Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
7b718769 NS |
2 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. |
3 | * All Rights Reserved. | |
1da177e4 | 4 | * |
7b718769 NS |
5 | * This program is free software; you can redistribute it and/or |
6 | * modify it under the terms of the GNU General Public License as | |
1da177e4 LT |
7 | * published by the Free Software Foundation. |
8 | * | |
7b718769 NS |
9 | * This program is distributed in the hope that it would be useful, |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | * GNU General Public License for more details. | |
1da177e4 | 13 | * |
7b718769 NS |
14 | * You should have received a copy of the GNU General Public License |
15 | * along with this program; if not, write the Free Software Foundation, | |
16 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
1da177e4 | 17 | */ |
1da177e4 | 18 | #include "xfs.h" |
a844f451 | 19 | #include "xfs_fs.h" |
70a9883c | 20 | #include "xfs_shared.h" |
239880ef DC |
21 | #include "xfs_format.h" |
22 | #include "xfs_log_format.h" | |
23 | #include "xfs_trans_resv.h" | |
a844f451 | 24 | #include "xfs_bit.h" |
1da177e4 | 25 | #include "xfs_sb.h" |
1da177e4 | 26 | #include "xfs_mount.h" |
3ab78df2 | 27 | #include "xfs_defer.h" |
57062787 | 28 | #include "xfs_da_format.h" |
9a2cc41c | 29 | #include "xfs_da_btree.h" |
1da177e4 | 30 | #include "xfs_inode.h" |
a4fbe6ab | 31 | #include "xfs_dir2.h" |
a844f451 | 32 | #include "xfs_ialloc.h" |
1da177e4 LT |
33 | #include "xfs_alloc.h" |
34 | #include "xfs_rtalloc.h" | |
35 | #include "xfs_bmap.h" | |
a4fbe6ab DC |
36 | #include "xfs_trans.h" |
37 | #include "xfs_trans_priv.h" | |
38 | #include "xfs_log.h" | |
1da177e4 | 39 | #include "xfs_error.h" |
1da177e4 LT |
40 | #include "xfs_quota.h" |
41 | #include "xfs_fsops.h" | |
0b1b213f | 42 | #include "xfs_trace.h" |
6d8b79cf | 43 | #include "xfs_icache.h" |
a31b1d3d | 44 | #include "xfs_sysfs.h" |
035e00ac | 45 | #include "xfs_rmap_btree.h" |
0b1b213f | 46 | |
1da177e4 | 47 | |
27174203 CH |
48 | static DEFINE_MUTEX(xfs_uuid_table_mutex); |
49 | static int xfs_uuid_table_size; | |
50 | static uuid_t *xfs_uuid_table; | |
51 | ||
af3b6382 DW |
52 | void |
53 | xfs_uuid_table_free(void) | |
54 | { | |
55 | if (xfs_uuid_table_size == 0) | |
56 | return; | |
57 | kmem_free(xfs_uuid_table); | |
58 | xfs_uuid_table = NULL; | |
59 | xfs_uuid_table_size = 0; | |
60 | } | |
61 | ||
27174203 CH |
62 | /* |
63 | * See if the UUID is unique among mounted XFS filesystems. | |
64 | * Mount fails if UUID is nil or a FS with the same UUID is already mounted. | |
65 | */ | |
66 | STATIC int | |
67 | xfs_uuid_mount( | |
68 | struct xfs_mount *mp) | |
69 | { | |
70 | uuid_t *uuid = &mp->m_sb.sb_uuid; | |
71 | int hole, i; | |
72 | ||
73 | if (mp->m_flags & XFS_MOUNT_NOUUID) | |
74 | return 0; | |
75 | ||
76 | if (uuid_is_nil(uuid)) { | |
0b932ccc | 77 | xfs_warn(mp, "Filesystem has nil UUID - can't mount"); |
2451337d | 78 | return -EINVAL; |
27174203 CH |
79 | } |
80 | ||
81 | mutex_lock(&xfs_uuid_table_mutex); | |
82 | for (i = 0, hole = -1; i < xfs_uuid_table_size; i++) { | |
83 | if (uuid_is_nil(&xfs_uuid_table[i])) { | |
84 | hole = i; | |
85 | continue; | |
86 | } | |
87 | if (uuid_equal(uuid, &xfs_uuid_table[i])) | |
88 | goto out_duplicate; | |
89 | } | |
90 | ||
91 | if (hole < 0) { | |
92 | xfs_uuid_table = kmem_realloc(xfs_uuid_table, | |
93 | (xfs_uuid_table_size + 1) * sizeof(*xfs_uuid_table), | |
27174203 CH |
94 | KM_SLEEP); |
95 | hole = xfs_uuid_table_size++; | |
96 | } | |
97 | xfs_uuid_table[hole] = *uuid; | |
98 | mutex_unlock(&xfs_uuid_table_mutex); | |
99 | ||
100 | return 0; | |
101 | ||
102 | out_duplicate: | |
103 | mutex_unlock(&xfs_uuid_table_mutex); | |
021000e5 | 104 | xfs_warn(mp, "Filesystem has duplicate UUID %pU - can't mount", uuid); |
2451337d | 105 | return -EINVAL; |
27174203 CH |
106 | } |
107 | ||
108 | STATIC void | |
109 | xfs_uuid_unmount( | |
110 | struct xfs_mount *mp) | |
111 | { | |
112 | uuid_t *uuid = &mp->m_sb.sb_uuid; | |
113 | int i; | |
114 | ||
115 | if (mp->m_flags & XFS_MOUNT_NOUUID) | |
116 | return; | |
117 | ||
118 | mutex_lock(&xfs_uuid_table_mutex); | |
119 | for (i = 0; i < xfs_uuid_table_size; i++) { | |
120 | if (uuid_is_nil(&xfs_uuid_table[i])) | |
121 | continue; | |
122 | if (!uuid_equal(uuid, &xfs_uuid_table[i])) | |
123 | continue; | |
124 | memset(&xfs_uuid_table[i], 0, sizeof(uuid_t)); | |
125 | break; | |
126 | } | |
127 | ASSERT(i < xfs_uuid_table_size); | |
128 | mutex_unlock(&xfs_uuid_table_mutex); | |
129 | } | |
130 | ||
131 | ||
e176579e DC |
132 | STATIC void |
133 | __xfs_free_perag( | |
134 | struct rcu_head *head) | |
135 | { | |
136 | struct xfs_perag *pag = container_of(head, struct xfs_perag, rcu_head); | |
137 | ||
138 | ASSERT(atomic_read(&pag->pag_ref) == 0); | |
139 | kmem_free(pag); | |
140 | } | |
141 | ||
1da177e4 | 142 | /* |
e176579e | 143 | * Free up the per-ag resources associated with the mount structure. |
1da177e4 | 144 | */ |
c962fb79 | 145 | STATIC void |
ff4f038c | 146 | xfs_free_perag( |
745f6919 | 147 | xfs_mount_t *mp) |
1da177e4 | 148 | { |
1c1c6ebc DC |
149 | xfs_agnumber_t agno; |
150 | struct xfs_perag *pag; | |
151 | ||
152 | for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) { | |
153 | spin_lock(&mp->m_perag_lock); | |
154 | pag = radix_tree_delete(&mp->m_perag_tree, agno); | |
155 | spin_unlock(&mp->m_perag_lock); | |
e176579e | 156 | ASSERT(pag); |
f83282a8 | 157 | ASSERT(atomic_read(&pag->pag_ref) == 0); |
e176579e | 158 | call_rcu(&pag->rcu_head, __xfs_free_perag); |
1da177e4 | 159 | } |
1da177e4 LT |
160 | } |
161 | ||
4cc929ee NS |
162 | /* |
163 | * Check size of device based on the (data/realtime) block count. | |
164 | * Note: this check is used by the growfs code as well as mount. | |
165 | */ | |
166 | int | |
167 | xfs_sb_validate_fsb_count( | |
168 | xfs_sb_t *sbp, | |
169 | __uint64_t nblocks) | |
170 | { | |
171 | ASSERT(PAGE_SHIFT >= sbp->sb_blocklog); | |
172 | ASSERT(sbp->sb_blocklog >= BBSHIFT); | |
173 | ||
d5cf09ba | 174 | /* Limited by ULONG_MAX of page cache index */ |
09cbfeaf | 175 | if (nblocks >> (PAGE_SHIFT - sbp->sb_blocklog) > ULONG_MAX) |
2451337d | 176 | return -EFBIG; |
4cc929ee NS |
177 | return 0; |
178 | } | |
1da177e4 | 179 | |
1c1c6ebc | 180 | int |
c11e2c36 | 181 | xfs_initialize_perag( |
c11e2c36 | 182 | xfs_mount_t *mp, |
1c1c6ebc DC |
183 | xfs_agnumber_t agcount, |
184 | xfs_agnumber_t *maxagi) | |
1da177e4 | 185 | { |
2d2194f6 | 186 | xfs_agnumber_t index; |
8b26c582 | 187 | xfs_agnumber_t first_initialised = 0; |
1da177e4 | 188 | xfs_perag_t *pag; |
8b26c582 | 189 | int error = -ENOMEM; |
1da177e4 | 190 | |
1c1c6ebc DC |
191 | /* |
192 | * Walk the current per-ag tree so we don't try to initialise AGs | |
193 | * that already exist (growfs case). Allocate and insert all the | |
194 | * AGs we don't find ready for initialisation. | |
195 | */ | |
196 | for (index = 0; index < agcount; index++) { | |
197 | pag = xfs_perag_get(mp, index); | |
198 | if (pag) { | |
199 | xfs_perag_put(pag); | |
200 | continue; | |
201 | } | |
8b26c582 DC |
202 | if (!first_initialised) |
203 | first_initialised = index; | |
fb3b504a | 204 | |
1c1c6ebc DC |
205 | pag = kmem_zalloc(sizeof(*pag), KM_MAYFAIL); |
206 | if (!pag) | |
8b26c582 | 207 | goto out_unwind; |
fb3b504a CH |
208 | pag->pag_agno = index; |
209 | pag->pag_mount = mp; | |
1a427ab0 | 210 | spin_lock_init(&pag->pag_ici_lock); |
69b491c2 | 211 | mutex_init(&pag->pag_ici_reclaim_lock); |
fb3b504a | 212 | INIT_RADIX_TREE(&pag->pag_ici_root, GFP_ATOMIC); |
74f75a0c DC |
213 | spin_lock_init(&pag->pag_buf_lock); |
214 | pag->pag_buf_tree = RB_ROOT; | |
fb3b504a | 215 | |
1c1c6ebc | 216 | if (radix_tree_preload(GFP_NOFS)) |
8b26c582 | 217 | goto out_unwind; |
fb3b504a | 218 | |
1c1c6ebc DC |
219 | spin_lock(&mp->m_perag_lock); |
220 | if (radix_tree_insert(&mp->m_perag_tree, index, pag)) { | |
221 | BUG(); | |
222 | spin_unlock(&mp->m_perag_lock); | |
8b26c582 DC |
223 | radix_tree_preload_end(); |
224 | error = -EEXIST; | |
225 | goto out_unwind; | |
1c1c6ebc DC |
226 | } |
227 | spin_unlock(&mp->m_perag_lock); | |
228 | radix_tree_preload_end(); | |
229 | } | |
230 | ||
12c3f05c | 231 | index = xfs_set_inode_alloc(mp, agcount); |
fb3b504a | 232 | |
1c1c6ebc DC |
233 | if (maxagi) |
234 | *maxagi = index; | |
8018026e DW |
235 | |
236 | mp->m_ag_prealloc_blocks = xfs_prealloc_blocks(mp); | |
1c1c6ebc | 237 | return 0; |
8b26c582 DC |
238 | |
239 | out_unwind: | |
240 | kmem_free(pag); | |
241 | for (; index > first_initialised; index--) { | |
242 | pag = radix_tree_delete(&mp->m_perag_tree, index); | |
243 | kmem_free(pag); | |
244 | } | |
245 | return error; | |
1da177e4 LT |
246 | } |
247 | ||
1da177e4 LT |
248 | /* |
249 | * xfs_readsb | |
250 | * | |
251 | * Does the initial read of the superblock. | |
252 | */ | |
253 | int | |
ff55068c DC |
254 | xfs_readsb( |
255 | struct xfs_mount *mp, | |
256 | int flags) | |
1da177e4 LT |
257 | { |
258 | unsigned int sector_size; | |
04a1e6c5 DC |
259 | struct xfs_buf *bp; |
260 | struct xfs_sb *sbp = &mp->m_sb; | |
1da177e4 | 261 | int error; |
af34e09d | 262 | int loud = !(flags & XFS_MFSI_QUIET); |
daba5427 | 263 | const struct xfs_buf_ops *buf_ops; |
1da177e4 LT |
264 | |
265 | ASSERT(mp->m_sb_bp == NULL); | |
266 | ASSERT(mp->m_ddev_targp != NULL); | |
267 | ||
daba5427 ES |
268 | /* |
269 | * For the initial read, we must guess at the sector | |
270 | * size based on the block device. It's enough to | |
271 | * get the sb_sectsize out of the superblock and | |
272 | * then reread with the proper length. | |
273 | * We don't verify it yet, because it may not be complete. | |
274 | */ | |
275 | sector_size = xfs_getsize_buftarg(mp->m_ddev_targp); | |
276 | buf_ops = NULL; | |
277 | ||
1da177e4 | 278 | /* |
c891c30a BF |
279 | * Allocate a (locked) buffer to hold the superblock. This will be kept |
280 | * around at all times to optimize access to the superblock. Therefore, | |
281 | * set XBF_NO_IOACCT to make sure it doesn't hold the buftarg count | |
282 | * elevated. | |
1da177e4 | 283 | */ |
26af6552 | 284 | reread: |
ba372674 | 285 | error = xfs_buf_read_uncached(mp->m_ddev_targp, XFS_SB_DADDR, |
c891c30a BF |
286 | BTOBB(sector_size), XBF_NO_IOACCT, &bp, |
287 | buf_ops); | |
ba372674 | 288 | if (error) { |
eab4e633 | 289 | if (loud) |
e721f504 | 290 | xfs_warn(mp, "SB validate failed with error %d.", error); |
ac75a1f7 | 291 | /* bad CRC means corrupted metadata */ |
2451337d DC |
292 | if (error == -EFSBADCRC) |
293 | error = -EFSCORRUPTED; | |
ba372674 | 294 | return error; |
eab4e633 | 295 | } |
1da177e4 LT |
296 | |
297 | /* | |
298 | * Initialize the mount structure from the superblock. | |
1da177e4 | 299 | */ |
556b8883 | 300 | xfs_sb_from_disk(sbp, XFS_BUF_TO_SBP(bp)); |
556b8883 DC |
301 | |
302 | /* | |
303 | * If we haven't validated the superblock, do so now before we try | |
304 | * to check the sector size and reread the superblock appropriately. | |
305 | */ | |
306 | if (sbp->sb_magicnum != XFS_SB_MAGIC) { | |
307 | if (loud) | |
308 | xfs_warn(mp, "Invalid superblock magic number"); | |
2451337d | 309 | error = -EINVAL; |
556b8883 DC |
310 | goto release_buf; |
311 | } | |
ff55068c | 312 | |
1da177e4 LT |
313 | /* |
314 | * We must be able to do sector-sized and sector-aligned IO. | |
315 | */ | |
04a1e6c5 | 316 | if (sector_size > sbp->sb_sectsize) { |
af34e09d DC |
317 | if (loud) |
318 | xfs_warn(mp, "device supports %u byte sectors (not %u)", | |
04a1e6c5 | 319 | sector_size, sbp->sb_sectsize); |
2451337d | 320 | error = -ENOSYS; |
26af6552 | 321 | goto release_buf; |
1da177e4 LT |
322 | } |
323 | ||
daba5427 | 324 | if (buf_ops == NULL) { |
556b8883 DC |
325 | /* |
326 | * Re-read the superblock so the buffer is correctly sized, | |
327 | * and properly verified. | |
328 | */ | |
1da177e4 | 329 | xfs_buf_relse(bp); |
04a1e6c5 | 330 | sector_size = sbp->sb_sectsize; |
daba5427 | 331 | buf_ops = loud ? &xfs_sb_buf_ops : &xfs_sb_quiet_buf_ops; |
26af6552 | 332 | goto reread; |
1da177e4 LT |
333 | } |
334 | ||
5681ca40 | 335 | xfs_reinit_percpu_counters(mp); |
8d280b98 | 336 | |
04a1e6c5 DC |
337 | /* no need to be quiet anymore, so reset the buf ops */ |
338 | bp->b_ops = &xfs_sb_buf_ops; | |
339 | ||
1da177e4 | 340 | mp->m_sb_bp = bp; |
26af6552 | 341 | xfs_buf_unlock(bp); |
1da177e4 LT |
342 | return 0; |
343 | ||
26af6552 DC |
344 | release_buf: |
345 | xfs_buf_relse(bp); | |
1da177e4 LT |
346 | return error; |
347 | } | |
348 | ||
1da177e4 | 349 | /* |
0771fb45 | 350 | * Update alignment values based on mount options and sb values |
1da177e4 | 351 | */ |
0771fb45 | 352 | STATIC int |
7884bc86 | 353 | xfs_update_alignment(xfs_mount_t *mp) |
1da177e4 | 354 | { |
1da177e4 | 355 | xfs_sb_t *sbp = &(mp->m_sb); |
1da177e4 | 356 | |
4249023a | 357 | if (mp->m_dalign) { |
1da177e4 LT |
358 | /* |
359 | * If stripe unit and stripe width are not multiples | |
360 | * of the fs blocksize turn off alignment. | |
361 | */ | |
362 | if ((BBTOB(mp->m_dalign) & mp->m_blockmask) || | |
363 | (BBTOB(mp->m_swidth) & mp->m_blockmask)) { | |
39a45d84 JL |
364 | xfs_warn(mp, |
365 | "alignment check failed: sunit/swidth vs. blocksize(%d)", | |
366 | sbp->sb_blocksize); | |
2451337d | 367 | return -EINVAL; |
1da177e4 LT |
368 | } else { |
369 | /* | |
370 | * Convert the stripe unit and width to FSBs. | |
371 | */ | |
372 | mp->m_dalign = XFS_BB_TO_FSBT(mp, mp->m_dalign); | |
373 | if (mp->m_dalign && (sbp->sb_agblocks % mp->m_dalign)) { | |
53487786 | 374 | xfs_warn(mp, |
39a45d84 JL |
375 | "alignment check failed: sunit/swidth vs. agsize(%d)", |
376 | sbp->sb_agblocks); | |
2451337d | 377 | return -EINVAL; |
1da177e4 LT |
378 | } else if (mp->m_dalign) { |
379 | mp->m_swidth = XFS_BB_TO_FSBT(mp, mp->m_swidth); | |
380 | } else { | |
39a45d84 JL |
381 | xfs_warn(mp, |
382 | "alignment check failed: sunit(%d) less than bsize(%d)", | |
383 | mp->m_dalign, sbp->sb_blocksize); | |
2451337d | 384 | return -EINVAL; |
1da177e4 LT |
385 | } |
386 | } | |
387 | ||
388 | /* | |
389 | * Update superblock with new values | |
390 | * and log changes | |
391 | */ | |
62118709 | 392 | if (xfs_sb_version_hasdalign(sbp)) { |
1da177e4 LT |
393 | if (sbp->sb_unit != mp->m_dalign) { |
394 | sbp->sb_unit = mp->m_dalign; | |
61e63ecb | 395 | mp->m_update_sb = true; |
1da177e4 LT |
396 | } |
397 | if (sbp->sb_width != mp->m_swidth) { | |
398 | sbp->sb_width = mp->m_swidth; | |
61e63ecb | 399 | mp->m_update_sb = true; |
1da177e4 | 400 | } |
34d7f603 JL |
401 | } else { |
402 | xfs_warn(mp, | |
403 | "cannot change alignment: superblock does not support data alignment"); | |
2451337d | 404 | return -EINVAL; |
1da177e4 LT |
405 | } |
406 | } else if ((mp->m_flags & XFS_MOUNT_NOALIGN) != XFS_MOUNT_NOALIGN && | |
62118709 | 407 | xfs_sb_version_hasdalign(&mp->m_sb)) { |
1da177e4 LT |
408 | mp->m_dalign = sbp->sb_unit; |
409 | mp->m_swidth = sbp->sb_width; | |
410 | } | |
411 | ||
0771fb45 ES |
412 | return 0; |
413 | } | |
1da177e4 | 414 | |
0771fb45 ES |
415 | /* |
416 | * Set the maximum inode count for this filesystem | |
417 | */ | |
418 | STATIC void | |
419 | xfs_set_maxicount(xfs_mount_t *mp) | |
420 | { | |
421 | xfs_sb_t *sbp = &(mp->m_sb); | |
422 | __uint64_t icount; | |
1da177e4 | 423 | |
0771fb45 ES |
424 | if (sbp->sb_imax_pct) { |
425 | /* | |
426 | * Make sure the maximum inode count is a multiple | |
427 | * of the units we allocate inodes in. | |
1da177e4 | 428 | */ |
1da177e4 LT |
429 | icount = sbp->sb_dblocks * sbp->sb_imax_pct; |
430 | do_div(icount, 100); | |
431 | do_div(icount, mp->m_ialloc_blks); | |
432 | mp->m_maxicount = (icount * mp->m_ialloc_blks) << | |
433 | sbp->sb_inopblog; | |
0771fb45 | 434 | } else { |
1da177e4 | 435 | mp->m_maxicount = 0; |
1da177e4 | 436 | } |
0771fb45 ES |
437 | } |
438 | ||
439 | /* | |
440 | * Set the default minimum read and write sizes unless | |
441 | * already specified in a mount option. | |
442 | * We use smaller I/O sizes when the file system | |
443 | * is being used for NFS service (wsync mount option). | |
444 | */ | |
445 | STATIC void | |
446 | xfs_set_rw_sizes(xfs_mount_t *mp) | |
447 | { | |
448 | xfs_sb_t *sbp = &(mp->m_sb); | |
449 | int readio_log, writeio_log; | |
1da177e4 | 450 | |
1da177e4 LT |
451 | if (!(mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)) { |
452 | if (mp->m_flags & XFS_MOUNT_WSYNC) { | |
453 | readio_log = XFS_WSYNC_READIO_LOG; | |
454 | writeio_log = XFS_WSYNC_WRITEIO_LOG; | |
455 | } else { | |
456 | readio_log = XFS_READIO_LOG_LARGE; | |
457 | writeio_log = XFS_WRITEIO_LOG_LARGE; | |
458 | } | |
459 | } else { | |
460 | readio_log = mp->m_readio_log; | |
461 | writeio_log = mp->m_writeio_log; | |
462 | } | |
463 | ||
1da177e4 LT |
464 | if (sbp->sb_blocklog > readio_log) { |
465 | mp->m_readio_log = sbp->sb_blocklog; | |
466 | } else { | |
467 | mp->m_readio_log = readio_log; | |
468 | } | |
469 | mp->m_readio_blocks = 1 << (mp->m_readio_log - sbp->sb_blocklog); | |
470 | if (sbp->sb_blocklog > writeio_log) { | |
471 | mp->m_writeio_log = sbp->sb_blocklog; | |
472 | } else { | |
473 | mp->m_writeio_log = writeio_log; | |
474 | } | |
475 | mp->m_writeio_blocks = 1 << (mp->m_writeio_log - sbp->sb_blocklog); | |
0771fb45 | 476 | } |
1da177e4 | 477 | |
055388a3 DC |
478 | /* |
479 | * precalculate the low space thresholds for dynamic speculative preallocation. | |
480 | */ | |
481 | void | |
482 | xfs_set_low_space_thresholds( | |
483 | struct xfs_mount *mp) | |
484 | { | |
485 | int i; | |
486 | ||
487 | for (i = 0; i < XFS_LOWSP_MAX; i++) { | |
488 | __uint64_t space = mp->m_sb.sb_dblocks; | |
489 | ||
490 | do_div(space, 100); | |
491 | mp->m_low_space[i] = space * (i + 1); | |
492 | } | |
493 | } | |
494 | ||
495 | ||
0771fb45 ES |
496 | /* |
497 | * Set whether we're using inode alignment. | |
498 | */ | |
499 | STATIC void | |
500 | xfs_set_inoalignment(xfs_mount_t *mp) | |
501 | { | |
62118709 | 502 | if (xfs_sb_version_hasalign(&mp->m_sb) && |
1da177e4 LT |
503 | mp->m_sb.sb_inoalignmt >= |
504 | XFS_B_TO_FSBT(mp, mp->m_inode_cluster_size)) | |
505 | mp->m_inoalign_mask = mp->m_sb.sb_inoalignmt - 1; | |
506 | else | |
507 | mp->m_inoalign_mask = 0; | |
508 | /* | |
509 | * If we are using stripe alignment, check whether | |
510 | * the stripe unit is a multiple of the inode alignment | |
511 | */ | |
512 | if (mp->m_dalign && mp->m_inoalign_mask && | |
513 | !(mp->m_dalign & mp->m_inoalign_mask)) | |
514 | mp->m_sinoalign = mp->m_dalign; | |
515 | else | |
516 | mp->m_sinoalign = 0; | |
0771fb45 ES |
517 | } |
518 | ||
519 | /* | |
0471f62e | 520 | * Check that the data (and log if separate) is an ok size. |
0771fb45 ES |
521 | */ |
522 | STATIC int | |
ba372674 DC |
523 | xfs_check_sizes( |
524 | struct xfs_mount *mp) | |
0771fb45 | 525 | { |
ba372674 | 526 | struct xfs_buf *bp; |
0771fb45 | 527 | xfs_daddr_t d; |
ba372674 | 528 | int error; |
0771fb45 | 529 | |
1da177e4 LT |
530 | d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks); |
531 | if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_dblocks) { | |
0b932ccc | 532 | xfs_warn(mp, "filesystem size mismatch detected"); |
2451337d | 533 | return -EFBIG; |
1da177e4 | 534 | } |
ba372674 | 535 | error = xfs_buf_read_uncached(mp->m_ddev_targp, |
1922c949 | 536 | d - XFS_FSS_TO_BB(mp, 1), |
ba372674 DC |
537 | XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL); |
538 | if (error) { | |
0b932ccc | 539 | xfs_warn(mp, "last sector read failed"); |
ba372674 | 540 | return error; |
1da177e4 | 541 | } |
1922c949 | 542 | xfs_buf_relse(bp); |
1da177e4 | 543 | |
ba372674 DC |
544 | if (mp->m_logdev_targp == mp->m_ddev_targp) |
545 | return 0; | |
546 | ||
547 | d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_logblocks); | |
548 | if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_logblocks) { | |
549 | xfs_warn(mp, "log size mismatch detected"); | |
550 | return -EFBIG; | |
551 | } | |
552 | error = xfs_buf_read_uncached(mp->m_logdev_targp, | |
1922c949 | 553 | d - XFS_FSB_TO_BB(mp, 1), |
ba372674 DC |
554 | XFS_FSB_TO_BB(mp, 1), 0, &bp, NULL); |
555 | if (error) { | |
556 | xfs_warn(mp, "log device read failed"); | |
557 | return error; | |
0771fb45 | 558 | } |
ba372674 | 559 | xfs_buf_relse(bp); |
0771fb45 ES |
560 | return 0; |
561 | } | |
562 | ||
7d095257 CH |
563 | /* |
564 | * Clear the quotaflags in memory and in the superblock. | |
565 | */ | |
566 | int | |
567 | xfs_mount_reset_sbqflags( | |
568 | struct xfs_mount *mp) | |
569 | { | |
7d095257 CH |
570 | mp->m_qflags = 0; |
571 | ||
61e63ecb | 572 | /* It is OK to look at sb_qflags in the mount path without m_sb_lock. */ |
7d095257 CH |
573 | if (mp->m_sb.sb_qflags == 0) |
574 | return 0; | |
575 | spin_lock(&mp->m_sb_lock); | |
576 | mp->m_sb.sb_qflags = 0; | |
577 | spin_unlock(&mp->m_sb_lock); | |
578 | ||
61e63ecb | 579 | if (!xfs_fs_writable(mp, SB_FREEZE_WRITE)) |
7d095257 CH |
580 | return 0; |
581 | ||
61e63ecb | 582 | return xfs_sync_sb(mp, false); |
7d095257 CH |
583 | } |
584 | ||
d5db0f97 ES |
585 | __uint64_t |
586 | xfs_default_resblks(xfs_mount_t *mp) | |
587 | { | |
588 | __uint64_t resblks; | |
589 | ||
590 | /* | |
8babd8a2 DC |
591 | * We default to 5% or 8192 fsbs of space reserved, whichever is |
592 | * smaller. This is intended to cover concurrent allocation | |
593 | * transactions when we initially hit enospc. These each require a 4 | |
594 | * block reservation. Hence by default we cover roughly 2000 concurrent | |
595 | * allocation reservations. | |
d5db0f97 ES |
596 | */ |
597 | resblks = mp->m_sb.sb_dblocks; | |
598 | do_div(resblks, 20); | |
8babd8a2 | 599 | resblks = min_t(__uint64_t, resblks, 8192); |
d5db0f97 ES |
600 | return resblks; |
601 | } | |
602 | ||
0771fb45 | 603 | /* |
0771fb45 ES |
604 | * This function does the following on an initial mount of a file system: |
605 | * - reads the superblock from disk and init the mount struct | |
606 | * - if we're a 32-bit kernel, do a size check on the superblock | |
607 | * so we don't mount terabyte filesystems | |
608 | * - init mount struct realtime fields | |
609 | * - allocate inode hash table for fs | |
610 | * - init directory manager | |
611 | * - perform recovery and init the log manager | |
612 | */ | |
613 | int | |
614 | xfs_mountfs( | |
f0b2efad | 615 | struct xfs_mount *mp) |
0771fb45 | 616 | { |
f0b2efad BF |
617 | struct xfs_sb *sbp = &(mp->m_sb); |
618 | struct xfs_inode *rip; | |
619 | __uint64_t resblks; | |
620 | uint quotamount = 0; | |
621 | uint quotaflags = 0; | |
622 | int error = 0; | |
0771fb45 | 623 | |
ff55068c | 624 | xfs_sb_mount_common(mp, sbp); |
0771fb45 | 625 | |
ee1c0908 | 626 | /* |
074e427b DC |
627 | * Check for a mismatched features2 values. Older kernels read & wrote |
628 | * into the wrong sb offset for sb_features2 on some platforms due to | |
629 | * xfs_sb_t not being 64bit size aligned when sb_features2 was added, | |
630 | * which made older superblock reading/writing routines swap it as a | |
631 | * 64-bit value. | |
ee1c0908 | 632 | * |
e6957ea4 ES |
633 | * For backwards compatibility, we make both slots equal. |
634 | * | |
074e427b DC |
635 | * If we detect a mismatched field, we OR the set bits into the existing |
636 | * features2 field in case it has already been modified; we don't want | |
637 | * to lose any features. We then update the bad location with the ORed | |
638 | * value so that older kernels will see any features2 flags. The | |
639 | * superblock writeback code ensures the new sb_features2 is copied to | |
640 | * sb_bad_features2 before it is logged or written to disk. | |
ee1c0908 | 641 | */ |
e6957ea4 | 642 | if (xfs_sb_has_mismatched_features2(sbp)) { |
0b932ccc | 643 | xfs_warn(mp, "correcting sb_features alignment problem"); |
ee1c0908 | 644 | sbp->sb_features2 |= sbp->sb_bad_features2; |
61e63ecb | 645 | mp->m_update_sb = true; |
e6957ea4 ES |
646 | |
647 | /* | |
648 | * Re-check for ATTR2 in case it was found in bad_features2 | |
649 | * slot. | |
650 | */ | |
7c12f296 TS |
651 | if (xfs_sb_version_hasattr2(&mp->m_sb) && |
652 | !(mp->m_flags & XFS_MOUNT_NOATTR2)) | |
e6957ea4 | 653 | mp->m_flags |= XFS_MOUNT_ATTR2; |
7c12f296 TS |
654 | } |
655 | ||
656 | if (xfs_sb_version_hasattr2(&mp->m_sb) && | |
657 | (mp->m_flags & XFS_MOUNT_NOATTR2)) { | |
658 | xfs_sb_version_removeattr2(&mp->m_sb); | |
61e63ecb | 659 | mp->m_update_sb = true; |
e6957ea4 | 660 | |
7c12f296 TS |
661 | /* update sb_versionnum for the clearing of the morebits */ |
662 | if (!sbp->sb_features2) | |
61e63ecb | 663 | mp->m_update_sb = true; |
ee1c0908 DC |
664 | } |
665 | ||
263997a6 DC |
666 | /* always use v2 inodes by default now */ |
667 | if (!(mp->m_sb.sb_versionnum & XFS_SB_VERSION_NLINKBIT)) { | |
668 | mp->m_sb.sb_versionnum |= XFS_SB_VERSION_NLINKBIT; | |
61e63ecb | 669 | mp->m_update_sb = true; |
263997a6 DC |
670 | } |
671 | ||
0771fb45 ES |
672 | /* |
673 | * Check if sb_agblocks is aligned at stripe boundary | |
674 | * If sb_agblocks is NOT aligned turn off m_dalign since | |
675 | * allocator alignment is within an ag, therefore ag has | |
676 | * to be aligned at stripe boundary. | |
677 | */ | |
7884bc86 | 678 | error = xfs_update_alignment(mp); |
0771fb45 | 679 | if (error) |
f9057e3d | 680 | goto out; |
0771fb45 ES |
681 | |
682 | xfs_alloc_compute_maxlevels(mp); | |
683 | xfs_bmap_compute_maxlevels(mp, XFS_DATA_FORK); | |
684 | xfs_bmap_compute_maxlevels(mp, XFS_ATTR_FORK); | |
685 | xfs_ialloc_compute_maxlevels(mp); | |
035e00ac | 686 | xfs_rmapbt_compute_maxlevels(mp); |
0771fb45 ES |
687 | |
688 | xfs_set_maxicount(mp); | |
689 | ||
e6b3bb78 CM |
690 | /* enable fail_at_unmount as default */ |
691 | mp->m_fail_unmount = 1; | |
692 | ||
a31b1d3d | 693 | error = xfs_sysfs_init(&mp->m_kobj, &xfs_mp_ktype, NULL, mp->m_fsname); |
27174203 CH |
694 | if (error) |
695 | goto out; | |
1da177e4 | 696 | |
225e4635 BD |
697 | error = xfs_sysfs_init(&mp->m_stats.xs_kobj, &xfs_stats_ktype, |
698 | &mp->m_kobj, "stats"); | |
a31b1d3d BF |
699 | if (error) |
700 | goto out_remove_sysfs; | |
701 | ||
192852be | 702 | error = xfs_error_sysfs_init(mp); |
225e4635 BD |
703 | if (error) |
704 | goto out_del_stats; | |
705 | ||
192852be CM |
706 | |
707 | error = xfs_uuid_mount(mp); | |
708 | if (error) | |
709 | goto out_remove_error_sysfs; | |
710 | ||
0771fb45 ES |
711 | /* |
712 | * Set the minimum read and write sizes | |
713 | */ | |
714 | xfs_set_rw_sizes(mp); | |
715 | ||
055388a3 DC |
716 | /* set the low space thresholds for dynamic preallocation */ |
717 | xfs_set_low_space_thresholds(mp); | |
718 | ||
0771fb45 ES |
719 | /* |
720 | * Set the inode cluster size. | |
721 | * This may still be overridden by the file system | |
722 | * block size if it is larger than the chosen cluster size. | |
8f80587b DC |
723 | * |
724 | * For v5 filesystems, scale the cluster size with the inode size to | |
725 | * keep a constant ratio of inode per cluster buffer, but only if mkfs | |
726 | * has set the inode alignment value appropriately for larger cluster | |
727 | * sizes. | |
0771fb45 ES |
728 | */ |
729 | mp->m_inode_cluster_size = XFS_INODE_BIG_CLUSTER_SIZE; | |
8f80587b DC |
730 | if (xfs_sb_version_hascrc(&mp->m_sb)) { |
731 | int new_size = mp->m_inode_cluster_size; | |
732 | ||
733 | new_size *= mp->m_sb.sb_inodesize / XFS_DINODE_MIN_SIZE; | |
734 | if (mp->m_sb.sb_inoalignmt >= XFS_B_TO_FSBT(mp, new_size)) | |
735 | mp->m_inode_cluster_size = new_size; | |
8f80587b | 736 | } |
0771fb45 | 737 | |
e5376fc1 BF |
738 | /* |
739 | * If enabled, sparse inode chunk alignment is expected to match the | |
740 | * cluster size. Full inode chunk alignment must match the chunk size, | |
741 | * but that is checked on sb read verification... | |
742 | */ | |
743 | if (xfs_sb_version_hassparseinodes(&mp->m_sb) && | |
744 | mp->m_sb.sb_spino_align != | |
745 | XFS_B_TO_FSBT(mp, mp->m_inode_cluster_size)) { | |
746 | xfs_warn(mp, | |
747 | "Sparse inode block alignment (%u) must match cluster size (%llu).", | |
748 | mp->m_sb.sb_spino_align, | |
749 | XFS_B_TO_FSBT(mp, mp->m_inode_cluster_size)); | |
750 | error = -EINVAL; | |
751 | goto out_remove_uuid; | |
752 | } | |
753 | ||
0771fb45 ES |
754 | /* |
755 | * Set inode alignment fields | |
756 | */ | |
757 | xfs_set_inoalignment(mp); | |
758 | ||
759 | /* | |
c2bfbc9b | 760 | * Check that the data (and log if separate) is an ok size. |
0771fb45 | 761 | */ |
4249023a | 762 | error = xfs_check_sizes(mp); |
0771fb45 | 763 | if (error) |
f9057e3d | 764 | goto out_remove_uuid; |
0771fb45 | 765 | |
1da177e4 LT |
766 | /* |
767 | * Initialize realtime fields in the mount structure | |
768 | */ | |
0771fb45 ES |
769 | error = xfs_rtmount_init(mp); |
770 | if (error) { | |
0b932ccc | 771 | xfs_warn(mp, "RT mount failed"); |
f9057e3d | 772 | goto out_remove_uuid; |
1da177e4 LT |
773 | } |
774 | ||
1da177e4 LT |
775 | /* |
776 | * Copies the low order bits of the timestamp and the randomly | |
777 | * set "sequence" number out of a UUID. | |
778 | */ | |
779 | uuid_getnodeuniq(&sbp->sb_uuid, mp->m_fixedfsid); | |
780 | ||
1da177e4 LT |
781 | mp->m_dmevmask = 0; /* not persistent; set after each mount */ |
782 | ||
0650b554 DC |
783 | error = xfs_da_mount(mp); |
784 | if (error) { | |
785 | xfs_warn(mp, "Failed dir/attr init: %d", error); | |
786 | goto out_remove_uuid; | |
787 | } | |
1da177e4 LT |
788 | |
789 | /* | |
790 | * Initialize the precomputed transaction reservations values. | |
791 | */ | |
792 | xfs_trans_init(mp); | |
793 | ||
1da177e4 LT |
794 | /* |
795 | * Allocate and initialize the per-ag data. | |
796 | */ | |
1c1c6ebc | 797 | spin_lock_init(&mp->m_perag_lock); |
9b98b6f3 | 798 | INIT_RADIX_TREE(&mp->m_perag_tree, GFP_ATOMIC); |
1c1c6ebc DC |
799 | error = xfs_initialize_perag(mp, sbp->sb_agcount, &mp->m_maxagi); |
800 | if (error) { | |
0b932ccc | 801 | xfs_warn(mp, "Failed per-ag init: %d", error); |
0650b554 | 802 | goto out_free_dir; |
1c1c6ebc | 803 | } |
1da177e4 | 804 | |
f9057e3d | 805 | if (!sbp->sb_logblocks) { |
0b932ccc | 806 | xfs_warn(mp, "no log defined"); |
f9057e3d | 807 | XFS_ERROR_REPORT("xfs_mountfs", XFS_ERRLEVEL_LOW, mp); |
2451337d | 808 | error = -EFSCORRUPTED; |
f9057e3d CH |
809 | goto out_free_perag; |
810 | } | |
811 | ||
1da177e4 | 812 | /* |
f0b2efad BF |
813 | * Log's mount-time initialization. The first part of recovery can place |
814 | * some items on the AIL, to be handled when recovery is finished or | |
815 | * cancelled. | |
1da177e4 | 816 | */ |
f9057e3d CH |
817 | error = xfs_log_mount(mp, mp->m_logdev_targp, |
818 | XFS_FSB_TO_DADDR(mp, sbp->sb_logstart), | |
819 | XFS_FSB_TO_BB(mp, sbp->sb_logblocks)); | |
820 | if (error) { | |
0b932ccc | 821 | xfs_warn(mp, "log mount failed"); |
d4f3512b | 822 | goto out_fail_wait; |
1da177e4 LT |
823 | } |
824 | ||
92821e2b DC |
825 | /* |
826 | * Now the log is mounted, we know if it was an unclean shutdown or | |
827 | * not. If it was, with the first phase of recovery has completed, we | |
828 | * have consistent AG blocks on disk. We have not recovered EFIs yet, | |
829 | * but they are recovered transactionally in the second recovery phase | |
830 | * later. | |
831 | * | |
832 | * Hence we can safely re-initialise incore superblock counters from | |
833 | * the per-ag data. These may not be correct if the filesystem was not | |
834 | * cleanly unmounted, so we need to wait for recovery to finish before | |
835 | * doing this. | |
836 | * | |
837 | * If the filesystem was cleanly unmounted, then we can trust the | |
838 | * values in the superblock to be correct and we don't need to do | |
839 | * anything here. | |
840 | * | |
841 | * If we are currently making the filesystem, the initialisation will | |
842 | * fail as the perag data is in an undefined state. | |
843 | */ | |
92821e2b DC |
844 | if (xfs_sb_version_haslazysbcount(&mp->m_sb) && |
845 | !XFS_LAST_UNMOUNT_WAS_CLEAN(mp) && | |
846 | !mp->m_sb.sb_inprogress) { | |
847 | error = xfs_initialize_perag_data(mp, sbp->sb_agcount); | |
f9057e3d | 848 | if (error) |
6eee8972 | 849 | goto out_log_dealloc; |
92821e2b | 850 | } |
f9057e3d | 851 | |
1da177e4 LT |
852 | /* |
853 | * Get and sanity-check the root inode. | |
854 | * Save the pointer to it in the mount structure. | |
855 | */ | |
7b6259e7 | 856 | error = xfs_iget(mp, NULL, sbp->sb_rootino, 0, XFS_ILOCK_EXCL, &rip); |
1da177e4 | 857 | if (error) { |
0b932ccc | 858 | xfs_warn(mp, "failed to read root inode"); |
f9057e3d | 859 | goto out_log_dealloc; |
1da177e4 LT |
860 | } |
861 | ||
862 | ASSERT(rip != NULL); | |
1da177e4 | 863 | |
c19b3b05 | 864 | if (unlikely(!S_ISDIR(VFS_I(rip)->i_mode))) { |
0b932ccc | 865 | xfs_warn(mp, "corrupted root inode %llu: not a directory", |
b6574520 | 866 | (unsigned long long)rip->i_ino); |
1da177e4 LT |
867 | xfs_iunlock(rip, XFS_ILOCK_EXCL); |
868 | XFS_ERROR_REPORT("xfs_mountfs_int(2)", XFS_ERRLEVEL_LOW, | |
869 | mp); | |
2451337d | 870 | error = -EFSCORRUPTED; |
f9057e3d | 871 | goto out_rele_rip; |
1da177e4 LT |
872 | } |
873 | mp->m_rootip = rip; /* save it */ | |
874 | ||
875 | xfs_iunlock(rip, XFS_ILOCK_EXCL); | |
876 | ||
877 | /* | |
878 | * Initialize realtime inode pointers in the mount structure | |
879 | */ | |
0771fb45 ES |
880 | error = xfs_rtmount_inodes(mp); |
881 | if (error) { | |
1da177e4 LT |
882 | /* |
883 | * Free up the root inode. | |
884 | */ | |
0b932ccc | 885 | xfs_warn(mp, "failed to read RT inodes"); |
f9057e3d | 886 | goto out_rele_rip; |
1da177e4 LT |
887 | } |
888 | ||
889 | /* | |
7884bc86 CH |
890 | * If this is a read-only mount defer the superblock updates until |
891 | * the next remount into writeable mode. Otherwise we would never | |
892 | * perform the update e.g. for the root filesystem. | |
1da177e4 | 893 | */ |
61e63ecb DC |
894 | if (mp->m_update_sb && !(mp->m_flags & XFS_MOUNT_RDONLY)) { |
895 | error = xfs_sync_sb(mp, false); | |
e5720eec | 896 | if (error) { |
0b932ccc | 897 | xfs_warn(mp, "failed to write sb changes"); |
b93b6e43 | 898 | goto out_rtunmount; |
e5720eec DC |
899 | } |
900 | } | |
1da177e4 LT |
901 | |
902 | /* | |
903 | * Initialise the XFS quota management subsystem for this mount | |
904 | */ | |
7d095257 CH |
905 | if (XFS_IS_QUOTA_RUNNING(mp)) { |
906 | error = xfs_qm_newmount(mp, "amount, "aflags); | |
907 | if (error) | |
908 | goto out_rtunmount; | |
909 | } else { | |
910 | ASSERT(!XFS_IS_QUOTA_ON(mp)); | |
911 | ||
912 | /* | |
913 | * If a file system had quotas running earlier, but decided to | |
914 | * mount without -o uquota/pquota/gquota options, revoke the | |
915 | * quotachecked license. | |
916 | */ | |
917 | if (mp->m_sb.sb_qflags & XFS_ALL_QUOTA_ACCT) { | |
0b932ccc | 918 | xfs_notice(mp, "resetting quota flags"); |
7d095257 CH |
919 | error = xfs_mount_reset_sbqflags(mp); |
920 | if (error) | |
a70a4fa5 | 921 | goto out_rtunmount; |
7d095257 CH |
922 | } |
923 | } | |
1da177e4 LT |
924 | |
925 | /* | |
f0b2efad BF |
926 | * Finish recovering the file system. This part needed to be delayed |
927 | * until after the root and real-time bitmap inodes were consistently | |
928 | * read in. | |
1da177e4 | 929 | */ |
4249023a | 930 | error = xfs_log_mount_finish(mp); |
1da177e4 | 931 | if (error) { |
0b932ccc | 932 | xfs_warn(mp, "log mount finish failed"); |
b93b6e43 | 933 | goto out_rtunmount; |
1da177e4 LT |
934 | } |
935 | ||
ddeb14f4 DC |
936 | /* |
937 | * Now the log is fully replayed, we can transition to full read-only | |
938 | * mode for read-only mounts. This will sync all the metadata and clean | |
939 | * the log so that the recovery we just performed does not have to be | |
940 | * replayed again on the next mount. | |
941 | * | |
942 | * We use the same quiesce mechanism as the rw->ro remount, as they are | |
943 | * semantically identical operations. | |
944 | */ | |
945 | if ((mp->m_flags & (XFS_MOUNT_RDONLY|XFS_MOUNT_NORECOVERY)) == | |
946 | XFS_MOUNT_RDONLY) { | |
947 | xfs_quiesce_attr(mp); | |
948 | } | |
949 | ||
1da177e4 LT |
950 | /* |
951 | * Complete the quota initialisation, post-log-replay component. | |
952 | */ | |
7d095257 CH |
953 | if (quotamount) { |
954 | ASSERT(mp->m_qflags == 0); | |
955 | mp->m_qflags = quotaflags; | |
956 | ||
957 | xfs_qm_mount_quotas(mp); | |
958 | } | |
959 | ||
84e1e99f DC |
960 | /* |
961 | * Now we are mounted, reserve a small amount of unused space for | |
962 | * privileged transactions. This is needed so that transaction | |
963 | * space required for critical operations can dip into this pool | |
964 | * when at ENOSPC. This is needed for operations like create with | |
965 | * attr, unwritten extent conversion at ENOSPC, etc. Data allocations | |
966 | * are not allowed to use this reserved space. | |
8babd8a2 DC |
967 | * |
968 | * This may drive us straight to ENOSPC on mount, but that implies | |
969 | * we were already there on the last unmount. Warn if this occurs. | |
84e1e99f | 970 | */ |
d5db0f97 ES |
971 | if (!(mp->m_flags & XFS_MOUNT_RDONLY)) { |
972 | resblks = xfs_default_resblks(mp); | |
973 | error = xfs_reserve_blocks(mp, &resblks, NULL); | |
974 | if (error) | |
0b932ccc DC |
975 | xfs_warn(mp, |
976 | "Unable to allocate reserve blocks. Continuing without reserve pool."); | |
d5db0f97 | 977 | } |
84e1e99f | 978 | |
1da177e4 LT |
979 | return 0; |
980 | ||
b93b6e43 CH |
981 | out_rtunmount: |
982 | xfs_rtunmount_inodes(mp); | |
f9057e3d | 983 | out_rele_rip: |
43355099 | 984 | IRELE(rip); |
0ae120f8 BF |
985 | cancel_delayed_work_sync(&mp->m_reclaim_work); |
986 | xfs_reclaim_inodes(mp, SYNC_WAIT); | |
f9057e3d | 987 | out_log_dealloc: |
e6b3bb78 | 988 | mp->m_flags |= XFS_MOUNT_UNMOUNTING; |
f0b2efad | 989 | xfs_log_mount_cancel(mp); |
d4f3512b DC |
990 | out_fail_wait: |
991 | if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) | |
992 | xfs_wait_buftarg(mp->m_logdev_targp); | |
993 | xfs_wait_buftarg(mp->m_ddev_targp); | |
f9057e3d | 994 | out_free_perag: |
ff4f038c | 995 | xfs_free_perag(mp); |
0650b554 DC |
996 | out_free_dir: |
997 | xfs_da_unmount(mp); | |
f9057e3d | 998 | out_remove_uuid: |
27174203 | 999 | xfs_uuid_unmount(mp); |
192852be CM |
1000 | out_remove_error_sysfs: |
1001 | xfs_error_sysfs_del(mp); | |
225e4635 BD |
1002 | out_del_stats: |
1003 | xfs_sysfs_del(&mp->m_stats.xs_kobj); | |
a31b1d3d BF |
1004 | out_remove_sysfs: |
1005 | xfs_sysfs_del(&mp->m_kobj); | |
f9057e3d | 1006 | out: |
1da177e4 LT |
1007 | return error; |
1008 | } | |
1009 | ||
1010 | /* | |
1da177e4 LT |
1011 | * This flushes out the inodes,dquots and the superblock, unmounts the |
1012 | * log and makes sure that incore structures are freed. | |
1013 | */ | |
41b5c2e7 CH |
1014 | void |
1015 | xfs_unmountfs( | |
1016 | struct xfs_mount *mp) | |
1da177e4 | 1017 | { |
41b5c2e7 CH |
1018 | __uint64_t resblks; |
1019 | int error; | |
1da177e4 | 1020 | |
579b62fa BF |
1021 | cancel_delayed_work_sync(&mp->m_eofblocks_work); |
1022 | ||
7d095257 | 1023 | xfs_qm_unmount_quotas(mp); |
b93b6e43 | 1024 | xfs_rtunmount_inodes(mp); |
77508ec8 CH |
1025 | IRELE(mp->m_rootip); |
1026 | ||
641c56fb DC |
1027 | /* |
1028 | * We can potentially deadlock here if we have an inode cluster | |
9da096fd | 1029 | * that has been freed has its buffer still pinned in memory because |
641c56fb DC |
1030 | * the transaction is still sitting in a iclog. The stale inodes |
1031 | * on that buffer will have their flush locks held until the | |
1032 | * transaction hits the disk and the callbacks run. the inode | |
1033 | * flush takes the flush lock unconditionally and with nothing to | |
1034 | * push out the iclog we will never get that unlocked. hence we | |
1035 | * need to force the log first. | |
1036 | */ | |
a14a348b | 1037 | xfs_log_force(mp, XFS_LOG_SYNC); |
c854363e | 1038 | |
e6b3bb78 CM |
1039 | /* |
1040 | * We now need to tell the world we are unmounting. This will allow | |
1041 | * us to detect that the filesystem is going away and we should error | |
1042 | * out anything that we have been retrying in the background. This will | |
1043 | * prevent neverending retries in AIL pushing from hanging the unmount. | |
1044 | */ | |
1045 | mp->m_flags |= XFS_MOUNT_UNMOUNTING; | |
1046 | ||
c854363e | 1047 | /* |
211e4d43 CH |
1048 | * Flush all pending changes from the AIL. |
1049 | */ | |
1050 | xfs_ail_push_all_sync(mp->m_ail); | |
1051 | ||
1052 | /* | |
1053 | * And reclaim all inodes. At this point there should be no dirty | |
7e18530b DC |
1054 | * inodes and none should be pinned or locked, but use synchronous |
1055 | * reclaim just to be sure. We can stop background inode reclaim | |
1056 | * here as well if it is still running. | |
c854363e | 1057 | */ |
7e18530b | 1058 | cancel_delayed_work_sync(&mp->m_reclaim_work); |
c854363e | 1059 | xfs_reclaim_inodes(mp, SYNC_WAIT); |
1da177e4 | 1060 | |
7d095257 | 1061 | xfs_qm_unmount(mp); |
a357a121 | 1062 | |
84e1e99f DC |
1063 | /* |
1064 | * Unreserve any blocks we have so that when we unmount we don't account | |
1065 | * the reserved free space as used. This is really only necessary for | |
1066 | * lazy superblock counting because it trusts the incore superblock | |
9da096fd | 1067 | * counters to be absolutely correct on clean unmount. |
84e1e99f DC |
1068 | * |
1069 | * We don't bother correcting this elsewhere for lazy superblock | |
1070 | * counting because on mount of an unclean filesystem we reconstruct the | |
1071 | * correct counter value and this is irrelevant. | |
1072 | * | |
1073 | * For non-lazy counter filesystems, this doesn't matter at all because | |
1074 | * we only every apply deltas to the superblock and hence the incore | |
1075 | * value does not matter.... | |
1076 | */ | |
1077 | resblks = 0; | |
714082bc DC |
1078 | error = xfs_reserve_blocks(mp, &resblks, NULL); |
1079 | if (error) | |
0b932ccc | 1080 | xfs_warn(mp, "Unable to free reserved block pool. " |
714082bc DC |
1081 | "Freespace may not be correct on next mount."); |
1082 | ||
adab0f67 | 1083 | error = xfs_log_sbcount(mp); |
e5720eec | 1084 | if (error) |
0b932ccc | 1085 | xfs_warn(mp, "Unable to update superblock counters. " |
e5720eec | 1086 | "Freespace may not be correct on next mount."); |
87c7bec7 | 1087 | |
225e4635 | 1088 | |
21b699c8 | 1089 | xfs_log_unmount(mp); |
0650b554 | 1090 | xfs_da_unmount(mp); |
27174203 | 1091 | xfs_uuid_unmount(mp); |
1da177e4 | 1092 | |
1550d0b0 | 1093 | #if defined(DEBUG) |
0ce4cfd4 | 1094 | xfs_errortag_clearall(mp, 0); |
1da177e4 | 1095 | #endif |
ff4f038c | 1096 | xfs_free_perag(mp); |
a31b1d3d | 1097 | |
192852be | 1098 | xfs_error_sysfs_del(mp); |
225e4635 | 1099 | xfs_sysfs_del(&mp->m_stats.xs_kobj); |
a31b1d3d | 1100 | xfs_sysfs_del(&mp->m_kobj); |
1da177e4 LT |
1101 | } |
1102 | ||
91ee575f BF |
1103 | /* |
1104 | * Determine whether modifications can proceed. The caller specifies the minimum | |
1105 | * freeze level for which modifications should not be allowed. This allows | |
1106 | * certain operations to proceed while the freeze sequence is in progress, if | |
1107 | * necessary. | |
1108 | */ | |
1109 | bool | |
1110 | xfs_fs_writable( | |
1111 | struct xfs_mount *mp, | |
1112 | int level) | |
92821e2b | 1113 | { |
91ee575f BF |
1114 | ASSERT(level > SB_UNFROZEN); |
1115 | if ((mp->m_super->s_writers.frozen >= level) || | |
1116 | XFS_FORCED_SHUTDOWN(mp) || (mp->m_flags & XFS_MOUNT_RDONLY)) | |
1117 | return false; | |
1118 | ||
1119 | return true; | |
92821e2b DC |
1120 | } |
1121 | ||
1122 | /* | |
b2ce3974 AE |
1123 | * xfs_log_sbcount |
1124 | * | |
adab0f67 | 1125 | * Sync the superblock counters to disk. |
b2ce3974 | 1126 | * |
91ee575f BF |
1127 | * Note this code can be called during the process of freezing, so we use the |
1128 | * transaction allocator that does not block when the transaction subsystem is | |
1129 | * in its frozen state. | |
92821e2b DC |
1130 | */ |
1131 | int | |
adab0f67 | 1132 | xfs_log_sbcount(xfs_mount_t *mp) |
92821e2b | 1133 | { |
91ee575f BF |
1134 | /* allow this to proceed during the freeze sequence... */ |
1135 | if (!xfs_fs_writable(mp, SB_FREEZE_COMPLETE)) | |
92821e2b DC |
1136 | return 0; |
1137 | ||
92821e2b DC |
1138 | /* |
1139 | * we don't need to do this if we are updating the superblock | |
1140 | * counters on every modification. | |
1141 | */ | |
1142 | if (!xfs_sb_version_haslazysbcount(&mp->m_sb)) | |
1143 | return 0; | |
1144 | ||
61e63ecb | 1145 | return xfs_sync_sb(mp, true); |
92821e2b DC |
1146 | } |
1147 | ||
8c1903d3 DC |
1148 | /* |
1149 | * Deltas for the inode count are +/-64, hence we use a large batch size | |
1150 | * of 128 so we don't need to take the counter lock on every update. | |
1151 | */ | |
1152 | #define XFS_ICOUNT_BATCH 128 | |
501ab323 DC |
1153 | int |
1154 | xfs_mod_icount( | |
1155 | struct xfs_mount *mp, | |
1156 | int64_t delta) | |
1157 | { | |
8c1903d3 DC |
1158 | __percpu_counter_add(&mp->m_icount, delta, XFS_ICOUNT_BATCH); |
1159 | if (__percpu_counter_compare(&mp->m_icount, 0, XFS_ICOUNT_BATCH) < 0) { | |
501ab323 DC |
1160 | ASSERT(0); |
1161 | percpu_counter_add(&mp->m_icount, -delta); | |
1162 | return -EINVAL; | |
1163 | } | |
1164 | return 0; | |
1165 | } | |
1166 | ||
e88b64ea DC |
1167 | int |
1168 | xfs_mod_ifree( | |
1169 | struct xfs_mount *mp, | |
1170 | int64_t delta) | |
1171 | { | |
1172 | percpu_counter_add(&mp->m_ifree, delta); | |
1173 | if (percpu_counter_compare(&mp->m_ifree, 0) < 0) { | |
1174 | ASSERT(0); | |
1175 | percpu_counter_add(&mp->m_ifree, -delta); | |
1176 | return -EINVAL; | |
1177 | } | |
1178 | return 0; | |
1179 | } | |
0d485ada | 1180 | |
8c1903d3 DC |
1181 | /* |
1182 | * Deltas for the block count can vary from 1 to very large, but lock contention | |
1183 | * only occurs on frequent small block count updates such as in the delayed | |
1184 | * allocation path for buffered writes (page a time updates). Hence we set | |
1185 | * a large batch count (1024) to minimise global counter updates except when | |
1186 | * we get near to ENOSPC and we have to be very accurate with our updates. | |
1187 | */ | |
1188 | #define XFS_FDBLOCKS_BATCH 1024 | |
0d485ada DC |
1189 | int |
1190 | xfs_mod_fdblocks( | |
1191 | struct xfs_mount *mp, | |
1192 | int64_t delta, | |
1193 | bool rsvd) | |
1194 | { | |
1195 | int64_t lcounter; | |
1196 | long long res_used; | |
1197 | s32 batch; | |
1198 | ||
1199 | if (delta > 0) { | |
1200 | /* | |
1201 | * If the reserve pool is depleted, put blocks back into it | |
1202 | * first. Most of the time the pool is full. | |
1203 | */ | |
1204 | if (likely(mp->m_resblks == mp->m_resblks_avail)) { | |
1205 | percpu_counter_add(&mp->m_fdblocks, delta); | |
1206 | return 0; | |
1207 | } | |
1208 | ||
1209 | spin_lock(&mp->m_sb_lock); | |
1210 | res_used = (long long)(mp->m_resblks - mp->m_resblks_avail); | |
1211 | ||
1212 | if (res_used > delta) { | |
1213 | mp->m_resblks_avail += delta; | |
1214 | } else { | |
1215 | delta -= res_used; | |
1216 | mp->m_resblks_avail = mp->m_resblks; | |
1217 | percpu_counter_add(&mp->m_fdblocks, delta); | |
1218 | } | |
1219 | spin_unlock(&mp->m_sb_lock); | |
1220 | return 0; | |
1221 | } | |
1222 | ||
1223 | /* | |
1224 | * Taking blocks away, need to be more accurate the closer we | |
1225 | * are to zero. | |
1226 | * | |
0d485ada DC |
1227 | * If the counter has a value of less than 2 * max batch size, |
1228 | * then make everything serialise as we are real close to | |
1229 | * ENOSPC. | |
1230 | */ | |
8c1903d3 DC |
1231 | if (__percpu_counter_compare(&mp->m_fdblocks, 2 * XFS_FDBLOCKS_BATCH, |
1232 | XFS_FDBLOCKS_BATCH) < 0) | |
0d485ada DC |
1233 | batch = 1; |
1234 | else | |
8c1903d3 | 1235 | batch = XFS_FDBLOCKS_BATCH; |
0d485ada DC |
1236 | |
1237 | __percpu_counter_add(&mp->m_fdblocks, delta, batch); | |
52548852 | 1238 | if (__percpu_counter_compare(&mp->m_fdblocks, mp->m_alloc_set_aside, |
8c1903d3 | 1239 | XFS_FDBLOCKS_BATCH) >= 0) { |
0d485ada DC |
1240 | /* we had space! */ |
1241 | return 0; | |
1242 | } | |
1243 | ||
1244 | /* | |
1245 | * lock up the sb for dipping into reserves before releasing the space | |
1246 | * that took us to ENOSPC. | |
1247 | */ | |
1248 | spin_lock(&mp->m_sb_lock); | |
1249 | percpu_counter_add(&mp->m_fdblocks, -delta); | |
1250 | if (!rsvd) | |
1251 | goto fdblocks_enospc; | |
1252 | ||
1253 | lcounter = (long long)mp->m_resblks_avail + delta; | |
1254 | if (lcounter >= 0) { | |
1255 | mp->m_resblks_avail = lcounter; | |
1256 | spin_unlock(&mp->m_sb_lock); | |
1257 | return 0; | |
1258 | } | |
1259 | printk_once(KERN_WARNING | |
1260 | "Filesystem \"%s\": reserve blocks depleted! " | |
1261 | "Consider increasing reserve pool size.", | |
1262 | mp->m_fsname); | |
1263 | fdblocks_enospc: | |
1264 | spin_unlock(&mp->m_sb_lock); | |
1265 | return -ENOSPC; | |
1266 | } | |
1267 | ||
bab98bbe DC |
1268 | int |
1269 | xfs_mod_frextents( | |
1270 | struct xfs_mount *mp, | |
1271 | int64_t delta) | |
1272 | { | |
1273 | int64_t lcounter; | |
1274 | int ret = 0; | |
1275 | ||
1276 | spin_lock(&mp->m_sb_lock); | |
1277 | lcounter = mp->m_sb.sb_frextents + delta; | |
1278 | if (lcounter < 0) | |
1279 | ret = -ENOSPC; | |
1280 | else | |
1281 | mp->m_sb.sb_frextents = lcounter; | |
1282 | spin_unlock(&mp->m_sb_lock); | |
1283 | return ret; | |
1284 | } | |
1285 | ||
1da177e4 LT |
1286 | /* |
1287 | * xfs_getsb() is called to obtain the buffer for the superblock. | |
1288 | * The buffer is returned locked and read in from disk. | |
1289 | * The buffer should be released with a call to xfs_brelse(). | |
1290 | * | |
1291 | * If the flags parameter is BUF_TRYLOCK, then we'll only return | |
1292 | * the superblock buffer if it can be locked without sleeping. | |
1293 | * If it can't then we'll return NULL. | |
1294 | */ | |
0c842ad4 | 1295 | struct xfs_buf * |
1da177e4 | 1296 | xfs_getsb( |
0c842ad4 CH |
1297 | struct xfs_mount *mp, |
1298 | int flags) | |
1da177e4 | 1299 | { |
0c842ad4 | 1300 | struct xfs_buf *bp = mp->m_sb_bp; |
1da177e4 | 1301 | |
0c842ad4 CH |
1302 | if (!xfs_buf_trylock(bp)) { |
1303 | if (flags & XBF_TRYLOCK) | |
1da177e4 | 1304 | return NULL; |
0c842ad4 | 1305 | xfs_buf_lock(bp); |
1da177e4 | 1306 | } |
0c842ad4 | 1307 | |
72790aa1 | 1308 | xfs_buf_hold(bp); |
b0388bf1 | 1309 | ASSERT(bp->b_flags & XBF_DONE); |
014c2544 | 1310 | return bp; |
1da177e4 LT |
1311 | } |
1312 | ||
1313 | /* | |
1314 | * Used to free the superblock along various error paths. | |
1315 | */ | |
1316 | void | |
1317 | xfs_freesb( | |
26af6552 | 1318 | struct xfs_mount *mp) |
1da177e4 | 1319 | { |
26af6552 | 1320 | struct xfs_buf *bp = mp->m_sb_bp; |
1da177e4 | 1321 | |
26af6552 | 1322 | xfs_buf_lock(bp); |
1da177e4 | 1323 | mp->m_sb_bp = NULL; |
26af6552 | 1324 | xfs_buf_relse(bp); |
1da177e4 LT |
1325 | } |
1326 | ||
dda35b8f CH |
1327 | /* |
1328 | * If the underlying (data/log/rt) device is readonly, there are some | |
1329 | * operations that cannot proceed. | |
1330 | */ | |
1331 | int | |
1332 | xfs_dev_is_read_only( | |
1333 | struct xfs_mount *mp, | |
1334 | char *message) | |
1335 | { | |
1336 | if (xfs_readonly_buftarg(mp->m_ddev_targp) || | |
1337 | xfs_readonly_buftarg(mp->m_logdev_targp) || | |
1338 | (mp->m_rtdev_targp && xfs_readonly_buftarg(mp->m_rtdev_targp))) { | |
0b932ccc DC |
1339 | xfs_notice(mp, "%s required on read-only device.", message); |
1340 | xfs_notice(mp, "write access unavailable, cannot proceed."); | |
2451337d | 1341 | return -EROFS; |
dda35b8f CH |
1342 | } |
1343 | return 0; | |
1344 | } |