projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
xfs: rmap btree requires more reserved free space
[deliverable/linux.git] / fs / xfs / xfs_fsops.c
1 /*
2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
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.
8 *
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.
13 *
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
17 */
18 #include "xfs.h"
19 #include "xfs_fs.h"
20 #include "xfs_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
24 #include "xfs_sb.h"
25 #include "xfs_mount.h"
26 #include "xfs_defer.h"
27 #include "xfs_da_format.h"
28 #include "xfs_da_btree.h"
29 #include "xfs_inode.h"
30 #include "xfs_trans.h"
31 #include "xfs_inode_item.h"
32 #include "xfs_error.h"
33 #include "xfs_btree.h"
34 #include "xfs_alloc_btree.h"
35 #include "xfs_alloc.h"
36 #include "xfs_rmap_btree.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"
42 #include "xfs_trace.h"
43 #include "xfs_log.h"
44 #include "xfs_filestream.h"
45 #include "xfs_rmap.h"
46
47 /*
48 * File system operations
49 */
50
51 int
52 xfs_fs_geometry(
53 xfs_mount_t *mp,
54 xfs_fsop_geom_t *geo,
55 int new_version)
56 {
57
58 memset(geo, 0, sizeof(*geo));
59
60 geo->blocksize = mp->m_sb.sb_blocksize;
61 geo->rtextsize = mp->m_sb.sb_rextsize;
62 geo->agblocks = mp->m_sb.sb_agblocks;
63 geo->agcount = mp->m_sb.sb_agcount;
64 geo->logblocks = mp->m_sb.sb_logblocks;
65 geo->sectsize = mp->m_sb.sb_sectsize;
66 geo->inodesize = mp->m_sb.sb_inodesize;
67 geo->imaxpct = mp->m_sb.sb_imax_pct;
68 geo->datablocks = mp->m_sb.sb_dblocks;
69 geo->rtblocks = mp->m_sb.sb_rblocks;
70 geo->rtextents = mp->m_sb.sb_rextents;
71 geo->logstart = mp->m_sb.sb_logstart;
72 ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid));
73 memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid));
74 if (new_version >= 2) {
75 geo->sunit = mp->m_sb.sb_unit;
76 geo->swidth = mp->m_sb.sb_width;
77 }
78 if (new_version >= 3) {
79 geo->version = XFS_FSOP_GEOM_VERSION;
80 geo->flags = XFS_FSOP_GEOM_FLAGS_NLINK |
81 XFS_FSOP_GEOM_FLAGS_DIRV2 |
82 (xfs_sb_version_hasattr(&mp->m_sb) ?
83 XFS_FSOP_GEOM_FLAGS_ATTR : 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_hasextflgbit(&mp->m_sb) ?
91 XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) |
92 (xfs_sb_version_hassector(&mp->m_sb) ?
93 XFS_FSOP_GEOM_FLAGS_SECTOR : 0) |
94 (xfs_sb_version_hasasciici(&mp->m_sb) ?
95 XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) |
96 (xfs_sb_version_haslazysbcount(&mp->m_sb) ?
97 XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) |
98 (xfs_sb_version_hasattr2(&mp->m_sb) ?
99 XFS_FSOP_GEOM_FLAGS_ATTR2 : 0) |
100 (xfs_sb_version_hasprojid32bit(&mp->m_sb) ?
101 XFS_FSOP_GEOM_FLAGS_PROJID32 : 0) |
102 (xfs_sb_version_hascrc(&mp->m_sb) ?
103 XFS_FSOP_GEOM_FLAGS_V5SB : 0) |
104 (xfs_sb_version_hasftype(&mp->m_sb) ?
105 XFS_FSOP_GEOM_FLAGS_FTYPE : 0) |
106 (xfs_sb_version_hasfinobt(&mp->m_sb) ?
107 XFS_FSOP_GEOM_FLAGS_FINOBT : 0) |
108 (xfs_sb_version_hassparseinodes(&mp->m_sb) ?
109 XFS_FSOP_GEOM_FLAGS_SPINODES : 0);
110 geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
111 mp->m_sb.sb_logsectsize : BBSIZE;
112 geo->rtsectsize = mp->m_sb.sb_blocksize;
113 geo->dirblocksize = mp->m_dir_geo->blksize;
114 }
115 if (new_version >= 4) {
116 geo->flags |=
117 (xfs_sb_version_haslogv2(&mp->m_sb) ?
118 XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
119 geo->logsunit = mp->m_sb.sb_logsunit;
120 }
121 return 0;
122 }
123
124 static struct xfs_buf *
125 xfs_growfs_get_hdr_buf(
126 struct xfs_mount *mp,
127 xfs_daddr_t blkno,
128 size_t numblks,
129 int flags,
130 const struct xfs_buf_ops *ops)
131 {
132 struct xfs_buf *bp;
133
134 bp = xfs_buf_get_uncached(mp->m_ddev_targp, numblks, flags);
135 if (!bp)
136 return NULL;
137
138 xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
139 bp->b_bn = blkno;
140 bp->b_maps[0].bm_bn = blkno;
141 bp->b_ops = ops;
142
143 return bp;
144 }
145
146 static int
147 xfs_growfs_data_private(
148 xfs_mount_t *mp, /* mount point for filesystem */
149 xfs_growfs_data_t *in) /* growfs data input struct */
150 {
151 xfs_agf_t *agf;
152 struct xfs_agfl *agfl;
153 xfs_agi_t *agi;
154 xfs_agnumber_t agno;
155 xfs_extlen_t agsize;
156 xfs_extlen_t tmpsize;
157 xfs_alloc_rec_t *arec;
158 xfs_buf_t *bp;
159 int bucket;
160 int dpct;
161 int error, saved_error = 0;
162 xfs_agnumber_t nagcount;
163 xfs_agnumber_t nagimax = 0;
164 xfs_rfsblock_t nb, nb_mod;
165 xfs_rfsblock_t new;
166 xfs_rfsblock_t nfree;
167 xfs_agnumber_t oagcount;
168 int pct;
169 xfs_trans_t *tp;
170
171 nb = in->newblocks;
172 pct = in->imaxpct;
173 if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
174 return -EINVAL;
175 if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
176 return error;
177 dpct = pct - mp->m_sb.sb_imax_pct;
178 error = xfs_buf_read_uncached(mp->m_ddev_targp,
179 XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
180 XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
181 if (error)
182 return error;
183 xfs_buf_relse(bp);
184
185 new = nb; /* use new as a temporary here */
186 nb_mod = do_div(new, mp->m_sb.sb_agblocks);
187 nagcount = new + (nb_mod != 0);
188 if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
189 nagcount--;
190 nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
191 if (nb < mp->m_sb.sb_dblocks)
192 return -EINVAL;
193 }
194 new = nb - mp->m_sb.sb_dblocks;
195 oagcount = mp->m_sb.sb_agcount;
196
197 /* allocate the new per-ag structures */
198 if (nagcount > oagcount) {
199 error = xfs_initialize_perag(mp, nagcount, &nagimax);
200 if (error)
201 return error;
202 }
203
204 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_growdata,
205 XFS_GROWFS_SPACE_RES(mp), 0, XFS_TRANS_RESERVE, &tp);
206 if (error)
207 return error;
208
209 /*
210 * Write new AG headers to disk. Non-transactional, but written
211 * synchronously so they are completed prior to the growfs transaction
212 * being logged.
213 */
214 nfree = 0;
215 for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
216 __be32 *agfl_bno;
217
218 /*
219 * AG freespace header block
220 */
221 bp = xfs_growfs_get_hdr_buf(mp,
222 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
223 XFS_FSS_TO_BB(mp, 1), 0,
224 &xfs_agf_buf_ops);
225 if (!bp) {
226 error = -ENOMEM;
227 goto error0;
228 }
229
230 agf = XFS_BUF_TO_AGF(bp);
231 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
232 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
233 agf->agf_seqno = cpu_to_be32(agno);
234 if (agno == nagcount - 1)
235 agsize =
236 nb -
237 (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
238 else
239 agsize = mp->m_sb.sb_agblocks;
240 agf->agf_length = cpu_to_be32(agsize);
241 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
242 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
243 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
244 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
245 if (xfs_sb_version_hasrmapbt(&mp->m_sb)) {
246 agf->agf_roots[XFS_BTNUM_RMAPi] =
247 cpu_to_be32(XFS_RMAP_BLOCK(mp));
248 agf->agf_levels[XFS_BTNUM_RMAPi] = cpu_to_be32(1);
249 }
250
251 agf->agf_flfirst = cpu_to_be32(1);
252 agf->agf_fllast = 0;
253 agf->agf_flcount = 0;
254 tmpsize = agsize - mp->m_ag_prealloc_blocks;
255 agf->agf_freeblks = cpu_to_be32(tmpsize);
256 agf->agf_longest = cpu_to_be32(tmpsize);
257 if (xfs_sb_version_hascrc(&mp->m_sb))
258 uuid_copy(&agf->agf_uuid, &mp->m_sb.sb_meta_uuid);
259
260 error = xfs_bwrite(bp);
261 xfs_buf_relse(bp);
262 if (error)
263 goto error0;
264
265 /*
266 * AG freelist header block
267 */
268 bp = xfs_growfs_get_hdr_buf(mp,
269 XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
270 XFS_FSS_TO_BB(mp, 1), 0,
271 &xfs_agfl_buf_ops);
272 if (!bp) {
273 error = -ENOMEM;
274 goto error0;
275 }
276
277 agfl = XFS_BUF_TO_AGFL(bp);
278 if (xfs_sb_version_hascrc(&mp->m_sb)) {
279 agfl->agfl_magicnum = cpu_to_be32(XFS_AGFL_MAGIC);
280 agfl->agfl_seqno = cpu_to_be32(agno);
281 uuid_copy(&agfl->agfl_uuid, &mp->m_sb.sb_meta_uuid);
282 }
283
284 agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, bp);
285 for (bucket = 0; bucket < XFS_AGFL_SIZE(mp); bucket++)
286 agfl_bno[bucket] = cpu_to_be32(NULLAGBLOCK);
287
288 error = xfs_bwrite(bp);
289 xfs_buf_relse(bp);
290 if (error)
291 goto error0;
292
293 /*
294 * AG inode header block
295 */
296 bp = xfs_growfs_get_hdr_buf(mp,
297 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
298 XFS_FSS_TO_BB(mp, 1), 0,
299 &xfs_agi_buf_ops);
300 if (!bp) {
301 error = -ENOMEM;
302 goto error0;
303 }
304
305 agi = XFS_BUF_TO_AGI(bp);
306 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
307 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
308 agi->agi_seqno = cpu_to_be32(agno);
309 agi->agi_length = cpu_to_be32(agsize);
310 agi->agi_count = 0;
311 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
312 agi->agi_level = cpu_to_be32(1);
313 agi->agi_freecount = 0;
314 agi->agi_newino = cpu_to_be32(NULLAGINO);
315 agi->agi_dirino = cpu_to_be32(NULLAGINO);
316 if (xfs_sb_version_hascrc(&mp->m_sb))
317 uuid_copy(&agi->agi_uuid, &mp->m_sb.sb_meta_uuid);
318 if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
319 agi->agi_free_root = cpu_to_be32(XFS_FIBT_BLOCK(mp));
320 agi->agi_free_level = cpu_to_be32(1);
321 }
322 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
323 agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
324
325 error = xfs_bwrite(bp);
326 xfs_buf_relse(bp);
327 if (error)
328 goto error0;
329
330 /*
331 * BNO btree root block
332 */
333 bp = xfs_growfs_get_hdr_buf(mp,
334 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
335 BTOBB(mp->m_sb.sb_blocksize), 0,
336 &xfs_allocbt_buf_ops);
337
338 if (!bp) {
339 error = -ENOMEM;
340 goto error0;
341 }
342
343 if (xfs_sb_version_hascrc(&mp->m_sb))
344 xfs_btree_init_block(mp, bp, XFS_ABTB_CRC_MAGIC, 0, 1,
345 agno, XFS_BTREE_CRC_BLOCKS);
346 else
347 xfs_btree_init_block(mp, bp, XFS_ABTB_MAGIC, 0, 1,
348 agno, 0);
349
350 arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
351 arec->ar_startblock = cpu_to_be32(mp->m_ag_prealloc_blocks);
352 arec->ar_blockcount = cpu_to_be32(
353 agsize - be32_to_cpu(arec->ar_startblock));
354
355 error = xfs_bwrite(bp);
356 xfs_buf_relse(bp);
357 if (error)
358 goto error0;
359
360 /*
361 * CNT btree root block
362 */
363 bp = xfs_growfs_get_hdr_buf(mp,
364 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
365 BTOBB(mp->m_sb.sb_blocksize), 0,
366 &xfs_allocbt_buf_ops);
367 if (!bp) {
368 error = -ENOMEM;
369 goto error0;
370 }
371
372 if (xfs_sb_version_hascrc(&mp->m_sb))
373 xfs_btree_init_block(mp, bp, XFS_ABTC_CRC_MAGIC, 0, 1,
374 agno, XFS_BTREE_CRC_BLOCKS);
375 else
376 xfs_btree_init_block(mp, bp, XFS_ABTC_MAGIC, 0, 1,
377 agno, 0);
378
379 arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
380 arec->ar_startblock = cpu_to_be32(mp->m_ag_prealloc_blocks);
381 arec->ar_blockcount = cpu_to_be32(
382 agsize - be32_to_cpu(arec->ar_startblock));
383 nfree += be32_to_cpu(arec->ar_blockcount);
384
385 error = xfs_bwrite(bp);
386 xfs_buf_relse(bp);
387 if (error)
388 goto error0;
389
390 /* RMAP btree root block */
391 if (xfs_sb_version_hasrmapbt(&mp->m_sb)) {
392 struct xfs_rmap_rec *rrec;
393 struct xfs_btree_block *block;
394
395 bp = xfs_growfs_get_hdr_buf(mp,
396 XFS_AGB_TO_DADDR(mp, agno, XFS_RMAP_BLOCK(mp)),
397 BTOBB(mp->m_sb.sb_blocksize), 0,
398 &xfs_rmapbt_buf_ops);
399 if (!bp) {
400 error = -ENOMEM;
401 goto error0;
402 }
403
404 xfs_btree_init_block(mp, bp, XFS_RMAP_CRC_MAGIC, 0, 0,
405 agno, XFS_BTREE_CRC_BLOCKS);
406 block = XFS_BUF_TO_BLOCK(bp);
407
408
409 /*
410 * mark the AG header regions as static metadata The BNO
411 * btree block is the first block after the headers, so
412 * it's location defines the size of region the static
413 * metadata consumes.
414 *
415 * Note: unlike mkfs, we never have to account for log
416 * space when growing the data regions
417 */
418 rrec = XFS_RMAP_REC_ADDR(block, 1);
419 rrec->rm_startblock = 0;
420 rrec->rm_blockcount = cpu_to_be32(XFS_BNO_BLOCK(mp));
421 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_FS);
422 rrec->rm_offset = 0;
423 be16_add_cpu(&block->bb_numrecs, 1);
424
425 /* account freespace btree root blocks */
426 rrec = XFS_RMAP_REC_ADDR(block, 2);
427 rrec->rm_startblock = cpu_to_be32(XFS_BNO_BLOCK(mp));
428 rrec->rm_blockcount = cpu_to_be32(2);
429 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_AG);
430 rrec->rm_offset = 0;
431 be16_add_cpu(&block->bb_numrecs, 1);
432
433 /* account inode btree root blocks */
434 rrec = XFS_RMAP_REC_ADDR(block, 3);
435 rrec->rm_startblock = cpu_to_be32(XFS_IBT_BLOCK(mp));
436 rrec->rm_blockcount = cpu_to_be32(XFS_RMAP_BLOCK(mp) -
437 XFS_IBT_BLOCK(mp));
438 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_INOBT);
439 rrec->rm_offset = 0;
440 be16_add_cpu(&block->bb_numrecs, 1);
441
442 /* account for rmap btree root */
443 rrec = XFS_RMAP_REC_ADDR(block, 4);
444 rrec->rm_startblock = cpu_to_be32(XFS_RMAP_BLOCK(mp));
445 rrec->rm_blockcount = cpu_to_be32(1);
446 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_AG);
447 rrec->rm_offset = 0;
448 be16_add_cpu(&block->bb_numrecs, 1);
449
450 error = xfs_bwrite(bp);
451 xfs_buf_relse(bp);
452 if (error)
453 goto error0;
454 }
455
456 /*
457 * INO btree root block
458 */
459 bp = xfs_growfs_get_hdr_buf(mp,
460 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
461 BTOBB(mp->m_sb.sb_blocksize), 0,
462 &xfs_inobt_buf_ops);
463 if (!bp) {
464 error = -ENOMEM;
465 goto error0;
466 }
467
468 if (xfs_sb_version_hascrc(&mp->m_sb))
469 xfs_btree_init_block(mp, bp, XFS_IBT_CRC_MAGIC, 0, 0,
470 agno, XFS_BTREE_CRC_BLOCKS);
471 else
472 xfs_btree_init_block(mp, bp, XFS_IBT_MAGIC, 0, 0,
473 agno, 0);
474
475 error = xfs_bwrite(bp);
476 xfs_buf_relse(bp);
477 if (error)
478 goto error0;
479
480 /*
481 * FINO btree root block
482 */
483 if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
484 bp = xfs_growfs_get_hdr_buf(mp,
485 XFS_AGB_TO_DADDR(mp, agno, XFS_FIBT_BLOCK(mp)),
486 BTOBB(mp->m_sb.sb_blocksize), 0,
487 &xfs_inobt_buf_ops);
488 if (!bp) {
489 error = -ENOMEM;
490 goto error0;
491 }
492
493 if (xfs_sb_version_hascrc(&mp->m_sb))
494 xfs_btree_init_block(mp, bp, XFS_FIBT_CRC_MAGIC,
495 0, 0, agno,
496 XFS_BTREE_CRC_BLOCKS);
497 else
498 xfs_btree_init_block(mp, bp, XFS_FIBT_MAGIC, 0,
499 0, agno, 0);
500
501 error = xfs_bwrite(bp);
502 xfs_buf_relse(bp);
503 if (error)
504 goto error0;
505 }
506
507 }
508 xfs_trans_agblocks_delta(tp, nfree);
509 /*
510 * There are new blocks in the old last a.g.
511 */
512 if (new) {
513 struct xfs_owner_info oinfo;
514
515 /*
516 * Change the agi length.
517 */
518 error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
519 if (error) {
520 goto error0;
521 }
522 ASSERT(bp);
523 agi = XFS_BUF_TO_AGI(bp);
524 be32_add_cpu(&agi->agi_length, new);
525 ASSERT(nagcount == oagcount ||
526 be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
527 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
528 /*
529 * Change agf length.
530 */
531 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
532 if (error) {
533 goto error0;
534 }
535 ASSERT(bp);
536 agf = XFS_BUF_TO_AGF(bp);
537 be32_add_cpu(&agf->agf_length, new);
538 ASSERT(be32_to_cpu(agf->agf_length) ==
539 be32_to_cpu(agi->agi_length));
540
541 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
542
543 /*
544 * Free the new space.
545 *
546 * XFS_RMAP_OWN_NULL is used here to tell the rmap btree that
547 * this doesn't actually exist in the rmap btree.
548 */
549 xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_NULL);
550 error = xfs_free_extent(tp,
551 XFS_AGB_TO_FSB(mp, agno,
552 be32_to_cpu(agf->agf_length) - new),
553 new, &oinfo);
554 if (error)
555 goto error0;
556 }
557
558 /*
559 * Update changed superblock fields transactionally. These are not
560 * seen by the rest of the world until the transaction commit applies
561 * them atomically to the superblock.
562 */
563 if (nagcount > oagcount)
564 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
565 if (nb > mp->m_sb.sb_dblocks)
566 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
567 nb - mp->m_sb.sb_dblocks);
568 if (nfree)
569 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
570 if (dpct)
571 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
572 xfs_trans_set_sync(tp);
573 error = xfs_trans_commit(tp);
574 if (error)
575 return error;
576
577 /* New allocation groups fully initialized, so update mount struct */
578 if (nagimax)
579 mp->m_maxagi = nagimax;
580 if (mp->m_sb.sb_imax_pct) {
581 __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
582 do_div(icount, 100);
583 mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
584 } else
585 mp->m_maxicount = 0;
586 xfs_set_low_space_thresholds(mp);
587 mp->m_alloc_set_aside = xfs_alloc_set_aside(mp);
588
589 /* update secondary superblocks. */
590 for (agno = 1; agno < nagcount; agno++) {
591 error = 0;
592 /*
593 * new secondary superblocks need to be zeroed, not read from
594 * disk as the contents of the new area we are growing into is
595 * completely unknown.
596 */
597 if (agno < oagcount) {
598 error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
599 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
600 XFS_FSS_TO_BB(mp, 1), 0, &bp,
601 &xfs_sb_buf_ops);
602 } else {
603 bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
604 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
605 XFS_FSS_TO_BB(mp, 1), 0);
606 if (bp) {
607 bp->b_ops = &xfs_sb_buf_ops;
608 xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
609 } else
610 error = -ENOMEM;
611 }
612
613 /*
614 * If we get an error reading or writing alternate superblocks,
615 * continue. xfs_repair chooses the "best" superblock based
616 * on most matches; if we break early, we'll leave more
617 * superblocks un-updated than updated, and xfs_repair may
618 * pick them over the properly-updated primary.
619 */
620 if (error) {
621 xfs_warn(mp,
622 "error %d reading secondary superblock for ag %d",
623 error, agno);
624 saved_error = error;
625 continue;
626 }
627 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb);
628
629 error = xfs_bwrite(bp);
630 xfs_buf_relse(bp);
631 if (error) {
632 xfs_warn(mp,
633 "write error %d updating secondary superblock for ag %d",
634 error, agno);
635 saved_error = error;
636 continue;
637 }
638 }
639 return saved_error ? saved_error : error;
640
641 error0:
642 xfs_trans_cancel(tp);
643 return error;
644 }
645
646 static int
647 xfs_growfs_log_private(
648 xfs_mount_t *mp, /* mount point for filesystem */
649 xfs_growfs_log_t *in) /* growfs log input struct */
650 {
651 xfs_extlen_t nb;
652
653 nb = in->newblocks;
654 if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
655 return -EINVAL;
656 if (nb == mp->m_sb.sb_logblocks &&
657 in->isint == (mp->m_sb.sb_logstart != 0))
658 return -EINVAL;
659 /*
660 * Moving the log is hard, need new interfaces to sync
661 * the log first, hold off all activity while moving it.
662 * Can have shorter or longer log in the same space,
663 * or transform internal to external log or vice versa.
664 */
665 return -ENOSYS;
666 }
667
668 /*
669 * protected versions of growfs function acquire and release locks on the mount
670 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
671 * XFS_IOC_FSGROWFSRT
672 */
673
674
675 int
676 xfs_growfs_data(
677 xfs_mount_t *mp,
678 xfs_growfs_data_t *in)
679 {
680 int error;
681
682 if (!capable(CAP_SYS_ADMIN))
683 return -EPERM;
684 if (!mutex_trylock(&mp->m_growlock))
685 return -EWOULDBLOCK;
686 error = xfs_growfs_data_private(mp, in);
687 /*
688 * Increment the generation unconditionally, the error could be from
689 * updating the secondary superblocks, in which case the new size
690 * is live already.
691 */
692 mp->m_generation++;
693 mutex_unlock(&mp->m_growlock);
694 return error;
695 }
696
697 int
698 xfs_growfs_log(
699 xfs_mount_t *mp,
700 xfs_growfs_log_t *in)
701 {
702 int error;
703
704 if (!capable(CAP_SYS_ADMIN))
705 return -EPERM;
706 if (!mutex_trylock(&mp->m_growlock))
707 return -EWOULDBLOCK;
708 error = xfs_growfs_log_private(mp, in);
709 mutex_unlock(&mp->m_growlock);
710 return error;
711 }
712
713 /*
714 * exported through ioctl XFS_IOC_FSCOUNTS
715 */
716
717 int
718 xfs_fs_counts(
719 xfs_mount_t *mp,
720 xfs_fsop_counts_t *cnt)
721 {
722 cnt->allocino = percpu_counter_read_positive(&mp->m_icount);
723 cnt->freeino = percpu_counter_read_positive(&mp->m_ifree);
724 cnt->freedata = percpu_counter_read_positive(&mp->m_fdblocks) -
725 mp->m_alloc_set_aside;
726
727 spin_lock(&mp->m_sb_lock);
728 cnt->freertx = mp->m_sb.sb_frextents;
729 spin_unlock(&mp->m_sb_lock);
730 return 0;
731 }
732
733 /*
734 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
735 *
736 * xfs_reserve_blocks is called to set m_resblks
737 * in the in-core mount table. The number of unused reserved blocks
738 * is kept in m_resblks_avail.
739 *
740 * Reserve the requested number of blocks if available. Otherwise return
741 * as many as possible to satisfy the request. The actual number
742 * reserved are returned in outval
743 *
744 * A null inval pointer indicates that only the current reserved blocks
745 * available should be returned no settings are changed.
746 */
747
748 int
749 xfs_reserve_blocks(
750 xfs_mount_t *mp,
751 __uint64_t *inval,
752 xfs_fsop_resblks_t *outval)
753 {
754 __int64_t lcounter, delta;
755 __int64_t fdblks_delta = 0;
756 __uint64_t request;
757 __int64_t free;
758 int error = 0;
759
760 /* If inval is null, report current values and return */
761 if (inval == (__uint64_t *)NULL) {
762 if (!outval)
763 return -EINVAL;
764 outval->resblks = mp->m_resblks;
765 outval->resblks_avail = mp->m_resblks_avail;
766 return 0;
767 }
768
769 request = *inval;
770
771 /*
772 * With per-cpu counters, this becomes an interesting problem. we need
773 * to work out if we are freeing or allocation blocks first, then we can
774 * do the modification as necessary.
775 *
776 * We do this under the m_sb_lock so that if we are near ENOSPC, we will
777 * hold out any changes while we work out what to do. This means that
778 * the amount of free space can change while we do this, so we need to
779 * retry if we end up trying to reserve more space than is available.
780 */
781 spin_lock(&mp->m_sb_lock);
782
783 /*
784 * If our previous reservation was larger than the current value,
785 * then move any unused blocks back to the free pool. Modify the resblks
786 * counters directly since we shouldn't have any problems unreserving
787 * space.
788 */
789 if (mp->m_resblks > request) {
790 lcounter = mp->m_resblks_avail - request;
791 if (lcounter > 0) { /* release unused blocks */
792 fdblks_delta = lcounter;
793 mp->m_resblks_avail -= lcounter;
794 }
795 mp->m_resblks = request;
796 if (fdblks_delta) {
797 spin_unlock(&mp->m_sb_lock);
798 error = xfs_mod_fdblocks(mp, fdblks_delta, 0);
799 spin_lock(&mp->m_sb_lock);
800 }
801
802 goto out;
803 }
804
805 /*
806 * If the request is larger than the current reservation, reserve the
807 * blocks before we update the reserve counters. Sample m_fdblocks and
808 * perform a partial reservation if the request exceeds free space.
809 */
810 error = -ENOSPC;
811 do {
812 free = percpu_counter_sum(&mp->m_fdblocks) -
813 mp->m_alloc_set_aside;
814 if (!free)
815 break;
816
817 delta = request - mp->m_resblks;
818 lcounter = free - delta;
819 if (lcounter < 0)
820 /* We can't satisfy the request, just get what we can */
821 fdblks_delta = free;
822 else
823 fdblks_delta = delta;
824
825 /*
826 * We'll either succeed in getting space from the free block
827 * count or we'll get an ENOSPC. If we get a ENOSPC, it means
828 * things changed while we were calculating fdblks_delta and so
829 * we should try again to see if there is anything left to
830 * reserve.
831 *
832 * Don't set the reserved flag here - we don't want to reserve
833 * the extra reserve blocks from the reserve.....
834 */
835 spin_unlock(&mp->m_sb_lock);
836 error = xfs_mod_fdblocks(mp, -fdblks_delta, 0);
837 spin_lock(&mp->m_sb_lock);
838 } while (error == -ENOSPC);
839
840 /*
841 * Update the reserve counters if blocks have been successfully
842 * allocated.
843 */
844 if (!error && fdblks_delta) {
845 mp->m_resblks += fdblks_delta;
846 mp->m_resblks_avail += fdblks_delta;
847 }
848
849 out:
850 if (outval) {
851 outval->resblks = mp->m_resblks;
852 outval->resblks_avail = mp->m_resblks_avail;
853 }
854
855 spin_unlock(&mp->m_sb_lock);
856 return error;
857 }
858
859 int
860 xfs_fs_goingdown(
861 xfs_mount_t *mp,
862 __uint32_t inflags)
863 {
864 switch (inflags) {
865 case XFS_FSOP_GOING_FLAGS_DEFAULT: {
866 struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
867
868 if (sb && !IS_ERR(sb)) {
869 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
870 thaw_bdev(sb->s_bdev, sb);
871 }
872
873 break;
874 }
875 case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
876 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
877 break;
878 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
879 xfs_force_shutdown(mp,
880 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
881 break;
882 default:
883 return -EINVAL;
884 }
885
886 return 0;
887 }
888
889 /*
890 * Force a shutdown of the filesystem instantly while keeping the filesystem
891 * consistent. We don't do an unmount here; just shutdown the shop, make sure
892 * that absolutely nothing persistent happens to this filesystem after this
893 * point.
894 */
895 void
896 xfs_do_force_shutdown(
897 xfs_mount_t *mp,
898 int flags,
899 char *fname,
900 int lnnum)
901 {
902 int logerror;
903
904 logerror = flags & SHUTDOWN_LOG_IO_ERROR;
905
906 if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
907 xfs_notice(mp,
908 "%s(0x%x) called from line %d of file %s. Return address = 0x%p",
909 __func__, flags, lnnum, fname, __return_address);
910 }
911 /*
912 * No need to duplicate efforts.
913 */
914 if (XFS_FORCED_SHUTDOWN(mp) && !logerror)
915 return;
916
917 /*
918 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
919 * queue up anybody new on the log reservations, and wakes up
920 * everybody who's sleeping on log reservations to tell them
921 * the bad news.
922 */
923 if (xfs_log_force_umount(mp, logerror))
924 return;
925
926 if (flags & SHUTDOWN_CORRUPT_INCORE) {
927 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_CORRUPT,
928 "Corruption of in-memory data detected. Shutting down filesystem");
929 if (XFS_ERRLEVEL_HIGH <= xfs_error_level)
930 xfs_stack_trace();
931 } else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
932 if (logerror) {
933 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_LOGERROR,
934 "Log I/O Error Detected. Shutting down filesystem");
935 } else if (flags & SHUTDOWN_DEVICE_REQ) {
936 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
937 "All device paths lost. Shutting down filesystem");
938 } else if (!(flags & SHUTDOWN_REMOTE_REQ)) {
939 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
940 "I/O Error Detected. Shutting down filesystem");
941 }
942 }
943 if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
944 xfs_alert(mp,
945 "Please umount the filesystem and rectify the problem(s)");
946 }
947 }
Linux kernel - Deliverables
This page took 0.051687 seconds and 5 git commands to generate.