staging: most: fix hdm-dim2 build error
[linux-2.6/btrfs-unstable.git] / fs / xfs / xfs_super.c
blob1fb16562c159947ac27adae43f6abb4f195e1cfa
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_sb.h"
25 #include "xfs_mount.h"
26 #include "xfs_da_format.h"
27 #include "xfs_inode.h"
28 #include "xfs_btree.h"
29 #include "xfs_bmap.h"
30 #include "xfs_alloc.h"
31 #include "xfs_error.h"
32 #include "xfs_fsops.h"
33 #include "xfs_trans.h"
34 #include "xfs_buf_item.h"
35 #include "xfs_log.h"
36 #include "xfs_log_priv.h"
37 #include "xfs_da_btree.h"
38 #include "xfs_dir2.h"
39 #include "xfs_extfree_item.h"
40 #include "xfs_mru_cache.h"
41 #include "xfs_inode_item.h"
42 #include "xfs_icache.h"
43 #include "xfs_trace.h"
44 #include "xfs_icreate_item.h"
45 #include "xfs_filestream.h"
46 #include "xfs_quota.h"
47 #include "xfs_sysfs.h"
49 #include <linux/namei.h>
50 #include <linux/init.h>
51 #include <linux/slab.h>
52 #include <linux/mount.h>
53 #include <linux/mempool.h>
54 #include <linux/writeback.h>
55 #include <linux/kthread.h>
56 #include <linux/freezer.h>
57 #include <linux/parser.h>
59 static const struct super_operations xfs_super_operations;
60 static kmem_zone_t *xfs_ioend_zone;
61 mempool_t *xfs_ioend_pool;
63 static struct kset *xfs_kset; /* top-level xfs sysfs dir */
64 #ifdef DEBUG
65 static struct xfs_kobj xfs_dbg_kobj; /* global debug sysfs attrs */
66 #endif
68 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
69 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
70 #define MNTOPT_LOGDEV "logdev" /* log device */
71 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
72 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
73 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
74 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
75 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
76 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
77 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
78 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
79 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
80 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
81 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
82 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
83 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
84 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
85 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
86 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
87 * unwritten extent conversion */
88 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
89 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
90 #define MNTOPT_32BITINODE "inode32" /* inode allocation limited to
91 * XFS_MAXINUMBER_32 */
92 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
93 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
94 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
95 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
96 * in stat(). */
97 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
98 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
99 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
100 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
101 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
102 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
103 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
104 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
105 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
106 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
107 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
108 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
109 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
110 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
111 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
112 #define MNTOPT_DISCARD "discard" /* Discard unused blocks */
113 #define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */
115 #define MNTOPT_DAX "dax" /* Enable direct access to bdev pages */
118 * Table driven mount option parser.
120 * Currently only used for remount, but it will be used for mount
121 * in the future, too.
123 enum {
124 Opt_barrier,
125 Opt_nobarrier,
126 Opt_inode64,
127 Opt_inode32,
128 Opt_err
131 static const match_table_t tokens = {
132 {Opt_barrier, "barrier"},
133 {Opt_nobarrier, "nobarrier"},
134 {Opt_inode64, "inode64"},
135 {Opt_inode32, "inode32"},
136 {Opt_err, NULL}
140 STATIC unsigned long
141 suffix_kstrtoint(char *s, unsigned int base, int *res)
143 int last, shift_left_factor = 0, _res;
144 char *value = s;
146 last = strlen(value) - 1;
147 if (value[last] == 'K' || value[last] == 'k') {
148 shift_left_factor = 10;
149 value[last] = '\0';
151 if (value[last] == 'M' || value[last] == 'm') {
152 shift_left_factor = 20;
153 value[last] = '\0';
155 if (value[last] == 'G' || value[last] == 'g') {
156 shift_left_factor = 30;
157 value[last] = '\0';
160 if (kstrtoint(s, base, &_res))
161 return -EINVAL;
162 *res = _res << shift_left_factor;
163 return 0;
167 * This function fills in xfs_mount_t fields based on mount args.
168 * Note: the superblock has _not_ yet been read in.
170 * Note that this function leaks the various device name allocations on
171 * failure. The caller takes care of them.
173 STATIC int
174 xfs_parseargs(
175 struct xfs_mount *mp,
176 char *options)
178 struct super_block *sb = mp->m_super;
179 char *this_char, *value;
180 int dsunit = 0;
181 int dswidth = 0;
182 int iosize = 0;
183 __uint8_t iosizelog = 0;
186 * set up the mount name first so all the errors will refer to the
187 * correct device.
189 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
190 if (!mp->m_fsname)
191 return -ENOMEM;
192 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
195 * Copy binary VFS mount flags we are interested in.
197 if (sb->s_flags & MS_RDONLY)
198 mp->m_flags |= XFS_MOUNT_RDONLY;
199 if (sb->s_flags & MS_DIRSYNC)
200 mp->m_flags |= XFS_MOUNT_DIRSYNC;
201 if (sb->s_flags & MS_SYNCHRONOUS)
202 mp->m_flags |= XFS_MOUNT_WSYNC;
205 * Set some default flags that could be cleared by the mount option
206 * parsing.
208 mp->m_flags |= XFS_MOUNT_BARRIER;
209 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
212 * These can be overridden by the mount option parsing.
214 mp->m_logbufs = -1;
215 mp->m_logbsize = -1;
217 if (!options)
218 goto done;
220 while ((this_char = strsep(&options, ",")) != NULL) {
221 if (!*this_char)
222 continue;
223 if ((value = strchr(this_char, '=')) != NULL)
224 *value++ = 0;
226 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
227 if (!value || !*value) {
228 xfs_warn(mp, "%s option requires an argument",
229 this_char);
230 return -EINVAL;
232 if (kstrtoint(value, 10, &mp->m_logbufs))
233 return -EINVAL;
234 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
235 if (!value || !*value) {
236 xfs_warn(mp, "%s option requires an argument",
237 this_char);
238 return -EINVAL;
240 if (suffix_kstrtoint(value, 10, &mp->m_logbsize))
241 return -EINVAL;
242 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
243 if (!value || !*value) {
244 xfs_warn(mp, "%s option requires an argument",
245 this_char);
246 return -EINVAL;
248 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
249 if (!mp->m_logname)
250 return -ENOMEM;
251 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
252 xfs_warn(mp, "%s option not allowed on this system",
253 this_char);
254 return -EINVAL;
255 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
256 if (!value || !*value) {
257 xfs_warn(mp, "%s option requires an argument",
258 this_char);
259 return -EINVAL;
261 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
262 if (!mp->m_rtname)
263 return -ENOMEM;
264 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
265 if (!value || !*value) {
266 xfs_warn(mp, "%s option requires an argument",
267 this_char);
268 return -EINVAL;
270 if (kstrtoint(value, 10, &iosize))
271 return -EINVAL;
272 iosizelog = ffs(iosize) - 1;
273 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
274 if (!value || !*value) {
275 xfs_warn(mp, "%s option requires an argument",
276 this_char);
277 return -EINVAL;
279 if (suffix_kstrtoint(value, 10, &iosize))
280 return -EINVAL;
281 iosizelog = ffs(iosize) - 1;
282 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
283 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
284 mp->m_flags |= XFS_MOUNT_GRPID;
285 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
286 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
287 mp->m_flags &= ~XFS_MOUNT_GRPID;
288 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
289 mp->m_flags |= XFS_MOUNT_WSYNC;
290 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
291 mp->m_flags |= XFS_MOUNT_NORECOVERY;
292 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
293 mp->m_flags |= XFS_MOUNT_NOALIGN;
294 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
295 mp->m_flags |= XFS_MOUNT_SWALLOC;
296 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
297 if (!value || !*value) {
298 xfs_warn(mp, "%s option requires an argument",
299 this_char);
300 return -EINVAL;
302 if (kstrtoint(value, 10, &dsunit))
303 return -EINVAL;
304 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
305 if (!value || !*value) {
306 xfs_warn(mp, "%s option requires an argument",
307 this_char);
308 return -EINVAL;
310 if (kstrtoint(value, 10, &dswidth))
311 return -EINVAL;
312 } else if (!strcmp(this_char, MNTOPT_32BITINODE)) {
313 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
314 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
315 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
316 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
317 mp->m_flags |= XFS_MOUNT_NOUUID;
318 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
319 mp->m_flags |= XFS_MOUNT_BARRIER;
320 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
321 mp->m_flags &= ~XFS_MOUNT_BARRIER;
322 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
323 mp->m_flags |= XFS_MOUNT_IKEEP;
324 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
325 mp->m_flags &= ~XFS_MOUNT_IKEEP;
326 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
327 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
328 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
329 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
330 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
331 mp->m_flags |= XFS_MOUNT_ATTR2;
332 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
333 mp->m_flags &= ~XFS_MOUNT_ATTR2;
334 mp->m_flags |= XFS_MOUNT_NOATTR2;
335 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
336 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
337 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
338 mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT;
339 mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD;
340 mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE;
341 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
342 !strcmp(this_char, MNTOPT_UQUOTA) ||
343 !strcmp(this_char, MNTOPT_USRQUOTA)) {
344 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
345 XFS_UQUOTA_ENFD);
346 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
347 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
348 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
349 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
350 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
351 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
352 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
353 XFS_PQUOTA_ENFD);
354 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
355 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
356 mp->m_qflags &= ~XFS_PQUOTA_ENFD;
357 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
358 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
359 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
360 XFS_GQUOTA_ENFD);
361 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
362 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
363 mp->m_qflags &= ~XFS_GQUOTA_ENFD;
364 } else if (!strcmp(this_char, MNTOPT_DISCARD)) {
365 mp->m_flags |= XFS_MOUNT_DISCARD;
366 } else if (!strcmp(this_char, MNTOPT_NODISCARD)) {
367 mp->m_flags &= ~XFS_MOUNT_DISCARD;
368 #ifdef CONFIG_FS_DAX
369 } else if (!strcmp(this_char, MNTOPT_DAX)) {
370 mp->m_flags |= XFS_MOUNT_DAX;
371 #endif
372 } else {
373 xfs_warn(mp, "unknown mount option [%s].", this_char);
374 return -EINVAL;
379 * no recovery flag requires a read-only mount
381 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
382 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
383 xfs_warn(mp, "no-recovery mounts must be read-only.");
384 return -EINVAL;
387 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
388 xfs_warn(mp,
389 "sunit and swidth options incompatible with the noalign option");
390 return -EINVAL;
393 #ifndef CONFIG_XFS_QUOTA
394 if (XFS_IS_QUOTA_RUNNING(mp)) {
395 xfs_warn(mp, "quota support not available in this kernel.");
396 return -EINVAL;
398 #endif
400 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
401 xfs_warn(mp, "sunit and swidth must be specified together");
402 return -EINVAL;
405 if (dsunit && (dswidth % dsunit != 0)) {
406 xfs_warn(mp,
407 "stripe width (%d) must be a multiple of the stripe unit (%d)",
408 dswidth, dsunit);
409 return -EINVAL;
412 done:
413 if (dsunit && !(mp->m_flags & XFS_MOUNT_NOALIGN)) {
415 * At this point the superblock has not been read
416 * in, therefore we do not know the block size.
417 * Before the mount call ends we will convert
418 * these to FSBs.
420 mp->m_dalign = dsunit;
421 mp->m_swidth = dswidth;
424 if (mp->m_logbufs != -1 &&
425 mp->m_logbufs != 0 &&
426 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
427 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
428 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
429 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
430 return -EINVAL;
432 if (mp->m_logbsize != -1 &&
433 mp->m_logbsize != 0 &&
434 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
435 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
436 !is_power_of_2(mp->m_logbsize))) {
437 xfs_warn(mp,
438 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
439 mp->m_logbsize);
440 return -EINVAL;
443 if (iosizelog) {
444 if (iosizelog > XFS_MAX_IO_LOG ||
445 iosizelog < XFS_MIN_IO_LOG) {
446 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]",
447 iosizelog, XFS_MIN_IO_LOG,
448 XFS_MAX_IO_LOG);
449 return -EINVAL;
452 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
453 mp->m_readio_log = iosizelog;
454 mp->m_writeio_log = iosizelog;
457 return 0;
460 struct proc_xfs_info {
461 uint64_t flag;
462 char *str;
465 STATIC int
466 xfs_showargs(
467 struct xfs_mount *mp,
468 struct seq_file *m)
470 static struct proc_xfs_info xfs_info_set[] = {
471 /* the few simple ones we can get from the mount struct */
472 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
473 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
474 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
475 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
476 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
477 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
478 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
479 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
480 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
481 { XFS_MOUNT_DISCARD, "," MNTOPT_DISCARD },
482 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_32BITINODE },
483 { XFS_MOUNT_DAX, "," MNTOPT_DAX },
484 { 0, NULL }
486 static struct proc_xfs_info xfs_info_unset[] = {
487 /* the few simple ones we can get from the mount struct */
488 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
489 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
490 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
491 { 0, NULL }
493 struct proc_xfs_info *xfs_infop;
495 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
496 if (mp->m_flags & xfs_infop->flag)
497 seq_puts(m, xfs_infop->str);
499 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
500 if (!(mp->m_flags & xfs_infop->flag))
501 seq_puts(m, xfs_infop->str);
504 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
505 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
506 (int)(1 << mp->m_writeio_log) >> 10);
508 if (mp->m_logbufs > 0)
509 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
510 if (mp->m_logbsize > 0)
511 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
513 if (mp->m_logname)
514 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
515 if (mp->m_rtname)
516 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
518 if (mp->m_dalign > 0)
519 seq_printf(m, "," MNTOPT_SUNIT "=%d",
520 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
521 if (mp->m_swidth > 0)
522 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
523 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
525 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
526 seq_puts(m, "," MNTOPT_USRQUOTA);
527 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
528 seq_puts(m, "," MNTOPT_UQUOTANOENF);
530 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
531 if (mp->m_qflags & XFS_PQUOTA_ENFD)
532 seq_puts(m, "," MNTOPT_PRJQUOTA);
533 else
534 seq_puts(m, "," MNTOPT_PQUOTANOENF);
536 if (mp->m_qflags & XFS_GQUOTA_ACCT) {
537 if (mp->m_qflags & XFS_GQUOTA_ENFD)
538 seq_puts(m, "," MNTOPT_GRPQUOTA);
539 else
540 seq_puts(m, "," MNTOPT_GQUOTANOENF);
543 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
544 seq_puts(m, "," MNTOPT_NOQUOTA);
546 return 0;
548 __uint64_t
549 xfs_max_file_offset(
550 unsigned int blockshift)
552 unsigned int pagefactor = 1;
553 unsigned int bitshift = BITS_PER_LONG - 1;
555 /* Figure out maximum filesize, on Linux this can depend on
556 * the filesystem blocksize (on 32 bit platforms).
557 * __block_write_begin does this in an [unsigned] long...
558 * page->index << (PAGE_CACHE_SHIFT - bbits)
559 * So, for page sized blocks (4K on 32 bit platforms),
560 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
561 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
562 * but for smaller blocksizes it is less (bbits = log2 bsize).
563 * Note1: get_block_t takes a long (implicit cast from above)
564 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
565 * can optionally convert the [unsigned] long from above into
566 * an [unsigned] long long.
569 #if BITS_PER_LONG == 32
570 # if defined(CONFIG_LBDAF)
571 ASSERT(sizeof(sector_t) == 8);
572 pagefactor = PAGE_CACHE_SIZE;
573 bitshift = BITS_PER_LONG;
574 # else
575 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
576 # endif
577 #endif
579 return (((__uint64_t)pagefactor) << bitshift) - 1;
583 * xfs_set_inode32() and xfs_set_inode64() are passed an agcount
584 * because in the growfs case, mp->m_sb.sb_agcount is not updated
585 * yet to the potentially higher ag count.
587 xfs_agnumber_t
588 xfs_set_inode32(struct xfs_mount *mp, xfs_agnumber_t agcount)
590 xfs_agnumber_t index = 0;
591 xfs_agnumber_t maxagi = 0;
592 xfs_sb_t *sbp = &mp->m_sb;
593 xfs_agnumber_t max_metadata;
594 xfs_agino_t agino;
595 xfs_ino_t ino;
596 xfs_perag_t *pag;
598 /* Calculate how much should be reserved for inodes to meet
599 * the max inode percentage.
601 if (mp->m_maxicount) {
602 __uint64_t icount;
604 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
605 do_div(icount, 100);
606 icount += sbp->sb_agblocks - 1;
607 do_div(icount, sbp->sb_agblocks);
608 max_metadata = icount;
609 } else {
610 max_metadata = agcount;
613 agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks - 1, 0);
615 for (index = 0; index < agcount; index++) {
616 ino = XFS_AGINO_TO_INO(mp, index, agino);
618 if (ino > XFS_MAXINUMBER_32) {
619 pag = xfs_perag_get(mp, index);
620 pag->pagi_inodeok = 0;
621 pag->pagf_metadata = 0;
622 xfs_perag_put(pag);
623 continue;
626 pag = xfs_perag_get(mp, index);
627 pag->pagi_inodeok = 1;
628 maxagi++;
629 if (index < max_metadata)
630 pag->pagf_metadata = 1;
631 xfs_perag_put(pag);
633 mp->m_flags |= (XFS_MOUNT_32BITINODES |
634 XFS_MOUNT_SMALL_INUMS);
636 return maxagi;
639 xfs_agnumber_t
640 xfs_set_inode64(struct xfs_mount *mp, xfs_agnumber_t agcount)
642 xfs_agnumber_t index = 0;
644 for (index = 0; index < agcount; index++) {
645 struct xfs_perag *pag;
647 pag = xfs_perag_get(mp, index);
648 pag->pagi_inodeok = 1;
649 pag->pagf_metadata = 0;
650 xfs_perag_put(pag);
653 /* There is no need for lock protection on m_flags,
654 * the rw_semaphore of the VFS superblock is locked
655 * during mount/umount/remount operations, so this is
656 * enough to avoid concurency on the m_flags field
658 mp->m_flags &= ~(XFS_MOUNT_32BITINODES |
659 XFS_MOUNT_SMALL_INUMS);
660 return index;
663 STATIC int
664 xfs_blkdev_get(
665 xfs_mount_t *mp,
666 const char *name,
667 struct block_device **bdevp)
669 int error = 0;
671 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
672 mp);
673 if (IS_ERR(*bdevp)) {
674 error = PTR_ERR(*bdevp);
675 xfs_warn(mp, "Invalid device [%s], error=%d", name, error);
678 return error;
681 STATIC void
682 xfs_blkdev_put(
683 struct block_device *bdev)
685 if (bdev)
686 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
689 void
690 xfs_blkdev_issue_flush(
691 xfs_buftarg_t *buftarg)
693 blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL);
696 STATIC void
697 xfs_close_devices(
698 struct xfs_mount *mp)
700 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
701 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
702 xfs_free_buftarg(mp, mp->m_logdev_targp);
703 xfs_blkdev_put(logdev);
705 if (mp->m_rtdev_targp) {
706 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
707 xfs_free_buftarg(mp, mp->m_rtdev_targp);
708 xfs_blkdev_put(rtdev);
710 xfs_free_buftarg(mp, mp->m_ddev_targp);
714 * The file system configurations are:
715 * (1) device (partition) with data and internal log
716 * (2) logical volume with data and log subvolumes.
717 * (3) logical volume with data, log, and realtime subvolumes.
719 * We only have to handle opening the log and realtime volumes here if
720 * they are present. The data subvolume has already been opened by
721 * get_sb_bdev() and is stored in sb->s_bdev.
723 STATIC int
724 xfs_open_devices(
725 struct xfs_mount *mp)
727 struct block_device *ddev = mp->m_super->s_bdev;
728 struct block_device *logdev = NULL, *rtdev = NULL;
729 int error;
732 * Open real time and log devices - order is important.
734 if (mp->m_logname) {
735 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
736 if (error)
737 goto out;
740 if (mp->m_rtname) {
741 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
742 if (error)
743 goto out_close_logdev;
745 if (rtdev == ddev || rtdev == logdev) {
746 xfs_warn(mp,
747 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
748 error = -EINVAL;
749 goto out_close_rtdev;
754 * Setup xfs_mount buffer target pointers
756 error = -ENOMEM;
757 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev);
758 if (!mp->m_ddev_targp)
759 goto out_close_rtdev;
761 if (rtdev) {
762 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev);
763 if (!mp->m_rtdev_targp)
764 goto out_free_ddev_targ;
767 if (logdev && logdev != ddev) {
768 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev);
769 if (!mp->m_logdev_targp)
770 goto out_free_rtdev_targ;
771 } else {
772 mp->m_logdev_targp = mp->m_ddev_targp;
775 return 0;
777 out_free_rtdev_targ:
778 if (mp->m_rtdev_targp)
779 xfs_free_buftarg(mp, mp->m_rtdev_targp);
780 out_free_ddev_targ:
781 xfs_free_buftarg(mp, mp->m_ddev_targp);
782 out_close_rtdev:
783 xfs_blkdev_put(rtdev);
784 out_close_logdev:
785 if (logdev && logdev != ddev)
786 xfs_blkdev_put(logdev);
787 out:
788 return error;
792 * Setup xfs_mount buffer target pointers based on superblock
794 STATIC int
795 xfs_setup_devices(
796 struct xfs_mount *mp)
798 int error;
800 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize);
801 if (error)
802 return error;
804 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
805 unsigned int log_sector_size = BBSIZE;
807 if (xfs_sb_version_hassector(&mp->m_sb))
808 log_sector_size = mp->m_sb.sb_logsectsize;
809 error = xfs_setsize_buftarg(mp->m_logdev_targp,
810 log_sector_size);
811 if (error)
812 return error;
814 if (mp->m_rtdev_targp) {
815 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
816 mp->m_sb.sb_sectsize);
817 if (error)
818 return error;
821 return 0;
824 STATIC int
825 xfs_init_mount_workqueues(
826 struct xfs_mount *mp)
828 mp->m_buf_workqueue = alloc_workqueue("xfs-buf/%s",
829 WQ_MEM_RECLAIM|WQ_FREEZABLE, 1, mp->m_fsname);
830 if (!mp->m_buf_workqueue)
831 goto out;
833 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s",
834 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
835 if (!mp->m_data_workqueue)
836 goto out_destroy_buf;
838 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
839 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
840 if (!mp->m_unwritten_workqueue)
841 goto out_destroy_data_iodone_queue;
843 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s",
844 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
845 if (!mp->m_cil_workqueue)
846 goto out_destroy_unwritten;
848 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
849 WQ_FREEZABLE, 0, mp->m_fsname);
850 if (!mp->m_reclaim_workqueue)
851 goto out_destroy_cil;
853 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
854 WQ_FREEZABLE|WQ_HIGHPRI, 0, mp->m_fsname);
855 if (!mp->m_log_workqueue)
856 goto out_destroy_reclaim;
858 mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s",
859 WQ_FREEZABLE, 0, mp->m_fsname);
860 if (!mp->m_eofblocks_workqueue)
861 goto out_destroy_log;
863 return 0;
865 out_destroy_log:
866 destroy_workqueue(mp->m_log_workqueue);
867 out_destroy_reclaim:
868 destroy_workqueue(mp->m_reclaim_workqueue);
869 out_destroy_cil:
870 destroy_workqueue(mp->m_cil_workqueue);
871 out_destroy_unwritten:
872 destroy_workqueue(mp->m_unwritten_workqueue);
873 out_destroy_data_iodone_queue:
874 destroy_workqueue(mp->m_data_workqueue);
875 out_destroy_buf:
876 destroy_workqueue(mp->m_buf_workqueue);
877 out:
878 return -ENOMEM;
881 STATIC void
882 xfs_destroy_mount_workqueues(
883 struct xfs_mount *mp)
885 destroy_workqueue(mp->m_eofblocks_workqueue);
886 destroy_workqueue(mp->m_log_workqueue);
887 destroy_workqueue(mp->m_reclaim_workqueue);
888 destroy_workqueue(mp->m_cil_workqueue);
889 destroy_workqueue(mp->m_data_workqueue);
890 destroy_workqueue(mp->m_unwritten_workqueue);
891 destroy_workqueue(mp->m_buf_workqueue);
895 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
896 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
897 * for IO to complete so that we effectively throttle multiple callers to the
898 * rate at which IO is completing.
900 void
901 xfs_flush_inodes(
902 struct xfs_mount *mp)
904 struct super_block *sb = mp->m_super;
906 if (down_read_trylock(&sb->s_umount)) {
907 sync_inodes_sb(sb);
908 up_read(&sb->s_umount);
912 /* Catch misguided souls that try to use this interface on XFS */
913 STATIC struct inode *
914 xfs_fs_alloc_inode(
915 struct super_block *sb)
917 BUG();
918 return NULL;
922 * Now that the generic code is guaranteed not to be accessing
923 * the linux inode, we can reclaim the inode.
925 STATIC void
926 xfs_fs_destroy_inode(
927 struct inode *inode)
929 struct xfs_inode *ip = XFS_I(inode);
931 trace_xfs_destroy_inode(ip);
933 XFS_STATS_INC(vn_reclaim);
935 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
938 * We should never get here with one of the reclaim flags already set.
940 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
941 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
944 * We always use background reclaim here because even if the
945 * inode is clean, it still may be under IO and hence we have
946 * to take the flush lock. The background reclaim path handles
947 * this more efficiently than we can here, so simply let background
948 * reclaim tear down all inodes.
950 xfs_inode_set_reclaim_tag(ip);
954 * Slab object creation initialisation for the XFS inode.
955 * This covers only the idempotent fields in the XFS inode;
956 * all other fields need to be initialised on allocation
957 * from the slab. This avoids the need to repeatedly initialise
958 * fields in the xfs inode that left in the initialise state
959 * when freeing the inode.
961 STATIC void
962 xfs_fs_inode_init_once(
963 void *inode)
965 struct xfs_inode *ip = inode;
967 memset(ip, 0, sizeof(struct xfs_inode));
969 /* vfs inode */
970 inode_init_once(VFS_I(ip));
972 /* xfs inode */
973 atomic_set(&ip->i_pincount, 0);
974 spin_lock_init(&ip->i_flags_lock);
976 mrlock_init(&ip->i_mmaplock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
977 "xfsino", ip->i_ino);
978 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
979 "xfsino", ip->i_ino);
982 STATIC void
983 xfs_fs_evict_inode(
984 struct inode *inode)
986 xfs_inode_t *ip = XFS_I(inode);
988 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
990 trace_xfs_evict_inode(ip);
992 truncate_inode_pages_final(&inode->i_data);
993 clear_inode(inode);
994 XFS_STATS_INC(vn_rele);
995 XFS_STATS_INC(vn_remove);
997 xfs_inactive(ip);
1001 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1002 * serialised against cache hits here via the inode->i_lock and igrab() in
1003 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1004 * racing with us, and it avoids needing to grab a spinlock here for every inode
1005 * we drop the final reference on.
1007 STATIC int
1008 xfs_fs_drop_inode(
1009 struct inode *inode)
1011 struct xfs_inode *ip = XFS_I(inode);
1013 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);
1016 STATIC void
1017 xfs_free_fsname(
1018 struct xfs_mount *mp)
1020 kfree(mp->m_fsname);
1021 kfree(mp->m_rtname);
1022 kfree(mp->m_logname);
1025 STATIC int
1026 xfs_fs_sync_fs(
1027 struct super_block *sb,
1028 int wait)
1030 struct xfs_mount *mp = XFS_M(sb);
1033 * Doing anything during the async pass would be counterproductive.
1035 if (!wait)
1036 return 0;
1038 xfs_log_force(mp, XFS_LOG_SYNC);
1039 if (laptop_mode) {
1041 * The disk must be active because we're syncing.
1042 * We schedule log work now (now that the disk is
1043 * active) instead of later (when it might not be).
1045 flush_delayed_work(&mp->m_log->l_work);
1048 return 0;
1051 STATIC int
1052 xfs_fs_statfs(
1053 struct dentry *dentry,
1054 struct kstatfs *statp)
1056 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1057 xfs_sb_t *sbp = &mp->m_sb;
1058 struct xfs_inode *ip = XFS_I(d_inode(dentry));
1059 __uint64_t fakeinos, id;
1060 __uint64_t icount;
1061 __uint64_t ifree;
1062 __uint64_t fdblocks;
1063 xfs_extlen_t lsize;
1064 __int64_t ffree;
1066 statp->f_type = XFS_SB_MAGIC;
1067 statp->f_namelen = MAXNAMELEN - 1;
1069 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1070 statp->f_fsid.val[0] = (u32)id;
1071 statp->f_fsid.val[1] = (u32)(id >> 32);
1073 icount = percpu_counter_sum(&mp->m_icount);
1074 ifree = percpu_counter_sum(&mp->m_ifree);
1075 fdblocks = percpu_counter_sum(&mp->m_fdblocks);
1077 spin_lock(&mp->m_sb_lock);
1078 statp->f_bsize = sbp->sb_blocksize;
1079 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1080 statp->f_blocks = sbp->sb_dblocks - lsize;
1081 spin_unlock(&mp->m_sb_lock);
1083 statp->f_bfree = fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1084 statp->f_bavail = statp->f_bfree;
1086 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1087 statp->f_files = MIN(icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1088 if (mp->m_maxicount)
1089 statp->f_files = min_t(typeof(statp->f_files),
1090 statp->f_files,
1091 mp->m_maxicount);
1093 /* If sb_icount overshot maxicount, report actual allocation */
1094 statp->f_files = max_t(typeof(statp->f_files),
1095 statp->f_files,
1096 sbp->sb_icount);
1098 /* make sure statp->f_ffree does not underflow */
1099 ffree = statp->f_files - (icount - ifree);
1100 statp->f_ffree = max_t(__int64_t, ffree, 0);
1103 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
1104 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) ==
1105 (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))
1106 xfs_qm_statvfs(ip, statp);
1107 return 0;
1110 STATIC void
1111 xfs_save_resvblks(struct xfs_mount *mp)
1113 __uint64_t resblks = 0;
1115 mp->m_resblks_save = mp->m_resblks;
1116 xfs_reserve_blocks(mp, &resblks, NULL);
1119 STATIC void
1120 xfs_restore_resvblks(struct xfs_mount *mp)
1122 __uint64_t resblks;
1124 if (mp->m_resblks_save) {
1125 resblks = mp->m_resblks_save;
1126 mp->m_resblks_save = 0;
1127 } else
1128 resblks = xfs_default_resblks(mp);
1130 xfs_reserve_blocks(mp, &resblks, NULL);
1134 * Trigger writeback of all the dirty metadata in the file system.
1136 * This ensures that the metadata is written to their location on disk rather
1137 * than just existing in transactions in the log. This means after a quiesce
1138 * there is no log replay required to write the inodes to disk - this is the
1139 * primary difference between a sync and a quiesce.
1141 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1142 * it is started again when appropriate.
1144 static void
1145 xfs_quiesce_attr(
1146 struct xfs_mount *mp)
1148 int error = 0;
1150 /* wait for all modifications to complete */
1151 while (atomic_read(&mp->m_active_trans) > 0)
1152 delay(100);
1154 /* force the log to unpin objects from the now complete transactions */
1155 xfs_log_force(mp, XFS_LOG_SYNC);
1157 /* reclaim inodes to do any IO before the freeze completes */
1158 xfs_reclaim_inodes(mp, 0);
1159 xfs_reclaim_inodes(mp, SYNC_WAIT);
1161 /* Push the superblock and write an unmount record */
1162 error = xfs_log_sbcount(mp);
1163 if (error)
1164 xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
1165 "Frozen image may not be consistent.");
1167 * Just warn here till VFS can correctly support
1168 * read-only remount without racing.
1170 WARN_ON(atomic_read(&mp->m_active_trans) != 0);
1172 xfs_log_quiesce(mp);
1175 STATIC int
1176 xfs_fs_remount(
1177 struct super_block *sb,
1178 int *flags,
1179 char *options)
1181 struct xfs_mount *mp = XFS_M(sb);
1182 xfs_sb_t *sbp = &mp->m_sb;
1183 substring_t args[MAX_OPT_ARGS];
1184 char *p;
1185 int error;
1187 sync_filesystem(sb);
1188 while ((p = strsep(&options, ",")) != NULL) {
1189 int token;
1191 if (!*p)
1192 continue;
1194 token = match_token(p, tokens, args);
1195 switch (token) {
1196 case Opt_barrier:
1197 mp->m_flags |= XFS_MOUNT_BARRIER;
1198 break;
1199 case Opt_nobarrier:
1200 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1201 break;
1202 case Opt_inode64:
1203 mp->m_maxagi = xfs_set_inode64(mp, sbp->sb_agcount);
1204 break;
1205 case Opt_inode32:
1206 mp->m_maxagi = xfs_set_inode32(mp, sbp->sb_agcount);
1207 break;
1208 default:
1210 * Logically we would return an error here to prevent
1211 * users from believing they might have changed
1212 * mount options using remount which can't be changed.
1214 * But unfortunately mount(8) adds all options from
1215 * mtab and fstab to the mount arguments in some cases
1216 * so we can't blindly reject options, but have to
1217 * check for each specified option if it actually
1218 * differs from the currently set option and only
1219 * reject it if that's the case.
1221 * Until that is implemented we return success for
1222 * every remount request, and silently ignore all
1223 * options that we can't actually change.
1225 #if 0
1226 xfs_info(mp,
1227 "mount option \"%s\" not supported for remount", p);
1228 return -EINVAL;
1229 #else
1230 break;
1231 #endif
1235 /* ro -> rw */
1236 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1237 if (mp->m_flags & XFS_MOUNT_NORECOVERY) {
1238 xfs_warn(mp,
1239 "ro->rw transition prohibited on norecovery mount");
1240 return -EINVAL;
1243 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1246 * If this is the first remount to writeable state we
1247 * might have some superblock changes to update.
1249 if (mp->m_update_sb) {
1250 error = xfs_sync_sb(mp, false);
1251 if (error) {
1252 xfs_warn(mp, "failed to write sb changes");
1253 return error;
1255 mp->m_update_sb = false;
1259 * Fill out the reserve pool if it is empty. Use the stashed
1260 * value if it is non-zero, otherwise go with the default.
1262 xfs_restore_resvblks(mp);
1263 xfs_log_work_queue(mp);
1266 /* rw -> ro */
1267 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1269 * Before we sync the metadata, we need to free up the reserve
1270 * block pool so that the used block count in the superblock on
1271 * disk is correct at the end of the remount. Stash the current
1272 * reserve pool size so that if we get remounted rw, we can
1273 * return it to the same size.
1275 xfs_save_resvblks(mp);
1276 xfs_quiesce_attr(mp);
1277 mp->m_flags |= XFS_MOUNT_RDONLY;
1280 return 0;
1284 * Second stage of a freeze. The data is already frozen so we only
1285 * need to take care of the metadata. Once that's done sync the superblock
1286 * to the log to dirty it in case of a crash while frozen. This ensures that we
1287 * will recover the unlinked inode lists on the next mount.
1289 STATIC int
1290 xfs_fs_freeze(
1291 struct super_block *sb)
1293 struct xfs_mount *mp = XFS_M(sb);
1295 xfs_save_resvblks(mp);
1296 xfs_quiesce_attr(mp);
1297 return xfs_sync_sb(mp, true);
1300 STATIC int
1301 xfs_fs_unfreeze(
1302 struct super_block *sb)
1304 struct xfs_mount *mp = XFS_M(sb);
1306 xfs_restore_resvblks(mp);
1307 xfs_log_work_queue(mp);
1308 return 0;
1311 STATIC int
1312 xfs_fs_show_options(
1313 struct seq_file *m,
1314 struct dentry *root)
1316 return xfs_showargs(XFS_M(root->d_sb), m);
1320 * This function fills in xfs_mount_t fields based on mount args.
1321 * Note: the superblock _has_ now been read in.
1323 STATIC int
1324 xfs_finish_flags(
1325 struct xfs_mount *mp)
1327 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1329 /* Fail a mount where the logbuf is smaller than the log stripe */
1330 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1331 if (mp->m_logbsize <= 0 &&
1332 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1333 mp->m_logbsize = mp->m_sb.sb_logsunit;
1334 } else if (mp->m_logbsize > 0 &&
1335 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1336 xfs_warn(mp,
1337 "logbuf size must be greater than or equal to log stripe size");
1338 return -EINVAL;
1340 } else {
1341 /* Fail a mount if the logbuf is larger than 32K */
1342 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1343 xfs_warn(mp,
1344 "logbuf size for version 1 logs must be 16K or 32K");
1345 return -EINVAL;
1350 * V5 filesystems always use attr2 format for attributes.
1352 if (xfs_sb_version_hascrc(&mp->m_sb) &&
1353 (mp->m_flags & XFS_MOUNT_NOATTR2)) {
1354 xfs_warn(mp,
1355 "Cannot mount a V5 filesystem as %s. %s is always enabled for V5 filesystems.",
1356 MNTOPT_NOATTR2, MNTOPT_ATTR2);
1357 return -EINVAL;
1361 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1362 * told by noattr2 to turn it off
1364 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1365 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1366 mp->m_flags |= XFS_MOUNT_ATTR2;
1369 * prohibit r/w mounts of read-only filesystems
1371 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1372 xfs_warn(mp,
1373 "cannot mount a read-only filesystem as read-write");
1374 return -EROFS;
1377 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
1378 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) &&
1379 !xfs_sb_version_has_pquotino(&mp->m_sb)) {
1380 xfs_warn(mp,
1381 "Super block does not support project and group quota together");
1382 return -EINVAL;
1385 return 0;
1388 static int
1389 xfs_init_percpu_counters(
1390 struct xfs_mount *mp)
1392 int error;
1394 error = percpu_counter_init(&mp->m_icount, 0, GFP_KERNEL);
1395 if (error)
1396 return -ENOMEM;
1398 error = percpu_counter_init(&mp->m_ifree, 0, GFP_KERNEL);
1399 if (error)
1400 goto free_icount;
1402 error = percpu_counter_init(&mp->m_fdblocks, 0, GFP_KERNEL);
1403 if (error)
1404 goto free_ifree;
1406 return 0;
1408 free_ifree:
1409 percpu_counter_destroy(&mp->m_ifree);
1410 free_icount:
1411 percpu_counter_destroy(&mp->m_icount);
1412 return -ENOMEM;
1415 void
1416 xfs_reinit_percpu_counters(
1417 struct xfs_mount *mp)
1419 percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount);
1420 percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree);
1421 percpu_counter_set(&mp->m_fdblocks, mp->m_sb.sb_fdblocks);
1424 static void
1425 xfs_destroy_percpu_counters(
1426 struct xfs_mount *mp)
1428 percpu_counter_destroy(&mp->m_icount);
1429 percpu_counter_destroy(&mp->m_ifree);
1430 percpu_counter_destroy(&mp->m_fdblocks);
1433 STATIC int
1434 xfs_fs_fill_super(
1435 struct super_block *sb,
1436 void *data,
1437 int silent)
1439 struct inode *root;
1440 struct xfs_mount *mp = NULL;
1441 int flags = 0, error = -ENOMEM;
1443 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1444 if (!mp)
1445 goto out;
1447 spin_lock_init(&mp->m_sb_lock);
1448 mutex_init(&mp->m_growlock);
1449 atomic_set(&mp->m_active_trans, 0);
1450 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1451 INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
1452 mp->m_kobj.kobject.kset = xfs_kset;
1454 mp->m_super = sb;
1455 sb->s_fs_info = mp;
1457 error = xfs_parseargs(mp, (char *)data);
1458 if (error)
1459 goto out_free_fsname;
1461 sb_min_blocksize(sb, BBSIZE);
1462 sb->s_xattr = xfs_xattr_handlers;
1463 sb->s_export_op = &xfs_export_operations;
1464 #ifdef CONFIG_XFS_QUOTA
1465 sb->s_qcop = &xfs_quotactl_operations;
1466 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
1467 #endif
1468 sb->s_op = &xfs_super_operations;
1470 if (silent)
1471 flags |= XFS_MFSI_QUIET;
1473 error = xfs_open_devices(mp);
1474 if (error)
1475 goto out_free_fsname;
1477 error = xfs_init_mount_workqueues(mp);
1478 if (error)
1479 goto out_close_devices;
1481 error = xfs_init_percpu_counters(mp);
1482 if (error)
1483 goto out_destroy_workqueues;
1485 error = xfs_readsb(mp, flags);
1486 if (error)
1487 goto out_destroy_counters;
1489 error = xfs_finish_flags(mp);
1490 if (error)
1491 goto out_free_sb;
1493 error = xfs_setup_devices(mp);
1494 if (error)
1495 goto out_free_sb;
1497 error = xfs_filestream_mount(mp);
1498 if (error)
1499 goto out_free_sb;
1502 * we must configure the block size in the superblock before we run the
1503 * full mount process as the mount process can lookup and cache inodes.
1505 sb->s_magic = XFS_SB_MAGIC;
1506 sb->s_blocksize = mp->m_sb.sb_blocksize;
1507 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1508 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1509 sb->s_max_links = XFS_MAXLINK;
1510 sb->s_time_gran = 1;
1511 set_posix_acl_flag(sb);
1513 /* version 5 superblocks support inode version counters. */
1514 if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
1515 sb->s_flags |= MS_I_VERSION;
1517 if (mp->m_flags & XFS_MOUNT_DAX) {
1518 xfs_warn(mp,
1519 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1520 if (sb->s_blocksize != PAGE_SIZE) {
1521 xfs_alert(mp,
1522 "Filesystem block size invalid for DAX Turning DAX off.");
1523 mp->m_flags &= ~XFS_MOUNT_DAX;
1524 } else if (!sb->s_bdev->bd_disk->fops->direct_access) {
1525 xfs_alert(mp,
1526 "Block device does not support DAX Turning DAX off.");
1527 mp->m_flags &= ~XFS_MOUNT_DAX;
1531 error = xfs_mountfs(mp);
1532 if (error)
1533 goto out_filestream_unmount;
1535 root = igrab(VFS_I(mp->m_rootip));
1536 if (!root) {
1537 error = -ENOENT;
1538 goto out_unmount;
1540 sb->s_root = d_make_root(root);
1541 if (!sb->s_root) {
1542 error = -ENOMEM;
1543 goto out_unmount;
1546 return 0;
1548 out_filestream_unmount:
1549 xfs_filestream_unmount(mp);
1550 out_free_sb:
1551 xfs_freesb(mp);
1552 out_destroy_counters:
1553 xfs_destroy_percpu_counters(mp);
1554 out_destroy_workqueues:
1555 xfs_destroy_mount_workqueues(mp);
1556 out_close_devices:
1557 xfs_close_devices(mp);
1558 out_free_fsname:
1559 xfs_free_fsname(mp);
1560 kfree(mp);
1561 out:
1562 return error;
1564 out_unmount:
1565 xfs_filestream_unmount(mp);
1566 xfs_unmountfs(mp);
1567 goto out_free_sb;
1570 STATIC void
1571 xfs_fs_put_super(
1572 struct super_block *sb)
1574 struct xfs_mount *mp = XFS_M(sb);
1576 xfs_notice(mp, "Unmounting Filesystem");
1577 xfs_filestream_unmount(mp);
1578 xfs_unmountfs(mp);
1580 xfs_freesb(mp);
1581 xfs_destroy_percpu_counters(mp);
1582 xfs_destroy_mount_workqueues(mp);
1583 xfs_close_devices(mp);
1584 xfs_free_fsname(mp);
1585 kfree(mp);
1588 STATIC struct dentry *
1589 xfs_fs_mount(
1590 struct file_system_type *fs_type,
1591 int flags,
1592 const char *dev_name,
1593 void *data)
1595 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1598 static long
1599 xfs_fs_nr_cached_objects(
1600 struct super_block *sb,
1601 struct shrink_control *sc)
1603 return xfs_reclaim_inodes_count(XFS_M(sb));
1606 static long
1607 xfs_fs_free_cached_objects(
1608 struct super_block *sb,
1609 struct shrink_control *sc)
1611 return xfs_reclaim_inodes_nr(XFS_M(sb), sc->nr_to_scan);
1614 static const struct super_operations xfs_super_operations = {
1615 .alloc_inode = xfs_fs_alloc_inode,
1616 .destroy_inode = xfs_fs_destroy_inode,
1617 .evict_inode = xfs_fs_evict_inode,
1618 .drop_inode = xfs_fs_drop_inode,
1619 .put_super = xfs_fs_put_super,
1620 .sync_fs = xfs_fs_sync_fs,
1621 .freeze_fs = xfs_fs_freeze,
1622 .unfreeze_fs = xfs_fs_unfreeze,
1623 .statfs = xfs_fs_statfs,
1624 .remount_fs = xfs_fs_remount,
1625 .show_options = xfs_fs_show_options,
1626 .nr_cached_objects = xfs_fs_nr_cached_objects,
1627 .free_cached_objects = xfs_fs_free_cached_objects,
1630 static struct file_system_type xfs_fs_type = {
1631 .owner = THIS_MODULE,
1632 .name = "xfs",
1633 .mount = xfs_fs_mount,
1634 .kill_sb = kill_block_super,
1635 .fs_flags = FS_REQUIRES_DEV,
1637 MODULE_ALIAS_FS("xfs");
1639 STATIC int __init
1640 xfs_init_zones(void)
1643 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1644 if (!xfs_ioend_zone)
1645 goto out;
1647 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1648 xfs_ioend_zone);
1649 if (!xfs_ioend_pool)
1650 goto out_destroy_ioend_zone;
1652 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1653 "xfs_log_ticket");
1654 if (!xfs_log_ticket_zone)
1655 goto out_destroy_ioend_pool;
1657 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1658 "xfs_bmap_free_item");
1659 if (!xfs_bmap_free_item_zone)
1660 goto out_destroy_log_ticket_zone;
1662 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1663 "xfs_btree_cur");
1664 if (!xfs_btree_cur_zone)
1665 goto out_destroy_bmap_free_item_zone;
1667 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1668 "xfs_da_state");
1669 if (!xfs_da_state_zone)
1670 goto out_destroy_btree_cur_zone;
1672 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1673 if (!xfs_ifork_zone)
1674 goto out_destroy_da_state_zone;
1676 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1677 if (!xfs_trans_zone)
1678 goto out_destroy_ifork_zone;
1680 xfs_log_item_desc_zone =
1681 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1682 "xfs_log_item_desc");
1683 if (!xfs_log_item_desc_zone)
1684 goto out_destroy_trans_zone;
1687 * The size of the zone allocated buf log item is the maximum
1688 * size possible under XFS. This wastes a little bit of memory,
1689 * but it is much faster.
1691 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item),
1692 "xfs_buf_item");
1693 if (!xfs_buf_item_zone)
1694 goto out_destroy_log_item_desc_zone;
1696 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1697 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1698 sizeof(xfs_extent_t))), "xfs_efd_item");
1699 if (!xfs_efd_zone)
1700 goto out_destroy_buf_item_zone;
1702 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1703 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1704 sizeof(xfs_extent_t))), "xfs_efi_item");
1705 if (!xfs_efi_zone)
1706 goto out_destroy_efd_zone;
1708 xfs_inode_zone =
1709 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1710 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1711 xfs_fs_inode_init_once);
1712 if (!xfs_inode_zone)
1713 goto out_destroy_efi_zone;
1715 xfs_ili_zone =
1716 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1717 KM_ZONE_SPREAD, NULL);
1718 if (!xfs_ili_zone)
1719 goto out_destroy_inode_zone;
1720 xfs_icreate_zone = kmem_zone_init(sizeof(struct xfs_icreate_item),
1721 "xfs_icr");
1722 if (!xfs_icreate_zone)
1723 goto out_destroy_ili_zone;
1725 return 0;
1727 out_destroy_ili_zone:
1728 kmem_zone_destroy(xfs_ili_zone);
1729 out_destroy_inode_zone:
1730 kmem_zone_destroy(xfs_inode_zone);
1731 out_destroy_efi_zone:
1732 kmem_zone_destroy(xfs_efi_zone);
1733 out_destroy_efd_zone:
1734 kmem_zone_destroy(xfs_efd_zone);
1735 out_destroy_buf_item_zone:
1736 kmem_zone_destroy(xfs_buf_item_zone);
1737 out_destroy_log_item_desc_zone:
1738 kmem_zone_destroy(xfs_log_item_desc_zone);
1739 out_destroy_trans_zone:
1740 kmem_zone_destroy(xfs_trans_zone);
1741 out_destroy_ifork_zone:
1742 kmem_zone_destroy(xfs_ifork_zone);
1743 out_destroy_da_state_zone:
1744 kmem_zone_destroy(xfs_da_state_zone);
1745 out_destroy_btree_cur_zone:
1746 kmem_zone_destroy(xfs_btree_cur_zone);
1747 out_destroy_bmap_free_item_zone:
1748 kmem_zone_destroy(xfs_bmap_free_item_zone);
1749 out_destroy_log_ticket_zone:
1750 kmem_zone_destroy(xfs_log_ticket_zone);
1751 out_destroy_ioend_pool:
1752 mempool_destroy(xfs_ioend_pool);
1753 out_destroy_ioend_zone:
1754 kmem_zone_destroy(xfs_ioend_zone);
1755 out:
1756 return -ENOMEM;
1759 STATIC void
1760 xfs_destroy_zones(void)
1763 * Make sure all delayed rcu free are flushed before we
1764 * destroy caches.
1766 rcu_barrier();
1767 kmem_zone_destroy(xfs_icreate_zone);
1768 kmem_zone_destroy(xfs_ili_zone);
1769 kmem_zone_destroy(xfs_inode_zone);
1770 kmem_zone_destroy(xfs_efi_zone);
1771 kmem_zone_destroy(xfs_efd_zone);
1772 kmem_zone_destroy(xfs_buf_item_zone);
1773 kmem_zone_destroy(xfs_log_item_desc_zone);
1774 kmem_zone_destroy(xfs_trans_zone);
1775 kmem_zone_destroy(xfs_ifork_zone);
1776 kmem_zone_destroy(xfs_da_state_zone);
1777 kmem_zone_destroy(xfs_btree_cur_zone);
1778 kmem_zone_destroy(xfs_bmap_free_item_zone);
1779 kmem_zone_destroy(xfs_log_ticket_zone);
1780 mempool_destroy(xfs_ioend_pool);
1781 kmem_zone_destroy(xfs_ioend_zone);
1785 STATIC int __init
1786 xfs_init_workqueues(void)
1789 * The allocation workqueue can be used in memory reclaim situations
1790 * (writepage path), and parallelism is only limited by the number of
1791 * AGs in all the filesystems mounted. Hence use the default large
1792 * max_active value for this workqueue.
1794 xfs_alloc_wq = alloc_workqueue("xfsalloc",
1795 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0);
1796 if (!xfs_alloc_wq)
1797 return -ENOMEM;
1799 return 0;
1802 STATIC void
1803 xfs_destroy_workqueues(void)
1805 destroy_workqueue(xfs_alloc_wq);
1808 STATIC int __init
1809 init_xfs_fs(void)
1811 int error;
1813 printk(KERN_INFO XFS_VERSION_STRING " with "
1814 XFS_BUILD_OPTIONS " enabled\n");
1816 xfs_dir_startup();
1818 error = xfs_init_zones();
1819 if (error)
1820 goto out;
1822 error = xfs_init_workqueues();
1823 if (error)
1824 goto out_destroy_zones;
1826 error = xfs_mru_cache_init();
1827 if (error)
1828 goto out_destroy_wq;
1830 error = xfs_buf_init();
1831 if (error)
1832 goto out_mru_cache_uninit;
1834 error = xfs_init_procfs();
1835 if (error)
1836 goto out_buf_terminate;
1838 error = xfs_sysctl_register();
1839 if (error)
1840 goto out_cleanup_procfs;
1842 xfs_kset = kset_create_and_add("xfs", NULL, fs_kobj);
1843 if (!xfs_kset) {
1844 error = -ENOMEM;
1845 goto out_sysctl_unregister;;
1848 #ifdef DEBUG
1849 xfs_dbg_kobj.kobject.kset = xfs_kset;
1850 error = xfs_sysfs_init(&xfs_dbg_kobj, &xfs_dbg_ktype, NULL, "debug");
1851 if (error)
1852 goto out_kset_unregister;
1853 #endif
1855 error = xfs_qm_init();
1856 if (error)
1857 goto out_remove_kobj;
1859 error = register_filesystem(&xfs_fs_type);
1860 if (error)
1861 goto out_qm_exit;
1862 return 0;
1864 out_qm_exit:
1865 xfs_qm_exit();
1866 out_remove_kobj:
1867 #ifdef DEBUG
1868 xfs_sysfs_del(&xfs_dbg_kobj);
1869 out_kset_unregister:
1870 #endif
1871 kset_unregister(xfs_kset);
1872 out_sysctl_unregister:
1873 xfs_sysctl_unregister();
1874 out_cleanup_procfs:
1875 xfs_cleanup_procfs();
1876 out_buf_terminate:
1877 xfs_buf_terminate();
1878 out_mru_cache_uninit:
1879 xfs_mru_cache_uninit();
1880 out_destroy_wq:
1881 xfs_destroy_workqueues();
1882 out_destroy_zones:
1883 xfs_destroy_zones();
1884 out:
1885 return error;
1888 STATIC void __exit
1889 exit_xfs_fs(void)
1891 xfs_qm_exit();
1892 unregister_filesystem(&xfs_fs_type);
1893 #ifdef DEBUG
1894 xfs_sysfs_del(&xfs_dbg_kobj);
1895 #endif
1896 kset_unregister(xfs_kset);
1897 xfs_sysctl_unregister();
1898 xfs_cleanup_procfs();
1899 xfs_buf_terminate();
1900 xfs_mru_cache_uninit();
1901 xfs_destroy_workqueues();
1902 xfs_destroy_zones();
1905 module_init(init_xfs_fs);
1906 module_exit(exit_xfs_fs);
1908 MODULE_AUTHOR("Silicon Graphics, Inc.");
1909 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1910 MODULE_LICENSE("GPL");