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" |
1946b91c | 46 | #include "xfs_refcount_btree.h" |
174edb0e | 47 | #include "xfs_reflink.h" |
ebf55872 | 48 | #include "xfs_extent_busy.h" |
0b1b213f | 49 | |
1da177e4 | 50 | |
27174203 CH |
51 | static DEFINE_MUTEX(xfs_uuid_table_mutex); |
52 | static int xfs_uuid_table_size; | |
53 | static uuid_t *xfs_uuid_table; | |
54 | ||
af3b6382 DW |
55 | void |
56 | xfs_uuid_table_free(void) | |
57 | { | |
58 | if (xfs_uuid_table_size == 0) | |
59 | return; | |
60 | kmem_free(xfs_uuid_table); | |
61 | xfs_uuid_table = NULL; | |
62 | xfs_uuid_table_size = 0; | |
63 | } | |
64 | ||
27174203 CH |
65 | /* |
66 | * See if the UUID is unique among mounted XFS filesystems. | |
67 | * Mount fails if UUID is nil or a FS with the same UUID is already mounted. | |
68 | */ | |
69 | STATIC int | |
70 | xfs_uuid_mount( | |
71 | struct xfs_mount *mp) | |
72 | { | |
73 | uuid_t *uuid = &mp->m_sb.sb_uuid; | |
74 | int hole, i; | |
75 | ||
8f720d9f | 76 | /* Publish UUID in struct super_block */ |
85787090 | 77 | uuid_copy(&mp->m_super->s_uuid, uuid); |
8f720d9f | 78 | |
27174203 CH |
79 | if (mp->m_flags & XFS_MOUNT_NOUUID) |
80 | return 0; | |
81 | ||
d905fdaa AG |
82 | if (uuid_is_null(uuid)) { |
83 | xfs_warn(mp, "Filesystem has null UUID - can't mount"); | |
2451337d | 84 | return -EINVAL; |
27174203 CH |
85 | } |
86 | ||
87 | mutex_lock(&xfs_uuid_table_mutex); | |
88 | for (i = 0, hole = -1; i < xfs_uuid_table_size; i++) { | |
d905fdaa | 89 | if (uuid_is_null(&xfs_uuid_table[i])) { |
27174203 CH |
90 | hole = i; |
91 | continue; | |
92 | } | |
93 | if (uuid_equal(uuid, &xfs_uuid_table[i])) | |
94 | goto out_duplicate; | |
95 | } | |
96 | ||
97 | if (hole < 0) { | |
98 | xfs_uuid_table = kmem_realloc(xfs_uuid_table, | |
99 | (xfs_uuid_table_size + 1) * sizeof(*xfs_uuid_table), | |
27174203 CH |
100 | KM_SLEEP); |
101 | hole = xfs_uuid_table_size++; | |
102 | } | |
103 | xfs_uuid_table[hole] = *uuid; | |
104 | mutex_unlock(&xfs_uuid_table_mutex); | |
105 | ||
106 | return 0; | |
107 | ||
108 | out_duplicate: | |
109 | mutex_unlock(&xfs_uuid_table_mutex); | |
021000e5 | 110 | xfs_warn(mp, "Filesystem has duplicate UUID %pU - can't mount", uuid); |
2451337d | 111 | return -EINVAL; |
27174203 CH |
112 | } |
113 | ||
114 | STATIC void | |
115 | xfs_uuid_unmount( | |
116 | struct xfs_mount *mp) | |
117 | { | |
118 | uuid_t *uuid = &mp->m_sb.sb_uuid; | |
119 | int i; | |
120 | ||
121 | if (mp->m_flags & XFS_MOUNT_NOUUID) | |
122 | return; | |
123 | ||
124 | mutex_lock(&xfs_uuid_table_mutex); | |
125 | for (i = 0; i < xfs_uuid_table_size; i++) { | |
d905fdaa | 126 | if (uuid_is_null(&xfs_uuid_table[i])) |
27174203 CH |
127 | continue; |
128 | if (!uuid_equal(uuid, &xfs_uuid_table[i])) | |
129 | continue; | |
130 | memset(&xfs_uuid_table[i], 0, sizeof(uuid_t)); | |
131 | break; | |
132 | } | |
133 | ASSERT(i < xfs_uuid_table_size); | |
134 | mutex_unlock(&xfs_uuid_table_mutex); | |
135 | } | |
136 | ||
137 | ||
e176579e DC |
138 | STATIC void |
139 | __xfs_free_perag( | |
140 | struct rcu_head *head) | |
141 | { | |
142 | struct xfs_perag *pag = container_of(head, struct xfs_perag, rcu_head); | |
143 | ||
144 | ASSERT(atomic_read(&pag->pag_ref) == 0); | |
145 | kmem_free(pag); | |
146 | } | |
147 | ||
1da177e4 | 148 | /* |
e176579e | 149 | * Free up the per-ag resources associated with the mount structure. |
1da177e4 | 150 | */ |
c962fb79 | 151 | STATIC void |
ff4f038c | 152 | xfs_free_perag( |
745f6919 | 153 | xfs_mount_t *mp) |
1da177e4 | 154 | { |
1c1c6ebc DC |
155 | xfs_agnumber_t agno; |
156 | struct xfs_perag *pag; | |
157 | ||
158 | for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) { | |
159 | spin_lock(&mp->m_perag_lock); | |
160 | pag = radix_tree_delete(&mp->m_perag_tree, agno); | |
161 | spin_unlock(&mp->m_perag_lock); | |
e176579e | 162 | ASSERT(pag); |
f83282a8 | 163 | ASSERT(atomic_read(&pag->pag_ref) == 0); |
6031e73a | 164 | xfs_buf_hash_destroy(pag); |
1da06189 | 165 | mutex_destroy(&pag->pag_ici_reclaim_lock); |
e176579e | 166 | call_rcu(&pag->rcu_head, __xfs_free_perag); |
1da177e4 | 167 | } |
1da177e4 LT |
168 | } |
169 | ||
4cc929ee NS |
170 | /* |
171 | * Check size of device based on the (data/realtime) block count. | |
172 | * Note: this check is used by the growfs code as well as mount. | |
173 | */ | |
174 | int | |
175 | xfs_sb_validate_fsb_count( | |
176 | xfs_sb_t *sbp, | |
c8ce540d | 177 | uint64_t nblocks) |
4cc929ee NS |
178 | { |
179 | ASSERT(PAGE_SHIFT >= sbp->sb_blocklog); | |
180 | ASSERT(sbp->sb_blocklog >= BBSHIFT); | |
181 | ||
d5cf09ba | 182 | /* Limited by ULONG_MAX of page cache index */ |
09cbfeaf | 183 | if (nblocks >> (PAGE_SHIFT - sbp->sb_blocklog) > ULONG_MAX) |
2451337d | 184 | return -EFBIG; |
4cc929ee NS |
185 | return 0; |
186 | } | |
1da177e4 | 187 | |
1c1c6ebc | 188 | int |
c11e2c36 | 189 | xfs_initialize_perag( |
c11e2c36 | 190 | xfs_mount_t *mp, |
1c1c6ebc DC |
191 | xfs_agnumber_t agcount, |
192 | xfs_agnumber_t *maxagi) | |
1da177e4 | 193 | { |
2d2194f6 | 194 | xfs_agnumber_t index; |
b20fe473 | 195 | xfs_agnumber_t first_initialised = NULLAGNUMBER; |
1da177e4 | 196 | xfs_perag_t *pag; |
8b26c582 | 197 | int error = -ENOMEM; |
1da177e4 | 198 | |
1c1c6ebc DC |
199 | /* |
200 | * Walk the current per-ag tree so we don't try to initialise AGs | |
201 | * that already exist (growfs case). Allocate and insert all the | |
202 | * AGs we don't find ready for initialisation. | |
203 | */ | |
204 | for (index = 0; index < agcount; index++) { | |
205 | pag = xfs_perag_get(mp, index); | |
206 | if (pag) { | |
207 | xfs_perag_put(pag); | |
208 | continue; | |
209 | } | |
fb3b504a | 210 | |
1c1c6ebc DC |
211 | pag = kmem_zalloc(sizeof(*pag), KM_MAYFAIL); |
212 | if (!pag) | |
b20fe473 | 213 | goto out_unwind_new_pags; |
fb3b504a CH |
214 | pag->pag_agno = index; |
215 | pag->pag_mount = mp; | |
1a427ab0 | 216 | spin_lock_init(&pag->pag_ici_lock); |
69b491c2 | 217 | mutex_init(&pag->pag_ici_reclaim_lock); |
fb3b504a | 218 | INIT_RADIX_TREE(&pag->pag_ici_root, GFP_ATOMIC); |
6031e73a | 219 | if (xfs_buf_hash_init(pag)) |
b20fe473 | 220 | goto out_free_pag; |
ebf55872 | 221 | init_waitqueue_head(&pag->pagb_wait); |
fb3b504a | 222 | |
1c1c6ebc | 223 | if (radix_tree_preload(GFP_NOFS)) |
b20fe473 | 224 | goto out_hash_destroy; |
fb3b504a | 225 | |
1c1c6ebc DC |
226 | spin_lock(&mp->m_perag_lock); |
227 | if (radix_tree_insert(&mp->m_perag_tree, index, pag)) { | |
228 | BUG(); | |
229 | spin_unlock(&mp->m_perag_lock); | |
8b26c582 DC |
230 | radix_tree_preload_end(); |
231 | error = -EEXIST; | |
b20fe473 | 232 | goto out_hash_destroy; |
1c1c6ebc DC |
233 | } |
234 | spin_unlock(&mp->m_perag_lock); | |
235 | radix_tree_preload_end(); | |
b20fe473 BD |
236 | /* first new pag is fully initialized */ |
237 | if (first_initialised == NULLAGNUMBER) | |
238 | first_initialised = index; | |
1c1c6ebc DC |
239 | } |
240 | ||
12c3f05c | 241 | index = xfs_set_inode_alloc(mp, agcount); |
fb3b504a | 242 | |
1c1c6ebc DC |
243 | if (maxagi) |
244 | *maxagi = index; | |
8018026e DW |
245 | |
246 | mp->m_ag_prealloc_blocks = xfs_prealloc_blocks(mp); | |
1c1c6ebc | 247 | return 0; |
8b26c582 | 248 | |
b20fe473 | 249 | out_hash_destroy: |
6031e73a | 250 | xfs_buf_hash_destroy(pag); |
b20fe473 | 251 | out_free_pag: |
1da06189 | 252 | mutex_destroy(&pag->pag_ici_reclaim_lock); |
8b26c582 | 253 | kmem_free(pag); |
b20fe473 BD |
254 | out_unwind_new_pags: |
255 | /* unwind any prior newly initialized pags */ | |
256 | for (index = first_initialised; index < agcount; index++) { | |
8b26c582 | 257 | pag = radix_tree_delete(&mp->m_perag_tree, index); |
b20fe473 BD |
258 | if (!pag) |
259 | break; | |
6031e73a | 260 | xfs_buf_hash_destroy(pag); |
1da06189 | 261 | mutex_destroy(&pag->pag_ici_reclaim_lock); |
8b26c582 DC |
262 | kmem_free(pag); |
263 | } | |
264 | return error; | |
1da177e4 LT |
265 | } |
266 | ||
1da177e4 LT |
267 | /* |
268 | * xfs_readsb | |
269 | * | |
270 | * Does the initial read of the superblock. | |
271 | */ | |
272 | int | |
ff55068c DC |
273 | xfs_readsb( |
274 | struct xfs_mount *mp, | |
275 | int flags) | |
1da177e4 LT |
276 | { |
277 | unsigned int sector_size; | |
04a1e6c5 DC |
278 | struct xfs_buf *bp; |
279 | struct xfs_sb *sbp = &mp->m_sb; | |
1da177e4 | 280 | int error; |
af34e09d | 281 | int loud = !(flags & XFS_MFSI_QUIET); |
daba5427 | 282 | const struct xfs_buf_ops *buf_ops; |
1da177e4 LT |
283 | |
284 | ASSERT(mp->m_sb_bp == NULL); | |
285 | ASSERT(mp->m_ddev_targp != NULL); | |
286 | ||
daba5427 ES |
287 | /* |
288 | * For the initial read, we must guess at the sector | |
289 | * size based on the block device. It's enough to | |
290 | * get the sb_sectsize out of the superblock and | |
291 | * then reread with the proper length. | |
292 | * We don't verify it yet, because it may not be complete. | |
293 | */ | |
294 | sector_size = xfs_getsize_buftarg(mp->m_ddev_targp); | |
295 | buf_ops = NULL; | |
296 | ||
1da177e4 | 297 | /* |
c891c30a BF |
298 | * Allocate a (locked) buffer to hold the superblock. This will be kept |
299 | * around at all times to optimize access to the superblock. Therefore, | |
300 | * set XBF_NO_IOACCT to make sure it doesn't hold the buftarg count | |
301 | * elevated. | |
1da177e4 | 302 | */ |
26af6552 | 303 | reread: |
ba372674 | 304 | error = xfs_buf_read_uncached(mp->m_ddev_targp, XFS_SB_DADDR, |
c891c30a BF |
305 | BTOBB(sector_size), XBF_NO_IOACCT, &bp, |
306 | buf_ops); | |
ba372674 | 307 | if (error) { |
eab4e633 | 308 | if (loud) |
e721f504 | 309 | xfs_warn(mp, "SB validate failed with error %d.", error); |
ac75a1f7 | 310 | /* bad CRC means corrupted metadata */ |
2451337d DC |
311 | if (error == -EFSBADCRC) |
312 | error = -EFSCORRUPTED; | |
ba372674 | 313 | return error; |
eab4e633 | 314 | } |
1da177e4 LT |
315 | |
316 | /* | |
317 | * Initialize the mount structure from the superblock. | |
1da177e4 | 318 | */ |
556b8883 | 319 | xfs_sb_from_disk(sbp, XFS_BUF_TO_SBP(bp)); |
556b8883 DC |
320 | |
321 | /* | |
322 | * If we haven't validated the superblock, do so now before we try | |
323 | * to check the sector size and reread the superblock appropriately. | |
324 | */ | |
325 | if (sbp->sb_magicnum != XFS_SB_MAGIC) { | |
326 | if (loud) | |
327 | xfs_warn(mp, "Invalid superblock magic number"); | |
2451337d | 328 | error = -EINVAL; |
556b8883 DC |
329 | goto release_buf; |
330 | } | |
ff55068c | 331 | |
1da177e4 LT |
332 | /* |
333 | * We must be able to do sector-sized and sector-aligned IO. | |
334 | */ | |
04a1e6c5 | 335 | if (sector_size > sbp->sb_sectsize) { |
af34e09d DC |
336 | if (loud) |
337 | xfs_warn(mp, "device supports %u byte sectors (not %u)", | |
04a1e6c5 | 338 | sector_size, sbp->sb_sectsize); |
2451337d | 339 | error = -ENOSYS; |
26af6552 | 340 | goto release_buf; |
1da177e4 LT |
341 | } |
342 | ||
daba5427 | 343 | if (buf_ops == NULL) { |
556b8883 DC |
344 | /* |
345 | * Re-read the superblock so the buffer is correctly sized, | |
346 | * and properly verified. | |
347 | */ | |
1da177e4 | 348 | xfs_buf_relse(bp); |
04a1e6c5 | 349 | sector_size = sbp->sb_sectsize; |
daba5427 | 350 | buf_ops = loud ? &xfs_sb_buf_ops : &xfs_sb_quiet_buf_ops; |
26af6552 | 351 | goto reread; |
1da177e4 LT |
352 | } |
353 | ||
5681ca40 | 354 | xfs_reinit_percpu_counters(mp); |
8d280b98 | 355 | |
04a1e6c5 DC |
356 | /* no need to be quiet anymore, so reset the buf ops */ |
357 | bp->b_ops = &xfs_sb_buf_ops; | |
358 | ||
1da177e4 | 359 | mp->m_sb_bp = bp; |
26af6552 | 360 | xfs_buf_unlock(bp); |
1da177e4 LT |
361 | return 0; |
362 | ||
26af6552 DC |
363 | release_buf: |
364 | xfs_buf_relse(bp); | |
1da177e4 LT |
365 | return error; |
366 | } | |
367 | ||
1da177e4 | 368 | /* |
0771fb45 | 369 | * Update alignment values based on mount options and sb values |
1da177e4 | 370 | */ |
0771fb45 | 371 | STATIC int |
7884bc86 | 372 | xfs_update_alignment(xfs_mount_t *mp) |
1da177e4 | 373 | { |
1da177e4 | 374 | xfs_sb_t *sbp = &(mp->m_sb); |
1da177e4 | 375 | |
4249023a | 376 | if (mp->m_dalign) { |
1da177e4 LT |
377 | /* |
378 | * If stripe unit and stripe width are not multiples | |
379 | * of the fs blocksize turn off alignment. | |
380 | */ | |
381 | if ((BBTOB(mp->m_dalign) & mp->m_blockmask) || | |
382 | (BBTOB(mp->m_swidth) & mp->m_blockmask)) { | |
39a45d84 JL |
383 | xfs_warn(mp, |
384 | "alignment check failed: sunit/swidth vs. blocksize(%d)", | |
385 | sbp->sb_blocksize); | |
2451337d | 386 | return -EINVAL; |
1da177e4 LT |
387 | } else { |
388 | /* | |
389 | * Convert the stripe unit and width to FSBs. | |
390 | */ | |
391 | mp->m_dalign = XFS_BB_TO_FSBT(mp, mp->m_dalign); | |
392 | if (mp->m_dalign && (sbp->sb_agblocks % mp->m_dalign)) { | |
53487786 | 393 | xfs_warn(mp, |
39a45d84 JL |
394 | "alignment check failed: sunit/swidth vs. agsize(%d)", |
395 | sbp->sb_agblocks); | |
2451337d | 396 | return -EINVAL; |
1da177e4 LT |
397 | } else if (mp->m_dalign) { |
398 | mp->m_swidth = XFS_BB_TO_FSBT(mp, mp->m_swidth); | |
399 | } else { | |
39a45d84 JL |
400 | xfs_warn(mp, |
401 | "alignment check failed: sunit(%d) less than bsize(%d)", | |
402 | mp->m_dalign, sbp->sb_blocksize); | |
2451337d | 403 | return -EINVAL; |
1da177e4 LT |
404 | } |
405 | } | |
406 | ||
407 | /* | |
408 | * Update superblock with new values | |
409 | * and log changes | |
410 | */ | |
62118709 | 411 | if (xfs_sb_version_hasdalign(sbp)) { |
1da177e4 LT |
412 | if (sbp->sb_unit != mp->m_dalign) { |
413 | sbp->sb_unit = mp->m_dalign; | |
61e63ecb | 414 | mp->m_update_sb = true; |
1da177e4 LT |
415 | } |
416 | if (sbp->sb_width != mp->m_swidth) { | |
417 | sbp->sb_width = mp->m_swidth; | |
61e63ecb | 418 | mp->m_update_sb = true; |
1da177e4 | 419 | } |
34d7f603 JL |
420 | } else { |
421 | xfs_warn(mp, | |
422 | "cannot change alignment: superblock does not support data alignment"); | |
2451337d | 423 | return -EINVAL; |
1da177e4 LT |
424 | } |
425 | } else if ((mp->m_flags & XFS_MOUNT_NOALIGN) != XFS_MOUNT_NOALIGN && | |
62118709 | 426 | xfs_sb_version_hasdalign(&mp->m_sb)) { |
1da177e4 LT |
427 | mp->m_dalign = sbp->sb_unit; |
428 | mp->m_swidth = sbp->sb_width; | |
429 | } | |
430 | ||
0771fb45 ES |
431 | return 0; |
432 | } | |
1da177e4 | 433 | |
0771fb45 ES |
434 | /* |
435 | * Set the maximum inode count for this filesystem | |
436 | */ | |
437 | STATIC void | |
438 | xfs_set_maxicount(xfs_mount_t *mp) | |
439 | { | |
440 | xfs_sb_t *sbp = &(mp->m_sb); | |
c8ce540d | 441 | uint64_t icount; |
1da177e4 | 442 | |
0771fb45 ES |
443 | if (sbp->sb_imax_pct) { |
444 | /* | |
445 | * Make sure the maximum inode count is a multiple | |
446 | * of the units we allocate inodes in. | |
1da177e4 | 447 | */ |
1da177e4 LT |
448 | icount = sbp->sb_dblocks * sbp->sb_imax_pct; |
449 | do_div(icount, 100); | |
450 | do_div(icount, mp->m_ialloc_blks); | |
451 | mp->m_maxicount = (icount * mp->m_ialloc_blks) << | |
452 | sbp->sb_inopblog; | |
0771fb45 | 453 | } else { |
1da177e4 | 454 | mp->m_maxicount = 0; |
1da177e4 | 455 | } |
0771fb45 ES |
456 | } |
457 | ||
458 | /* | |
459 | * Set the default minimum read and write sizes unless | |
460 | * already specified in a mount option. | |
461 | * We use smaller I/O sizes when the file system | |
462 | * is being used for NFS service (wsync mount option). | |
463 | */ | |
464 | STATIC void | |
465 | xfs_set_rw_sizes(xfs_mount_t *mp) | |
466 | { | |
467 | xfs_sb_t *sbp = &(mp->m_sb); | |
468 | int readio_log, writeio_log; | |
1da177e4 | 469 | |
1da177e4 LT |
470 | if (!(mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)) { |
471 | if (mp->m_flags & XFS_MOUNT_WSYNC) { | |
472 | readio_log = XFS_WSYNC_READIO_LOG; | |
473 | writeio_log = XFS_WSYNC_WRITEIO_LOG; | |
474 | } else { | |
475 | readio_log = XFS_READIO_LOG_LARGE; | |
476 | writeio_log = XFS_WRITEIO_LOG_LARGE; | |
477 | } | |
478 | } else { | |
479 | readio_log = mp->m_readio_log; | |
480 | writeio_log = mp->m_writeio_log; | |
481 | } | |
482 | ||
1da177e4 LT |
483 | if (sbp->sb_blocklog > readio_log) { |
484 | mp->m_readio_log = sbp->sb_blocklog; | |
485 | } else { | |
486 | mp->m_readio_log = readio_log; | |
487 | } | |
488 | mp->m_readio_blocks = 1 << (mp->m_readio_log - sbp->sb_blocklog); | |
489 | if (sbp->sb_blocklog > writeio_log) { | |
490 | mp->m_writeio_log = sbp->sb_blocklog; | |
491 | } else { | |
492 | mp->m_writeio_log = writeio_log; | |
493 | } | |
494 | mp->m_writeio_blocks = 1 << (mp->m_writeio_log - sbp->sb_blocklog); | |
0771fb45 | 495 | } |
1da177e4 | 496 | |
055388a3 DC |
497 | /* |
498 | * precalculate the low space thresholds for dynamic speculative preallocation. | |
499 | */ | |
500 | void | |
501 | xfs_set_low_space_thresholds( | |
502 | struct xfs_mount *mp) | |
503 | { | |
504 | int i; | |
505 | ||
506 | for (i = 0; i < XFS_LOWSP_MAX; i++) { | |
c8ce540d | 507 | uint64_t space = mp->m_sb.sb_dblocks; |
055388a3 DC |
508 | |
509 | do_div(space, 100); | |
510 | mp->m_low_space[i] = space * (i + 1); | |
511 | } | |
512 | } | |
513 | ||
514 | ||
0771fb45 ES |
515 | /* |
516 | * Set whether we're using inode alignment. | |
517 | */ | |
518 | STATIC void | |
519 | xfs_set_inoalignment(xfs_mount_t *mp) | |
520 | { | |
62118709 | 521 | if (xfs_sb_version_hasalign(&mp->m_sb) && |
d5825712 | 522 | mp->m_sb.sb_inoalignmt >= xfs_icluster_size_fsb(mp)) |
1da177e4 LT |
523 | mp->m_inoalign_mask = mp->m_sb.sb_inoalignmt - 1; |
524 | else | |
525 | mp->m_inoalign_mask = 0; | |
526 | /* | |
527 | * If we are using stripe alignment, check whether | |
528 | * the stripe unit is a multiple of the inode alignment | |
529 | */ | |
530 | if (mp->m_dalign && mp->m_inoalign_mask && | |
531 | !(mp->m_dalign & mp->m_inoalign_mask)) | |
532 | mp->m_sinoalign = mp->m_dalign; | |
533 | else | |
534 | mp->m_sinoalign = 0; | |
0771fb45 ES |
535 | } |
536 | ||
537 | /* | |
0471f62e | 538 | * Check that the data (and log if separate) is an ok size. |
0771fb45 ES |
539 | */ |
540 | STATIC int | |
ba372674 DC |
541 | xfs_check_sizes( |
542 | struct xfs_mount *mp) | |
0771fb45 | 543 | { |
ba372674 | 544 | struct xfs_buf *bp; |
0771fb45 | 545 | xfs_daddr_t d; |
ba372674 | 546 | int error; |
0771fb45 | 547 | |
1da177e4 LT |
548 | d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks); |
549 | if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_dblocks) { | |
0b932ccc | 550 | xfs_warn(mp, "filesystem size mismatch detected"); |
2451337d | 551 | return -EFBIG; |
1da177e4 | 552 | } |
ba372674 | 553 | error = xfs_buf_read_uncached(mp->m_ddev_targp, |
1922c949 | 554 | d - XFS_FSS_TO_BB(mp, 1), |
ba372674 DC |
555 | XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL); |
556 | if (error) { | |
0b932ccc | 557 | xfs_warn(mp, "last sector read failed"); |
ba372674 | 558 | return error; |
1da177e4 | 559 | } |
1922c949 | 560 | xfs_buf_relse(bp); |
1da177e4 | 561 | |
ba372674 DC |
562 | if (mp->m_logdev_targp == mp->m_ddev_targp) |
563 | return 0; | |
564 | ||
565 | d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_logblocks); | |
566 | if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_logblocks) { | |
567 | xfs_warn(mp, "log size mismatch detected"); | |
568 | return -EFBIG; | |
569 | } | |
570 | error = xfs_buf_read_uncached(mp->m_logdev_targp, | |
1922c949 | 571 | d - XFS_FSB_TO_BB(mp, 1), |
ba372674 DC |
572 | XFS_FSB_TO_BB(mp, 1), 0, &bp, NULL); |
573 | if (error) { | |
574 | xfs_warn(mp, "log device read failed"); | |
575 | return error; | |
0771fb45 | 576 | } |
ba372674 | 577 | xfs_buf_relse(bp); |
0771fb45 ES |
578 | return 0; |
579 | } | |
580 | ||
7d095257 CH |
581 | /* |
582 | * Clear the quotaflags in memory and in the superblock. | |
583 | */ | |
584 | int | |
585 | xfs_mount_reset_sbqflags( | |
586 | struct xfs_mount *mp) | |
587 | { | |
7d095257 CH |
588 | mp->m_qflags = 0; |
589 | ||
61e63ecb | 590 | /* It is OK to look at sb_qflags in the mount path without m_sb_lock. */ |
7d095257 CH |
591 | if (mp->m_sb.sb_qflags == 0) |
592 | return 0; | |
593 | spin_lock(&mp->m_sb_lock); | |
594 | mp->m_sb.sb_qflags = 0; | |
595 | spin_unlock(&mp->m_sb_lock); | |
596 | ||
61e63ecb | 597 | if (!xfs_fs_writable(mp, SB_FREEZE_WRITE)) |
7d095257 CH |
598 | return 0; |
599 | ||
61e63ecb | 600 | return xfs_sync_sb(mp, false); |
7d095257 CH |
601 | } |
602 | ||
c8ce540d | 603 | uint64_t |
d5db0f97 ES |
604 | xfs_default_resblks(xfs_mount_t *mp) |
605 | { | |
c8ce540d | 606 | uint64_t resblks; |
d5db0f97 ES |
607 | |
608 | /* | |
8babd8a2 DC |
609 | * We default to 5% or 8192 fsbs of space reserved, whichever is |
610 | * smaller. This is intended to cover concurrent allocation | |
611 | * transactions when we initially hit enospc. These each require a 4 | |
612 | * block reservation. Hence by default we cover roughly 2000 concurrent | |
613 | * allocation reservations. | |
d5db0f97 ES |
614 | */ |
615 | resblks = mp->m_sb.sb_dblocks; | |
616 | do_div(resblks, 20); | |
c8ce540d | 617 | resblks = min_t(uint64_t, resblks, 8192); |
d5db0f97 ES |
618 | return resblks; |
619 | } | |
620 | ||
0771fb45 | 621 | /* |
0771fb45 ES |
622 | * This function does the following on an initial mount of a file system: |
623 | * - reads the superblock from disk and init the mount struct | |
624 | * - if we're a 32-bit kernel, do a size check on the superblock | |
625 | * so we don't mount terabyte filesystems | |
626 | * - init mount struct realtime fields | |
627 | * - allocate inode hash table for fs | |
628 | * - init directory manager | |
629 | * - perform recovery and init the log manager | |
630 | */ | |
631 | int | |
632 | xfs_mountfs( | |
f0b2efad | 633 | struct xfs_mount *mp) |
0771fb45 | 634 | { |
f0b2efad BF |
635 | struct xfs_sb *sbp = &(mp->m_sb); |
636 | struct xfs_inode *rip; | |
c8ce540d | 637 | uint64_t resblks; |
f0b2efad BF |
638 | uint quotamount = 0; |
639 | uint quotaflags = 0; | |
640 | int error = 0; | |
0771fb45 | 641 | |
ff55068c | 642 | xfs_sb_mount_common(mp, sbp); |
0771fb45 | 643 | |
ee1c0908 | 644 | /* |
074e427b DC |
645 | * Check for a mismatched features2 values. Older kernels read & wrote |
646 | * into the wrong sb offset for sb_features2 on some platforms due to | |
647 | * xfs_sb_t not being 64bit size aligned when sb_features2 was added, | |
648 | * which made older superblock reading/writing routines swap it as a | |
649 | * 64-bit value. | |
ee1c0908 | 650 | * |
e6957ea4 ES |
651 | * For backwards compatibility, we make both slots equal. |
652 | * | |
074e427b DC |
653 | * If we detect a mismatched field, we OR the set bits into the existing |
654 | * features2 field in case it has already been modified; we don't want | |
655 | * to lose any features. We then update the bad location with the ORed | |
656 | * value so that older kernels will see any features2 flags. The | |
657 | * superblock writeback code ensures the new sb_features2 is copied to | |
658 | * sb_bad_features2 before it is logged or written to disk. | |
ee1c0908 | 659 | */ |
e6957ea4 | 660 | if (xfs_sb_has_mismatched_features2(sbp)) { |
0b932ccc | 661 | xfs_warn(mp, "correcting sb_features alignment problem"); |
ee1c0908 | 662 | sbp->sb_features2 |= sbp->sb_bad_features2; |
61e63ecb | 663 | mp->m_update_sb = true; |
e6957ea4 ES |
664 | |
665 | /* | |
666 | * Re-check for ATTR2 in case it was found in bad_features2 | |
667 | * slot. | |
668 | */ | |
7c12f296 TS |
669 | if (xfs_sb_version_hasattr2(&mp->m_sb) && |
670 | !(mp->m_flags & XFS_MOUNT_NOATTR2)) | |
e6957ea4 | 671 | mp->m_flags |= XFS_MOUNT_ATTR2; |
7c12f296 TS |
672 | } |
673 | ||
674 | if (xfs_sb_version_hasattr2(&mp->m_sb) && | |
675 | (mp->m_flags & XFS_MOUNT_NOATTR2)) { | |
676 | xfs_sb_version_removeattr2(&mp->m_sb); | |
61e63ecb | 677 | mp->m_update_sb = true; |
e6957ea4 | 678 | |
7c12f296 TS |
679 | /* update sb_versionnum for the clearing of the morebits */ |
680 | if (!sbp->sb_features2) | |
61e63ecb | 681 | mp->m_update_sb = true; |
ee1c0908 DC |
682 | } |
683 | ||
263997a6 DC |
684 | /* always use v2 inodes by default now */ |
685 | if (!(mp->m_sb.sb_versionnum & XFS_SB_VERSION_NLINKBIT)) { | |
686 | mp->m_sb.sb_versionnum |= XFS_SB_VERSION_NLINKBIT; | |
61e63ecb | 687 | mp->m_update_sb = true; |
263997a6 DC |
688 | } |
689 | ||
0771fb45 ES |
690 | /* |
691 | * Check if sb_agblocks is aligned at stripe boundary | |
692 | * If sb_agblocks is NOT aligned turn off m_dalign since | |
693 | * allocator alignment is within an ag, therefore ag has | |
694 | * to be aligned at stripe boundary. | |
695 | */ | |
7884bc86 | 696 | error = xfs_update_alignment(mp); |
0771fb45 | 697 | if (error) |
f9057e3d | 698 | goto out; |
0771fb45 ES |
699 | |
700 | xfs_alloc_compute_maxlevels(mp); | |
701 | xfs_bmap_compute_maxlevels(mp, XFS_DATA_FORK); | |
702 | xfs_bmap_compute_maxlevels(mp, XFS_ATTR_FORK); | |
703 | xfs_ialloc_compute_maxlevels(mp); | |
035e00ac | 704 | xfs_rmapbt_compute_maxlevels(mp); |
1946b91c | 705 | xfs_refcountbt_compute_maxlevels(mp); |
0771fb45 ES |
706 | |
707 | xfs_set_maxicount(mp); | |
708 | ||
e6b3bb78 | 709 | /* enable fail_at_unmount as default */ |
749f24f3 | 710 | mp->m_fail_unmount = true; |
e6b3bb78 | 711 | |
a31b1d3d | 712 | error = xfs_sysfs_init(&mp->m_kobj, &xfs_mp_ktype, NULL, mp->m_fsname); |
27174203 CH |
713 | if (error) |
714 | goto out; | |
1da177e4 | 715 | |
225e4635 BD |
716 | error = xfs_sysfs_init(&mp->m_stats.xs_kobj, &xfs_stats_ktype, |
717 | &mp->m_kobj, "stats"); | |
a31b1d3d BF |
718 | if (error) |
719 | goto out_remove_sysfs; | |
720 | ||
192852be | 721 | error = xfs_error_sysfs_init(mp); |
225e4635 BD |
722 | if (error) |
723 | goto out_del_stats; | |
724 | ||
31965ef3 DW |
725 | error = xfs_errortag_init(mp); |
726 | if (error) | |
727 | goto out_remove_error_sysfs; | |
192852be CM |
728 | |
729 | error = xfs_uuid_mount(mp); | |
730 | if (error) | |
31965ef3 | 731 | goto out_remove_errortag; |
192852be | 732 | |
0771fb45 ES |
733 | /* |
734 | * Set the minimum read and write sizes | |
735 | */ | |
736 | xfs_set_rw_sizes(mp); | |
737 | ||
055388a3 DC |
738 | /* set the low space thresholds for dynamic preallocation */ |
739 | xfs_set_low_space_thresholds(mp); | |
740 | ||
0771fb45 ES |
741 | /* |
742 | * Set the inode cluster size. | |
743 | * This may still be overridden by the file system | |
744 | * block size if it is larger than the chosen cluster size. | |
8f80587b DC |
745 | * |
746 | * For v5 filesystems, scale the cluster size with the inode size to | |
747 | * keep a constant ratio of inode per cluster buffer, but only if mkfs | |
748 | * has set the inode alignment value appropriately for larger cluster | |
749 | * sizes. | |
0771fb45 ES |
750 | */ |
751 | mp->m_inode_cluster_size = XFS_INODE_BIG_CLUSTER_SIZE; | |
8f80587b DC |
752 | if (xfs_sb_version_hascrc(&mp->m_sb)) { |
753 | int new_size = mp->m_inode_cluster_size; | |
754 | ||
755 | new_size *= mp->m_sb.sb_inodesize / XFS_DINODE_MIN_SIZE; | |
756 | if (mp->m_sb.sb_inoalignmt >= XFS_B_TO_FSBT(mp, new_size)) | |
757 | mp->m_inode_cluster_size = new_size; | |
8f80587b | 758 | } |
0771fb45 | 759 | |
e5376fc1 BF |
760 | /* |
761 | * If enabled, sparse inode chunk alignment is expected to match the | |
762 | * cluster size. Full inode chunk alignment must match the chunk size, | |
763 | * but that is checked on sb read verification... | |
764 | */ | |
765 | if (xfs_sb_version_hassparseinodes(&mp->m_sb) && | |
766 | mp->m_sb.sb_spino_align != | |
767 | XFS_B_TO_FSBT(mp, mp->m_inode_cluster_size)) { | |
768 | xfs_warn(mp, | |
769 | "Sparse inode block alignment (%u) must match cluster size (%llu).", | |
770 | mp->m_sb.sb_spino_align, | |
771 | XFS_B_TO_FSBT(mp, mp->m_inode_cluster_size)); | |
772 | error = -EINVAL; | |
773 | goto out_remove_uuid; | |
774 | } | |
775 | ||
0771fb45 ES |
776 | /* |
777 | * Set inode alignment fields | |
778 | */ | |
779 | xfs_set_inoalignment(mp); | |
780 | ||
781 | /* | |
c2bfbc9b | 782 | * Check that the data (and log if separate) is an ok size. |
0771fb45 | 783 | */ |
4249023a | 784 | error = xfs_check_sizes(mp); |
0771fb45 | 785 | if (error) |
f9057e3d | 786 | goto out_remove_uuid; |
0771fb45 | 787 | |
1da177e4 LT |
788 | /* |
789 | * Initialize realtime fields in the mount structure | |
790 | */ | |
0771fb45 ES |
791 | error = xfs_rtmount_init(mp); |
792 | if (error) { | |
0b932ccc | 793 | xfs_warn(mp, "RT mount failed"); |
f9057e3d | 794 | goto out_remove_uuid; |
1da177e4 LT |
795 | } |
796 | ||
1da177e4 LT |
797 | /* |
798 | * Copies the low order bits of the timestamp and the randomly | |
799 | * set "sequence" number out of a UUID. | |
800 | */ | |
cb0ba6cc CH |
801 | mp->m_fixedfsid[0] = |
802 | (get_unaligned_be16(&sbp->sb_uuid.b[8]) << 16) | | |
803 | get_unaligned_be16(&sbp->sb_uuid.b[4]); | |
804 | mp->m_fixedfsid[1] = get_unaligned_be32(&sbp->sb_uuid.b[0]); | |
1da177e4 | 805 | |
0650b554 DC |
806 | error = xfs_da_mount(mp); |
807 | if (error) { | |
808 | xfs_warn(mp, "Failed dir/attr init: %d", error); | |
809 | goto out_remove_uuid; | |
810 | } | |
1da177e4 LT |
811 | |
812 | /* | |
813 | * Initialize the precomputed transaction reservations values. | |
814 | */ | |
815 | xfs_trans_init(mp); | |
816 | ||
1da177e4 LT |
817 | /* |
818 | * Allocate and initialize the per-ag data. | |
819 | */ | |
1c1c6ebc DC |
820 | error = xfs_initialize_perag(mp, sbp->sb_agcount, &mp->m_maxagi); |
821 | if (error) { | |
0b932ccc | 822 | xfs_warn(mp, "Failed per-ag init: %d", error); |
0650b554 | 823 | goto out_free_dir; |
1c1c6ebc | 824 | } |
1da177e4 | 825 | |
f9057e3d | 826 | if (!sbp->sb_logblocks) { |
0b932ccc | 827 | xfs_warn(mp, "no log defined"); |
f9057e3d | 828 | XFS_ERROR_REPORT("xfs_mountfs", XFS_ERRLEVEL_LOW, mp); |
2451337d | 829 | error = -EFSCORRUPTED; |
f9057e3d CH |
830 | goto out_free_perag; |
831 | } | |
832 | ||
1da177e4 | 833 | /* |
f0b2efad BF |
834 | * Log's mount-time initialization. The first part of recovery can place |
835 | * some items on the AIL, to be handled when recovery is finished or | |
836 | * cancelled. | |
1da177e4 | 837 | */ |
f9057e3d CH |
838 | error = xfs_log_mount(mp, mp->m_logdev_targp, |
839 | XFS_FSB_TO_DADDR(mp, sbp->sb_logstart), | |
840 | XFS_FSB_TO_BB(mp, sbp->sb_logblocks)); | |
841 | if (error) { | |
0b932ccc | 842 | xfs_warn(mp, "log mount failed"); |
d4f3512b | 843 | goto out_fail_wait; |
1da177e4 LT |
844 | } |
845 | ||
92821e2b DC |
846 | /* |
847 | * Now the log is mounted, we know if it was an unclean shutdown or | |
848 | * not. If it was, with the first phase of recovery has completed, we | |
849 | * have consistent AG blocks on disk. We have not recovered EFIs yet, | |
850 | * but they are recovered transactionally in the second recovery phase | |
851 | * later. | |
852 | * | |
853 | * Hence we can safely re-initialise incore superblock counters from | |
854 | * the per-ag data. These may not be correct if the filesystem was not | |
855 | * cleanly unmounted, so we need to wait for recovery to finish before | |
856 | * doing this. | |
857 | * | |
858 | * If the filesystem was cleanly unmounted, then we can trust the | |
859 | * values in the superblock to be correct and we don't need to do | |
860 | * anything here. | |
861 | * | |
862 | * If we are currently making the filesystem, the initialisation will | |
863 | * fail as the perag data is in an undefined state. | |
864 | */ | |
92821e2b DC |
865 | if (xfs_sb_version_haslazysbcount(&mp->m_sb) && |
866 | !XFS_LAST_UNMOUNT_WAS_CLEAN(mp) && | |
867 | !mp->m_sb.sb_inprogress) { | |
868 | error = xfs_initialize_perag_data(mp, sbp->sb_agcount); | |
f9057e3d | 869 | if (error) |
6eee8972 | 870 | goto out_log_dealloc; |
92821e2b | 871 | } |
f9057e3d | 872 | |
1da177e4 LT |
873 | /* |
874 | * Get and sanity-check the root inode. | |
875 | * Save the pointer to it in the mount structure. | |
876 | */ | |
7b6259e7 | 877 | error = xfs_iget(mp, NULL, sbp->sb_rootino, 0, XFS_ILOCK_EXCL, &rip); |
1da177e4 | 878 | if (error) { |
0b932ccc | 879 | xfs_warn(mp, "failed to read root inode"); |
f9057e3d | 880 | goto out_log_dealloc; |
1da177e4 LT |
881 | } |
882 | ||
883 | ASSERT(rip != NULL); | |
1da177e4 | 884 | |
c19b3b05 | 885 | if (unlikely(!S_ISDIR(VFS_I(rip)->i_mode))) { |
0b932ccc | 886 | xfs_warn(mp, "corrupted root inode %llu: not a directory", |
b6574520 | 887 | (unsigned long long)rip->i_ino); |
1da177e4 LT |
888 | xfs_iunlock(rip, XFS_ILOCK_EXCL); |
889 | XFS_ERROR_REPORT("xfs_mountfs_int(2)", XFS_ERRLEVEL_LOW, | |
890 | mp); | |
2451337d | 891 | error = -EFSCORRUPTED; |
f9057e3d | 892 | goto out_rele_rip; |
1da177e4 LT |
893 | } |
894 | mp->m_rootip = rip; /* save it */ | |
895 | ||
896 | xfs_iunlock(rip, XFS_ILOCK_EXCL); | |
897 | ||
898 | /* | |
899 | * Initialize realtime inode pointers in the mount structure | |
900 | */ | |
0771fb45 ES |
901 | error = xfs_rtmount_inodes(mp); |
902 | if (error) { | |
1da177e4 LT |
903 | /* |
904 | * Free up the root inode. | |
905 | */ | |
0b932ccc | 906 | xfs_warn(mp, "failed to read RT inodes"); |
f9057e3d | 907 | goto out_rele_rip; |
1da177e4 LT |
908 | } |
909 | ||
910 | /* | |
7884bc86 CH |
911 | * If this is a read-only mount defer the superblock updates until |
912 | * the next remount into writeable mode. Otherwise we would never | |
913 | * perform the update e.g. for the root filesystem. | |
1da177e4 | 914 | */ |
61e63ecb DC |
915 | if (mp->m_update_sb && !(mp->m_flags & XFS_MOUNT_RDONLY)) { |
916 | error = xfs_sync_sb(mp, false); | |
e5720eec | 917 | if (error) { |
0b932ccc | 918 | xfs_warn(mp, "failed to write sb changes"); |
b93b6e43 | 919 | goto out_rtunmount; |
e5720eec DC |
920 | } |
921 | } | |
1da177e4 LT |
922 | |
923 | /* | |
924 | * Initialise the XFS quota management subsystem for this mount | |
925 | */ | |
7d095257 CH |
926 | if (XFS_IS_QUOTA_RUNNING(mp)) { |
927 | error = xfs_qm_newmount(mp, "amount, "aflags); | |
928 | if (error) | |
929 | goto out_rtunmount; | |
930 | } else { | |
931 | ASSERT(!XFS_IS_QUOTA_ON(mp)); | |
932 | ||
933 | /* | |
934 | * If a file system had quotas running earlier, but decided to | |
935 | * mount without -o uquota/pquota/gquota options, revoke the | |
936 | * quotachecked license. | |
937 | */ | |
938 | if (mp->m_sb.sb_qflags & XFS_ALL_QUOTA_ACCT) { | |
0b932ccc | 939 | xfs_notice(mp, "resetting quota flags"); |
7d095257 CH |
940 | error = xfs_mount_reset_sbqflags(mp); |
941 | if (error) | |
a70a4fa5 | 942 | goto out_rtunmount; |
7d095257 CH |
943 | } |
944 | } | |
1da177e4 LT |
945 | |
946 | /* | |
f0b2efad BF |
947 | * Finish recovering the file system. This part needed to be delayed |
948 | * until after the root and real-time bitmap inodes were consistently | |
949 | * read in. | |
1da177e4 | 950 | */ |
4249023a | 951 | error = xfs_log_mount_finish(mp); |
1da177e4 | 952 | if (error) { |
0b932ccc | 953 | xfs_warn(mp, "log mount finish failed"); |
b93b6e43 | 954 | goto out_rtunmount; |
1da177e4 LT |
955 | } |
956 | ||
ddeb14f4 DC |
957 | /* |
958 | * Now the log is fully replayed, we can transition to full read-only | |
959 | * mode for read-only mounts. This will sync all the metadata and clean | |
960 | * the log so that the recovery we just performed does not have to be | |
961 | * replayed again on the next mount. | |
962 | * | |
963 | * We use the same quiesce mechanism as the rw->ro remount, as they are | |
964 | * semantically identical operations. | |
965 | */ | |
966 | if ((mp->m_flags & (XFS_MOUNT_RDONLY|XFS_MOUNT_NORECOVERY)) == | |
967 | XFS_MOUNT_RDONLY) { | |
968 | xfs_quiesce_attr(mp); | |
969 | } | |
970 | ||
1da177e4 LT |
971 | /* |
972 | * Complete the quota initialisation, post-log-replay component. | |
973 | */ | |
7d095257 CH |
974 | if (quotamount) { |
975 | ASSERT(mp->m_qflags == 0); | |
976 | mp->m_qflags = quotaflags; | |
977 | ||
978 | xfs_qm_mount_quotas(mp); | |
979 | } | |
980 | ||
84e1e99f DC |
981 | /* |
982 | * Now we are mounted, reserve a small amount of unused space for | |
983 | * privileged transactions. This is needed so that transaction | |
984 | * space required for critical operations can dip into this pool | |
985 | * when at ENOSPC. This is needed for operations like create with | |
986 | * attr, unwritten extent conversion at ENOSPC, etc. Data allocations | |
987 | * are not allowed to use this reserved space. | |
8babd8a2 DC |
988 | * |
989 | * This may drive us straight to ENOSPC on mount, but that implies | |
990 | * we were already there on the last unmount. Warn if this occurs. | |
84e1e99f | 991 | */ |
d5db0f97 ES |
992 | if (!(mp->m_flags & XFS_MOUNT_RDONLY)) { |
993 | resblks = xfs_default_resblks(mp); | |
994 | error = xfs_reserve_blocks(mp, &resblks, NULL); | |
995 | if (error) | |
0b932ccc DC |
996 | xfs_warn(mp, |
997 | "Unable to allocate reserve blocks. Continuing without reserve pool."); | |
174edb0e DW |
998 | |
999 | /* Recover any CoW blocks that never got remapped. */ | |
1000 | error = xfs_reflink_recover_cow(mp); | |
1001 | if (error) { | |
1002 | xfs_err(mp, | |
1003 | "Error %d recovering leftover CoW allocations.", error); | |
1004 | xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); | |
1005 | goto out_quota; | |
1006 | } | |
84d69619 DW |
1007 | |
1008 | /* Reserve AG blocks for future btree expansion. */ | |
1009 | error = xfs_fs_reserve_ag_blocks(mp); | |
1010 | if (error && error != -ENOSPC) | |
1011 | goto out_agresv; | |
d5db0f97 | 1012 | } |
84e1e99f | 1013 | |
1da177e4 LT |
1014 | return 0; |
1015 | ||
84d69619 DW |
1016 | out_agresv: |
1017 | xfs_fs_unreserve_ag_blocks(mp); | |
174edb0e DW |
1018 | out_quota: |
1019 | xfs_qm_unmount_quotas(mp); | |
b93b6e43 CH |
1020 | out_rtunmount: |
1021 | xfs_rtunmount_inodes(mp); | |
f9057e3d | 1022 | out_rele_rip: |
43355099 | 1023 | IRELE(rip); |
77aff8c7 DW |
1024 | /* Clean out dquots that might be in memory after quotacheck. */ |
1025 | xfs_qm_unmount(mp); | |
2d1d1da3 DW |
1026 | /* |
1027 | * Cancel all delayed reclaim work and reclaim the inodes directly. | |
1028 | * We have to do this /after/ rtunmount and qm_unmount because those | |
1029 | * two will have scheduled delayed reclaim for the rt/quota inodes. | |
1030 | * | |
1031 | * This is slightly different from the unmountfs call sequence | |
1032 | * because we could be tearing down a partially set up mount. In | |
1033 | * particular, if log_mount_finish fails we bail out without calling | |
1034 | * qm_unmount_quotas and therefore rely on qm_unmount to release the | |
1035 | * quota inodes. | |
1036 | */ | |
1037 | cancel_delayed_work_sync(&mp->m_reclaim_work); | |
1038 | xfs_reclaim_inodes(mp, SYNC_WAIT); | |
f9057e3d | 1039 | out_log_dealloc: |
e6b3bb78 | 1040 | mp->m_flags |= XFS_MOUNT_UNMOUNTING; |
f0b2efad | 1041 | xfs_log_mount_cancel(mp); |
d4f3512b DC |
1042 | out_fail_wait: |
1043 | if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) | |
1044 | xfs_wait_buftarg(mp->m_logdev_targp); | |
1045 | xfs_wait_buftarg(mp->m_ddev_targp); | |
f9057e3d | 1046 | out_free_perag: |
ff4f038c | 1047 | xfs_free_perag(mp); |
0650b554 DC |
1048 | out_free_dir: |
1049 | xfs_da_unmount(mp); | |
f9057e3d | 1050 | out_remove_uuid: |
27174203 | 1051 | xfs_uuid_unmount(mp); |
31965ef3 DW |
1052 | out_remove_errortag: |
1053 | xfs_errortag_del(mp); | |
192852be CM |
1054 | out_remove_error_sysfs: |
1055 | xfs_error_sysfs_del(mp); | |
225e4635 BD |
1056 | out_del_stats: |
1057 | xfs_sysfs_del(&mp->m_stats.xs_kobj); | |
a31b1d3d BF |
1058 | out_remove_sysfs: |
1059 | xfs_sysfs_del(&mp->m_kobj); | |
f9057e3d | 1060 | out: |
1da177e4 LT |
1061 | return error; |
1062 | } | |
1063 | ||
1064 | /* | |
1da177e4 LT |
1065 | * This flushes out the inodes,dquots and the superblock, unmounts the |
1066 | * log and makes sure that incore structures are freed. | |
1067 | */ | |
41b5c2e7 CH |
1068 | void |
1069 | xfs_unmountfs( | |
1070 | struct xfs_mount *mp) | |
1da177e4 | 1071 | { |
c8ce540d | 1072 | uint64_t resblks; |
41b5c2e7 | 1073 | int error; |
1da177e4 | 1074 | |
579b62fa | 1075 | cancel_delayed_work_sync(&mp->m_eofblocks_work); |
83104d44 | 1076 | cancel_delayed_work_sync(&mp->m_cowblocks_work); |
579b62fa | 1077 | |
84d69619 | 1078 | xfs_fs_unreserve_ag_blocks(mp); |
7d095257 | 1079 | xfs_qm_unmount_quotas(mp); |
b93b6e43 | 1080 | xfs_rtunmount_inodes(mp); |
77508ec8 CH |
1081 | IRELE(mp->m_rootip); |
1082 | ||
641c56fb DC |
1083 | /* |
1084 | * We can potentially deadlock here if we have an inode cluster | |
9da096fd | 1085 | * that has been freed has its buffer still pinned in memory because |
641c56fb DC |
1086 | * the transaction is still sitting in a iclog. The stale inodes |
1087 | * on that buffer will have their flush locks held until the | |
1088 | * transaction hits the disk and the callbacks run. the inode | |
1089 | * flush takes the flush lock unconditionally and with nothing to | |
1090 | * push out the iclog we will never get that unlocked. hence we | |
1091 | * need to force the log first. | |
1092 | */ | |
a14a348b | 1093 | xfs_log_force(mp, XFS_LOG_SYNC); |
c854363e | 1094 | |
ebf55872 CH |
1095 | /* |
1096 | * Wait for all busy extents to be freed, including completion of | |
1097 | * any discard operation. | |
1098 | */ | |
1099 | xfs_extent_busy_wait_all(mp); | |
4560e78f | 1100 | flush_workqueue(xfs_discard_wq); |
ebf55872 | 1101 | |
e6b3bb78 CM |
1102 | /* |
1103 | * We now need to tell the world we are unmounting. This will allow | |
1104 | * us to detect that the filesystem is going away and we should error | |
1105 | * out anything that we have been retrying in the background. This will | |
1106 | * prevent neverending retries in AIL pushing from hanging the unmount. | |
1107 | */ | |
1108 | mp->m_flags |= XFS_MOUNT_UNMOUNTING; | |
1109 | ||
c854363e | 1110 | /* |
211e4d43 CH |
1111 | * Flush all pending changes from the AIL. |
1112 | */ | |
1113 | xfs_ail_push_all_sync(mp->m_ail); | |
1114 | ||
1115 | /* | |
1116 | * And reclaim all inodes. At this point there should be no dirty | |
7e18530b DC |
1117 | * inodes and none should be pinned or locked, but use synchronous |
1118 | * reclaim just to be sure. We can stop background inode reclaim | |
1119 | * here as well if it is still running. | |
c854363e | 1120 | */ |
7e18530b | 1121 | cancel_delayed_work_sync(&mp->m_reclaim_work); |
c854363e | 1122 | xfs_reclaim_inodes(mp, SYNC_WAIT); |
1da177e4 | 1123 | |
7d095257 | 1124 | xfs_qm_unmount(mp); |
a357a121 | 1125 | |
84e1e99f DC |
1126 | /* |
1127 | * Unreserve any blocks we have so that when we unmount we don't account | |
1128 | * the reserved free space as used. This is really only necessary for | |
1129 | * lazy superblock counting because it trusts the incore superblock | |
9da096fd | 1130 | * counters to be absolutely correct on clean unmount. |
84e1e99f DC |
1131 | * |
1132 | * We don't bother correcting this elsewhere for lazy superblock | |
1133 | * counting because on mount of an unclean filesystem we reconstruct the | |
1134 | * correct counter value and this is irrelevant. | |
1135 | * | |
1136 | * For non-lazy counter filesystems, this doesn't matter at all because | |
1137 | * we only every apply deltas to the superblock and hence the incore | |
1138 | * value does not matter.... | |
1139 | */ | |
1140 | resblks = 0; | |
714082bc DC |
1141 | error = xfs_reserve_blocks(mp, &resblks, NULL); |
1142 | if (error) | |
0b932ccc | 1143 | xfs_warn(mp, "Unable to free reserved block pool. " |
714082bc DC |
1144 | "Freespace may not be correct on next mount."); |
1145 | ||
adab0f67 | 1146 | error = xfs_log_sbcount(mp); |
e5720eec | 1147 | if (error) |
0b932ccc | 1148 | xfs_warn(mp, "Unable to update superblock counters. " |
e5720eec | 1149 | "Freespace may not be correct on next mount."); |
87c7bec7 | 1150 | |
225e4635 | 1151 | |
21b699c8 | 1152 | xfs_log_unmount(mp); |
0650b554 | 1153 | xfs_da_unmount(mp); |
27174203 | 1154 | xfs_uuid_unmount(mp); |
1da177e4 | 1155 | |
1550d0b0 | 1156 | #if defined(DEBUG) |
31965ef3 | 1157 | xfs_errortag_clearall(mp); |
1da177e4 | 1158 | #endif |
ff4f038c | 1159 | xfs_free_perag(mp); |
a31b1d3d | 1160 | |
31965ef3 | 1161 | xfs_errortag_del(mp); |
192852be | 1162 | xfs_error_sysfs_del(mp); |
225e4635 | 1163 | xfs_sysfs_del(&mp->m_stats.xs_kobj); |
a31b1d3d | 1164 | xfs_sysfs_del(&mp->m_kobj); |
1da177e4 LT |
1165 | } |
1166 | ||
91ee575f BF |
1167 | /* |
1168 | * Determine whether modifications can proceed. The caller specifies the minimum | |
1169 | * freeze level for which modifications should not be allowed. This allows | |
1170 | * certain operations to proceed while the freeze sequence is in progress, if | |
1171 | * necessary. | |
1172 | */ | |
1173 | bool | |
1174 | xfs_fs_writable( | |
1175 | struct xfs_mount *mp, | |
1176 | int level) | |
92821e2b | 1177 | { |
91ee575f BF |
1178 | ASSERT(level > SB_UNFROZEN); |
1179 | if ((mp->m_super->s_writers.frozen >= level) || | |
1180 | XFS_FORCED_SHUTDOWN(mp) || (mp->m_flags & XFS_MOUNT_RDONLY)) | |
1181 | return false; | |
1182 | ||
1183 | return true; | |
92821e2b DC |
1184 | } |
1185 | ||
1186 | /* | |
b2ce3974 AE |
1187 | * xfs_log_sbcount |
1188 | * | |
adab0f67 | 1189 | * Sync the superblock counters to disk. |
b2ce3974 | 1190 | * |
91ee575f BF |
1191 | * Note this code can be called during the process of freezing, so we use the |
1192 | * transaction allocator that does not block when the transaction subsystem is | |
1193 | * in its frozen state. | |
92821e2b DC |
1194 | */ |
1195 | int | |
adab0f67 | 1196 | xfs_log_sbcount(xfs_mount_t *mp) |
92821e2b | 1197 | { |
91ee575f BF |
1198 | /* allow this to proceed during the freeze sequence... */ |
1199 | if (!xfs_fs_writable(mp, SB_FREEZE_COMPLETE)) | |
92821e2b DC |
1200 | return 0; |
1201 | ||
92821e2b DC |
1202 | /* |
1203 | * we don't need to do this if we are updating the superblock | |
1204 | * counters on every modification. | |
1205 | */ | |
1206 | if (!xfs_sb_version_haslazysbcount(&mp->m_sb)) | |
1207 | return 0; | |
1208 | ||
61e63ecb | 1209 | return xfs_sync_sb(mp, true); |
92821e2b DC |
1210 | } |
1211 | ||
8c1903d3 DC |
1212 | /* |
1213 | * Deltas for the inode count are +/-64, hence we use a large batch size | |
1214 | * of 128 so we don't need to take the counter lock on every update. | |
1215 | */ | |
1216 | #define XFS_ICOUNT_BATCH 128 | |
501ab323 DC |
1217 | int |
1218 | xfs_mod_icount( | |
1219 | struct xfs_mount *mp, | |
1220 | int64_t delta) | |
1221 | { | |
104b4e51 | 1222 | percpu_counter_add_batch(&mp->m_icount, delta, XFS_ICOUNT_BATCH); |
8c1903d3 | 1223 | if (__percpu_counter_compare(&mp->m_icount, 0, XFS_ICOUNT_BATCH) < 0) { |
501ab323 DC |
1224 | ASSERT(0); |
1225 | percpu_counter_add(&mp->m_icount, -delta); | |
1226 | return -EINVAL; | |
1227 | } | |
1228 | return 0; | |
1229 | } | |
1230 | ||
e88b64ea DC |
1231 | int |
1232 | xfs_mod_ifree( | |
1233 | struct xfs_mount *mp, | |
1234 | int64_t delta) | |
1235 | { | |
1236 | percpu_counter_add(&mp->m_ifree, delta); | |
1237 | if (percpu_counter_compare(&mp->m_ifree, 0) < 0) { | |
1238 | ASSERT(0); | |
1239 | percpu_counter_add(&mp->m_ifree, -delta); | |
1240 | return -EINVAL; | |
1241 | } | |
1242 | return 0; | |
1243 | } | |
0d485ada | 1244 | |
8c1903d3 DC |
1245 | /* |
1246 | * Deltas for the block count can vary from 1 to very large, but lock contention | |
1247 | * only occurs on frequent small block count updates such as in the delayed | |
1248 | * allocation path for buffered writes (page a time updates). Hence we set | |
1249 | * a large batch count (1024) to minimise global counter updates except when | |
1250 | * we get near to ENOSPC and we have to be very accurate with our updates. | |
1251 | */ | |
1252 | #define XFS_FDBLOCKS_BATCH 1024 | |
0d485ada DC |
1253 | int |
1254 | xfs_mod_fdblocks( | |
1255 | struct xfs_mount *mp, | |
1256 | int64_t delta, | |
1257 | bool rsvd) | |
1258 | { | |
1259 | int64_t lcounter; | |
1260 | long long res_used; | |
1261 | s32 batch; | |
1262 | ||
1263 | if (delta > 0) { | |
1264 | /* | |
1265 | * If the reserve pool is depleted, put blocks back into it | |
1266 | * first. Most of the time the pool is full. | |
1267 | */ | |
1268 | if (likely(mp->m_resblks == mp->m_resblks_avail)) { | |
1269 | percpu_counter_add(&mp->m_fdblocks, delta); | |
1270 | return 0; | |
1271 | } | |
1272 | ||
1273 | spin_lock(&mp->m_sb_lock); | |
1274 | res_used = (long long)(mp->m_resblks - mp->m_resblks_avail); | |
1275 | ||
1276 | if (res_used > delta) { | |
1277 | mp->m_resblks_avail += delta; | |
1278 | } else { | |
1279 | delta -= res_used; | |
1280 | mp->m_resblks_avail = mp->m_resblks; | |
1281 | percpu_counter_add(&mp->m_fdblocks, delta); | |
1282 | } | |
1283 | spin_unlock(&mp->m_sb_lock); | |
1284 | return 0; | |
1285 | } | |
1286 | ||
1287 | /* | |
1288 | * Taking blocks away, need to be more accurate the closer we | |
1289 | * are to zero. | |
1290 | * | |
0d485ada DC |
1291 | * If the counter has a value of less than 2 * max batch size, |
1292 | * then make everything serialise as we are real close to | |
1293 | * ENOSPC. | |
1294 | */ | |
8c1903d3 DC |
1295 | if (__percpu_counter_compare(&mp->m_fdblocks, 2 * XFS_FDBLOCKS_BATCH, |
1296 | XFS_FDBLOCKS_BATCH) < 0) | |
0d485ada DC |
1297 | batch = 1; |
1298 | else | |
8c1903d3 | 1299 | batch = XFS_FDBLOCKS_BATCH; |
0d485ada | 1300 | |
104b4e51 | 1301 | percpu_counter_add_batch(&mp->m_fdblocks, delta, batch); |
52548852 | 1302 | if (__percpu_counter_compare(&mp->m_fdblocks, mp->m_alloc_set_aside, |
8c1903d3 | 1303 | XFS_FDBLOCKS_BATCH) >= 0) { |
0d485ada DC |
1304 | /* we had space! */ |
1305 | return 0; | |
1306 | } | |
1307 | ||
1308 | /* | |
1309 | * lock up the sb for dipping into reserves before releasing the space | |
1310 | * that took us to ENOSPC. | |
1311 | */ | |
1312 | spin_lock(&mp->m_sb_lock); | |
1313 | percpu_counter_add(&mp->m_fdblocks, -delta); | |
1314 | if (!rsvd) | |
1315 | goto fdblocks_enospc; | |
1316 | ||
1317 | lcounter = (long long)mp->m_resblks_avail + delta; | |
1318 | if (lcounter >= 0) { | |
1319 | mp->m_resblks_avail = lcounter; | |
1320 | spin_unlock(&mp->m_sb_lock); | |
1321 | return 0; | |
1322 | } | |
1323 | printk_once(KERN_WARNING | |
1324 | "Filesystem \"%s\": reserve blocks depleted! " | |
1325 | "Consider increasing reserve pool size.", | |
1326 | mp->m_fsname); | |
1327 | fdblocks_enospc: | |
1328 | spin_unlock(&mp->m_sb_lock); | |
1329 | return -ENOSPC; | |
1330 | } | |
1331 | ||
bab98bbe DC |
1332 | int |
1333 | xfs_mod_frextents( | |
1334 | struct xfs_mount *mp, | |
1335 | int64_t delta) | |
1336 | { | |
1337 | int64_t lcounter; | |
1338 | int ret = 0; | |
1339 | ||
1340 | spin_lock(&mp->m_sb_lock); | |
1341 | lcounter = mp->m_sb.sb_frextents + delta; | |
1342 | if (lcounter < 0) | |
1343 | ret = -ENOSPC; | |
1344 | else | |
1345 | mp->m_sb.sb_frextents = lcounter; | |
1346 | spin_unlock(&mp->m_sb_lock); | |
1347 | return ret; | |
1348 | } | |
1349 | ||
1da177e4 LT |
1350 | /* |
1351 | * xfs_getsb() is called to obtain the buffer for the superblock. | |
1352 | * The buffer is returned locked and read in from disk. | |
1353 | * The buffer should be released with a call to xfs_brelse(). | |
1354 | * | |
1355 | * If the flags parameter is BUF_TRYLOCK, then we'll only return | |
1356 | * the superblock buffer if it can be locked without sleeping. | |
1357 | * If it can't then we'll return NULL. | |
1358 | */ | |
0c842ad4 | 1359 | struct xfs_buf * |
1da177e4 | 1360 | xfs_getsb( |
0c842ad4 CH |
1361 | struct xfs_mount *mp, |
1362 | int flags) | |
1da177e4 | 1363 | { |
0c842ad4 | 1364 | struct xfs_buf *bp = mp->m_sb_bp; |
1da177e4 | 1365 | |
0c842ad4 CH |
1366 | if (!xfs_buf_trylock(bp)) { |
1367 | if (flags & XBF_TRYLOCK) | |
1da177e4 | 1368 | return NULL; |
0c842ad4 | 1369 | xfs_buf_lock(bp); |
1da177e4 | 1370 | } |
0c842ad4 | 1371 | |
72790aa1 | 1372 | xfs_buf_hold(bp); |
b0388bf1 | 1373 | ASSERT(bp->b_flags & XBF_DONE); |
014c2544 | 1374 | return bp; |
1da177e4 LT |
1375 | } |
1376 | ||
1377 | /* | |
1378 | * Used to free the superblock along various error paths. | |
1379 | */ | |
1380 | void | |
1381 | xfs_freesb( | |
26af6552 | 1382 | struct xfs_mount *mp) |
1da177e4 | 1383 | { |
26af6552 | 1384 | struct xfs_buf *bp = mp->m_sb_bp; |
1da177e4 | 1385 | |
26af6552 | 1386 | xfs_buf_lock(bp); |
1da177e4 | 1387 | mp->m_sb_bp = NULL; |
26af6552 | 1388 | xfs_buf_relse(bp); |
1da177e4 LT |
1389 | } |
1390 | ||
dda35b8f CH |
1391 | /* |
1392 | * If the underlying (data/log/rt) device is readonly, there are some | |
1393 | * operations that cannot proceed. | |
1394 | */ | |
1395 | int | |
1396 | xfs_dev_is_read_only( | |
1397 | struct xfs_mount *mp, | |
1398 | char *message) | |
1399 | { | |
1400 | if (xfs_readonly_buftarg(mp->m_ddev_targp) || | |
1401 | xfs_readonly_buftarg(mp->m_logdev_targp) || | |
1402 | (mp->m_rtdev_targp && xfs_readonly_buftarg(mp->m_rtdev_targp))) { | |
0b932ccc DC |
1403 | xfs_notice(mp, "%s required on read-only device.", message); |
1404 | xfs_notice(mp, "write access unavailable, cannot proceed."); | |
2451337d | 1405 | return -EROFS; |
dda35b8f CH |
1406 | } |
1407 | return 0; | |
1408 | } |