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_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 xfs_btree_sblock_t *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 error = xfs_read_buf(mp, 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, &bp);
156 new = nb; /* use new as a temporary here */
157 nb_mod = do_div(new, mp->m_sb.sb_agblocks);
158 nagcount = new + (nb_mod != 0);
159 if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
161 nb = nagcount * mp->m_sb.sb_agblocks;
162 if (nb < mp->m_sb.sb_dblocks)
163 return XFS_ERROR(EINVAL);
165 new = nb - mp->m_sb.sb_dblocks;
166 oagcount = mp->m_sb.sb_agcount;
167 if (nagcount > oagcount) {
168 xfs_filestream_flush(mp);
169 down_write(&mp->m_peraglock);
170 mp->m_perag = kmem_realloc(mp->m_perag,
171 sizeof(xfs_perag_t) * nagcount,
172 sizeof(xfs_perag_t) * oagcount,
174 memset(&mp->m_perag[oagcount], 0,
175 (nagcount - oagcount) * sizeof(xfs_perag_t));
176 mp->m_flags |= XFS_MOUNT_32BITINODES;
177 nagimax = xfs_initialize_perag(mp, nagcount);
178 up_write(&mp->m_peraglock);
180 tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
181 tp->t_flags |= XFS_TRANS_RESERVE;
182 if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp),
183 XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) {
184 xfs_trans_cancel(tp, 0);
189 for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
191 * AG freelist header block
193 bp = xfs_buf_get(mp->m_ddev_targp,
194 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
195 XFS_FSS_TO_BB(mp, 1), 0);
196 agf = XFS_BUF_TO_AGF(bp);
197 memset(agf, 0, mp->m_sb.sb_sectsize);
198 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
199 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
200 agf->agf_seqno = cpu_to_be32(agno);
201 if (agno == nagcount - 1)
204 (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
206 agsize = mp->m_sb.sb_agblocks;
207 agf->agf_length = cpu_to_be32(agsize);
208 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
209 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
210 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
211 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
212 agf->agf_flfirst = 0;
213 agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
214 agf->agf_flcount = 0;
215 tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
216 agf->agf_freeblks = cpu_to_be32(tmpsize);
217 agf->agf_longest = cpu_to_be32(tmpsize);
218 error = xfs_bwrite(mp, bp);
223 * AG inode header block
225 bp = xfs_buf_get(mp->m_ddev_targp,
226 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
227 XFS_FSS_TO_BB(mp, 1), 0);
228 agi = XFS_BUF_TO_AGI(bp);
229 memset(agi, 0, mp->m_sb.sb_sectsize);
230 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
231 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
232 agi->agi_seqno = cpu_to_be32(agno);
233 agi->agi_length = cpu_to_be32(agsize);
235 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
236 agi->agi_level = cpu_to_be32(1);
237 agi->agi_freecount = 0;
238 agi->agi_newino = cpu_to_be32(NULLAGINO);
239 agi->agi_dirino = cpu_to_be32(NULLAGINO);
240 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
241 agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
242 error = xfs_bwrite(mp, bp);
247 * BNO btree root block
249 bp = xfs_buf_get(mp->m_ddev_targp,
250 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
251 BTOBB(mp->m_sb.sb_blocksize), 0);
252 block = XFS_BUF_TO_SBLOCK(bp);
253 memset(block, 0, mp->m_sb.sb_blocksize);
254 block->bb_magic = cpu_to_be32(XFS_ABTB_MAGIC);
256 block->bb_numrecs = cpu_to_be16(1);
257 block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
258 block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
259 arec = XFS_BTREE_REC_ADDR(xfs_alloc, block, 1);
260 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
261 arec->ar_blockcount = cpu_to_be32(
262 agsize - be32_to_cpu(arec->ar_startblock));
263 error = xfs_bwrite(mp, bp);
268 * CNT btree root block
270 bp = xfs_buf_get(mp->m_ddev_targp,
271 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
272 BTOBB(mp->m_sb.sb_blocksize), 0);
273 block = XFS_BUF_TO_SBLOCK(bp);
274 memset(block, 0, mp->m_sb.sb_blocksize);
275 block->bb_magic = cpu_to_be32(XFS_ABTC_MAGIC);
277 block->bb_numrecs = cpu_to_be16(1);
278 block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
279 block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
280 arec = XFS_BTREE_REC_ADDR(xfs_alloc, block, 1);
281 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
282 arec->ar_blockcount = cpu_to_be32(
283 agsize - be32_to_cpu(arec->ar_startblock));
284 nfree += be32_to_cpu(arec->ar_blockcount);
285 error = xfs_bwrite(mp, bp);
290 * INO btree root block
292 bp = xfs_buf_get(mp->m_ddev_targp,
293 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
294 BTOBB(mp->m_sb.sb_blocksize), 0);
295 block = XFS_BUF_TO_SBLOCK(bp);
296 memset(block, 0, mp->m_sb.sb_blocksize);
297 block->bb_magic = cpu_to_be32(XFS_IBT_MAGIC);
299 block->bb_numrecs = 0;
300 block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
301 block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
302 error = xfs_bwrite(mp, bp);
307 xfs_trans_agblocks_delta(tp, nfree);
309 * There are new blocks in the old last a.g.
313 * Change the agi length.
315 error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
320 agi = XFS_BUF_TO_AGI(bp);
321 be32_add_cpu(&agi->agi_length, new);
322 ASSERT(nagcount == oagcount ||
323 be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
324 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
328 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
333 agf = XFS_BUF_TO_AGF(bp);
334 be32_add_cpu(&agf->agf_length, new);
335 ASSERT(be32_to_cpu(agf->agf_length) ==
336 be32_to_cpu(agi->agi_length));
337 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
339 * Free the new space.
341 error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
342 be32_to_cpu(agf->agf_length) - new), new);
347 if (nagcount > oagcount)
348 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
349 if (nb > mp->m_sb.sb_dblocks)
350 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
351 nb - mp->m_sb.sb_dblocks);
353 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
355 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
356 error = xfs_trans_commit(tp, 0);
360 /* New allocation groups fully initialized, so update mount struct */
362 mp->m_maxagi = nagimax;
363 if (mp->m_sb.sb_imax_pct) {
364 __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
366 mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
369 for (agno = 1; agno < nagcount; agno++) {
370 error = xfs_read_buf(mp, mp->m_ddev_targp,
371 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
372 XFS_FSS_TO_BB(mp, 1), 0, &bp);
374 xfs_fs_cmn_err(CE_WARN, mp,
375 "error %d reading secondary superblock for ag %d",
379 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
381 * If we get an error writing out the alternate superblocks,
382 * just issue a warning and continue. The real work is
383 * already done and committed.
385 if (!(error = xfs_bwrite(mp, bp))) {
388 xfs_fs_cmn_err(CE_WARN, mp,
389 "write error %d updating secondary superblock for ag %d",
391 break; /* no point in continuing */
397 xfs_trans_cancel(tp, XFS_TRANS_ABORT);
402 xfs_growfs_log_private(
403 xfs_mount_t *mp, /* mount point for filesystem */
404 xfs_growfs_log_t *in) /* growfs log input struct */
409 if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
410 return XFS_ERROR(EINVAL);
411 if (nb == mp->m_sb.sb_logblocks &&
412 in->isint == (mp->m_sb.sb_logstart != 0))
413 return XFS_ERROR(EINVAL);
415 * Moving the log is hard, need new interfaces to sync
416 * the log first, hold off all activity while moving it.
417 * Can have shorter or longer log in the same space,
418 * or transform internal to external log or vice versa.
420 return XFS_ERROR(ENOSYS);
424 * protected versions of growfs function acquire and release locks on the mount
425 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
433 xfs_growfs_data_t *in)
436 if (!mutex_trylock(&mp->m_growlock))
437 return XFS_ERROR(EWOULDBLOCK);
438 error = xfs_growfs_data_private(mp, in);
439 mutex_unlock(&mp->m_growlock);
446 xfs_growfs_log_t *in)
449 if (!mutex_trylock(&mp->m_growlock))
450 return XFS_ERROR(EWOULDBLOCK);
451 error = xfs_growfs_log_private(mp, in);
452 mutex_unlock(&mp->m_growlock);
457 * exported through ioctl XFS_IOC_FSCOUNTS
463 xfs_fsop_counts_t *cnt)
465 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
466 spin_lock(&mp->m_sb_lock);
467 cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
468 cnt->freertx = mp->m_sb.sb_frextents;
469 cnt->freeino = mp->m_sb.sb_ifree;
470 cnt->allocino = mp->m_sb.sb_icount;
471 spin_unlock(&mp->m_sb_lock);
476 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
478 * xfs_reserve_blocks is called to set m_resblks
479 * in the in-core mount table. The number of unused reserved blocks
480 * is kept in m_resblks_avail.
482 * Reserve the requested number of blocks if available. Otherwise return
483 * as many as possible to satisfy the request. The actual number
484 * reserved are returned in outval
486 * A null inval pointer indicates that only the current reserved blocks
487 * available should be returned no settings are changed.
494 xfs_fsop_resblks_t *outval)
496 __int64_t lcounter, delta, fdblks_delta;
499 /* If inval is null, report current values and return */
500 if (inval == (__uint64_t *)NULL) {
503 outval->resblks = mp->m_resblks;
504 outval->resblks_avail = mp->m_resblks_avail;
511 * With per-cpu counters, this becomes an interesting
512 * problem. we needto work out if we are freeing or allocation
513 * blocks first, then we can do the modification as necessary.
515 * We do this under the m_sb_lock so that if we are near
516 * ENOSPC, we will hold out any changes while we work out
517 * what to do. This means that the amount of free space can
518 * change while we do this, so we need to retry if we end up
519 * trying to reserve more space than is available.
521 * We also use the xfs_mod_incore_sb() interface so that we
522 * don't have to care about whether per cpu counter are
523 * enabled, disabled or even compiled in....
526 spin_lock(&mp->m_sb_lock);
527 xfs_icsb_sync_counters_locked(mp, 0);
530 * If our previous reservation was larger than the current value,
531 * then move any unused blocks back to the free pool.
534 if (mp->m_resblks > request) {
535 lcounter = mp->m_resblks_avail - request;
536 if (lcounter > 0) { /* release unused blocks */
537 fdblks_delta = lcounter;
538 mp->m_resblks_avail -= lcounter;
540 mp->m_resblks = request;
544 free = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
546 goto out; /* ENOSPC and fdblks_delta = 0 */
548 delta = request - mp->m_resblks;
549 lcounter = free - delta;
551 /* We can't satisfy the request, just get what we can */
552 mp->m_resblks += free;
553 mp->m_resblks_avail += free;
554 fdblks_delta = -free;
556 fdblks_delta = -delta;
557 mp->m_resblks = request;
558 mp->m_resblks_avail += delta;
563 outval->resblks = mp->m_resblks;
564 outval->resblks_avail = mp->m_resblks_avail;
566 spin_unlock(&mp->m_sb_lock);
570 * If we are putting blocks back here, m_resblks_avail is
571 * already at it's max so this will put it in the free pool.
573 * If we need space, we'll either succeed in getting it
574 * from the free block count or we'll get an enospc. If
575 * we get a ENOSPC, it means things changed while we were
576 * calculating fdblks_delta and so we should try again to
577 * see if there is anything left to reserve.
579 * Don't set the reserved flag here - we don't want to reserve
580 * the extra reserve blocks from the reserve.....
583 error = xfs_mod_incore_sb(mp, XFS_SBS_FDBLOCKS, fdblks_delta, 0);
597 tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1);
598 if (xfs_trans_reserve(tp, 0, XFS_ICHANGE_LOG_RES(mp), 0, 0, 0)) {
599 xfs_trans_cancel(tp, 0);
604 xfs_ilock(ip, XFS_ILOCK_EXCL);
606 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
607 xfs_trans_ihold(tp, ip);
608 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
609 xfs_trans_set_sync(tp);
610 xfs_trans_commit(tp, 0);
612 xfs_iunlock(ip, XFS_ILOCK_EXCL);
621 case XFS_FSOP_GOING_FLAGS_DEFAULT: {
622 struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
624 if (sb && !IS_ERR(sb)) {
625 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
626 thaw_bdev(sb->s_bdev, sb);
631 case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
632 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
634 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
635 xfs_force_shutdown(mp,
636 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
639 return XFS_ERROR(EINVAL);