]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - fs/xfs/xfs_fsops.c
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
xfs: growfs: use uncached buffers for new headers
[mirror_ubuntu-jammy-kernel.git] / fs / xfs / xfs_fsops.c
1 /*
2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
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
7 * published by the Free Software Foundation.
8 *
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.
13 *
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
17 */
18 #include "xfs.h"
19 #include "xfs_fs.h"
20 #include "xfs_types.h"
21 #include "xfs_log.h"
22 #include "xfs_trans.h"
23 #include "xfs_sb.h"
24 #include "xfs_ag.h"
25 #include "xfs_mount.h"
26 #include "xfs_bmap_btree.h"
27 #include "xfs_alloc_btree.h"
28 #include "xfs_ialloc_btree.h"
29 #include "xfs_dinode.h"
30 #include "xfs_inode.h"
31 #include "xfs_inode_item.h"
32 #include "xfs_btree.h"
33 #include "xfs_error.h"
34 #include "xfs_alloc.h"
35 #include "xfs_ialloc.h"
36 #include "xfs_fsops.h"
37 #include "xfs_itable.h"
38 #include "xfs_trans_space.h"
39 #include "xfs_rtalloc.h"
40 #include "xfs_filestream.h"
41 #include "xfs_trace.h"
42
43 /*
44 * File system operations
45 */
46
47 int
48 xfs_fs_geometry(
49 xfs_mount_t *mp,
50 xfs_fsop_geom_t *geo,
51 int new_version)
52 {
53
54 memset(geo, 0, sizeof(*geo));
55
56 geo->blocksize = mp->m_sb.sb_blocksize;
57 geo->rtextsize = mp->m_sb.sb_rextsize;
58 geo->agblocks = mp->m_sb.sb_agblocks;
59 geo->agcount = mp->m_sb.sb_agcount;
60 geo->logblocks = mp->m_sb.sb_logblocks;
61 geo->sectsize = mp->m_sb.sb_sectsize;
62 geo->inodesize = mp->m_sb.sb_inodesize;
63 geo->imaxpct = mp->m_sb.sb_imax_pct;
64 geo->datablocks = mp->m_sb.sb_dblocks;
65 geo->rtblocks = mp->m_sb.sb_rblocks;
66 geo->rtextents = mp->m_sb.sb_rextents;
67 geo->logstart = mp->m_sb.sb_logstart;
68 ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid));
69 memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid));
70 if (new_version >= 2) {
71 geo->sunit = mp->m_sb.sb_unit;
72 geo->swidth = mp->m_sb.sb_width;
73 }
74 if (new_version >= 3) {
75 geo->version = XFS_FSOP_GEOM_VERSION;
76 geo->flags =
77 (xfs_sb_version_hasattr(&mp->m_sb) ?
78 XFS_FSOP_GEOM_FLAGS_ATTR : 0) |
79 (xfs_sb_version_hasnlink(&mp->m_sb) ?
80 XFS_FSOP_GEOM_FLAGS_NLINK : 0) |
81 (xfs_sb_version_hasquota(&mp->m_sb) ?
82 XFS_FSOP_GEOM_FLAGS_QUOTA : 0) |
83 (xfs_sb_version_hasalign(&mp->m_sb) ?
84 XFS_FSOP_GEOM_FLAGS_IALIGN : 0) |
85 (xfs_sb_version_hasdalign(&mp->m_sb) ?
86 XFS_FSOP_GEOM_FLAGS_DALIGN : 0) |
87 (xfs_sb_version_hasshared(&mp->m_sb) ?
88 XFS_FSOP_GEOM_FLAGS_SHARED : 0) |
89 (xfs_sb_version_hasextflgbit(&mp->m_sb) ?
90 XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) |
91 (xfs_sb_version_hasdirv2(&mp->m_sb) ?
92 XFS_FSOP_GEOM_FLAGS_DIRV2 : 0) |
93 (xfs_sb_version_hassector(&mp->m_sb) ?
94 XFS_FSOP_GEOM_FLAGS_SECTOR : 0) |
95 (xfs_sb_version_hasasciici(&mp->m_sb) ?
96 XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) |
97 (xfs_sb_version_haslazysbcount(&mp->m_sb) ?
98 XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) |
99 (xfs_sb_version_hasattr2(&mp->m_sb) ?
100 XFS_FSOP_GEOM_FLAGS_ATTR2 : 0) |
101 (xfs_sb_version_hasprojid32bit(&mp->m_sb) ?
102 XFS_FSOP_GEOM_FLAGS_PROJID32 : 0);
103 geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
104 mp->m_sb.sb_logsectsize : BBSIZE;
105 geo->rtsectsize = mp->m_sb.sb_blocksize;
106 geo->dirblocksize = mp->m_dirblksize;
107 }
108 if (new_version >= 4) {
109 geo->flags |=
110 (xfs_sb_version_haslogv2(&mp->m_sb) ?
111 XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
112 geo->logsunit = mp->m_sb.sb_logsunit;
113 }
114 return 0;
115 }
116
117 static struct xfs_buf *
118 xfs_growfs_get_hdr_buf(
119 struct xfs_mount *mp,
120 xfs_daddr_t blkno,
121 size_t numblks,
122 int flags)
123 {
124 struct xfs_buf *bp;
125
126 bp = xfs_buf_get_uncached(mp->m_ddev_targp, numblks, flags);
127 if (!bp)
128 return NULL;
129
130 xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
131 bp->b_bn = blkno;
132 bp->b_maps[0].bm_bn = blkno;
133
134 return bp;
135 }
136
137 static int
138 xfs_growfs_data_private(
139 xfs_mount_t *mp, /* mount point for filesystem */
140 xfs_growfs_data_t *in) /* growfs data input struct */
141 {
142 xfs_agf_t *agf;
143 xfs_agi_t *agi;
144 xfs_agnumber_t agno;
145 xfs_extlen_t agsize;
146 xfs_extlen_t tmpsize;
147 xfs_alloc_rec_t *arec;
148 xfs_buf_t *bp;
149 int bucket;
150 int dpct;
151 int error;
152 xfs_agnumber_t nagcount;
153 xfs_agnumber_t nagimax = 0;
154 xfs_rfsblock_t nb, nb_mod;
155 xfs_rfsblock_t new;
156 xfs_rfsblock_t nfree;
157 xfs_agnumber_t oagcount;
158 int pct;
159 xfs_trans_t *tp;
160
161 nb = in->newblocks;
162 pct = in->imaxpct;
163 if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
164 return XFS_ERROR(EINVAL);
165 if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
166 return error;
167 dpct = pct - mp->m_sb.sb_imax_pct;
168 bp = xfs_buf_read_uncached(mp->m_ddev_targp,
169 XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
170 XFS_FSS_TO_BB(mp, 1), 0);
171 if (!bp)
172 return EIO;
173 xfs_buf_relse(bp);
174
175 new = nb; /* use new as a temporary here */
176 nb_mod = do_div(new, mp->m_sb.sb_agblocks);
177 nagcount = new + (nb_mod != 0);
178 if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
179 nagcount--;
180 nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
181 if (nb < mp->m_sb.sb_dblocks)
182 return XFS_ERROR(EINVAL);
183 }
184 new = nb - mp->m_sb.sb_dblocks;
185 oagcount = mp->m_sb.sb_agcount;
186
187 /* allocate the new per-ag structures */
188 if (nagcount > oagcount) {
189 error = xfs_initialize_perag(mp, nagcount, &nagimax);
190 if (error)
191 return error;
192 }
193
194 tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
195 tp->t_flags |= XFS_TRANS_RESERVE;
196 if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp),
197 XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) {
198 xfs_trans_cancel(tp, 0);
199 return error;
200 }
201
202 /*
203 * Write new AG headers to disk. Non-transactional, but written
204 * synchronously so they are completed prior to the growfs transaction
205 * being logged.
206 */
207 nfree = 0;
208 for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
209 /*
210 * AG freelist header block
211 */
212 bp = xfs_growfs_get_hdr_buf(mp,
213 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
214 XFS_FSS_TO_BB(mp, 1), 0);
215 if (!bp) {
216 error = ENOMEM;
217 goto error0;
218 }
219
220 agf = XFS_BUF_TO_AGF(bp);
221 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
222 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
223 agf->agf_seqno = cpu_to_be32(agno);
224 if (agno == nagcount - 1)
225 agsize =
226 nb -
227 (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
228 else
229 agsize = mp->m_sb.sb_agblocks;
230 agf->agf_length = cpu_to_be32(agsize);
231 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
232 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
233 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
234 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
235 agf->agf_flfirst = 0;
236 agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
237 agf->agf_flcount = 0;
238 tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
239 agf->agf_freeblks = cpu_to_be32(tmpsize);
240 agf->agf_longest = cpu_to_be32(tmpsize);
241 error = xfs_bwrite(bp);
242 xfs_buf_relse(bp);
243 if (error)
244 goto error0;
245
246 /*
247 * AG inode header block
248 */
249 bp = xfs_growfs_get_hdr_buf(mp,
250 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
251 XFS_FSS_TO_BB(mp, 1), 0);
252 if (!bp) {
253 error = ENOMEM;
254 goto error0;
255 }
256
257 agi = XFS_BUF_TO_AGI(bp);
258 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
259 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
260 agi->agi_seqno = cpu_to_be32(agno);
261 agi->agi_length = cpu_to_be32(agsize);
262 agi->agi_count = 0;
263 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
264 agi->agi_level = cpu_to_be32(1);
265 agi->agi_freecount = 0;
266 agi->agi_newino = cpu_to_be32(NULLAGINO);
267 agi->agi_dirino = cpu_to_be32(NULLAGINO);
268 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
269 agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
270 error = xfs_bwrite(bp);
271 xfs_buf_relse(bp);
272 if (error)
273 goto error0;
274
275 /*
276 * BNO btree root block
277 */
278 bp = xfs_growfs_get_hdr_buf(mp,
279 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
280 BTOBB(mp->m_sb.sb_blocksize), 0);
281
282 if (!bp) {
283 error = ENOMEM;
284 goto error0;
285 }
286
287 xfs_btree_init_block(mp, bp, XFS_ABTB_MAGIC, 0, 1, 0);
288 arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
289 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
290 arec->ar_blockcount = cpu_to_be32(
291 agsize - be32_to_cpu(arec->ar_startblock));
292
293 error = xfs_bwrite(bp);
294 xfs_buf_relse(bp);
295 if (error)
296 goto error0;
297
298 /*
299 * CNT btree root block
300 */
301 bp = xfs_growfs_get_hdr_buf(mp,
302 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
303 BTOBB(mp->m_sb.sb_blocksize), 0);
304 if (!bp) {
305 error = ENOMEM;
306 goto error0;
307 }
308
309 xfs_btree_init_block(mp, bp, XFS_ABTC_MAGIC, 0, 1, 0);
310 arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
311 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
312 arec->ar_blockcount = cpu_to_be32(
313 agsize - be32_to_cpu(arec->ar_startblock));
314 nfree += be32_to_cpu(arec->ar_blockcount);
315
316 error = xfs_bwrite(bp);
317 xfs_buf_relse(bp);
318 if (error)
319 goto error0;
320
321 /*
322 * INO btree root block
323 */
324 bp = xfs_growfs_get_hdr_buf(mp,
325 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
326 BTOBB(mp->m_sb.sb_blocksize), 0);
327 if (!bp) {
328 error = ENOMEM;
329 goto error0;
330 }
331
332 xfs_btree_init_block(mp, bp, XFS_IBT_MAGIC, 0, 0, 0);
333
334 error = xfs_bwrite(bp);
335 xfs_buf_relse(bp);
336 if (error)
337 goto error0;
338 }
339 xfs_trans_agblocks_delta(tp, nfree);
340 /*
341 * There are new blocks in the old last a.g.
342 */
343 if (new) {
344 /*
345 * Change the agi length.
346 */
347 error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
348 if (error) {
349 goto error0;
350 }
351 ASSERT(bp);
352 agi = XFS_BUF_TO_AGI(bp);
353 be32_add_cpu(&agi->agi_length, new);
354 ASSERT(nagcount == oagcount ||
355 be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
356 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
357 /*
358 * Change agf length.
359 */
360 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
361 if (error) {
362 goto error0;
363 }
364 ASSERT(bp);
365 agf = XFS_BUF_TO_AGF(bp);
366 be32_add_cpu(&agf->agf_length, new);
367 ASSERT(be32_to_cpu(agf->agf_length) ==
368 be32_to_cpu(agi->agi_length));
369
370 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
371 /*
372 * Free the new space.
373 */
374 error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
375 be32_to_cpu(agf->agf_length) - new), new);
376 if (error) {
377 goto error0;
378 }
379 }
380
381 /*
382 * Update changed superblock fields transactionally. These are not
383 * seen by the rest of the world until the transaction commit applies
384 * them atomically to the superblock.
385 */
386 if (nagcount > oagcount)
387 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
388 if (nb > mp->m_sb.sb_dblocks)
389 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
390 nb - mp->m_sb.sb_dblocks);
391 if (nfree)
392 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
393 if (dpct)
394 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
395 error = xfs_trans_commit(tp, 0);
396 if (error)
397 return error;
398
399 /* New allocation groups fully initialized, so update mount struct */
400 if (nagimax)
401 mp->m_maxagi = nagimax;
402 if (mp->m_sb.sb_imax_pct) {
403 __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
404 do_div(icount, 100);
405 mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
406 } else
407 mp->m_maxicount = 0;
408 xfs_set_low_space_thresholds(mp);
409
410 /* update secondary superblocks. */
411 for (agno = 1; agno < nagcount; agno++) {
412 error = 0;
413 /*
414 * new secondary superblocks need to be zeroed, not read from
415 * disk as the contents of the new area we are growing into is
416 * completely unknown.
417 */
418 if (agno < oagcount) {
419 error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
420 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
421 XFS_FSS_TO_BB(mp, 1), 0, &bp);
422 } else {
423 bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
424 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
425 XFS_FSS_TO_BB(mp, 1), 0);
426 if (bp)
427 xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
428 else
429 error = ENOMEM;
430 }
431
432 if (error) {
433 xfs_warn(mp,
434 "error %d reading secondary superblock for ag %d",
435 error, agno);
436 break;
437 }
438 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
439 /*
440 * If we get an error writing out the alternate superblocks,
441 * just issue a warning and continue. The real work is
442 * already done and committed.
443 */
444 error = xfs_bwrite(bp);
445 xfs_buf_relse(bp);
446 if (error) {
447 xfs_warn(mp,
448 "write error %d updating secondary superblock for ag %d",
449 error, agno);
450 break; /* no point in continuing */
451 }
452 }
453 return error;
454
455 error0:
456 xfs_trans_cancel(tp, XFS_TRANS_ABORT);
457 return error;
458 }
459
460 static int
461 xfs_growfs_log_private(
462 xfs_mount_t *mp, /* mount point for filesystem */
463 xfs_growfs_log_t *in) /* growfs log input struct */
464 {
465 xfs_extlen_t nb;
466
467 nb = in->newblocks;
468 if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
469 return XFS_ERROR(EINVAL);
470 if (nb == mp->m_sb.sb_logblocks &&
471 in->isint == (mp->m_sb.sb_logstart != 0))
472 return XFS_ERROR(EINVAL);
473 /*
474 * Moving the log is hard, need new interfaces to sync
475 * the log first, hold off all activity while moving it.
476 * Can have shorter or longer log in the same space,
477 * or transform internal to external log or vice versa.
478 */
479 return XFS_ERROR(ENOSYS);
480 }
481
482 /*
483 * protected versions of growfs function acquire and release locks on the mount
484 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
485 * XFS_IOC_FSGROWFSRT
486 */
487
488
489 int
490 xfs_growfs_data(
491 xfs_mount_t *mp,
492 xfs_growfs_data_t *in)
493 {
494 int error;
495
496 if (!capable(CAP_SYS_ADMIN))
497 return XFS_ERROR(EPERM);
498 if (!mutex_trylock(&mp->m_growlock))
499 return XFS_ERROR(EWOULDBLOCK);
500 error = xfs_growfs_data_private(mp, in);
501 mutex_unlock(&mp->m_growlock);
502 return error;
503 }
504
505 int
506 xfs_growfs_log(
507 xfs_mount_t *mp,
508 xfs_growfs_log_t *in)
509 {
510 int error;
511
512 if (!capable(CAP_SYS_ADMIN))
513 return XFS_ERROR(EPERM);
514 if (!mutex_trylock(&mp->m_growlock))
515 return XFS_ERROR(EWOULDBLOCK);
516 error = xfs_growfs_log_private(mp, in);
517 mutex_unlock(&mp->m_growlock);
518 return error;
519 }
520
521 /*
522 * exported through ioctl XFS_IOC_FSCOUNTS
523 */
524
525 int
526 xfs_fs_counts(
527 xfs_mount_t *mp,
528 xfs_fsop_counts_t *cnt)
529 {
530 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
531 spin_lock(&mp->m_sb_lock);
532 cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
533 cnt->freertx = mp->m_sb.sb_frextents;
534 cnt->freeino = mp->m_sb.sb_ifree;
535 cnt->allocino = mp->m_sb.sb_icount;
536 spin_unlock(&mp->m_sb_lock);
537 return 0;
538 }
539
540 /*
541 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
542 *
543 * xfs_reserve_blocks is called to set m_resblks
544 * in the in-core mount table. The number of unused reserved blocks
545 * is kept in m_resblks_avail.
546 *
547 * Reserve the requested number of blocks if available. Otherwise return
548 * as many as possible to satisfy the request. The actual number
549 * reserved are returned in outval
550 *
551 * A null inval pointer indicates that only the current reserved blocks
552 * available should be returned no settings are changed.
553 */
554
555 int
556 xfs_reserve_blocks(
557 xfs_mount_t *mp,
558 __uint64_t *inval,
559 xfs_fsop_resblks_t *outval)
560 {
561 __int64_t lcounter, delta, fdblks_delta;
562 __uint64_t request;
563
564 /* If inval is null, report current values and return */
565 if (inval == (__uint64_t *)NULL) {
566 if (!outval)
567 return EINVAL;
568 outval->resblks = mp->m_resblks;
569 outval->resblks_avail = mp->m_resblks_avail;
570 return 0;
571 }
572
573 request = *inval;
574
575 /*
576 * With per-cpu counters, this becomes an interesting
577 * problem. we needto work out if we are freeing or allocation
578 * blocks first, then we can do the modification as necessary.
579 *
580 * We do this under the m_sb_lock so that if we are near
581 * ENOSPC, we will hold out any changes while we work out
582 * what to do. This means that the amount of free space can
583 * change while we do this, so we need to retry if we end up
584 * trying to reserve more space than is available.
585 *
586 * We also use the xfs_mod_incore_sb() interface so that we
587 * don't have to care about whether per cpu counter are
588 * enabled, disabled or even compiled in....
589 */
590 retry:
591 spin_lock(&mp->m_sb_lock);
592 xfs_icsb_sync_counters_locked(mp, 0);
593
594 /*
595 * If our previous reservation was larger than the current value,
596 * then move any unused blocks back to the free pool.
597 */
598 fdblks_delta = 0;
599 if (mp->m_resblks > request) {
600 lcounter = mp->m_resblks_avail - request;
601 if (lcounter > 0) { /* release unused blocks */
602 fdblks_delta = lcounter;
603 mp->m_resblks_avail -= lcounter;
604 }
605 mp->m_resblks = request;
606 } else {
607 __int64_t free;
608
609 free = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
610 if (!free)
611 goto out; /* ENOSPC and fdblks_delta = 0 */
612
613 delta = request - mp->m_resblks;
614 lcounter = free - delta;
615 if (lcounter < 0) {
616 /* We can't satisfy the request, just get what we can */
617 mp->m_resblks += free;
618 mp->m_resblks_avail += free;
619 fdblks_delta = -free;
620 } else {
621 fdblks_delta = -delta;
622 mp->m_resblks = request;
623 mp->m_resblks_avail += delta;
624 }
625 }
626 out:
627 if (outval) {
628 outval->resblks = mp->m_resblks;
629 outval->resblks_avail = mp->m_resblks_avail;
630 }
631 spin_unlock(&mp->m_sb_lock);
632
633 if (fdblks_delta) {
634 /*
635 * If we are putting blocks back here, m_resblks_avail is
636 * already at its max so this will put it in the free pool.
637 *
638 * If we need space, we'll either succeed in getting it
639 * from the free block count or we'll get an enospc. If
640 * we get a ENOSPC, it means things changed while we were
641 * calculating fdblks_delta and so we should try again to
642 * see if there is anything left to reserve.
643 *
644 * Don't set the reserved flag here - we don't want to reserve
645 * the extra reserve blocks from the reserve.....
646 */
647 int error;
648 error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
649 fdblks_delta, 0);
650 if (error == ENOSPC)
651 goto retry;
652 }
653 return 0;
654 }
655
656 /*
657 * Dump a transaction into the log that contains no real change. This is needed
658 * to be able to make the log dirty or stamp the current tail LSN into the log
659 * during the covering operation.
660 *
661 * We cannot use an inode here for this - that will push dirty state back up
662 * into the VFS and then periodic inode flushing will prevent log covering from
663 * making progress. Hence we log a field in the superblock instead and use a
664 * synchronous transaction to ensure the superblock is immediately unpinned
665 * and can be written back.
666 */
667 int
668 xfs_fs_log_dummy(
669 xfs_mount_t *mp)
670 {
671 xfs_trans_t *tp;
672 int error;
673
674 tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP);
675 error = xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0,
676 XFS_DEFAULT_LOG_COUNT);
677 if (error) {
678 xfs_trans_cancel(tp, 0);
679 return error;
680 }
681
682 /* log the UUID because it is an unchanging field */
683 xfs_mod_sb(tp, XFS_SB_UUID);
684 xfs_trans_set_sync(tp);
685 return xfs_trans_commit(tp, 0);
686 }
687
688 int
689 xfs_fs_goingdown(
690 xfs_mount_t *mp,
691 __uint32_t inflags)
692 {
693 switch (inflags) {
694 case XFS_FSOP_GOING_FLAGS_DEFAULT: {
695 struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
696
697 if (sb && !IS_ERR(sb)) {
698 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
699 thaw_bdev(sb->s_bdev, sb);
700 }
701
702 break;
703 }
704 case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
705 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
706 break;
707 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
708 xfs_force_shutdown(mp,
709 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
710 break;
711 default:
712 return XFS_ERROR(EINVAL);
713 }
714
715 return 0;
716 }
717
718 /*
719 * Force a shutdown of the filesystem instantly while keeping the filesystem
720 * consistent. We don't do an unmount here; just shutdown the shop, make sure
721 * that absolutely nothing persistent happens to this filesystem after this
722 * point.
723 */
724 void
725 xfs_do_force_shutdown(
726 xfs_mount_t *mp,
727 int flags,
728 char *fname,
729 int lnnum)
730 {
731 int logerror;
732
733 logerror = flags & SHUTDOWN_LOG_IO_ERROR;
734
735 if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
736 xfs_notice(mp,
737 "%s(0x%x) called from line %d of file %s. Return address = 0x%p",
738 __func__, flags, lnnum, fname, __return_address);
739 }
740 /*
741 * No need to duplicate efforts.
742 */
743 if (XFS_FORCED_SHUTDOWN(mp) && !logerror)
744 return;
745
746 /*
747 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
748 * queue up anybody new on the log reservations, and wakes up
749 * everybody who's sleeping on log reservations to tell them
750 * the bad news.
751 */
752 if (xfs_log_force_umount(mp, logerror))
753 return;
754
755 if (flags & SHUTDOWN_CORRUPT_INCORE) {
756 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_CORRUPT,
757 "Corruption of in-memory data detected. Shutting down filesystem");
758 if (XFS_ERRLEVEL_HIGH <= xfs_error_level)
759 xfs_stack_trace();
760 } else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
761 if (logerror) {
762 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_LOGERROR,
763 "Log I/O Error Detected. Shutting down filesystem");
764 } else if (flags & SHUTDOWN_DEVICE_REQ) {
765 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
766 "All device paths lost. Shutting down filesystem");
767 } else if (!(flags & SHUTDOWN_REMOTE_REQ)) {
768 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
769 "I/O Error Detected. Shutting down filesystem");
770 }
771 }
772 if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
773 xfs_alert(mp,
774 "Please umount the filesystem and rectify the problem(s)");
775 }
776 }
Ubuntu Jammy Linux kernel mirror