?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
[linux-2.6.git] / fs / xfs / xfs_super.c
blobf317488263dd975eb85eb44d9746abfb974ef46f
1 /*
2 * Copyright (c) 2000-2006 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
19 #include "xfs.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_inum.h"
25 #include "xfs_sb.h"
26 #include "xfs_ag.h"
27 #include "xfs_mount.h"
28 #include "xfs_da_format.h"
29 #include "xfs_inode.h"
30 #include "xfs_btree.h"
31 #include "xfs_bmap.h"
32 #include "xfs_alloc.h"
33 #include "xfs_error.h"
34 #include "xfs_fsops.h"
35 #include "xfs_trans.h"
36 #include "xfs_buf_item.h"
37 #include "xfs_log.h"
38 #include "xfs_log_priv.h"
39 #include "xfs_da_btree.h"
40 #include "xfs_dir2.h"
41 #include "xfs_extfree_item.h"
42 #include "xfs_mru_cache.h"
43 #include "xfs_inode_item.h"
44 #include "xfs_icache.h"
45 #include "xfs_trace.h"
46 #include "xfs_icreate_item.h"
47 #include "xfs_dinode.h"
48 #include "xfs_filestream.h"
49 #include "xfs_quota.h"
50
51 #include <linux/namei.h>
52 #include <linux/init.h>
53 #include <linux/slab.h>
54 #include <linux/mount.h>
55 #include <linux/mempool.h>
56 #include <linux/writeback.h>
57 #include <linux/kthread.h>
58 #include <linux/freezer.h>
59 #include <linux/parser.h>
60
61 static const struct super_operations xfs_super_operations;
62 static kmem_zone_t *xfs_ioend_zone;
63 mempool_t *xfs_ioend_pool;
64
65 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
66 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
67 #define MNTOPT_LOGDEV "logdev" /* log device */
68 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
69 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
70 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
71 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
72 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
73 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
74 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
75 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
76 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
77 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
78 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
79 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
80 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
81 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
82 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
83 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
84 * unwritten extent conversion */
85 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
86 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
87 #define MNTOPT_32BITINODE "inode32" /* inode allocation limited to
88 * XFS_MAXINUMBER_32 */
89 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
90 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
91 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
92 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
93 * in stat(). */
94 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
95 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
96 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
97 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
98 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
99 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
100 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
101 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
102 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
103 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
104 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
105 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
106 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
107 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
108 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
109 #define MNTOPT_DELAYLOG "delaylog" /* Delayed logging enabled */
110 #define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed logging disabled */
111 #define MNTOPT_DISCARD "discard" /* Discard unused blocks */
112 #define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */
113
114 /*
115 * Table driven mount option parser.
116 *
117 * Currently only used for remount, but it will be used for mount
118 * in the future, too.
119 */
120 enum {
121 Opt_barrier,
122 Opt_nobarrier,
123 Opt_inode64,
124 Opt_inode32,
125 Opt_err
126 };
127
128 static const match_table_t tokens = {
129 {Opt_barrier, "barrier"},
130 {Opt_nobarrier, "nobarrier"},
131 {Opt_inode64, "inode64"},
132 {Opt_inode32, "inode32"},
133 {Opt_err, NULL}
134 };
135
136
137 STATIC unsigned long
138 suffix_kstrtoint(char *s, unsigned int base, int *res)
139 {
140 int last, shift_left_factor = 0, _res;
141 char *value = s;
142
143 last = strlen(value) - 1;
144 if (value[last] == 'K' || value[last] == 'k') {
145 shift_left_factor = 10;
146 value[last] = '\0';
147 }
148 if (value[last] == 'M' || value[last] == 'm') {
149 shift_left_factor = 20;
150 value[last] = '\0';
151 }
152 if (value[last] == 'G' || value[last] == 'g') {
153 shift_left_factor = 30;
154 value[last] = '\0';
155 }
156
157 if (kstrtoint(s, base, &_res))
158 return -EINVAL;
159 *res = _res << shift_left_factor;
160 return 0;
161 }
162
163 /*
164 * This function fills in xfs_mount_t fields based on mount args.
165 * Note: the superblock has _not_ yet been read in.
166 *
167 * Note that this function leaks the various device name allocations on
168 * failure. The caller takes care of them.
169 */
170 STATIC int
171 xfs_parseargs(
172 struct xfs_mount *mp,
173 char *options)
174 {
175 struct super_block *sb = mp->m_super;
176 char *this_char, *value;
177 int dsunit = 0;
178 int dswidth = 0;
179 int iosize = 0;
180 __uint8_t iosizelog = 0;
181
182 /*
183 * set up the mount name first so all the errors will refer to the
184 * correct device.
185 */
186 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
187 if (!mp->m_fsname)
188 return ENOMEM;
189 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
190
191 /*
192 * Copy binary VFS mount flags we are interested in.
193 */
194 if (sb->s_flags & MS_RDONLY)
195 mp->m_flags |= XFS_MOUNT_RDONLY;
196 if (sb->s_flags & MS_DIRSYNC)
197 mp->m_flags |= XFS_MOUNT_DIRSYNC;
198 if (sb->s_flags & MS_SYNCHRONOUS)
199 mp->m_flags |= XFS_MOUNT_WSYNC;
200
201 /*
202 * Set some default flags that could be cleared by the mount option
203 * parsing.
204 */
205 mp->m_flags |= XFS_MOUNT_BARRIER;
206 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
207 #if !XFS_BIG_INUMS
208 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
209 #endif
210
211 /*
212 * These can be overridden by the mount option parsing.
213 */
214 mp->m_logbufs = -1;
215 mp->m_logbsize = -1;
216
217 if (!options)
218 goto done;
219
220 while ((this_char = strsep(&options, ",")) != NULL) {
221 if (!*this_char)
222 continue;
223 if ((value = strchr(this_char, '=')) != NULL)
224 *value++ = 0;
225
226 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
227 if (!value || !*value) {
228 xfs_warn(mp, "%s option requires an argument",
229 this_char);
230 return EINVAL;
231 }
232 if (kstrtoint(value, 10, &mp->m_logbufs))
233 return EINVAL;
234 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
235 if (!value || !*value) {
236 xfs_warn(mp, "%s option requires an argument",
237 this_char);
238 return EINVAL;
239 }
240 if (suffix_kstrtoint(value, 10, &mp->m_logbsize))
241 return EINVAL;
242 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
243 if (!value || !*value) {
244 xfs_warn(mp, "%s option requires an argument",
245 this_char);
246 return EINVAL;
247 }
248 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
249 if (!mp->m_logname)
250 return ENOMEM;
251 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
252 xfs_warn(mp, "%s option not allowed on this system",
253 this_char);
254 return EINVAL;
255 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
256 if (!value || !*value) {
257 xfs_warn(mp, "%s option requires an argument",
258 this_char);
259 return EINVAL;
260 }
261 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
262 if (!mp->m_rtname)
263 return ENOMEM;
264 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
265 if (!value || !*value) {
266 xfs_warn(mp, "%s option requires an argument",
267 this_char);
268 return EINVAL;
269 }
270 if (kstrtoint(value, 10, &iosize))
271 return EINVAL;
272 iosizelog = ffs(iosize) - 1;
273 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
274 if (!value || !*value) {
275 xfs_warn(mp, "%s option requires an argument",
276 this_char);
277 return EINVAL;
278 }
279 if (suffix_kstrtoint(value, 10, &iosize))
280 return EINVAL;
281 iosizelog = ffs(iosize) - 1;
282 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
283 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
284 mp->m_flags |= XFS_MOUNT_GRPID;
285 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
286 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
287 mp->m_flags &= ~XFS_MOUNT_GRPID;
288 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
289 mp->m_flags |= XFS_MOUNT_WSYNC;
290 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
291 mp->m_flags |= XFS_MOUNT_NORECOVERY;
292 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
293 mp->m_flags |= XFS_MOUNT_NOALIGN;
294 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
295 mp->m_flags |= XFS_MOUNT_SWALLOC;
296 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
297 if (!value || !*value) {
298 xfs_warn(mp, "%s option requires an argument",
299 this_char);
300 return EINVAL;
301 }
302 if (kstrtoint(value, 10, &dsunit))
303 return EINVAL;
304 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
305 if (!value || !*value) {
306 xfs_warn(mp, "%s option requires an argument",
307 this_char);
308 return EINVAL;
309 }
310 if (kstrtoint(value, 10, &dswidth))
311 return EINVAL;
312 } else if (!strcmp(this_char, MNTOPT_32BITINODE)) {
313 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
314 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
315 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
316 #if !XFS_BIG_INUMS
317 xfs_warn(mp, "%s option not allowed on this system",
318 this_char);
319 return EINVAL;
320 #endif
321 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
322 mp->m_flags |= XFS_MOUNT_NOUUID;
323 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
324 mp->m_flags |= XFS_MOUNT_BARRIER;
325 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
326 mp->m_flags &= ~XFS_MOUNT_BARRIER;
327 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
328 mp->m_flags |= XFS_MOUNT_IKEEP;
329 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
330 mp->m_flags &= ~XFS_MOUNT_IKEEP;
331 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
332 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
333 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
334 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
335 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
336 mp->m_flags |= XFS_MOUNT_ATTR2;
337 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
338 mp->m_flags &= ~XFS_MOUNT_ATTR2;
339 mp->m_flags |= XFS_MOUNT_NOATTR2;
340 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
341 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
342 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
343 mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT;
344 mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD;
345 mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE;
346 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
347 !strcmp(this_char, MNTOPT_UQUOTA) ||
348 !strcmp(this_char, MNTOPT_USRQUOTA)) {
349 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
350 XFS_UQUOTA_ENFD);
351 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
352 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
353 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
354 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
355 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
356 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
357 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
358 XFS_PQUOTA_ENFD);
359 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
360 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
361 mp->m_qflags &= ~XFS_PQUOTA_ENFD;
362 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
363 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
364 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
365 XFS_GQUOTA_ENFD);
366 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
367 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
368 mp->m_qflags &= ~XFS_GQUOTA_ENFD;
369 } else if (!strcmp(this_char, MNTOPT_DELAYLOG)) {
370 xfs_warn(mp,
371 "delaylog is the default now, option is deprecated.");
372 } else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) {
373 xfs_warn(mp,
374 "nodelaylog support has been removed, option is deprecated.");
375 } else if (!strcmp(this_char, MNTOPT_DISCARD)) {
376 mp->m_flags |= XFS_MOUNT_DISCARD;
377 } else if (!strcmp(this_char, MNTOPT_NODISCARD)) {
378 mp->m_flags &= ~XFS_MOUNT_DISCARD;
379 } else if (!strcmp(this_char, "ihashsize")) {
380 xfs_warn(mp,
381 "ihashsize no longer used, option is deprecated.");
382 } else if (!strcmp(this_char, "osyncisdsync")) {
383 xfs_warn(mp,
384 "osyncisdsync has no effect, option is deprecated.");
385 } else if (!strcmp(this_char, "osyncisosync")) {
386 xfs_warn(mp,
387 "osyncisosync has no effect, option is deprecated.");
388 } else if (!strcmp(this_char, "irixsgid")) {
389 xfs_warn(mp,
390 "irixsgid is now a sysctl(2) variable, option is deprecated.");
391 } else {
392 xfs_warn(mp, "unknown mount option [%s].", this_char);
393 return EINVAL;
394 }
395 }
396
397 /*
398 * no recovery flag requires a read-only mount
399 */
400 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
401 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
402 xfs_warn(mp, "no-recovery mounts must be read-only.");
403 return EINVAL;
404 }
405
406 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
407 xfs_warn(mp,
408 "sunit and swidth options incompatible with the noalign option");
409 return EINVAL;
410 }
411
412 #ifndef CONFIG_XFS_QUOTA
413 if (XFS_IS_QUOTA_RUNNING(mp)) {
414 xfs_warn(mp, "quota support not available in this kernel.");
415 return EINVAL;
416 }
417 #endif
418
419 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
420 xfs_warn(mp, "sunit and swidth must be specified together");
421 return EINVAL;
422 }
423
424 if (dsunit && (dswidth % dsunit != 0)) {
425 xfs_warn(mp,
426 "stripe width (%d) must be a multiple of the stripe unit (%d)",
427 dswidth, dsunit);
428 return EINVAL;
429 }
430
431 done:
432 if (dsunit && !(mp->m_flags & XFS_MOUNT_NOALIGN)) {
433 /*
434 * At this point the superblock has not been read
435 * in, therefore we do not know the block size.
436 * Before the mount call ends we will convert
437 * these to FSBs.
438 */
439 mp->m_dalign = dsunit;
440 mp->m_swidth = dswidth;
441 }
442
443 if (mp->m_logbufs != -1 &&
444 mp->m_logbufs != 0 &&
445 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
446 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
447 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
448 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
449 return XFS_ERROR(EINVAL);
450 }
451 if (mp->m_logbsize != -1 &&
452 mp->m_logbsize != 0 &&
453 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
454 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
455 !is_power_of_2(mp->m_logbsize))) {
456 xfs_warn(mp,
457 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
458 mp->m_logbsize);
459 return XFS_ERROR(EINVAL);
460 }
461
462 if (iosizelog) {
463 if (iosizelog > XFS_MAX_IO_LOG ||
464 iosizelog < XFS_MIN_IO_LOG) {
465 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]",
466 iosizelog, XFS_MIN_IO_LOG,
467 XFS_MAX_IO_LOG);
468 return XFS_ERROR(EINVAL);
469 }
470
471 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
472 mp->m_readio_log = iosizelog;
473 mp->m_writeio_log = iosizelog;
474 }
475
476 return 0;
477 }
478
479 struct proc_xfs_info {
480 int flag;
481 char *str;
482 };
483
484 STATIC int
485 xfs_showargs(
486 struct xfs_mount *mp,
487 struct seq_file *m)
488 {
489 static struct proc_xfs_info xfs_info_set[] = {
490 /* the few simple ones we can get from the mount struct */
491 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
492 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
493 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
494 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
495 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
496 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
497 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
498 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
499 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
500 { XFS_MOUNT_DISCARD, "," MNTOPT_DISCARD },
501 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_32BITINODE },
502 { 0, NULL }
503 };
504 static struct proc_xfs_info xfs_info_unset[] = {
505 /* the few simple ones we can get from the mount struct */
506 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
507 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
508 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
509 { 0, NULL }
510 };
511 struct proc_xfs_info *xfs_infop;
512
513 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
514 if (mp->m_flags & xfs_infop->flag)
515 seq_puts(m, xfs_infop->str);
516 }
517 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
518 if (!(mp->m_flags & xfs_infop->flag))
519 seq_puts(m, xfs_infop->str);
520 }
521
522 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
523 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
524 (int)(1 << mp->m_writeio_log) >> 10);
525
526 if (mp->m_logbufs > 0)
527 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
528 if (mp->m_logbsize > 0)
529 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
530
531 if (mp->m_logname)
532 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
533 if (mp->m_rtname)
534 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
535
536 if (mp->m_dalign > 0)
537 seq_printf(m, "," MNTOPT_SUNIT "=%d",
538 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
539 if (mp->m_swidth > 0)
540 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
541 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
542
543 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
544 seq_puts(m, "," MNTOPT_USRQUOTA);
545 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
546 seq_puts(m, "," MNTOPT_UQUOTANOENF);
547
548 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
549 if (mp->m_qflags & XFS_PQUOTA_ENFD)
550 seq_puts(m, "," MNTOPT_PRJQUOTA);
551 else
552 seq_puts(m, "," MNTOPT_PQUOTANOENF);
553 }
554 if (mp->m_qflags & XFS_GQUOTA_ACCT) {
555 if (mp->m_qflags & XFS_GQUOTA_ENFD)
556 seq_puts(m, "," MNTOPT_GRPQUOTA);
557 else
558 seq_puts(m, "," MNTOPT_GQUOTANOENF);
559 }
560
561 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
562 seq_puts(m, "," MNTOPT_NOQUOTA);
563
564 return 0;
565 }
566 __uint64_t
567 xfs_max_file_offset(
568 unsigned int blockshift)
569 {
570 unsigned int pagefactor = 1;
571 unsigned int bitshift = BITS_PER_LONG - 1;
572
573 /* Figure out maximum filesize, on Linux this can depend on
574 * the filesystem blocksize (on 32 bit platforms).
575 * __block_write_begin does this in an [unsigned] long...
576 * page->index << (PAGE_CACHE_SHIFT - bbits)
577 * So, for page sized blocks (4K on 32 bit platforms),
578 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
579 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
580 * but for smaller blocksizes it is less (bbits = log2 bsize).
581 * Note1: get_block_t takes a long (implicit cast from above)
582 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
583 * can optionally convert the [unsigned] long from above into
584 * an [unsigned] long long.
585 */
586
587 #if BITS_PER_LONG == 32
588 # if defined(CONFIG_LBDAF)
589 ASSERT(sizeof(sector_t) == 8);
590 pagefactor = PAGE_CACHE_SIZE;
591 bitshift = BITS_PER_LONG;
592 # else
593 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
594 # endif
595 #endif
596
597 return (((__uint64_t)pagefactor) << bitshift) - 1;
598 }
599
600 xfs_agnumber_t
601 xfs_set_inode32(struct xfs_mount *mp)
602 {
603 xfs_agnumber_t index = 0;
604 xfs_agnumber_t maxagi = 0;
605 xfs_sb_t *sbp = &mp->m_sb;
606 xfs_agnumber_t max_metadata;
607 xfs_agino_t agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks -1, 0);
608 xfs_ino_t ino = XFS_AGINO_TO_INO(mp, sbp->sb_agcount -1, agino);
609 xfs_perag_t *pag;
610
611 /* Calculate how much should be reserved for inodes to meet
612 * the max inode percentage.
613 */
614 if (mp->m_maxicount) {
615 __uint64_t icount;
616
617 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
618 do_div(icount, 100);
619 icount += sbp->sb_agblocks - 1;
620 do_div(icount, sbp->sb_agblocks);
621 max_metadata = icount;
622 } else {
623 max_metadata = sbp->sb_agcount;
624 }
625
626 for (index = 0; index < sbp->sb_agcount; index++) {
627 ino = XFS_AGINO_TO_INO(mp, index, agino);
628
629 if (ino > XFS_MAXINUMBER_32) {
630 pag = xfs_perag_get(mp, index);
631 pag->pagi_inodeok = 0;
632 pag->pagf_metadata = 0;
633 xfs_perag_put(pag);
634 continue;
635 }
636
637 pag = xfs_perag_get(mp, index);
638 pag->pagi_inodeok = 1;
639 maxagi++;
640 if (index < max_metadata)
641 pag->pagf_metadata = 1;
642 xfs_perag_put(pag);
643 }
644 mp->m_flags |= (XFS_MOUNT_32BITINODES |
645 XFS_MOUNT_SMALL_INUMS);
646
647 return maxagi;
648 }
649
650 xfs_agnumber_t
651 xfs_set_inode64(struct xfs_mount *mp)
652 {
653 xfs_agnumber_t index = 0;
654
655 for (index = 0; index < mp->m_sb.sb_agcount; index++) {
656 struct xfs_perag *pag;
657
658 pag = xfs_perag_get(mp, index);
659 pag->pagi_inodeok = 1;
660 pag->pagf_metadata = 0;
661 xfs_perag_put(pag);
662 }
663
664 /* There is no need for lock protection on m_flags,
665 * the rw_semaphore of the VFS superblock is locked
666 * during mount/umount/remount operations, so this is
667 * enough to avoid concurency on the m_flags field
668 */
669 mp->m_flags &= ~(XFS_MOUNT_32BITINODES |
670 XFS_MOUNT_SMALL_INUMS);
671 return index;
672 }
673
674 STATIC int
675 xfs_blkdev_get(
676 xfs_mount_t *mp,
677 const char *name,
678 struct block_device **bdevp)
679 {
680 int error = 0;
681
682 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
683 mp);
684 if (IS_ERR(*bdevp)) {
685 error = PTR_ERR(*bdevp);
686 xfs_warn(mp, "Invalid device [%s], error=%d\n", name, error);
687 }
688
689 return -error;
690 }
691
692 STATIC void
693 xfs_blkdev_put(
694 struct block_device *bdev)
695 {
696 if (bdev)
697 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
698 }
699
700 void
701 xfs_blkdev_issue_flush(
702 xfs_buftarg_t *buftarg)
703 {
704 blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL);
705 }
706
707 STATIC void
708 xfs_close_devices(
709 struct xfs_mount *mp)
710 {
711 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
712 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
713 xfs_free_buftarg(mp, mp->m_logdev_targp);
714 xfs_blkdev_put(logdev);
715 }
716 if (mp->m_rtdev_targp) {
717 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
718 xfs_free_buftarg(mp, mp->m_rtdev_targp);
719 xfs_blkdev_put(rtdev);
720 }
721 xfs_free_buftarg(mp, mp->m_ddev_targp);
722 }
723
724 /*
725 * The file system configurations are:
726 * (1) device (partition) with data and internal log
727 * (2) logical volume with data and log subvolumes.
728 * (3) logical volume with data, log, and realtime subvolumes.
729 *
730 * We only have to handle opening the log and realtime volumes here if
731 * they are present. The data subvolume has already been opened by
732 * get_sb_bdev() and is stored in sb->s_bdev.
733 */
734 STATIC int
735 xfs_open_devices(
736 struct xfs_mount *mp)
737 {
738 struct block_device *ddev = mp->m_super->s_bdev;
739 struct block_device *logdev = NULL, *rtdev = NULL;
740 int error;
741
742 /*
743 * Open real time and log devices - order is important.
744 */
745 if (mp->m_logname) {
746 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
747 if (error)
748 goto out;
749 }
750
751 if (mp->m_rtname) {
752 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
753 if (error)
754 goto out_close_logdev;
755
756 if (rtdev == ddev || rtdev == logdev) {
757 xfs_warn(mp,
758 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
759 error = EINVAL;
760 goto out_close_rtdev;
761 }
762 }
763
764 /*
765 * Setup xfs_mount buffer target pointers
766 */
767 error = ENOMEM;
768 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, 0, mp->m_fsname);
769 if (!mp->m_ddev_targp)
770 goto out_close_rtdev;
771
772 if (rtdev) {
773 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, 1,
774 mp->m_fsname);
775 if (!mp->m_rtdev_targp)
776 goto out_free_ddev_targ;
777 }
778
779 if (logdev && logdev != ddev) {
780 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, 1,
781 mp->m_fsname);
782 if (!mp->m_logdev_targp)
783 goto out_free_rtdev_targ;
784 } else {
785 mp->m_logdev_targp = mp->m_ddev_targp;
786 }
787
788 return 0;
789
790 out_free_rtdev_targ:
791 if (mp->m_rtdev_targp)
792 xfs_free_buftarg(mp, mp->m_rtdev_targp);
793 out_free_ddev_targ:
794 xfs_free_buftarg(mp, mp->m_ddev_targp);
795 out_close_rtdev:
796 if (rtdev)
797 xfs_blkdev_put(rtdev);
798 out_close_logdev:
799 if (logdev && logdev != ddev)
800 xfs_blkdev_put(logdev);
801 out:
802 return error;
803 }
804
805 /*
806 * Setup xfs_mount buffer target pointers based on superblock
807 */
808 STATIC int
809 xfs_setup_devices(
810 struct xfs_mount *mp)
811 {
812 int error;
813
814 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
815 mp->m_sb.sb_sectsize);
816 if (error)
817 return error;
818
819 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
820 unsigned int log_sector_size = BBSIZE;
821
822 if (xfs_sb_version_hassector(&mp->m_sb))
823 log_sector_size = mp->m_sb.sb_logsectsize;
824 error = xfs_setsize_buftarg(mp->m_logdev_targp,
825 mp->m_sb.sb_blocksize,
826 log_sector_size);
827 if (error)
828 return error;
829 }
830 if (mp->m_rtdev_targp) {
831 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
832 mp->m_sb.sb_blocksize,
833 mp->m_sb.sb_sectsize);
834 if (error)
835 return error;
836 }
837
838 return 0;
839 }
840
841 STATIC int
842 xfs_init_mount_workqueues(
843 struct xfs_mount *mp)
844 {
845 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s",
846 WQ_MEM_RECLAIM, 0, mp->m_fsname);
847 if (!mp->m_data_workqueue)
848 goto out;
849
850 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
851 WQ_MEM_RECLAIM, 0, mp->m_fsname);
852 if (!mp->m_unwritten_workqueue)
853 goto out_destroy_data_iodone_queue;
854
855 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s",
856 WQ_MEM_RECLAIM, 0, mp->m_fsname);
857 if (!mp->m_cil_workqueue)
858 goto out_destroy_unwritten;
859
860 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
861 0, 0, mp->m_fsname);
862 if (!mp->m_reclaim_workqueue)
863 goto out_destroy_cil;
864
865 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
866 0, 0, mp->m_fsname);
867 if (!mp->m_log_workqueue)
868 goto out_destroy_reclaim;
869
870 mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s",
871 0, 0, mp->m_fsname);
872 if (!mp->m_eofblocks_workqueue)
873 goto out_destroy_log;
874
875 return 0;
876
877 out_destroy_log:
878 destroy_workqueue(mp->m_log_workqueue);
879 out_destroy_reclaim:
880 destroy_workqueue(mp->m_reclaim_workqueue);
881 out_destroy_cil:
882 destroy_workqueue(mp->m_cil_workqueue);
883 out_destroy_unwritten:
884 destroy_workqueue(mp->m_unwritten_workqueue);
885 out_destroy_data_iodone_queue:
886 destroy_workqueue(mp->m_data_workqueue);
887 out:
888 return -ENOMEM;
889 }
890
891 STATIC void
892 xfs_destroy_mount_workqueues(
893 struct xfs_mount *mp)
894 {
895 destroy_workqueue(mp->m_eofblocks_workqueue);
896 destroy_workqueue(mp->m_log_workqueue);
897 destroy_workqueue(mp->m_reclaim_workqueue);
898 destroy_workqueue(mp->m_cil_workqueue);
899 destroy_workqueue(mp->m_data_workqueue);
900 destroy_workqueue(mp->m_unwritten_workqueue);
901 }
902
903 /*
904 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
905 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
906 * for IO to complete so that we effectively throttle multiple callers to the
907 * rate at which IO is completing.
908 */
909 void
910 xfs_flush_inodes(
911 struct xfs_mount *mp)
912 {
913 struct super_block *sb = mp->m_super;
914
915 if (down_read_trylock(&sb->s_umount)) {
916 sync_inodes_sb(sb, jiffies);
917 up_read(&sb->s_umount);
918 }
919 }
920
921 /* Catch misguided souls that try to use this interface on XFS */
922 STATIC struct inode *
923 xfs_fs_alloc_inode(
924 struct super_block *sb)
925 {
926 BUG();
927 return NULL;
928 }
929
930 /*
931 * Now that the generic code is guaranteed not to be accessing
932 * the linux inode, we can reclaim the inode.
933 */
934 STATIC void
935 xfs_fs_destroy_inode(
936 struct inode *inode)
937 {
938 struct xfs_inode *ip = XFS_I(inode);
939
940 trace_xfs_destroy_inode(ip);
941
942 XFS_STATS_INC(vn_reclaim);
943
944 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
945
946 /*
947 * We should never get here with one of the reclaim flags already set.
948 */
949 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
950 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
951
952 /*
953 * We always use background reclaim here because even if the
954 * inode is clean, it still may be under IO and hence we have
955 * to take the flush lock. The background reclaim path handles
956 * this more efficiently than we can here, so simply let background
957 * reclaim tear down all inodes.
958 */
959 xfs_inode_set_reclaim_tag(ip);
960 }
961
962 /*
963 * Slab object creation initialisation for the XFS inode.
964 * This covers only the idempotent fields in the XFS inode;
965 * all other fields need to be initialised on allocation
966 * from the slab. This avoids the need to repeatedly initialise
967 * fields in the xfs inode that left in the initialise state
968 * when freeing the inode.
969 */
970 STATIC void
971 xfs_fs_inode_init_once(
972 void *inode)
973 {
974 struct xfs_inode *ip = inode;
975
976 memset(ip, 0, sizeof(struct xfs_inode));
977
978 /* vfs inode */
979 inode_init_once(VFS_I(ip));
980
981 /* xfs inode */
982 atomic_set(&ip->i_pincount, 0);
983 spin_lock_init(&ip->i_flags_lock);
984
985 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
986 "xfsino", ip->i_ino);
987 }
988
989 STATIC void
990 xfs_fs_evict_inode(
991 struct inode *inode)
992 {
993 xfs_inode_t *ip = XFS_I(inode);
994
995 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
996
997 trace_xfs_evict_inode(ip);
998
999 truncate_inode_pages(&inode->i_data, 0);
1000 clear_inode(inode);
1001 XFS_STATS_INC(vn_rele);
1002 XFS_STATS_INC(vn_remove);
1003 XFS_STATS_DEC(vn_active);
1004
1005 xfs_inactive(ip);
1006 }
1007
1008 /*
1009 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1010 * serialised against cache hits here via the inode->i_lock and igrab() in
1011 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1012 * racing with us, and it avoids needing to grab a spinlock here for every inode
1013 * we drop the final reference on.
1014 */
1015 STATIC int
1016 xfs_fs_drop_inode(
1017 struct inode *inode)
1018 {
1019 struct xfs_inode *ip = XFS_I(inode);
1020
1021 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);
1022 }
1023
1024 STATIC void
1025 xfs_free_fsname(
1026 struct xfs_mount *mp)
1027 {
1028 kfree(mp->m_fsname);
1029 kfree(mp->m_rtname);
1030 kfree(mp->m_logname);
1031 }
1032
1033 STATIC void
1034 xfs_fs_put_super(
1035 struct super_block *sb)
1036 {
1037 struct xfs_mount *mp = XFS_M(sb);
1038
1039 xfs_filestream_unmount(mp);
1040 xfs_unmountfs(mp);
1041
1042 xfs_freesb(mp);
1043 xfs_icsb_destroy_counters(mp);
1044 xfs_destroy_mount_workqueues(mp);
1045 xfs_close_devices(mp);
1046 xfs_free_fsname(mp);
1047 kfree(mp);
1048 }
1049
1050 STATIC int
1051 xfs_fs_sync_fs(
1052 struct super_block *sb,
1053 int wait)
1054 {
1055 struct xfs_mount *mp = XFS_M(sb);
1056
1057 /*
1058 * Doing anything during the async pass would be counterproductive.
1059 */
1060 if (!wait)
1061 return 0;
1062
1063 xfs_log_force(mp, XFS_LOG_SYNC);
1064 if (laptop_mode) {
1065 /*
1066 * The disk must be active because we're syncing.
1067 * We schedule log work now (now that the disk is
1068 * active) instead of later (when it might not be).
1069 */
1070 flush_delayed_work(&mp->m_log->l_work);
1071 }
1072
1073 return 0;
1074 }
1075
1076 STATIC int
1077 xfs_fs_statfs(
1078 struct dentry *dentry,
1079 struct kstatfs *statp)
1080 {
1081 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1082 xfs_sb_t *sbp = &mp->m_sb;
1083 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1084 __uint64_t fakeinos, id;
1085 xfs_extlen_t lsize;
1086 __int64_t ffree;
1087
1088 statp->f_type = XFS_SB_MAGIC;
1089 statp->f_namelen = MAXNAMELEN - 1;
1090
1091 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1092 statp->f_fsid.val[0] = (u32)id;
1093 statp->f_fsid.val[1] = (u32)(id >> 32);
1094
1095 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1096
1097 spin_lock(&mp->m_sb_lock);
1098 statp->f_bsize = sbp->sb_blocksize;
1099 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1100 statp->f_blocks = sbp->sb_dblocks - lsize;
1101 statp->f_bfree = statp->f_bavail =
1102 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1103 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1104 statp->f_files =
1105 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1106 if (mp->m_maxicount)
1107 statp->f_files = min_t(typeof(statp->f_files),
1108 statp->f_files,
1109 mp->m_maxicount);
1110
1111 /* make sure statp->f_ffree does not underflow */
1112 ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1113 statp->f_ffree = max_t(__int64_t, ffree, 0);
1114
1115 spin_unlock(&mp->m_sb_lock);
1116
1117 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
1118 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) ==
1119 (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))
1120 xfs_qm_statvfs(ip, statp);
1121 return 0;
1122 }
1123
1124 STATIC void
1125 xfs_save_resvblks(struct xfs_mount *mp)
1126 {
1127 __uint64_t resblks = 0;
1128
1129 mp->m_resblks_save = mp->m_resblks;
1130 xfs_reserve_blocks(mp, &resblks, NULL);
1131 }
1132
1133 STATIC void
1134 xfs_restore_resvblks(struct xfs_mount *mp)
1135 {
1136 __uint64_t resblks;
1137
1138 if (mp->m_resblks_save) {
1139 resblks = mp->m_resblks_save;
1140 mp->m_resblks_save = 0;
1141 } else
1142 resblks = xfs_default_resblks(mp);
1143
1144 xfs_reserve_blocks(mp, &resblks, NULL);
1145 }
1146
1147 /*
1148 * Trigger writeback of all the dirty metadata in the file system.
1149 *
1150 * This ensures that the metadata is written to their location on disk rather
1151 * than just existing in transactions in the log. This means after a quiesce
1152 * there is no log replay required to write the inodes to disk - this is the
1153 * primary difference between a sync and a quiesce.
1154 *
1155 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1156 * it is started again when appropriate.
1157 */
1158 static void
1159 xfs_quiesce_attr(
1160 struct xfs_mount *mp)
1161 {
1162 int error = 0;
1163
1164 /* wait for all modifications to complete */
1165 while (atomic_read(&mp->m_active_trans) > 0)
1166 delay(100);
1167
1168 /* force the log to unpin objects from the now complete transactions */
1169 xfs_log_force(mp, XFS_LOG_SYNC);
1170
1171 /* reclaim inodes to do any IO before the freeze completes */
1172 xfs_reclaim_inodes(mp, 0);
1173 xfs_reclaim_inodes(mp, SYNC_WAIT);
1174
1175 /* Push the superblock and write an unmount record */
1176 error = xfs_log_sbcount(mp);
1177 if (error)
1178 xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
1179 "Frozen image may not be consistent.");
1180 /*
1181 * Just warn here till VFS can correctly support
1182 * read-only remount without racing.
1183 */
1184 WARN_ON(atomic_read(&mp->m_active_trans) != 0);
1185
1186 xfs_log_quiesce(mp);
1187 }
1188
1189 STATIC int
1190 xfs_fs_remount(
1191 struct super_block *sb,
1192 int *flags,
1193 char *options)
1194 {
1195 struct xfs_mount *mp = XFS_M(sb);
1196 substring_t args[MAX_OPT_ARGS];
1197 char *p;
1198 int error;
1199
1200 while ((p = strsep(&options, ",")) != NULL) {
1201 int token;
1202
1203 if (!*p)
1204 continue;
1205
1206 token = match_token(p, tokens, args);
1207 switch (token) {
1208 case Opt_barrier:
1209 mp->m_flags |= XFS_MOUNT_BARRIER;
1210 break;
1211 case Opt_nobarrier:
1212 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1213 break;
1214 case Opt_inode64:
1215 mp->m_maxagi = xfs_set_inode64(mp);
1216 break;
1217 case Opt_inode32:
1218 mp->m_maxagi = xfs_set_inode32(mp);
1219 break;
1220 default:
1221 /*
1222 * Logically we would return an error here to prevent
1223 * users from believing they might have changed
1224 * mount options using remount which can't be changed.
1225 *
1226 * But unfortunately mount(8) adds all options from
1227 * mtab and fstab to the mount arguments in some cases
1228 * so we can't blindly reject options, but have to
1229 * check for each specified option if it actually
1230 * differs from the currently set option and only
1231 * reject it if that's the case.
1232 *
1233 * Until that is implemented we return success for
1234 * every remount request, and silently ignore all
1235 * options that we can't actually change.
1236 */
1237 #if 0
1238 xfs_info(mp,
1239 "mount option \"%s\" not supported for remount", p);
1240 return -EINVAL;
1241 #else
1242 break;
1243 #endif
1244 }
1245 }
1246
1247 /* ro -> rw */
1248 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1249 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1250
1251 /*
1252 * If this is the first remount to writeable state we
1253 * might have some superblock changes to update.
1254 */
1255 if (mp->m_update_flags) {
1256 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1257 if (error) {
1258 xfs_warn(mp, "failed to write sb changes");
1259 return error;
1260 }
1261 mp->m_update_flags = 0;
1262 }
1263
1264 /*
1265 * Fill out the reserve pool if it is empty. Use the stashed
1266 * value if it is non-zero, otherwise go with the default.
1267 */
1268 xfs_restore_resvblks(mp);
1269 xfs_log_work_queue(mp);
1270 }
1271
1272 /* rw -> ro */
1273 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1274 /*
1275 * Before we sync the metadata, we need to free up the reserve
1276 * block pool so that the used block count in the superblock on
1277 * disk is correct at the end of the remount. Stash the current
1278 * reserve pool size so that if we get remounted rw, we can
1279 * return it to the same size.
1280 */
1281 xfs_save_resvblks(mp);
1282 xfs_quiesce_attr(mp);
1283 mp->m_flags |= XFS_MOUNT_RDONLY;
1284 }
1285
1286 return 0;
1287 }
1288
1289 /*
1290 * Second stage of a freeze. The data is already frozen so we only
1291 * need to take care of the metadata. Once that's done write a dummy
1292 * record to dirty the log in case of a crash while frozen.
1293 */
1294 STATIC int
1295 xfs_fs_freeze(
1296 struct super_block *sb)
1297 {
1298 struct xfs_mount *mp = XFS_M(sb);
1299
1300 xfs_save_resvblks(mp);
1301 xfs_quiesce_attr(mp);
1302 return -xfs_fs_log_dummy(mp);
1303 }
1304
1305 STATIC int
1306 xfs_fs_unfreeze(
1307 struct super_block *sb)
1308 {
1309 struct xfs_mount *mp = XFS_M(sb);
1310
1311 xfs_restore_resvblks(mp);
1312 xfs_log_work_queue(mp);
1313 return 0;
1314 }
1315
1316 STATIC int
1317 xfs_fs_show_options(
1318 struct seq_file *m,
1319 struct dentry *root)
1320 {
1321 return -xfs_showargs(XFS_M(root->d_sb), m);
1322 }
1323
1324 /*
1325 * This function fills in xfs_mount_t fields based on mount args.
1326 * Note: the superblock _has_ now been read in.
1327 */
1328 STATIC int
1329 xfs_finish_flags(
1330 struct xfs_mount *mp)
1331 {
1332 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1333
1334 /* Fail a mount where the logbuf is smaller than the log stripe */
1335 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1336 if (mp->m_logbsize <= 0 &&
1337 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1338 mp->m_logbsize = mp->m_sb.sb_logsunit;
1339 } else if (mp->m_logbsize > 0 &&
1340 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1341 xfs_warn(mp,
1342 "logbuf size must be greater than or equal to log stripe size");
1343 return XFS_ERROR(EINVAL);
1344 }
1345 } else {
1346 /* Fail a mount if the logbuf is larger than 32K */
1347 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1348 xfs_warn(mp,
1349 "logbuf size for version 1 logs must be 16K or 32K");
1350 return XFS_ERROR(EINVAL);
1351 }
1352 }
1353
1354 /*
1355 * V5 filesystems always use attr2 format for attributes.
1356 */
1357 if (xfs_sb_version_hascrc(&mp->m_sb) &&
1358 (mp->m_flags & XFS_MOUNT_NOATTR2)) {
1359 xfs_warn(mp,
1360 "Cannot mount a V5 filesystem as %s. %s is always enabled for V5 filesystems.",
1361 MNTOPT_NOATTR2, MNTOPT_ATTR2);
1362 return XFS_ERROR(EINVAL);
1363 }
1364
1365 /*
1366 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1367 * told by noattr2 to turn it off
1368 */
1369 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1370 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1371 mp->m_flags |= XFS_MOUNT_ATTR2;
1372
1373 /*
1374 * prohibit r/w mounts of read-only filesystems
1375 */
1376 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1377 xfs_warn(mp,
1378 "cannot mount a read-only filesystem as read-write");
1379 return XFS_ERROR(EROFS);
1380 }
1381
1382 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
1383 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) &&
1384 !xfs_sb_version_has_pquotino(&mp->m_sb)) {
1385 xfs_warn(mp,
1386 "Super block does not support project and group quota together");
1387 return XFS_ERROR(EINVAL);
1388 }
1389
1390 return 0;
1391 }
1392
1393 STATIC int
1394 xfs_fs_fill_super(
1395 struct super_block *sb,
1396 void *data,
1397 int silent)
1398 {
1399 struct inode *root;
1400 struct xfs_mount *mp = NULL;
1401 int flags = 0, error = ENOMEM;
1402
1403 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1404 if (!mp)
1405 goto out;
1406
1407 spin_lock_init(&mp->m_sb_lock);
1408 mutex_init(&mp->m_growlock);
1409 atomic_set(&mp->m_active_trans, 0);
1410 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1411 INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
1412
1413 mp->m_super = sb;
1414 sb->s_fs_info = mp;
1415
1416 error = xfs_parseargs(mp, (char *)data);
1417 if (error)
1418 goto out_free_fsname;
1419
1420 sb_min_blocksize(sb, BBSIZE);
1421 sb->s_xattr = xfs_xattr_handlers;
1422 sb->s_export_op = &xfs_export_operations;
1423 #ifdef CONFIG_XFS_QUOTA
1424 sb->s_qcop = &xfs_quotactl_operations;
1425 #endif
1426 sb->s_op = &xfs_super_operations;
1427
1428 if (silent)
1429 flags |= XFS_MFSI_QUIET;
1430
1431 error = xfs_open_devices(mp);
1432 if (error)
1433 goto out_free_fsname;
1434
1435 error = xfs_init_mount_workqueues(mp);
1436 if (error)
1437 goto out_close_devices;
1438
1439 error = xfs_icsb_init_counters(mp);
1440 if (error)
1441 goto out_destroy_workqueues;
1442
1443 error = xfs_readsb(mp, flags);
1444 if (error)
1445 goto out_destroy_counters;
1446
1447 error = xfs_finish_flags(mp);
1448 if (error)
1449 goto out_free_sb;
1450
1451 error = xfs_setup_devices(mp);
1452 if (error)
1453 goto out_free_sb;
1454
1455 error = xfs_filestream_mount(mp);
1456 if (error)
1457 goto out_free_sb;
1458
1459 /*
1460 * we must configure the block size in the superblock before we run the
1461 * full mount process as the mount process can lookup and cache inodes.
1462 */
1463 sb->s_magic = XFS_SB_MAGIC;
1464 sb->s_blocksize = mp->m_sb.sb_blocksize;
1465 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1466 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1467 sb->s_max_links = XFS_MAXLINK;
1468 sb->s_time_gran = 1;
1469 set_posix_acl_flag(sb);
1470
1471 /* version 5 superblocks support inode version counters. */
1472 if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
1473 sb->s_flags |= MS_I_VERSION;
1474
1475 error = xfs_mountfs(mp);
1476 if (error)
1477 goto out_filestream_unmount;
1478
1479 root = igrab(VFS_I(mp->m_rootip));
1480 if (!root) {
1481 error = ENOENT;
1482 goto out_unmount;
1483 }
1484 sb->s_root = d_make_root(root);
1485 if (!sb->s_root) {
1486 error = ENOMEM;
1487 goto out_unmount;
1488 }
1489
1490 return 0;
1491
1492 out_filestream_unmount:
1493 xfs_filestream_unmount(mp);
1494 out_free_sb:
1495 xfs_freesb(mp);
1496 out_destroy_counters:
1497 xfs_icsb_destroy_counters(mp);
1498 out_destroy_workqueues:
1499 xfs_destroy_mount_workqueues(mp);
1500 out_close_devices:
1501 xfs_close_devices(mp);
1502 out_free_fsname:
1503 xfs_free_fsname(mp);
1504 kfree(mp);
1505 out:
1506 return -error;
1507
1508 out_unmount:
1509 xfs_filestream_unmount(mp);
1510 xfs_unmountfs(mp);
1511 goto out_free_sb;
1512 }
1513
1514 STATIC struct dentry *
1515 xfs_fs_mount(
1516 struct file_system_type *fs_type,
1517 int flags,
1518 const char *dev_name,
1519 void *data)
1520 {
1521 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1522 }
1523
1524 static long
1525 xfs_fs_nr_cached_objects(
1526 struct super_block *sb,
1527 int nid)
1528 {
1529 return xfs_reclaim_inodes_count(XFS_M(sb));
1530 }
1531
1532 static long
1533 xfs_fs_free_cached_objects(
1534 struct super_block *sb,
1535 long nr_to_scan,
1536 int nid)
1537 {
1538 return xfs_reclaim_inodes_nr(XFS_M(sb), nr_to_scan);
1539 }
1540
1541 static const struct super_operations xfs_super_operations = {
1542 .alloc_inode = xfs_fs_alloc_inode,
1543 .destroy_inode = xfs_fs_destroy_inode,
1544 .evict_inode = xfs_fs_evict_inode,
1545 .drop_inode = xfs_fs_drop_inode,
1546 .put_super = xfs_fs_put_super,
1547 .sync_fs = xfs_fs_sync_fs,
1548 .freeze_fs = xfs_fs_freeze,
1549 .unfreeze_fs = xfs_fs_unfreeze,
1550 .statfs = xfs_fs_statfs,
1551 .remount_fs = xfs_fs_remount,
1552 .show_options = xfs_fs_show_options,
1553 .nr_cached_objects = xfs_fs_nr_cached_objects,
1554 .free_cached_objects = xfs_fs_free_cached_objects,
1555 };
1556
1557 static struct file_system_type xfs_fs_type = {
1558 .owner = THIS_MODULE,
1559 .name = "xfs",
1560 .mount = xfs_fs_mount,
1561 .kill_sb = kill_block_super,
1562 .fs_flags = FS_REQUIRES_DEV,
1563 };
1564 MODULE_ALIAS_FS("xfs");
1565
1566 STATIC int __init
1567 xfs_init_zones(void)
1568 {
1569
1570 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1571 if (!xfs_ioend_zone)
1572 goto out;
1573
1574 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1575 xfs_ioend_zone);
1576 if (!xfs_ioend_pool)
1577 goto out_destroy_ioend_zone;
1578
1579 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1580 "xfs_log_ticket");
1581 if (!xfs_log_ticket_zone)
1582 goto out_destroy_ioend_pool;
1583
1584 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1585 "xfs_bmap_free_item");
1586 if (!xfs_bmap_free_item_zone)
1587 goto out_destroy_log_ticket_zone;
1588
1589 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1590 "xfs_btree_cur");
1591 if (!xfs_btree_cur_zone)
1592 goto out_destroy_bmap_free_item_zone;
1593
1594 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1595 "xfs_da_state");
1596 if (!xfs_da_state_zone)
1597 goto out_destroy_btree_cur_zone;
1598
1599 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1600 if (!xfs_ifork_zone)
1601 goto out_destroy_da_state_zone;
1602
1603 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1604 if (!xfs_trans_zone)
1605 goto out_destroy_ifork_zone;
1606
1607 xfs_log_item_desc_zone =
1608 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1609 "xfs_log_item_desc");
1610 if (!xfs_log_item_desc_zone)
1611 goto out_destroy_trans_zone;
1612
1613 /*
1614 * The size of the zone allocated buf log item is the maximum
1615 * size possible under XFS. This wastes a little bit of memory,
1616 * but it is much faster.
1617 */
1618 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item),
1619 "xfs_buf_item");
1620 if (!xfs_buf_item_zone)
1621 goto out_destroy_log_item_desc_zone;
1622
1623 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1624 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1625 sizeof(xfs_extent_t))), "xfs_efd_item");
1626 if (!xfs_efd_zone)
1627 goto out_destroy_buf_item_zone;
1628
1629 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1630 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1631 sizeof(xfs_extent_t))), "xfs_efi_item");
1632 if (!xfs_efi_zone)
1633 goto out_destroy_efd_zone;
1634
1635 xfs_inode_zone =
1636 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1637 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1638 xfs_fs_inode_init_once);
1639 if (!xfs_inode_zone)
1640 goto out_destroy_efi_zone;
1641
1642 xfs_ili_zone =
1643 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1644 KM_ZONE_SPREAD, NULL);
1645 if (!xfs_ili_zone)
1646 goto out_destroy_inode_zone;
1647 xfs_icreate_zone = kmem_zone_init(sizeof(struct xfs_icreate_item),
1648 "xfs_icr");
1649 if (!xfs_icreate_zone)
1650 goto out_destroy_ili_zone;
1651
1652 return 0;
1653
1654 out_destroy_ili_zone:
1655 kmem_zone_destroy(xfs_ili_zone);
1656 out_destroy_inode_zone:
1657 kmem_zone_destroy(xfs_inode_zone);
1658 out_destroy_efi_zone:
1659 kmem_zone_destroy(xfs_efi_zone);
1660 out_destroy_efd_zone:
1661 kmem_zone_destroy(xfs_efd_zone);
1662 out_destroy_buf_item_zone:
1663 kmem_zone_destroy(xfs_buf_item_zone);
1664 out_destroy_log_item_desc_zone:
1665 kmem_zone_destroy(xfs_log_item_desc_zone);
1666 out_destroy_trans_zone:
1667 kmem_zone_destroy(xfs_trans_zone);
1668 out_destroy_ifork_zone:
1669 kmem_zone_destroy(xfs_ifork_zone);
1670 out_destroy_da_state_zone:
1671 kmem_zone_destroy(xfs_da_state_zone);
1672 out_destroy_btree_cur_zone:
1673 kmem_zone_destroy(xfs_btree_cur_zone);
1674 out_destroy_bmap_free_item_zone:
1675 kmem_zone_destroy(xfs_bmap_free_item_zone);
1676 out_destroy_log_ticket_zone:
1677 kmem_zone_destroy(xfs_log_ticket_zone);
1678 out_destroy_ioend_pool:
1679 mempool_destroy(xfs_ioend_pool);
1680 out_destroy_ioend_zone:
1681 kmem_zone_destroy(xfs_ioend_zone);
1682 out:
1683 return -ENOMEM;
1684 }
1685
1686 STATIC void
1687 xfs_destroy_zones(void)
1688 {
1689 /*
1690 * Make sure all delayed rcu free are flushed before we
1691 * destroy caches.
1692 */
1693 rcu_barrier();
1694 kmem_zone_destroy(xfs_icreate_zone);
1695 kmem_zone_destroy(xfs_ili_zone);
1696 kmem_zone_destroy(xfs_inode_zone);
1697 kmem_zone_destroy(xfs_efi_zone);
1698 kmem_zone_destroy(xfs_efd_zone);
1699 kmem_zone_destroy(xfs_buf_item_zone);
1700 kmem_zone_destroy(xfs_log_item_desc_zone);
1701 kmem_zone_destroy(xfs_trans_zone);
1702 kmem_zone_destroy(xfs_ifork_zone);
1703 kmem_zone_destroy(xfs_da_state_zone);
1704 kmem_zone_destroy(xfs_btree_cur_zone);
1705 kmem_zone_destroy(xfs_bmap_free_item_zone);
1706 kmem_zone_destroy(xfs_log_ticket_zone);
1707 mempool_destroy(xfs_ioend_pool);
1708 kmem_zone_destroy(xfs_ioend_zone);
1709
1710 }
1711
1712 STATIC int __init
1713 xfs_init_workqueues(void)
1714 {
1715 /*
1716 * The allocation workqueue can be used in memory reclaim situations
1717 * (writepage path), and parallelism is only limited by the number of
1718 * AGs in all the filesystems mounted. Hence use the default large
1719 * max_active value for this workqueue.
1720 */
1721 xfs_alloc_wq = alloc_workqueue("xfsalloc", WQ_MEM_RECLAIM, 0);
1722 if (!xfs_alloc_wq)
1723 return -ENOMEM;
1724
1725 return 0;
1726 }
1727
1728 STATIC void
1729 xfs_destroy_workqueues(void)
1730 {
1731 destroy_workqueue(xfs_alloc_wq);
1732 }
1733
1734 STATIC int __init
1735 init_xfs_fs(void)
1736 {
1737 int error;
1738
1739 printk(KERN_INFO XFS_VERSION_STRING " with "
1740 XFS_BUILD_OPTIONS " enabled\n");
1741
1742 xfs_dir_startup();
1743
1744 error = xfs_init_zones();
1745 if (error)
1746 goto out;
1747
1748 error = xfs_init_workqueues();
1749 if (error)
1750 goto out_destroy_zones;
1751
1752 error = xfs_mru_cache_init();
1753 if (error)
1754 goto out_destroy_wq;
1755
1756 error = xfs_filestream_init();
1757 if (error)
1758 goto out_mru_cache_uninit;
1759
1760 error = xfs_buf_init();
1761 if (error)
1762 goto out_filestream_uninit;
1763
1764 error = xfs_init_procfs();
1765 if (error)
1766 goto out_buf_terminate;
1767
1768 error = xfs_sysctl_register();
1769 if (error)
1770 goto out_cleanup_procfs;
1771
1772 error = xfs_qm_init();
1773 if (error)
1774 goto out_sysctl_unregister;
1775
1776 error = register_filesystem(&xfs_fs_type);
1777 if (error)
1778 goto out_qm_exit;
1779 return 0;
1780
1781 out_qm_exit:
1782 xfs_qm_exit();
1783 out_sysctl_unregister:
1784 xfs_sysctl_unregister();
1785 out_cleanup_procfs:
1786 xfs_cleanup_procfs();
1787 out_buf_terminate:
1788 xfs_buf_terminate();
1789 out_filestream_uninit:
1790 xfs_filestream_uninit();
1791 out_mru_cache_uninit:
1792 xfs_mru_cache_uninit();
1793 out_destroy_wq:
1794 xfs_destroy_workqueues();
1795 out_destroy_zones:
1796 xfs_destroy_zones();
1797 out:
1798 return error;
1799 }
1800
1801 STATIC void __exit
1802 exit_xfs_fs(void)
1803 {
1804 xfs_qm_exit();
1805 unregister_filesystem(&xfs_fs_type);
1806 xfs_sysctl_unregister();
1807 xfs_cleanup_procfs();
1808 xfs_buf_terminate();
1809 xfs_filestream_uninit();
1810 xfs_mru_cache_uninit();
1811 xfs_destroy_workqueues();
1812 xfs_destroy_zones();
1813 }
1814
1815 module_init(init_xfs_fs);
1816 module_exit(exit_xfs_fs);
1817
1818 MODULE_AUTHOR("Silicon Graphics, Inc.");
1819 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1820 MODULE_LICENSE("GPL");
Linus' kernel tree