2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
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
20 #include "xfs_types.h"
24 #include "xfs_trans.h"
28 #include "xfs_dmapi.h"
29 #include "xfs_mount.h"
30 #include "xfs_bmap_btree.h"
31 #include "xfs_alloc_btree.h"
32 #include "xfs_ialloc_btree.h"
33 #include "xfs_dir2_sf.h"
34 #include "xfs_attr_sf.h"
35 #include "xfs_dinode.h"
36 #include "xfs_inode.h"
37 #include "xfs_inode_item.h"
38 #include "xfs_btree.h"
39 #include "xfs_error.h"
40 #include "xfs_alloc.h"
41 #include "xfs_ialloc.h"
42 #include "xfs_fsops.h"
43 #include "xfs_itable.h"
44 #include "xfs_trans_space.h"
45 #include "xfs_rtalloc.h"
47 #include "xfs_filestream.h"
48 #include "xfs_trace.h"
51 * File system operations
60 geo->blocksize = mp->m_sb.sb_blocksize;
61 geo->rtextsize = mp->m_sb.sb_rextsize;
62 geo->agblocks = mp->m_sb.sb_agblocks;
63 geo->agcount = mp->m_sb.sb_agcount;
64 geo->logblocks = mp->m_sb.sb_logblocks;
65 geo->sectsize = mp->m_sb.sb_sectsize;
66 geo->inodesize = mp->m_sb.sb_inodesize;
67 geo->imaxpct = mp->m_sb.sb_imax_pct;
68 geo->datablocks = mp->m_sb.sb_dblocks;
69 geo->rtblocks = mp->m_sb.sb_rblocks;
70 geo->rtextents = mp->m_sb.sb_rextents;
71 geo->logstart = mp->m_sb.sb_logstart;
72 ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid));
73 memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid));
74 if (new_version >= 2) {
75 geo->sunit = mp->m_sb.sb_unit;
76 geo->swidth = mp->m_sb.sb_width;
78 if (new_version >= 3) {
79 geo->version = XFS_FSOP_GEOM_VERSION;
81 (xfs_sb_version_hasattr(&mp->m_sb) ?
82 XFS_FSOP_GEOM_FLAGS_ATTR : 0) |
83 (xfs_sb_version_hasnlink(&mp->m_sb) ?
84 XFS_FSOP_GEOM_FLAGS_NLINK : 0) |
85 (xfs_sb_version_hasquota(&mp->m_sb) ?
86 XFS_FSOP_GEOM_FLAGS_QUOTA : 0) |
87 (xfs_sb_version_hasalign(&mp->m_sb) ?
88 XFS_FSOP_GEOM_FLAGS_IALIGN : 0) |
89 (xfs_sb_version_hasdalign(&mp->m_sb) ?
90 XFS_FSOP_GEOM_FLAGS_DALIGN : 0) |
91 (xfs_sb_version_hasshared(&mp->m_sb) ?
92 XFS_FSOP_GEOM_FLAGS_SHARED : 0) |
93 (xfs_sb_version_hasextflgbit(&mp->m_sb) ?
94 XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) |
95 (xfs_sb_version_hasdirv2(&mp->m_sb) ?
96 XFS_FSOP_GEOM_FLAGS_DIRV2 : 0) |
97 (xfs_sb_version_hassector(&mp->m_sb) ?
98 XFS_FSOP_GEOM_FLAGS_SECTOR : 0) |
99 (xfs_sb_version_hasasciici(&mp->m_sb) ?
100 XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) |
101 (xfs_sb_version_haslazysbcount(&mp->m_sb) ?
102 XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) |
103 (xfs_sb_version_hasattr2(&mp->m_sb) ?
104 XFS_FSOP_GEOM_FLAGS_ATTR2 : 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) {
112 (xfs_sb_version_haslogv2(&mp->m_sb) ?
113 XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
114 geo->logsunit = mp->m_sb.sb_logsunit;
120 xfs_growfs_data_private(
121 xfs_mount_t *mp, /* mount point for filesystem */
122 xfs_growfs_data_t *in) /* growfs data input struct */
128 xfs_extlen_t tmpsize;
129 xfs_alloc_rec_t *arec;
130 struct xfs_btree_block *block;
135 xfs_agnumber_t nagcount;
136 xfs_agnumber_t nagimax = 0;
137 xfs_rfsblock_t nb, nb_mod;
139 xfs_rfsblock_t nfree;
140 xfs_agnumber_t oagcount;
146 if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
147 return XFS_ERROR(EINVAL);
148 if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
150 dpct = pct - mp->m_sb.sb_imax_pct;
151 error = xfs_read_buf(mp, mp->m_ddev_targp,
152 XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
153 XFS_FSS_TO_BB(mp, 1), 0, &bp);
159 new = nb; /* use new as a temporary here */
160 nb_mod = do_div(new, mp->m_sb.sb_agblocks);
161 nagcount = new + (nb_mod != 0);
162 if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
164 nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
165 if (nb < mp->m_sb.sb_dblocks)
166 return XFS_ERROR(EINVAL);
168 new = nb - mp->m_sb.sb_dblocks;
169 oagcount = mp->m_sb.sb_agcount;
171 /* allocate the new per-ag structures */
172 if (nagcount > oagcount) {
173 error = xfs_initialize_perag(mp, nagcount, &nagimax);
178 tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
179 tp->t_flags |= XFS_TRANS_RESERVE;
180 if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp),
181 XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) {
182 xfs_trans_cancel(tp, 0);
187 * Write new AG headers to disk. Non-transactional, but written
188 * synchronously so they are completed prior to the growfs transaction
192 for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
194 * AG freelist header block
196 bp = xfs_buf_get(mp->m_ddev_targp,
197 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
198 XFS_FSS_TO_BB(mp, 1), XBF_LOCK | XBF_MAPPED);
199 agf = XFS_BUF_TO_AGF(bp);
200 memset(agf, 0, mp->m_sb.sb_sectsize);
201 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
202 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
203 agf->agf_seqno = cpu_to_be32(agno);
204 if (agno == nagcount - 1)
207 (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
209 agsize = mp->m_sb.sb_agblocks;
210 agf->agf_length = cpu_to_be32(agsize);
211 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
212 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
213 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
214 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
215 agf->agf_flfirst = 0;
216 agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
217 agf->agf_flcount = 0;
218 tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
219 agf->agf_freeblks = cpu_to_be32(tmpsize);
220 agf->agf_longest = cpu_to_be32(tmpsize);
221 error = xfs_bwrite(mp, bp);
226 * AG inode header block
228 bp = xfs_buf_get(mp->m_ddev_targp,
229 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
230 XFS_FSS_TO_BB(mp, 1), XBF_LOCK | XBF_MAPPED);
231 agi = XFS_BUF_TO_AGI(bp);
232 memset(agi, 0, mp->m_sb.sb_sectsize);
233 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
234 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
235 agi->agi_seqno = cpu_to_be32(agno);
236 agi->agi_length = cpu_to_be32(agsize);
238 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
239 agi->agi_level = cpu_to_be32(1);
240 agi->agi_freecount = 0;
241 agi->agi_newino = cpu_to_be32(NULLAGINO);
242 agi->agi_dirino = cpu_to_be32(NULLAGINO);
243 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
244 agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
245 error = xfs_bwrite(mp, bp);
250 * BNO btree root block
252 bp = xfs_buf_get(mp->m_ddev_targp,
253 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
254 BTOBB(mp->m_sb.sb_blocksize),
255 XBF_LOCK | XBF_MAPPED);
256 block = XFS_BUF_TO_BLOCK(bp);
257 memset(block, 0, mp->m_sb.sb_blocksize);
258 block->bb_magic = cpu_to_be32(XFS_ABTB_MAGIC);
260 block->bb_numrecs = cpu_to_be16(1);
261 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
262 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
263 arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
264 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
265 arec->ar_blockcount = cpu_to_be32(
266 agsize - be32_to_cpu(arec->ar_startblock));
267 error = xfs_bwrite(mp, bp);
272 * CNT btree root block
274 bp = xfs_buf_get(mp->m_ddev_targp,
275 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
276 BTOBB(mp->m_sb.sb_blocksize),
277 XBF_LOCK | XBF_MAPPED);
278 block = XFS_BUF_TO_BLOCK(bp);
279 memset(block, 0, mp->m_sb.sb_blocksize);
280 block->bb_magic = cpu_to_be32(XFS_ABTC_MAGIC);
282 block->bb_numrecs = cpu_to_be16(1);
283 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
284 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
285 arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
286 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
287 arec->ar_blockcount = cpu_to_be32(
288 agsize - be32_to_cpu(arec->ar_startblock));
289 nfree += be32_to_cpu(arec->ar_blockcount);
290 error = xfs_bwrite(mp, bp);
295 * INO btree root block
297 bp = xfs_buf_get(mp->m_ddev_targp,
298 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
299 BTOBB(mp->m_sb.sb_blocksize),
300 XBF_LOCK | XBF_MAPPED);
301 block = XFS_BUF_TO_BLOCK(bp);
302 memset(block, 0, mp->m_sb.sb_blocksize);
303 block->bb_magic = cpu_to_be32(XFS_IBT_MAGIC);
305 block->bb_numrecs = 0;
306 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
307 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
308 error = xfs_bwrite(mp, bp);
313 xfs_trans_agblocks_delta(tp, nfree);
315 * There are new blocks in the old last a.g.
319 * Change the agi length.
321 error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
326 agi = XFS_BUF_TO_AGI(bp);
327 be32_add_cpu(&agi->agi_length, new);
328 ASSERT(nagcount == oagcount ||
329 be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
330 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
334 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
339 agf = XFS_BUF_TO_AGF(bp);
340 be32_add_cpu(&agf->agf_length, new);
341 ASSERT(be32_to_cpu(agf->agf_length) ==
342 be32_to_cpu(agi->agi_length));
344 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
346 * Free the new space.
348 error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
349 be32_to_cpu(agf->agf_length) - new), new);
356 * Update changed superblock fields transactionally. These are not
357 * seen by the rest of the world until the transaction commit applies
358 * them atomically to the superblock.
360 if (nagcount > oagcount)
361 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
362 if (nb > mp->m_sb.sb_dblocks)
363 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
364 nb - mp->m_sb.sb_dblocks);
366 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
368 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
369 error = xfs_trans_commit(tp, 0);
373 /* New allocation groups fully initialized, so update mount struct */
375 mp->m_maxagi = nagimax;
376 if (mp->m_sb.sb_imax_pct) {
377 __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
379 mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
383 /* update secondary superblocks. */
384 for (agno = 1; agno < nagcount; agno++) {
385 error = xfs_read_buf(mp, mp->m_ddev_targp,
386 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
387 XFS_FSS_TO_BB(mp, 1), 0, &bp);
389 xfs_fs_cmn_err(CE_WARN, mp,
390 "error %d reading secondary superblock for ag %d",
394 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
396 * If we get an error writing out the alternate superblocks,
397 * just issue a warning and continue. The real work is
398 * already done and committed.
400 if (!(error = xfs_bwrite(mp, bp))) {
403 xfs_fs_cmn_err(CE_WARN, mp,
404 "write error %d updating secondary superblock for ag %d",
406 break; /* no point in continuing */
412 xfs_trans_cancel(tp, XFS_TRANS_ABORT);
417 xfs_growfs_log_private(
418 xfs_mount_t *mp, /* mount point for filesystem */
419 xfs_growfs_log_t *in) /* growfs log input struct */
424 if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
425 return XFS_ERROR(EINVAL);
426 if (nb == mp->m_sb.sb_logblocks &&
427 in->isint == (mp->m_sb.sb_logstart != 0))
428 return XFS_ERROR(EINVAL);
430 * Moving the log is hard, need new interfaces to sync
431 * the log first, hold off all activity while moving it.
432 * Can have shorter or longer log in the same space,
433 * or transform internal to external log or vice versa.
435 return XFS_ERROR(ENOSYS);
439 * protected versions of growfs function acquire and release locks on the mount
440 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
448 xfs_growfs_data_t *in)
452 if (!capable(CAP_SYS_ADMIN))
453 return XFS_ERROR(EPERM);
454 if (!mutex_trylock(&mp->m_growlock))
455 return XFS_ERROR(EWOULDBLOCK);
456 error = xfs_growfs_data_private(mp, in);
457 mutex_unlock(&mp->m_growlock);
464 xfs_growfs_log_t *in)
468 if (!capable(CAP_SYS_ADMIN))
469 return XFS_ERROR(EPERM);
470 if (!mutex_trylock(&mp->m_growlock))
471 return XFS_ERROR(EWOULDBLOCK);
472 error = xfs_growfs_log_private(mp, in);
473 mutex_unlock(&mp->m_growlock);
478 * exported through ioctl XFS_IOC_FSCOUNTS
484 xfs_fsop_counts_t *cnt)
486 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
487 spin_lock(&mp->m_sb_lock);
488 cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
489 cnt->freertx = mp->m_sb.sb_frextents;
490 cnt->freeino = mp->m_sb.sb_ifree;
491 cnt->allocino = mp->m_sb.sb_icount;
492 spin_unlock(&mp->m_sb_lock);
497 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
499 * xfs_reserve_blocks is called to set m_resblks
500 * in the in-core mount table. The number of unused reserved blocks
501 * is kept in m_resblks_avail.
503 * Reserve the requested number of blocks if available. Otherwise return
504 * as many as possible to satisfy the request. The actual number
505 * reserved are returned in outval
507 * A null inval pointer indicates that only the current reserved blocks
508 * available should be returned no settings are changed.
515 xfs_fsop_resblks_t *outval)
517 __int64_t lcounter, delta, fdblks_delta;
520 /* If inval is null, report current values and return */
521 if (inval == (__uint64_t *)NULL) {
524 outval->resblks = mp->m_resblks;
525 outval->resblks_avail = mp->m_resblks_avail;
532 * With per-cpu counters, this becomes an interesting
533 * problem. we needto work out if we are freeing or allocation
534 * blocks first, then we can do the modification as necessary.
536 * We do this under the m_sb_lock so that if we are near
537 * ENOSPC, we will hold out any changes while we work out
538 * what to do. This means that the amount of free space can
539 * change while we do this, so we need to retry if we end up
540 * trying to reserve more space than is available.
542 * We also use the xfs_mod_incore_sb() interface so that we
543 * don't have to care about whether per cpu counter are
544 * enabled, disabled or even compiled in....
547 spin_lock(&mp->m_sb_lock);
548 xfs_icsb_sync_counters_locked(mp, 0);
551 * If our previous reservation was larger than the current value,
552 * then move any unused blocks back to the free pool.
555 if (mp->m_resblks > request) {
556 lcounter = mp->m_resblks_avail - request;
557 if (lcounter > 0) { /* release unused blocks */
558 fdblks_delta = lcounter;
559 mp->m_resblks_avail -= lcounter;
561 mp->m_resblks = request;
565 free = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
567 goto out; /* ENOSPC and fdblks_delta = 0 */
569 delta = request - mp->m_resblks;
570 lcounter = free - delta;
572 /* We can't satisfy the request, just get what we can */
573 mp->m_resblks += free;
574 mp->m_resblks_avail += free;
575 fdblks_delta = -free;
577 fdblks_delta = -delta;
578 mp->m_resblks = request;
579 mp->m_resblks_avail += delta;
584 outval->resblks = mp->m_resblks;
585 outval->resblks_avail = mp->m_resblks_avail;
587 spin_unlock(&mp->m_sb_lock);
591 * If we are putting blocks back here, m_resblks_avail is
592 * already at its max so this will put it in the free pool.
594 * If we need space, we'll either succeed in getting it
595 * from the free block count or we'll get an enospc. If
596 * we get a ENOSPC, it means things changed while we were
597 * calculating fdblks_delta and so we should try again to
598 * see if there is anything left to reserve.
600 * Don't set the reserved flag here - we don't want to reserve
601 * the extra reserve blocks from the reserve.....
604 error = xfs_mod_incore_sb(mp, XFS_SBS_FDBLOCKS, fdblks_delta, 0);
619 tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP);
620 error = xfs_trans_reserve(tp, 0, XFS_ICHANGE_LOG_RES(mp), 0, 0, 0);
622 xfs_trans_cancel(tp, 0);
627 xfs_ilock(ip, XFS_ILOCK_EXCL);
629 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
630 xfs_trans_ihold(tp, ip);
631 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
632 xfs_trans_set_sync(tp);
633 error = xfs_trans_commit(tp, 0);
635 xfs_iunlock(ip, XFS_ILOCK_EXCL);
645 case XFS_FSOP_GOING_FLAGS_DEFAULT: {
646 struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
648 if (sb && !IS_ERR(sb)) {
649 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
650 thaw_bdev(sb->s_bdev, sb);
655 case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
656 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
658 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
659 xfs_force_shutdown(mp,
660 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
663 return XFS_ERROR(EINVAL);