Merge tag 'berlin-dt-3.19-2' of git://git.infradead.org/users/hesselba/linux-berlin...
[linux-2.6/btrfs-unstable.git] / fs / xfs / xfs_super.c
blob9f622feda6a43bb32ea1983779a48d20b1446391
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_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 #include "xfs_sysfs.h"
52 #include <linux/namei.h>
53 #include <linux/init.h>
54 #include <linux/slab.h>
55 #include <linux/mount.h>
56 #include <linux/mempool.h>
57 #include <linux/writeback.h>
58 #include <linux/kthread.h>
59 #include <linux/freezer.h>
60 #include <linux/parser.h>
62 static const struct super_operations xfs_super_operations;
63 static kmem_zone_t *xfs_ioend_zone;
64 mempool_t *xfs_ioend_pool;
66 static struct kset *xfs_kset; /* top-level xfs sysfs dir */
67 #ifdef DEBUG
68 static struct xfs_kobj xfs_dbg_kobj; /* global debug sysfs attrs */
69 #endif
71 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
72 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
73 #define MNTOPT_LOGDEV "logdev" /* log device */
74 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
75 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
76 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
77 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
78 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
79 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
80 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
81 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
82 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
83 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
84 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
85 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
86 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
87 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
88 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
89 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
90 * unwritten extent conversion */
91 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
92 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
93 #define MNTOPT_32BITINODE "inode32" /* inode allocation limited to
94 * XFS_MAXINUMBER_32 */
95 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
96 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
97 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
98 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
99 * in stat(). */
100 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
101 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
102 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
103 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
104 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
105 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
106 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
107 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
108 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
109 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
110 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
111 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
112 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
113 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
114 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
115 #define MNTOPT_DELAYLOG "delaylog" /* Delayed logging enabled */
116 #define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed logging disabled */
117 #define MNTOPT_DISCARD "discard" /* Discard unused blocks */
118 #define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */
121 * Table driven mount option parser.
123 * Currently only used for remount, but it will be used for mount
124 * in the future, too.
126 enum {
127 Opt_barrier,
128 Opt_nobarrier,
129 Opt_inode64,
130 Opt_inode32,
131 Opt_err
134 static const match_table_t tokens = {
135 {Opt_barrier, "barrier"},
136 {Opt_nobarrier, "nobarrier"},
137 {Opt_inode64, "inode64"},
138 {Opt_inode32, "inode32"},
139 {Opt_err, NULL}
143 STATIC unsigned long
144 suffix_kstrtoint(char *s, unsigned int base, int *res)
146 int last, shift_left_factor = 0, _res;
147 char *value = s;
149 last = strlen(value) - 1;
150 if (value[last] == 'K' || value[last] == 'k') {
151 shift_left_factor = 10;
152 value[last] = '\0';
154 if (value[last] == 'M' || value[last] == 'm') {
155 shift_left_factor = 20;
156 value[last] = '\0';
158 if (value[last] == 'G' || value[last] == 'g') {
159 shift_left_factor = 30;
160 value[last] = '\0';
163 if (kstrtoint(s, base, &_res))
164 return -EINVAL;
165 *res = _res << shift_left_factor;
166 return 0;
170 * This function fills in xfs_mount_t fields based on mount args.
171 * Note: the superblock has _not_ yet been read in.
173 * Note that this function leaks the various device name allocations on
174 * failure. The caller takes care of them.
176 STATIC int
177 xfs_parseargs(
178 struct xfs_mount *mp,
179 char *options)
181 struct super_block *sb = mp->m_super;
182 char *this_char, *value;
183 int dsunit = 0;
184 int dswidth = 0;
185 int iosize = 0;
186 __uint8_t iosizelog = 0;
189 * set up the mount name first so all the errors will refer to the
190 * correct device.
192 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
193 if (!mp->m_fsname)
194 return -ENOMEM;
195 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
198 * Copy binary VFS mount flags we are interested in.
200 if (sb->s_flags & MS_RDONLY)
201 mp->m_flags |= XFS_MOUNT_RDONLY;
202 if (sb->s_flags & MS_DIRSYNC)
203 mp->m_flags |= XFS_MOUNT_DIRSYNC;
204 if (sb->s_flags & MS_SYNCHRONOUS)
205 mp->m_flags |= XFS_MOUNT_WSYNC;
208 * Set some default flags that could be cleared by the mount option
209 * parsing.
211 mp->m_flags |= XFS_MOUNT_BARRIER;
212 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
215 * These can be overridden by the mount option parsing.
217 mp->m_logbufs = -1;
218 mp->m_logbsize = -1;
220 if (!options)
221 goto done;
223 while ((this_char = strsep(&options, ",")) != NULL) {
224 if (!*this_char)
225 continue;
226 if ((value = strchr(this_char, '=')) != NULL)
227 *value++ = 0;
229 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
230 if (!value || !*value) {
231 xfs_warn(mp, "%s option requires an argument",
232 this_char);
233 return -EINVAL;
235 if (kstrtoint(value, 10, &mp->m_logbufs))
236 return -EINVAL;
237 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
238 if (!value || !*value) {
239 xfs_warn(mp, "%s option requires an argument",
240 this_char);
241 return -EINVAL;
243 if (suffix_kstrtoint(value, 10, &mp->m_logbsize))
244 return -EINVAL;
245 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
246 if (!value || !*value) {
247 xfs_warn(mp, "%s option requires an argument",
248 this_char);
249 return -EINVAL;
251 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
252 if (!mp->m_logname)
253 return -ENOMEM;
254 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
255 xfs_warn(mp, "%s option not allowed on this system",
256 this_char);
257 return -EINVAL;
258 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
259 if (!value || !*value) {
260 xfs_warn(mp, "%s option requires an argument",
261 this_char);
262 return -EINVAL;
264 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
265 if (!mp->m_rtname)
266 return -ENOMEM;
267 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
268 if (!value || !*value) {
269 xfs_warn(mp, "%s option requires an argument",
270 this_char);
271 return -EINVAL;
273 if (kstrtoint(value, 10, &iosize))
274 return -EINVAL;
275 iosizelog = ffs(iosize) - 1;
276 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
277 if (!value || !*value) {
278 xfs_warn(mp, "%s option requires an argument",
279 this_char);
280 return -EINVAL;
282 if (suffix_kstrtoint(value, 10, &iosize))
283 return -EINVAL;
284 iosizelog = ffs(iosize) - 1;
285 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
286 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
287 mp->m_flags |= XFS_MOUNT_GRPID;
288 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
289 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
290 mp->m_flags &= ~XFS_MOUNT_GRPID;
291 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
292 mp->m_flags |= XFS_MOUNT_WSYNC;
293 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
294 mp->m_flags |= XFS_MOUNT_NORECOVERY;
295 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
296 mp->m_flags |= XFS_MOUNT_NOALIGN;
297 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
298 mp->m_flags |= XFS_MOUNT_SWALLOC;
299 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
300 if (!value || !*value) {
301 xfs_warn(mp, "%s option requires an argument",
302 this_char);
303 return -EINVAL;
305 if (kstrtoint(value, 10, &dsunit))
306 return -EINVAL;
307 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
308 if (!value || !*value) {
309 xfs_warn(mp, "%s option requires an argument",
310 this_char);
311 return -EINVAL;
313 if (kstrtoint(value, 10, &dswidth))
314 return -EINVAL;
315 } else if (!strcmp(this_char, MNTOPT_32BITINODE)) {
316 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
317 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
318 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
319 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
320 mp->m_flags |= XFS_MOUNT_NOUUID;
321 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
322 mp->m_flags |= XFS_MOUNT_BARRIER;
323 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
324 mp->m_flags &= ~XFS_MOUNT_BARRIER;
325 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
326 mp->m_flags |= XFS_MOUNT_IKEEP;
327 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
328 mp->m_flags &= ~XFS_MOUNT_IKEEP;
329 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
330 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
331 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
332 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
333 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
334 mp->m_flags |= XFS_MOUNT_ATTR2;
335 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
336 mp->m_flags &= ~XFS_MOUNT_ATTR2;
337 mp->m_flags |= XFS_MOUNT_NOATTR2;
338 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
339 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
340 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
341 mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT;
342 mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD;
343 mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE;
344 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
345 !strcmp(this_char, MNTOPT_UQUOTA) ||
346 !strcmp(this_char, MNTOPT_USRQUOTA)) {
347 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
348 XFS_UQUOTA_ENFD);
349 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
350 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
351 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
352 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
353 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
354 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
355 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
356 XFS_PQUOTA_ENFD);
357 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
358 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
359 mp->m_qflags &= ~XFS_PQUOTA_ENFD;
360 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
361 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
362 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
363 XFS_GQUOTA_ENFD);
364 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
365 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
366 mp->m_qflags &= ~XFS_GQUOTA_ENFD;
367 } else if (!strcmp(this_char, MNTOPT_DELAYLOG)) {
368 xfs_warn(mp,
369 "delaylog is the default now, option is deprecated.");
370 } else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) {
371 xfs_warn(mp,
372 "nodelaylog support has been removed, option is deprecated.");
373 } else if (!strcmp(this_char, MNTOPT_DISCARD)) {
374 mp->m_flags |= XFS_MOUNT_DISCARD;
375 } else if (!strcmp(this_char, MNTOPT_NODISCARD)) {
376 mp->m_flags &= ~XFS_MOUNT_DISCARD;
377 } else if (!strcmp(this_char, "ihashsize")) {
378 xfs_warn(mp,
379 "ihashsize no longer used, option is deprecated.");
380 } else if (!strcmp(this_char, "osyncisdsync")) {
381 xfs_warn(mp,
382 "osyncisdsync has no effect, option is deprecated.");
383 } else if (!strcmp(this_char, "osyncisosync")) {
384 xfs_warn(mp,
385 "osyncisosync has no effect, option is deprecated.");
386 } else if (!strcmp(this_char, "irixsgid")) {
387 xfs_warn(mp,
388 "irixsgid is now a sysctl(2) variable, option is deprecated.");
389 } else {
390 xfs_warn(mp, "unknown mount option [%s].", this_char);
391 return -EINVAL;
396 * no recovery flag requires a read-only mount
398 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
399 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
400 xfs_warn(mp, "no-recovery mounts must be read-only.");
401 return -EINVAL;
404 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
405 xfs_warn(mp,
406 "sunit and swidth options incompatible with the noalign option");
407 return -EINVAL;
410 #ifndef CONFIG_XFS_QUOTA
411 if (XFS_IS_QUOTA_RUNNING(mp)) {
412 xfs_warn(mp, "quota support not available in this kernel.");
413 return -EINVAL;
415 #endif
417 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
418 xfs_warn(mp, "sunit and swidth must be specified together");
419 return -EINVAL;
422 if (dsunit && (dswidth % dsunit != 0)) {
423 xfs_warn(mp,
424 "stripe width (%d) must be a multiple of the stripe unit (%d)",
425 dswidth, dsunit);
426 return -EINVAL;
429 done:
430 if (dsunit && !(mp->m_flags & XFS_MOUNT_NOALIGN)) {
432 * At this point the superblock has not been read
433 * in, therefore we do not know the block size.
434 * Before the mount call ends we will convert
435 * these to FSBs.
437 mp->m_dalign = dsunit;
438 mp->m_swidth = dswidth;
441 if (mp->m_logbufs != -1 &&
442 mp->m_logbufs != 0 &&
443 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
444 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
445 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
446 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
447 return -EINVAL;
449 if (mp->m_logbsize != -1 &&
450 mp->m_logbsize != 0 &&
451 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
452 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
453 !is_power_of_2(mp->m_logbsize))) {
454 xfs_warn(mp,
455 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
456 mp->m_logbsize);
457 return -EINVAL;
460 if (iosizelog) {
461 if (iosizelog > XFS_MAX_IO_LOG ||
462 iosizelog < XFS_MIN_IO_LOG) {
463 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]",
464 iosizelog, XFS_MIN_IO_LOG,
465 XFS_MAX_IO_LOG);
466 return -EINVAL;
469 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
470 mp->m_readio_log = iosizelog;
471 mp->m_writeio_log = iosizelog;
474 return 0;
477 struct proc_xfs_info {
478 int flag;
479 char *str;
482 STATIC int
483 xfs_showargs(
484 struct xfs_mount *mp,
485 struct seq_file *m)
487 static struct proc_xfs_info xfs_info_set[] = {
488 /* the few simple ones we can get from the mount struct */
489 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
490 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
491 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
492 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
493 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
494 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
495 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
496 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
497 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
498 { XFS_MOUNT_DISCARD, "," MNTOPT_DISCARD },
499 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_32BITINODE },
500 { 0, NULL }
502 static struct proc_xfs_info xfs_info_unset[] = {
503 /* the few simple ones we can get from the mount struct */
504 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
505 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
506 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
507 { 0, NULL }
509 struct proc_xfs_info *xfs_infop;
511 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
512 if (mp->m_flags & xfs_infop->flag)
513 seq_puts(m, xfs_infop->str);
515 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
516 if (!(mp->m_flags & xfs_infop->flag))
517 seq_puts(m, xfs_infop->str);
520 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
521 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
522 (int)(1 << mp->m_writeio_log) >> 10);
524 if (mp->m_logbufs > 0)
525 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
526 if (mp->m_logbsize > 0)
527 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
529 if (mp->m_logname)
530 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
531 if (mp->m_rtname)
532 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
534 if (mp->m_dalign > 0)
535 seq_printf(m, "," MNTOPT_SUNIT "=%d",
536 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
537 if (mp->m_swidth > 0)
538 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
539 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
541 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
542 seq_puts(m, "," MNTOPT_USRQUOTA);
543 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
544 seq_puts(m, "," MNTOPT_UQUOTANOENF);
546 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
547 if (mp->m_qflags & XFS_PQUOTA_ENFD)
548 seq_puts(m, "," MNTOPT_PRJQUOTA);
549 else
550 seq_puts(m, "," MNTOPT_PQUOTANOENF);
552 if (mp->m_qflags & XFS_GQUOTA_ACCT) {
553 if (mp->m_qflags & XFS_GQUOTA_ENFD)
554 seq_puts(m, "," MNTOPT_GRPQUOTA);
555 else
556 seq_puts(m, "," MNTOPT_GQUOTANOENF);
559 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
560 seq_puts(m, "," MNTOPT_NOQUOTA);
562 return 0;
564 __uint64_t
565 xfs_max_file_offset(
566 unsigned int blockshift)
568 unsigned int pagefactor = 1;
569 unsigned int bitshift = BITS_PER_LONG - 1;
571 /* Figure out maximum filesize, on Linux this can depend on
572 * the filesystem blocksize (on 32 bit platforms).
573 * __block_write_begin does this in an [unsigned] long...
574 * page->index << (PAGE_CACHE_SHIFT - bbits)
575 * So, for page sized blocks (4K on 32 bit platforms),
576 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
577 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
578 * but for smaller blocksizes it is less (bbits = log2 bsize).
579 * Note1: get_block_t takes a long (implicit cast from above)
580 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
581 * can optionally convert the [unsigned] long from above into
582 * an [unsigned] long long.
585 #if BITS_PER_LONG == 32
586 # if defined(CONFIG_LBDAF)
587 ASSERT(sizeof(sector_t) == 8);
588 pagefactor = PAGE_CACHE_SIZE;
589 bitshift = BITS_PER_LONG;
590 # else
591 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
592 # endif
593 #endif
595 return (((__uint64_t)pagefactor) << bitshift) - 1;
599 * xfs_set_inode32() and xfs_set_inode64() are passed an agcount
600 * because in the growfs case, mp->m_sb.sb_agcount is not updated
601 * yet to the potentially higher ag count.
603 xfs_agnumber_t
604 xfs_set_inode32(struct xfs_mount *mp, xfs_agnumber_t agcount)
606 xfs_agnumber_t index = 0;
607 xfs_agnumber_t maxagi = 0;
608 xfs_sb_t *sbp = &mp->m_sb;
609 xfs_agnumber_t max_metadata;
610 xfs_agino_t agino;
611 xfs_ino_t ino;
612 xfs_perag_t *pag;
614 /* Calculate how much should be reserved for inodes to meet
615 * the max inode percentage.
617 if (mp->m_maxicount) {
618 __uint64_t icount;
620 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
621 do_div(icount, 100);
622 icount += sbp->sb_agblocks - 1;
623 do_div(icount, sbp->sb_agblocks);
624 max_metadata = icount;
625 } else {
626 max_metadata = agcount;
629 agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks - 1, 0);
631 for (index = 0; index < 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, xfs_agnumber_t agcount)
658 xfs_agnumber_t index = 0;
660 for (index = 0; index < 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);
774 if (!mp->m_ddev_targp)
775 goto out_close_rtdev;
777 if (rtdev) {
778 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev);
779 if (!mp->m_rtdev_targp)
780 goto out_free_ddev_targ;
783 if (logdev && logdev != ddev) {
784 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev);
785 if (!mp->m_logdev_targp)
786 goto out_free_rtdev_targ;
787 } else {
788 mp->m_logdev_targp = mp->m_ddev_targp;
791 return 0;
793 out_free_rtdev_targ:
794 if (mp->m_rtdev_targp)
795 xfs_free_buftarg(mp, mp->m_rtdev_targp);
796 out_free_ddev_targ:
797 xfs_free_buftarg(mp, mp->m_ddev_targp);
798 out_close_rtdev:
799 if (rtdev)
800 xfs_blkdev_put(rtdev);
801 out_close_logdev:
802 if (logdev && logdev != ddev)
803 xfs_blkdev_put(logdev);
804 out:
805 return error;
809 * Setup xfs_mount buffer target pointers based on superblock
811 STATIC int
812 xfs_setup_devices(
813 struct xfs_mount *mp)
815 int error;
817 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize);
818 if (error)
819 return error;
821 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
822 unsigned int log_sector_size = BBSIZE;
824 if (xfs_sb_version_hassector(&mp->m_sb))
825 log_sector_size = mp->m_sb.sb_logsectsize;
826 error = xfs_setsize_buftarg(mp->m_logdev_targp,
827 log_sector_size);
828 if (error)
829 return error;
831 if (mp->m_rtdev_targp) {
832 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
833 mp->m_sb.sb_sectsize);
834 if (error)
835 return error;
838 return 0;
841 STATIC int
842 xfs_init_mount_workqueues(
843 struct xfs_mount *mp)
845 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s",
846 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
847 if (!mp->m_data_workqueue)
848 goto out;
850 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
851 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
852 if (!mp->m_unwritten_workqueue)
853 goto out_destroy_data_iodone_queue;
855 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s",
856 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
857 if (!mp->m_cil_workqueue)
858 goto out_destroy_unwritten;
860 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
861 WQ_FREEZABLE, 0, mp->m_fsname);
862 if (!mp->m_reclaim_workqueue)
863 goto out_destroy_cil;
865 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
866 WQ_FREEZABLE, 0, mp->m_fsname);
867 if (!mp->m_log_workqueue)
868 goto out_destroy_reclaim;
870 mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s",
871 WQ_FREEZABLE, 0, mp->m_fsname);
872 if (!mp->m_eofblocks_workqueue)
873 goto out_destroy_log;
875 return 0;
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;
891 STATIC void
892 xfs_destroy_mount_workqueues(
893 struct xfs_mount *mp)
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);
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.
909 void
910 xfs_flush_inodes(
911 struct xfs_mount *mp)
913 struct super_block *sb = mp->m_super;
915 if (down_read_trylock(&sb->s_umount)) {
916 sync_inodes_sb(sb);
917 up_read(&sb->s_umount);
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)
926 BUG();
927 return NULL;
931 * Now that the generic code is guaranteed not to be accessing
932 * the linux inode, we can reclaim the inode.
934 STATIC void
935 xfs_fs_destroy_inode(
936 struct inode *inode)
938 struct xfs_inode *ip = XFS_I(inode);
940 trace_xfs_destroy_inode(ip);
942 XFS_STATS_INC(vn_reclaim);
944 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
947 * We should never get here with one of the reclaim flags already set.
949 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
950 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
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.
959 xfs_inode_set_reclaim_tag(ip);
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.
970 STATIC void
971 xfs_fs_inode_init_once(
972 void *inode)
974 struct xfs_inode *ip = inode;
976 memset(ip, 0, sizeof(struct xfs_inode));
978 /* vfs inode */
979 inode_init_once(VFS_I(ip));
981 /* xfs inode */
982 atomic_set(&ip->i_pincount, 0);
983 spin_lock_init(&ip->i_flags_lock);
985 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
986 "xfsino", ip->i_ino);
989 STATIC void
990 xfs_fs_evict_inode(
991 struct inode *inode)
993 xfs_inode_t *ip = XFS_I(inode);
995 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
997 trace_xfs_evict_inode(ip);
999 truncate_inode_pages_final(&inode->i_data);
1000 clear_inode(inode);
1001 XFS_STATS_INC(vn_rele);
1002 XFS_STATS_INC(vn_remove);
1003 XFS_STATS_DEC(vn_active);
1005 xfs_inactive(ip);
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.
1015 STATIC int
1016 xfs_fs_drop_inode(
1017 struct inode *inode)
1019 struct xfs_inode *ip = XFS_I(inode);
1021 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);
1024 STATIC void
1025 xfs_free_fsname(
1026 struct xfs_mount *mp)
1028 kfree(mp->m_fsname);
1029 kfree(mp->m_rtname);
1030 kfree(mp->m_logname);
1033 STATIC void
1034 xfs_fs_put_super(
1035 struct super_block *sb)
1037 struct xfs_mount *mp = XFS_M(sb);
1039 xfs_filestream_unmount(mp);
1040 xfs_unmountfs(mp);
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);
1050 STATIC int
1051 xfs_fs_sync_fs(
1052 struct super_block *sb,
1053 int wait)
1055 struct xfs_mount *mp = XFS_M(sb);
1058 * Doing anything during the async pass would be counterproductive.
1060 if (!wait)
1061 return 0;
1063 xfs_log_force(mp, XFS_LOG_SYNC);
1064 if (laptop_mode) {
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).
1070 flush_delayed_work(&mp->m_log->l_work);
1073 return 0;
1076 STATIC int
1077 xfs_fs_statfs(
1078 struct dentry *dentry,
1079 struct kstatfs *statp)
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;
1088 statp->f_type = XFS_SB_MAGIC;
1089 statp->f_namelen = MAXNAMELEN - 1;
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);
1095 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
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);
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);
1115 spin_unlock(&mp->m_sb_lock);
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;
1124 STATIC void
1125 xfs_save_resvblks(struct xfs_mount *mp)
1127 __uint64_t resblks = 0;
1129 mp->m_resblks_save = mp->m_resblks;
1130 xfs_reserve_blocks(mp, &resblks, NULL);
1133 STATIC void
1134 xfs_restore_resvblks(struct xfs_mount *mp)
1136 __uint64_t resblks;
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);
1144 xfs_reserve_blocks(mp, &resblks, NULL);
1148 * Trigger writeback of all the dirty metadata in the file system.
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.
1155 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1156 * it is started again when appropriate.
1158 static void
1159 xfs_quiesce_attr(
1160 struct xfs_mount *mp)
1162 int error = 0;
1164 /* wait for all modifications to complete */
1165 while (atomic_read(&mp->m_active_trans) > 0)
1166 delay(100);
1168 /* force the log to unpin objects from the now complete transactions */
1169 xfs_log_force(mp, XFS_LOG_SYNC);
1171 /* reclaim inodes to do any IO before the freeze completes */
1172 xfs_reclaim_inodes(mp, 0);
1173 xfs_reclaim_inodes(mp, SYNC_WAIT);
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.");
1181 * Just warn here till VFS can correctly support
1182 * read-only remount without racing.
1184 WARN_ON(atomic_read(&mp->m_active_trans) != 0);
1186 xfs_log_quiesce(mp);
1189 STATIC int
1190 xfs_fs_remount(
1191 struct super_block *sb,
1192 int *flags,
1193 char *options)
1195 struct xfs_mount *mp = XFS_M(sb);
1196 xfs_sb_t *sbp = &mp->m_sb;
1197 substring_t args[MAX_OPT_ARGS];
1198 char *p;
1199 int error;
1201 sync_filesystem(sb);
1202 while ((p = strsep(&options, ",")) != NULL) {
1203 int token;
1205 if (!*p)
1206 continue;
1208 token = match_token(p, tokens, args);
1209 switch (token) {
1210 case Opt_barrier:
1211 mp->m_flags |= XFS_MOUNT_BARRIER;
1212 break;
1213 case Opt_nobarrier:
1214 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1215 break;
1216 case Opt_inode64:
1217 mp->m_maxagi = xfs_set_inode64(mp, sbp->sb_agcount);
1218 break;
1219 case Opt_inode32:
1220 mp->m_maxagi = xfs_set_inode32(mp, sbp->sb_agcount);
1221 break;
1222 default:
1224 * Logically we would return an error here to prevent
1225 * users from believing they might have changed
1226 * mount options using remount which can't be changed.
1228 * But unfortunately mount(8) adds all options from
1229 * mtab and fstab to the mount arguments in some cases
1230 * so we can't blindly reject options, but have to
1231 * check for each specified option if it actually
1232 * differs from the currently set option and only
1233 * reject it if that's the case.
1235 * Until that is implemented we return success for
1236 * every remount request, and silently ignore all
1237 * options that we can't actually change.
1239 #if 0
1240 xfs_info(mp,
1241 "mount option \"%s\" not supported for remount", p);
1242 return -EINVAL;
1243 #else
1244 break;
1245 #endif
1249 /* ro -> rw */
1250 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1251 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1254 * If this is the first remount to writeable state we
1255 * might have some superblock changes to update.
1257 if (mp->m_update_flags) {
1258 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1259 if (error) {
1260 xfs_warn(mp, "failed to write sb changes");
1261 return error;
1263 mp->m_update_flags = 0;
1267 * Fill out the reserve pool if it is empty. Use the stashed
1268 * value if it is non-zero, otherwise go with the default.
1270 xfs_restore_resvblks(mp);
1271 xfs_log_work_queue(mp);
1274 /* rw -> ro */
1275 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1277 * Before we sync the metadata, we need to free up the reserve
1278 * block pool so that the used block count in the superblock on
1279 * disk is correct at the end of the remount. Stash the current
1280 * reserve pool size so that if we get remounted rw, we can
1281 * return it to the same size.
1283 xfs_save_resvblks(mp);
1284 xfs_quiesce_attr(mp);
1285 mp->m_flags |= XFS_MOUNT_RDONLY;
1288 return 0;
1292 * Second stage of a freeze. The data is already frozen so we only
1293 * need to take care of the metadata. Once that's done write a dummy
1294 * record to dirty the log in case of a crash while frozen.
1296 STATIC int
1297 xfs_fs_freeze(
1298 struct super_block *sb)
1300 struct xfs_mount *mp = XFS_M(sb);
1302 xfs_save_resvblks(mp);
1303 xfs_quiesce_attr(mp);
1304 return xfs_fs_log_dummy(mp);
1307 STATIC int
1308 xfs_fs_unfreeze(
1309 struct super_block *sb)
1311 struct xfs_mount *mp = XFS_M(sb);
1313 xfs_restore_resvblks(mp);
1314 xfs_log_work_queue(mp);
1315 return 0;
1318 STATIC int
1319 xfs_fs_show_options(
1320 struct seq_file *m,
1321 struct dentry *root)
1323 return xfs_showargs(XFS_M(root->d_sb), m);
1327 * This function fills in xfs_mount_t fields based on mount args.
1328 * Note: the superblock _has_ now been read in.
1330 STATIC int
1331 xfs_finish_flags(
1332 struct xfs_mount *mp)
1334 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1336 /* Fail a mount where the logbuf is smaller than the log stripe */
1337 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1338 if (mp->m_logbsize <= 0 &&
1339 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1340 mp->m_logbsize = mp->m_sb.sb_logsunit;
1341 } else if (mp->m_logbsize > 0 &&
1342 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1343 xfs_warn(mp,
1344 "logbuf size must be greater than or equal to log stripe size");
1345 return -EINVAL;
1347 } else {
1348 /* Fail a mount if the logbuf is larger than 32K */
1349 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1350 xfs_warn(mp,
1351 "logbuf size for version 1 logs must be 16K or 32K");
1352 return -EINVAL;
1357 * V5 filesystems always use attr2 format for attributes.
1359 if (xfs_sb_version_hascrc(&mp->m_sb) &&
1360 (mp->m_flags & XFS_MOUNT_NOATTR2)) {
1361 xfs_warn(mp,
1362 "Cannot mount a V5 filesystem as %s. %s is always enabled for V5 filesystems.",
1363 MNTOPT_NOATTR2, MNTOPT_ATTR2);
1364 return -EINVAL;
1368 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1369 * told by noattr2 to turn it off
1371 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1372 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1373 mp->m_flags |= XFS_MOUNT_ATTR2;
1376 * prohibit r/w mounts of read-only filesystems
1378 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1379 xfs_warn(mp,
1380 "cannot mount a read-only filesystem as read-write");
1381 return -EROFS;
1384 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
1385 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) &&
1386 !xfs_sb_version_has_pquotino(&mp->m_sb)) {
1387 xfs_warn(mp,
1388 "Super block does not support project and group quota together");
1389 return -EINVAL;
1392 return 0;
1395 STATIC int
1396 xfs_fs_fill_super(
1397 struct super_block *sb,
1398 void *data,
1399 int silent)
1401 struct inode *root;
1402 struct xfs_mount *mp = NULL;
1403 int flags = 0, error = -ENOMEM;
1405 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1406 if (!mp)
1407 goto out;
1409 spin_lock_init(&mp->m_sb_lock);
1410 mutex_init(&mp->m_growlock);
1411 atomic_set(&mp->m_active_trans, 0);
1412 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1413 INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
1414 mp->m_kobj.kobject.kset = xfs_kset;
1416 mp->m_super = sb;
1417 sb->s_fs_info = mp;
1419 error = xfs_parseargs(mp, (char *)data);
1420 if (error)
1421 goto out_free_fsname;
1423 sb_min_blocksize(sb, BBSIZE);
1424 sb->s_xattr = xfs_xattr_handlers;
1425 sb->s_export_op = &xfs_export_operations;
1426 #ifdef CONFIG_XFS_QUOTA
1427 sb->s_qcop = &xfs_quotactl_operations;
1428 #endif
1429 sb->s_op = &xfs_super_operations;
1431 if (silent)
1432 flags |= XFS_MFSI_QUIET;
1434 error = xfs_open_devices(mp);
1435 if (error)
1436 goto out_free_fsname;
1438 error = xfs_init_mount_workqueues(mp);
1439 if (error)
1440 goto out_close_devices;
1442 error = xfs_icsb_init_counters(mp);
1443 if (error)
1444 goto out_destroy_workqueues;
1446 error = xfs_readsb(mp, flags);
1447 if (error)
1448 goto out_destroy_counters;
1450 error = xfs_finish_flags(mp);
1451 if (error)
1452 goto out_free_sb;
1454 error = xfs_setup_devices(mp);
1455 if (error)
1456 goto out_free_sb;
1458 error = xfs_filestream_mount(mp);
1459 if (error)
1460 goto out_free_sb;
1463 * we must configure the block size in the superblock before we run the
1464 * full mount process as the mount process can lookup and cache inodes.
1466 sb->s_magic = XFS_SB_MAGIC;
1467 sb->s_blocksize = mp->m_sb.sb_blocksize;
1468 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1469 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1470 sb->s_max_links = XFS_MAXLINK;
1471 sb->s_time_gran = 1;
1472 set_posix_acl_flag(sb);
1474 /* version 5 superblocks support inode version counters. */
1475 if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
1476 sb->s_flags |= MS_I_VERSION;
1478 error = xfs_mountfs(mp);
1479 if (error)
1480 goto out_filestream_unmount;
1482 root = igrab(VFS_I(mp->m_rootip));
1483 if (!root) {
1484 error = -ENOENT;
1485 goto out_unmount;
1487 sb->s_root = d_make_root(root);
1488 if (!sb->s_root) {
1489 error = -ENOMEM;
1490 goto out_unmount;
1493 return 0;
1495 out_filestream_unmount:
1496 xfs_filestream_unmount(mp);
1497 out_free_sb:
1498 xfs_freesb(mp);
1499 out_destroy_counters:
1500 xfs_icsb_destroy_counters(mp);
1501 out_destroy_workqueues:
1502 xfs_destroy_mount_workqueues(mp);
1503 out_close_devices:
1504 xfs_close_devices(mp);
1505 out_free_fsname:
1506 xfs_free_fsname(mp);
1507 kfree(mp);
1508 out:
1509 return error;
1511 out_unmount:
1512 xfs_filestream_unmount(mp);
1513 xfs_unmountfs(mp);
1514 goto out_free_sb;
1517 STATIC struct dentry *
1518 xfs_fs_mount(
1519 struct file_system_type *fs_type,
1520 int flags,
1521 const char *dev_name,
1522 void *data)
1524 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1527 static long
1528 xfs_fs_nr_cached_objects(
1529 struct super_block *sb,
1530 int nid)
1532 return xfs_reclaim_inodes_count(XFS_M(sb));
1535 static long
1536 xfs_fs_free_cached_objects(
1537 struct super_block *sb,
1538 long nr_to_scan,
1539 int nid)
1541 return xfs_reclaim_inodes_nr(XFS_M(sb), nr_to_scan);
1544 static const struct super_operations xfs_super_operations = {
1545 .alloc_inode = xfs_fs_alloc_inode,
1546 .destroy_inode = xfs_fs_destroy_inode,
1547 .evict_inode = xfs_fs_evict_inode,
1548 .drop_inode = xfs_fs_drop_inode,
1549 .put_super = xfs_fs_put_super,
1550 .sync_fs = xfs_fs_sync_fs,
1551 .freeze_fs = xfs_fs_freeze,
1552 .unfreeze_fs = xfs_fs_unfreeze,
1553 .statfs = xfs_fs_statfs,
1554 .remount_fs = xfs_fs_remount,
1555 .show_options = xfs_fs_show_options,
1556 .nr_cached_objects = xfs_fs_nr_cached_objects,
1557 .free_cached_objects = xfs_fs_free_cached_objects,
1560 static struct file_system_type xfs_fs_type = {
1561 .owner = THIS_MODULE,
1562 .name = "xfs",
1563 .mount = xfs_fs_mount,
1564 .kill_sb = kill_block_super,
1565 .fs_flags = FS_REQUIRES_DEV,
1567 MODULE_ALIAS_FS("xfs");
1569 STATIC int __init
1570 xfs_init_zones(void)
1573 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1574 if (!xfs_ioend_zone)
1575 goto out;
1577 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1578 xfs_ioend_zone);
1579 if (!xfs_ioend_pool)
1580 goto out_destroy_ioend_zone;
1582 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1583 "xfs_log_ticket");
1584 if (!xfs_log_ticket_zone)
1585 goto out_destroy_ioend_pool;
1587 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1588 "xfs_bmap_free_item");
1589 if (!xfs_bmap_free_item_zone)
1590 goto out_destroy_log_ticket_zone;
1592 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1593 "xfs_btree_cur");
1594 if (!xfs_btree_cur_zone)
1595 goto out_destroy_bmap_free_item_zone;
1597 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1598 "xfs_da_state");
1599 if (!xfs_da_state_zone)
1600 goto out_destroy_btree_cur_zone;
1602 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1603 if (!xfs_ifork_zone)
1604 goto out_destroy_da_state_zone;
1606 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1607 if (!xfs_trans_zone)
1608 goto out_destroy_ifork_zone;
1610 xfs_log_item_desc_zone =
1611 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1612 "xfs_log_item_desc");
1613 if (!xfs_log_item_desc_zone)
1614 goto out_destroy_trans_zone;
1617 * The size of the zone allocated buf log item is the maximum
1618 * size possible under XFS. This wastes a little bit of memory,
1619 * but it is much faster.
1621 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item),
1622 "xfs_buf_item");
1623 if (!xfs_buf_item_zone)
1624 goto out_destroy_log_item_desc_zone;
1626 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1627 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1628 sizeof(xfs_extent_t))), "xfs_efd_item");
1629 if (!xfs_efd_zone)
1630 goto out_destroy_buf_item_zone;
1632 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1633 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1634 sizeof(xfs_extent_t))), "xfs_efi_item");
1635 if (!xfs_efi_zone)
1636 goto out_destroy_efd_zone;
1638 xfs_inode_zone =
1639 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1640 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1641 xfs_fs_inode_init_once);
1642 if (!xfs_inode_zone)
1643 goto out_destroy_efi_zone;
1645 xfs_ili_zone =
1646 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1647 KM_ZONE_SPREAD, NULL);
1648 if (!xfs_ili_zone)
1649 goto out_destroy_inode_zone;
1650 xfs_icreate_zone = kmem_zone_init(sizeof(struct xfs_icreate_item),
1651 "xfs_icr");
1652 if (!xfs_icreate_zone)
1653 goto out_destroy_ili_zone;
1655 return 0;
1657 out_destroy_ili_zone:
1658 kmem_zone_destroy(xfs_ili_zone);
1659 out_destroy_inode_zone:
1660 kmem_zone_destroy(xfs_inode_zone);
1661 out_destroy_efi_zone:
1662 kmem_zone_destroy(xfs_efi_zone);
1663 out_destroy_efd_zone:
1664 kmem_zone_destroy(xfs_efd_zone);
1665 out_destroy_buf_item_zone:
1666 kmem_zone_destroy(xfs_buf_item_zone);
1667 out_destroy_log_item_desc_zone:
1668 kmem_zone_destroy(xfs_log_item_desc_zone);
1669 out_destroy_trans_zone:
1670 kmem_zone_destroy(xfs_trans_zone);
1671 out_destroy_ifork_zone:
1672 kmem_zone_destroy(xfs_ifork_zone);
1673 out_destroy_da_state_zone:
1674 kmem_zone_destroy(xfs_da_state_zone);
1675 out_destroy_btree_cur_zone:
1676 kmem_zone_destroy(xfs_btree_cur_zone);
1677 out_destroy_bmap_free_item_zone:
1678 kmem_zone_destroy(xfs_bmap_free_item_zone);
1679 out_destroy_log_ticket_zone:
1680 kmem_zone_destroy(xfs_log_ticket_zone);
1681 out_destroy_ioend_pool:
1682 mempool_destroy(xfs_ioend_pool);
1683 out_destroy_ioend_zone:
1684 kmem_zone_destroy(xfs_ioend_zone);
1685 out:
1686 return -ENOMEM;
1689 STATIC void
1690 xfs_destroy_zones(void)
1693 * Make sure all delayed rcu free are flushed before we
1694 * destroy caches.
1696 rcu_barrier();
1697 kmem_zone_destroy(xfs_icreate_zone);
1698 kmem_zone_destroy(xfs_ili_zone);
1699 kmem_zone_destroy(xfs_inode_zone);
1700 kmem_zone_destroy(xfs_efi_zone);
1701 kmem_zone_destroy(xfs_efd_zone);
1702 kmem_zone_destroy(xfs_buf_item_zone);
1703 kmem_zone_destroy(xfs_log_item_desc_zone);
1704 kmem_zone_destroy(xfs_trans_zone);
1705 kmem_zone_destroy(xfs_ifork_zone);
1706 kmem_zone_destroy(xfs_da_state_zone);
1707 kmem_zone_destroy(xfs_btree_cur_zone);
1708 kmem_zone_destroy(xfs_bmap_free_item_zone);
1709 kmem_zone_destroy(xfs_log_ticket_zone);
1710 mempool_destroy(xfs_ioend_pool);
1711 kmem_zone_destroy(xfs_ioend_zone);
1715 STATIC int __init
1716 xfs_init_workqueues(void)
1719 * The allocation workqueue can be used in memory reclaim situations
1720 * (writepage path), and parallelism is only limited by the number of
1721 * AGs in all the filesystems mounted. Hence use the default large
1722 * max_active value for this workqueue.
1724 xfs_alloc_wq = alloc_workqueue("xfsalloc",
1725 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0);
1726 if (!xfs_alloc_wq)
1727 return -ENOMEM;
1729 return 0;
1732 STATIC void
1733 xfs_destroy_workqueues(void)
1735 destroy_workqueue(xfs_alloc_wq);
1738 STATIC int __init
1739 init_xfs_fs(void)
1741 int error;
1743 printk(KERN_INFO XFS_VERSION_STRING " with "
1744 XFS_BUILD_OPTIONS " enabled\n");
1746 xfs_dir_startup();
1748 error = xfs_init_zones();
1749 if (error)
1750 goto out;
1752 error = xfs_init_workqueues();
1753 if (error)
1754 goto out_destroy_zones;
1756 error = xfs_mru_cache_init();
1757 if (error)
1758 goto out_destroy_wq;
1760 error = xfs_buf_init();
1761 if (error)
1762 goto out_mru_cache_uninit;
1764 error = xfs_init_procfs();
1765 if (error)
1766 goto out_buf_terminate;
1768 error = xfs_sysctl_register();
1769 if (error)
1770 goto out_cleanup_procfs;
1772 xfs_kset = kset_create_and_add("xfs", NULL, fs_kobj);
1773 if (!xfs_kset) {
1774 error = -ENOMEM;
1775 goto out_sysctl_unregister;;
1778 #ifdef DEBUG
1779 xfs_dbg_kobj.kobject.kset = xfs_kset;
1780 error = xfs_sysfs_init(&xfs_dbg_kobj, &xfs_dbg_ktype, NULL, "debug");
1781 if (error)
1782 goto out_kset_unregister;
1783 #endif
1785 error = xfs_qm_init();
1786 if (error)
1787 goto out_remove_kobj;
1789 error = register_filesystem(&xfs_fs_type);
1790 if (error)
1791 goto out_qm_exit;
1792 return 0;
1794 out_qm_exit:
1795 xfs_qm_exit();
1796 out_remove_kobj:
1797 #ifdef DEBUG
1798 xfs_sysfs_del(&xfs_dbg_kobj);
1799 out_kset_unregister:
1800 #endif
1801 kset_unregister(xfs_kset);
1802 out_sysctl_unregister:
1803 xfs_sysctl_unregister();
1804 out_cleanup_procfs:
1805 xfs_cleanup_procfs();
1806 out_buf_terminate:
1807 xfs_buf_terminate();
1808 out_mru_cache_uninit:
1809 xfs_mru_cache_uninit();
1810 out_destroy_wq:
1811 xfs_destroy_workqueues();
1812 out_destroy_zones:
1813 xfs_destroy_zones();
1814 out:
1815 return error;
1818 STATIC void __exit
1819 exit_xfs_fs(void)
1821 xfs_qm_exit();
1822 unregister_filesystem(&xfs_fs_type);
1823 #ifdef DEBUG
1824 xfs_sysfs_del(&xfs_dbg_kobj);
1825 #endif
1826 kset_unregister(xfs_kset);
1827 xfs_sysctl_unregister();
1828 xfs_cleanup_procfs();
1829 xfs_buf_terminate();
1830 xfs_mru_cache_uninit();
1831 xfs_destroy_workqueues();
1832 xfs_destroy_zones();
1835 module_init(init_xfs_fs);
1836 module_exit(exit_xfs_fs);
1838 MODULE_AUTHOR("Silicon Graphics, Inc.");
1839 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1840 MODULE_LICENSE("GPL");