?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
mm: fix exec activate_mm vs TLB shootdown and lazy tlb switching race
[linux/fpc-iii.git] / fs / xfs / xfs_fsops.c
blob40783a313df957b3cd4ce45f83864ceb87dc1162
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 #include "xfs_ag_resv.h"
47
48 /*
49 * File system operations
50 */
51
52 int
53 xfs_fs_geometry(
54 xfs_mount_t *mp,
55 xfs_fsop_geom_t *geo,
56 int new_version)
57 {
58
59 memset(geo, 0, sizeof(*geo));
60
61 geo->blocksize = mp->m_sb.sb_blocksize;
62 geo->rtextsize = mp->m_sb.sb_rextsize;
63 geo->agblocks = mp->m_sb.sb_agblocks;
64 geo->agcount = mp->m_sb.sb_agcount;
65 geo->logblocks = mp->m_sb.sb_logblocks;
66 geo->sectsize = mp->m_sb.sb_sectsize;
67 geo->inodesize = mp->m_sb.sb_inodesize;
68 geo->imaxpct = mp->m_sb.sb_imax_pct;
69 geo->datablocks = mp->m_sb.sb_dblocks;
70 geo->rtblocks = mp->m_sb.sb_rblocks;
71 geo->rtextents = mp->m_sb.sb_rextents;
72 geo->logstart = mp->m_sb.sb_logstart;
73 ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid));
74 memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid));
75 if (new_version >= 2) {
76 geo->sunit = mp->m_sb.sb_unit;
77 geo->swidth = mp->m_sb.sb_width;
78 }
79 if (new_version >= 3) {
80 geo->version = XFS_FSOP_GEOM_VERSION;
81 geo->flags = XFS_FSOP_GEOM_FLAGS_NLINK |
82 XFS_FSOP_GEOM_FLAGS_DIRV2 |
83 (xfs_sb_version_hasattr(&mp->m_sb) ?
84 XFS_FSOP_GEOM_FLAGS_ATTR : 0) |
85 (xfs_sb_version_hasquota(&mp->m_sb) ?
86 XFS_FSOP_GEOM_FLAGS_QUOTA : 0) |
87 (xfs_sb_version_hasalign(&mp->m_sb) ?
88 XFS_FSOP_GEOM_FLAGS_IALIGN : 0) |
89 (xfs_sb_version_hasdalign(&mp->m_sb) ?
90 XFS_FSOP_GEOM_FLAGS_DALIGN : 0) |
91 (xfs_sb_version_hasextflgbit(&mp->m_sb) ?
92 XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) |
93 (xfs_sb_version_hassector(&mp->m_sb) ?
94 XFS_FSOP_GEOM_FLAGS_SECTOR : 0) |
95 (xfs_sb_version_hasasciici(&mp->m_sb) ?
96 XFS_FSOP_GEOM_FLAGS_DIRV2CI : 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 (xfs_sb_version_hasprojid32bit(&mp->m_sb) ?
102 XFS_FSOP_GEOM_FLAGS_PROJID32 : 0) |
103 (xfs_sb_version_hascrc(&mp->m_sb) ?
104 XFS_FSOP_GEOM_FLAGS_V5SB : 0) |
105 (xfs_sb_version_hasftype(&mp->m_sb) ?
106 XFS_FSOP_GEOM_FLAGS_FTYPE : 0) |
107 (xfs_sb_version_hasfinobt(&mp->m_sb) ?
108 XFS_FSOP_GEOM_FLAGS_FINOBT : 0) |
109 (xfs_sb_version_hassparseinodes(&mp->m_sb) ?
110 XFS_FSOP_GEOM_FLAGS_SPINODES : 0) |
111 (xfs_sb_version_hasrmapbt(&mp->m_sb) ?
112 XFS_FSOP_GEOM_FLAGS_RMAPBT : 0) |
113 (xfs_sb_version_hasreflink(&mp->m_sb) ?
114 XFS_FSOP_GEOM_FLAGS_REFLINK : 0);
115 geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
116 mp->m_sb.sb_logsectsize : BBSIZE;
117 geo->rtsectsize = mp->m_sb.sb_blocksize;
118 geo->dirblocksize = mp->m_dir_geo->blksize;
119 }
120 if (new_version >= 4) {
121 geo->flags |=
122 (xfs_sb_version_haslogv2(&mp->m_sb) ?
123 XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
124 geo->logsunit = mp->m_sb.sb_logsunit;
125 }
126 return 0;
127 }
128
129 static struct xfs_buf *
130 xfs_growfs_get_hdr_buf(
131 struct xfs_mount *mp,
132 xfs_daddr_t blkno,
133 size_t numblks,
134 int flags,
135 const struct xfs_buf_ops *ops)
136 {
137 struct xfs_buf *bp;
138
139 bp = xfs_buf_get_uncached(mp->m_ddev_targp, numblks, flags);
140 if (!bp)
141 return NULL;
142
143 xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
144 bp->b_bn = blkno;
145 bp->b_maps[0].bm_bn = blkno;
146 bp->b_ops = ops;
147
148 return bp;
149 }
150
151 static int
152 xfs_growfs_data_private(
153 xfs_mount_t *mp, /* mount point for filesystem */
154 xfs_growfs_data_t *in) /* growfs data input struct */
155 {
156 xfs_agf_t *agf;
157 struct xfs_agfl *agfl;
158 xfs_agi_t *agi;
159 xfs_agnumber_t agno;
160 xfs_extlen_t agsize;
161 xfs_extlen_t tmpsize;
162 xfs_alloc_rec_t *arec;
163 xfs_buf_t *bp;
164 int bucket;
165 int dpct;
166 int error, saved_error = 0;
167 xfs_agnumber_t nagcount;
168 xfs_agnumber_t nagimax = 0;
169 xfs_rfsblock_t nb, nb_mod;
170 xfs_rfsblock_t new;
171 xfs_rfsblock_t nfree;
172 xfs_agnumber_t oagcount;
173 int pct;
174 xfs_trans_t *tp;
175
176 nb = in->newblocks;
177 pct = in->imaxpct;
178 if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
179 return -EINVAL;
180 if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
181 return error;
182 dpct = pct - mp->m_sb.sb_imax_pct;
183 error = xfs_buf_read_uncached(mp->m_ddev_targp,
184 XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
185 XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
186 if (error)
187 return error;
188 xfs_buf_relse(bp);
189
190 new = nb; /* use new as a temporary here */
191 nb_mod = do_div(new, mp->m_sb.sb_agblocks);
192 nagcount = new + (nb_mod != 0);
193 if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
194 nagcount--;
195 nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
196 if (nb < mp->m_sb.sb_dblocks)
197 return -EINVAL;
198 }
199 new = nb - mp->m_sb.sb_dblocks;
200 oagcount = mp->m_sb.sb_agcount;
201
202 /* allocate the new per-ag structures */
203 if (nagcount > oagcount) {
204 error = xfs_initialize_perag(mp, nagcount, &nagimax);
205 if (error)
206 return error;
207 }
208
209 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_growdata,
210 XFS_GROWFS_SPACE_RES(mp), 0, XFS_TRANS_RESERVE, &tp);
211 if (error)
212 return error;
213
214 /*
215 * Write new AG headers to disk. Non-transactional, but written
216 * synchronously so they are completed prior to the growfs transaction
217 * being logged.
218 */
219 nfree = 0;
220 for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
221 __be32 *agfl_bno;
222
223 /*
224 * AG freespace header block
225 */
226 bp = xfs_growfs_get_hdr_buf(mp,
227 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
228 XFS_FSS_TO_BB(mp, 1), 0,
229 &xfs_agf_buf_ops);
230 if (!bp) {
231 error = -ENOMEM;
232 goto error0;
233 }
234
235 agf = XFS_BUF_TO_AGF(bp);
236 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
237 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
238 agf->agf_seqno = cpu_to_be32(agno);
239 if (agno == nagcount - 1)
240 agsize =
241 nb -
242 (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
243 else
244 agsize = mp->m_sb.sb_agblocks;
245 agf->agf_length = cpu_to_be32(agsize);
246 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
247 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
248 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
249 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
250 if (xfs_sb_version_hasrmapbt(&mp->m_sb)) {
251 agf->agf_roots[XFS_BTNUM_RMAPi] =
252 cpu_to_be32(XFS_RMAP_BLOCK(mp));
253 agf->agf_levels[XFS_BTNUM_RMAPi] = cpu_to_be32(1);
254 agf->agf_rmap_blocks = cpu_to_be32(1);
255 }
256
257 agf->agf_flfirst = cpu_to_be32(1);
258 agf->agf_fllast = 0;
259 agf->agf_flcount = 0;
260 tmpsize = agsize - mp->m_ag_prealloc_blocks;
261 agf->agf_freeblks = cpu_to_be32(tmpsize);
262 agf->agf_longest = cpu_to_be32(tmpsize);
263 if (xfs_sb_version_hascrc(&mp->m_sb))
264 uuid_copy(&agf->agf_uuid, &mp->m_sb.sb_meta_uuid);
265 if (xfs_sb_version_hasreflink(&mp->m_sb)) {
266 agf->agf_refcount_root = cpu_to_be32(
267 xfs_refc_block(mp));
268 agf->agf_refcount_level = cpu_to_be32(1);
269 agf->agf_refcount_blocks = cpu_to_be32(1);
270 }
271
272 error = xfs_bwrite(bp);
273 xfs_buf_relse(bp);
274 if (error)
275 goto error0;
276
277 /*
278 * AG freelist header block
279 */
280 bp = xfs_growfs_get_hdr_buf(mp,
281 XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
282 XFS_FSS_TO_BB(mp, 1), 0,
283 &xfs_agfl_buf_ops);
284 if (!bp) {
285 error = -ENOMEM;
286 goto error0;
287 }
288
289 agfl = XFS_BUF_TO_AGFL(bp);
290 if (xfs_sb_version_hascrc(&mp->m_sb)) {
291 agfl->agfl_magicnum = cpu_to_be32(XFS_AGFL_MAGIC);
292 agfl->agfl_seqno = cpu_to_be32(agno);
293 uuid_copy(&agfl->agfl_uuid, &mp->m_sb.sb_meta_uuid);
294 }
295
296 agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, bp);
297 for (bucket = 0; bucket < xfs_agfl_size(mp); bucket++)
298 agfl_bno[bucket] = cpu_to_be32(NULLAGBLOCK);
299
300 error = xfs_bwrite(bp);
301 xfs_buf_relse(bp);
302 if (error)
303 goto error0;
304
305 /*
306 * AG inode header block
307 */
308 bp = xfs_growfs_get_hdr_buf(mp,
309 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
310 XFS_FSS_TO_BB(mp, 1), 0,
311 &xfs_agi_buf_ops);
312 if (!bp) {
313 error = -ENOMEM;
314 goto error0;
315 }
316
317 agi = XFS_BUF_TO_AGI(bp);
318 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
319 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
320 agi->agi_seqno = cpu_to_be32(agno);
321 agi->agi_length = cpu_to_be32(agsize);
322 agi->agi_count = 0;
323 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
324 agi->agi_level = cpu_to_be32(1);
325 agi->agi_freecount = 0;
326 agi->agi_newino = cpu_to_be32(NULLAGINO);
327 agi->agi_dirino = cpu_to_be32(NULLAGINO);
328 if (xfs_sb_version_hascrc(&mp->m_sb))
329 uuid_copy(&agi->agi_uuid, &mp->m_sb.sb_meta_uuid);
330 if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
331 agi->agi_free_root = cpu_to_be32(XFS_FIBT_BLOCK(mp));
332 agi->agi_free_level = cpu_to_be32(1);
333 }
334 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
335 agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
336
337 error = xfs_bwrite(bp);
338 xfs_buf_relse(bp);
339 if (error)
340 goto error0;
341
342 /*
343 * BNO btree root block
344 */
345 bp = xfs_growfs_get_hdr_buf(mp,
346 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
347 BTOBB(mp->m_sb.sb_blocksize), 0,
348 &xfs_allocbt_buf_ops);
349
350 if (!bp) {
351 error = -ENOMEM;
352 goto error0;
353 }
354
355 xfs_btree_init_block(mp, bp, XFS_BTNUM_BNO, 0, 1, agno, 0);
356
357 arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
358 arec->ar_startblock = cpu_to_be32(mp->m_ag_prealloc_blocks);
359 arec->ar_blockcount = cpu_to_be32(
360 agsize - be32_to_cpu(arec->ar_startblock));
361
362 error = xfs_bwrite(bp);
363 xfs_buf_relse(bp);
364 if (error)
365 goto error0;
366
367 /*
368 * CNT btree root block
369 */
370 bp = xfs_growfs_get_hdr_buf(mp,
371 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
372 BTOBB(mp->m_sb.sb_blocksize), 0,
373 &xfs_allocbt_buf_ops);
374 if (!bp) {
375 error = -ENOMEM;
376 goto error0;
377 }
378
379 xfs_btree_init_block(mp, bp, XFS_BTNUM_CNT, 0, 1, agno, 0);
380
381 arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
382 arec->ar_startblock = cpu_to_be32(mp->m_ag_prealloc_blocks);
383 arec->ar_blockcount = cpu_to_be32(
384 agsize - be32_to_cpu(arec->ar_startblock));
385 nfree += be32_to_cpu(arec->ar_blockcount);
386
387 error = xfs_bwrite(bp);
388 xfs_buf_relse(bp);
389 if (error)
390 goto error0;
391
392 /* RMAP btree root block */
393 if (xfs_sb_version_hasrmapbt(&mp->m_sb)) {
394 struct xfs_rmap_rec *rrec;
395 struct xfs_btree_block *block;
396
397 bp = xfs_growfs_get_hdr_buf(mp,
398 XFS_AGB_TO_DADDR(mp, agno, XFS_RMAP_BLOCK(mp)),
399 BTOBB(mp->m_sb.sb_blocksize), 0,
400 &xfs_rmapbt_buf_ops);
401 if (!bp) {
402 error = -ENOMEM;
403 goto error0;
404 }
405
406 xfs_btree_init_block(mp, bp, XFS_BTNUM_RMAP, 0, 0,
407 agno, 0);
408 block = XFS_BUF_TO_BLOCK(bp);
409
410
411 /*
412 * mark the AG header regions as static metadata The BNO
413 * btree block is the first block after the headers, so
414 * it's location defines the size of region the static
415 * metadata consumes.
416 *
417 * Note: unlike mkfs, we never have to account for log
418 * space when growing the data regions
419 */
420 rrec = XFS_RMAP_REC_ADDR(block, 1);
421 rrec->rm_startblock = 0;
422 rrec->rm_blockcount = cpu_to_be32(XFS_BNO_BLOCK(mp));
423 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_FS);
424 rrec->rm_offset = 0;
425 be16_add_cpu(&block->bb_numrecs, 1);
426
427 /* account freespace btree root blocks */
428 rrec = XFS_RMAP_REC_ADDR(block, 2);
429 rrec->rm_startblock = cpu_to_be32(XFS_BNO_BLOCK(mp));
430 rrec->rm_blockcount = cpu_to_be32(2);
431 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_AG);
432 rrec->rm_offset = 0;
433 be16_add_cpu(&block->bb_numrecs, 1);
434
435 /* account inode btree root blocks */
436 rrec = XFS_RMAP_REC_ADDR(block, 3);
437 rrec->rm_startblock = cpu_to_be32(XFS_IBT_BLOCK(mp));
438 rrec->rm_blockcount = cpu_to_be32(XFS_RMAP_BLOCK(mp) -
439 XFS_IBT_BLOCK(mp));
440 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_INOBT);
441 rrec->rm_offset = 0;
442 be16_add_cpu(&block->bb_numrecs, 1);
443
444 /* account for rmap btree root */
445 rrec = XFS_RMAP_REC_ADDR(block, 4);
446 rrec->rm_startblock = cpu_to_be32(XFS_RMAP_BLOCK(mp));
447 rrec->rm_blockcount = cpu_to_be32(1);
448 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_AG);
449 rrec->rm_offset = 0;
450 be16_add_cpu(&block->bb_numrecs, 1);
451
452 /* account for refc btree root */
453 if (xfs_sb_version_hasreflink(&mp->m_sb)) {
454 rrec = XFS_RMAP_REC_ADDR(block, 5);
455 rrec->rm_startblock = cpu_to_be32(
456 xfs_refc_block(mp));
457 rrec->rm_blockcount = cpu_to_be32(1);
458 rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_REFC);
459 rrec->rm_offset = 0;
460 be16_add_cpu(&block->bb_numrecs, 1);
461 }
462
463 error = xfs_bwrite(bp);
464 xfs_buf_relse(bp);
465 if (error)
466 goto error0;
467 }
468
469 /*
470 * INO btree root block
471 */
472 bp = xfs_growfs_get_hdr_buf(mp,
473 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
474 BTOBB(mp->m_sb.sb_blocksize), 0,
475 &xfs_inobt_buf_ops);
476 if (!bp) {
477 error = -ENOMEM;
478 goto error0;
479 }
480
481 xfs_btree_init_block(mp, bp, XFS_BTNUM_INO , 0, 0, agno, 0);
482
483 error = xfs_bwrite(bp);
484 xfs_buf_relse(bp);
485 if (error)
486 goto error0;
487
488 /*
489 * FINO btree root block
490 */
491 if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
492 bp = xfs_growfs_get_hdr_buf(mp,
493 XFS_AGB_TO_DADDR(mp, agno, XFS_FIBT_BLOCK(mp)),
494 BTOBB(mp->m_sb.sb_blocksize), 0,
495 &xfs_inobt_buf_ops);
496 if (!bp) {
497 error = -ENOMEM;
498 goto error0;
499 }
500
501 xfs_btree_init_block(mp, bp, XFS_BTNUM_FINO,
502 0, 0, agno, 0);
503
504 error = xfs_bwrite(bp);
505 xfs_buf_relse(bp);
506 if (error)
507 goto error0;
508 }
509
510 /*
511 * refcount btree root block
512 */
513 if (xfs_sb_version_hasreflink(&mp->m_sb)) {
514 bp = xfs_growfs_get_hdr_buf(mp,
515 XFS_AGB_TO_DADDR(mp, agno, xfs_refc_block(mp)),
516 BTOBB(mp->m_sb.sb_blocksize), 0,
517 &xfs_refcountbt_buf_ops);
518 if (!bp) {
519 error = -ENOMEM;
520 goto error0;
521 }
522
523 xfs_btree_init_block(mp, bp, XFS_BTNUM_REFC,
524 0, 0, agno, 0);
525
526 error = xfs_bwrite(bp);
527 xfs_buf_relse(bp);
528 if (error)
529 goto error0;
530 }
531 }
532 xfs_trans_agblocks_delta(tp, nfree);
533 /*
534 * There are new blocks in the old last a.g.
535 */
536 if (new) {
537 struct xfs_owner_info oinfo;
538
539 /*
540 * Change the agi length.
541 */
542 error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
543 if (error) {
544 goto error0;
545 }
546 ASSERT(bp);
547 agi = XFS_BUF_TO_AGI(bp);
548 be32_add_cpu(&agi->agi_length, new);
549 ASSERT(nagcount == oagcount ||
550 be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
551 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
552 /*
553 * Change agf length.
554 */
555 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
556 if (error) {
557 goto error0;
558 }
559 ASSERT(bp);
560 agf = XFS_BUF_TO_AGF(bp);
561 be32_add_cpu(&agf->agf_length, new);
562 ASSERT(be32_to_cpu(agf->agf_length) ==
563 be32_to_cpu(agi->agi_length));
564
565 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
566
567 /*
568 * Free the new space.
569 *
570 * XFS_RMAP_OWN_NULL is used here to tell the rmap btree that
571 * this doesn't actually exist in the rmap btree.
572 */
573 xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_NULL);
574 error = xfs_free_extent(tp,
575 XFS_AGB_TO_FSB(mp, agno,
576 be32_to_cpu(agf->agf_length) - new),
577 new, &oinfo, XFS_AG_RESV_NONE);
578 if (error)
579 goto error0;
580 }
581
582 /*
583 * Update changed superblock fields transactionally. These are not
584 * seen by the rest of the world until the transaction commit applies
585 * them atomically to the superblock.
586 */
587 if (nagcount > oagcount)
588 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
589 if (nb > mp->m_sb.sb_dblocks)
590 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
591 nb - mp->m_sb.sb_dblocks);
592 if (nfree)
593 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
594 if (dpct)
595 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
596 xfs_trans_set_sync(tp);
597 error = xfs_trans_commit(tp);
598 if (error)
599 return error;
600
601 /* New allocation groups fully initialized, so update mount struct */
602 if (nagimax)
603 mp->m_maxagi = nagimax;
604 if (mp->m_sb.sb_imax_pct) {
605 uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
606 do_div(icount, 100);
607 mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
608 } else
609 mp->m_maxicount = 0;
610 xfs_set_low_space_thresholds(mp);
611 mp->m_alloc_set_aside = xfs_alloc_set_aside(mp);
612
613 /*
614 * If we expanded the last AG, free the per-AG reservation
615 * so we can reinitialize it with the new size.
616 */
617 if (new) {
618 struct xfs_perag *pag;
619
620 pag = xfs_perag_get(mp, agno);
621 error = xfs_ag_resv_free(pag);
622 xfs_perag_put(pag);
623 if (error)
624 goto out;
625 }
626
627 /* Reserve AG metadata blocks. */
628 error = xfs_fs_reserve_ag_blocks(mp);
629 if (error && error != -ENOSPC)
630 goto out;
631
632 /* update secondary superblocks. */
633 for (agno = 1; agno < nagcount; agno++) {
634 error = 0;
635 /*
636 * new secondary superblocks need to be zeroed, not read from
637 * disk as the contents of the new area we are growing into is
638 * completely unknown.
639 */
640 if (agno < oagcount) {
641 error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
642 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
643 XFS_FSS_TO_BB(mp, 1), 0, &bp,
644 &xfs_sb_buf_ops);
645 } else {
646 bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
647 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
648 XFS_FSS_TO_BB(mp, 1), 0);
649 if (bp) {
650 bp->b_ops = &xfs_sb_buf_ops;
651 xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
652 } else
653 error = -ENOMEM;
654 }
655
656 /*
657 * If we get an error reading or writing alternate superblocks,
658 * continue. xfs_repair chooses the "best" superblock based
659 * on most matches; if we break early, we'll leave more
660 * superblocks un-updated than updated, and xfs_repair may
661 * pick them over the properly-updated primary.
662 */
663 if (error) {
664 xfs_warn(mp,
665 "error %d reading secondary superblock for ag %d",
666 error, agno);
667 saved_error = error;
668 continue;
669 }
670 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb);
671
672 error = xfs_bwrite(bp);
673 xfs_buf_relse(bp);
674 if (error) {
675 xfs_warn(mp,
676 "write error %d updating secondary superblock for ag %d",
677 error, agno);
678 saved_error = error;
679 continue;
680 }
681 }
682
683 out:
684 return saved_error ? saved_error : error;
685
686 error0:
687 xfs_trans_cancel(tp);
688 return error;
689 }
690
691 static int
692 xfs_growfs_log_private(
693 xfs_mount_t *mp, /* mount point for filesystem */
694 xfs_growfs_log_t *in) /* growfs log input struct */
695 {
696 xfs_extlen_t nb;
697
698 nb = in->newblocks;
699 if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
700 return -EINVAL;
701 if (nb == mp->m_sb.sb_logblocks &&
702 in->isint == (mp->m_sb.sb_logstart != 0))
703 return -EINVAL;
704 /*
705 * Moving the log is hard, need new interfaces to sync
706 * the log first, hold off all activity while moving it.
707 * Can have shorter or longer log in the same space,
708 * or transform internal to external log or vice versa.
709 */
710 return -ENOSYS;
711 }
712
713 /*
714 * protected versions of growfs function acquire and release locks on the mount
715 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
716 * XFS_IOC_FSGROWFSRT
717 */
718
719
720 int
721 xfs_growfs_data(
722 xfs_mount_t *mp,
723 xfs_growfs_data_t *in)
724 {
725 int error;
726
727 if (!capable(CAP_SYS_ADMIN))
728 return -EPERM;
729 if (!mutex_trylock(&mp->m_growlock))
730 return -EWOULDBLOCK;
731 error = xfs_growfs_data_private(mp, in);
732 /*
733 * Increment the generation unconditionally, the error could be from
734 * updating the secondary superblocks, in which case the new size
735 * is live already.
736 */
737 mp->m_generation++;
738 mutex_unlock(&mp->m_growlock);
739 return error;
740 }
741
742 int
743 xfs_growfs_log(
744 xfs_mount_t *mp,
745 xfs_growfs_log_t *in)
746 {
747 int error;
748
749 if (!capable(CAP_SYS_ADMIN))
750 return -EPERM;
751 if (!mutex_trylock(&mp->m_growlock))
752 return -EWOULDBLOCK;
753 error = xfs_growfs_log_private(mp, in);
754 mutex_unlock(&mp->m_growlock);
755 return error;
756 }
757
758 /*
759 * exported through ioctl XFS_IOC_FSCOUNTS
760 */
761
762 int
763 xfs_fs_counts(
764 xfs_mount_t *mp,
765 xfs_fsop_counts_t *cnt)
766 {
767 cnt->allocino = percpu_counter_read_positive(&mp->m_icount);
768 cnt->freeino = percpu_counter_read_positive(&mp->m_ifree);
769 cnt->freedata = percpu_counter_read_positive(&mp->m_fdblocks) -
770 mp->m_alloc_set_aside;
771
772 spin_lock(&mp->m_sb_lock);
773 cnt->freertx = mp->m_sb.sb_frextents;
774 spin_unlock(&mp->m_sb_lock);
775 return 0;
776 }
777
778 /*
779 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
780 *
781 * xfs_reserve_blocks is called to set m_resblks
782 * in the in-core mount table. The number of unused reserved blocks
783 * is kept in m_resblks_avail.
784 *
785 * Reserve the requested number of blocks if available. Otherwise return
786 * as many as possible to satisfy the request. The actual number
787 * reserved are returned in outval
788 *
789 * A null inval pointer indicates that only the current reserved blocks
790 * available should be returned no settings are changed.
791 */
792
793 int
794 xfs_reserve_blocks(
795 xfs_mount_t *mp,
796 uint64_t *inval,
797 xfs_fsop_resblks_t *outval)
798 {
799 int64_t lcounter, delta;
800 int64_t fdblks_delta = 0;
801 uint64_t request;
802 int64_t free;
803 int error = 0;
804
805 /* If inval is null, report current values and return */
806 if (inval == (uint64_t *)NULL) {
807 if (!outval)
808 return -EINVAL;
809 outval->resblks = mp->m_resblks;
810 outval->resblks_avail = mp->m_resblks_avail;
811 return 0;
812 }
813
814 request = *inval;
815
816 /*
817 * With per-cpu counters, this becomes an interesting problem. we need
818 * to work out if we are freeing or allocation blocks first, then we can
819 * do the modification as necessary.
820 *
821 * We do this under the m_sb_lock so that if we are near ENOSPC, we will
822 * hold out any changes while we work out what to do. This means that
823 * the amount of free space can change while we do this, so we need to
824 * retry if we end up trying to reserve more space than is available.
825 */
826 spin_lock(&mp->m_sb_lock);
827
828 /*
829 * If our previous reservation was larger than the current value,
830 * then move any unused blocks back to the free pool. Modify the resblks
831 * counters directly since we shouldn't have any problems unreserving
832 * space.
833 */
834 if (mp->m_resblks > request) {
835 lcounter = mp->m_resblks_avail - request;
836 if (lcounter > 0) { /* release unused blocks */
837 fdblks_delta = lcounter;
838 mp->m_resblks_avail -= lcounter;
839 }
840 mp->m_resblks = request;
841 if (fdblks_delta) {
842 spin_unlock(&mp->m_sb_lock);
843 error = xfs_mod_fdblocks(mp, fdblks_delta, 0);
844 spin_lock(&mp->m_sb_lock);
845 }
846
847 goto out;
848 }
849
850 /*
851 * If the request is larger than the current reservation, reserve the
852 * blocks before we update the reserve counters. Sample m_fdblocks and
853 * perform a partial reservation if the request exceeds free space.
854 */
855 error = -ENOSPC;
856 do {
857 free = percpu_counter_sum(&mp->m_fdblocks) -
858 mp->m_alloc_set_aside;
859 if (!free)
860 break;
861
862 delta = request - mp->m_resblks;
863 lcounter = free - delta;
864 if (lcounter < 0)
865 /* We can't satisfy the request, just get what we can */
866 fdblks_delta = free;
867 else
868 fdblks_delta = delta;
869
870 /*
871 * We'll either succeed in getting space from the free block
872 * count or we'll get an ENOSPC. If we get a ENOSPC, it means
873 * things changed while we were calculating fdblks_delta and so
874 * we should try again to see if there is anything left to
875 * reserve.
876 *
877 * Don't set the reserved flag here - we don't want to reserve
878 * the extra reserve blocks from the reserve.....
879 */
880 spin_unlock(&mp->m_sb_lock);
881 error = xfs_mod_fdblocks(mp, -fdblks_delta, 0);
882 spin_lock(&mp->m_sb_lock);
883 } while (error == -ENOSPC);
884
885 /*
886 * Update the reserve counters if blocks have been successfully
887 * allocated.
888 */
889 if (!error && fdblks_delta) {
890 mp->m_resblks += fdblks_delta;
891 mp->m_resblks_avail += fdblks_delta;
892 }
893
894 out:
895 if (outval) {
896 outval->resblks = mp->m_resblks;
897 outval->resblks_avail = mp->m_resblks_avail;
898 }
899
900 spin_unlock(&mp->m_sb_lock);
901 return error;
902 }
903
904 int
905 xfs_fs_goingdown(
906 xfs_mount_t *mp,
907 uint32_t inflags)
908 {
909 switch (inflags) {
910 case XFS_FSOP_GOING_FLAGS_DEFAULT: {
911 struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
912
913 if (sb && !IS_ERR(sb)) {
914 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
915 thaw_bdev(sb->s_bdev, sb);
916 }
917
918 break;
919 }
920 case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
921 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
922 break;
923 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
924 xfs_force_shutdown(mp,
925 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
926 break;
927 default:
928 return -EINVAL;
929 }
930
931 return 0;
932 }
933
934 /*
935 * Force a shutdown of the filesystem instantly while keeping the filesystem
936 * consistent. We don't do an unmount here; just shutdown the shop, make sure
937 * that absolutely nothing persistent happens to this filesystem after this
938 * point.
939 */
940 void
941 xfs_do_force_shutdown(
942 xfs_mount_t *mp,
943 int flags,
944 char *fname,
945 int lnnum)
946 {
947 int logerror;
948
949 logerror = flags & SHUTDOWN_LOG_IO_ERROR;
950
951 if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
952 xfs_notice(mp,
953 "%s(0x%x) called from line %d of file %s. Return address = 0x%p",
954 __func__, flags, lnnum, fname, __return_address);
955 }
956 /*
957 * No need to duplicate efforts.
958 */
959 if (XFS_FORCED_SHUTDOWN(mp) && !logerror)
960 return;
961
962 /*
963 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
964 * queue up anybody new on the log reservations, and wakes up
965 * everybody who's sleeping on log reservations to tell them
966 * the bad news.
967 */
968 if (xfs_log_force_umount(mp, logerror))
969 return;
970
971 if (flags & SHUTDOWN_CORRUPT_INCORE) {
972 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_CORRUPT,
973 "Corruption of in-memory data detected. Shutting down filesystem");
974 if (XFS_ERRLEVEL_HIGH <= xfs_error_level)
975 xfs_stack_trace();
976 } else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
977 if (logerror) {
978 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_LOGERROR,
979 "Log I/O Error Detected. Shutting down filesystem");
980 } else if (flags & SHUTDOWN_DEVICE_REQ) {
981 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
982 "All device paths lost. Shutting down filesystem");
983 } else if (!(flags & SHUTDOWN_REMOTE_REQ)) {
984 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
985 "I/O Error Detected. Shutting down filesystem");
986 }
987 }
988 if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
989 xfs_alert(mp,
990 "Please umount the filesystem and rectify the problem(s)");
991 }
992 }
993
994 /*
995 * Reserve free space for per-AG metadata.
996 */
997 int
998 xfs_fs_reserve_ag_blocks(
999 struct xfs_mount *mp)
1000 {
1001 xfs_agnumber_t agno;
1002 struct xfs_perag *pag;
1003 int error = 0;
1004 int err2;
1005
1006 for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
1007 pag = xfs_perag_get(mp, agno);
1008 err2 = xfs_ag_resv_init(pag);
1009 xfs_perag_put(pag);
1010 if (err2 && !error)
1011 error = err2;
1012 }
1013
1014 if (error && error != -ENOSPC) {
1015 xfs_warn(mp,
1016 "Error %d reserving per-AG metadata reserve pool.", error);
1017 xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
1018 }
1019
1020 return error;
1021 }
1022
1023 /*
1024 * Free space reserved for per-AG metadata.
1025 */
1026 int
1027 xfs_fs_unreserve_ag_blocks(
1028 struct xfs_mount *mp)
1029 {
1030 xfs_agnumber_t agno;
1031 struct xfs_perag *pag;
1032 int error = 0;
1033 int err2;
1034
1035 for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
1036 pag = xfs_perag_get(mp, agno);
1037 err2 = xfs_ag_resv_free(pag);
1038 xfs_perag_put(pag);
1039 if (err2 && !error)
1040 error = err2;
1041 }
1042
1043 if (error)
1044 xfs_warn(mp,
1045 "Error %d freeing per-AG metadata reserve pool.", error);
1046
1047 return error;
1048 }
Linux with added FPC-III support