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"
27 #include "xfs_mount.h"
28 #include "xfs_bmap_btree.h"
29 #include "xfs_alloc_btree.h"
30 #include "xfs_ialloc_btree.h"
31 #include "xfs_dinode.h"
32 #include "xfs_inode.h"
33 #include "xfs_inode_item.h"
34 #include "xfs_btree.h"
35 #include "xfs_error.h"
36 #include "xfs_alloc.h"
37 #include "xfs_ialloc.h"
38 #include "xfs_fsops.h"
39 #include "xfs_itable.h"
40 #include "xfs_trans_space.h"
41 #include "xfs_rtalloc.h"
43 #include "xfs_filestream.h"
44 #include "xfs_trace.h"
47 * File system operations
57 memset(geo, 0, sizeof(*geo));
59 geo->blocksize = mp->m_sb.sb_blocksize;
60 geo->rtextsize = mp->m_sb.sb_rextsize;
61 geo->agblocks = mp->m_sb.sb_agblocks;
62 geo->agcount = mp->m_sb.sb_agcount;
63 geo->logblocks = mp->m_sb.sb_logblocks;
64 geo->sectsize = mp->m_sb.sb_sectsize;
65 geo->inodesize = mp->m_sb.sb_inodesize;
66 geo->imaxpct = mp->m_sb.sb_imax_pct;
67 geo->datablocks = mp->m_sb.sb_dblocks;
68 geo->rtblocks = mp->m_sb.sb_rblocks;
69 geo->rtextents = mp->m_sb.sb_rextents;
70 geo->logstart = mp->m_sb.sb_logstart;
71 ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid));
72 memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid));
73 if (new_version >= 2) {
74 geo->sunit = mp->m_sb.sb_unit;
75 geo->swidth = mp->m_sb.sb_width;
77 if (new_version >= 3) {
78 geo->version = XFS_FSOP_GEOM_VERSION;
80 (xfs_sb_version_hasattr(&mp->m_sb) ?
81 XFS_FSOP_GEOM_FLAGS_ATTR : 0) |
82 (xfs_sb_version_hasnlink(&mp->m_sb) ?
83 XFS_FSOP_GEOM_FLAGS_NLINK : 0) |
84 (xfs_sb_version_hasquota(&mp->m_sb) ?
85 XFS_FSOP_GEOM_FLAGS_QUOTA : 0) |
86 (xfs_sb_version_hasalign(&mp->m_sb) ?
87 XFS_FSOP_GEOM_FLAGS_IALIGN : 0) |
88 (xfs_sb_version_hasdalign(&mp->m_sb) ?
89 XFS_FSOP_GEOM_FLAGS_DALIGN : 0) |
90 (xfs_sb_version_hasshared(&mp->m_sb) ?
91 XFS_FSOP_GEOM_FLAGS_SHARED : 0) |
92 (xfs_sb_version_hasextflgbit(&mp->m_sb) ?
93 XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) |
94 (xfs_sb_version_hasdirv2(&mp->m_sb) ?
95 XFS_FSOP_GEOM_FLAGS_DIRV2 : 0) |
96 (xfs_sb_version_hassector(&mp->m_sb) ?
97 XFS_FSOP_GEOM_FLAGS_SECTOR : 0) |
98 (xfs_sb_version_hasasciici(&mp->m_sb) ?
99 XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) |
100 (xfs_sb_version_haslazysbcount(&mp->m_sb) ?
101 XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) |
102 (xfs_sb_version_hasattr2(&mp->m_sb) ?
103 XFS_FSOP_GEOM_FLAGS_ATTR2 : 0);
104 geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
105 mp->m_sb.sb_logsectsize : BBSIZE;
106 geo->rtsectsize = mp->m_sb.sb_blocksize;
107 geo->dirblocksize = mp->m_dirblksize;
109 if (new_version >= 4) {
111 (xfs_sb_version_haslogv2(&mp->m_sb) ?
112 XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
113 geo->logsunit = mp->m_sb.sb_logsunit;
119 xfs_growfs_data_private(
120 xfs_mount_t *mp, /* mount point for filesystem */
121 xfs_growfs_data_t *in) /* growfs data input struct */
127 xfs_extlen_t tmpsize;
128 xfs_alloc_rec_t *arec;
129 struct xfs_btree_block *block;
134 xfs_agnumber_t nagcount;
135 xfs_agnumber_t nagimax = 0;
136 xfs_rfsblock_t nb, nb_mod;
138 xfs_rfsblock_t nfree;
139 xfs_agnumber_t oagcount;
145 if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
146 return XFS_ERROR(EINVAL);
147 if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
149 dpct = pct - mp->m_sb.sb_imax_pct;
150 bp = xfs_buf_read_uncached(mp->m_ddev_targp,
151 XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
152 XFS_FSS_TO_BB(mp, 1), 0);
157 new = nb; /* use new as a temporary here */
158 nb_mod = do_div(new, mp->m_sb.sb_agblocks);
159 nagcount = new + (nb_mod != 0);
160 if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
162 nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
163 if (nb < mp->m_sb.sb_dblocks)
164 return XFS_ERROR(EINVAL);
166 new = nb - mp->m_sb.sb_dblocks;
167 oagcount = mp->m_sb.sb_agcount;
169 /* allocate the new per-ag structures */
170 if (nagcount > oagcount) {
171 error = xfs_initialize_perag(mp, nagcount, &nagimax);
176 tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
177 tp->t_flags |= XFS_TRANS_RESERVE;
178 if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp),
179 XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) {
180 xfs_trans_cancel(tp, 0);
185 * Write new AG headers to disk. Non-transactional, but written
186 * synchronously so they are completed prior to the growfs transaction
190 for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
192 * AG freelist header block
194 bp = xfs_buf_get(mp->m_ddev_targp,
195 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
196 XFS_FSS_TO_BB(mp, 1), XBF_MAPPED);
201 agf = XFS_BUF_TO_AGF(bp);
202 memset(agf, 0, mp->m_sb.sb_sectsize);
203 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
204 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
205 agf->agf_seqno = cpu_to_be32(agno);
206 if (agno == nagcount - 1)
209 (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
211 agsize = mp->m_sb.sb_agblocks;
212 agf->agf_length = cpu_to_be32(agsize);
213 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
214 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
215 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
216 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
217 agf->agf_flfirst = 0;
218 agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
219 agf->agf_flcount = 0;
220 tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
221 agf->agf_freeblks = cpu_to_be32(tmpsize);
222 agf->agf_longest = cpu_to_be32(tmpsize);
223 error = xfs_bwrite(bp);
229 * AG inode header block
231 bp = xfs_buf_get(mp->m_ddev_targp,
232 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
233 XFS_FSS_TO_BB(mp, 1), XBF_MAPPED);
238 agi = XFS_BUF_TO_AGI(bp);
239 memset(agi, 0, mp->m_sb.sb_sectsize);
240 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
241 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
242 agi->agi_seqno = cpu_to_be32(agno);
243 agi->agi_length = cpu_to_be32(agsize);
245 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
246 agi->agi_level = cpu_to_be32(1);
247 agi->agi_freecount = 0;
248 agi->agi_newino = cpu_to_be32(NULLAGINO);
249 agi->agi_dirino = cpu_to_be32(NULLAGINO);
250 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
251 agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
252 error = xfs_bwrite(bp);
258 * BNO btree root block
260 bp = xfs_buf_get(mp->m_ddev_targp,
261 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
262 BTOBB(mp->m_sb.sb_blocksize), XBF_MAPPED);
267 block = XFS_BUF_TO_BLOCK(bp);
268 memset(block, 0, mp->m_sb.sb_blocksize);
269 block->bb_magic = cpu_to_be32(XFS_ABTB_MAGIC);
271 block->bb_numrecs = cpu_to_be16(1);
272 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
273 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
274 arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
275 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
276 arec->ar_blockcount = cpu_to_be32(
277 agsize - be32_to_cpu(arec->ar_startblock));
278 error = xfs_bwrite(bp);
284 * CNT btree root block
286 bp = xfs_buf_get(mp->m_ddev_targp,
287 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
288 BTOBB(mp->m_sb.sb_blocksize), XBF_MAPPED);
293 block = XFS_BUF_TO_BLOCK(bp);
294 memset(block, 0, mp->m_sb.sb_blocksize);
295 block->bb_magic = cpu_to_be32(XFS_ABTC_MAGIC);
297 block->bb_numrecs = cpu_to_be16(1);
298 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
299 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
300 arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
301 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
302 arec->ar_blockcount = cpu_to_be32(
303 agsize - be32_to_cpu(arec->ar_startblock));
304 nfree += be32_to_cpu(arec->ar_blockcount);
305 error = xfs_bwrite(bp);
311 * INO btree root block
313 bp = xfs_buf_get(mp->m_ddev_targp,
314 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
315 BTOBB(mp->m_sb.sb_blocksize), XBF_MAPPED);
320 block = XFS_BUF_TO_BLOCK(bp);
321 memset(block, 0, mp->m_sb.sb_blocksize);
322 block->bb_magic = cpu_to_be32(XFS_IBT_MAGIC);
324 block->bb_numrecs = 0;
325 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
326 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
327 error = xfs_bwrite(bp);
332 xfs_trans_agblocks_delta(tp, nfree);
334 * There are new blocks in the old last a.g.
338 * Change the agi length.
340 error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
345 agi = XFS_BUF_TO_AGI(bp);
346 be32_add_cpu(&agi->agi_length, new);
347 ASSERT(nagcount == oagcount ||
348 be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
349 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
353 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
358 agf = XFS_BUF_TO_AGF(bp);
359 be32_add_cpu(&agf->agf_length, new);
360 ASSERT(be32_to_cpu(agf->agf_length) ==
361 be32_to_cpu(agi->agi_length));
363 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
365 * Free the new space.
367 error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
368 be32_to_cpu(agf->agf_length) - new), new);
375 * Update changed superblock fields transactionally. These are not
376 * seen by the rest of the world until the transaction commit applies
377 * them atomically to the superblock.
379 if (nagcount > oagcount)
380 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
381 if (nb > mp->m_sb.sb_dblocks)
382 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
383 nb - mp->m_sb.sb_dblocks);
385 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
387 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
388 error = xfs_trans_commit(tp, 0);
392 /* New allocation groups fully initialized, so update mount struct */
394 mp->m_maxagi = nagimax;
395 if (mp->m_sb.sb_imax_pct) {
396 __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
398 mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
401 xfs_set_low_space_thresholds(mp);
403 /* update secondary superblocks. */
404 for (agno = 1; agno < nagcount; agno++) {
405 error = xfs_read_buf(mp, mp->m_ddev_targp,
406 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
407 XFS_FSS_TO_BB(mp, 1), 0, &bp);
410 "error %d reading secondary superblock for ag %d",
414 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
416 * If we get an error writing out the alternate superblocks,
417 * just issue a warning and continue. The real work is
418 * already done and committed.
420 error = xfs_bwrite(bp);
424 "write error %d updating secondary superblock for ag %d",
426 break; /* no point in continuing */
432 xfs_trans_cancel(tp, XFS_TRANS_ABORT);
437 xfs_growfs_log_private(
438 xfs_mount_t *mp, /* mount point for filesystem */
439 xfs_growfs_log_t *in) /* growfs log input struct */
444 if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
445 return XFS_ERROR(EINVAL);
446 if (nb == mp->m_sb.sb_logblocks &&
447 in->isint == (mp->m_sb.sb_logstart != 0))
448 return XFS_ERROR(EINVAL);
450 * Moving the log is hard, need new interfaces to sync
451 * the log first, hold off all activity while moving it.
452 * Can have shorter or longer log in the same space,
453 * or transform internal to external log or vice versa.
455 return XFS_ERROR(ENOSYS);
459 * protected versions of growfs function acquire and release locks on the mount
460 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
468 xfs_growfs_data_t *in)
472 if (!capable(CAP_SYS_ADMIN))
473 return XFS_ERROR(EPERM);
474 if (!mutex_trylock(&mp->m_growlock))
475 return XFS_ERROR(EWOULDBLOCK);
476 error = xfs_growfs_data_private(mp, in);
477 mutex_unlock(&mp->m_growlock);
484 xfs_growfs_log_t *in)
488 if (!capable(CAP_SYS_ADMIN))
489 return XFS_ERROR(EPERM);
490 if (!mutex_trylock(&mp->m_growlock))
491 return XFS_ERROR(EWOULDBLOCK);
492 error = xfs_growfs_log_private(mp, in);
493 mutex_unlock(&mp->m_growlock);
498 * exported through ioctl XFS_IOC_FSCOUNTS
504 xfs_fsop_counts_t *cnt)
506 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
507 spin_lock(&mp->m_sb_lock);
508 cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
509 cnt->freertx = mp->m_sb.sb_frextents;
510 cnt->freeino = mp->m_sb.sb_ifree;
511 cnt->allocino = mp->m_sb.sb_icount;
512 spin_unlock(&mp->m_sb_lock);
517 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
519 * xfs_reserve_blocks is called to set m_resblks
520 * in the in-core mount table. The number of unused reserved blocks
521 * is kept in m_resblks_avail.
523 * Reserve the requested number of blocks if available. Otherwise return
524 * as many as possible to satisfy the request. The actual number
525 * reserved are returned in outval
527 * A null inval pointer indicates that only the current reserved blocks
528 * available should be returned no settings are changed.
535 xfs_fsop_resblks_t *outval)
537 __int64_t lcounter, delta, fdblks_delta;
540 /* If inval is null, report current values and return */
541 if (inval == (__uint64_t *)NULL) {
544 outval->resblks = mp->m_resblks;
545 outval->resblks_avail = mp->m_resblks_avail;
552 * With per-cpu counters, this becomes an interesting
553 * problem. we needto work out if we are freeing or allocation
554 * blocks first, then we can do the modification as necessary.
556 * We do this under the m_sb_lock so that if we are near
557 * ENOSPC, we will hold out any changes while we work out
558 * what to do. This means that the amount of free space can
559 * change while we do this, so we need to retry if we end up
560 * trying to reserve more space than is available.
562 * We also use the xfs_mod_incore_sb() interface so that we
563 * don't have to care about whether per cpu counter are
564 * enabled, disabled or even compiled in....
567 spin_lock(&mp->m_sb_lock);
568 xfs_icsb_sync_counters_locked(mp, 0);
571 * If our previous reservation was larger than the current value,
572 * then move any unused blocks back to the free pool.
575 if (mp->m_resblks > request) {
576 lcounter = mp->m_resblks_avail - request;
577 if (lcounter > 0) { /* release unused blocks */
578 fdblks_delta = lcounter;
579 mp->m_resblks_avail -= lcounter;
581 mp->m_resblks = request;
585 free = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
587 goto out; /* ENOSPC and fdblks_delta = 0 */
589 delta = request - mp->m_resblks;
590 lcounter = free - delta;
592 /* We can't satisfy the request, just get what we can */
593 mp->m_resblks += free;
594 mp->m_resblks_avail += free;
595 fdblks_delta = -free;
597 fdblks_delta = -delta;
598 mp->m_resblks = request;
599 mp->m_resblks_avail += delta;
604 outval->resblks = mp->m_resblks;
605 outval->resblks_avail = mp->m_resblks_avail;
607 spin_unlock(&mp->m_sb_lock);
611 * If we are putting blocks back here, m_resblks_avail is
612 * already at its max so this will put it in the free pool.
614 * If we need space, we'll either succeed in getting it
615 * from the free block count or we'll get an enospc. If
616 * we get a ENOSPC, it means things changed while we were
617 * calculating fdblks_delta and so we should try again to
618 * see if there is anything left to reserve.
620 * Don't set the reserved flag here - we don't want to reserve
621 * the extra reserve blocks from the reserve.....
624 error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
633 * Dump a transaction into the log that contains no real change. This is needed
634 * to be able to make the log dirty or stamp the current tail LSN into the log
635 * during the covering operation.
637 * We cannot use an inode here for this - that will push dirty state back up
638 * into the VFS and then periodic inode flushing will prevent log covering from
639 * making progress. Hence we log a field in the superblock instead and use a
640 * synchronous transaction to ensure the superblock is immediately unpinned
641 * and can be written back.
650 tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP);
651 error = xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0,
652 XFS_DEFAULT_LOG_COUNT);
654 xfs_trans_cancel(tp, 0);
658 /* log the UUID because it is an unchanging field */
659 xfs_mod_sb(tp, XFS_SB_UUID);
660 xfs_trans_set_sync(tp);
661 return xfs_trans_commit(tp, 0);
670 case XFS_FSOP_GOING_FLAGS_DEFAULT: {
671 struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
673 if (sb && !IS_ERR(sb)) {
674 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
675 thaw_bdev(sb->s_bdev, sb);
680 case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
681 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
683 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
684 xfs_force_shutdown(mp,
685 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
688 return XFS_ERROR(EINVAL);