lock = PTHREAD_MUTEX_INITIALIZER; | |
35 | #endif | |
36 | ||
2bd0ea18 NS |
37 | char *progname = "libxfs"; /* default, changed by each tool */ |
38 | ||
f1b058f9 | 39 | struct cache *libxfs_bcache; /* global buffer cache */ |
9f38f08d | 40 | int libxfs_bhash_size; /* #buckets in bcache */ |
f1b058f9 | 41 | |
167137fe | 42 | int use_xfs_buf_lock; /* global flag: use xfs_buf locks for MT */ |
d0572de5 | 43 | |
00ff2b10 | 44 | static int nextfakedev = -1; /* device number to give to next fake device */ |
2bd0ea18 | 45 | |
2bd0ea18 NS |
46 | /* |
47 | * Checks whether a given device has a mounted, writable | |
48 | * filesystem, returns 1 if it does & fatal (just warns | |
49 | * if not fatal, but allows us to proceed). | |
5000d01d | 50 | * |
2bd0ea18 NS |
51 | * Useful to tools which will produce uncertain results |
52 | * if the filesystem is active - repair, check, logprint. | |
53 | */ | |
54 | static int | |
55 | check_isactive(char *name, char *block, int fatal) | |
56 | { | |
f594a0d1 | 57 | struct stat st; |
2bd0ea18 | 58 | |
f594a0d1 | 59 | if (stat(block, &st) < 0) |
9440d84d | 60 | return 0; |
fc8202ba | 61 | if ((st.st_mode & S_IFMT) != S_IFBLK) |
9440d84d | 62 | return 0; |
93d9f139 | 63 | if (platform_check_ismounted(name, block, &st, 0) == 0) |
9440d84d | 64 | return 0; |
7f510afb ES |
65 | if (platform_check_iswritable(name, block, &st)) |
66 | return fatal ? 1 : 0; | |
67 | return 0; | |
2bd0ea18 NS |
68 | } |
69 | ||
fc83c757 CH |
70 | static int |
71 | check_open( | |
72 | struct libxfs_init *xi, | |
73 | struct libxfs_dev *dev) | |
2bd0ea18 | 74 | { |
fc83c757 | 75 | struct stat stbuf; |
7eb6693f | 76 | |
fc83c757 CH |
77 | if (stat(dev->name, &stbuf) < 0) { |
78 | perror(dev->name); | |
79 | return 0; | |
80 | } | |
81 | if (!(xi->flags & LIBXFS_ISREADONLY) && | |
82 | !(xi->flags & LIBXFS_ISINACTIVE) && | |
83 | platform_check_ismounted(dev->name, dev->name, NULL, 1)) | |
84 | return 0; | |
2bd0ea18 | 85 | |
fc83c757 CH |
86 | if ((xi->flags & LIBXFS_ISINACTIVE) && |
87 | check_isactive(dev->name, dev->name, !!(xi->flags & | |
88 | (LIBXFS_ISREADONLY | LIBXFS_DANGEROUSLY)))) | |
89 | return 0; | |
90 | ||
91 | return 1; | |
92 | } | |
93 | ||
94 | static bool | |
95 | libxfs_device_open( | |
96 | struct libxfs_init *xi, | |
97 | struct libxfs_dev *dev) | |
98 | { | |
99 | struct stat statb; | |
100 | int flags; | |
101 | ||
102 | dev->fd = -1; | |
103 | ||
104 | if (!dev->name) | |
105 | return true; | |
106 | if (!dev->isfile && !check_open(xi, dev)) | |
107 | return false; | |
108 | ||
109 | if (xi->flags & LIBXFS_ISREADONLY) | |
110 | flags = O_RDONLY; | |
111 | else | |
112 | flags = O_RDWR; | |
113 | ||
114 | if (dev->create) { | |
115 | flags |= O_CREAT | O_TRUNC; | |
116 | } else { | |
117 | if (xi->flags & LIBXFS_EXCLUSIVELY) | |
118 | flags |= O_EXCL; | |
119 | if ((xi->flags & LIBXFS_DIRECT) && platform_direct_blockdev()) | |
120 | flags |= O_DIRECT; | |
121 | } | |
b74a1f6a | 122 | |
fc83c757 CH |
123 | retry: |
124 | dev->fd = open(dev->name, flags, 0666); | |
125 | if (dev->fd < 0) { | |
126 | if (errno == EINVAL && (flags & O_DIRECT)) { | |
127 | flags &= ~O_DIRECT; | |
b74a1f6a | 128 | goto retry; |
fc83c757 | 129 | } |
9440d84d | 130 | fprintf(stderr, _("%s: cannot open %s: %s\n"), |
fc83c757 | 131 | progname, dev->name, strerror(errno)); |
2bd0ea18 NS |
132 | exit(1); |
133 | } | |
134 | ||
fc83c757 | 135 | if (fstat(dev->fd, &statb) < 0) { |
9440d84d | 136 | fprintf(stderr, _("%s: cannot stat %s: %s\n"), |
fc83c757 | 137 | progname, dev->name, strerror(errno)); |
2bd0ea18 NS |
138 | exit(1); |
139 | } | |
a33a9e62 | 140 | |
fc83c757 CH |
141 | if (!(xi->flags & LIBXFS_ISREADONLY) && |
142 | xi->setblksize && | |
143 | (statb.st_mode & S_IFMT) == S_IFBLK) { | |
a3106f32 CH |
144 | /* |
145 | * Try to use the given explicit blocksize. Failure to set the | |
146 | * block size is only fatal for direct I/O. | |
147 | */ | |
fc83c757 CH |
148 | platform_set_blocksize(dev->fd, dev->name, statb.st_rdev, |
149 | xi->setblksize, flags & O_DIRECT); | |
edd45774 | 150 | } |
2bd0ea18 | 151 | |
a33a9e62 | 152 | /* |
7b47b1bc CH |
153 | * Get the device number from the stat buf - unless we're not opening a |
154 | * real device, in which case choose a new fake device number. | |
2bd0ea18 | 155 | */ |
7b47b1bc | 156 | if (statb.st_rdev) |
fc83c757 CH |
157 | dev->dev = statb.st_rdev; |
158 | else | |
159 | dev->dev = nextfakedev--; | |
160 | platform_findsizes(dev->name, dev->fd, &dev->size, &dev->bsize); | |
161 | return true; | |
2bd0ea18 NS |
162 | } |
163 | ||
4f112cb1 | 164 | static void |
fc83c757 CH |
165 | libxfs_device_close( |
166 | struct libxfs_dev *dev) | |
2bd0ea18 | 167 | { |
fc83c757 | 168 | int ret; |
2bd0ea18 | 169 | |
fc83c757 | 170 | ret = platform_flush_device(dev->fd, dev->dev); |
7b47b1bc CH |
171 | if (ret) { |
172 | ret = -errno; | |
173 | fprintf(stderr, | |
023ba280 | 174 | _("%s: flush of device %lld failed, err=%d"), |
7b47b1bc CH |
175 | progname, (long long)dev, ret); |
176 | } | |
fc83c757 | 177 | close(dev->fd); |
c781939c | 178 | |
fc83c757 CH |
179 | dev->fd = -1; |
180 | dev->dev = 0; | |
c781939c RC |
181 | } |
182 | ||
7a326ce0 | 183 | /* |
2e1394fc | 184 | * Initialize/destroy all of the cache allocators we use. |
7a326ce0 ES |
185 | */ |
186 | static void | |
2e1394fc | 187 | init_caches(void) |
7a326ce0 | 188 | { |
7d10d094 DW |
189 | int error; |
190 | ||
2e1394fc DW |
191 | /* initialise cache allocation */ |
192 | xfs_buf_cache = kmem_cache_init(sizeof(struct xfs_buf), "xfs_buffer"); | |
193 | xfs_inode_cache = kmem_cache_init(sizeof(struct xfs_inode), "xfs_inode"); | |
194 | xfs_ifork_cache = kmem_cache_init(sizeof(struct xfs_ifork), "xfs_ifork"); | |
195 | xfs_ili_cache = kmem_cache_init( | |
7a326ce0 | 196 | sizeof(struct xfs_inode_log_item),"xfs_inode_log_item"); |
2e1394fc | 197 | xfs_buf_item_cache = kmem_cache_init( |
7a326ce0 | 198 | sizeof(struct xfs_buf_log_item), "xfs_buf_log_item"); |
1577541c DW |
199 | error = xfs_defer_init_item_caches(); |
200 | if (error) { | |
201 | fprintf(stderr, "Could not allocate defer init item caches.\n"); | |
202 | abort(); | |
203 | } | |
2e1394fc | 204 | xfs_da_state_cache = kmem_cache_init( |
7a326ce0 | 205 | sizeof(struct xfs_da_state), "xfs_da_state"); |
7d10d094 DW |
206 | error = xfs_btree_init_cur_caches(); |
207 | if (error) { | |
208 | fprintf(stderr, "Could not allocate btree cursor caches.\n"); | |
209 | abort(); | |
210 | } | |
7d84b02d | 211 | xfs_extfree_item_cache = kmem_cache_init( |
7a326ce0 | 212 | sizeof(struct xfs_extent_free_item), |
7d84b02d | 213 | "xfs_extfree_item"); |
2e1394fc | 214 | xfs_trans_cache = kmem_cache_init( |
7a326ce0 ES |
215 | sizeof(struct xfs_trans), "xfs_trans"); |
216 | } | |
217 | ||
218 | static int | |
2e1394fc | 219 | destroy_caches(void) |
7a326ce0 ES |
220 | { |
221 | int leaked = 0; | |
222 | ||
2e1394fc DW |
223 | leaked += kmem_cache_destroy(xfs_buf_cache); |
224 | leaked += kmem_cache_destroy(xfs_ili_cache); | |
225 | leaked += kmem_cache_destroy(xfs_inode_cache); | |
226 | leaked += kmem_cache_destroy(xfs_ifork_cache); | |
227 | leaked += kmem_cache_destroy(xfs_buf_item_cache); | |
228 | leaked += kmem_cache_destroy(xfs_da_state_cache); | |
1577541c | 229 | xfs_defer_destroy_item_caches(); |
7d10d094 | 230 | xfs_btree_destroy_cur_caches(); |
7d84b02d | 231 | leaked += kmem_cache_destroy(xfs_extfree_item_cache); |
2e1394fc | 232 | leaked += kmem_cache_destroy(xfs_trans_cache); |
7a326ce0 ES |
233 | |
234 | return leaked; | |
235 | } | |
236 | ||
a9468486 DW |
237 | static void |
238 | libxfs_close_devices( | |
01dcfd9e | 239 | struct libxfs_init *li) |
a9468486 | 240 | { |
fc83c757 CH |
241 | if (li->data.dev) |
242 | libxfs_device_close(&li->data); | |
243 | if (li->log.dev && li->log.dev != li->data.dev) | |
244 | libxfs_device_close(&li->log); | |
245 | if (li->rt.dev) | |
246 | libxfs_device_close(&li->rt); | |
a9468486 DW |
247 | } |
248 | ||
2bd0ea18 NS |
249 | /* |
250 | * libxfs initialization. | |
251 | * Caller gets a 0 on failure (and we print a message), 1 on success. | |
252 | */ | |
253 | int | |
01dcfd9e | 254 | libxfs_init(struct libxfs_init *a) |
2bd0ea18 | 255 | { |
d57269fa | 256 | xfs_check_ondisk_structs(); |
e4da1b16 DC |
257 | rcu_init(); |
258 | rcu_register_thread(); | |
bacd44a5 AE |
259 | radix_tree_init(); |
260 | ||
fc83c757 CH |
261 | if (!libxfs_device_open(a, &a->data)) |
262 | goto done; | |
263 | if (!libxfs_device_open(a, &a->log)) | |
264 | goto done; | |
265 | if (!libxfs_device_open(a, &a->rt)) | |
266 | goto done; | |
b6e08bf3 | 267 | |
9f38f08d MV |
268 | if (!libxfs_bhash_size) |
269 | libxfs_bhash_size = LIBXFS_BHASHSIZE(sbp); | |
ba9ecd40 DC |
270 | libxfs_bcache = cache_init(a->bcache_flags, libxfs_bhash_size, |
271 | &libxfs_bcache_operations); | |
23d88955 | 272 | use_xfs_buf_lock = a->flags & LIBXFS_USEBUFLOCK; |
7a326ce0 | 273 | xfs_dir_startup(); |
2e1394fc | 274 | init_caches(); |
732f5b90 | 275 | return 1; |
a9468486 | 276 | |
732f5b90 CH |
277 | done: |
278 | libxfs_close_devices(a); | |
279 | rcu_unregister_thread(); | |
280 | return 0; | |
2bd0ea18 NS |
281 | } |
282 | ||
283 | ||
b391b7cd NS |
284 | /* |
285 | * Initialize realtime fields in the mount structure. | |
286 | */ | |
287 | static int | |
288 | rtmount_init( | |
2420d095 | 289 | xfs_mount_t *mp) /* file system mount structure */ |
b391b7cd | 290 | { |
31079e67 | 291 | struct xfs_buf *bp; /* buffer for last block of subvolume */ |
b391b7cd | 292 | xfs_daddr_t d; /* address of last block of subvolume */ |
113af235 | 293 | unsigned int rsumblocks; |
31079e67 | 294 | int error; |
b391b7cd | 295 | |
575f24e5 | 296 | if (mp->m_sb.sb_rblocks == 0) |
b391b7cd | 297 | return 0; |
4aaeedc4 | 298 | |
eefdf2ab | 299 | if (xfs_has_reflink(mp)) { |
4aaeedc4 DW |
300 | fprintf(stderr, |
301 | _("%s: Reflink not compatible with realtime device. Please try a newer xfsprogs.\n"), | |
302 | progname); | |
303 | return -1; | |
304 | } | |
305 | ||
eefdf2ab | 306 | if (xfs_has_rmapbt(mp)) { |
4aaeedc4 DW |
307 | fprintf(stderr, |
308 | _("%s: Reverse mapping btree not compatible with realtime device. Please try a newer xfsprogs.\n"), | |
309 | progname); | |
310 | return -1; | |
311 | } | |
312 | ||
2420d095 | 313 | if (mp->m_rtdev_targp->bt_bdev == 0 && !xfs_is_debugger(mp)) { |
9440d84d | 314 | fprintf(stderr, _("%s: filesystem has a realtime subvolume\n"), |
b391b7cd NS |
315 | progname); |
316 | return -1; | |
317 | } | |
575f24e5 | 318 | mp->m_rsumlevels = mp->m_sb.sb_rextslog + 1; |
113af235 DW |
319 | rsumblocks = xfs_rtsummary_blockcount(mp, mp->m_rsumlevels, |
320 | mp->m_sb.sb_rbmblocks); | |
321 | mp->m_rsumsize = XFS_FSB_TO_B(mp, rsumblocks); | |
b391b7cd | 322 | mp->m_rbmip = mp->m_rsumip = NULL; |
39798eb5 NS |
323 | |
324 | /* | |
325 | * Allow debugger to be run without the realtime device present. | |
326 | */ | |
2420d095 | 327 | if (xfs_is_debugger(mp)) |
39798eb5 NS |
328 | return 0; |
329 | ||
b391b7cd NS |
330 | /* |
331 | * Check that the realtime section is an ok size. | |
332 | */ | |
333 | d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_rblocks); | |
334 | if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_rblocks) { | |
9440d84d NS |
335 | fprintf(stderr, _("%s: realtime init - %llu != %llu\n"), |
336 | progname, (unsigned long long) XFS_BB_TO_FSB(mp, d), | |
b391b7cd NS |
337 | (unsigned long long) mp->m_sb.sb_rblocks); |
338 | return -1; | |
339 | } | |
31079e67 DW |
340 | error = libxfs_buf_read(mp->m_rtdev, d - XFS_FSB_TO_BB(mp, 1), |
341 | XFS_FSB_TO_BB(mp, 1), 0, &bp, NULL); | |
342 | if (error) { | |
9440d84d NS |
343 | fprintf(stderr, _("%s: realtime size check failed\n"), |
344 | progname); | |
b391b7cd NS |
345 | return -1; |
346 | } | |
e02ba985 | 347 | libxfs_buf_relse(bp); |
b391b7cd NS |
348 | return 0; |
349 | } | |
350 | ||
03dc2ef2 DC |
351 | static bool |
352 | xfs_set_inode_alloc_perag( | |
353 | struct xfs_perag *pag, | |
354 | xfs_ino_t ino, | |
355 | xfs_agnumber_t max_metadata) | |
356 | { | |
357 | if (!xfs_is_inode32(pag->pag_mount)) { | |
358 | set_bit(XFS_AGSTATE_ALLOWS_INODES, &pag->pag_opstate); | |
359 | clear_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate); | |
360 | return false; | |
361 | } | |
362 | ||
363 | if (ino > XFS_MAXINUMBER_32) { | |
364 | clear_bit(XFS_AGSTATE_ALLOWS_INODES, &pag->pag_opstate); | |
365 | clear_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate); | |
366 | return false; | |
367 | } | |
368 | ||
369 | set_bit(XFS_AGSTATE_ALLOWS_INODES, &pag->pag_opstate); | |
370 | if (pag->pag_agno < max_metadata) | |
371 | set_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate); | |
372 | else | |
373 | clear_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate); | |
374 | return true; | |
375 | } | |
376 | ||
b9ee1227 DW |
377 | /* |
378 | * Set parameters for inode allocation heuristics, taking into account | |
379 | * filesystem size and inode32/inode64 mount options; i.e. specifically | |
380 | * whether or not XFS_MOUNT_SMALL_INUMS is set. | |
381 | * | |
382 | * Inode allocation patterns are altered only if inode32 is requested | |
383 | * (XFS_MOUNT_SMALL_INUMS), and the filesystem is sufficiently large. | |
384 | * If altered, XFS_MOUNT_32BITINODES is set as well. | |
385 | * | |
386 | * An agcount independent of that in the mount structure is provided | |
387 | * because in the growfs case, mp->m_sb.sb_agcount is not yet updated | |
388 | * to the potentially higher ag count. | |
389 | * | |
390 | * Returns the maximum AG index which may contain inodes. | |
ed8f5980 DW |
391 | * |
392 | * NOTE: userspace has no concept of "inode32" and so xfs_has_small_inums | |
393 | * is always false, and much of this code is a no-op. | |
b9ee1227 DW |
394 | */ |
395 | xfs_agnumber_t | |
396 | xfs_set_inode_alloc( | |
397 | struct xfs_mount *mp, | |
398 | xfs_agnumber_t agcount) | |
399 | { | |
400 | xfs_agnumber_t index; | |
401 | xfs_agnumber_t maxagi = 0; | |
402 | xfs_sb_t *sbp = &mp->m_sb; | |
403 | xfs_agnumber_t max_metadata; | |
404 | xfs_agino_t agino; | |
405 | xfs_ino_t ino; | |
406 | ||
407 | /* | |
408 | * Calculate how much should be reserved for inodes to meet | |
409 | * the max inode percentage. Used only for inode32. | |
410 | */ | |
411 | if (M_IGEO(mp)->maxicount) { | |
412 | uint64_t icount; | |
413 | ||
414 | icount = sbp->sb_dblocks * sbp->sb_imax_pct; | |
415 | do_div(icount, 100); | |
416 | icount += sbp->sb_agblocks - 1; | |
417 | do_div(icount, sbp->sb_agblocks); | |
418 | max_metadata = icount; | |
419 | } else { | |
420 | max_metadata = agcount; | |
421 | } | |
422 | ||
423 | /* Get the last possible inode in the filesystem */ | |
424 | agino = XFS_AGB_TO_AGINO(mp, sbp->sb_agblocks - 1); | |
425 | ino = XFS_AGINO_TO_INO(mp, agcount - 1, agino); | |
426 | ||
427 | /* | |
428 | * If user asked for no more than 32-bit inodes, and the fs is | |
429 | * sufficiently large, set XFS_MOUNT_32BITINODES if we must alter | |
430 | * the allocator to accommodate the request. | |
431 | */ | |
ed8f5980 | 432 | if (xfs_has_small_inums(mp) && ino > XFS_MAXINUMBER_32) |
03dc2ef2 | 433 | set_bit(XFS_OPSTATE_INODE32, &mp->m_opstate); |
ed8f5980 | 434 | else |
03dc2ef2 | 435 | clear_bit(XFS_OPSTATE_INODE32, &mp->m_opstate); |
b9ee1227 DW |
436 | |
437 | for (index = 0; index < agcount; index++) { | |
438 | struct xfs_perag *pag; | |
439 | ||
440 | ino = XFS_AGINO_TO_INO(mp, index, agino); | |
441 | ||
442 | pag = xfs_perag_get(mp, index); | |
03dc2ef2 DC |
443 | if (xfs_set_inode_alloc_perag(pag, ino, max_metadata)) |
444 | maxagi++; | |
b9ee1227 DW |
445 | xfs_perag_put(pag); |
446 | } | |
447 | ||
0ee9753e | 448 | return xfs_is_inode32(mp) ? maxagi : agcount; |
b9ee1227 DW |
449 | } |
450 | ||
75c8b434 DC |
451 | static struct xfs_buftarg * |
452 | libxfs_buftarg_alloc( | |
453 | struct xfs_mount *mp, | |
fc83c757 | 454 | struct libxfs_dev *dev, |
704e4cef | 455 | unsigned long write_fails) |
75c8b434 DC |
456 | { |
457 | struct xfs_buftarg *btp; | |
458 | ||
459 | btp = malloc(sizeof(*btp)); | |
460 | if (!btp) { | |
461 | fprintf(stderr, _("%s: buftarg init failed\n"), | |
462 | progname); | |
463 | exit(1); | |
464 | } | |
465 | btp->bt_mount = mp; | |
fc83c757 CH |
466 | btp->bt_bdev = dev->dev; |
467 | btp->bt_bdev_fd = dev->fd; | |
c335b673 | 468 | btp->flags = 0; |
704e4cef DW |
469 | if (write_fails) { |
470 | btp->writes_left = write_fails; | |
471 | btp->flags |= XFS_BUFTARG_INJECT_WRITE_FAIL; | |
472 | } | |
473 | pthread_mutex_init(&btp->lock, NULL); | |
c335b673 | 474 | |
75c8b434 DC |
475 | return btp; |
476 | } | |
477 | ||
704e4cef DW |
478 | enum libxfs_write_failure_nums { |
479 | WF_DATA = 0, | |
480 | WF_LOG, | |
481 | WF_RT, | |
482 | WF_MAX_OPTS, | |
483 | }; | |
484 | ||
485 | static char *wf_opts[] = { | |
486 | [WF_DATA] = "ddev", | |
487 | [WF_LOG] = "logdev", | |
488 | [WF_RT] = "rtdev", | |
489 | [WF_MAX_OPTS] = NULL, | |
490 | }; | |
491 | ||
75c8b434 DC |
492 | void |
493 | libxfs_buftarg_init( | |
494 | struct xfs_mount *mp, | |
ca8cc76e | 495 | struct libxfs_init *xi) |
75c8b434 | 496 | { |
704e4cef DW |
497 | char *p = getenv("LIBXFS_DEBUG_WRITE_CRASH"); |
498 | unsigned long dfail = 0, lfail = 0, rfail = 0; | |
499 | ||
500 | /* Simulate utility crash after a certain number of writes. */ | |
501 | while (p && *p) { | |
502 | char *val; | |
503 | ||
504 | switch (getsubopt(&p, wf_opts, &val)) { | |
505 | case WF_DATA: | |
506 | if (!val) { | |
507 | fprintf(stderr, | |
508 | _("ddev write fail requires a parameter\n")); | |
509 | exit(1); | |
510 | } | |
511 | dfail = strtoul(val, NULL, 0); | |
512 | break; | |
513 | case WF_LOG: | |
514 | if (!val) { | |
515 | fprintf(stderr, | |
516 | _("logdev write fail requires a parameter\n")); | |
517 | exit(1); | |
518 | } | |
519 | lfail = strtoul(val, NULL, 0); | |
520 | break; | |
521 | case WF_RT: | |
522 | if (!val) { | |
523 | fprintf(stderr, | |
524 | _("rtdev write fail requires a parameter\n")); | |
525 | exit(1); | |
526 | } | |
527 | rfail = strtoul(val, NULL, 0); | |
528 | break; | |
529 | default: | |
530 | fprintf(stderr, _("unknown write fail type %s\n"), | |
531 | val); | |
532 | exit(1); | |
533 | break; | |
534 | } | |
535 | } | |
536 | ||
75c8b434 DC |
537 | if (mp->m_ddev_targp) { |
538 | /* should already have all buftargs initialised */ | |
fc83c757 | 539 | if (mp->m_ddev_targp->bt_bdev != xi->data.dev || |
75c8b434 DC |
540 | mp->m_ddev_targp->bt_mount != mp) { |
541 | fprintf(stderr, | |
542 | _("%s: bad buftarg reinit, ddev\n"), | |
543 | progname); | |
544 | exit(1); | |
545 | } | |
fc83c757 | 546 | if (!xi->log.dev || xi->log.dev == xi->data.dev) { |
75c8b434 DC |
547 | if (mp->m_logdev_targp != mp->m_ddev_targp) { |
548 | fprintf(stderr, | |
549 | _("%s: bad buftarg reinit, ldev mismatch\n"), | |
550 | progname); | |
551 | exit(1); | |
552 | } | |
fc83c757 | 553 | } else if (mp->m_logdev_targp->bt_bdev != xi->log.dev || |
75c8b434 DC |
554 | mp->m_logdev_targp->bt_mount != mp) { |
555 | fprintf(stderr, | |
556 | _("%s: bad buftarg reinit, logdev\n"), | |
557 | progname); | |
558 | exit(1); | |
559 | } | |
fc83c757 CH |
560 | if (xi->rt.dev && |
561 | (mp->m_rtdev_targp->bt_bdev != xi->rt.dev || | |
ca8cc76e | 562 | mp->m_rtdev_targp->bt_mount != mp)) { |
75c8b434 DC |
563 | fprintf(stderr, |
564 | _("%s: bad buftarg reinit, rtdev\n"), | |
565 | progname); | |
566 | exit(1); | |
567 | } | |
568 | return; | |
569 | } | |
570 | ||
fc83c757 CH |
571 | mp->m_ddev_targp = libxfs_buftarg_alloc(mp, &xi->data, dfail); |
572 | if (!xi->log.dev || xi->log.dev == xi->data.dev) | |
75c8b434 DC |
573 | mp->m_logdev_targp = mp->m_ddev_targp; |
574 | else | |
fc83c757 CH |
575 | mp->m_logdev_targp = libxfs_buftarg_alloc(mp, &xi->log, lfail); |
576 | mp->m_rtdev_targp = libxfs_buftarg_alloc(mp, &xi->rt, rfail); | |
75c8b434 DC |
577 | } |
578 | ||
6afce48f DW |
579 | /* Compute maximum possible height for per-AG btree types for this fs. */ |
580 | static inline void | |
581 | xfs_agbtree_compute_maxlevels( | |
582 | struct xfs_mount *mp) | |
583 | { | |
584 | unsigned int levels; | |
585 | ||
586 | levels = max(mp->m_alloc_maxlevels, M_IGEO(mp)->inobt_maxlevels); | |
587 | levels = max(levels, mp->m_rmap_maxlevels); | |
588 | mp->m_agbtree_maxlevels = max(levels, mp->m_refc_maxlevels); | |
589 | } | |
590 | ||
7aeffc87 DW |
591 | /* Compute maximum possible height of all btrees. */ |
592 | void | |
593 | libxfs_compute_all_maxlevels( | |
594 | struct xfs_mount *mp) | |
595 | { | |
596 | xfs_alloc_compute_maxlevels(mp); | |
597 | xfs_bmap_compute_maxlevels(mp, XFS_DATA_FORK); | |
598 | xfs_bmap_compute_maxlevels(mp, XFS_ATTR_FORK); | |
599 | xfs_ialloc_setup_geometry(mp); | |
600 | xfs_rmapbt_compute_maxlevels(mp); | |
601 | xfs_refcountbt_compute_maxlevels(mp); | |
602 | ||
603 | xfs_agbtree_compute_maxlevels(mp); | |
604 | } | |
605 | ||
33f3aac8 DC |
606 | /* |
607 | * precalculate the low space thresholds for dynamic speculative preallocation. | |
608 | */ | |
609 | static void | |
610 | xfs_set_low_space_thresholds( | |
611 | struct xfs_mount *mp) | |
612 | { | |
613 | uint64_t dblocks = mp->m_sb.sb_dblocks; | |
614 | int i; | |
615 | ||
616 | do_div(dblocks, 100); | |
617 | ||
618 | for (i = 0; i < XFS_LOWSP_MAX; i++) | |
619 | mp->m_low_space[i] = dblocks * (i + 1); | |
620 | } | |
621 | ||
2bd0ea18 NS |
622 | /* |
623 | * Mount structure initialization, provides a filled-in xfs_mount_t | |
624 | * such that the numerous XFS_* macros can be used. If dev is zero, | |
625 | * no IO will be performed (no size checks, read root inodes). | |
626 | */ | |
d855bce8 | 627 | struct xfs_mount * |
2bd0ea18 | 628 | libxfs_mount( |
d855bce8 DW |
629 | struct xfs_mount *mp, |
630 | struct xfs_sb *sb, | |
ddd9942b | 631 | struct libxfs_init *xi, |
ed8f5980 | 632 | unsigned int flags) |
2bd0ea18 | 633 | { |
d855bce8 DW |
634 | struct xfs_buf *bp; |
635 | struct xfs_sb *sbp; | |
636 | xfs_daddr_t d; | |
d855bce8 | 637 | int error; |
2bd0ea18 | 638 | |
3bc1fdd4 | 639 | mp->m_features = xfs_sb_version_to_features(sb); |
2420d095 DW |
640 | if (flags & LIBXFS_MOUNT_DEBUGGER) |
641 | xfs_set_debugger(mp); | |
e42c53f3 | 642 | if (flags & LIBXFS_MOUNT_REPORT_CORRUPTION) |
2420d095 | 643 | xfs_set_reporting_corruption(mp); |
ca8cc76e | 644 | libxfs_buftarg_init(mp, xi); |
75c8b434 | 645 | |
f747f7dd | 646 | mp->m_finobt_nores = true; |
0ee9753e | 647 | xfs_set_inode32(mp); |
2bd0ea18 | 648 | mp->m_sb = *sb; |
56b2de80 | 649 | INIT_RADIX_TREE(&mp->m_perag_tree, GFP_KERNEL); |
686bddf9 DC |
650 | sbp = &mp->m_sb; |
651 | spin_lock_init(&mp->m_sb_lock); | |
652 | spin_lock_init(&mp->m_agirotor_lock); | |
2bd0ea18 | 653 | |
4896e6c8 | 654 | xfs_sb_mount_common(mp, sb); |
2bd0ea18 | 655 | |
949c0f10 NS |
656 | /* |
657 | * Set whether we're using stripe alignment. | |
658 | */ | |
2660e653 | 659 | if (xfs_has_dalign(mp)) { |
949c0f10 NS |
660 | mp->m_dalign = sbp->sb_unit; |
661 | mp->m_swidth = sbp->sb_width; | |
662 | } | |
663 | ||
7aeffc87 | 664 | libxfs_compute_all_maxlevels(mp); |
6afce48f | 665 | |
2bd0ea18 NS |
666 | /* |
667 | * Check that the data (and log if separate) are an ok size. | |
668 | */ | |
9440d84d | 669 | d = (xfs_daddr_t) XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks); |
2bd0ea18 | 670 | if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_dblocks) { |
9440d84d | 671 | fprintf(stderr, _("%s: size check failed\n"), progname); |
2420d095 | 672 | if (!xfs_is_debugger(mp)) |
4ca431fc | 673 | return NULL; |
2bd0ea18 NS |
674 | } |
675 | ||
ff105f75 DC |
676 | /* |
677 | * We automatically convert v1 inodes to v2 inodes now, so if | |
678 | * the NLINK bit is not set we can't operate on the filesystem. | |
679 | */ | |
680 | if (!(sbp->sb_versionnum & XFS_SB_VERSION_NLINKBIT)) { | |
681 | ||
682 | fprintf(stderr, _( | |
683 | "%s: V1 inodes unsupported. Please try an older xfsprogs.\n"), | |
684 | progname); | |
685 | exit(1); | |
686 | } | |
687 | ||
688 | /* Check for supported directory formats */ | |
689 | if (!(sbp->sb_versionnum & XFS_SB_VERSION_DIRV2BIT)) { | |
9a048535 DC |
690 | |
691 | fprintf(stderr, _( | |
692 | "%s: V1 directories unsupported. Please try an older xfsprogs.\n"), | |
693 | progname); | |
694 | exit(1); | |
5e656dbb | 695 | } |
2bd0ea18 | 696 | |
ff105f75 DC |
697 | /* check for unsupported other features */ |
698 | if (!xfs_sb_good_version(sbp)) { | |
699 | fprintf(stderr, _( | |
700 | "%s: Unsupported features detected. Please try a newer xfsprogs.\n"), | |
701 | progname); | |
702 | exit(1); | |
703 | } | |
704 | ||
705 | xfs_da_mount(mp); | |
706 | ||
2bd0ea18 | 707 | /* Initialize the precomputed transaction reservations values */ |
5e656dbb | 708 | xfs_trans_init(mp); |
2bd0ea18 | 709 | |
fc83c757 | 710 | if (xi->data.dev == 0) /* maxtrres, we have no device so leave now */ |
2bd0ea18 NS |
711 | return mp; |
712 | ||
d855bce8 | 713 | /* device size checks must pass unless we're a debugger. */ |
31079e67 DW |
714 | error = libxfs_buf_read(mp->m_dev, d - XFS_FSS_TO_BB(mp, 1), |
715 | XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL); | |
716 | if (error) { | |
9440d84d | 717 | fprintf(stderr, _("%s: data size check failed\n"), progname); |
2420d095 | 718 | if (!xfs_is_debugger(mp)) |
4ca431fc | 719 | return NULL; |
32244196 | 720 | } else |
e02ba985 | 721 | libxfs_buf_relse(bp); |
2bd0ea18 | 722 | |
ab434d12 DC |
723 | if (mp->m_logdev_targp->bt_bdev && |
724 | mp->m_logdev_targp->bt_bdev != mp->m_ddev_targp->bt_bdev) { | |
9440d84d | 725 | d = (xfs_daddr_t) XFS_FSB_TO_BB(mp, mp->m_sb.sb_logblocks); |
31079e67 DW |
726 | if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_logblocks || |
727 | libxfs_buf_read(mp->m_logdev_targp, | |
728 | d - XFS_FSB_TO_BB(mp, 1), XFS_FSB_TO_BB(mp, 1), | |
729 | 0, &bp, NULL)) { | |
9440d84d | 730 | fprintf(stderr, _("%s: log size checks failed\n"), |
2bd0ea18 | 731 | progname); |
2420d095 | 732 | if (!xfs_is_debugger(mp)) |
4ca431fc | 733 | return NULL; |
2bd0ea18 | 734 | } |
32244196 | 735 | if (bp) |
e02ba985 | 736 | libxfs_buf_relse(bp); |
2bd0ea18 NS |
737 | } |
738 | ||
33f3aac8 DC |
739 | xfs_set_low_space_thresholds(mp); |
740 | ||
2bd0ea18 | 741 | /* Initialize realtime fields in the mount structure */ |
2420d095 | 742 | if (rtmount_init(mp)) { |
9440d84d NS |
743 | fprintf(stderr, _("%s: realtime device init failed\n"), |
744 | progname); | |
4ca431fc | 745 | return NULL; |
2bd0ea18 NS |
746 | } |
747 | ||
a547152d ES |
748 | /* |
749 | * libxfs_initialize_perag will allocate a perag structure for each ag. | |
750 | * If agcount is corrupted and insanely high, this will OOM the box. | |
751 | * If the agount seems (arbitrarily) high, try to read what would be | |
752 | * the last AG, and if that fails for a relatively high agcount, just | |
753 | * read the first one and let the user know to check the geometry. | |
754 | */ | |
755 | if (sbp->sb_agcount > 1000000) { | |
31079e67 | 756 | error = libxfs_buf_read(mp->m_dev, |
a547152d | 757 | XFS_AG_DADDR(mp, sbp->sb_agcount - 1, 0), 1, |
31079e67 DW |
758 | 0, &bp, NULL); |
759 | if (error) { | |
a547152d ES |
760 | fprintf(stderr, _("%s: read of AG %u failed\n"), |
761 | progname, sbp->sb_agcount); | |
2420d095 | 762 | if (!xfs_is_debugger(mp)) |
a547152d ES |
763 | return NULL; |
764 | fprintf(stderr, _("%s: limiting reads to AG 0\n"), | |
765 | progname); | |
766 | sbp->sb_agcount = 1; | |
31079e67 DW |
767 | } else |
768 | libxfs_buf_relse(bp); | |
a547152d ES |
769 | } |
770 | ||
83af0d13 DC |
771 | error = libxfs_initialize_perag(mp, sbp->sb_agcount, sbp->sb_dblocks, |
772 | &mp->m_maxagi); | |
56b2de80 DC |
773 | if (error) { |
774 | fprintf(stderr, _("%s: perag init failed\n"), | |
775 | progname); | |
2bd0ea18 NS |
776 | exit(1); |
777 | } | |
2420d095 | 778 | xfs_set_perag_data_loaded(mp); |
2bd0ea18 | 779 | |
2bd0ea18 NS |
780 | return mp; |
781 | } | |
782 | ||
f1b058f9 NS |
783 | void |
784 | libxfs_rtmount_destroy(xfs_mount_t *mp) | |
785 | { | |
786 | if (mp->m_rsumip) | |
31845e4c | 787 | libxfs_irele(mp->m_rsumip); |
f1b058f9 | 788 | if (mp->m_rbmip) |
31845e4c | 789 | libxfs_irele(mp->m_rbmip); |
f1b058f9 NS |
790 | mp->m_rsumip = mp->m_rbmip = NULL; |
791 | } | |
792 | ||
c335b673 DW |
793 | /* Flush a device and report on writes that didn't make it to stable storage. */ |
794 | static inline int | |
795 | libxfs_flush_buftarg( | |
796 | struct xfs_buftarg *btp, | |
797 | const char *buftarg_descr) | |
798 | { | |
799 | int error = 0; | |
800 | int err2; | |
801 | ||
802 | /* | |
803 | * Write verifier failures are evidence of a buggy program. Make sure | |
804 | * that this state is always reported to the caller. | |
805 | */ | |
806 | if (btp->flags & XFS_BUFTARG_CORRUPT_WRITE) { | |
807 | fprintf(stderr, | |
808 | _("%s: Refusing to write a corrupt buffer to the %s!\n"), | |
809 | progname, buftarg_descr); | |
810 | error = -EFSCORRUPTED; | |
811 | } | |
812 | ||
813 | if (btp->flags & XFS_BUFTARG_LOST_WRITE) { | |
814 | fprintf(stderr, | |
815 | _("%s: Lost a write to the %s!\n"), | |
816 | progname, buftarg_descr); | |
817 | if (!error) | |
818 | error = -EIO; | |
819 | } | |
820 | ||
821 | err2 = libxfs_blkdev_issue_flush(btp); | |
822 | if (err2) { | |
823 | fprintf(stderr, | |
824 | _("%s: Flushing the %s failed, err=%d!\n"), | |
825 | progname, buftarg_descr, -err2); | |
826 | } | |
827 | if (!error) | |
828 | error = err2; | |
829 | ||
830 | return error; | |
831 | } | |
832 | ||
833 | /* | |
834 | * Flush all dirty buffers to stable storage and report on writes that didn't | |
835 | * make it to stable storage. | |
836 | */ | |
a7348c58 | 837 | int |
c335b673 DW |
838 | libxfs_flush_mount( |
839 | struct xfs_mount *mp) | |
840 | { | |
841 | int error = 0; | |
842 | int err2; | |
843 | ||
844 | /* | |
a7348c58 DW |
845 | * Flush the buffer cache to write all dirty buffers to disk. Buffers |
846 | * that fail write verification will cause the CORRUPT_WRITE flag to be | |
847 | * set in the buftarg. Buffers that cannot be written will cause the | |
848 | * LOST_WRITE flag to be set in the buftarg. Once that's done, | |
849 | * instruct the disks to persist their write caches. | |
c335b673 | 850 | */ |
a7348c58 | 851 | libxfs_bcache_flush(); |
c335b673 DW |
852 | |
853 | /* Flush all kernel and disk write caches, and report failures. */ | |
854 | if (mp->m_ddev_targp) { | |
855 | err2 = libxfs_flush_buftarg(mp->m_ddev_targp, _("data device")); | |
856 | if (!error) | |
857 | error = err2; | |
858 | } | |
859 | ||
860 | if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { | |
861 | err2 = libxfs_flush_buftarg(mp->m_logdev_targp, | |
862 | _("log device")); | |
863 | if (!error) | |
864 | error = err2; | |
865 | } | |
866 | ||
867 | if (mp->m_rtdev_targp) { | |
868 | err2 = libxfs_flush_buftarg(mp->m_rtdev_targp, | |
869 | _("realtime device")); | |
870 | if (!error) | |
871 | error = err2; | |
872 | } | |
873 | ||
874 | return error; | |
875 | } | |
876 | ||
2bd0ea18 | 877 | /* |
9440d84d | 878 | * Release any resource obtained during a mount. |
2bd0ea18 | 879 | */ |
c335b673 DW |
880 | int |
881 | libxfs_umount( | |
882 | struct xfs_mount *mp) | |
2bd0ea18 | 883 | { |
c335b673 | 884 | int error; |
56b2de80 | 885 | |
f1b058f9 | 886 | libxfs_rtmount_destroy(mp); |
c335b673 | 887 | |
a7348c58 DW |
888 | /* |
889 | * Purge the buffer cache to write all dirty buffers to disk and free | |
890 | * all incore buffers, then pick up the outcome when we tell the disks | |
891 | * to persist their write caches. | |
892 | */ | |
893 | libxfs_bcache_purge(); | |
c335b673 | 894 | error = libxfs_flush_mount(mp); |
f1b058f9 | 895 | |
7bf9cd9d DW |
896 | /* |
897 | * Only try to free the per-AG structures if we set them up in the | |
898 | * first place. | |
899 | */ | |
2420d095 | 900 | if (xfs_is_perag_data_loaded(mp)) |
4bcd30f6 | 901 | libxfs_free_perag(mp); |
4334e2e8 ES |
902 | |
903 | kmem_free(mp->m_attr_geo); | |
904 | kmem_free(mp->m_dir_geo); | |
905 | ||
906 | kmem_free(mp->m_rtdev_targp); | |
907 | if (mp->m_logdev_targp != mp->m_ddev_targp) | |
908 | kmem_free(mp->m_logdev_targp); | |
909 | kmem_free(mp->m_ddev_targp); | |
f8149110 | 910 | |
c335b673 | 911 | return error; |
2bd0ea18 | 912 | } |
f1b058f9 NS |
913 | |
914 | /* | |
915 | * Release any global resources used by libxfs. | |
916 | */ | |
917 | void | |
a9468486 | 918 | libxfs_destroy( |
01dcfd9e | 919 | struct libxfs_init *li) |
f1b058f9 | 920 | { |
a9468486 DW |
921 | int leaked; |
922 | ||
923 | libxfs_close_devices(li); | |
44488491 | 924 | |
2e1394fc | 925 | /* Free everything from the buffer cache before freeing buffer cache */ |
864028ed ES |
926 | libxfs_bcache_purge(); |
927 | libxfs_bcache_free(); | |
f1b058f9 | 928 | cache_destroy(libxfs_bcache); |
2e1394fc | 929 | leaked = destroy_caches(); |
e4da1b16 | 930 | rcu_unregister_thread(); |
44488491 ES |
931 | if (getenv("LIBXFS_LEAK_CHECK") && leaked) |
932 | exit(1); | |
f1b058f9 | 933 | } |
9f38f08d | 934 | |
b74a1f6a NS |
935 | int |
936 | libxfs_device_alignment(void) | |
937 | { | |
938 | return platform_align_blockdev(); | |
939 | } | |
940 | ||
9f38f08d | 941 | void |
b6281496 | 942 | libxfs_report(FILE *fp) |
9f38f08d | 943 | { |
cb5b3ef4 MV |
944 | time_t t; |
945 | char *c; | |
946 | ||
b6281496 | 947 | cache_report(fp, "libxfs_bcache", libxfs_bcache); |
cb5b3ef4 MV |
948 | |
949 | t = time(NULL); | |
950 | c = asctime(localtime(&t)); | |
951 | fprintf(fp, "%s", c); | |
952 | } |