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, mp->m_ddev_targp,
151 XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
152 BBTOB(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_LOCK | XBF_MAPPED);
197 agf = XFS_BUF_TO_AGF(bp);
198 memset(agf, 0, mp->m_sb.sb_sectsize);
199 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
200 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
201 agf->agf_seqno = cpu_to_be32(agno);
202 if (agno == nagcount - 1)
205 (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
207 agsize = mp->m_sb.sb_agblocks;
208 agf->agf_length = cpu_to_be32(agsize);
209 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
210 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
211 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
212 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
213 agf->agf_flfirst = 0;
214 agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
215 agf->agf_flcount = 0;
216 tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
217 agf->agf_freeblks = cpu_to_be32(tmpsize);
218 agf->agf_longest = cpu_to_be32(tmpsize);
219 error = xfs_bwrite(bp);
225 * AG inode header block
227 bp = xfs_buf_get(mp->m_ddev_targp,
228 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
229 XFS_FSS_TO_BB(mp, 1), XBF_LOCK | XBF_MAPPED);
230 agi = XFS_BUF_TO_AGI(bp);
231 memset(agi, 0, mp->m_sb.sb_sectsize);
232 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
233 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
234 agi->agi_seqno = cpu_to_be32(agno);
235 agi->agi_length = cpu_to_be32(agsize);
237 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
238 agi->agi_level = cpu_to_be32(1);
239 agi->agi_freecount = 0;
240 agi->agi_newino = cpu_to_be32(NULLAGINO);
241 agi->agi_dirino = cpu_to_be32(NULLAGINO);
242 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
243 agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
244 error = xfs_bwrite(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(bp);
273 * CNT btree root block
275 bp = xfs_buf_get(mp->m_ddev_targp,
276 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
277 BTOBB(mp->m_sb.sb_blocksize),
278 XBF_LOCK | XBF_MAPPED);
279 block = XFS_BUF_TO_BLOCK(bp);
280 memset(block, 0, mp->m_sb.sb_blocksize);
281 block->bb_magic = cpu_to_be32(XFS_ABTC_MAGIC);
283 block->bb_numrecs = cpu_to_be16(1);
284 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
285 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
286 arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
287 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
288 arec->ar_blockcount = cpu_to_be32(
289 agsize - be32_to_cpu(arec->ar_startblock));
290 nfree += be32_to_cpu(arec->ar_blockcount);
291 error = xfs_bwrite(bp);
297 * INO btree root block
299 bp = xfs_buf_get(mp->m_ddev_targp,
300 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
301 BTOBB(mp->m_sb.sb_blocksize),
302 XBF_LOCK | XBF_MAPPED);
303 block = XFS_BUF_TO_BLOCK(bp);
304 memset(block, 0, mp->m_sb.sb_blocksize);
305 block->bb_magic = cpu_to_be32(XFS_IBT_MAGIC);
307 block->bb_numrecs = 0;
308 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
309 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
310 error = xfs_bwrite(bp);
315 xfs_trans_agblocks_delta(tp, nfree);
317 * There are new blocks in the old last a.g.
321 * Change the agi length.
323 error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
328 agi = XFS_BUF_TO_AGI(bp);
329 be32_add_cpu(&agi->agi_length, new);
330 ASSERT(nagcount == oagcount ||
331 be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
332 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
336 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
341 agf = XFS_BUF_TO_AGF(bp);
342 be32_add_cpu(&agf->agf_length, new);
343 ASSERT(be32_to_cpu(agf->agf_length) ==
344 be32_to_cpu(agi->agi_length));
346 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
348 * Free the new space.
350 error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
351 be32_to_cpu(agf->agf_length) - new), new);
358 * Update changed superblock fields transactionally. These are not
359 * seen by the rest of the world until the transaction commit applies
360 * them atomically to the superblock.
362 if (nagcount > oagcount)
363 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
364 if (nb > mp->m_sb.sb_dblocks)
365 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
366 nb - mp->m_sb.sb_dblocks);
368 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
370 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
371 error = xfs_trans_commit(tp, 0);
375 /* New allocation groups fully initialized, so update mount struct */
377 mp->m_maxagi = nagimax;
378 if (mp->m_sb.sb_imax_pct) {
379 __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
381 mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
384 xfs_set_low_space_thresholds(mp);
386 /* update secondary superblocks. */
387 for (agno = 1; agno < nagcount; agno++) {
388 error = xfs_read_buf(mp, mp->m_ddev_targp,
389 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
390 XFS_FSS_TO_BB(mp, 1), 0, &bp);
393 "error %d reading secondary superblock for ag %d",
397 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
399 * If we get an error writing out the alternate superblocks,
400 * just issue a warning and continue. The real work is
401 * already done and committed.
403 error = xfs_bwrite(bp);
407 "write error %d updating secondary superblock for ag %d",
409 break; /* no point in continuing */
415 xfs_trans_cancel(tp, XFS_TRANS_ABORT);
420 xfs_growfs_log_private(
421 xfs_mount_t *mp, /* mount point for filesystem */
422 xfs_growfs_log_t *in) /* growfs log input struct */
427 if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
428 return XFS_ERROR(EINVAL);
429 if (nb == mp->m_sb.sb_logblocks &&
430 in->isint == (mp->m_sb.sb_logstart != 0))
431 return XFS_ERROR(EINVAL);
433 * Moving the log is hard, need new interfaces to sync
434 * the log first, hold off all activity while moving it.
435 * Can have shorter or longer log in the same space,
436 * or transform internal to external log or vice versa.
438 return XFS_ERROR(ENOSYS);
442 * protected versions of growfs function acquire and release locks on the mount
443 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
451 xfs_growfs_data_t *in)
455 if (!capable(CAP_SYS_ADMIN))
456 return XFS_ERROR(EPERM);
457 if (!mutex_trylock(&mp->m_growlock))
458 return XFS_ERROR(EWOULDBLOCK);
459 error = xfs_growfs_data_private(mp, in);
460 mutex_unlock(&mp->m_growlock);
467 xfs_growfs_log_t *in)
471 if (!capable(CAP_SYS_ADMIN))
472 return XFS_ERROR(EPERM);
473 if (!mutex_trylock(&mp->m_growlock))
474 return XFS_ERROR(EWOULDBLOCK);
475 error = xfs_growfs_log_private(mp, in);
476 mutex_unlock(&mp->m_growlock);
481 * exported through ioctl XFS_IOC_FSCOUNTS
487 xfs_fsop_counts_t *cnt)
489 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
490 spin_lock(&mp->m_sb_lock);
491 cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
492 cnt->freertx = mp->m_sb.sb_frextents;
493 cnt->freeino = mp->m_sb.sb_ifree;
494 cnt->allocino = mp->m_sb.sb_icount;
495 spin_unlock(&mp->m_sb_lock);
500 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
502 * xfs_reserve_blocks is called to set m_resblks
503 * in the in-core mount table. The number of unused reserved blocks
504 * is kept in m_resblks_avail.
506 * Reserve the requested number of blocks if available. Otherwise return
507 * as many as possible to satisfy the request. The actual number
508 * reserved are returned in outval
510 * A null inval pointer indicates that only the current reserved blocks
511 * available should be returned no settings are changed.
518 xfs_fsop_resblks_t *outval)
520 __int64_t lcounter, delta, fdblks_delta;
523 /* If inval is null, report current values and return */
524 if (inval == (__uint64_t *)NULL) {
527 outval->resblks = mp->m_resblks;
528 outval->resblks_avail = mp->m_resblks_avail;
535 * With per-cpu counters, this becomes an interesting
536 * problem. we needto work out if we are freeing or allocation
537 * blocks first, then we can do the modification as necessary.
539 * We do this under the m_sb_lock so that if we are near
540 * ENOSPC, we will hold out any changes while we work out
541 * what to do. This means that the amount of free space can
542 * change while we do this, so we need to retry if we end up
543 * trying to reserve more space than is available.
545 * We also use the xfs_mod_incore_sb() interface so that we
546 * don't have to care about whether per cpu counter are
547 * enabled, disabled or even compiled in....
550 spin_lock(&mp->m_sb_lock);
551 xfs_icsb_sync_counters_locked(mp, 0);
554 * If our previous reservation was larger than the current value,
555 * then move any unused blocks back to the free pool.
558 if (mp->m_resblks > request) {
559 lcounter = mp->m_resblks_avail - request;
560 if (lcounter > 0) { /* release unused blocks */
561 fdblks_delta = lcounter;
562 mp->m_resblks_avail -= lcounter;
564 mp->m_resblks = request;
568 free = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
570 goto out; /* ENOSPC and fdblks_delta = 0 */
572 delta = request - mp->m_resblks;
573 lcounter = free - delta;
575 /* We can't satisfy the request, just get what we can */
576 mp->m_resblks += free;
577 mp->m_resblks_avail += free;
578 fdblks_delta = -free;
580 fdblks_delta = -delta;
581 mp->m_resblks = request;
582 mp->m_resblks_avail += delta;
587 outval->resblks = mp->m_resblks;
588 outval->resblks_avail = mp->m_resblks_avail;
590 spin_unlock(&mp->m_sb_lock);
594 * If we are putting blocks back here, m_resblks_avail is
595 * already at its max so this will put it in the free pool.
597 * If we need space, we'll either succeed in getting it
598 * from the free block count or we'll get an enospc. If
599 * we get a ENOSPC, it means things changed while we were
600 * calculating fdblks_delta and so we should try again to
601 * see if there is anything left to reserve.
603 * Don't set the reserved flag here - we don't want to reserve
604 * the extra reserve blocks from the reserve.....
607 error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
616 * Dump a transaction into the log that contains no real change. This is needed
617 * to be able to make the log dirty or stamp the current tail LSN into the log
618 * during the covering operation.
620 * We cannot use an inode here for this - that will push dirty state back up
621 * into the VFS and then periodic inode flushing will prevent log covering from
622 * making progress. Hence we log a field in the superblock instead and use a
623 * synchronous transaction to ensure the superblock is immediately unpinned
624 * and can be written back.
633 tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP);
634 error = xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0,
635 XFS_DEFAULT_LOG_COUNT);
637 xfs_trans_cancel(tp, 0);
641 /* log the UUID because it is an unchanging field */
642 xfs_mod_sb(tp, XFS_SB_UUID);
643 xfs_trans_set_sync(tp);
644 return xfs_trans_commit(tp, 0);
653 case XFS_FSOP_GOING_FLAGS_DEFAULT: {
654 struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
656 if (sb && !IS_ERR(sb)) {
657 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
658 thaw_bdev(sb->s_bdev, sb);
663 case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
664 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
666 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
667 xfs_force_shutdown(mp,
668 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
671 return XFS_ERROR(EINVAL);