x86, delay: tsc based udelay should have rdtsc_barrier
[linux-2.6/mini2440.git] / fs / xfs / linux-2.6 / xfs_super.c
bloba220d36f789b3454c90fa0a4fe71924c4b186b62
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
18 #include "xfs.h"
19 #include "xfs_bit.h"
20 #include "xfs_log.h"
21 #include "xfs_inum.h"
22 #include "xfs_trans.h"
23 #include "xfs_sb.h"
24 #include "xfs_ag.h"
25 #include "xfs_dir2.h"
26 #include "xfs_alloc.h"
27 #include "xfs_dmapi.h"
28 #include "xfs_quota.h"
29 #include "xfs_mount.h"
30 #include "xfs_bmap_btree.h"
31 #include "xfs_alloc_btree.h"
32 #include "xfs_ialloc_btree.h"
33 #include "xfs_dir2_sf.h"
34 #include "xfs_attr_sf.h"
35 #include "xfs_dinode.h"
36 #include "xfs_inode.h"
37 #include "xfs_btree.h"
38 #include "xfs_btree_trace.h"
39 #include "xfs_ialloc.h"
40 #include "xfs_bmap.h"
41 #include "xfs_rtalloc.h"
42 #include "xfs_error.h"
43 #include "xfs_itable.h"
44 #include "xfs_fsops.h"
45 #include "xfs_rw.h"
46 #include "xfs_attr.h"
47 #include "xfs_buf_item.h"
48 #include "xfs_utils.h"
49 #include "xfs_vnodeops.h"
50 #include "xfs_version.h"
51 #include "xfs_log_priv.h"
52 #include "xfs_trans_priv.h"
53 #include "xfs_filestream.h"
54 #include "xfs_da_btree.h"
55 #include "xfs_dir2_trace.h"
56 #include "xfs_extfree_item.h"
57 #include "xfs_mru_cache.h"
58 #include "xfs_inode_item.h"
59 #include "xfs_sync.h"
61 #include <linux/namei.h>
62 #include <linux/init.h>
63 #include <linux/mount.h>
64 #include <linux/mempool.h>
65 #include <linux/writeback.h>
66 #include <linux/kthread.h>
67 #include <linux/freezer.h>
68 #include <linux/parser.h>
70 static struct super_operations xfs_super_operations;
71 static kmem_zone_t *xfs_ioend_zone;
72 mempool_t *xfs_ioend_pool;
74 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
75 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
76 #define MNTOPT_LOGDEV "logdev" /* log device */
77 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
78 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
79 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
80 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
81 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
82 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
83 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
84 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
85 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
86 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
87 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
88 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
89 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
90 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
91 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
92 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
93 * unwritten extent conversion */
94 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
95 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
96 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
97 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
98 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
99 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
100 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
101 * in stat(). */
102 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
103 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
104 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
105 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
106 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
107 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
108 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
109 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
110 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
111 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
112 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
113 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
114 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
115 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
116 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
117 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
118 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
119 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
122 * Table driven mount option parser.
124 * Currently only used for remount, but it will be used for mount
125 * in the future, too.
127 enum {
128 Opt_barrier, Opt_nobarrier, Opt_err
131 static const match_table_t tokens = {
132 {Opt_barrier, "barrier"},
133 {Opt_nobarrier, "nobarrier"},
134 {Opt_err, NULL}
138 STATIC unsigned long
139 suffix_strtoul(char *s, char **endp, unsigned int base)
141 int last, shift_left_factor = 0;
142 char *value = s;
144 last = strlen(value) - 1;
145 if (value[last] == 'K' || value[last] == 'k') {
146 shift_left_factor = 10;
147 value[last] = '\0';
149 if (value[last] == 'M' || value[last] == 'm') {
150 shift_left_factor = 20;
151 value[last] = '\0';
153 if (value[last] == 'G' || value[last] == 'g') {
154 shift_left_factor = 30;
155 value[last] = '\0';
158 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
162 * This function fills in xfs_mount_t fields based on mount args.
163 * Note: the superblock has _not_ yet been read in.
165 * Note that this function leaks the various device name allocations on
166 * failure. The caller takes care of them.
168 STATIC int
169 xfs_parseargs(
170 struct xfs_mount *mp,
171 char *options,
172 char **mtpt)
174 struct super_block *sb = mp->m_super;
175 char *this_char, *value, *eov;
176 int dsunit = 0;
177 int dswidth = 0;
178 int iosize = 0;
179 int dmapi_implies_ikeep = 1;
180 __uint8_t iosizelog = 0;
183 * Copy binary VFS mount flags we are interested in.
185 if (sb->s_flags & MS_RDONLY)
186 mp->m_flags |= XFS_MOUNT_RDONLY;
187 if (sb->s_flags & MS_DIRSYNC)
188 mp->m_flags |= XFS_MOUNT_DIRSYNC;
189 if (sb->s_flags & MS_SYNCHRONOUS)
190 mp->m_flags |= XFS_MOUNT_WSYNC;
193 * Set some default flags that could be cleared by the mount option
194 * parsing.
196 mp->m_flags |= XFS_MOUNT_BARRIER;
197 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
198 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
201 * These can be overridden by the mount option parsing.
203 mp->m_logbufs = -1;
204 mp->m_logbsize = -1;
206 if (!options)
207 goto done;
209 while ((this_char = strsep(&options, ",")) != NULL) {
210 if (!*this_char)
211 continue;
212 if ((value = strchr(this_char, '=')) != NULL)
213 *value++ = 0;
215 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
216 if (!value || !*value) {
217 cmn_err(CE_WARN,
218 "XFS: %s option requires an argument",
219 this_char);
220 return EINVAL;
222 mp->m_logbufs = simple_strtoul(value, &eov, 10);
223 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
224 if (!value || !*value) {
225 cmn_err(CE_WARN,
226 "XFS: %s option requires an argument",
227 this_char);
228 return EINVAL;
230 mp->m_logbsize = suffix_strtoul(value, &eov, 10);
231 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
232 if (!value || !*value) {
233 cmn_err(CE_WARN,
234 "XFS: %s option requires an argument",
235 this_char);
236 return EINVAL;
238 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
239 if (!mp->m_logname)
240 return ENOMEM;
241 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
242 if (!value || !*value) {
243 cmn_err(CE_WARN,
244 "XFS: %s option requires an argument",
245 this_char);
246 return EINVAL;
248 *mtpt = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
249 if (!*mtpt)
250 return ENOMEM;
251 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
252 if (!value || !*value) {
253 cmn_err(CE_WARN,
254 "XFS: %s option requires an argument",
255 this_char);
256 return EINVAL;
258 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
259 if (!mp->m_rtname)
260 return ENOMEM;
261 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
262 if (!value || !*value) {
263 cmn_err(CE_WARN,
264 "XFS: %s option requires an argument",
265 this_char);
266 return EINVAL;
268 iosize = simple_strtoul(value, &eov, 10);
269 iosizelog = ffs(iosize) - 1;
270 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
271 if (!value || !*value) {
272 cmn_err(CE_WARN,
273 "XFS: %s option requires an argument",
274 this_char);
275 return EINVAL;
277 iosize = suffix_strtoul(value, &eov, 10);
278 iosizelog = ffs(iosize) - 1;
279 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
280 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
281 mp->m_flags |= XFS_MOUNT_GRPID;
282 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
283 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
284 mp->m_flags &= ~XFS_MOUNT_GRPID;
285 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
286 mp->m_flags |= XFS_MOUNT_WSYNC;
287 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
288 mp->m_flags |= XFS_MOUNT_OSYNCISOSYNC;
289 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
290 mp->m_flags |= XFS_MOUNT_NORECOVERY;
291 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
292 mp->m_flags |= XFS_MOUNT_NOALIGN;
293 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
294 mp->m_flags |= XFS_MOUNT_SWALLOC;
295 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
296 if (!value || !*value) {
297 cmn_err(CE_WARN,
298 "XFS: %s option requires an argument",
299 this_char);
300 return EINVAL;
302 dsunit = simple_strtoul(value, &eov, 10);
303 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
304 if (!value || !*value) {
305 cmn_err(CE_WARN,
306 "XFS: %s option requires an argument",
307 this_char);
308 return EINVAL;
310 dswidth = simple_strtoul(value, &eov, 10);
311 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
312 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
313 #if !XFS_BIG_INUMS
314 cmn_err(CE_WARN,
315 "XFS: %s option not allowed on this system",
316 this_char);
317 return EINVAL;
318 #endif
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 dmapi_implies_ikeep = 0;
329 mp->m_flags &= ~XFS_MOUNT_IKEEP;
330 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
331 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
332 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
333 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
334 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
335 mp->m_flags |= XFS_MOUNT_ATTR2;
336 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
337 mp->m_flags &= ~XFS_MOUNT_ATTR2;
338 mp->m_flags |= XFS_MOUNT_NOATTR2;
339 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
340 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
341 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
342 mp->m_qflags &= ~(XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
343 XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
344 XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
345 XFS_UQUOTA_ENFD | XFS_OQUOTA_ENFD);
346 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
347 !strcmp(this_char, MNTOPT_UQUOTA) ||
348 !strcmp(this_char, MNTOPT_USRQUOTA)) {
349 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
350 XFS_UQUOTA_ENFD);
351 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
352 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
353 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
354 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
355 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
356 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
357 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
358 XFS_OQUOTA_ENFD);
359 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
360 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
361 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
362 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
363 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
364 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
365 XFS_OQUOTA_ENFD);
366 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
367 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
368 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
369 } else if (!strcmp(this_char, MNTOPT_DMAPI)) {
370 mp->m_flags |= XFS_MOUNT_DMAPI;
371 } else if (!strcmp(this_char, MNTOPT_XDSM)) {
372 mp->m_flags |= XFS_MOUNT_DMAPI;
373 } else if (!strcmp(this_char, MNTOPT_DMI)) {
374 mp->m_flags |= XFS_MOUNT_DMAPI;
375 } else if (!strcmp(this_char, "ihashsize")) {
376 cmn_err(CE_WARN,
377 "XFS: ihashsize no longer used, option is deprecated.");
378 } else if (!strcmp(this_char, "osyncisdsync")) {
379 /* no-op, this is now the default */
380 cmn_err(CE_WARN,
381 "XFS: osyncisdsync is now the default, option is deprecated.");
382 } else if (!strcmp(this_char, "irixsgid")) {
383 cmn_err(CE_WARN,
384 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
385 } else {
386 cmn_err(CE_WARN,
387 "XFS: unknown mount option [%s].", this_char);
388 return EINVAL;
393 * no recovery flag requires a read-only mount
395 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
396 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
397 cmn_err(CE_WARN, "XFS: no-recovery mounts must be read-only.");
398 return EINVAL;
401 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
402 cmn_err(CE_WARN,
403 "XFS: sunit and swidth options incompatible with the noalign option");
404 return EINVAL;
407 #ifndef CONFIG_XFS_QUOTA
408 if (XFS_IS_QUOTA_RUNNING(mp)) {
409 cmn_err(CE_WARN,
410 "XFS: quota support not available in this kernel.");
411 return EINVAL;
413 #endif
415 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
416 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE))) {
417 cmn_err(CE_WARN,
418 "XFS: cannot mount with both project and group quota");
419 return EINVAL;
422 if ((mp->m_flags & XFS_MOUNT_DMAPI) && (!*mtpt || *mtpt[0] == '\0')) {
423 printk("XFS: %s option needs the mount point option as well\n",
424 MNTOPT_DMAPI);
425 return EINVAL;
428 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
429 cmn_err(CE_WARN,
430 "XFS: sunit and swidth must be specified together");
431 return EINVAL;
434 if (dsunit && (dswidth % dsunit != 0)) {
435 cmn_err(CE_WARN,
436 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
437 dswidth, dsunit);
438 return EINVAL;
442 * Applications using DMI filesystems often expect the
443 * inode generation number to be monotonically increasing.
444 * If we delete inode chunks we break this assumption, so
445 * keep unused inode chunks on disk for DMI filesystems
446 * until we come up with a better solution.
447 * Note that if "ikeep" or "noikeep" mount options are
448 * supplied, then they are honored.
450 if ((mp->m_flags & XFS_MOUNT_DMAPI) && dmapi_implies_ikeep)
451 mp->m_flags |= XFS_MOUNT_IKEEP;
453 done:
454 if (!(mp->m_flags & XFS_MOUNT_NOALIGN)) {
456 * At this point the superblock has not been read
457 * in, therefore we do not know the block size.
458 * Before the mount call ends we will convert
459 * these to FSBs.
461 if (dsunit) {
462 mp->m_dalign = dsunit;
463 mp->m_flags |= XFS_MOUNT_RETERR;
466 if (dswidth)
467 mp->m_swidth = dswidth;
470 if (mp->m_logbufs != -1 &&
471 mp->m_logbufs != 0 &&
472 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
473 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
474 cmn_err(CE_WARN,
475 "XFS: invalid logbufs value: %d [not %d-%d]",
476 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
477 return XFS_ERROR(EINVAL);
479 if (mp->m_logbsize != -1 &&
480 mp->m_logbsize != 0 &&
481 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
482 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
483 !is_power_of_2(mp->m_logbsize))) {
484 cmn_err(CE_WARN,
485 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
486 mp->m_logbsize);
487 return XFS_ERROR(EINVAL);
490 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
491 if (!mp->m_fsname)
492 return ENOMEM;
493 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
495 if (iosizelog) {
496 if (iosizelog > XFS_MAX_IO_LOG ||
497 iosizelog < XFS_MIN_IO_LOG) {
498 cmn_err(CE_WARN,
499 "XFS: invalid log iosize: %d [not %d-%d]",
500 iosizelog, XFS_MIN_IO_LOG,
501 XFS_MAX_IO_LOG);
502 return XFS_ERROR(EINVAL);
505 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
506 mp->m_readio_log = iosizelog;
507 mp->m_writeio_log = iosizelog;
510 return 0;
513 struct proc_xfs_info {
514 int flag;
515 char *str;
518 STATIC int
519 xfs_showargs(
520 struct xfs_mount *mp,
521 struct seq_file *m)
523 static struct proc_xfs_info xfs_info_set[] = {
524 /* the few simple ones we can get from the mount struct */
525 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
526 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
527 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
528 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
529 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
530 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
531 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
532 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
533 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
534 { XFS_MOUNT_DMAPI, "," MNTOPT_DMAPI },
535 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
536 { 0, NULL }
538 static struct proc_xfs_info xfs_info_unset[] = {
539 /* the few simple ones we can get from the mount struct */
540 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
541 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
542 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
543 { 0, NULL }
545 struct proc_xfs_info *xfs_infop;
547 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
548 if (mp->m_flags & xfs_infop->flag)
549 seq_puts(m, xfs_infop->str);
551 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
552 if (!(mp->m_flags & xfs_infop->flag))
553 seq_puts(m, xfs_infop->str);
556 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
557 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
558 (int)(1 << mp->m_writeio_log) >> 10);
560 if (mp->m_logbufs > 0)
561 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
562 if (mp->m_logbsize > 0)
563 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
565 if (mp->m_logname)
566 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
567 if (mp->m_rtname)
568 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
570 if (mp->m_dalign > 0)
571 seq_printf(m, "," MNTOPT_SUNIT "=%d",
572 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
573 if (mp->m_swidth > 0)
574 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
575 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
577 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
578 seq_puts(m, "," MNTOPT_USRQUOTA);
579 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
580 seq_puts(m, "," MNTOPT_UQUOTANOENF);
582 if (mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
583 seq_puts(m, "," MNTOPT_PRJQUOTA);
584 else if (mp->m_qflags & XFS_PQUOTA_ACCT)
585 seq_puts(m, "," MNTOPT_PQUOTANOENF);
587 if (mp->m_qflags & (XFS_GQUOTA_ACCT|XFS_OQUOTA_ENFD))
588 seq_puts(m, "," MNTOPT_GRPQUOTA);
589 else if (mp->m_qflags & XFS_GQUOTA_ACCT)
590 seq_puts(m, "," MNTOPT_GQUOTANOENF);
592 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
593 seq_puts(m, "," MNTOPT_NOQUOTA);
595 return 0;
597 __uint64_t
598 xfs_max_file_offset(
599 unsigned int blockshift)
601 unsigned int pagefactor = 1;
602 unsigned int bitshift = BITS_PER_LONG - 1;
604 /* Figure out maximum filesize, on Linux this can depend on
605 * the filesystem blocksize (on 32 bit platforms).
606 * __block_prepare_write does this in an [unsigned] long...
607 * page->index << (PAGE_CACHE_SHIFT - bbits)
608 * So, for page sized blocks (4K on 32 bit platforms),
609 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
610 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
611 * but for smaller blocksizes it is less (bbits = log2 bsize).
612 * Note1: get_block_t takes a long (implicit cast from above)
613 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
614 * can optionally convert the [unsigned] long from above into
615 * an [unsigned] long long.
618 #if BITS_PER_LONG == 32
619 # if defined(CONFIG_LBDAF)
620 ASSERT(sizeof(sector_t) == 8);
621 pagefactor = PAGE_CACHE_SIZE;
622 bitshift = BITS_PER_LONG;
623 # else
624 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
625 # endif
626 #endif
628 return (((__uint64_t)pagefactor) << bitshift) - 1;
631 STATIC int
632 xfs_blkdev_get(
633 xfs_mount_t *mp,
634 const char *name,
635 struct block_device **bdevp)
637 int error = 0;
639 *bdevp = open_bdev_exclusive(name, FMODE_READ|FMODE_WRITE, mp);
640 if (IS_ERR(*bdevp)) {
641 error = PTR_ERR(*bdevp);
642 printk("XFS: Invalid device [%s], error=%d\n", name, error);
645 return -error;
648 STATIC void
649 xfs_blkdev_put(
650 struct block_device *bdev)
652 if (bdev)
653 close_bdev_exclusive(bdev, FMODE_READ|FMODE_WRITE);
657 * Try to write out the superblock using barriers.
659 STATIC int
660 xfs_barrier_test(
661 xfs_mount_t *mp)
663 xfs_buf_t *sbp = xfs_getsb(mp, 0);
664 int error;
666 XFS_BUF_UNDONE(sbp);
667 XFS_BUF_UNREAD(sbp);
668 XFS_BUF_UNDELAYWRITE(sbp);
669 XFS_BUF_WRITE(sbp);
670 XFS_BUF_UNASYNC(sbp);
671 XFS_BUF_ORDERED(sbp);
673 xfsbdstrat(mp, sbp);
674 error = xfs_iowait(sbp);
677 * Clear all the flags we set and possible error state in the
678 * buffer. We only did the write to try out whether barriers
679 * worked and shouldn't leave any traces in the superblock
680 * buffer.
682 XFS_BUF_DONE(sbp);
683 XFS_BUF_ERROR(sbp, 0);
684 XFS_BUF_UNORDERED(sbp);
686 xfs_buf_relse(sbp);
687 return error;
690 void
691 xfs_mountfs_check_barriers(xfs_mount_t *mp)
693 int error;
695 if (mp->m_logdev_targp != mp->m_ddev_targp) {
696 xfs_fs_cmn_err(CE_NOTE, mp,
697 "Disabling barriers, not supported with external log device");
698 mp->m_flags &= ~XFS_MOUNT_BARRIER;
699 return;
702 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
703 xfs_fs_cmn_err(CE_NOTE, mp,
704 "Disabling barriers, underlying device is readonly");
705 mp->m_flags &= ~XFS_MOUNT_BARRIER;
706 return;
709 error = xfs_barrier_test(mp);
710 if (error) {
711 xfs_fs_cmn_err(CE_NOTE, mp,
712 "Disabling barriers, trial barrier write failed");
713 mp->m_flags &= ~XFS_MOUNT_BARRIER;
714 return;
718 void
719 xfs_blkdev_issue_flush(
720 xfs_buftarg_t *buftarg)
722 blkdev_issue_flush(buftarg->bt_bdev, NULL);
725 STATIC void
726 xfs_close_devices(
727 struct xfs_mount *mp)
729 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
730 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
731 xfs_free_buftarg(mp, mp->m_logdev_targp);
732 xfs_blkdev_put(logdev);
734 if (mp->m_rtdev_targp) {
735 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
736 xfs_free_buftarg(mp, mp->m_rtdev_targp);
737 xfs_blkdev_put(rtdev);
739 xfs_free_buftarg(mp, mp->m_ddev_targp);
743 * The file system configurations are:
744 * (1) device (partition) with data and internal log
745 * (2) logical volume with data and log subvolumes.
746 * (3) logical volume with data, log, and realtime subvolumes.
748 * We only have to handle opening the log and realtime volumes here if
749 * they are present. The data subvolume has already been opened by
750 * get_sb_bdev() and is stored in sb->s_bdev.
752 STATIC int
753 xfs_open_devices(
754 struct xfs_mount *mp)
756 struct block_device *ddev = mp->m_super->s_bdev;
757 struct block_device *logdev = NULL, *rtdev = NULL;
758 int error;
761 * Open real time and log devices - order is important.
763 if (mp->m_logname) {
764 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
765 if (error)
766 goto out;
769 if (mp->m_rtname) {
770 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
771 if (error)
772 goto out_close_logdev;
774 if (rtdev == ddev || rtdev == logdev) {
775 cmn_err(CE_WARN,
776 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
777 error = EINVAL;
778 goto out_close_rtdev;
783 * Setup xfs_mount buffer target pointers
785 error = ENOMEM;
786 mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0);
787 if (!mp->m_ddev_targp)
788 goto out_close_rtdev;
790 if (rtdev) {
791 mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1);
792 if (!mp->m_rtdev_targp)
793 goto out_free_ddev_targ;
796 if (logdev && logdev != ddev) {
797 mp->m_logdev_targp = xfs_alloc_buftarg(logdev, 1);
798 if (!mp->m_logdev_targp)
799 goto out_free_rtdev_targ;
800 } else {
801 mp->m_logdev_targp = mp->m_ddev_targp;
804 return 0;
806 out_free_rtdev_targ:
807 if (mp->m_rtdev_targp)
808 xfs_free_buftarg(mp, mp->m_rtdev_targp);
809 out_free_ddev_targ:
810 xfs_free_buftarg(mp, mp->m_ddev_targp);
811 out_close_rtdev:
812 if (rtdev)
813 xfs_blkdev_put(rtdev);
814 out_close_logdev:
815 if (logdev && logdev != ddev)
816 xfs_blkdev_put(logdev);
817 out:
818 return error;
822 * Setup xfs_mount buffer target pointers based on superblock
824 STATIC int
825 xfs_setup_devices(
826 struct xfs_mount *mp)
828 int error;
830 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
831 mp->m_sb.sb_sectsize);
832 if (error)
833 return error;
835 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
836 unsigned int log_sector_size = BBSIZE;
838 if (xfs_sb_version_hassector(&mp->m_sb))
839 log_sector_size = mp->m_sb.sb_logsectsize;
840 error = xfs_setsize_buftarg(mp->m_logdev_targp,
841 mp->m_sb.sb_blocksize,
842 log_sector_size);
843 if (error)
844 return error;
846 if (mp->m_rtdev_targp) {
847 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
848 mp->m_sb.sb_blocksize,
849 mp->m_sb.sb_sectsize);
850 if (error)
851 return error;
854 return 0;
858 * XFS AIL push thread support
860 void
861 xfsaild_wakeup(
862 struct xfs_ail *ailp,
863 xfs_lsn_t threshold_lsn)
865 ailp->xa_target = threshold_lsn;
866 wake_up_process(ailp->xa_task);
869 STATIC int
870 xfsaild(
871 void *data)
873 struct xfs_ail *ailp = data;
874 xfs_lsn_t last_pushed_lsn = 0;
875 long tout = 0;
877 while (!kthread_should_stop()) {
878 if (tout)
879 schedule_timeout_interruptible(msecs_to_jiffies(tout));
880 tout = 1000;
882 /* swsusp */
883 try_to_freeze();
885 ASSERT(ailp->xa_mount->m_log);
886 if (XFS_FORCED_SHUTDOWN(ailp->xa_mount))
887 continue;
889 tout = xfsaild_push(ailp, &last_pushed_lsn);
892 return 0;
893 } /* xfsaild */
896 xfsaild_start(
897 struct xfs_ail *ailp)
899 ailp->xa_target = 0;
900 ailp->xa_task = kthread_run(xfsaild, ailp, "xfsaild");
901 if (IS_ERR(ailp->xa_task))
902 return -PTR_ERR(ailp->xa_task);
903 return 0;
906 void
907 xfsaild_stop(
908 struct xfs_ail *ailp)
910 kthread_stop(ailp->xa_task);
914 /* Catch misguided souls that try to use this interface on XFS */
915 STATIC struct inode *
916 xfs_fs_alloc_inode(
917 struct super_block *sb)
919 BUG();
920 return NULL;
924 * Now that the generic code is guaranteed not to be accessing
925 * the linux inode, we can reclaim the inode.
927 STATIC void
928 xfs_fs_destroy_inode(
929 struct inode *inode)
931 xfs_inode_t *ip = XFS_I(inode);
933 XFS_STATS_INC(vn_reclaim);
934 if (xfs_reclaim(ip))
935 panic("%s: cannot reclaim 0x%p\n", __func__, inode);
939 * Slab object creation initialisation for the XFS inode.
940 * This covers only the idempotent fields in the XFS inode;
941 * all other fields need to be initialised on allocation
942 * from the slab. This avoids the need to repeatedly intialise
943 * fields in the xfs inode that left in the initialise state
944 * when freeing the inode.
946 STATIC void
947 xfs_fs_inode_init_once(
948 void *inode)
950 struct xfs_inode *ip = inode;
952 memset(ip, 0, sizeof(struct xfs_inode));
954 /* vfs inode */
955 inode_init_once(VFS_I(ip));
957 /* xfs inode */
958 atomic_set(&ip->i_iocount, 0);
959 atomic_set(&ip->i_pincount, 0);
960 spin_lock_init(&ip->i_flags_lock);
961 init_waitqueue_head(&ip->i_ipin_wait);
963 * Because we want to use a counting completion, complete
964 * the flush completion once to allow a single access to
965 * the flush completion without blocking.
967 init_completion(&ip->i_flush);
968 complete(&ip->i_flush);
970 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
971 "xfsino", ip->i_ino);
972 mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
976 * Attempt to flush the inode, this will actually fail
977 * if the inode is pinned, but we dirty the inode again
978 * at the point when it is unpinned after a log write,
979 * since this is when the inode itself becomes flushable.
981 STATIC int
982 xfs_fs_write_inode(
983 struct inode *inode,
984 int sync)
986 struct xfs_inode *ip = XFS_I(inode);
987 struct xfs_mount *mp = ip->i_mount;
988 int error = 0;
990 xfs_itrace_entry(ip);
992 if (XFS_FORCED_SHUTDOWN(mp))
993 return XFS_ERROR(EIO);
995 if (sync) {
996 error = xfs_wait_on_pages(ip, 0, -1);
997 if (error)
998 goto out;
1002 * Bypass inodes which have already been cleaned by
1003 * the inode flush clustering code inside xfs_iflush
1005 if (xfs_inode_clean(ip))
1006 goto out;
1009 * We make this non-blocking if the inode is contended, return
1010 * EAGAIN to indicate to the caller that they did not succeed.
1011 * This prevents the flush path from blocking on inodes inside
1012 * another operation right now, they get caught later by xfs_sync.
1014 if (sync) {
1015 xfs_ilock(ip, XFS_ILOCK_SHARED);
1016 xfs_iflock(ip);
1018 error = xfs_iflush(ip, XFS_IFLUSH_SYNC);
1019 } else {
1020 error = EAGAIN;
1021 if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED))
1022 goto out;
1023 if (xfs_ipincount(ip) || !xfs_iflock_nowait(ip))
1024 goto out_unlock;
1026 error = xfs_iflush(ip, XFS_IFLUSH_ASYNC_NOBLOCK);
1029 out_unlock:
1030 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1031 out:
1033 * if we failed to write out the inode then mark
1034 * it dirty again so we'll try again later.
1036 if (error)
1037 xfs_mark_inode_dirty_sync(ip);
1038 return -error;
1041 STATIC void
1042 xfs_fs_clear_inode(
1043 struct inode *inode)
1045 xfs_inode_t *ip = XFS_I(inode);
1047 xfs_itrace_entry(ip);
1048 XFS_STATS_INC(vn_rele);
1049 XFS_STATS_INC(vn_remove);
1050 XFS_STATS_DEC(vn_active);
1052 xfs_inactive(ip);
1055 STATIC void
1056 xfs_free_fsname(
1057 struct xfs_mount *mp)
1059 kfree(mp->m_fsname);
1060 kfree(mp->m_rtname);
1061 kfree(mp->m_logname);
1064 STATIC void
1065 xfs_fs_put_super(
1066 struct super_block *sb)
1068 struct xfs_mount *mp = XFS_M(sb);
1069 struct xfs_inode *rip = mp->m_rootip;
1070 int unmount_event_flags = 0;
1072 xfs_syncd_stop(mp);
1074 if (!(sb->s_flags & MS_RDONLY)) {
1076 * XXX(hch): this should be SYNC_WAIT.
1078 * Or more likely not needed at all because the VFS is already
1079 * calling ->sync_fs after shutting down all filestem
1080 * operations and just before calling ->put_super.
1082 xfs_sync_data(mp, 0);
1083 xfs_sync_attr(mp, 0);
1086 #ifdef HAVE_DMAPI
1087 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1088 unmount_event_flags =
1089 (mp->m_dmevmask & (1 << DM_EVENT_UNMOUNT)) ?
1090 0 : DM_FLAGS_UNWANTED;
1092 * Ignore error from dmapi here, first unmount is not allowed
1093 * to fail anyway, and second we wouldn't want to fail a
1094 * unmount because of dmapi.
1096 XFS_SEND_PREUNMOUNT(mp, rip, DM_RIGHT_NULL, rip, DM_RIGHT_NULL,
1097 NULL, NULL, 0, 0, unmount_event_flags);
1099 #endif
1102 * Blow away any referenced inode in the filestreams cache.
1103 * This can and will cause log traffic as inodes go inactive
1104 * here.
1106 xfs_filestream_unmount(mp);
1108 XFS_bflush(mp->m_ddev_targp);
1110 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1111 XFS_SEND_UNMOUNT(mp, rip, DM_RIGHT_NULL, 0, 0,
1112 unmount_event_flags);
1115 xfs_unmountfs(mp);
1116 xfs_freesb(mp);
1117 xfs_icsb_destroy_counters(mp);
1118 xfs_close_devices(mp);
1119 xfs_dmops_put(mp);
1120 xfs_free_fsname(mp);
1121 kfree(mp);
1124 STATIC int
1125 xfs_fs_sync_super(
1126 struct super_block *sb,
1127 int wait)
1129 struct xfs_mount *mp = XFS_M(sb);
1130 int error;
1133 * Treat a sync operation like a freeze. This is to work
1134 * around a race in sync_inodes() which works in two phases
1135 * - an asynchronous flush, which can write out an inode
1136 * without waiting for file size updates to complete, and a
1137 * synchronous flush, which wont do anything because the
1138 * async flush removed the inode's dirty flag. Also
1139 * sync_inodes() will not see any files that just have
1140 * outstanding transactions to be flushed because we don't
1141 * dirty the Linux inode until after the transaction I/O
1142 * completes.
1144 if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE))
1145 error = xfs_quiesce_data(mp);
1146 else
1147 error = xfs_sync_fsdata(mp, 0);
1149 if (unlikely(laptop_mode)) {
1150 int prev_sync_seq = mp->m_sync_seq;
1153 * The disk must be active because we're syncing.
1154 * We schedule xfssyncd now (now that the disk is
1155 * active) instead of later (when it might not be).
1157 wake_up_process(mp->m_sync_task);
1159 * We have to wait for the sync iteration to complete.
1160 * If we don't, the disk activity caused by the sync
1161 * will come after the sync is completed, and that
1162 * triggers another sync from laptop mode.
1164 wait_event(mp->m_wait_single_sync_task,
1165 mp->m_sync_seq != prev_sync_seq);
1168 return -error;
1171 STATIC int
1172 xfs_fs_statfs(
1173 struct dentry *dentry,
1174 struct kstatfs *statp)
1176 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1177 xfs_sb_t *sbp = &mp->m_sb;
1178 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1179 __uint64_t fakeinos, id;
1180 xfs_extlen_t lsize;
1182 statp->f_type = XFS_SB_MAGIC;
1183 statp->f_namelen = MAXNAMELEN - 1;
1185 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1186 statp->f_fsid.val[0] = (u32)id;
1187 statp->f_fsid.val[1] = (u32)(id >> 32);
1189 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1191 spin_lock(&mp->m_sb_lock);
1192 statp->f_bsize = sbp->sb_blocksize;
1193 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1194 statp->f_blocks = sbp->sb_dblocks - lsize;
1195 statp->f_bfree = statp->f_bavail =
1196 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1197 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1198 statp->f_files =
1199 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1200 if (mp->m_maxicount)
1201 statp->f_files = min_t(typeof(statp->f_files),
1202 statp->f_files,
1203 mp->m_maxicount);
1204 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1205 spin_unlock(&mp->m_sb_lock);
1207 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) ||
1208 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))) ==
1209 (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
1210 xfs_qm_statvfs(ip, statp);
1211 return 0;
1214 STATIC int
1215 xfs_fs_remount(
1216 struct super_block *sb,
1217 int *flags,
1218 char *options)
1220 struct xfs_mount *mp = XFS_M(sb);
1221 substring_t args[MAX_OPT_ARGS];
1222 char *p;
1223 int error;
1225 while ((p = strsep(&options, ",")) != NULL) {
1226 int token;
1228 if (!*p)
1229 continue;
1231 token = match_token(p, tokens, args);
1232 switch (token) {
1233 case Opt_barrier:
1234 mp->m_flags |= XFS_MOUNT_BARRIER;
1237 * Test if barriers are actually working if we can,
1238 * else delay this check until the filesystem is
1239 * marked writeable.
1241 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1242 xfs_mountfs_check_barriers(mp);
1243 break;
1244 case Opt_nobarrier:
1245 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1246 break;
1247 default:
1249 * Logically we would return an error here to prevent
1250 * users from believing they might have changed
1251 * mount options using remount which can't be changed.
1253 * But unfortunately mount(8) adds all options from
1254 * mtab and fstab to the mount arguments in some cases
1255 * so we can't blindly reject options, but have to
1256 * check for each specified option if it actually
1257 * differs from the currently set option and only
1258 * reject it if that's the case.
1260 * Until that is implemented we return success for
1261 * every remount request, and silently ignore all
1262 * options that we can't actually change.
1264 #if 0
1265 printk(KERN_INFO
1266 "XFS: mount option \"%s\" not supported for remount\n", p);
1267 return -EINVAL;
1268 #else
1269 break;
1270 #endif
1274 /* ro -> rw */
1275 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1276 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1277 if (mp->m_flags & XFS_MOUNT_BARRIER)
1278 xfs_mountfs_check_barriers(mp);
1281 * If this is the first remount to writeable state we
1282 * might have some superblock changes to update.
1284 if (mp->m_update_flags) {
1285 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1286 if (error) {
1287 cmn_err(CE_WARN,
1288 "XFS: failed to write sb changes");
1289 return error;
1291 mp->m_update_flags = 0;
1295 /* rw -> ro */
1296 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1297 xfs_quiesce_data(mp);
1298 xfs_quiesce_attr(mp);
1299 mp->m_flags |= XFS_MOUNT_RDONLY;
1302 return 0;
1306 * Second stage of a freeze. The data is already frozen so we only
1307 * need to take care of the metadata. Once that's done write a dummy
1308 * record to dirty the log in case of a crash while frozen.
1310 STATIC int
1311 xfs_fs_freeze(
1312 struct super_block *sb)
1314 struct xfs_mount *mp = XFS_M(sb);
1316 xfs_quiesce_attr(mp);
1317 return -xfs_fs_log_dummy(mp);
1320 STATIC int
1321 xfs_fs_show_options(
1322 struct seq_file *m,
1323 struct vfsmount *mnt)
1325 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1329 * This function fills in xfs_mount_t fields based on mount args.
1330 * Note: the superblock _has_ now been read in.
1332 STATIC int
1333 xfs_finish_flags(
1334 struct xfs_mount *mp)
1336 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1338 /* Fail a mount where the logbuf is smaller than the log stripe */
1339 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1340 if (mp->m_logbsize <= 0 &&
1341 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1342 mp->m_logbsize = mp->m_sb.sb_logsunit;
1343 } else if (mp->m_logbsize > 0 &&
1344 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1345 cmn_err(CE_WARN,
1346 "XFS: logbuf size must be greater than or equal to log stripe size");
1347 return XFS_ERROR(EINVAL);
1349 } else {
1350 /* Fail a mount if the logbuf is larger than 32K */
1351 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1352 cmn_err(CE_WARN,
1353 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1354 return XFS_ERROR(EINVAL);
1359 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1360 * told by noattr2 to turn it off
1362 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1363 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1364 mp->m_flags |= XFS_MOUNT_ATTR2;
1367 * prohibit r/w mounts of read-only filesystems
1369 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1370 cmn_err(CE_WARN,
1371 "XFS: cannot mount a read-only filesystem as read-write");
1372 return XFS_ERROR(EROFS);
1375 return 0;
1378 STATIC int
1379 xfs_fs_fill_super(
1380 struct super_block *sb,
1381 void *data,
1382 int silent)
1384 struct inode *root;
1385 struct xfs_mount *mp = NULL;
1386 int flags = 0, error = ENOMEM;
1387 char *mtpt = NULL;
1389 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1390 if (!mp)
1391 goto out;
1393 spin_lock_init(&mp->m_sb_lock);
1394 mutex_init(&mp->m_growlock);
1395 atomic_set(&mp->m_active_trans, 0);
1396 INIT_LIST_HEAD(&mp->m_sync_list);
1397 spin_lock_init(&mp->m_sync_lock);
1398 init_waitqueue_head(&mp->m_wait_single_sync_task);
1400 mp->m_super = sb;
1401 sb->s_fs_info = mp;
1403 error = xfs_parseargs(mp, (char *)data, &mtpt);
1404 if (error)
1405 goto out_free_fsname;
1407 sb_min_blocksize(sb, BBSIZE);
1408 sb->s_xattr = xfs_xattr_handlers;
1409 sb->s_export_op = &xfs_export_operations;
1410 #ifdef CONFIG_XFS_QUOTA
1411 sb->s_qcop = &xfs_quotactl_operations;
1412 #endif
1413 sb->s_op = &xfs_super_operations;
1415 error = xfs_dmops_get(mp);
1416 if (error)
1417 goto out_free_fsname;
1419 if (silent)
1420 flags |= XFS_MFSI_QUIET;
1422 error = xfs_open_devices(mp);
1423 if (error)
1424 goto out_put_dmops;
1426 if (xfs_icsb_init_counters(mp))
1427 mp->m_flags |= XFS_MOUNT_NO_PERCPU_SB;
1429 error = xfs_readsb(mp, flags);
1430 if (error)
1431 goto out_destroy_counters;
1433 error = xfs_finish_flags(mp);
1434 if (error)
1435 goto out_free_sb;
1437 error = xfs_setup_devices(mp);
1438 if (error)
1439 goto out_free_sb;
1441 if (mp->m_flags & XFS_MOUNT_BARRIER)
1442 xfs_mountfs_check_barriers(mp);
1444 error = xfs_filestream_mount(mp);
1445 if (error)
1446 goto out_free_sb;
1448 error = xfs_mountfs(mp);
1449 if (error)
1450 goto out_filestream_unmount;
1452 XFS_SEND_MOUNT(mp, DM_RIGHT_NULL, mtpt, mp->m_fsname);
1454 sb->s_magic = XFS_SB_MAGIC;
1455 sb->s_blocksize = mp->m_sb.sb_blocksize;
1456 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1457 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1458 sb->s_time_gran = 1;
1459 set_posix_acl_flag(sb);
1461 root = igrab(VFS_I(mp->m_rootip));
1462 if (!root) {
1463 error = ENOENT;
1464 goto fail_unmount;
1466 if (is_bad_inode(root)) {
1467 error = EINVAL;
1468 goto fail_vnrele;
1470 sb->s_root = d_alloc_root(root);
1471 if (!sb->s_root) {
1472 error = ENOMEM;
1473 goto fail_vnrele;
1476 error = xfs_syncd_init(mp);
1477 if (error)
1478 goto fail_vnrele;
1480 kfree(mtpt);
1482 xfs_itrace_exit(XFS_I(sb->s_root->d_inode));
1483 return 0;
1485 out_filestream_unmount:
1486 xfs_filestream_unmount(mp);
1487 out_free_sb:
1488 xfs_freesb(mp);
1489 out_destroy_counters:
1490 xfs_icsb_destroy_counters(mp);
1491 xfs_close_devices(mp);
1492 out_put_dmops:
1493 xfs_dmops_put(mp);
1494 out_free_fsname:
1495 xfs_free_fsname(mp);
1496 kfree(mtpt);
1497 kfree(mp);
1498 out:
1499 return -error;
1501 fail_vnrele:
1502 if (sb->s_root) {
1503 dput(sb->s_root);
1504 sb->s_root = NULL;
1505 } else {
1506 iput(root);
1509 fail_unmount:
1511 * Blow away any referenced inode in the filestreams cache.
1512 * This can and will cause log traffic as inodes go inactive
1513 * here.
1515 xfs_filestream_unmount(mp);
1517 XFS_bflush(mp->m_ddev_targp);
1519 xfs_unmountfs(mp);
1520 goto out_free_sb;
1523 STATIC int
1524 xfs_fs_get_sb(
1525 struct file_system_type *fs_type,
1526 int flags,
1527 const char *dev_name,
1528 void *data,
1529 struct vfsmount *mnt)
1531 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1532 mnt);
1535 static struct super_operations xfs_super_operations = {
1536 .alloc_inode = xfs_fs_alloc_inode,
1537 .destroy_inode = xfs_fs_destroy_inode,
1538 .write_inode = xfs_fs_write_inode,
1539 .clear_inode = xfs_fs_clear_inode,
1540 .put_super = xfs_fs_put_super,
1541 .sync_fs = xfs_fs_sync_super,
1542 .freeze_fs = xfs_fs_freeze,
1543 .statfs = xfs_fs_statfs,
1544 .remount_fs = xfs_fs_remount,
1545 .show_options = xfs_fs_show_options,
1548 static struct file_system_type xfs_fs_type = {
1549 .owner = THIS_MODULE,
1550 .name = "xfs",
1551 .get_sb = xfs_fs_get_sb,
1552 .kill_sb = kill_block_super,
1553 .fs_flags = FS_REQUIRES_DEV,
1556 STATIC int __init
1557 xfs_alloc_trace_bufs(void)
1559 #ifdef XFS_ALLOC_TRACE
1560 xfs_alloc_trace_buf = ktrace_alloc(XFS_ALLOC_TRACE_SIZE, KM_MAYFAIL);
1561 if (!xfs_alloc_trace_buf)
1562 goto out;
1563 #endif
1564 #ifdef XFS_BMAP_TRACE
1565 xfs_bmap_trace_buf = ktrace_alloc(XFS_BMAP_TRACE_SIZE, KM_MAYFAIL);
1566 if (!xfs_bmap_trace_buf)
1567 goto out_free_alloc_trace;
1568 #endif
1569 #ifdef XFS_BTREE_TRACE
1570 xfs_allocbt_trace_buf = ktrace_alloc(XFS_ALLOCBT_TRACE_SIZE,
1571 KM_MAYFAIL);
1572 if (!xfs_allocbt_trace_buf)
1573 goto out_free_bmap_trace;
1575 xfs_inobt_trace_buf = ktrace_alloc(XFS_INOBT_TRACE_SIZE, KM_MAYFAIL);
1576 if (!xfs_inobt_trace_buf)
1577 goto out_free_allocbt_trace;
1579 xfs_bmbt_trace_buf = ktrace_alloc(XFS_BMBT_TRACE_SIZE, KM_MAYFAIL);
1580 if (!xfs_bmbt_trace_buf)
1581 goto out_free_inobt_trace;
1582 #endif
1583 #ifdef XFS_ATTR_TRACE
1584 xfs_attr_trace_buf = ktrace_alloc(XFS_ATTR_TRACE_SIZE, KM_MAYFAIL);
1585 if (!xfs_attr_trace_buf)
1586 goto out_free_bmbt_trace;
1587 #endif
1588 #ifdef XFS_DIR2_TRACE
1589 xfs_dir2_trace_buf = ktrace_alloc(XFS_DIR2_GTRACE_SIZE, KM_MAYFAIL);
1590 if (!xfs_dir2_trace_buf)
1591 goto out_free_attr_trace;
1592 #endif
1594 return 0;
1596 #ifdef XFS_DIR2_TRACE
1597 out_free_attr_trace:
1598 #endif
1599 #ifdef XFS_ATTR_TRACE
1600 ktrace_free(xfs_attr_trace_buf);
1601 out_free_bmbt_trace:
1602 #endif
1603 #ifdef XFS_BTREE_TRACE
1604 ktrace_free(xfs_bmbt_trace_buf);
1605 out_free_inobt_trace:
1606 ktrace_free(xfs_inobt_trace_buf);
1607 out_free_allocbt_trace:
1608 ktrace_free(xfs_allocbt_trace_buf);
1609 out_free_bmap_trace:
1610 #endif
1611 #ifdef XFS_BMAP_TRACE
1612 ktrace_free(xfs_bmap_trace_buf);
1613 out_free_alloc_trace:
1614 #endif
1615 #ifdef XFS_ALLOC_TRACE
1616 ktrace_free(xfs_alloc_trace_buf);
1617 out:
1618 #endif
1619 return -ENOMEM;
1622 STATIC void
1623 xfs_free_trace_bufs(void)
1625 #ifdef XFS_DIR2_TRACE
1626 ktrace_free(xfs_dir2_trace_buf);
1627 #endif
1628 #ifdef XFS_ATTR_TRACE
1629 ktrace_free(xfs_attr_trace_buf);
1630 #endif
1631 #ifdef XFS_BTREE_TRACE
1632 ktrace_free(xfs_bmbt_trace_buf);
1633 ktrace_free(xfs_inobt_trace_buf);
1634 ktrace_free(xfs_allocbt_trace_buf);
1635 #endif
1636 #ifdef XFS_BMAP_TRACE
1637 ktrace_free(xfs_bmap_trace_buf);
1638 #endif
1639 #ifdef XFS_ALLOC_TRACE
1640 ktrace_free(xfs_alloc_trace_buf);
1641 #endif
1644 STATIC int __init
1645 xfs_init_zones(void)
1648 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1649 if (!xfs_ioend_zone)
1650 goto out;
1652 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1653 xfs_ioend_zone);
1654 if (!xfs_ioend_pool)
1655 goto out_destroy_ioend_zone;
1657 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1658 "xfs_log_ticket");
1659 if (!xfs_log_ticket_zone)
1660 goto out_destroy_ioend_pool;
1662 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1663 "xfs_bmap_free_item");
1664 if (!xfs_bmap_free_item_zone)
1665 goto out_destroy_log_ticket_zone;
1667 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1668 "xfs_btree_cur");
1669 if (!xfs_btree_cur_zone)
1670 goto out_destroy_bmap_free_item_zone;
1672 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1673 "xfs_da_state");
1674 if (!xfs_da_state_zone)
1675 goto out_destroy_btree_cur_zone;
1677 xfs_dabuf_zone = kmem_zone_init(sizeof(xfs_dabuf_t), "xfs_dabuf");
1678 if (!xfs_dabuf_zone)
1679 goto out_destroy_da_state_zone;
1681 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1682 if (!xfs_ifork_zone)
1683 goto out_destroy_dabuf_zone;
1685 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1686 if (!xfs_trans_zone)
1687 goto out_destroy_ifork_zone;
1690 * The size of the zone allocated buf log item is the maximum
1691 * size possible under XFS. This wastes a little bit of memory,
1692 * but it is much faster.
1694 xfs_buf_item_zone = kmem_zone_init((sizeof(xfs_buf_log_item_t) +
1695 (((XFS_MAX_BLOCKSIZE / XFS_BLI_CHUNK) /
1696 NBWORD) * sizeof(int))), "xfs_buf_item");
1697 if (!xfs_buf_item_zone)
1698 goto out_destroy_trans_zone;
1700 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1701 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1702 sizeof(xfs_extent_t))), "xfs_efd_item");
1703 if (!xfs_efd_zone)
1704 goto out_destroy_buf_item_zone;
1706 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1707 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1708 sizeof(xfs_extent_t))), "xfs_efi_item");
1709 if (!xfs_efi_zone)
1710 goto out_destroy_efd_zone;
1712 xfs_inode_zone =
1713 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1714 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1715 xfs_fs_inode_init_once);
1716 if (!xfs_inode_zone)
1717 goto out_destroy_efi_zone;
1719 xfs_ili_zone =
1720 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1721 KM_ZONE_SPREAD, NULL);
1722 if (!xfs_ili_zone)
1723 goto out_destroy_inode_zone;
1725 return 0;
1727 out_destroy_inode_zone:
1728 kmem_zone_destroy(xfs_inode_zone);
1729 out_destroy_efi_zone:
1730 kmem_zone_destroy(xfs_efi_zone);
1731 out_destroy_efd_zone:
1732 kmem_zone_destroy(xfs_efd_zone);
1733 out_destroy_buf_item_zone:
1734 kmem_zone_destroy(xfs_buf_item_zone);
1735 out_destroy_trans_zone:
1736 kmem_zone_destroy(xfs_trans_zone);
1737 out_destroy_ifork_zone:
1738 kmem_zone_destroy(xfs_ifork_zone);
1739 out_destroy_dabuf_zone:
1740 kmem_zone_destroy(xfs_dabuf_zone);
1741 out_destroy_da_state_zone:
1742 kmem_zone_destroy(xfs_da_state_zone);
1743 out_destroy_btree_cur_zone:
1744 kmem_zone_destroy(xfs_btree_cur_zone);
1745 out_destroy_bmap_free_item_zone:
1746 kmem_zone_destroy(xfs_bmap_free_item_zone);
1747 out_destroy_log_ticket_zone:
1748 kmem_zone_destroy(xfs_log_ticket_zone);
1749 out_destroy_ioend_pool:
1750 mempool_destroy(xfs_ioend_pool);
1751 out_destroy_ioend_zone:
1752 kmem_zone_destroy(xfs_ioend_zone);
1753 out:
1754 return -ENOMEM;
1757 STATIC void
1758 xfs_destroy_zones(void)
1760 kmem_zone_destroy(xfs_ili_zone);
1761 kmem_zone_destroy(xfs_inode_zone);
1762 kmem_zone_destroy(xfs_efi_zone);
1763 kmem_zone_destroy(xfs_efd_zone);
1764 kmem_zone_destroy(xfs_buf_item_zone);
1765 kmem_zone_destroy(xfs_trans_zone);
1766 kmem_zone_destroy(xfs_ifork_zone);
1767 kmem_zone_destroy(xfs_dabuf_zone);
1768 kmem_zone_destroy(xfs_da_state_zone);
1769 kmem_zone_destroy(xfs_btree_cur_zone);
1770 kmem_zone_destroy(xfs_bmap_free_item_zone);
1771 kmem_zone_destroy(xfs_log_ticket_zone);
1772 mempool_destroy(xfs_ioend_pool);
1773 kmem_zone_destroy(xfs_ioend_zone);
1777 STATIC int __init
1778 init_xfs_fs(void)
1780 int error;
1782 printk(KERN_INFO XFS_VERSION_STRING " with "
1783 XFS_BUILD_OPTIONS " enabled\n");
1785 ktrace_init(64);
1786 xfs_ioend_init();
1787 xfs_dir_startup();
1789 error = xfs_init_zones();
1790 if (error)
1791 goto out;
1793 error = xfs_alloc_trace_bufs();
1794 if (error)
1795 goto out_destroy_zones;
1797 error = xfs_mru_cache_init();
1798 if (error)
1799 goto out_free_trace_buffers;
1801 error = xfs_filestream_init();
1802 if (error)
1803 goto out_mru_cache_uninit;
1805 error = xfs_buf_init();
1806 if (error)
1807 goto out_filestream_uninit;
1809 error = xfs_init_procfs();
1810 if (error)
1811 goto out_buf_terminate;
1813 error = xfs_sysctl_register();
1814 if (error)
1815 goto out_cleanup_procfs;
1817 vfs_initquota();
1819 error = register_filesystem(&xfs_fs_type);
1820 if (error)
1821 goto out_sysctl_unregister;
1822 return 0;
1824 out_sysctl_unregister:
1825 xfs_sysctl_unregister();
1826 out_cleanup_procfs:
1827 xfs_cleanup_procfs();
1828 out_buf_terminate:
1829 xfs_buf_terminate();
1830 out_filestream_uninit:
1831 xfs_filestream_uninit();
1832 out_mru_cache_uninit:
1833 xfs_mru_cache_uninit();
1834 out_free_trace_buffers:
1835 xfs_free_trace_bufs();
1836 out_destroy_zones:
1837 xfs_destroy_zones();
1838 out:
1839 return error;
1842 STATIC void __exit
1843 exit_xfs_fs(void)
1845 vfs_exitquota();
1846 unregister_filesystem(&xfs_fs_type);
1847 xfs_sysctl_unregister();
1848 xfs_cleanup_procfs();
1849 xfs_buf_terminate();
1850 xfs_filestream_uninit();
1851 xfs_mru_cache_uninit();
1852 xfs_free_trace_bufs();
1853 xfs_destroy_zones();
1854 ktrace_uninit();
1857 module_init(init_xfs_fs);
1858 module_exit(exit_xfs_fs);
1860 MODULE_AUTHOR("Silicon Graphics, Inc.");
1861 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1862 MODULE_LICENSE("GPL");