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"
49 * File system operations
58 geo->blocksize = mp->m_sb.sb_blocksize;
59 geo->rtextsize = mp->m_sb.sb_rextsize;
60 geo->agblocks = mp->m_sb.sb_agblocks;
61 geo->agcount = mp->m_sb.sb_agcount;
62 geo->logblocks = mp->m_sb.sb_logblocks;
63 geo->sectsize = mp->m_sb.sb_sectsize;
64 geo->inodesize = mp->m_sb.sb_inodesize;
65 geo->imaxpct = mp->m_sb.sb_imax_pct;
66 geo->datablocks = mp->m_sb.sb_dblocks;
67 geo->rtblocks = mp->m_sb.sb_rblocks;
68 geo->rtextents = mp->m_sb.sb_rextents;
69 geo->logstart = mp->m_sb.sb_logstart;
70 ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid));
71 memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid));
72 if (new_version >= 2) {
73 geo->sunit = mp->m_sb.sb_unit;
74 geo->swidth = mp->m_sb.sb_width;
76 if (new_version >= 3) {
77 geo->version = XFS_FSOP_GEOM_VERSION;
79 (XFS_SB_VERSION_HASATTR(&mp->m_sb) ?
80 XFS_FSOP_GEOM_FLAGS_ATTR : 0) |
81 (XFS_SB_VERSION_HASNLINK(&mp->m_sb) ?
82 XFS_FSOP_GEOM_FLAGS_NLINK : 0) |
83 (XFS_SB_VERSION_HASQUOTA(&mp->m_sb) ?
84 XFS_FSOP_GEOM_FLAGS_QUOTA : 0) |
85 (XFS_SB_VERSION_HASALIGN(&mp->m_sb) ?
86 XFS_FSOP_GEOM_FLAGS_IALIGN : 0) |
87 (XFS_SB_VERSION_HASDALIGN(&mp->m_sb) ?
88 XFS_FSOP_GEOM_FLAGS_DALIGN : 0) |
89 (XFS_SB_VERSION_HASSHARED(&mp->m_sb) ?
90 XFS_FSOP_GEOM_FLAGS_SHARED : 0) |
91 (XFS_SB_VERSION_HASEXTFLGBIT(&mp->m_sb) ?
92 XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) |
93 (XFS_SB_VERSION_HASDIRV2(&mp->m_sb) ?
94 XFS_FSOP_GEOM_FLAGS_DIRV2 : 0) |
95 (XFS_SB_VERSION_HASSECTOR(&mp->m_sb) ?
96 XFS_FSOP_GEOM_FLAGS_SECTOR : 0) |
97 (xfs_sb_version_haslazysbcount(&mp->m_sb) ?
98 XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) |
99 (XFS_SB_VERSION_HASATTR2(&mp->m_sb) ?
100 XFS_FSOP_GEOM_FLAGS_ATTR2 : 0);
101 geo->logsectsize = XFS_SB_VERSION_HASSECTOR(&mp->m_sb) ?
102 mp->m_sb.sb_logsectsize : BBSIZE;
103 geo->rtsectsize = mp->m_sb.sb_blocksize;
104 geo->dirblocksize = mp->m_dirblksize;
106 if (new_version >= 4) {
108 (XFS_SB_VERSION_HASLOGV2(&mp->m_sb) ?
109 XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
110 geo->logsunit = mp->m_sb.sb_logsunit;
116 xfs_growfs_data_private(
117 xfs_mount_t *mp, /* mount point for filesystem */
118 xfs_growfs_data_t *in) /* growfs data input struct */
124 xfs_extlen_t tmpsize;
125 xfs_alloc_rec_t *arec;
126 xfs_btree_sblock_t *block;
131 xfs_agnumber_t nagcount;
132 xfs_agnumber_t nagimax = 0;
133 xfs_rfsblock_t nb, nb_mod;
135 xfs_rfsblock_t nfree;
136 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 down_write(&mp->m_peraglock);
169 mp->m_perag = kmem_realloc(mp->m_perag,
170 sizeof(xfs_perag_t) * nagcount,
171 sizeof(xfs_perag_t) * oagcount,
173 memset(&mp->m_perag[oagcount], 0,
174 (nagcount - oagcount) * sizeof(xfs_perag_t));
175 mp->m_flags |= XFS_MOUNT_32BITINODES;
176 nagimax = xfs_initialize_perag(XFS_MTOVFS(mp), mp, nagcount);
177 up_write(&mp->m_peraglock);
179 tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
180 if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp),
181 XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) {
182 xfs_trans_cancel(tp, 0);
187 for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
189 * AG freelist header block
191 bp = xfs_buf_get(mp->m_ddev_targp,
192 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
193 XFS_FSS_TO_BB(mp, 1), 0);
194 agf = XFS_BUF_TO_AGF(bp);
195 memset(agf, 0, mp->m_sb.sb_sectsize);
196 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
197 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
198 agf->agf_seqno = cpu_to_be32(agno);
199 if (agno == nagcount - 1)
202 (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
204 agsize = mp->m_sb.sb_agblocks;
205 agf->agf_length = cpu_to_be32(agsize);
206 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
207 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
208 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
209 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
210 agf->agf_flfirst = 0;
211 agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
212 agf->agf_flcount = 0;
213 tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
214 agf->agf_freeblks = cpu_to_be32(tmpsize);
215 agf->agf_longest = cpu_to_be32(tmpsize);
216 error = xfs_bwrite(mp, bp);
221 * AG inode header block
223 bp = xfs_buf_get(mp->m_ddev_targp,
224 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
225 XFS_FSS_TO_BB(mp, 1), 0);
226 agi = XFS_BUF_TO_AGI(bp);
227 memset(agi, 0, mp->m_sb.sb_sectsize);
228 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
229 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
230 agi->agi_seqno = cpu_to_be32(agno);
231 agi->agi_length = cpu_to_be32(agsize);
233 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
234 agi->agi_level = cpu_to_be32(1);
235 agi->agi_freecount = 0;
236 agi->agi_newino = cpu_to_be32(NULLAGINO);
237 agi->agi_dirino = cpu_to_be32(NULLAGINO);
238 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
239 agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
240 error = xfs_bwrite(mp, bp);
245 * BNO btree root block
247 bp = xfs_buf_get(mp->m_ddev_targp,
248 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
249 BTOBB(mp->m_sb.sb_blocksize), 0);
250 block = XFS_BUF_TO_SBLOCK(bp);
251 memset(block, 0, mp->m_sb.sb_blocksize);
252 block->bb_magic = cpu_to_be32(XFS_ABTB_MAGIC);
254 block->bb_numrecs = cpu_to_be16(1);
255 block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
256 block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
257 arec = XFS_BTREE_REC_ADDR(xfs_alloc, block, 1);
258 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
259 arec->ar_blockcount = cpu_to_be32(
260 agsize - be32_to_cpu(arec->ar_startblock));
261 error = xfs_bwrite(mp, bp);
266 * CNT btree root block
268 bp = xfs_buf_get(mp->m_ddev_targp,
269 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
270 BTOBB(mp->m_sb.sb_blocksize), 0);
271 block = XFS_BUF_TO_SBLOCK(bp);
272 memset(block, 0, mp->m_sb.sb_blocksize);
273 block->bb_magic = cpu_to_be32(XFS_ABTC_MAGIC);
275 block->bb_numrecs = cpu_to_be16(1);
276 block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
277 block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
278 arec = XFS_BTREE_REC_ADDR(xfs_alloc, block, 1);
279 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
280 arec->ar_blockcount = cpu_to_be32(
281 agsize - be32_to_cpu(arec->ar_startblock));
282 nfree += be32_to_cpu(arec->ar_blockcount);
283 error = xfs_bwrite(mp, bp);
288 * INO btree root block
290 bp = xfs_buf_get(mp->m_ddev_targp,
291 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
292 BTOBB(mp->m_sb.sb_blocksize), 0);
293 block = XFS_BUF_TO_SBLOCK(bp);
294 memset(block, 0, mp->m_sb.sb_blocksize);
295 block->bb_magic = cpu_to_be32(XFS_IBT_MAGIC);
297 block->bb_numrecs = 0;
298 block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
299 block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
300 error = xfs_bwrite(mp, bp);
305 xfs_trans_agblocks_delta(tp, nfree);
307 * There are new blocks in the old last a.g.
311 * Change the agi length.
313 error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
318 agi = XFS_BUF_TO_AGI(bp);
319 be32_add(&agi->agi_length, new);
320 ASSERT(nagcount == oagcount ||
321 be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
322 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
326 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
331 agf = XFS_BUF_TO_AGF(bp);
332 be32_add(&agf->agf_length, new);
333 ASSERT(be32_to_cpu(agf->agf_length) ==
334 be32_to_cpu(agi->agi_length));
335 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
337 * Free the new space.
339 error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
340 be32_to_cpu(agf->agf_length) - new), new);
345 if (nagcount > oagcount)
346 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
347 if (nb > mp->m_sb.sb_dblocks)
348 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
349 nb - mp->m_sb.sb_dblocks);
351 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
353 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
354 error = xfs_trans_commit(tp, 0);
358 /* New allocation groups fully initialized, so update mount struct */
360 mp->m_maxagi = nagimax;
361 if (mp->m_sb.sb_imax_pct) {
362 __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
364 mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
367 for (agno = 1; agno < nagcount; agno++) {
368 error = xfs_read_buf(mp, mp->m_ddev_targp,
369 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
370 XFS_FSS_TO_BB(mp, 1), 0, &bp);
372 xfs_fs_cmn_err(CE_WARN, mp,
373 "error %d reading secondary superblock for ag %d",
377 sbp = XFS_BUF_TO_SBP(bp);
378 xfs_xlatesb(sbp, &mp->m_sb, -1, XFS_SB_ALL_BITS);
380 * If we get an error writing out the alternate superblocks,
381 * just issue a warning and continue. The real work is
382 * already done and committed.
384 if (!(error = xfs_bwrite(mp, bp))) {
387 xfs_fs_cmn_err(CE_WARN, mp,
388 "write error %d updating secondary superblock for ag %d",
390 break; /* no point in continuing */
396 xfs_trans_cancel(tp, XFS_TRANS_ABORT);
401 xfs_growfs_log_private(
402 xfs_mount_t *mp, /* mount point for filesystem */
403 xfs_growfs_log_t *in) /* growfs log input struct */
408 if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
409 return XFS_ERROR(EINVAL);
410 if (nb == mp->m_sb.sb_logblocks &&
411 in->isint == (mp->m_sb.sb_logstart != 0))
412 return XFS_ERROR(EINVAL);
414 * Moving the log is hard, need new interfaces to sync
415 * the log first, hold off all activity while moving it.
416 * Can have shorter or longer log in the same space,
417 * or transform internal to external log or vice versa.
419 return XFS_ERROR(ENOSYS);
423 * protected versions of growfs function acquire and release locks on the mount
424 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
432 xfs_growfs_data_t *in)
435 if (!cpsema(&mp->m_growlock))
436 return XFS_ERROR(EWOULDBLOCK);
437 error = xfs_growfs_data_private(mp, in);
438 vsema(&mp->m_growlock);
445 xfs_growfs_log_t *in)
448 if (!cpsema(&mp->m_growlock))
449 return XFS_ERROR(EWOULDBLOCK);
450 error = xfs_growfs_log_private(mp, in);
451 vsema(&mp->m_growlock);
456 * exported through ioctl XFS_IOC_FSCOUNTS
462 xfs_fsop_counts_t *cnt)
466 xfs_icsb_sync_counters_flags(mp, XFS_ICSB_LAZY_COUNT);
468 cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
469 cnt->freertx = mp->m_sb.sb_frextents;
470 cnt->freeino = mp->m_sb.sb_ifree;
471 cnt->allocino = mp->m_sb.sb_icount;
472 XFS_SB_UNLOCK(mp, s);
477 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
479 * xfs_reserve_blocks is called to set m_resblks
480 * in the in-core mount table. The number of unused reserved blocks
481 * is kept in m_resblks_avail.
483 * Reserve the requested number of blocks if available. Otherwise return
484 * as many as possible to satisfy the request. The actual number
485 * reserved are returned in outval
487 * A null inval pointer indicates that only the current reserved blocks
488 * available should be returned no settings are changed.
495 xfs_fsop_resblks_t *outval)
497 __int64_t lcounter, delta, fdblks_delta;
501 /* If inval is null, report current values and return */
503 if (inval == (__uint64_t *)NULL) {
504 outval->resblks = mp->m_resblks;
505 outval->resblks_avail = mp->m_resblks_avail;
512 * With per-cpu counters, this becomes an interesting
513 * problem. we needto work out if we are freeing or allocation
514 * blocks first, then we can do the modification as necessary.
516 * We do this under the XFS_SB_LOCK so that if we are near
517 * ENOSPC, we will hold out any changes while we work out
518 * what to do. This means that the amount of free space can
519 * change while we do this, so we need to retry if we end up
520 * trying to reserve more space than is available.
522 * We also use the xfs_mod_incore_sb() interface so that we
523 * don't have to care about whether per cpu counter are
524 * enabled, disabled or even compiled in....
528 xfs_icsb_sync_counters_flags(mp, XFS_ICSB_SB_LOCKED);
531 * If our previous reservation was larger than the current value,
532 * then move any unused blocks back to the free pool.
535 if (mp->m_resblks > request) {
536 lcounter = mp->m_resblks_avail - request;
537 if (lcounter > 0) { /* release unused blocks */
538 fdblks_delta = lcounter;
539 mp->m_resblks_avail -= lcounter;
541 mp->m_resblks = request;
545 free = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
547 goto out; /* ENOSPC and fdblks_delta = 0 */
549 delta = request - mp->m_resblks;
550 lcounter = free - delta;
552 /* We can't satisfy the request, just get what we can */
553 mp->m_resblks += free;
554 mp->m_resblks_avail += free;
555 fdblks_delta = -free;
556 mp->m_sb.sb_fdblocks = XFS_ALLOC_SET_ASIDE(mp);
558 fdblks_delta = -delta;
559 mp->m_sb.sb_fdblocks =
560 lcounter + XFS_ALLOC_SET_ASIDE(mp);
561 mp->m_resblks = request;
562 mp->m_resblks_avail += delta;
566 outval->resblks = mp->m_resblks;
567 outval->resblks_avail = mp->m_resblks_avail;
568 XFS_SB_UNLOCK(mp, s);
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);
600 tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1);
601 if (xfs_trans_reserve(tp, 0, XFS_ICHANGE_LOG_RES(mp), 0, 0, 0)) {
602 xfs_trans_cancel(tp, 0);
607 xfs_ilock(ip, XFS_ILOCK_EXCL);
609 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
610 xfs_trans_ihold(tp, ip);
611 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
612 xfs_trans_set_sync(tp);
613 xfs_trans_commit(tp, 0);
615 xfs_iunlock(ip, XFS_ILOCK_EXCL);
624 case XFS_FSOP_GOING_FLAGS_DEFAULT: {
625 struct bhv_vfs *vfsp = XFS_MTOVFS(mp);
626 struct super_block *sb = freeze_bdev(vfsp->vfs_super->s_bdev);
628 if (sb && !IS_ERR(sb)) {
629 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
630 thaw_bdev(sb->s_bdev, sb);
635 case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
636 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
638 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
639 xfs_force_shutdown(mp,
640 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
643 return XFS_ERROR(EINVAL);