Linux 3.12.28
[linux/fpc-iii.git] / fs / xfs / xfs_super.c
blob15188cc9944919e275e43e4978056efa91801343
1 /*
2 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
3 * All Rights Reserved.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
19 #include "xfs.h"
20 #include "xfs_format.h"
21 #include "xfs_log.h"
22 #include "xfs_inum.h"
23 #include "xfs_trans.h"
24 #include "xfs_sb.h"
25 #include "xfs_ag.h"
26 #include "xfs_alloc.h"
27 #include "xfs_quota.h"
28 #include "xfs_mount.h"
29 #include "xfs_bmap_btree.h"
30 #include "xfs_alloc_btree.h"
31 #include "xfs_ialloc_btree.h"
32 #include "xfs_dinode.h"
33 #include "xfs_inode.h"
34 #include "xfs_btree.h"
35 #include "xfs_ialloc.h"
36 #include "xfs_bmap.h"
37 #include "xfs_rtalloc.h"
38 #include "xfs_error.h"
39 #include "xfs_itable.h"
40 #include "xfs_fsops.h"
41 #include "xfs_attr.h"
42 #include "xfs_buf_item.h"
43 #include "xfs_log_priv.h"
44 #include "xfs_trans_priv.h"
45 #include "xfs_filestream.h"
46 #include "xfs_da_btree.h"
47 #include "xfs_dir2_format.h"
48 #include "xfs_dir2.h"
49 #include "xfs_extfree_item.h"
50 #include "xfs_mru_cache.h"
51 #include "xfs_inode_item.h"
52 #include "xfs_icache.h"
53 #include "xfs_trace.h"
54 #include "xfs_icreate_item.h"
56 #include <linux/namei.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
59 #include <linux/mount.h>
60 #include <linux/mempool.h>
61 #include <linux/writeback.h>
62 #include <linux/kthread.h>
63 #include <linux/freezer.h>
64 #include <linux/parser.h>
66 static const struct super_operations xfs_super_operations;
67 static kmem_zone_t *xfs_ioend_zone;
68 mempool_t *xfs_ioend_pool;
70 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
71 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
72 #define MNTOPT_LOGDEV "logdev" /* log device */
73 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
74 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
75 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
76 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
77 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
78 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
79 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
80 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
81 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
82 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
83 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
84 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
85 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
86 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
87 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
88 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
89 * unwritten extent conversion */
90 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
91 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
92 #define MNTOPT_32BITINODE "inode32" /* inode allocation limited to
93 * XFS_MAXINUMBER_32 */
94 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
95 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
96 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
97 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
98 * in stat(). */
99 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
100 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
101 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
102 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
103 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
104 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
105 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
106 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
107 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
108 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
109 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
110 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
111 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
112 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
113 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
114 #define MNTOPT_DELAYLOG "delaylog" /* Delayed logging enabled */
115 #define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed logging disabled */
116 #define MNTOPT_DISCARD "discard" /* Discard unused blocks */
117 #define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */
120 * Table driven mount option parser.
122 * Currently only used for remount, but it will be used for mount
123 * in the future, too.
125 enum {
126 Opt_barrier,
127 Opt_nobarrier,
128 Opt_inode64,
129 Opt_inode32,
130 Opt_err
133 static const match_table_t tokens = {
134 {Opt_barrier, "barrier"},
135 {Opt_nobarrier, "nobarrier"},
136 {Opt_inode64, "inode64"},
137 {Opt_inode32, "inode32"},
138 {Opt_err, NULL}
142 STATIC unsigned long
143 suffix_kstrtoint(char *s, unsigned int base, int *res)
145 int last, shift_left_factor = 0, _res;
146 char *value = s;
148 last = strlen(value) - 1;
149 if (value[last] == 'K' || value[last] == 'k') {
150 shift_left_factor = 10;
151 value[last] = '\0';
153 if (value[last] == 'M' || value[last] == 'm') {
154 shift_left_factor = 20;
155 value[last] = '\0';
157 if (value[last] == 'G' || value[last] == 'g') {
158 shift_left_factor = 30;
159 value[last] = '\0';
162 if (kstrtoint(s, base, &_res))
163 return -EINVAL;
164 *res = _res << shift_left_factor;
165 return 0;
169 * This function fills in xfs_mount_t fields based on mount args.
170 * Note: the superblock has _not_ yet been read in.
172 * Note that this function leaks the various device name allocations on
173 * failure. The caller takes care of them.
175 STATIC int
176 xfs_parseargs(
177 struct xfs_mount *mp,
178 char *options)
180 struct super_block *sb = mp->m_super;
181 char *this_char, *value;
182 int dsunit = 0;
183 int dswidth = 0;
184 int iosize = 0;
185 __uint8_t iosizelog = 0;
188 * set up the mount name first so all the errors will refer to the
189 * correct device.
191 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
192 if (!mp->m_fsname)
193 return ENOMEM;
194 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
197 * Copy binary VFS mount flags we are interested in.
199 if (sb->s_flags & MS_RDONLY)
200 mp->m_flags |= XFS_MOUNT_RDONLY;
201 if (sb->s_flags & MS_DIRSYNC)
202 mp->m_flags |= XFS_MOUNT_DIRSYNC;
203 if (sb->s_flags & MS_SYNCHRONOUS)
204 mp->m_flags |= XFS_MOUNT_WSYNC;
207 * Set some default flags that could be cleared by the mount option
208 * parsing.
210 mp->m_flags |= XFS_MOUNT_BARRIER;
211 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
212 #if !XFS_BIG_INUMS
213 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
214 #endif
217 * These can be overridden by the mount option parsing.
219 mp->m_logbufs = -1;
220 mp->m_logbsize = -1;
222 if (!options)
223 goto done;
225 while ((this_char = strsep(&options, ",")) != NULL) {
226 if (!*this_char)
227 continue;
228 if ((value = strchr(this_char, '=')) != NULL)
229 *value++ = 0;
231 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
232 if (!value || !*value) {
233 xfs_warn(mp, "%s option requires an argument",
234 this_char);
235 return EINVAL;
237 if (kstrtoint(value, 10, &mp->m_logbufs))
238 return EINVAL;
239 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
240 if (!value || !*value) {
241 xfs_warn(mp, "%s option requires an argument",
242 this_char);
243 return EINVAL;
245 if (suffix_kstrtoint(value, 10, &mp->m_logbsize))
246 return EINVAL;
247 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
248 if (!value || !*value) {
249 xfs_warn(mp, "%s option requires an argument",
250 this_char);
251 return EINVAL;
253 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
254 if (!mp->m_logname)
255 return ENOMEM;
256 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
257 xfs_warn(mp, "%s option not allowed on this system",
258 this_char);
259 return EINVAL;
260 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
261 if (!value || !*value) {
262 xfs_warn(mp, "%s option requires an argument",
263 this_char);
264 return EINVAL;
266 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
267 if (!mp->m_rtname)
268 return ENOMEM;
269 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
270 if (!value || !*value) {
271 xfs_warn(mp, "%s option requires an argument",
272 this_char);
273 return EINVAL;
275 if (kstrtoint(value, 10, &iosize))
276 return EINVAL;
277 iosizelog = ffs(iosize) - 1;
278 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
279 if (!value || !*value) {
280 xfs_warn(mp, "%s option requires an argument",
281 this_char);
282 return EINVAL;
284 if (suffix_kstrtoint(value, 10, &iosize))
285 return EINVAL;
286 iosizelog = ffs(iosize) - 1;
287 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
288 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
289 mp->m_flags |= XFS_MOUNT_GRPID;
290 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
291 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
292 mp->m_flags &= ~XFS_MOUNT_GRPID;
293 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
294 mp->m_flags |= XFS_MOUNT_WSYNC;
295 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
296 mp->m_flags |= XFS_MOUNT_NORECOVERY;
297 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
298 mp->m_flags |= XFS_MOUNT_NOALIGN;
299 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
300 mp->m_flags |= XFS_MOUNT_SWALLOC;
301 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
302 if (!value || !*value) {
303 xfs_warn(mp, "%s option requires an argument",
304 this_char);
305 return EINVAL;
307 if (kstrtoint(value, 10, &dsunit))
308 return EINVAL;
309 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
310 if (!value || !*value) {
311 xfs_warn(mp, "%s option requires an argument",
312 this_char);
313 return EINVAL;
315 if (kstrtoint(value, 10, &dswidth))
316 return EINVAL;
317 } else if (!strcmp(this_char, MNTOPT_32BITINODE)) {
318 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
319 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
320 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
321 #if !XFS_BIG_INUMS
322 xfs_warn(mp, "%s option not allowed on this system",
323 this_char);
324 return EINVAL;
325 #endif
326 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
327 mp->m_flags |= XFS_MOUNT_NOUUID;
328 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
329 mp->m_flags |= XFS_MOUNT_BARRIER;
330 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
331 mp->m_flags &= ~XFS_MOUNT_BARRIER;
332 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
333 mp->m_flags |= XFS_MOUNT_IKEEP;
334 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
335 mp->m_flags &= ~XFS_MOUNT_IKEEP;
336 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
337 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
338 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
339 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
340 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
341 mp->m_flags |= XFS_MOUNT_ATTR2;
342 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
343 mp->m_flags &= ~XFS_MOUNT_ATTR2;
344 mp->m_flags |= XFS_MOUNT_NOATTR2;
345 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
346 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
347 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
348 mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT;
349 mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD;
350 mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE;
351 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
352 !strcmp(this_char, MNTOPT_UQUOTA) ||
353 !strcmp(this_char, MNTOPT_USRQUOTA)) {
354 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
355 XFS_UQUOTA_ENFD);
356 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
357 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
358 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
359 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
360 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
361 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
362 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
363 XFS_PQUOTA_ENFD);
364 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
365 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
366 mp->m_qflags &= ~XFS_PQUOTA_ENFD;
367 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
368 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
369 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
370 XFS_GQUOTA_ENFD);
371 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
372 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
373 mp->m_qflags &= ~XFS_GQUOTA_ENFD;
374 } else if (!strcmp(this_char, MNTOPT_DELAYLOG)) {
375 xfs_warn(mp,
376 "delaylog is the default now, option is deprecated.");
377 } else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) {
378 xfs_warn(mp,
379 "nodelaylog support has been removed, option is deprecated.");
380 } else if (!strcmp(this_char, MNTOPT_DISCARD)) {
381 mp->m_flags |= XFS_MOUNT_DISCARD;
382 } else if (!strcmp(this_char, MNTOPT_NODISCARD)) {
383 mp->m_flags &= ~XFS_MOUNT_DISCARD;
384 } else if (!strcmp(this_char, "ihashsize")) {
385 xfs_warn(mp,
386 "ihashsize no longer used, option is deprecated.");
387 } else if (!strcmp(this_char, "osyncisdsync")) {
388 xfs_warn(mp,
389 "osyncisdsync has no effect, option is deprecated.");
390 } else if (!strcmp(this_char, "osyncisosync")) {
391 xfs_warn(mp,
392 "osyncisosync has no effect, option is deprecated.");
393 } else if (!strcmp(this_char, "irixsgid")) {
394 xfs_warn(mp,
395 "irixsgid is now a sysctl(2) variable, option is deprecated.");
396 } else {
397 xfs_warn(mp, "unknown mount option [%s].", this_char);
398 return EINVAL;
403 * no recovery flag requires a read-only mount
405 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
406 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
407 xfs_warn(mp, "no-recovery mounts must be read-only.");
408 return EINVAL;
411 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
412 xfs_warn(mp,
413 "sunit and swidth options incompatible with the noalign option");
414 return EINVAL;
417 #ifndef CONFIG_XFS_QUOTA
418 if (XFS_IS_QUOTA_RUNNING(mp)) {
419 xfs_warn(mp, "quota support not available in this kernel.");
420 return EINVAL;
422 #endif
424 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
425 xfs_warn(mp, "sunit and swidth must be specified together");
426 return EINVAL;
429 if (dsunit && (dswidth % dsunit != 0)) {
430 xfs_warn(mp,
431 "stripe width (%d) must be a multiple of the stripe unit (%d)",
432 dswidth, dsunit);
433 return EINVAL;
436 done:
437 if (dsunit && !(mp->m_flags & XFS_MOUNT_NOALIGN)) {
439 * At this point the superblock has not been read
440 * in, therefore we do not know the block size.
441 * Before the mount call ends we will convert
442 * these to FSBs.
444 mp->m_dalign = dsunit;
445 mp->m_swidth = dswidth;
448 if (mp->m_logbufs != -1 &&
449 mp->m_logbufs != 0 &&
450 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
451 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
452 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
453 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
454 return XFS_ERROR(EINVAL);
456 if (mp->m_logbsize != -1 &&
457 mp->m_logbsize != 0 &&
458 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
459 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
460 !is_power_of_2(mp->m_logbsize))) {
461 xfs_warn(mp,
462 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
463 mp->m_logbsize);
464 return XFS_ERROR(EINVAL);
467 if (iosizelog) {
468 if (iosizelog > XFS_MAX_IO_LOG ||
469 iosizelog < XFS_MIN_IO_LOG) {
470 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]",
471 iosizelog, XFS_MIN_IO_LOG,
472 XFS_MAX_IO_LOG);
473 return XFS_ERROR(EINVAL);
476 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
477 mp->m_readio_log = iosizelog;
478 mp->m_writeio_log = iosizelog;
481 return 0;
484 struct proc_xfs_info {
485 int flag;
486 char *str;
489 STATIC int
490 xfs_showargs(
491 struct xfs_mount *mp,
492 struct seq_file *m)
494 static struct proc_xfs_info xfs_info_set[] = {
495 /* the few simple ones we can get from the mount struct */
496 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
497 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
498 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
499 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
500 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
501 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
502 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
503 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
504 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
505 { XFS_MOUNT_DISCARD, "," MNTOPT_DISCARD },
506 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_32BITINODE },
507 { 0, NULL }
509 static struct proc_xfs_info xfs_info_unset[] = {
510 /* the few simple ones we can get from the mount struct */
511 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
512 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
513 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
514 { 0, NULL }
516 struct proc_xfs_info *xfs_infop;
518 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
519 if (mp->m_flags & xfs_infop->flag)
520 seq_puts(m, xfs_infop->str);
522 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
523 if (!(mp->m_flags & xfs_infop->flag))
524 seq_puts(m, xfs_infop->str);
527 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
528 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
529 (int)(1 << mp->m_writeio_log) >> 10);
531 if (mp->m_logbufs > 0)
532 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
533 if (mp->m_logbsize > 0)
534 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
536 if (mp->m_logname)
537 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
538 if (mp->m_rtname)
539 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
541 if (mp->m_dalign > 0)
542 seq_printf(m, "," MNTOPT_SUNIT "=%d",
543 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
544 if (mp->m_swidth > 0)
545 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
546 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
548 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
549 seq_puts(m, "," MNTOPT_USRQUOTA);
550 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
551 seq_puts(m, "," MNTOPT_UQUOTANOENF);
553 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
554 if (mp->m_qflags & XFS_PQUOTA_ENFD)
555 seq_puts(m, "," MNTOPT_PRJQUOTA);
556 else
557 seq_puts(m, "," MNTOPT_PQUOTANOENF);
559 if (mp->m_qflags & XFS_GQUOTA_ACCT) {
560 if (mp->m_qflags & XFS_GQUOTA_ENFD)
561 seq_puts(m, "," MNTOPT_GRPQUOTA);
562 else
563 seq_puts(m, "," MNTOPT_GQUOTANOENF);
566 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
567 seq_puts(m, "," MNTOPT_NOQUOTA);
569 return 0;
571 __uint64_t
572 xfs_max_file_offset(
573 unsigned int blockshift)
575 unsigned int pagefactor = 1;
576 unsigned int bitshift = BITS_PER_LONG - 1;
578 /* Figure out maximum filesize, on Linux this can depend on
579 * the filesystem blocksize (on 32 bit platforms).
580 * __block_write_begin does this in an [unsigned] long...
581 * page->index << (PAGE_CACHE_SHIFT - bbits)
582 * So, for page sized blocks (4K on 32 bit platforms),
583 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
584 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
585 * but for smaller blocksizes it is less (bbits = log2 bsize).
586 * Note1: get_block_t takes a long (implicit cast from above)
587 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
588 * can optionally convert the [unsigned] long from above into
589 * an [unsigned] long long.
592 #if BITS_PER_LONG == 32
593 # if defined(CONFIG_LBDAF)
594 ASSERT(sizeof(sector_t) == 8);
595 pagefactor = PAGE_CACHE_SIZE;
596 bitshift = BITS_PER_LONG;
597 # else
598 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
599 # endif
600 #endif
602 return (((__uint64_t)pagefactor) << bitshift) - 1;
605 xfs_agnumber_t
606 xfs_set_inode32(struct xfs_mount *mp)
608 xfs_agnumber_t index = 0;
609 xfs_agnumber_t maxagi = 0;
610 xfs_sb_t *sbp = &mp->m_sb;
611 xfs_agnumber_t max_metadata;
612 xfs_agino_t agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks -1, 0);
613 xfs_ino_t ino = XFS_AGINO_TO_INO(mp, sbp->sb_agcount -1, agino);
614 xfs_perag_t *pag;
616 /* Calculate how much should be reserved for inodes to meet
617 * the max inode percentage.
619 if (mp->m_maxicount) {
620 __uint64_t icount;
622 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
623 do_div(icount, 100);
624 icount += sbp->sb_agblocks - 1;
625 do_div(icount, sbp->sb_agblocks);
626 max_metadata = icount;
627 } else {
628 max_metadata = sbp->sb_agcount;
631 for (index = 0; index < sbp->sb_agcount; index++) {
632 ino = XFS_AGINO_TO_INO(mp, index, agino);
634 if (ino > XFS_MAXINUMBER_32) {
635 pag = xfs_perag_get(mp, index);
636 pag->pagi_inodeok = 0;
637 pag->pagf_metadata = 0;
638 xfs_perag_put(pag);
639 continue;
642 pag = xfs_perag_get(mp, index);
643 pag->pagi_inodeok = 1;
644 maxagi++;
645 if (index < max_metadata)
646 pag->pagf_metadata = 1;
647 xfs_perag_put(pag);
649 mp->m_flags |= (XFS_MOUNT_32BITINODES |
650 XFS_MOUNT_SMALL_INUMS);
652 return maxagi;
655 xfs_agnumber_t
656 xfs_set_inode64(struct xfs_mount *mp)
658 xfs_agnumber_t index = 0;
660 for (index = 0; index < mp->m_sb.sb_agcount; index++) {
661 struct xfs_perag *pag;
663 pag = xfs_perag_get(mp, index);
664 pag->pagi_inodeok = 1;
665 pag->pagf_metadata = 0;
666 xfs_perag_put(pag);
669 /* There is no need for lock protection on m_flags,
670 * the rw_semaphore of the VFS superblock is locked
671 * during mount/umount/remount operations, so this is
672 * enough to avoid concurency on the m_flags field
674 mp->m_flags &= ~(XFS_MOUNT_32BITINODES |
675 XFS_MOUNT_SMALL_INUMS);
676 return index;
679 STATIC int
680 xfs_blkdev_get(
681 xfs_mount_t *mp,
682 const char *name,
683 struct block_device **bdevp)
685 int error = 0;
687 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
688 mp);
689 if (IS_ERR(*bdevp)) {
690 error = PTR_ERR(*bdevp);
691 xfs_warn(mp, "Invalid device [%s], error=%d\n", name, error);
694 return -error;
697 STATIC void
698 xfs_blkdev_put(
699 struct block_device *bdev)
701 if (bdev)
702 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
705 void
706 xfs_blkdev_issue_flush(
707 xfs_buftarg_t *buftarg)
709 blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL);
712 STATIC void
713 xfs_close_devices(
714 struct xfs_mount *mp)
716 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
717 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
718 xfs_free_buftarg(mp, mp->m_logdev_targp);
719 xfs_blkdev_put(logdev);
721 if (mp->m_rtdev_targp) {
722 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
723 xfs_free_buftarg(mp, mp->m_rtdev_targp);
724 xfs_blkdev_put(rtdev);
726 xfs_free_buftarg(mp, mp->m_ddev_targp);
730 * The file system configurations are:
731 * (1) device (partition) with data and internal log
732 * (2) logical volume with data and log subvolumes.
733 * (3) logical volume with data, log, and realtime subvolumes.
735 * We only have to handle opening the log and realtime volumes here if
736 * they are present. The data subvolume has already been opened by
737 * get_sb_bdev() and is stored in sb->s_bdev.
739 STATIC int
740 xfs_open_devices(
741 struct xfs_mount *mp)
743 struct block_device *ddev = mp->m_super->s_bdev;
744 struct block_device *logdev = NULL, *rtdev = NULL;
745 int error;
748 * Open real time and log devices - order is important.
750 if (mp->m_logname) {
751 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
752 if (error)
753 goto out;
756 if (mp->m_rtname) {
757 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
758 if (error)
759 goto out_close_logdev;
761 if (rtdev == ddev || rtdev == logdev) {
762 xfs_warn(mp,
763 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
764 error = EINVAL;
765 goto out_close_rtdev;
770 * Setup xfs_mount buffer target pointers
772 error = ENOMEM;
773 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, 0, mp->m_fsname);
774 if (!mp->m_ddev_targp)
775 goto out_close_rtdev;
777 if (rtdev) {
778 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, 1,
779 mp->m_fsname);
780 if (!mp->m_rtdev_targp)
781 goto out_free_ddev_targ;
784 if (logdev && logdev != ddev) {
785 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, 1,
786 mp->m_fsname);
787 if (!mp->m_logdev_targp)
788 goto out_free_rtdev_targ;
789 } else {
790 mp->m_logdev_targp = mp->m_ddev_targp;
793 return 0;
795 out_free_rtdev_targ:
796 if (mp->m_rtdev_targp)
797 xfs_free_buftarg(mp, mp->m_rtdev_targp);
798 out_free_ddev_targ:
799 xfs_free_buftarg(mp, mp->m_ddev_targp);
800 out_close_rtdev:
801 if (rtdev)
802 xfs_blkdev_put(rtdev);
803 out_close_logdev:
804 if (logdev && logdev != ddev)
805 xfs_blkdev_put(logdev);
806 out:
807 return error;
811 * Setup xfs_mount buffer target pointers based on superblock
813 STATIC int
814 xfs_setup_devices(
815 struct xfs_mount *mp)
817 int error;
819 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
820 mp->m_sb.sb_sectsize);
821 if (error)
822 return error;
824 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
825 unsigned int log_sector_size = BBSIZE;
827 if (xfs_sb_version_hassector(&mp->m_sb))
828 log_sector_size = mp->m_sb.sb_logsectsize;
829 error = xfs_setsize_buftarg(mp->m_logdev_targp,
830 mp->m_sb.sb_blocksize,
831 log_sector_size);
832 if (error)
833 return error;
835 if (mp->m_rtdev_targp) {
836 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
837 mp->m_sb.sb_blocksize,
838 mp->m_sb.sb_sectsize);
839 if (error)
840 return error;
843 return 0;
846 STATIC int
847 xfs_init_mount_workqueues(
848 struct xfs_mount *mp)
850 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s",
851 WQ_MEM_RECLAIM, 0, mp->m_fsname);
852 if (!mp->m_data_workqueue)
853 goto out;
855 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
856 WQ_MEM_RECLAIM, 0, mp->m_fsname);
857 if (!mp->m_unwritten_workqueue)
858 goto out_destroy_data_iodone_queue;
860 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s",
861 WQ_MEM_RECLAIM, 0, mp->m_fsname);
862 if (!mp->m_cil_workqueue)
863 goto out_destroy_unwritten;
865 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
866 0, 0, mp->m_fsname);
867 if (!mp->m_reclaim_workqueue)
868 goto out_destroy_cil;
870 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
871 0, 0, mp->m_fsname);
872 if (!mp->m_log_workqueue)
873 goto out_destroy_reclaim;
875 mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s",
876 0, 0, mp->m_fsname);
877 if (!mp->m_eofblocks_workqueue)
878 goto out_destroy_log;
880 return 0;
882 out_destroy_log:
883 destroy_workqueue(mp->m_log_workqueue);
884 out_destroy_reclaim:
885 destroy_workqueue(mp->m_reclaim_workqueue);
886 out_destroy_cil:
887 destroy_workqueue(mp->m_cil_workqueue);
888 out_destroy_unwritten:
889 destroy_workqueue(mp->m_unwritten_workqueue);
890 out_destroy_data_iodone_queue:
891 destroy_workqueue(mp->m_data_workqueue);
892 out:
893 return -ENOMEM;
896 STATIC void
897 xfs_destroy_mount_workqueues(
898 struct xfs_mount *mp)
900 destroy_workqueue(mp->m_eofblocks_workqueue);
901 destroy_workqueue(mp->m_log_workqueue);
902 destroy_workqueue(mp->m_reclaim_workqueue);
903 destroy_workqueue(mp->m_cil_workqueue);
904 destroy_workqueue(mp->m_data_workqueue);
905 destroy_workqueue(mp->m_unwritten_workqueue);
909 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
910 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
911 * for IO to complete so that we effectively throttle multiple callers to the
912 * rate at which IO is completing.
914 void
915 xfs_flush_inodes(
916 struct xfs_mount *mp)
918 struct super_block *sb = mp->m_super;
920 if (down_read_trylock(&sb->s_umount)) {
921 sync_inodes_sb(sb);
922 up_read(&sb->s_umount);
926 /* Catch misguided souls that try to use this interface on XFS */
927 STATIC struct inode *
928 xfs_fs_alloc_inode(
929 struct super_block *sb)
931 BUG();
932 return NULL;
936 * Now that the generic code is guaranteed not to be accessing
937 * the linux inode, we can reclaim the inode.
939 STATIC void
940 xfs_fs_destroy_inode(
941 struct inode *inode)
943 struct xfs_inode *ip = XFS_I(inode);
945 trace_xfs_destroy_inode(ip);
947 XFS_STATS_INC(vn_reclaim);
949 /* bad inode, get out here ASAP */
950 if (is_bad_inode(inode))
951 goto out_reclaim;
953 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
956 * We should never get here with one of the reclaim flags already set.
958 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
959 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
962 * We always use background reclaim here because even if the
963 * inode is clean, it still may be under IO and hence we have
964 * to take the flush lock. The background reclaim path handles
965 * this more efficiently than we can here, so simply let background
966 * reclaim tear down all inodes.
968 out_reclaim:
969 xfs_inode_set_reclaim_tag(ip);
973 * Slab object creation initialisation for the XFS inode.
974 * This covers only the idempotent fields in the XFS inode;
975 * all other fields need to be initialised on allocation
976 * from the slab. This avoids the need to repeatedly initialise
977 * fields in the xfs inode that left in the initialise state
978 * when freeing the inode.
980 STATIC void
981 xfs_fs_inode_init_once(
982 void *inode)
984 struct xfs_inode *ip = inode;
986 memset(ip, 0, sizeof(struct xfs_inode));
988 /* vfs inode */
989 inode_init_once(VFS_I(ip));
991 /* xfs inode */
992 atomic_set(&ip->i_pincount, 0);
993 spin_lock_init(&ip->i_flags_lock);
995 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
996 "xfsino", ip->i_ino);
999 STATIC void
1000 xfs_fs_evict_inode(
1001 struct inode *inode)
1003 xfs_inode_t *ip = XFS_I(inode);
1005 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
1007 trace_xfs_evict_inode(ip);
1009 truncate_inode_pages(&inode->i_data, 0);
1010 clear_inode(inode);
1011 XFS_STATS_INC(vn_rele);
1012 XFS_STATS_INC(vn_remove);
1013 XFS_STATS_DEC(vn_active);
1015 xfs_inactive(ip);
1019 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1020 * serialised against cache hits here via the inode->i_lock and igrab() in
1021 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1022 * racing with us, and it avoids needing to grab a spinlock here for every inode
1023 * we drop the final reference on.
1025 STATIC int
1026 xfs_fs_drop_inode(
1027 struct inode *inode)
1029 struct xfs_inode *ip = XFS_I(inode);
1031 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);
1034 STATIC void
1035 xfs_free_fsname(
1036 struct xfs_mount *mp)
1038 kfree(mp->m_fsname);
1039 kfree(mp->m_rtname);
1040 kfree(mp->m_logname);
1043 STATIC void
1044 xfs_fs_put_super(
1045 struct super_block *sb)
1047 struct xfs_mount *mp = XFS_M(sb);
1049 xfs_filestream_unmount(mp);
1050 xfs_unmountfs(mp);
1052 xfs_freesb(mp);
1053 xfs_icsb_destroy_counters(mp);
1054 xfs_destroy_mount_workqueues(mp);
1055 xfs_close_devices(mp);
1056 xfs_free_fsname(mp);
1057 kfree(mp);
1060 STATIC int
1061 xfs_fs_sync_fs(
1062 struct super_block *sb,
1063 int wait)
1065 struct xfs_mount *mp = XFS_M(sb);
1068 * Doing anything during the async pass would be counterproductive.
1070 if (!wait)
1071 return 0;
1073 xfs_log_force(mp, XFS_LOG_SYNC);
1074 if (laptop_mode) {
1076 * The disk must be active because we're syncing.
1077 * We schedule log work now (now that the disk is
1078 * active) instead of later (when it might not be).
1080 flush_delayed_work(&mp->m_log->l_work);
1083 return 0;
1086 STATIC int
1087 xfs_fs_statfs(
1088 struct dentry *dentry,
1089 struct kstatfs *statp)
1091 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1092 xfs_sb_t *sbp = &mp->m_sb;
1093 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1094 __uint64_t fakeinos, id;
1095 xfs_extlen_t lsize;
1096 __int64_t ffree;
1098 statp->f_type = XFS_SB_MAGIC;
1099 statp->f_namelen = MAXNAMELEN - 1;
1101 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1102 statp->f_fsid.val[0] = (u32)id;
1103 statp->f_fsid.val[1] = (u32)(id >> 32);
1105 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1107 spin_lock(&mp->m_sb_lock);
1108 statp->f_bsize = sbp->sb_blocksize;
1109 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1110 statp->f_blocks = sbp->sb_dblocks - lsize;
1111 statp->f_bfree = statp->f_bavail =
1112 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1113 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1114 statp->f_files =
1115 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1116 if (mp->m_maxicount)
1117 statp->f_files = min_t(typeof(statp->f_files),
1118 statp->f_files,
1119 mp->m_maxicount);
1121 /* make sure statp->f_ffree does not underflow */
1122 ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1123 statp->f_ffree = max_t(__int64_t, ffree, 0);
1125 spin_unlock(&mp->m_sb_lock);
1127 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
1128 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) ==
1129 (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))
1130 xfs_qm_statvfs(ip, statp);
1131 return 0;
1134 STATIC void
1135 xfs_save_resvblks(struct xfs_mount *mp)
1137 __uint64_t resblks = 0;
1139 mp->m_resblks_save = mp->m_resblks;
1140 xfs_reserve_blocks(mp, &resblks, NULL);
1143 STATIC void
1144 xfs_restore_resvblks(struct xfs_mount *mp)
1146 __uint64_t resblks;
1148 if (mp->m_resblks_save) {
1149 resblks = mp->m_resblks_save;
1150 mp->m_resblks_save = 0;
1151 } else
1152 resblks = xfs_default_resblks(mp);
1154 xfs_reserve_blocks(mp, &resblks, NULL);
1158 * Trigger writeback of all the dirty metadata in the file system.
1160 * This ensures that the metadata is written to their location on disk rather
1161 * than just existing in transactions in the log. This means after a quiesce
1162 * there is no log replay required to write the inodes to disk - this is the
1163 * primary difference between a sync and a quiesce.
1165 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1166 * it is started again when appropriate.
1168 void
1169 xfs_quiesce_attr(
1170 struct xfs_mount *mp)
1172 int error = 0;
1174 /* wait for all modifications to complete */
1175 while (atomic_read(&mp->m_active_trans) > 0)
1176 delay(100);
1178 /* force the log to unpin objects from the now complete transactions */
1179 xfs_log_force(mp, XFS_LOG_SYNC);
1181 /* reclaim inodes to do any IO before the freeze completes */
1182 xfs_reclaim_inodes(mp, 0);
1183 xfs_reclaim_inodes(mp, SYNC_WAIT);
1185 /* Push the superblock and write an unmount record */
1186 error = xfs_log_sbcount(mp);
1187 if (error)
1188 xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
1189 "Frozen image may not be consistent.");
1191 * Just warn here till VFS can correctly support
1192 * read-only remount without racing.
1194 WARN_ON(atomic_read(&mp->m_active_trans) != 0);
1196 xfs_log_quiesce(mp);
1199 STATIC int
1200 xfs_fs_remount(
1201 struct super_block *sb,
1202 int *flags,
1203 char *options)
1205 struct xfs_mount *mp = XFS_M(sb);
1206 substring_t args[MAX_OPT_ARGS];
1207 char *p;
1208 int error;
1210 while ((p = strsep(&options, ",")) != NULL) {
1211 int token;
1213 if (!*p)
1214 continue;
1216 token = match_token(p, tokens, args);
1217 switch (token) {
1218 case Opt_barrier:
1219 mp->m_flags |= XFS_MOUNT_BARRIER;
1220 break;
1221 case Opt_nobarrier:
1222 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1223 break;
1224 case Opt_inode64:
1225 mp->m_maxagi = xfs_set_inode64(mp);
1226 break;
1227 case Opt_inode32:
1228 mp->m_maxagi = xfs_set_inode32(mp);
1229 break;
1230 default:
1232 * Logically we would return an error here to prevent
1233 * users from believing they might have changed
1234 * mount options using remount which can't be changed.
1236 * But unfortunately mount(8) adds all options from
1237 * mtab and fstab to the mount arguments in some cases
1238 * so we can't blindly reject options, but have to
1239 * check for each specified option if it actually
1240 * differs from the currently set option and only
1241 * reject it if that's the case.
1243 * Until that is implemented we return success for
1244 * every remount request, and silently ignore all
1245 * options that we can't actually change.
1247 #if 0
1248 xfs_info(mp,
1249 "mount option \"%s\" not supported for remount\n", p);
1250 return -EINVAL;
1251 #else
1252 break;
1253 #endif
1257 /* ro -> rw */
1258 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1259 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1262 * If this is the first remount to writeable state we
1263 * might have some superblock changes to update.
1265 if (mp->m_update_flags) {
1266 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1267 if (error) {
1268 xfs_warn(mp, "failed to write sb changes");
1269 return error;
1271 mp->m_update_flags = 0;
1275 * Fill out the reserve pool if it is empty. Use the stashed
1276 * value if it is non-zero, otherwise go with the default.
1278 xfs_restore_resvblks(mp);
1279 xfs_log_work_queue(mp);
1282 /* rw -> ro */
1283 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1285 * Before we sync the metadata, we need to free up the reserve
1286 * block pool so that the used block count in the superblock on
1287 * disk is correct at the end of the remount. Stash the current
1288 * reserve pool size so that if we get remounted rw, we can
1289 * return it to the same size.
1291 xfs_save_resvblks(mp);
1292 xfs_quiesce_attr(mp);
1293 mp->m_flags |= XFS_MOUNT_RDONLY;
1296 return 0;
1300 * Second stage of a freeze. The data is already frozen so we only
1301 * need to take care of the metadata. Once that's done write a dummy
1302 * record to dirty the log in case of a crash while frozen.
1304 STATIC int
1305 xfs_fs_freeze(
1306 struct super_block *sb)
1308 struct xfs_mount *mp = XFS_M(sb);
1310 xfs_save_resvblks(mp);
1311 xfs_quiesce_attr(mp);
1312 return -xfs_fs_log_dummy(mp);
1315 STATIC int
1316 xfs_fs_unfreeze(
1317 struct super_block *sb)
1319 struct xfs_mount *mp = XFS_M(sb);
1321 xfs_restore_resvblks(mp);
1322 xfs_log_work_queue(mp);
1323 return 0;
1326 STATIC int
1327 xfs_fs_show_options(
1328 struct seq_file *m,
1329 struct dentry *root)
1331 return -xfs_showargs(XFS_M(root->d_sb), m);
1335 * This function fills in xfs_mount_t fields based on mount args.
1336 * Note: the superblock _has_ now been read in.
1338 STATIC int
1339 xfs_finish_flags(
1340 struct xfs_mount *mp)
1342 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1344 /* Fail a mount where the logbuf is smaller than the log stripe */
1345 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1346 if (mp->m_logbsize <= 0 &&
1347 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1348 mp->m_logbsize = mp->m_sb.sb_logsunit;
1349 } else if (mp->m_logbsize > 0 &&
1350 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1351 xfs_warn(mp,
1352 "logbuf size must be greater than or equal to log stripe size");
1353 return XFS_ERROR(EINVAL);
1355 } else {
1356 /* Fail a mount if the logbuf is larger than 32K */
1357 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1358 xfs_warn(mp,
1359 "logbuf size for version 1 logs must be 16K or 32K");
1360 return XFS_ERROR(EINVAL);
1365 * V5 filesystems always use attr2 format for attributes.
1367 if (xfs_sb_version_hascrc(&mp->m_sb) &&
1368 (mp->m_flags & XFS_MOUNT_NOATTR2)) {
1369 xfs_warn(mp,
1370 "Cannot mount a V5 filesystem as %s. %s is always enabled for V5 filesystems.",
1371 MNTOPT_NOATTR2, MNTOPT_ATTR2);
1372 return XFS_ERROR(EINVAL);
1376 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1377 * told by noattr2 to turn it off
1379 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1380 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1381 mp->m_flags |= XFS_MOUNT_ATTR2;
1384 * prohibit r/w mounts of read-only filesystems
1386 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1387 xfs_warn(mp,
1388 "cannot mount a read-only filesystem as read-write");
1389 return XFS_ERROR(EROFS);
1392 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
1393 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) &&
1394 !xfs_sb_version_has_pquotino(&mp->m_sb)) {
1395 xfs_warn(mp,
1396 "Super block does not support project and group quota together");
1397 return XFS_ERROR(EINVAL);
1400 return 0;
1403 STATIC int
1404 xfs_fs_fill_super(
1405 struct super_block *sb,
1406 void *data,
1407 int silent)
1409 struct inode *root;
1410 struct xfs_mount *mp = NULL;
1411 int flags = 0, error = ENOMEM;
1413 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1414 if (!mp)
1415 goto out;
1417 spin_lock_init(&mp->m_sb_lock);
1418 mutex_init(&mp->m_growlock);
1419 atomic_set(&mp->m_active_trans, 0);
1420 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1421 INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
1423 mp->m_super = sb;
1424 sb->s_fs_info = mp;
1426 error = xfs_parseargs(mp, (char *)data);
1427 if (error)
1428 goto out_free_fsname;
1430 sb_min_blocksize(sb, BBSIZE);
1431 sb->s_xattr = xfs_xattr_handlers;
1432 sb->s_export_op = &xfs_export_operations;
1433 #ifdef CONFIG_XFS_QUOTA
1434 sb->s_qcop = &xfs_quotactl_operations;
1435 #endif
1436 sb->s_op = &xfs_super_operations;
1438 if (silent)
1439 flags |= XFS_MFSI_QUIET;
1441 error = xfs_open_devices(mp);
1442 if (error)
1443 goto out_free_fsname;
1445 error = xfs_init_mount_workqueues(mp);
1446 if (error)
1447 goto out_close_devices;
1449 error = xfs_icsb_init_counters(mp);
1450 if (error)
1451 goto out_destroy_workqueues;
1453 error = xfs_readsb(mp, flags);
1454 if (error)
1455 goto out_destroy_counters;
1457 error = xfs_finish_flags(mp);
1458 if (error)
1459 goto out_free_sb;
1461 error = xfs_setup_devices(mp);
1462 if (error)
1463 goto out_free_sb;
1465 error = xfs_filestream_mount(mp);
1466 if (error)
1467 goto out_free_sb;
1470 * we must configure the block size in the superblock before we run the
1471 * full mount process as the mount process can lookup and cache inodes.
1473 sb->s_magic = XFS_SB_MAGIC;
1474 sb->s_blocksize = mp->m_sb.sb_blocksize;
1475 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1476 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1477 sb->s_max_links = XFS_MAXLINK;
1478 sb->s_time_gran = 1;
1479 set_posix_acl_flag(sb);
1481 /* version 5 superblocks support inode version counters. */
1482 if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
1483 sb->s_flags |= MS_I_VERSION;
1485 error = xfs_mountfs(mp);
1486 if (error)
1487 goto out_filestream_unmount;
1489 root = igrab(VFS_I(mp->m_rootip));
1490 if (!root) {
1491 error = ENOENT;
1492 goto out_unmount;
1494 if (is_bad_inode(root)) {
1495 error = EINVAL;
1496 goto out_unmount;
1498 sb->s_root = d_make_root(root);
1499 if (!sb->s_root) {
1500 error = ENOMEM;
1501 goto out_unmount;
1504 return 0;
1506 out_filestream_unmount:
1507 xfs_filestream_unmount(mp);
1508 out_free_sb:
1509 xfs_freesb(mp);
1510 out_destroy_counters:
1511 xfs_icsb_destroy_counters(mp);
1512 out_destroy_workqueues:
1513 xfs_destroy_mount_workqueues(mp);
1514 out_close_devices:
1515 xfs_close_devices(mp);
1516 out_free_fsname:
1517 xfs_free_fsname(mp);
1518 kfree(mp);
1519 out:
1520 return -error;
1522 out_unmount:
1523 xfs_filestream_unmount(mp);
1524 xfs_unmountfs(mp);
1525 goto out_free_sb;
1528 STATIC struct dentry *
1529 xfs_fs_mount(
1530 struct file_system_type *fs_type,
1531 int flags,
1532 const char *dev_name,
1533 void *data)
1535 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1538 static long
1539 xfs_fs_nr_cached_objects(
1540 struct super_block *sb,
1541 int nid)
1543 return xfs_reclaim_inodes_count(XFS_M(sb));
1546 static long
1547 xfs_fs_free_cached_objects(
1548 struct super_block *sb,
1549 long nr_to_scan,
1550 int nid)
1552 return xfs_reclaim_inodes_nr(XFS_M(sb), nr_to_scan);
1555 static const struct super_operations xfs_super_operations = {
1556 .alloc_inode = xfs_fs_alloc_inode,
1557 .destroy_inode = xfs_fs_destroy_inode,
1558 .evict_inode = xfs_fs_evict_inode,
1559 .drop_inode = xfs_fs_drop_inode,
1560 .put_super = xfs_fs_put_super,
1561 .sync_fs = xfs_fs_sync_fs,
1562 .freeze_fs = xfs_fs_freeze,
1563 .unfreeze_fs = xfs_fs_unfreeze,
1564 .statfs = xfs_fs_statfs,
1565 .remount_fs = xfs_fs_remount,
1566 .show_options = xfs_fs_show_options,
1567 .nr_cached_objects = xfs_fs_nr_cached_objects,
1568 .free_cached_objects = xfs_fs_free_cached_objects,
1571 static struct file_system_type xfs_fs_type = {
1572 .owner = THIS_MODULE,
1573 .name = "xfs",
1574 .mount = xfs_fs_mount,
1575 .kill_sb = kill_block_super,
1576 .fs_flags = FS_REQUIRES_DEV,
1578 MODULE_ALIAS_FS("xfs");
1580 STATIC int __init
1581 xfs_init_zones(void)
1584 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1585 if (!xfs_ioend_zone)
1586 goto out;
1588 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1589 xfs_ioend_zone);
1590 if (!xfs_ioend_pool)
1591 goto out_destroy_ioend_zone;
1593 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1594 "xfs_log_ticket");
1595 if (!xfs_log_ticket_zone)
1596 goto out_destroy_ioend_pool;
1598 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1599 "xfs_bmap_free_item");
1600 if (!xfs_bmap_free_item_zone)
1601 goto out_destroy_log_ticket_zone;
1603 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1604 "xfs_btree_cur");
1605 if (!xfs_btree_cur_zone)
1606 goto out_destroy_bmap_free_item_zone;
1608 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1609 "xfs_da_state");
1610 if (!xfs_da_state_zone)
1611 goto out_destroy_btree_cur_zone;
1613 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1614 if (!xfs_ifork_zone)
1615 goto out_destroy_da_state_zone;
1617 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1618 if (!xfs_trans_zone)
1619 goto out_destroy_ifork_zone;
1621 xfs_log_item_desc_zone =
1622 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1623 "xfs_log_item_desc");
1624 if (!xfs_log_item_desc_zone)
1625 goto out_destroy_trans_zone;
1628 * The size of the zone allocated buf log item is the maximum
1629 * size possible under XFS. This wastes a little bit of memory,
1630 * but it is much faster.
1632 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item),
1633 "xfs_buf_item");
1634 if (!xfs_buf_item_zone)
1635 goto out_destroy_log_item_desc_zone;
1637 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1638 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1639 sizeof(xfs_extent_t))), "xfs_efd_item");
1640 if (!xfs_efd_zone)
1641 goto out_destroy_buf_item_zone;
1643 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1644 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1645 sizeof(xfs_extent_t))), "xfs_efi_item");
1646 if (!xfs_efi_zone)
1647 goto out_destroy_efd_zone;
1649 xfs_inode_zone =
1650 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1651 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1652 xfs_fs_inode_init_once);
1653 if (!xfs_inode_zone)
1654 goto out_destroy_efi_zone;
1656 xfs_ili_zone =
1657 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1658 KM_ZONE_SPREAD, NULL);
1659 if (!xfs_ili_zone)
1660 goto out_destroy_inode_zone;
1661 xfs_icreate_zone = kmem_zone_init(sizeof(struct xfs_icreate_item),
1662 "xfs_icr");
1663 if (!xfs_icreate_zone)
1664 goto out_destroy_ili_zone;
1666 return 0;
1668 out_destroy_ili_zone:
1669 kmem_zone_destroy(xfs_ili_zone);
1670 out_destroy_inode_zone:
1671 kmem_zone_destroy(xfs_inode_zone);
1672 out_destroy_efi_zone:
1673 kmem_zone_destroy(xfs_efi_zone);
1674 out_destroy_efd_zone:
1675 kmem_zone_destroy(xfs_efd_zone);
1676 out_destroy_buf_item_zone:
1677 kmem_zone_destroy(xfs_buf_item_zone);
1678 out_destroy_log_item_desc_zone:
1679 kmem_zone_destroy(xfs_log_item_desc_zone);
1680 out_destroy_trans_zone:
1681 kmem_zone_destroy(xfs_trans_zone);
1682 out_destroy_ifork_zone:
1683 kmem_zone_destroy(xfs_ifork_zone);
1684 out_destroy_da_state_zone:
1685 kmem_zone_destroy(xfs_da_state_zone);
1686 out_destroy_btree_cur_zone:
1687 kmem_zone_destroy(xfs_btree_cur_zone);
1688 out_destroy_bmap_free_item_zone:
1689 kmem_zone_destroy(xfs_bmap_free_item_zone);
1690 out_destroy_log_ticket_zone:
1691 kmem_zone_destroy(xfs_log_ticket_zone);
1692 out_destroy_ioend_pool:
1693 mempool_destroy(xfs_ioend_pool);
1694 out_destroy_ioend_zone:
1695 kmem_zone_destroy(xfs_ioend_zone);
1696 out:
1697 return -ENOMEM;
1700 STATIC void
1701 xfs_destroy_zones(void)
1704 * Make sure all delayed rcu free are flushed before we
1705 * destroy caches.
1707 rcu_barrier();
1708 kmem_zone_destroy(xfs_icreate_zone);
1709 kmem_zone_destroy(xfs_ili_zone);
1710 kmem_zone_destroy(xfs_inode_zone);
1711 kmem_zone_destroy(xfs_efi_zone);
1712 kmem_zone_destroy(xfs_efd_zone);
1713 kmem_zone_destroy(xfs_buf_item_zone);
1714 kmem_zone_destroy(xfs_log_item_desc_zone);
1715 kmem_zone_destroy(xfs_trans_zone);
1716 kmem_zone_destroy(xfs_ifork_zone);
1717 kmem_zone_destroy(xfs_da_state_zone);
1718 kmem_zone_destroy(xfs_btree_cur_zone);
1719 kmem_zone_destroy(xfs_bmap_free_item_zone);
1720 kmem_zone_destroy(xfs_log_ticket_zone);
1721 mempool_destroy(xfs_ioend_pool);
1722 kmem_zone_destroy(xfs_ioend_zone);
1726 STATIC int __init
1727 xfs_init_workqueues(void)
1730 * The allocation workqueue can be used in memory reclaim situations
1731 * (writepage path), and parallelism is only limited by the number of
1732 * AGs in all the filesystems mounted. Hence use the default large
1733 * max_active value for this workqueue.
1735 xfs_alloc_wq = alloc_workqueue("xfsalloc", WQ_MEM_RECLAIM, 0);
1736 if (!xfs_alloc_wq)
1737 return -ENOMEM;
1739 return 0;
1742 STATIC void
1743 xfs_destroy_workqueues(void)
1745 destroy_workqueue(xfs_alloc_wq);
1748 STATIC int __init
1749 init_xfs_fs(void)
1751 int error;
1753 printk(KERN_INFO XFS_VERSION_STRING " with "
1754 XFS_BUILD_OPTIONS " enabled\n");
1756 xfs_dir_startup();
1758 error = xfs_init_zones();
1759 if (error)
1760 goto out;
1762 error = xfs_init_workqueues();
1763 if (error)
1764 goto out_destroy_zones;
1766 error = xfs_mru_cache_init();
1767 if (error)
1768 goto out_destroy_wq;
1770 error = xfs_filestream_init();
1771 if (error)
1772 goto out_mru_cache_uninit;
1774 error = xfs_buf_init();
1775 if (error)
1776 goto out_filestream_uninit;
1778 error = xfs_init_procfs();
1779 if (error)
1780 goto out_buf_terminate;
1782 error = xfs_sysctl_register();
1783 if (error)
1784 goto out_cleanup_procfs;
1786 error = xfs_qm_init();
1787 if (error)
1788 goto out_sysctl_unregister;
1790 error = register_filesystem(&xfs_fs_type);
1791 if (error)
1792 goto out_qm_exit;
1793 return 0;
1795 out_qm_exit:
1796 xfs_qm_exit();
1797 out_sysctl_unregister:
1798 xfs_sysctl_unregister();
1799 out_cleanup_procfs:
1800 xfs_cleanup_procfs();
1801 out_buf_terminate:
1802 xfs_buf_terminate();
1803 out_filestream_uninit:
1804 xfs_filestream_uninit();
1805 out_mru_cache_uninit:
1806 xfs_mru_cache_uninit();
1807 out_destroy_wq:
1808 xfs_destroy_workqueues();
1809 out_destroy_zones:
1810 xfs_destroy_zones();
1811 out:
1812 return error;
1815 STATIC void __exit
1816 exit_xfs_fs(void)
1818 xfs_qm_exit();
1819 unregister_filesystem(&xfs_fs_type);
1820 xfs_sysctl_unregister();
1821 xfs_cleanup_procfs();
1822 xfs_buf_terminate();
1823 xfs_filestream_uninit();
1824 xfs_mru_cache_uninit();
1825 xfs_destroy_workqueues();
1826 xfs_destroy_zones();
1829 module_init(init_xfs_fs);
1830 module_exit(exit_xfs_fs);
1832 MODULE_AUTHOR("Silicon Graphics, Inc.");
1833 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1834 MODULE_LICENSE("GPL");