ARM: imx: Include linux/err.h
[linux-2.6.git] / fs / xfs / xfs_fsops.c
blobe64ee5288b86be2d0c0267b383d3f6f9297e60a1
1 /*
2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
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.
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.
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
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"
44 * File system operations
47 int
48 xfs_fs_geometry(
49 xfs_mount_t *mp,
50 xfs_fsop_geom_t *geo,
51 int new_version)
54 memset(geo, 0, sizeof(*geo));
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;
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 (xfs_sb_version_hascrc(&mp->m_sb) ?
104 XFS_FSOP_GEOM_FLAGS_V5SB : 0);
105 geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
106 mp->m_sb.sb_logsectsize : BBSIZE;
107 geo->rtsectsize = mp->m_sb.sb_blocksize;
108 geo->dirblocksize = mp->m_dirblksize;
110 if (new_version >= 4) {
111 geo->flags |=
112 (xfs_sb_version_haslogv2(&mp->m_sb) ?
113 XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
114 geo->logsunit = mp->m_sb.sb_logsunit;
116 return 0;
119 static struct xfs_buf *
120 xfs_growfs_get_hdr_buf(
121 struct xfs_mount *mp,
122 xfs_daddr_t blkno,
123 size_t numblks,
124 int flags,
125 const struct xfs_buf_ops *ops)
127 struct xfs_buf *bp;
129 bp = xfs_buf_get_uncached(mp->m_ddev_targp, numblks, flags);
130 if (!bp)
131 return NULL;
133 xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
134 bp->b_bn = blkno;
135 bp->b_maps[0].bm_bn = blkno;
136 bp->b_ops = ops;
138 return bp;
141 static int
142 xfs_growfs_data_private(
143 xfs_mount_t *mp, /* mount point for filesystem */
144 xfs_growfs_data_t *in) /* growfs data input struct */
146 xfs_agf_t *agf;
147 struct xfs_agfl *agfl;
148 xfs_agi_t *agi;
149 xfs_agnumber_t agno;
150 xfs_extlen_t agsize;
151 xfs_extlen_t tmpsize;
152 xfs_alloc_rec_t *arec;
153 xfs_buf_t *bp;
154 int bucket;
155 int dpct;
156 int error;
157 xfs_agnumber_t nagcount;
158 xfs_agnumber_t nagimax = 0;
159 xfs_rfsblock_t nb, nb_mod;
160 xfs_rfsblock_t new;
161 xfs_rfsblock_t nfree;
162 xfs_agnumber_t oagcount;
163 int pct;
164 xfs_trans_t *tp;
166 nb = in->newblocks;
167 pct = in->imaxpct;
168 if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
169 return XFS_ERROR(EINVAL);
170 if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
171 return error;
172 dpct = pct - mp->m_sb.sb_imax_pct;
173 bp = xfs_buf_read_uncached(mp->m_ddev_targp,
174 XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
175 XFS_FSS_TO_BB(mp, 1), 0, NULL);
176 if (!bp)
177 return EIO;
178 if (bp->b_error) {
179 error = bp->b_error;
180 xfs_buf_relse(bp);
181 return error;
183 xfs_buf_relse(bp);
185 new = nb; /* use new as a temporary here */
186 nb_mod = do_div(new, mp->m_sb.sb_agblocks);
187 nagcount = new + (nb_mod != 0);
188 if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
189 nagcount--;
190 nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
191 if (nb < mp->m_sb.sb_dblocks)
192 return XFS_ERROR(EINVAL);
194 new = nb - mp->m_sb.sb_dblocks;
195 oagcount = mp->m_sb.sb_agcount;
197 /* allocate the new per-ag structures */
198 if (nagcount > oagcount) {
199 error = xfs_initialize_perag(mp, nagcount, &nagimax);
200 if (error)
201 return error;
204 tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
205 tp->t_flags |= XFS_TRANS_RESERVE;
206 error = xfs_trans_reserve(tp, &M_RES(mp)->tr_growdata,
207 XFS_GROWFS_SPACE_RES(mp), 0);
208 if (error) {
209 xfs_trans_cancel(tp, 0);
210 return error;
214 * Write new AG headers to disk. Non-transactional, but written
215 * synchronously so they are completed prior to the growfs transaction
216 * being logged.
218 nfree = 0;
219 for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
221 * AG freespace header block
223 bp = xfs_growfs_get_hdr_buf(mp,
224 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
225 XFS_FSS_TO_BB(mp, 1), 0,
226 &xfs_agf_buf_ops);
227 if (!bp) {
228 error = ENOMEM;
229 goto error0;
232 agf = XFS_BUF_TO_AGF(bp);
233 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
234 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
235 agf->agf_seqno = cpu_to_be32(agno);
236 if (agno == nagcount - 1)
237 agsize =
238 nb -
239 (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
240 else
241 agsize = mp->m_sb.sb_agblocks;
242 agf->agf_length = cpu_to_be32(agsize);
243 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
244 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
245 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
246 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
247 agf->agf_flfirst = 0;
248 agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
249 agf->agf_flcount = 0;
250 tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
251 agf->agf_freeblks = cpu_to_be32(tmpsize);
252 agf->agf_longest = cpu_to_be32(tmpsize);
253 if (xfs_sb_version_hascrc(&mp->m_sb))
254 uuid_copy(&agf->agf_uuid, &mp->m_sb.sb_uuid);
256 error = xfs_bwrite(bp);
257 xfs_buf_relse(bp);
258 if (error)
259 goto error0;
262 * AG freelist header block
264 bp = xfs_growfs_get_hdr_buf(mp,
265 XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
266 XFS_FSS_TO_BB(mp, 1), 0,
267 &xfs_agfl_buf_ops);
268 if (!bp) {
269 error = ENOMEM;
270 goto error0;
273 agfl = XFS_BUF_TO_AGFL(bp);
274 if (xfs_sb_version_hascrc(&mp->m_sb)) {
275 agfl->agfl_magicnum = cpu_to_be32(XFS_AGFL_MAGIC);
276 agfl->agfl_seqno = cpu_to_be32(agno);
277 uuid_copy(&agfl->agfl_uuid, &mp->m_sb.sb_uuid);
279 for (bucket = 0; bucket < XFS_AGFL_SIZE(mp); bucket++)
280 agfl->agfl_bno[bucket] = cpu_to_be32(NULLAGBLOCK);
282 error = xfs_bwrite(bp);
283 xfs_buf_relse(bp);
284 if (error)
285 goto error0;
288 * AG inode header block
290 bp = xfs_growfs_get_hdr_buf(mp,
291 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
292 XFS_FSS_TO_BB(mp, 1), 0,
293 &xfs_agi_buf_ops);
294 if (!bp) {
295 error = ENOMEM;
296 goto error0;
299 agi = XFS_BUF_TO_AGI(bp);
300 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
301 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
302 agi->agi_seqno = cpu_to_be32(agno);
303 agi->agi_length = cpu_to_be32(agsize);
304 agi->agi_count = 0;
305 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
306 agi->agi_level = cpu_to_be32(1);
307 agi->agi_freecount = 0;
308 agi->agi_newino = cpu_to_be32(NULLAGINO);
309 agi->agi_dirino = cpu_to_be32(NULLAGINO);
310 if (xfs_sb_version_hascrc(&mp->m_sb))
311 uuid_copy(&agi->agi_uuid, &mp->m_sb.sb_uuid);
312 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
313 agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
315 error = xfs_bwrite(bp);
316 xfs_buf_relse(bp);
317 if (error)
318 goto error0;
321 * BNO btree root block
323 bp = xfs_growfs_get_hdr_buf(mp,
324 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
325 BTOBB(mp->m_sb.sb_blocksize), 0,
326 &xfs_allocbt_buf_ops);
328 if (!bp) {
329 error = ENOMEM;
330 goto error0;
333 if (xfs_sb_version_hascrc(&mp->m_sb))
334 xfs_btree_init_block(mp, bp, XFS_ABTB_CRC_MAGIC, 0, 1,
335 agno, XFS_BTREE_CRC_BLOCKS);
336 else
337 xfs_btree_init_block(mp, bp, XFS_ABTB_MAGIC, 0, 1,
338 agno, 0);
340 arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
341 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
342 arec->ar_blockcount = cpu_to_be32(
343 agsize - be32_to_cpu(arec->ar_startblock));
345 error = xfs_bwrite(bp);
346 xfs_buf_relse(bp);
347 if (error)
348 goto error0;
351 * CNT btree root block
353 bp = xfs_growfs_get_hdr_buf(mp,
354 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
355 BTOBB(mp->m_sb.sb_blocksize), 0,
356 &xfs_allocbt_buf_ops);
357 if (!bp) {
358 error = ENOMEM;
359 goto error0;
362 if (xfs_sb_version_hascrc(&mp->m_sb))
363 xfs_btree_init_block(mp, bp, XFS_ABTC_CRC_MAGIC, 0, 1,
364 agno, XFS_BTREE_CRC_BLOCKS);
365 else
366 xfs_btree_init_block(mp, bp, XFS_ABTC_MAGIC, 0, 1,
367 agno, 0);
369 arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
370 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
371 arec->ar_blockcount = cpu_to_be32(
372 agsize - be32_to_cpu(arec->ar_startblock));
373 nfree += be32_to_cpu(arec->ar_blockcount);
375 error = xfs_bwrite(bp);
376 xfs_buf_relse(bp);
377 if (error)
378 goto error0;
381 * INO btree root block
383 bp = xfs_growfs_get_hdr_buf(mp,
384 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
385 BTOBB(mp->m_sb.sb_blocksize), 0,
386 &xfs_inobt_buf_ops);
387 if (!bp) {
388 error = ENOMEM;
389 goto error0;
392 if (xfs_sb_version_hascrc(&mp->m_sb))
393 xfs_btree_init_block(mp, bp, XFS_IBT_CRC_MAGIC, 0, 0,
394 agno, XFS_BTREE_CRC_BLOCKS);
395 else
396 xfs_btree_init_block(mp, bp, XFS_IBT_MAGIC, 0, 0,
397 agno, 0);
399 error = xfs_bwrite(bp);
400 xfs_buf_relse(bp);
401 if (error)
402 goto error0;
404 xfs_trans_agblocks_delta(tp, nfree);
406 * There are new blocks in the old last a.g.
408 if (new) {
410 * Change the agi length.
412 error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
413 if (error) {
414 goto error0;
416 ASSERT(bp);
417 agi = XFS_BUF_TO_AGI(bp);
418 be32_add_cpu(&agi->agi_length, new);
419 ASSERT(nagcount == oagcount ||
420 be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
421 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
423 * Change agf length.
425 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
426 if (error) {
427 goto error0;
429 ASSERT(bp);
430 agf = XFS_BUF_TO_AGF(bp);
431 be32_add_cpu(&agf->agf_length, new);
432 ASSERT(be32_to_cpu(agf->agf_length) ==
433 be32_to_cpu(agi->agi_length));
435 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
437 * Free the new space.
439 error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
440 be32_to_cpu(agf->agf_length) - new), new);
441 if (error) {
442 goto error0;
447 * Update changed superblock fields transactionally. These are not
448 * seen by the rest of the world until the transaction commit applies
449 * them atomically to the superblock.
451 if (nagcount > oagcount)
452 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
453 if (nb > mp->m_sb.sb_dblocks)
454 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
455 nb - mp->m_sb.sb_dblocks);
456 if (nfree)
457 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
458 if (dpct)
459 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
460 error = xfs_trans_commit(tp, 0);
461 if (error)
462 return error;
464 /* New allocation groups fully initialized, so update mount struct */
465 if (nagimax)
466 mp->m_maxagi = nagimax;
467 if (mp->m_sb.sb_imax_pct) {
468 __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
469 do_div(icount, 100);
470 mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
471 } else
472 mp->m_maxicount = 0;
473 xfs_set_low_space_thresholds(mp);
475 /* update secondary superblocks. */
476 for (agno = 1; agno < nagcount; agno++) {
477 error = 0;
479 * new secondary superblocks need to be zeroed, not read from
480 * disk as the contents of the new area we are growing into is
481 * completely unknown.
483 if (agno < oagcount) {
484 error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
485 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
486 XFS_FSS_TO_BB(mp, 1), 0, &bp,
487 &xfs_sb_buf_ops);
488 } else {
489 bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
490 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
491 XFS_FSS_TO_BB(mp, 1), 0);
492 if (bp) {
493 bp->b_ops = &xfs_sb_buf_ops;
494 xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
495 } else
496 error = ENOMEM;
499 if (error) {
500 xfs_warn(mp,
501 "error %d reading secondary superblock for ag %d",
502 error, agno);
503 break;
505 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
508 * If we get an error writing out the alternate superblocks,
509 * just issue a warning and continue. The real work is
510 * already done and committed.
512 error = xfs_bwrite(bp);
513 xfs_buf_relse(bp);
514 if (error) {
515 xfs_warn(mp,
516 "write error %d updating secondary superblock for ag %d",
517 error, agno);
518 break; /* no point in continuing */
521 return error;
523 error0:
524 xfs_trans_cancel(tp, XFS_TRANS_ABORT);
525 return error;
528 static int
529 xfs_growfs_log_private(
530 xfs_mount_t *mp, /* mount point for filesystem */
531 xfs_growfs_log_t *in) /* growfs log input struct */
533 xfs_extlen_t nb;
535 nb = in->newblocks;
536 if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
537 return XFS_ERROR(EINVAL);
538 if (nb == mp->m_sb.sb_logblocks &&
539 in->isint == (mp->m_sb.sb_logstart != 0))
540 return XFS_ERROR(EINVAL);
542 * Moving the log is hard, need new interfaces to sync
543 * the log first, hold off all activity while moving it.
544 * Can have shorter or longer log in the same space,
545 * or transform internal to external log or vice versa.
547 return XFS_ERROR(ENOSYS);
551 * protected versions of growfs function acquire and release locks on the mount
552 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
553 * XFS_IOC_FSGROWFSRT
558 xfs_growfs_data(
559 xfs_mount_t *mp,
560 xfs_growfs_data_t *in)
562 int error;
564 if (!capable(CAP_SYS_ADMIN))
565 return XFS_ERROR(EPERM);
566 if (!mutex_trylock(&mp->m_growlock))
567 return XFS_ERROR(EWOULDBLOCK);
568 error = xfs_growfs_data_private(mp, in);
569 mutex_unlock(&mp->m_growlock);
570 return error;
574 xfs_growfs_log(
575 xfs_mount_t *mp,
576 xfs_growfs_log_t *in)
578 int error;
580 if (!capable(CAP_SYS_ADMIN))
581 return XFS_ERROR(EPERM);
582 if (!mutex_trylock(&mp->m_growlock))
583 return XFS_ERROR(EWOULDBLOCK);
584 error = xfs_growfs_log_private(mp, in);
585 mutex_unlock(&mp->m_growlock);
586 return error;
590 * exported through ioctl XFS_IOC_FSCOUNTS
594 xfs_fs_counts(
595 xfs_mount_t *mp,
596 xfs_fsop_counts_t *cnt)
598 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
599 spin_lock(&mp->m_sb_lock);
600 cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
601 cnt->freertx = mp->m_sb.sb_frextents;
602 cnt->freeino = mp->m_sb.sb_ifree;
603 cnt->allocino = mp->m_sb.sb_icount;
604 spin_unlock(&mp->m_sb_lock);
605 return 0;
609 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
611 * xfs_reserve_blocks is called to set m_resblks
612 * in the in-core mount table. The number of unused reserved blocks
613 * is kept in m_resblks_avail.
615 * Reserve the requested number of blocks if available. Otherwise return
616 * as many as possible to satisfy the request. The actual number
617 * reserved are returned in outval
619 * A null inval pointer indicates that only the current reserved blocks
620 * available should be returned no settings are changed.
624 xfs_reserve_blocks(
625 xfs_mount_t *mp,
626 __uint64_t *inval,
627 xfs_fsop_resblks_t *outval)
629 __int64_t lcounter, delta, fdblks_delta;
630 __uint64_t request;
632 /* If inval is null, report current values and return */
633 if (inval == (__uint64_t *)NULL) {
634 if (!outval)
635 return EINVAL;
636 outval->resblks = mp->m_resblks;
637 outval->resblks_avail = mp->m_resblks_avail;
638 return 0;
641 request = *inval;
644 * With per-cpu counters, this becomes an interesting
645 * problem. we needto work out if we are freeing or allocation
646 * blocks first, then we can do the modification as necessary.
648 * We do this under the m_sb_lock so that if we are near
649 * ENOSPC, we will hold out any changes while we work out
650 * what to do. This means that the amount of free space can
651 * change while we do this, so we need to retry if we end up
652 * trying to reserve more space than is available.
654 * We also use the xfs_mod_incore_sb() interface so that we
655 * don't have to care about whether per cpu counter are
656 * enabled, disabled or even compiled in....
658 retry:
659 spin_lock(&mp->m_sb_lock);
660 xfs_icsb_sync_counters_locked(mp, 0);
663 * If our previous reservation was larger than the current value,
664 * then move any unused blocks back to the free pool.
666 fdblks_delta = 0;
667 if (mp->m_resblks > request) {
668 lcounter = mp->m_resblks_avail - request;
669 if (lcounter > 0) { /* release unused blocks */
670 fdblks_delta = lcounter;
671 mp->m_resblks_avail -= lcounter;
673 mp->m_resblks = request;
674 } else {
675 __int64_t free;
677 free = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
678 if (!free)
679 goto out; /* ENOSPC and fdblks_delta = 0 */
681 delta = request - mp->m_resblks;
682 lcounter = free - delta;
683 if (lcounter < 0) {
684 /* We can't satisfy the request, just get what we can */
685 mp->m_resblks += free;
686 mp->m_resblks_avail += free;
687 fdblks_delta = -free;
688 } else {
689 fdblks_delta = -delta;
690 mp->m_resblks = request;
691 mp->m_resblks_avail += delta;
694 out:
695 if (outval) {
696 outval->resblks = mp->m_resblks;
697 outval->resblks_avail = mp->m_resblks_avail;
699 spin_unlock(&mp->m_sb_lock);
701 if (fdblks_delta) {
703 * If we are putting blocks back here, m_resblks_avail is
704 * already at its max so this will put it in the free pool.
706 * If we need space, we'll either succeed in getting it
707 * from the free block count or we'll get an enospc. If
708 * we get a ENOSPC, it means things changed while we were
709 * calculating fdblks_delta and so we should try again to
710 * see if there is anything left to reserve.
712 * Don't set the reserved flag here - we don't want to reserve
713 * the extra reserve blocks from the reserve.....
715 int error;
716 error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
717 fdblks_delta, 0);
718 if (error == ENOSPC)
719 goto retry;
721 return 0;
725 * Dump a transaction into the log that contains no real change. This is needed
726 * to be able to make the log dirty or stamp the current tail LSN into the log
727 * during the covering operation.
729 * We cannot use an inode here for this - that will push dirty state back up
730 * into the VFS and then periodic inode flushing will prevent log covering from
731 * making progress. Hence we log a field in the superblock instead and use a
732 * synchronous transaction to ensure the superblock is immediately unpinned
733 * and can be written back.
736 xfs_fs_log_dummy(
737 xfs_mount_t *mp)
739 xfs_trans_t *tp;
740 int error;
742 tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP);
743 error = xfs_trans_reserve(tp, &M_RES(mp)->tr_sb, 0, 0);
744 if (error) {
745 xfs_trans_cancel(tp, 0);
746 return error;
749 /* log the UUID because it is an unchanging field */
750 xfs_mod_sb(tp, XFS_SB_UUID);
751 xfs_trans_set_sync(tp);
752 return xfs_trans_commit(tp, 0);
756 xfs_fs_goingdown(
757 xfs_mount_t *mp,
758 __uint32_t inflags)
760 switch (inflags) {
761 case XFS_FSOP_GOING_FLAGS_DEFAULT: {
762 struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
764 if (sb && !IS_ERR(sb)) {
765 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
766 thaw_bdev(sb->s_bdev, sb);
769 break;
771 case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
772 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
773 break;
774 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
775 xfs_force_shutdown(mp,
776 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
777 break;
778 default:
779 return XFS_ERROR(EINVAL);
782 return 0;
786 * Force a shutdown of the filesystem instantly while keeping the filesystem
787 * consistent. We don't do an unmount here; just shutdown the shop, make sure
788 * that absolutely nothing persistent happens to this filesystem after this
789 * point.
791 void
792 xfs_do_force_shutdown(
793 xfs_mount_t *mp,
794 int flags,
795 char *fname,
796 int lnnum)
798 int logerror;
800 logerror = flags & SHUTDOWN_LOG_IO_ERROR;
802 if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
803 xfs_notice(mp,
804 "%s(0x%x) called from line %d of file %s. Return address = 0x%p",
805 __func__, flags, lnnum, fname, __return_address);
808 * No need to duplicate efforts.
810 if (XFS_FORCED_SHUTDOWN(mp) && !logerror)
811 return;
814 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
815 * queue up anybody new on the log reservations, and wakes up
816 * everybody who's sleeping on log reservations to tell them
817 * the bad news.
819 if (xfs_log_force_umount(mp, logerror))
820 return;
822 if (flags & SHUTDOWN_CORRUPT_INCORE) {
823 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_CORRUPT,
824 "Corruption of in-memory data detected. Shutting down filesystem");
825 if (XFS_ERRLEVEL_HIGH <= xfs_error_level)
826 xfs_stack_trace();
827 } else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
828 if (logerror) {
829 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_LOGERROR,
830 "Log I/O Error Detected. Shutting down filesystem");
831 } else if (flags & SHUTDOWN_DEVICE_REQ) {
832 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
833 "All device paths lost. Shutting down filesystem");
834 } else if (!(flags & SHUTDOWN_REMOTE_REQ)) {
835 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
836 "I/O Error Detected. Shutting down filesystem");
839 if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
840 xfs_alert(mp,
841 "Please umount the filesystem and rectify the problem(s)");