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"
23 #include "xfs_trans.h"
26 #include "xfs_mount.h"
27 #include "xfs_bmap_btree.h"
28 #include "xfs_alloc_btree.h"
29 #include "xfs_ialloc_btree.h"
30 #include "xfs_dinode.h"
31 #include "xfs_inode.h"
32 #include "xfs_inode_item.h"
33 #include "xfs_btree.h"
34 #include "xfs_error.h"
35 #include "xfs_alloc.h"
36 #include "xfs_ialloc.h"
37 #include "xfs_fsops.h"
38 #include "xfs_itable.h"
39 #include "xfs_trans_space.h"
40 #include "xfs_rtalloc.h"
41 #include "xfs_filestream.h"
42 #include "xfs_trace.h"
45 * File system operations
55 memset(geo, 0, sizeof(*geo));
57 geo->blocksize = mp->m_sb.sb_blocksize;
58 geo->rtextsize = mp->m_sb.sb_rextsize;
59 geo->agblocks = mp->m_sb.sb_agblocks;
60 geo->agcount = mp->m_sb.sb_agcount;
61 geo->logblocks = mp->m_sb.sb_logblocks;
62 geo->sectsize = mp->m_sb.sb_sectsize;
63 geo->inodesize = mp->m_sb.sb_inodesize;
64 geo->imaxpct = mp->m_sb.sb_imax_pct;
65 geo->datablocks = mp->m_sb.sb_dblocks;
66 geo->rtblocks = mp->m_sb.sb_rblocks;
67 geo->rtextents = mp->m_sb.sb_rextents;
68 geo->logstart = mp->m_sb.sb_logstart;
69 ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid));
70 memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid));
71 if (new_version >= 2) {
72 geo->sunit = mp->m_sb.sb_unit;
73 geo->swidth = mp->m_sb.sb_width;
75 if (new_version >= 3) {
76 geo->version = XFS_FSOP_GEOM_VERSION;
78 (xfs_sb_version_hasattr(&mp->m_sb) ?
79 XFS_FSOP_GEOM_FLAGS_ATTR : 0) |
80 (xfs_sb_version_hasnlink(&mp->m_sb) ?
81 XFS_FSOP_GEOM_FLAGS_NLINK : 0) |
82 (xfs_sb_version_hasquota(&mp->m_sb) ?
83 XFS_FSOP_GEOM_FLAGS_QUOTA : 0) |
84 (xfs_sb_version_hasalign(&mp->m_sb) ?
85 XFS_FSOP_GEOM_FLAGS_IALIGN : 0) |
86 (xfs_sb_version_hasdalign(&mp->m_sb) ?
87 XFS_FSOP_GEOM_FLAGS_DALIGN : 0) |
88 (xfs_sb_version_hasshared(&mp->m_sb) ?
89 XFS_FSOP_GEOM_FLAGS_SHARED : 0) |
90 (xfs_sb_version_hasextflgbit(&mp->m_sb) ?
91 XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) |
92 (xfs_sb_version_hasdirv2(&mp->m_sb) ?
93 XFS_FSOP_GEOM_FLAGS_DIRV2 : 0) |
94 (xfs_sb_version_hassector(&mp->m_sb) ?
95 XFS_FSOP_GEOM_FLAGS_SECTOR : 0) |
96 (xfs_sb_version_hasasciici(&mp->m_sb) ?
97 XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) |
98 (xfs_sb_version_haslazysbcount(&mp->m_sb) ?
99 XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) |
100 (xfs_sb_version_hasattr2(&mp->m_sb) ?
101 XFS_FSOP_GEOM_FLAGS_ATTR2 : 0);
102 geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
103 mp->m_sb.sb_logsectsize : BBSIZE;
104 geo->rtsectsize = mp->m_sb.sb_blocksize;
105 geo->dirblocksize = mp->m_dirblksize;
107 if (new_version >= 4) {
109 (xfs_sb_version_haslogv2(&mp->m_sb) ?
110 XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
111 geo->logsunit = mp->m_sb.sb_logsunit;
117 xfs_growfs_data_private(
118 xfs_mount_t *mp, /* mount point for filesystem */
119 xfs_growfs_data_t *in) /* growfs data input struct */
125 xfs_extlen_t tmpsize;
126 xfs_alloc_rec_t *arec;
127 struct xfs_btree_block *block;
132 xfs_agnumber_t nagcount;
133 xfs_agnumber_t nagimax = 0;
134 xfs_rfsblock_t nb, nb_mod;
136 xfs_rfsblock_t nfree;
137 xfs_agnumber_t oagcount;
143 if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
144 return XFS_ERROR(EINVAL);
145 if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
147 dpct = pct - mp->m_sb.sb_imax_pct;
148 bp = xfs_buf_read_uncached(mp->m_ddev_targp,
149 XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
150 XFS_FSS_TO_BB(mp, 1), 0);
155 new = nb; /* use new as a temporary here */
156 nb_mod = do_div(new, mp->m_sb.sb_agblocks);
157 nagcount = new + (nb_mod != 0);
158 if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
160 nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
161 if (nb < mp->m_sb.sb_dblocks)
162 return XFS_ERROR(EINVAL);
164 new = nb - mp->m_sb.sb_dblocks;
165 oagcount = mp->m_sb.sb_agcount;
167 /* allocate the new per-ag structures */
168 if (nagcount > oagcount) {
169 error = xfs_initialize_perag(mp, nagcount, &nagimax);
174 tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
175 tp->t_flags |= XFS_TRANS_RESERVE;
176 if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp),
177 XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) {
178 xfs_trans_cancel(tp, 0);
183 * Write new AG headers to disk. Non-transactional, but written
184 * synchronously so they are completed prior to the growfs transaction
188 for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
190 * AG freelist header block
192 bp = xfs_buf_get(mp->m_ddev_targp,
193 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
194 XFS_FSS_TO_BB(mp, 1), 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(bp);
227 * AG inode header block
229 bp = xfs_buf_get(mp->m_ddev_targp,
230 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
231 XFS_FSS_TO_BB(mp, 1), XBF_MAPPED);
236 agi = XFS_BUF_TO_AGI(bp);
237 memset(agi, 0, mp->m_sb.sb_sectsize);
238 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
239 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
240 agi->agi_seqno = cpu_to_be32(agno);
241 agi->agi_length = cpu_to_be32(agsize);
243 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
244 agi->agi_level = cpu_to_be32(1);
245 agi->agi_freecount = 0;
246 agi->agi_newino = cpu_to_be32(NULLAGINO);
247 agi->agi_dirino = cpu_to_be32(NULLAGINO);
248 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
249 agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
250 error = xfs_bwrite(bp);
256 * BNO btree root block
258 bp = xfs_buf_get(mp->m_ddev_targp,
259 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
260 BTOBB(mp->m_sb.sb_blocksize), XBF_MAPPED);
265 block = XFS_BUF_TO_BLOCK(bp);
266 memset(block, 0, mp->m_sb.sb_blocksize);
267 block->bb_magic = cpu_to_be32(XFS_ABTB_MAGIC);
269 block->bb_numrecs = cpu_to_be16(1);
270 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
271 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
272 arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
273 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
274 arec->ar_blockcount = cpu_to_be32(
275 agsize - be32_to_cpu(arec->ar_startblock));
276 error = xfs_bwrite(bp);
282 * CNT btree root block
284 bp = xfs_buf_get(mp->m_ddev_targp,
285 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
286 BTOBB(mp->m_sb.sb_blocksize), XBF_MAPPED);
291 block = XFS_BUF_TO_BLOCK(bp);
292 memset(block, 0, mp->m_sb.sb_blocksize);
293 block->bb_magic = cpu_to_be32(XFS_ABTC_MAGIC);
295 block->bb_numrecs = cpu_to_be16(1);
296 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
297 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
298 arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
299 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
300 arec->ar_blockcount = cpu_to_be32(
301 agsize - be32_to_cpu(arec->ar_startblock));
302 nfree += be32_to_cpu(arec->ar_blockcount);
303 error = xfs_bwrite(bp);
309 * INO btree root block
311 bp = xfs_buf_get(mp->m_ddev_targp,
312 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
313 BTOBB(mp->m_sb.sb_blocksize), XBF_MAPPED);
318 block = XFS_BUF_TO_BLOCK(bp);
319 memset(block, 0, mp->m_sb.sb_blocksize);
320 block->bb_magic = cpu_to_be32(XFS_IBT_MAGIC);
322 block->bb_numrecs = 0;
323 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
324 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
325 error = xfs_bwrite(bp);
330 xfs_trans_agblocks_delta(tp, nfree);
332 * There are new blocks in the old last a.g.
336 * Change the agi length.
338 error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
343 agi = XFS_BUF_TO_AGI(bp);
344 be32_add_cpu(&agi->agi_length, new);
345 ASSERT(nagcount == oagcount ||
346 be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
347 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
351 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
356 agf = XFS_BUF_TO_AGF(bp);
357 be32_add_cpu(&agf->agf_length, new);
358 ASSERT(be32_to_cpu(agf->agf_length) ==
359 be32_to_cpu(agi->agi_length));
361 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
363 * Free the new space.
365 error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
366 be32_to_cpu(agf->agf_length) - new), new);
373 * Update changed superblock fields transactionally. These are not
374 * seen by the rest of the world until the transaction commit applies
375 * them atomically to the superblock.
377 if (nagcount > oagcount)
378 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
379 if (nb > mp->m_sb.sb_dblocks)
380 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
381 nb - mp->m_sb.sb_dblocks);
383 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
385 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
386 error = xfs_trans_commit(tp, 0);
390 /* New allocation groups fully initialized, so update mount struct */
392 mp->m_maxagi = nagimax;
393 if (mp->m_sb.sb_imax_pct) {
394 __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
396 mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
399 xfs_set_low_space_thresholds(mp);
401 /* update secondary superblocks. */
402 for (agno = 1; agno < nagcount; agno++) {
403 error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
404 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
405 XFS_FSS_TO_BB(mp, 1), 0, &bp);
408 "error %d reading secondary superblock for ag %d",
412 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
414 * If we get an error writing out the alternate superblocks,
415 * just issue a warning and continue. The real work is
416 * already done and committed.
418 error = xfs_bwrite(bp);
422 "write error %d updating secondary superblock for ag %d",
424 break; /* no point in continuing */
430 xfs_trans_cancel(tp, XFS_TRANS_ABORT);
435 xfs_growfs_log_private(
436 xfs_mount_t *mp, /* mount point for filesystem */
437 xfs_growfs_log_t *in) /* growfs log input struct */
442 if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
443 return XFS_ERROR(EINVAL);
444 if (nb == mp->m_sb.sb_logblocks &&
445 in->isint == (mp->m_sb.sb_logstart != 0))
446 return XFS_ERROR(EINVAL);
448 * Moving the log is hard, need new interfaces to sync
449 * the log first, hold off all activity while moving it.
450 * Can have shorter or longer log in the same space,
451 * or transform internal to external log or vice versa.
453 return XFS_ERROR(ENOSYS);
457 * protected versions of growfs function acquire and release locks on the mount
458 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
466 xfs_growfs_data_t *in)
470 if (!capable(CAP_SYS_ADMIN))
471 return XFS_ERROR(EPERM);
472 if (!mutex_trylock(&mp->m_growlock))
473 return XFS_ERROR(EWOULDBLOCK);
474 error = xfs_growfs_data_private(mp, in);
475 mutex_unlock(&mp->m_growlock);
482 xfs_growfs_log_t *in)
486 if (!capable(CAP_SYS_ADMIN))
487 return XFS_ERROR(EPERM);
488 if (!mutex_trylock(&mp->m_growlock))
489 return XFS_ERROR(EWOULDBLOCK);
490 error = xfs_growfs_log_private(mp, in);
491 mutex_unlock(&mp->m_growlock);
496 * exported through ioctl XFS_IOC_FSCOUNTS
502 xfs_fsop_counts_t *cnt)
504 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
505 spin_lock(&mp->m_sb_lock);
506 cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
507 cnt->freertx = mp->m_sb.sb_frextents;
508 cnt->freeino = mp->m_sb.sb_ifree;
509 cnt->allocino = mp->m_sb.sb_icount;
510 spin_unlock(&mp->m_sb_lock);
515 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
517 * xfs_reserve_blocks is called to set m_resblks
518 * in the in-core mount table. The number of unused reserved blocks
519 * is kept in m_resblks_avail.
521 * Reserve the requested number of blocks if available. Otherwise return
522 * as many as possible to satisfy the request. The actual number
523 * reserved are returned in outval
525 * A null inval pointer indicates that only the current reserved blocks
526 * available should be returned no settings are changed.
533 xfs_fsop_resblks_t *outval)
535 __int64_t lcounter, delta, fdblks_delta;
538 /* If inval is null, report current values and return */
539 if (inval == (__uint64_t *)NULL) {
542 outval->resblks = mp->m_resblks;
543 outval->resblks_avail = mp->m_resblks_avail;
550 * With per-cpu counters, this becomes an interesting
551 * problem. we needto work out if we are freeing or allocation
552 * blocks first, then we can do the modification as necessary.
554 * We do this under the m_sb_lock so that if we are near
555 * ENOSPC, we will hold out any changes while we work out
556 * what to do. This means that the amount of free space can
557 * change while we do this, so we need to retry if we end up
558 * trying to reserve more space than is available.
560 * We also use the xfs_mod_incore_sb() interface so that we
561 * don't have to care about whether per cpu counter are
562 * enabled, disabled or even compiled in....
565 spin_lock(&mp->m_sb_lock);
566 xfs_icsb_sync_counters_locked(mp, 0);
569 * If our previous reservation was larger than the current value,
570 * then move any unused blocks back to the free pool.
573 if (mp->m_resblks > request) {
574 lcounter = mp->m_resblks_avail - request;
575 if (lcounter > 0) { /* release unused blocks */
576 fdblks_delta = lcounter;
577 mp->m_resblks_avail -= lcounter;
579 mp->m_resblks = request;
583 free = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
585 goto out; /* ENOSPC and fdblks_delta = 0 */
587 delta = request - mp->m_resblks;
588 lcounter = free - delta;
590 /* We can't satisfy the request, just get what we can */
591 mp->m_resblks += free;
592 mp->m_resblks_avail += free;
593 fdblks_delta = -free;
595 fdblks_delta = -delta;
596 mp->m_resblks = request;
597 mp->m_resblks_avail += delta;
602 outval->resblks = mp->m_resblks;
603 outval->resblks_avail = mp->m_resblks_avail;
605 spin_unlock(&mp->m_sb_lock);
609 * If we are putting blocks back here, m_resblks_avail is
610 * already at its max so this will put it in the free pool.
612 * If we need space, we'll either succeed in getting it
613 * from the free block count or we'll get an enospc. If
614 * we get a ENOSPC, it means things changed while we were
615 * calculating fdblks_delta and so we should try again to
616 * see if there is anything left to reserve.
618 * Don't set the reserved flag here - we don't want to reserve
619 * the extra reserve blocks from the reserve.....
622 error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
631 * Dump a transaction into the log that contains no real change. This is needed
632 * to be able to make the log dirty or stamp the current tail LSN into the log
633 * during the covering operation.
635 * We cannot use an inode here for this - that will push dirty state back up
636 * into the VFS and then periodic inode flushing will prevent log covering from
637 * making progress. Hence we log a field in the superblock instead and use a
638 * synchronous transaction to ensure the superblock is immediately unpinned
639 * and can be written back.
648 tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP);
649 error = xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0,
650 XFS_DEFAULT_LOG_COUNT);
652 xfs_trans_cancel(tp, 0);
656 /* log the UUID because it is an unchanging field */
657 xfs_mod_sb(tp, XFS_SB_UUID);
658 xfs_trans_set_sync(tp);
659 return xfs_trans_commit(tp, 0);
668 case XFS_FSOP_GOING_FLAGS_DEFAULT: {
669 struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
671 if (sb && !IS_ERR(sb)) {
672 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
673 thaw_bdev(sb->s_bdev, sb);
678 case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
679 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
681 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
682 xfs_force_shutdown(mp,
683 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
686 return XFS_ERROR(EINVAL);