xfs: make buffer read verication an IO completion function
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / xfs / xfs_fsops.c
CommitLineData
1da177e4 1/*
7b718769
NS
2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
1da177e4 4 *
7b718769
NS
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
1da177e4
LT
7 * published by the Free Software Foundation.
8 *
7b718769
NS
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.
1da177e4 13 *
7b718769
NS
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
1da177e4 17 */
1da177e4 18#include "xfs.h"
a844f451 19#include "xfs_fs.h"
1da177e4 20#include "xfs_types.h"
1da177e4
LT
21#include "xfs_log.h"
22#include "xfs_trans.h"
23#include "xfs_sb.h"
a844f451 24#include "xfs_ag.h"
1da177e4 25#include "xfs_mount.h"
1da177e4 26#include "xfs_bmap_btree.h"
a844f451 27#include "xfs_alloc_btree.h"
1da177e4 28#include "xfs_ialloc_btree.h"
a844f451
NS
29#include "xfs_dinode.h"
30#include "xfs_inode.h"
31#include "xfs_inode_item.h"
1da177e4
LT
32#include "xfs_btree.h"
33#include "xfs_error.h"
34#include "xfs_alloc.h"
35#include "xfs_ialloc.h"
36#include "xfs_fsops.h"
37#include "xfs_itable.h"
1da177e4
LT
38#include "xfs_trans_space.h"
39#include "xfs_rtalloc.h"
2a82b8be 40#include "xfs_filestream.h"
0b1b213f 41#include "xfs_trace.h"
1da177e4
LT
42
43/*
44 * File system operations
45 */
46
47int
48xfs_fs_geometry(
49 xfs_mount_t *mp,
50 xfs_fsop_geom_t *geo,
51 int new_version)
52{
c4d0c3b0
DR
53
54 memset(geo, 0, sizeof(*geo));
55
1da177e4
LT
56 geo->blocksize = mp->m_sb.sb_blocksize;
57 geo->rtextsize = mp->m_sb.sb_rextsize;
58 geo->agblocks = mp->m_sb.sb_agblocks;
59 geo->agcount = mp->m_sb.sb_agcount;
60 geo->logblocks = mp->m_sb.sb_logblocks;
61 geo->sectsize = mp->m_sb.sb_sectsize;
62 geo->inodesize = mp->m_sb.sb_inodesize;
63 geo->imaxpct = mp->m_sb.sb_imax_pct;
64 geo->datablocks = mp->m_sb.sb_dblocks;
65 geo->rtblocks = mp->m_sb.sb_rblocks;
66 geo->rtextents = mp->m_sb.sb_rextents;
67 geo->logstart = mp->m_sb.sb_logstart;
68 ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid));
69 memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid));
70 if (new_version >= 2) {
71 geo->sunit = mp->m_sb.sb_unit;
72 geo->swidth = mp->m_sb.sb_width;
73 }
74 if (new_version >= 3) {
75 geo->version = XFS_FSOP_GEOM_VERSION;
76 geo->flags =
62118709 77 (xfs_sb_version_hasattr(&mp->m_sb) ?
1da177e4 78 XFS_FSOP_GEOM_FLAGS_ATTR : 0) |
62118709 79 (xfs_sb_version_hasnlink(&mp->m_sb) ?
1da177e4 80 XFS_FSOP_GEOM_FLAGS_NLINK : 0) |
62118709 81 (xfs_sb_version_hasquota(&mp->m_sb) ?
1da177e4 82 XFS_FSOP_GEOM_FLAGS_QUOTA : 0) |
62118709 83 (xfs_sb_version_hasalign(&mp->m_sb) ?
1da177e4 84 XFS_FSOP_GEOM_FLAGS_IALIGN : 0) |
62118709 85 (xfs_sb_version_hasdalign(&mp->m_sb) ?
1da177e4 86 XFS_FSOP_GEOM_FLAGS_DALIGN : 0) |
62118709 87 (xfs_sb_version_hasshared(&mp->m_sb) ?
1da177e4 88 XFS_FSOP_GEOM_FLAGS_SHARED : 0) |
62118709 89 (xfs_sb_version_hasextflgbit(&mp->m_sb) ?
1da177e4 90 XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) |
62118709 91 (xfs_sb_version_hasdirv2(&mp->m_sb) ?
1da177e4 92 XFS_FSOP_GEOM_FLAGS_DIRV2 : 0) |
62118709 93 (xfs_sb_version_hassector(&mp->m_sb) ?
d8cc890d 94 XFS_FSOP_GEOM_FLAGS_SECTOR : 0) |
189f4bf2
BN
95 (xfs_sb_version_hasasciici(&mp->m_sb) ?
96 XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) |
92821e2b
DC
97 (xfs_sb_version_haslazysbcount(&mp->m_sb) ?
98 XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) |
62118709 99 (xfs_sb_version_hasattr2(&mp->m_sb) ?
69a58a43
ES
100 XFS_FSOP_GEOM_FLAGS_ATTR2 : 0) |
101 (xfs_sb_version_hasprojid32bit(&mp->m_sb) ?
102 XFS_FSOP_GEOM_FLAGS_PROJID32 : 0);
62118709 103 geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
1da177e4
LT
104 mp->m_sb.sb_logsectsize : BBSIZE;
105 geo->rtsectsize = mp->m_sb.sb_blocksize;
106 geo->dirblocksize = mp->m_dirblksize;
107 }
108 if (new_version >= 4) {
109 geo->flags |=
62118709 110 (xfs_sb_version_haslogv2(&mp->m_sb) ?
1da177e4
LT
111 XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
112 geo->logsunit = mp->m_sb.sb_logsunit;
113 }
114 return 0;
115}
116
fd23683c
DC
117static struct xfs_buf *
118xfs_growfs_get_hdr_buf(
119 struct xfs_mount *mp,
120 xfs_daddr_t blkno,
121 size_t numblks,
122 int flags)
123{
124 struct xfs_buf *bp;
125
126 bp = xfs_buf_get_uncached(mp->m_ddev_targp, numblks, flags);
127 if (!bp)
128 return NULL;
129
130 xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
131 bp->b_bn = blkno;
132 bp->b_maps[0].bm_bn = blkno;
133
134 return bp;
135}
136
1da177e4
LT
137static int
138xfs_growfs_data_private(
139 xfs_mount_t *mp, /* mount point for filesystem */
140 xfs_growfs_data_t *in) /* growfs data input struct */
141{
142 xfs_agf_t *agf;
de497688 143 struct xfs_agfl *agfl;
1da177e4
LT
144 xfs_agi_t *agi;
145 xfs_agnumber_t agno;
146 xfs_extlen_t agsize;
147 xfs_extlen_t tmpsize;
148 xfs_alloc_rec_t *arec;
1da177e4
LT
149 xfs_buf_t *bp;
150 int bucket;
151 int dpct;
152 int error;
153 xfs_agnumber_t nagcount;
154 xfs_agnumber_t nagimax = 0;
155 xfs_rfsblock_t nb, nb_mod;
156 xfs_rfsblock_t new;
157 xfs_rfsblock_t nfree;
158 xfs_agnumber_t oagcount;
159 int pct;
1da177e4
LT
160 xfs_trans_t *tp;
161
162 nb = in->newblocks;
163 pct = in->imaxpct;
164 if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
165 return XFS_ERROR(EINVAL);
4cc929ee
NS
166 if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
167 return error;
1da177e4 168 dpct = pct - mp->m_sb.sb_imax_pct;
e70b73f8 169 bp = xfs_buf_read_uncached(mp->m_ddev_targp,
1922c949 170 XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
c3f8fc73 171 XFS_FSS_TO_BB(mp, 1), 0, NULL);
1922c949
DC
172 if (!bp)
173 return EIO;
1da177e4
LT
174 xfs_buf_relse(bp);
175
176 new = nb; /* use new as a temporary here */
177 nb_mod = do_div(new, mp->m_sb.sb_agblocks);
178 nagcount = new + (nb_mod != 0);
179 if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
180 nagcount--;
e6da7c9f 181 nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
1da177e4
LT
182 if (nb < mp->m_sb.sb_dblocks)
183 return XFS_ERROR(EINVAL);
184 }
185 new = nb - mp->m_sb.sb_dblocks;
186 oagcount = mp->m_sb.sb_agcount;
0cc6eee1 187
1c1c6ebc
DC
188 /* allocate the new per-ag structures */
189 if (nagcount > oagcount) {
190 error = xfs_initialize_perag(mp, nagcount, &nagimax);
191 if (error)
192 return error;
1da177e4 193 }
1c1c6ebc 194
1da177e4 195 tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
84e1e99f 196 tp->t_flags |= XFS_TRANS_RESERVE;
1da177e4
LT
197 if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp),
198 XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) {
199 xfs_trans_cancel(tp, 0);
200 return error;
201 }
202
1c1c6ebc
DC
203 /*
204 * Write new AG headers to disk. Non-transactional, but written
205 * synchronously so they are completed prior to the growfs transaction
206 * being logged.
207 */
1da177e4
LT
208 nfree = 0;
209 for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
210 /*
de497688 211 * AG freespace header block
1da177e4 212 */
fd23683c
DC
213 bp = xfs_growfs_get_hdr_buf(mp,
214 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
215 XFS_FSS_TO_BB(mp, 1), 0);
b522950f
CS
216 if (!bp) {
217 error = ENOMEM;
218 goto error0;
219 }
fd23683c 220
1da177e4 221 agf = XFS_BUF_TO_AGF(bp);
16259e7d
CH
222 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
223 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
224 agf->agf_seqno = cpu_to_be32(agno);
1da177e4
LT
225 if (agno == nagcount - 1)
226 agsize =
227 nb -
228 (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
229 else
230 agsize = mp->m_sb.sb_agblocks;
16259e7d
CH
231 agf->agf_length = cpu_to_be32(agsize);
232 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
233 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
234 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
235 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
1da177e4 236 agf->agf_flfirst = 0;
16259e7d 237 agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
1da177e4
LT
238 agf->agf_flcount = 0;
239 tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
16259e7d
CH
240 agf->agf_freeblks = cpu_to_be32(tmpsize);
241 agf->agf_longest = cpu_to_be32(tmpsize);
c2b006c1
CH
242 error = xfs_bwrite(bp);
243 xfs_buf_relse(bp);
244 if (error)
1da177e4 245 goto error0;
c2b006c1 246
de497688
DC
247 /*
248 * AG freelist header block
249 */
250 bp = xfs_growfs_get_hdr_buf(mp,
251 XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
252 XFS_FSS_TO_BB(mp, 1), 0);
253 if (!bp) {
254 error = ENOMEM;
255 goto error0;
256 }
257
258 agfl = XFS_BUF_TO_AGFL(bp);
259 for (bucket = 0; bucket < XFS_AGFL_SIZE(mp); bucket++)
260 agfl->agfl_bno[bucket] = cpu_to_be32(NULLAGBLOCK);
261
262 error = xfs_bwrite(bp);
263 xfs_buf_relse(bp);
264 if (error)
265 goto error0;
266
1da177e4
LT
267 /*
268 * AG inode header block
269 */
fd23683c
DC
270 bp = xfs_growfs_get_hdr_buf(mp,
271 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
272 XFS_FSS_TO_BB(mp, 1), 0);
b522950f
CS
273 if (!bp) {
274 error = ENOMEM;
275 goto error0;
276 }
fd23683c 277
1da177e4 278 agi = XFS_BUF_TO_AGI(bp);
16259e7d
CH
279 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
280 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
281 agi->agi_seqno = cpu_to_be32(agno);
282 agi->agi_length = cpu_to_be32(agsize);
1da177e4 283 agi->agi_count = 0;
16259e7d
CH
284 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
285 agi->agi_level = cpu_to_be32(1);
1da177e4 286 agi->agi_freecount = 0;
16259e7d
CH
287 agi->agi_newino = cpu_to_be32(NULLAGINO);
288 agi->agi_dirino = cpu_to_be32(NULLAGINO);
1da177e4 289 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
16259e7d 290 agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
c2b006c1
CH
291 error = xfs_bwrite(bp);
292 xfs_buf_relse(bp);
293 if (error)
1da177e4 294 goto error0;
c2b006c1 295
1da177e4
LT
296 /*
297 * BNO btree root block
298 */
fd23683c
DC
299 bp = xfs_growfs_get_hdr_buf(mp,
300 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
301 BTOBB(mp->m_sb.sb_blocksize), 0);
302
b522950f
CS
303 if (!bp) {
304 error = ENOMEM;
305 goto error0;
306 }
b64f3a39 307
fd23683c 308 xfs_btree_init_block(mp, bp, XFS_ABTB_MAGIC, 0, 1, 0);
b64f3a39 309 arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
16259e7d
CH
310 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
311 arec->ar_blockcount = cpu_to_be32(
312 agsize - be32_to_cpu(arec->ar_startblock));
b64f3a39 313
c2b006c1
CH
314 error = xfs_bwrite(bp);
315 xfs_buf_relse(bp);
316 if (error)
1da177e4 317 goto error0;
c2b006c1 318
1da177e4
LT
319 /*
320 * CNT btree root block
321 */
fd23683c
DC
322 bp = xfs_growfs_get_hdr_buf(mp,
323 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
324 BTOBB(mp->m_sb.sb_blocksize), 0);
b522950f
CS
325 if (!bp) {
326 error = ENOMEM;
327 goto error0;
328 }
b64f3a39 329
fd23683c 330 xfs_btree_init_block(mp, bp, XFS_ABTC_MAGIC, 0, 1, 0);
b64f3a39 331 arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
16259e7d
CH
332 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
333 arec->ar_blockcount = cpu_to_be32(
334 agsize - be32_to_cpu(arec->ar_startblock));
335 nfree += be32_to_cpu(arec->ar_blockcount);
b64f3a39 336
c2b006c1
CH
337 error = xfs_bwrite(bp);
338 xfs_buf_relse(bp);
339 if (error)
1da177e4 340 goto error0;
c2b006c1 341
1da177e4
LT
342 /*
343 * INO btree root block
344 */
fd23683c
DC
345 bp = xfs_growfs_get_hdr_buf(mp,
346 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
347 BTOBB(mp->m_sb.sb_blocksize), 0);
b522950f
CS
348 if (!bp) {
349 error = ENOMEM;
350 goto error0;
351 }
fd23683c 352
b64f3a39
DC
353 xfs_btree_init_block(mp, bp, XFS_IBT_MAGIC, 0, 0, 0);
354
c2b006c1
CH
355 error = xfs_bwrite(bp);
356 xfs_buf_relse(bp);
357 if (error)
1da177e4 358 goto error0;
1da177e4
LT
359 }
360 xfs_trans_agblocks_delta(tp, nfree);
361 /*
362 * There are new blocks in the old last a.g.
363 */
364 if (new) {
365 /*
366 * Change the agi length.
367 */
368 error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
369 if (error) {
370 goto error0;
371 }
372 ASSERT(bp);
373 agi = XFS_BUF_TO_AGI(bp);
413d57c9 374 be32_add_cpu(&agi->agi_length, new);
1da177e4 375 ASSERT(nagcount == oagcount ||
16259e7d 376 be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
1da177e4
LT
377 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
378 /*
379 * Change agf length.
380 */
381 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
382 if (error) {
383 goto error0;
384 }
385 ASSERT(bp);
386 agf = XFS_BUF_TO_AGF(bp);
413d57c9 387 be32_add_cpu(&agf->agf_length, new);
16259e7d
CH
388 ASSERT(be32_to_cpu(agf->agf_length) ==
389 be32_to_cpu(agi->agi_length));
0b1b213f 390
0164af51 391 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
1da177e4
LT
392 /*
393 * Free the new space.
394 */
395 error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
16259e7d 396 be32_to_cpu(agf->agf_length) - new), new);
1da177e4
LT
397 if (error) {
398 goto error0;
399 }
400 }
1c1c6ebc
DC
401
402 /*
403 * Update changed superblock fields transactionally. These are not
404 * seen by the rest of the world until the transaction commit applies
405 * them atomically to the superblock.
406 */
1da177e4
LT
407 if (nagcount > oagcount)
408 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
409 if (nb > mp->m_sb.sb_dblocks)
410 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
411 nb - mp->m_sb.sb_dblocks);
412 if (nfree)
413 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
414 if (dpct)
415 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
1c72bf90 416 error = xfs_trans_commit(tp, 0);
1c1c6ebc 417 if (error)
1da177e4 418 return error;
1c1c6ebc 419
1da177e4
LT
420 /* New allocation groups fully initialized, so update mount struct */
421 if (nagimax)
422 mp->m_maxagi = nagimax;
423 if (mp->m_sb.sb_imax_pct) {
424 __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
425 do_div(icount, 100);
426 mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
427 } else
428 mp->m_maxicount = 0;
055388a3 429 xfs_set_low_space_thresholds(mp);
1c1c6ebc
DC
430
431 /* update secondary superblocks. */
1da177e4 432 for (agno = 1; agno < nagcount; agno++) {
1375cb65
DC
433 error = 0;
434 /*
435 * new secondary superblocks need to be zeroed, not read from
436 * disk as the contents of the new area we are growing into is
437 * completely unknown.
438 */
439 if (agno < oagcount) {
440 error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
1da177e4 441 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
c3f8fc73 442 XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
1375cb65
DC
443 } else {
444 bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
445 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
446 XFS_FSS_TO_BB(mp, 1), 0);
447 if (bp)
448 xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
449 else
450 error = ENOMEM;
451 }
452
1da177e4 453 if (error) {
53487786
DC
454 xfs_warn(mp,
455 "error %d reading secondary superblock for ag %d",
1da177e4
LT
456 error, agno);
457 break;
458 }
2bdf7cd0 459 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
1da177e4
LT
460 /*
461 * If we get an error writing out the alternate superblocks,
462 * just issue a warning and continue. The real work is
463 * already done and committed.
464 */
c2b006c1
CH
465 error = xfs_bwrite(bp);
466 xfs_buf_relse(bp);
467 if (error) {
53487786 468 xfs_warn(mp,
1da177e4
LT
469 "write error %d updating secondary superblock for ag %d",
470 error, agno);
471 break; /* no point in continuing */
472 }
473 }
1375cb65 474 return error;
1da177e4
LT
475
476 error0:
477 xfs_trans_cancel(tp, XFS_TRANS_ABORT);
478 return error;
479}
480
481static int
482xfs_growfs_log_private(
483 xfs_mount_t *mp, /* mount point for filesystem */
484 xfs_growfs_log_t *in) /* growfs log input struct */
485{
486 xfs_extlen_t nb;
487
488 nb = in->newblocks;
489 if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
490 return XFS_ERROR(EINVAL);
491 if (nb == mp->m_sb.sb_logblocks &&
492 in->isint == (mp->m_sb.sb_logstart != 0))
493 return XFS_ERROR(EINVAL);
494 /*
495 * Moving the log is hard, need new interfaces to sync
496 * the log first, hold off all activity while moving it.
497 * Can have shorter or longer log in the same space,
498 * or transform internal to external log or vice versa.
499 */
500 return XFS_ERROR(ENOSYS);
501}
502
503/*
504 * protected versions of growfs function acquire and release locks on the mount
505 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
506 * XFS_IOC_FSGROWFSRT
507 */
508
509
510int
511xfs_growfs_data(
512 xfs_mount_t *mp,
513 xfs_growfs_data_t *in)
514{
515 int error;
743bb465 516
517 if (!capable(CAP_SYS_ADMIN))
518 return XFS_ERROR(EPERM);
cc92e7ac 519 if (!mutex_trylock(&mp->m_growlock))
1da177e4
LT
520 return XFS_ERROR(EWOULDBLOCK);
521 error = xfs_growfs_data_private(mp, in);
cc92e7ac 522 mutex_unlock(&mp->m_growlock);
1da177e4
LT
523 return error;
524}
525
526int
527xfs_growfs_log(
528 xfs_mount_t *mp,
529 xfs_growfs_log_t *in)
530{
531 int error;
743bb465 532
533 if (!capable(CAP_SYS_ADMIN))
534 return XFS_ERROR(EPERM);
cc92e7ac 535 if (!mutex_trylock(&mp->m_growlock))
1da177e4
LT
536 return XFS_ERROR(EWOULDBLOCK);
537 error = xfs_growfs_log_private(mp, in);
cc92e7ac 538 mutex_unlock(&mp->m_growlock);
1da177e4
LT
539 return error;
540}
541
542/*
543 * exported through ioctl XFS_IOC_FSCOUNTS
544 */
545
546int
547xfs_fs_counts(
548 xfs_mount_t *mp,
549 xfs_fsop_counts_t *cnt)
550{
d4d90b57 551 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
3685c2a1 552 spin_lock(&mp->m_sb_lock);
4be536de 553 cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1da177e4
LT
554 cnt->freertx = mp->m_sb.sb_frextents;
555 cnt->freeino = mp->m_sb.sb_ifree;
556 cnt->allocino = mp->m_sb.sb_icount;
3685c2a1 557 spin_unlock(&mp->m_sb_lock);
1da177e4
LT
558 return 0;
559}
560
561/*
562 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
563 *
564 * xfs_reserve_blocks is called to set m_resblks
565 * in the in-core mount table. The number of unused reserved blocks
c41564b5 566 * is kept in m_resblks_avail.
1da177e4
LT
567 *
568 * Reserve the requested number of blocks if available. Otherwise return
569 * as many as possible to satisfy the request. The actual number
570 * reserved are returned in outval
571 *
572 * A null inval pointer indicates that only the current reserved blocks
573 * available should be returned no settings are changed.
574 */
575
576int
577xfs_reserve_blocks(
578 xfs_mount_t *mp,
579 __uint64_t *inval,
580 xfs_fsop_resblks_t *outval)
581{
dbcabad1 582 __int64_t lcounter, delta, fdblks_delta;
1da177e4 583 __uint64_t request;
1da177e4
LT
584
585 /* If inval is null, report current values and return */
1da177e4 586 if (inval == (__uint64_t *)NULL) {
84e1e99f
DC
587 if (!outval)
588 return EINVAL;
1da177e4
LT
589 outval->resblks = mp->m_resblks;
590 outval->resblks_avail = mp->m_resblks_avail;
014c2544 591 return 0;
1da177e4
LT
592 }
593
594 request = *inval;
dbcabad1
DC
595
596 /*
597 * With per-cpu counters, this becomes an interesting
598 * problem. we needto work out if we are freeing or allocation
599 * blocks first, then we can do the modification as necessary.
600 *
3685c2a1 601 * We do this under the m_sb_lock so that if we are near
dbcabad1
DC
602 * ENOSPC, we will hold out any changes while we work out
603 * what to do. This means that the amount of free space can
604 * change while we do this, so we need to retry if we end up
605 * trying to reserve more space than is available.
606 *
607 * We also use the xfs_mod_incore_sb() interface so that we
608 * don't have to care about whether per cpu counter are
609 * enabled, disabled or even compiled in....
610 */
611retry:
3685c2a1 612 spin_lock(&mp->m_sb_lock);
d4d90b57 613 xfs_icsb_sync_counters_locked(mp, 0);
1da177e4
LT
614
615 /*
616 * If our previous reservation was larger than the current value,
617 * then move any unused blocks back to the free pool.
618 */
dbcabad1 619 fdblks_delta = 0;
1da177e4
LT
620 if (mp->m_resblks > request) {
621 lcounter = mp->m_resblks_avail - request;
622 if (lcounter > 0) { /* release unused blocks */
dbcabad1 623 fdblks_delta = lcounter;
1da177e4
LT
624 mp->m_resblks_avail -= lcounter;
625 }
626 mp->m_resblks = request;
627 } else {
4be536de
DC
628 __int64_t free;
629
630 free = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
dbcabad1
DC
631 if (!free)
632 goto out; /* ENOSPC and fdblks_delta = 0 */
633
1da177e4 634 delta = request - mp->m_resblks;
4be536de 635 lcounter = free - delta;
1da177e4
LT
636 if (lcounter < 0) {
637 /* We can't satisfy the request, just get what we can */
4be536de
DC
638 mp->m_resblks += free;
639 mp->m_resblks_avail += free;
dbcabad1 640 fdblks_delta = -free;
1da177e4 641 } else {
dbcabad1 642 fdblks_delta = -delta;
1da177e4
LT
643 mp->m_resblks = request;
644 mp->m_resblks_avail += delta;
645 }
646 }
dbcabad1 647out:
84e1e99f
DC
648 if (outval) {
649 outval->resblks = mp->m_resblks;
650 outval->resblks_avail = mp->m_resblks_avail;
651 }
3685c2a1 652 spin_unlock(&mp->m_sb_lock);
dbcabad1
DC
653
654 if (fdblks_delta) {
655 /*
656 * If we are putting blocks back here, m_resblks_avail is
9da096fd 657 * already at its max so this will put it in the free pool.
dbcabad1
DC
658 *
659 * If we need space, we'll either succeed in getting it
660 * from the free block count or we'll get an enospc. If
661 * we get a ENOSPC, it means things changed while we were
662 * calculating fdblks_delta and so we should try again to
663 * see if there is anything left to reserve.
664 *
665 * Don't set the reserved flag here - we don't want to reserve
666 * the extra reserve blocks from the reserve.....
667 */
668 int error;
96540c78
CH
669 error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
670 fdblks_delta, 0);
dbcabad1
DC
671 if (error == ENOSPC)
672 goto retry;
673 }
014c2544 674 return 0;
1da177e4
LT
675}
676
1a387d3b
DC
677/*
678 * Dump a transaction into the log that contains no real change. This is needed
679 * to be able to make the log dirty or stamp the current tail LSN into the log
680 * during the covering operation.
681 *
682 * We cannot use an inode here for this - that will push dirty state back up
683 * into the VFS and then periodic inode flushing will prevent log covering from
c58efdb4
DC
684 * making progress. Hence we log a field in the superblock instead and use a
685 * synchronous transaction to ensure the superblock is immediately unpinned
686 * and can be written back.
1a387d3b 687 */
c4be0c1d 688int
421ad134 689xfs_fs_log_dummy(
c58efdb4 690 xfs_mount_t *mp)
e13a73f0 691{
421ad134 692 xfs_trans_t *tp;
c4be0c1d 693 int error;
e13a73f0 694
80641dc6 695 tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP);
1a387d3b
DC
696 error = xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0,
697 XFS_DEFAULT_LOG_COUNT);
c4be0c1d 698 if (error) {
e13a73f0 699 xfs_trans_cancel(tp, 0);
c4be0c1d 700 return error;
e13a73f0
CH
701 }
702
1a387d3b
DC
703 /* log the UUID because it is an unchanging field */
704 xfs_mod_sb(tp, XFS_SB_UUID);
c58efdb4 705 xfs_trans_set_sync(tp);
1a387d3b 706 return xfs_trans_commit(tp, 0);
e13a73f0
CH
707}
708
1da177e4
LT
709int
710xfs_fs_goingdown(
711 xfs_mount_t *mp,
712 __uint32_t inflags)
713{
714 switch (inflags) {
715 case XFS_FSOP_GOING_FLAGS_DEFAULT: {
b267ce99 716 struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
1da177e4 717
f33c6797 718 if (sb && !IS_ERR(sb)) {
7d04a335 719 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
1da177e4
LT
720 thaw_bdev(sb->s_bdev, sb);
721 }
189f4bf2 722
1da177e4
LT
723 break;
724 }
725 case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
7d04a335 726 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
1da177e4
LT
727 break;
728 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
7d04a335
NS
729 xfs_force_shutdown(mp,
730 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
1da177e4
LT
731 break;
732 default:
733 return XFS_ERROR(EINVAL);
734 }
735
736 return 0;
737}
2af51f3a
DC
738
739/*
740 * Force a shutdown of the filesystem instantly while keeping the filesystem
741 * consistent. We don't do an unmount here; just shutdown the shop, make sure
742 * that absolutely nothing persistent happens to this filesystem after this
743 * point.
744 */
745void
746xfs_do_force_shutdown(
747 xfs_mount_t *mp,
748 int flags,
749 char *fname,
750 int lnnum)
751{
752 int logerror;
753
754 logerror = flags & SHUTDOWN_LOG_IO_ERROR;
755
756 if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
757 xfs_notice(mp,
758 "%s(0x%x) called from line %d of file %s. Return address = 0x%p",
759 __func__, flags, lnnum, fname, __return_address);
760 }
761 /*
762 * No need to duplicate efforts.
763 */
764 if (XFS_FORCED_SHUTDOWN(mp) && !logerror)
765 return;
766
767 /*
768 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
769 * queue up anybody new on the log reservations, and wakes up
770 * everybody who's sleeping on log reservations to tell them
771 * the bad news.
772 */
773 if (xfs_log_force_umount(mp, logerror))
774 return;
775
776 if (flags & SHUTDOWN_CORRUPT_INCORE) {
777 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_CORRUPT,
778 "Corruption of in-memory data detected. Shutting down filesystem");
779 if (XFS_ERRLEVEL_HIGH <= xfs_error_level)
780 xfs_stack_trace();
781 } else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
782 if (logerror) {
783 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_LOGERROR,
784 "Log I/O Error Detected. Shutting down filesystem");
785 } else if (flags & SHUTDOWN_DEVICE_REQ) {
786 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
787 "All device paths lost. Shutting down filesystem");
788 } else if (!(flags & SHUTDOWN_REMOTE_REQ)) {
789 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
790 "I/O Error Detected. Shutting down filesystem");
791 }
792 }
793 if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
794 xfs_alert(mp,
795 "Please umount the filesystem and rectify the problem(s)");
796 }
797}