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"
50 * File system operations
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 xfs_btree_sblock_t *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 error = xfs_read_buf(mp, 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, &bp);
158 new = nb; /* use new as a temporary here */
159 nb_mod = do_div(new, mp->m_sb.sb_agblocks);
160 nagcount = new + (nb_mod != 0);
161 if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
163 nb = nagcount * mp->m_sb.sb_agblocks;
164 if (nb < mp->m_sb.sb_dblocks)
165 return XFS_ERROR(EINVAL);
167 new = nb - mp->m_sb.sb_dblocks;
168 oagcount = mp->m_sb.sb_agcount;
169 if (nagcount > oagcount) {
170 xfs_filestream_flush(mp);
171 down_write(&mp->m_peraglock);
172 mp->m_perag = kmem_realloc(mp->m_perag,
173 sizeof(xfs_perag_t) * nagcount,
174 sizeof(xfs_perag_t) * oagcount,
176 memset(&mp->m_perag[oagcount], 0,
177 (nagcount - oagcount) * sizeof(xfs_perag_t));
178 mp->m_flags |= XFS_MOUNT_32BITINODES;
179 nagimax = xfs_initialize_perag(mp, nagcount);
180 up_write(&mp->m_peraglock);
182 tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
183 tp->t_flags |= XFS_TRANS_RESERVE;
184 if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp),
185 XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) {
186 xfs_trans_cancel(tp, 0);
191 for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
193 * AG freelist header block
195 bp = xfs_buf_get(mp->m_ddev_targp,
196 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
197 XFS_FSS_TO_BB(mp, 1), 0);
198 agf = XFS_BUF_TO_AGF(bp);
199 memset(agf, 0, mp->m_sb.sb_sectsize);
200 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
201 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
202 agf->agf_seqno = cpu_to_be32(agno);
203 if (agno == nagcount - 1)
206 (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
208 agsize = mp->m_sb.sb_agblocks;
209 agf->agf_length = cpu_to_be32(agsize);
210 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
211 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
212 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
213 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
214 agf->agf_flfirst = 0;
215 agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
216 agf->agf_flcount = 0;
217 tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
218 agf->agf_freeblks = cpu_to_be32(tmpsize);
219 agf->agf_longest = cpu_to_be32(tmpsize);
220 error = xfs_bwrite(mp, 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), 0);
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(mp, bp);
249 * BNO btree root block
251 bp = xfs_buf_get(mp->m_ddev_targp,
252 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
253 BTOBB(mp->m_sb.sb_blocksize), 0);
254 block = XFS_BUF_TO_SBLOCK(bp);
255 memset(block, 0, mp->m_sb.sb_blocksize);
256 block->bb_magic = cpu_to_be32(XFS_ABTB_MAGIC);
258 block->bb_numrecs = cpu_to_be16(1);
259 block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
260 block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
261 arec = XFS_BTREE_REC_ADDR(xfs_alloc, block, 1);
262 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
263 arec->ar_blockcount = cpu_to_be32(
264 agsize - be32_to_cpu(arec->ar_startblock));
265 error = xfs_bwrite(mp, bp);
270 * CNT btree root block
272 bp = xfs_buf_get(mp->m_ddev_targp,
273 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
274 BTOBB(mp->m_sb.sb_blocksize), 0);
275 block = XFS_BUF_TO_SBLOCK(bp);
276 memset(block, 0, mp->m_sb.sb_blocksize);
277 block->bb_magic = cpu_to_be32(XFS_ABTC_MAGIC);
279 block->bb_numrecs = cpu_to_be16(1);
280 block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
281 block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
282 arec = XFS_BTREE_REC_ADDR(xfs_alloc, block, 1);
283 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
284 arec->ar_blockcount = cpu_to_be32(
285 agsize - be32_to_cpu(arec->ar_startblock));
286 nfree += be32_to_cpu(arec->ar_blockcount);
287 error = xfs_bwrite(mp, bp);
292 * INO btree root block
294 bp = xfs_buf_get(mp->m_ddev_targp,
295 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
296 BTOBB(mp->m_sb.sb_blocksize), 0);
297 block = XFS_BUF_TO_SBLOCK(bp);
298 memset(block, 0, mp->m_sb.sb_blocksize);
299 block->bb_magic = cpu_to_be32(XFS_IBT_MAGIC);
301 block->bb_numrecs = 0;
302 block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
303 block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
304 error = xfs_bwrite(mp, bp);
309 xfs_trans_agblocks_delta(tp, nfree);
311 * There are new blocks in the old last a.g.
315 * Change the agi length.
317 error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
322 agi = XFS_BUF_TO_AGI(bp);
323 be32_add_cpu(&agi->agi_length, new);
324 ASSERT(nagcount == oagcount ||
325 be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
326 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
330 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
335 agf = XFS_BUF_TO_AGF(bp);
336 be32_add_cpu(&agf->agf_length, new);
337 ASSERT(be32_to_cpu(agf->agf_length) ==
338 be32_to_cpu(agi->agi_length));
339 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
341 * Free the new space.
343 error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
344 be32_to_cpu(agf->agf_length) - new), new);
349 if (nagcount > oagcount)
350 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
351 if (nb > mp->m_sb.sb_dblocks)
352 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
353 nb - mp->m_sb.sb_dblocks);
355 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
357 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
358 error = xfs_trans_commit(tp, 0);
362 /* New allocation groups fully initialized, so update mount struct */
364 mp->m_maxagi = nagimax;
365 if (mp->m_sb.sb_imax_pct) {
366 __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
368 mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
371 for (agno = 1; agno < nagcount; agno++) {
372 error = xfs_read_buf(mp, mp->m_ddev_targp,
373 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
374 XFS_FSS_TO_BB(mp, 1), 0, &bp);
376 xfs_fs_cmn_err(CE_WARN, mp,
377 "error %d reading secondary superblock for ag %d",
381 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
383 * If we get an error writing out the alternate superblocks,
384 * just issue a warning and continue. The real work is
385 * already done and committed.
387 if (!(error = xfs_bwrite(mp, bp))) {
390 xfs_fs_cmn_err(CE_WARN, mp,
391 "write error %d updating secondary superblock for ag %d",
393 break; /* no point in continuing */
399 xfs_trans_cancel(tp, XFS_TRANS_ABORT);
404 xfs_growfs_log_private(
405 xfs_mount_t *mp, /* mount point for filesystem */
406 xfs_growfs_log_t *in) /* growfs log input struct */
411 if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
412 return XFS_ERROR(EINVAL);
413 if (nb == mp->m_sb.sb_logblocks &&
414 in->isint == (mp->m_sb.sb_logstart != 0))
415 return XFS_ERROR(EINVAL);
417 * Moving the log is hard, need new interfaces to sync
418 * the log first, hold off all activity while moving it.
419 * Can have shorter or longer log in the same space,
420 * or transform internal to external log or vice versa.
422 return XFS_ERROR(ENOSYS);
426 * protected versions of growfs function acquire and release locks on the mount
427 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
435 xfs_growfs_data_t *in)
438 if (!mutex_trylock(&mp->m_growlock))
439 return XFS_ERROR(EWOULDBLOCK);
440 error = xfs_growfs_data_private(mp, in);
441 mutex_unlock(&mp->m_growlock);
448 xfs_growfs_log_t *in)
451 if (!mutex_trylock(&mp->m_growlock))
452 return XFS_ERROR(EWOULDBLOCK);
453 error = xfs_growfs_log_private(mp, in);
454 mutex_unlock(&mp->m_growlock);
459 * exported through ioctl XFS_IOC_FSCOUNTS
465 xfs_fsop_counts_t *cnt)
467 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
468 spin_lock(&mp->m_sb_lock);
469 cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
470 cnt->freertx = mp->m_sb.sb_frextents;
471 cnt->freeino = mp->m_sb.sb_ifree;
472 cnt->allocino = mp->m_sb.sb_icount;
473 spin_unlock(&mp->m_sb_lock);
478 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
480 * xfs_reserve_blocks is called to set m_resblks
481 * in the in-core mount table. The number of unused reserved blocks
482 * is kept in m_resblks_avail.
484 * Reserve the requested number of blocks if available. Otherwise return
485 * as many as possible to satisfy the request. The actual number
486 * reserved are returned in outval
488 * A null inval pointer indicates that only the current reserved blocks
489 * available should be returned no settings are changed.
496 xfs_fsop_resblks_t *outval)
498 __int64_t lcounter, delta, fdblks_delta;
501 /* If inval is null, report current values and return */
502 if (inval == (__uint64_t *)NULL) {
505 outval->resblks = mp->m_resblks;
506 outval->resblks_avail = mp->m_resblks_avail;
513 * With per-cpu counters, this becomes an interesting
514 * problem. we needto work out if we are freeing or allocation
515 * blocks first, then we can do the modification as necessary.
517 * We do this under the m_sb_lock so that if we are near
518 * ENOSPC, we will hold out any changes while we work out
519 * what to do. This means that the amount of free space can
520 * change while we do this, so we need to retry if we end up
521 * trying to reserve more space than is available.
523 * We also use the xfs_mod_incore_sb() interface so that we
524 * don't have to care about whether per cpu counter are
525 * enabled, disabled or even compiled in....
528 spin_lock(&mp->m_sb_lock);
529 xfs_icsb_sync_counters_locked(mp, 0);
532 * If our previous reservation was larger than the current value,
533 * then move any unused blocks back to the free pool.
536 if (mp->m_resblks > request) {
537 lcounter = mp->m_resblks_avail - request;
538 if (lcounter > 0) { /* release unused blocks */
539 fdblks_delta = lcounter;
540 mp->m_resblks_avail -= lcounter;
542 mp->m_resblks = request;
546 free = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
548 goto out; /* ENOSPC and fdblks_delta = 0 */
550 delta = request - mp->m_resblks;
551 lcounter = free - delta;
553 /* We can't satisfy the request, just get what we can */
554 mp->m_resblks += free;
555 mp->m_resblks_avail += free;
556 fdblks_delta = -free;
558 fdblks_delta = -delta;
559 mp->m_resblks = request;
560 mp->m_resblks_avail += delta;
565 outval->resblks = mp->m_resblks;
566 outval->resblks_avail = mp->m_resblks_avail;
568 spin_unlock(&mp->m_sb_lock);
572 * If we are putting blocks back here, m_resblks_avail is
573 * already at it's max so this will put it in the free pool.
575 * If we need space, we'll either succeed in getting it
576 * from the free block count or we'll get an enospc. If
577 * we get a ENOSPC, it means things changed while we were
578 * calculating fdblks_delta and so we should try again to
579 * see if there is anything left to reserve.
581 * Don't set the reserved flag here - we don't want to reserve
582 * the extra reserve blocks from the reserve.....
585 error = xfs_mod_incore_sb(mp, XFS_SBS_FDBLOCKS, fdblks_delta, 0);
599 tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1);
600 if (xfs_trans_reserve(tp, 0, XFS_ICHANGE_LOG_RES(mp), 0, 0, 0)) {
601 xfs_trans_cancel(tp, 0);
606 xfs_ilock(ip, XFS_ILOCK_EXCL);
608 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
609 xfs_trans_ihold(tp, ip);
610 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
611 xfs_trans_set_sync(tp);
612 xfs_trans_commit(tp, 0);
614 xfs_iunlock(ip, XFS_ILOCK_EXCL);
623 case XFS_FSOP_GOING_FLAGS_DEFAULT: {
624 struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
626 if (sb && !IS_ERR(sb)) {
627 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
628 thaw_bdev(sb->s_bdev, sb);
633 case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
634 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
636 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
637 xfs_force_shutdown(mp,
638 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
641 return XFS_ERROR(EINVAL);