2 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
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
20 #include "xfs_format.h"
21 #include "xfs_shared.h"
24 #include "xfs_trans.h"
27 #include "xfs_alloc.h"
28 #include "xfs_quota.h"
29 #include "xfs_mount.h"
30 #include "xfs_da_format.h"
31 #include "xfs_bmap_btree.h"
32 #include "xfs_alloc_btree.h"
33 #include "xfs_ialloc_btree.h"
34 #include "xfs_dinode.h"
35 #include "xfs_inode.h"
36 #include "xfs_btree.h"
37 #include "xfs_ialloc.h"
39 #include "xfs_rtalloc.h"
40 #include "xfs_error.h"
41 #include "xfs_itable.h"
42 #include "xfs_fsops.h"
44 #include "xfs_buf_item.h"
45 #include "xfs_log_priv.h"
46 #include "xfs_trans_priv.h"
47 #include "xfs_filestream.h"
48 #include "xfs_da_btree.h"
50 #include "xfs_extfree_item.h"
51 #include "xfs_mru_cache.h"
52 #include "xfs_inode_item.h"
53 #include "xfs_icache.h"
54 #include "xfs_trace.h"
55 #include "xfs_icreate_item.h"
57 #include <linux/namei.h>
58 #include <linux/init.h>
59 #include <linux/slab.h>
60 #include <linux/mount.h>
61 #include <linux/mempool.h>
62 #include <linux/writeback.h>
63 #include <linux/kthread.h>
64 #include <linux/freezer.h>
65 #include <linux/parser.h>
67 static const struct super_operations xfs_super_operations
;
68 static kmem_zone_t
*xfs_ioend_zone
;
69 mempool_t
*xfs_ioend_pool
;
71 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
72 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
73 #define MNTOPT_LOGDEV "logdev" /* log device */
74 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
75 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
76 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
77 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
78 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
79 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
80 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
81 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
82 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
83 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
84 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
85 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
86 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
87 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
88 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
89 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
90 * unwritten extent conversion */
91 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
92 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
93 #define MNTOPT_32BITINODE "inode32" /* inode allocation limited to
94 * XFS_MAXINUMBER_32 */
95 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
96 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
97 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
98 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
100 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
101 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
102 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
103 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
104 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
105 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
106 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
107 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
108 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
109 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
110 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
111 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
112 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
113 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
114 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
115 #define MNTOPT_DELAYLOG "delaylog" /* Delayed logging enabled */
116 #define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed logging disabled */
117 #define MNTOPT_DISCARD "discard" /* Discard unused blocks */
118 #define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */
121 * Table driven mount option parser.
123 * Currently only used for remount, but it will be used for mount
124 * in the future, too.
134 static const match_table_t tokens
= {
135 {Opt_barrier
, "barrier"},
136 {Opt_nobarrier
, "nobarrier"},
137 {Opt_inode64
, "inode64"},
138 {Opt_inode32
, "inode32"},
144 suffix_kstrtoint(char *s
, unsigned int base
, int *res
)
146 int last
, shift_left_factor
= 0, _res
;
149 last
= strlen(value
) - 1;
150 if (value
[last
] == 'K' || value
[last
] == 'k') {
151 shift_left_factor
= 10;
154 if (value
[last
] == 'M' || value
[last
] == 'm') {
155 shift_left_factor
= 20;
158 if (value
[last
] == 'G' || value
[last
] == 'g') {
159 shift_left_factor
= 30;
163 if (kstrtoint(s
, base
, &_res
))
165 *res
= _res
<< shift_left_factor
;
170 * This function fills in xfs_mount_t fields based on mount args.
171 * Note: the superblock has _not_ yet been read in.
173 * Note that this function leaks the various device name allocations on
174 * failure. The caller takes care of them.
178 struct xfs_mount
*mp
,
181 struct super_block
*sb
= mp
->m_super
;
182 char *this_char
, *value
;
186 __uint8_t iosizelog
= 0;
189 * set up the mount name first so all the errors will refer to the
192 mp
->m_fsname
= kstrndup(sb
->s_id
, MAXNAMELEN
, GFP_KERNEL
);
195 mp
->m_fsname_len
= strlen(mp
->m_fsname
) + 1;
198 * Copy binary VFS mount flags we are interested in.
200 if (sb
->s_flags
& MS_RDONLY
)
201 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
202 if (sb
->s_flags
& MS_DIRSYNC
)
203 mp
->m_flags
|= XFS_MOUNT_DIRSYNC
;
204 if (sb
->s_flags
& MS_SYNCHRONOUS
)
205 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
208 * Set some default flags that could be cleared by the mount option
211 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
212 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
214 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
218 * These can be overridden by the mount option parsing.
226 while ((this_char
= strsep(&options
, ",")) != NULL
) {
229 if ((value
= strchr(this_char
, '=')) != NULL
)
232 if (!strcmp(this_char
, MNTOPT_LOGBUFS
)) {
233 if (!value
|| !*value
) {
234 xfs_warn(mp
, "%s option requires an argument",
238 if (kstrtoint(value
, 10, &mp
->m_logbufs
))
240 } else if (!strcmp(this_char
, MNTOPT_LOGBSIZE
)) {
241 if (!value
|| !*value
) {
242 xfs_warn(mp
, "%s option requires an argument",
246 if (suffix_kstrtoint(value
, 10, &mp
->m_logbsize
))
248 } else if (!strcmp(this_char
, MNTOPT_LOGDEV
)) {
249 if (!value
|| !*value
) {
250 xfs_warn(mp
, "%s option requires an argument",
254 mp
->m_logname
= kstrndup(value
, MAXNAMELEN
, GFP_KERNEL
);
257 } else if (!strcmp(this_char
, MNTOPT_MTPT
)) {
258 xfs_warn(mp
, "%s option not allowed on this system",
261 } else if (!strcmp(this_char
, MNTOPT_RTDEV
)) {
262 if (!value
|| !*value
) {
263 xfs_warn(mp
, "%s option requires an argument",
267 mp
->m_rtname
= kstrndup(value
, MAXNAMELEN
, GFP_KERNEL
);
270 } else if (!strcmp(this_char
, MNTOPT_BIOSIZE
)) {
271 if (!value
|| !*value
) {
272 xfs_warn(mp
, "%s option requires an argument",
276 if (kstrtoint(value
, 10, &iosize
))
278 iosizelog
= ffs(iosize
) - 1;
279 } else if (!strcmp(this_char
, MNTOPT_ALLOCSIZE
)) {
280 if (!value
|| !*value
) {
281 xfs_warn(mp
, "%s option requires an argument",
285 if (suffix_kstrtoint(value
, 10, &iosize
))
287 iosizelog
= ffs(iosize
) - 1;
288 } else if (!strcmp(this_char
, MNTOPT_GRPID
) ||
289 !strcmp(this_char
, MNTOPT_BSDGROUPS
)) {
290 mp
->m_flags
|= XFS_MOUNT_GRPID
;
291 } else if (!strcmp(this_char
, MNTOPT_NOGRPID
) ||
292 !strcmp(this_char
, MNTOPT_SYSVGROUPS
)) {
293 mp
->m_flags
&= ~XFS_MOUNT_GRPID
;
294 } else if (!strcmp(this_char
, MNTOPT_WSYNC
)) {
295 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
296 } else if (!strcmp(this_char
, MNTOPT_NORECOVERY
)) {
297 mp
->m_flags
|= XFS_MOUNT_NORECOVERY
;
298 } else if (!strcmp(this_char
, MNTOPT_NOALIGN
)) {
299 mp
->m_flags
|= XFS_MOUNT_NOALIGN
;
300 } else if (!strcmp(this_char
, MNTOPT_SWALLOC
)) {
301 mp
->m_flags
|= XFS_MOUNT_SWALLOC
;
302 } else if (!strcmp(this_char
, MNTOPT_SUNIT
)) {
303 if (!value
|| !*value
) {
304 xfs_warn(mp
, "%s option requires an argument",
308 if (kstrtoint(value
, 10, &dsunit
))
310 } else if (!strcmp(this_char
, MNTOPT_SWIDTH
)) {
311 if (!value
|| !*value
) {
312 xfs_warn(mp
, "%s option requires an argument",
316 if (kstrtoint(value
, 10, &dswidth
))
318 } else if (!strcmp(this_char
, MNTOPT_32BITINODE
)) {
319 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
320 } else if (!strcmp(this_char
, MNTOPT_64BITINODE
)) {
321 mp
->m_flags
&= ~XFS_MOUNT_SMALL_INUMS
;
323 xfs_warn(mp
, "%s option not allowed on this system",
327 } else if (!strcmp(this_char
, MNTOPT_NOUUID
)) {
328 mp
->m_flags
|= XFS_MOUNT_NOUUID
;
329 } else if (!strcmp(this_char
, MNTOPT_BARRIER
)) {
330 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
331 } else if (!strcmp(this_char
, MNTOPT_NOBARRIER
)) {
332 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
333 } else if (!strcmp(this_char
, MNTOPT_IKEEP
)) {
334 mp
->m_flags
|= XFS_MOUNT_IKEEP
;
335 } else if (!strcmp(this_char
, MNTOPT_NOIKEEP
)) {
336 mp
->m_flags
&= ~XFS_MOUNT_IKEEP
;
337 } else if (!strcmp(this_char
, MNTOPT_LARGEIO
)) {
338 mp
->m_flags
&= ~XFS_MOUNT_COMPAT_IOSIZE
;
339 } else if (!strcmp(this_char
, MNTOPT_NOLARGEIO
)) {
340 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
341 } else if (!strcmp(this_char
, MNTOPT_ATTR2
)) {
342 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
343 } else if (!strcmp(this_char
, MNTOPT_NOATTR2
)) {
344 mp
->m_flags
&= ~XFS_MOUNT_ATTR2
;
345 mp
->m_flags
|= XFS_MOUNT_NOATTR2
;
346 } else if (!strcmp(this_char
, MNTOPT_FILESTREAM
)) {
347 mp
->m_flags
|= XFS_MOUNT_FILESTREAMS
;
348 } else if (!strcmp(this_char
, MNTOPT_NOQUOTA
)) {
349 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACCT
;
350 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ENFD
;
351 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACTIVE
;
352 } else if (!strcmp(this_char
, MNTOPT_QUOTA
) ||
353 !strcmp(this_char
, MNTOPT_UQUOTA
) ||
354 !strcmp(this_char
, MNTOPT_USRQUOTA
)) {
355 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
|
357 } else if (!strcmp(this_char
, MNTOPT_QUOTANOENF
) ||
358 !strcmp(this_char
, MNTOPT_UQUOTANOENF
)) {
359 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
);
360 mp
->m_qflags
&= ~XFS_UQUOTA_ENFD
;
361 } else if (!strcmp(this_char
, MNTOPT_PQUOTA
) ||
362 !strcmp(this_char
, MNTOPT_PRJQUOTA
)) {
363 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
|
365 } else if (!strcmp(this_char
, MNTOPT_PQUOTANOENF
)) {
366 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
);
367 mp
->m_qflags
&= ~XFS_PQUOTA_ENFD
;
368 } else if (!strcmp(this_char
, MNTOPT_GQUOTA
) ||
369 !strcmp(this_char
, MNTOPT_GRPQUOTA
)) {
370 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
|
372 } else if (!strcmp(this_char
, MNTOPT_GQUOTANOENF
)) {
373 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
);
374 mp
->m_qflags
&= ~XFS_GQUOTA_ENFD
;
375 } else if (!strcmp(this_char
, MNTOPT_DELAYLOG
)) {
377 "delaylog is the default now, option is deprecated.");
378 } else if (!strcmp(this_char
, MNTOPT_NODELAYLOG
)) {
380 "nodelaylog support has been removed, option is deprecated.");
381 } else if (!strcmp(this_char
, MNTOPT_DISCARD
)) {
382 mp
->m_flags
|= XFS_MOUNT_DISCARD
;
383 } else if (!strcmp(this_char
, MNTOPT_NODISCARD
)) {
384 mp
->m_flags
&= ~XFS_MOUNT_DISCARD
;
385 } else if (!strcmp(this_char
, "ihashsize")) {
387 "ihashsize no longer used, option is deprecated.");
388 } else if (!strcmp(this_char
, "osyncisdsync")) {
390 "osyncisdsync has no effect, option is deprecated.");
391 } else if (!strcmp(this_char
, "osyncisosync")) {
393 "osyncisosync has no effect, option is deprecated.");
394 } else if (!strcmp(this_char
, "irixsgid")) {
396 "irixsgid is now a sysctl(2) variable, option is deprecated.");
398 xfs_warn(mp
, "unknown mount option [%s].", this_char
);
404 * no recovery flag requires a read-only mount
406 if ((mp
->m_flags
& XFS_MOUNT_NORECOVERY
) &&
407 !(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
408 xfs_warn(mp
, "no-recovery mounts must be read-only.");
412 if ((mp
->m_flags
& XFS_MOUNT_NOALIGN
) && (dsunit
|| dswidth
)) {
414 "sunit and swidth options incompatible with the noalign option");
418 #ifndef CONFIG_XFS_QUOTA
419 if (XFS_IS_QUOTA_RUNNING(mp
)) {
420 xfs_warn(mp
, "quota support not available in this kernel.");
425 if ((dsunit
&& !dswidth
) || (!dsunit
&& dswidth
)) {
426 xfs_warn(mp
, "sunit and swidth must be specified together");
430 if (dsunit
&& (dswidth
% dsunit
!= 0)) {
432 "stripe width (%d) must be a multiple of the stripe unit (%d)",
438 if (dsunit
&& !(mp
->m_flags
& XFS_MOUNT_NOALIGN
)) {
440 * At this point the superblock has not been read
441 * in, therefore we do not know the block size.
442 * Before the mount call ends we will convert
445 mp
->m_dalign
= dsunit
;
446 mp
->m_swidth
= dswidth
;
449 if (mp
->m_logbufs
!= -1 &&
450 mp
->m_logbufs
!= 0 &&
451 (mp
->m_logbufs
< XLOG_MIN_ICLOGS
||
452 mp
->m_logbufs
> XLOG_MAX_ICLOGS
)) {
453 xfs_warn(mp
, "invalid logbufs value: %d [not %d-%d]",
454 mp
->m_logbufs
, XLOG_MIN_ICLOGS
, XLOG_MAX_ICLOGS
);
455 return XFS_ERROR(EINVAL
);
457 if (mp
->m_logbsize
!= -1 &&
458 mp
->m_logbsize
!= 0 &&
459 (mp
->m_logbsize
< XLOG_MIN_RECORD_BSIZE
||
460 mp
->m_logbsize
> XLOG_MAX_RECORD_BSIZE
||
461 !is_power_of_2(mp
->m_logbsize
))) {
463 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
465 return XFS_ERROR(EINVAL
);
469 if (iosizelog
> XFS_MAX_IO_LOG
||
470 iosizelog
< XFS_MIN_IO_LOG
) {
471 xfs_warn(mp
, "invalid log iosize: %d [not %d-%d]",
472 iosizelog
, XFS_MIN_IO_LOG
,
474 return XFS_ERROR(EINVAL
);
477 mp
->m_flags
|= XFS_MOUNT_DFLT_IOSIZE
;
478 mp
->m_readio_log
= iosizelog
;
479 mp
->m_writeio_log
= iosizelog
;
485 struct proc_xfs_info
{
492 struct xfs_mount
*mp
,
495 static struct proc_xfs_info xfs_info_set
[] = {
496 /* the few simple ones we can get from the mount struct */
497 { XFS_MOUNT_IKEEP
, "," MNTOPT_IKEEP
},
498 { XFS_MOUNT_WSYNC
, "," MNTOPT_WSYNC
},
499 { XFS_MOUNT_NOALIGN
, "," MNTOPT_NOALIGN
},
500 { XFS_MOUNT_SWALLOC
, "," MNTOPT_SWALLOC
},
501 { XFS_MOUNT_NOUUID
, "," MNTOPT_NOUUID
},
502 { XFS_MOUNT_NORECOVERY
, "," MNTOPT_NORECOVERY
},
503 { XFS_MOUNT_ATTR2
, "," MNTOPT_ATTR2
},
504 { XFS_MOUNT_FILESTREAMS
, "," MNTOPT_FILESTREAM
},
505 { XFS_MOUNT_GRPID
, "," MNTOPT_GRPID
},
506 { XFS_MOUNT_DISCARD
, "," MNTOPT_DISCARD
},
507 { XFS_MOUNT_SMALL_INUMS
, "," MNTOPT_32BITINODE
},
510 static struct proc_xfs_info xfs_info_unset
[] = {
511 /* the few simple ones we can get from the mount struct */
512 { XFS_MOUNT_COMPAT_IOSIZE
, "," MNTOPT_LARGEIO
},
513 { XFS_MOUNT_BARRIER
, "," MNTOPT_NOBARRIER
},
514 { XFS_MOUNT_SMALL_INUMS
, "," MNTOPT_64BITINODE
},
517 struct proc_xfs_info
*xfs_infop
;
519 for (xfs_infop
= xfs_info_set
; xfs_infop
->flag
; xfs_infop
++) {
520 if (mp
->m_flags
& xfs_infop
->flag
)
521 seq_puts(m
, xfs_infop
->str
);
523 for (xfs_infop
= xfs_info_unset
; xfs_infop
->flag
; xfs_infop
++) {
524 if (!(mp
->m_flags
& xfs_infop
->flag
))
525 seq_puts(m
, xfs_infop
->str
);
528 if (mp
->m_flags
& XFS_MOUNT_DFLT_IOSIZE
)
529 seq_printf(m
, "," MNTOPT_ALLOCSIZE
"=%dk",
530 (int)(1 << mp
->m_writeio_log
) >> 10);
532 if (mp
->m_logbufs
> 0)
533 seq_printf(m
, "," MNTOPT_LOGBUFS
"=%d", mp
->m_logbufs
);
534 if (mp
->m_logbsize
> 0)
535 seq_printf(m
, "," MNTOPT_LOGBSIZE
"=%dk", mp
->m_logbsize
>> 10);
538 seq_printf(m
, "," MNTOPT_LOGDEV
"=%s", mp
->m_logname
);
540 seq_printf(m
, "," MNTOPT_RTDEV
"=%s", mp
->m_rtname
);
542 if (mp
->m_dalign
> 0)
543 seq_printf(m
, "," MNTOPT_SUNIT
"=%d",
544 (int)XFS_FSB_TO_BB(mp
, mp
->m_dalign
));
545 if (mp
->m_swidth
> 0)
546 seq_printf(m
, "," MNTOPT_SWIDTH
"=%d",
547 (int)XFS_FSB_TO_BB(mp
, mp
->m_swidth
));
549 if (mp
->m_qflags
& (XFS_UQUOTA_ACCT
|XFS_UQUOTA_ENFD
))
550 seq_puts(m
, "," MNTOPT_USRQUOTA
);
551 else if (mp
->m_qflags
& XFS_UQUOTA_ACCT
)
552 seq_puts(m
, "," MNTOPT_UQUOTANOENF
);
554 if (mp
->m_qflags
& XFS_PQUOTA_ACCT
) {
555 if (mp
->m_qflags
& XFS_PQUOTA_ENFD
)
556 seq_puts(m
, "," MNTOPT_PRJQUOTA
);
558 seq_puts(m
, "," MNTOPT_PQUOTANOENF
);
560 if (mp
->m_qflags
& XFS_GQUOTA_ACCT
) {
561 if (mp
->m_qflags
& XFS_GQUOTA_ENFD
)
562 seq_puts(m
, "," MNTOPT_GRPQUOTA
);
564 seq_puts(m
, "," MNTOPT_GQUOTANOENF
);
567 if (!(mp
->m_qflags
& XFS_ALL_QUOTA_ACCT
))
568 seq_puts(m
, "," MNTOPT_NOQUOTA
);
574 unsigned int blockshift
)
576 unsigned int pagefactor
= 1;
577 unsigned int bitshift
= BITS_PER_LONG
- 1;
579 /* Figure out maximum filesize, on Linux this can depend on
580 * the filesystem blocksize (on 32 bit platforms).
581 * __block_write_begin does this in an [unsigned] long...
582 * page->index << (PAGE_CACHE_SHIFT - bbits)
583 * So, for page sized blocks (4K on 32 bit platforms),
584 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
585 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
586 * but for smaller blocksizes it is less (bbits = log2 bsize).
587 * Note1: get_block_t takes a long (implicit cast from above)
588 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
589 * can optionally convert the [unsigned] long from above into
590 * an [unsigned] long long.
593 #if BITS_PER_LONG == 32
594 # if defined(CONFIG_LBDAF)
595 ASSERT(sizeof(sector_t
) == 8);
596 pagefactor
= PAGE_CACHE_SIZE
;
597 bitshift
= BITS_PER_LONG
;
599 pagefactor
= PAGE_CACHE_SIZE
>> (PAGE_CACHE_SHIFT
- blockshift
);
603 return (((__uint64_t
)pagefactor
) << bitshift
) - 1;
607 xfs_set_inode32(struct xfs_mount
*mp
)
609 xfs_agnumber_t index
= 0;
610 xfs_agnumber_t maxagi
= 0;
611 xfs_sb_t
*sbp
= &mp
->m_sb
;
612 xfs_agnumber_t max_metadata
;
613 xfs_agino_t agino
= XFS_OFFBNO_TO_AGINO(mp
, sbp
->sb_agblocks
-1, 0);
614 xfs_ino_t ino
= XFS_AGINO_TO_INO(mp
, sbp
->sb_agcount
-1, agino
);
617 /* Calculate how much should be reserved for inodes to meet
618 * the max inode percentage.
620 if (mp
->m_maxicount
) {
623 icount
= sbp
->sb_dblocks
* sbp
->sb_imax_pct
;
625 icount
+= sbp
->sb_agblocks
- 1;
626 do_div(icount
, sbp
->sb_agblocks
);
627 max_metadata
= icount
;
629 max_metadata
= sbp
->sb_agcount
;
632 for (index
= 0; index
< sbp
->sb_agcount
; index
++) {
633 ino
= XFS_AGINO_TO_INO(mp
, index
, agino
);
635 if (ino
> XFS_MAXINUMBER_32
) {
636 pag
= xfs_perag_get(mp
, index
);
637 pag
->pagi_inodeok
= 0;
638 pag
->pagf_metadata
= 0;
643 pag
= xfs_perag_get(mp
, index
);
644 pag
->pagi_inodeok
= 1;
646 if (index
< max_metadata
)
647 pag
->pagf_metadata
= 1;
650 mp
->m_flags
|= (XFS_MOUNT_32BITINODES
|
651 XFS_MOUNT_SMALL_INUMS
);
657 xfs_set_inode64(struct xfs_mount
*mp
)
659 xfs_agnumber_t index
= 0;
661 for (index
= 0; index
< mp
->m_sb
.sb_agcount
; index
++) {
662 struct xfs_perag
*pag
;
664 pag
= xfs_perag_get(mp
, index
);
665 pag
->pagi_inodeok
= 1;
666 pag
->pagf_metadata
= 0;
670 /* There is no need for lock protection on m_flags,
671 * the rw_semaphore of the VFS superblock is locked
672 * during mount/umount/remount operations, so this is
673 * enough to avoid concurency on the m_flags field
675 mp
->m_flags
&= ~(XFS_MOUNT_32BITINODES
|
676 XFS_MOUNT_SMALL_INUMS
);
684 struct block_device
**bdevp
)
688 *bdevp
= blkdev_get_by_path(name
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
,
690 if (IS_ERR(*bdevp
)) {
691 error
= PTR_ERR(*bdevp
);
692 xfs_warn(mp
, "Invalid device [%s], error=%d\n", name
, error
);
700 struct block_device
*bdev
)
703 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
707 xfs_blkdev_issue_flush(
708 xfs_buftarg_t
*buftarg
)
710 blkdev_issue_flush(buftarg
->bt_bdev
, GFP_NOFS
, NULL
);
715 struct xfs_mount
*mp
)
717 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
718 struct block_device
*logdev
= mp
->m_logdev_targp
->bt_bdev
;
719 xfs_free_buftarg(mp
, mp
->m_logdev_targp
);
720 xfs_blkdev_put(logdev
);
722 if (mp
->m_rtdev_targp
) {
723 struct block_device
*rtdev
= mp
->m_rtdev_targp
->bt_bdev
;
724 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
725 xfs_blkdev_put(rtdev
);
727 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
731 * The file system configurations are:
732 * (1) device (partition) with data and internal log
733 * (2) logical volume with data and log subvolumes.
734 * (3) logical volume with data, log, and realtime subvolumes.
736 * We only have to handle opening the log and realtime volumes here if
737 * they are present. The data subvolume has already been opened by
738 * get_sb_bdev() and is stored in sb->s_bdev.
742 struct xfs_mount
*mp
)
744 struct block_device
*ddev
= mp
->m_super
->s_bdev
;
745 struct block_device
*logdev
= NULL
, *rtdev
= NULL
;
749 * Open real time and log devices - order is important.
752 error
= xfs_blkdev_get(mp
, mp
->m_logname
, &logdev
);
758 error
= xfs_blkdev_get(mp
, mp
->m_rtname
, &rtdev
);
760 goto out_close_logdev
;
762 if (rtdev
== ddev
|| rtdev
== logdev
) {
764 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
766 goto out_close_rtdev
;
771 * Setup xfs_mount buffer target pointers
774 mp
->m_ddev_targp
= xfs_alloc_buftarg(mp
, ddev
, 0, mp
->m_fsname
);
775 if (!mp
->m_ddev_targp
)
776 goto out_close_rtdev
;
779 mp
->m_rtdev_targp
= xfs_alloc_buftarg(mp
, rtdev
, 1,
781 if (!mp
->m_rtdev_targp
)
782 goto out_free_ddev_targ
;
785 if (logdev
&& logdev
!= ddev
) {
786 mp
->m_logdev_targp
= xfs_alloc_buftarg(mp
, logdev
, 1,
788 if (!mp
->m_logdev_targp
)
789 goto out_free_rtdev_targ
;
791 mp
->m_logdev_targp
= mp
->m_ddev_targp
;
797 if (mp
->m_rtdev_targp
)
798 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
800 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
803 xfs_blkdev_put(rtdev
);
805 if (logdev
&& logdev
!= ddev
)
806 xfs_blkdev_put(logdev
);
812 * Setup xfs_mount buffer target pointers based on superblock
816 struct xfs_mount
*mp
)
820 error
= xfs_setsize_buftarg(mp
->m_ddev_targp
, mp
->m_sb
.sb_blocksize
,
821 mp
->m_sb
.sb_sectsize
);
825 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
826 unsigned int log_sector_size
= BBSIZE
;
828 if (xfs_sb_version_hassector(&mp
->m_sb
))
829 log_sector_size
= mp
->m_sb
.sb_logsectsize
;
830 error
= xfs_setsize_buftarg(mp
->m_logdev_targp
,
831 mp
->m_sb
.sb_blocksize
,
836 if (mp
->m_rtdev_targp
) {
837 error
= xfs_setsize_buftarg(mp
->m_rtdev_targp
,
838 mp
->m_sb
.sb_blocksize
,
839 mp
->m_sb
.sb_sectsize
);
848 xfs_init_mount_workqueues(
849 struct xfs_mount
*mp
)
851 mp
->m_data_workqueue
= alloc_workqueue("xfs-data/%s",
852 WQ_MEM_RECLAIM
, 0, mp
->m_fsname
);
853 if (!mp
->m_data_workqueue
)
856 mp
->m_unwritten_workqueue
= alloc_workqueue("xfs-conv/%s",
857 WQ_MEM_RECLAIM
, 0, mp
->m_fsname
);
858 if (!mp
->m_unwritten_workqueue
)
859 goto out_destroy_data_iodone_queue
;
861 mp
->m_cil_workqueue
= alloc_workqueue("xfs-cil/%s",
862 WQ_MEM_RECLAIM
, 0, mp
->m_fsname
);
863 if (!mp
->m_cil_workqueue
)
864 goto out_destroy_unwritten
;
866 mp
->m_reclaim_workqueue
= alloc_workqueue("xfs-reclaim/%s",
868 if (!mp
->m_reclaim_workqueue
)
869 goto out_destroy_cil
;
871 mp
->m_log_workqueue
= alloc_workqueue("xfs-log/%s",
873 if (!mp
->m_log_workqueue
)
874 goto out_destroy_reclaim
;
876 mp
->m_eofblocks_workqueue
= alloc_workqueue("xfs-eofblocks/%s",
878 if (!mp
->m_eofblocks_workqueue
)
879 goto out_destroy_log
;
884 destroy_workqueue(mp
->m_log_workqueue
);
886 destroy_workqueue(mp
->m_reclaim_workqueue
);
888 destroy_workqueue(mp
->m_cil_workqueue
);
889 out_destroy_unwritten
:
890 destroy_workqueue(mp
->m_unwritten_workqueue
);
891 out_destroy_data_iodone_queue
:
892 destroy_workqueue(mp
->m_data_workqueue
);
898 xfs_destroy_mount_workqueues(
899 struct xfs_mount
*mp
)
901 destroy_workqueue(mp
->m_eofblocks_workqueue
);
902 destroy_workqueue(mp
->m_log_workqueue
);
903 destroy_workqueue(mp
->m_reclaim_workqueue
);
904 destroy_workqueue(mp
->m_cil_workqueue
);
905 destroy_workqueue(mp
->m_data_workqueue
);
906 destroy_workqueue(mp
->m_unwritten_workqueue
);
910 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
911 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
912 * for IO to complete so that we effectively throttle multiple callers to the
913 * rate at which IO is completing.
917 struct xfs_mount
*mp
)
919 struct super_block
*sb
= mp
->m_super
;
921 if (down_read_trylock(&sb
->s_umount
)) {
923 up_read(&sb
->s_umount
);
927 /* Catch misguided souls that try to use this interface on XFS */
928 STATIC
struct inode
*
930 struct super_block
*sb
)
937 * Now that the generic code is guaranteed not to be accessing
938 * the linux inode, we can reclaim the inode.
941 xfs_fs_destroy_inode(
944 struct xfs_inode
*ip
= XFS_I(inode
);
946 trace_xfs_destroy_inode(ip
);
948 XFS_STATS_INC(vn_reclaim
);
950 ASSERT(XFS_FORCED_SHUTDOWN(ip
->i_mount
) || ip
->i_delayed_blks
== 0);
953 * We should never get here with one of the reclaim flags already set.
955 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIMABLE
));
956 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIM
));
959 * We always use background reclaim here because even if the
960 * inode is clean, it still may be under IO and hence we have
961 * to take the flush lock. The background reclaim path handles
962 * this more efficiently than we can here, so simply let background
963 * reclaim tear down all inodes.
965 xfs_inode_set_reclaim_tag(ip
);
969 * Slab object creation initialisation for the XFS inode.
970 * This covers only the idempotent fields in the XFS inode;
971 * all other fields need to be initialised on allocation
972 * from the slab. This avoids the need to repeatedly initialise
973 * fields in the xfs inode that left in the initialise state
974 * when freeing the inode.
977 xfs_fs_inode_init_once(
980 struct xfs_inode
*ip
= inode
;
982 memset(ip
, 0, sizeof(struct xfs_inode
));
985 inode_init_once(VFS_I(ip
));
988 atomic_set(&ip
->i_pincount
, 0);
989 spin_lock_init(&ip
->i_flags_lock
);
991 mrlock_init(&ip
->i_lock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
992 "xfsino", ip
->i_ino
);
999 xfs_inode_t
*ip
= XFS_I(inode
);
1001 ASSERT(!rwsem_is_locked(&ip
->i_iolock
.mr_lock
));
1003 trace_xfs_evict_inode(ip
);
1005 truncate_inode_pages(&inode
->i_data
, 0);
1007 XFS_STATS_INC(vn_rele
);
1008 XFS_STATS_INC(vn_remove
);
1009 XFS_STATS_DEC(vn_active
);
1015 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1016 * serialised against cache hits here via the inode->i_lock and igrab() in
1017 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1018 * racing with us, and it avoids needing to grab a spinlock here for every inode
1019 * we drop the final reference on.
1023 struct inode
*inode
)
1025 struct xfs_inode
*ip
= XFS_I(inode
);
1027 return generic_drop_inode(inode
) || (ip
->i_flags
& XFS_IDONTCACHE
);
1032 struct xfs_mount
*mp
)
1034 kfree(mp
->m_fsname
);
1035 kfree(mp
->m_rtname
);
1036 kfree(mp
->m_logname
);
1041 struct super_block
*sb
)
1043 struct xfs_mount
*mp
= XFS_M(sb
);
1045 xfs_filestream_unmount(mp
);
1049 xfs_icsb_destroy_counters(mp
);
1050 xfs_destroy_mount_workqueues(mp
);
1051 xfs_close_devices(mp
);
1052 xfs_free_fsname(mp
);
1058 struct super_block
*sb
,
1061 struct xfs_mount
*mp
= XFS_M(sb
);
1064 * Doing anything during the async pass would be counterproductive.
1069 xfs_log_force(mp
, XFS_LOG_SYNC
);
1072 * The disk must be active because we're syncing.
1073 * We schedule log work now (now that the disk is
1074 * active) instead of later (when it might not be).
1076 flush_delayed_work(&mp
->m_log
->l_work
);
1084 struct dentry
*dentry
,
1085 struct kstatfs
*statp
)
1087 struct xfs_mount
*mp
= XFS_M(dentry
->d_sb
);
1088 xfs_sb_t
*sbp
= &mp
->m_sb
;
1089 struct xfs_inode
*ip
= XFS_I(dentry
->d_inode
);
1090 __uint64_t fakeinos
, id
;
1094 statp
->f_type
= XFS_SB_MAGIC
;
1095 statp
->f_namelen
= MAXNAMELEN
- 1;
1097 id
= huge_encode_dev(mp
->m_ddev_targp
->bt_dev
);
1098 statp
->f_fsid
.val
[0] = (u32
)id
;
1099 statp
->f_fsid
.val
[1] = (u32
)(id
>> 32);
1101 xfs_icsb_sync_counters(mp
, XFS_ICSB_LAZY_COUNT
);
1103 spin_lock(&mp
->m_sb_lock
);
1104 statp
->f_bsize
= sbp
->sb_blocksize
;
1105 lsize
= sbp
->sb_logstart
? sbp
->sb_logblocks
: 0;
1106 statp
->f_blocks
= sbp
->sb_dblocks
- lsize
;
1107 statp
->f_bfree
= statp
->f_bavail
=
1108 sbp
->sb_fdblocks
- XFS_ALLOC_SET_ASIDE(mp
);
1109 fakeinos
= statp
->f_bfree
<< sbp
->sb_inopblog
;
1111 MIN(sbp
->sb_icount
+ fakeinos
, (__uint64_t
)XFS_MAXINUMBER
);
1112 if (mp
->m_maxicount
)
1113 statp
->f_files
= min_t(typeof(statp
->f_files
),
1117 /* make sure statp->f_ffree does not underflow */
1118 ffree
= statp
->f_files
- (sbp
->sb_icount
- sbp
->sb_ifree
);
1119 statp
->f_ffree
= max_t(__int64_t
, ffree
, 0);
1121 spin_unlock(&mp
->m_sb_lock
);
1123 if ((ip
->i_d
.di_flags
& XFS_DIFLAG_PROJINHERIT
) &&
1124 ((mp
->m_qflags
& (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))) ==
1125 (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))
1126 xfs_qm_statvfs(ip
, statp
);
1131 xfs_save_resvblks(struct xfs_mount
*mp
)
1133 __uint64_t resblks
= 0;
1135 mp
->m_resblks_save
= mp
->m_resblks
;
1136 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1140 xfs_restore_resvblks(struct xfs_mount
*mp
)
1144 if (mp
->m_resblks_save
) {
1145 resblks
= mp
->m_resblks_save
;
1146 mp
->m_resblks_save
= 0;
1148 resblks
= xfs_default_resblks(mp
);
1150 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1154 * Trigger writeback of all the dirty metadata in the file system.
1156 * This ensures that the metadata is written to their location on disk rather
1157 * than just existing in transactions in the log. This means after a quiesce
1158 * there is no log replay required to write the inodes to disk - this is the
1159 * primary difference between a sync and a quiesce.
1161 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1162 * it is started again when appropriate.
1166 struct xfs_mount
*mp
)
1170 /* wait for all modifications to complete */
1171 while (atomic_read(&mp
->m_active_trans
) > 0)
1174 /* force the log to unpin objects from the now complete transactions */
1175 xfs_log_force(mp
, XFS_LOG_SYNC
);
1177 /* reclaim inodes to do any IO before the freeze completes */
1178 xfs_reclaim_inodes(mp
, 0);
1179 xfs_reclaim_inodes(mp
, SYNC_WAIT
);
1181 /* Push the superblock and write an unmount record */
1182 error
= xfs_log_sbcount(mp
);
1184 xfs_warn(mp
, "xfs_attr_quiesce: failed to log sb changes. "
1185 "Frozen image may not be consistent.");
1187 * Just warn here till VFS can correctly support
1188 * read-only remount without racing.
1190 WARN_ON(atomic_read(&mp
->m_active_trans
) != 0);
1192 xfs_log_quiesce(mp
);
1197 struct super_block
*sb
,
1201 struct xfs_mount
*mp
= XFS_M(sb
);
1202 substring_t args
[MAX_OPT_ARGS
];
1206 while ((p
= strsep(&options
, ",")) != NULL
) {
1212 token
= match_token(p
, tokens
, args
);
1215 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
1218 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
1221 mp
->m_maxagi
= xfs_set_inode64(mp
);
1224 mp
->m_maxagi
= xfs_set_inode32(mp
);
1228 * Logically we would return an error here to prevent
1229 * users from believing they might have changed
1230 * mount options using remount which can't be changed.
1232 * But unfortunately mount(8) adds all options from
1233 * mtab and fstab to the mount arguments in some cases
1234 * so we can't blindly reject options, but have to
1235 * check for each specified option if it actually
1236 * differs from the currently set option and only
1237 * reject it if that's the case.
1239 * Until that is implemented we return success for
1240 * every remount request, and silently ignore all
1241 * options that we can't actually change.
1245 "mount option \"%s\" not supported for remount", p
);
1254 if ((mp
->m_flags
& XFS_MOUNT_RDONLY
) && !(*flags
& MS_RDONLY
)) {
1255 mp
->m_flags
&= ~XFS_MOUNT_RDONLY
;
1258 * If this is the first remount to writeable state we
1259 * might have some superblock changes to update.
1261 if (mp
->m_update_flags
) {
1262 error
= xfs_mount_log_sb(mp
, mp
->m_update_flags
);
1264 xfs_warn(mp
, "failed to write sb changes");
1267 mp
->m_update_flags
= 0;
1271 * Fill out the reserve pool if it is empty. Use the stashed
1272 * value if it is non-zero, otherwise go with the default.
1274 xfs_restore_resvblks(mp
);
1275 xfs_log_work_queue(mp
);
1279 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
) && (*flags
& MS_RDONLY
)) {
1281 * Before we sync the metadata, we need to free up the reserve
1282 * block pool so that the used block count in the superblock on
1283 * disk is correct at the end of the remount. Stash the current
1284 * reserve pool size so that if we get remounted rw, we can
1285 * return it to the same size.
1287 xfs_save_resvblks(mp
);
1288 xfs_quiesce_attr(mp
);
1289 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
1296 * Second stage of a freeze. The data is already frozen so we only
1297 * need to take care of the metadata. Once that's done write a dummy
1298 * record to dirty the log in case of a crash while frozen.
1302 struct super_block
*sb
)
1304 struct xfs_mount
*mp
= XFS_M(sb
);
1306 xfs_save_resvblks(mp
);
1307 xfs_quiesce_attr(mp
);
1308 return -xfs_fs_log_dummy(mp
);
1313 struct super_block
*sb
)
1315 struct xfs_mount
*mp
= XFS_M(sb
);
1317 xfs_restore_resvblks(mp
);
1318 xfs_log_work_queue(mp
);
1323 xfs_fs_show_options(
1325 struct dentry
*root
)
1327 return -xfs_showargs(XFS_M(root
->d_sb
), m
);
1331 * This function fills in xfs_mount_t fields based on mount args.
1332 * Note: the superblock _has_ now been read in.
1336 struct xfs_mount
*mp
)
1338 int ronly
= (mp
->m_flags
& XFS_MOUNT_RDONLY
);
1340 /* Fail a mount where the logbuf is smaller than the log stripe */
1341 if (xfs_sb_version_haslogv2(&mp
->m_sb
)) {
1342 if (mp
->m_logbsize
<= 0 &&
1343 mp
->m_sb
.sb_logsunit
> XLOG_BIG_RECORD_BSIZE
) {
1344 mp
->m_logbsize
= mp
->m_sb
.sb_logsunit
;
1345 } else if (mp
->m_logbsize
> 0 &&
1346 mp
->m_logbsize
< mp
->m_sb
.sb_logsunit
) {
1348 "logbuf size must be greater than or equal to log stripe size");
1349 return XFS_ERROR(EINVAL
);
1352 /* Fail a mount if the logbuf is larger than 32K */
1353 if (mp
->m_logbsize
> XLOG_BIG_RECORD_BSIZE
) {
1355 "logbuf size for version 1 logs must be 16K or 32K");
1356 return XFS_ERROR(EINVAL
);
1361 * V5 filesystems always use attr2 format for attributes.
1363 if (xfs_sb_version_hascrc(&mp
->m_sb
) &&
1364 (mp
->m_flags
& XFS_MOUNT_NOATTR2
)) {
1366 "Cannot mount a V5 filesystem as %s. %s is always enabled for V5 filesystems.",
1367 MNTOPT_NOATTR2
, MNTOPT_ATTR2
);
1368 return XFS_ERROR(EINVAL
);
1372 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1373 * told by noattr2 to turn it off
1375 if (xfs_sb_version_hasattr2(&mp
->m_sb
) &&
1376 !(mp
->m_flags
& XFS_MOUNT_NOATTR2
))
1377 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
1380 * prohibit r/w mounts of read-only filesystems
1382 if ((mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) && !ronly
) {
1384 "cannot mount a read-only filesystem as read-write");
1385 return XFS_ERROR(EROFS
);
1388 if ((mp
->m_qflags
& (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
)) &&
1389 (mp
->m_qflags
& (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
)) &&
1390 !xfs_sb_version_has_pquotino(&mp
->m_sb
)) {
1392 "Super block does not support project and group quota together");
1393 return XFS_ERROR(EINVAL
);
1401 struct super_block
*sb
,
1406 struct xfs_mount
*mp
= NULL
;
1407 int flags
= 0, error
= ENOMEM
;
1409 mp
= kzalloc(sizeof(struct xfs_mount
), GFP_KERNEL
);
1413 spin_lock_init(&mp
->m_sb_lock
);
1414 mutex_init(&mp
->m_growlock
);
1415 atomic_set(&mp
->m_active_trans
, 0);
1416 INIT_DELAYED_WORK(&mp
->m_reclaim_work
, xfs_reclaim_worker
);
1417 INIT_DELAYED_WORK(&mp
->m_eofblocks_work
, xfs_eofblocks_worker
);
1422 error
= xfs_parseargs(mp
, (char *)data
);
1424 goto out_free_fsname
;
1426 sb_min_blocksize(sb
, BBSIZE
);
1427 sb
->s_xattr
= xfs_xattr_handlers
;
1428 sb
->s_export_op
= &xfs_export_operations
;
1429 #ifdef CONFIG_XFS_QUOTA
1430 sb
->s_qcop
= &xfs_quotactl_operations
;
1432 sb
->s_op
= &xfs_super_operations
;
1435 flags
|= XFS_MFSI_QUIET
;
1437 error
= xfs_open_devices(mp
);
1439 goto out_free_fsname
;
1441 error
= xfs_init_mount_workqueues(mp
);
1443 goto out_close_devices
;
1445 error
= xfs_icsb_init_counters(mp
);
1447 goto out_destroy_workqueues
;
1449 error
= xfs_readsb(mp
, flags
);
1451 goto out_destroy_counters
;
1453 error
= xfs_finish_flags(mp
);
1457 error
= xfs_setup_devices(mp
);
1461 error
= xfs_filestream_mount(mp
);
1466 * we must configure the block size in the superblock before we run the
1467 * full mount process as the mount process can lookup and cache inodes.
1469 sb
->s_magic
= XFS_SB_MAGIC
;
1470 sb
->s_blocksize
= mp
->m_sb
.sb_blocksize
;
1471 sb
->s_blocksize_bits
= ffs(sb
->s_blocksize
) - 1;
1472 sb
->s_maxbytes
= xfs_max_file_offset(sb
->s_blocksize_bits
);
1473 sb
->s_max_links
= XFS_MAXLINK
;
1474 sb
->s_time_gran
= 1;
1475 set_posix_acl_flag(sb
);
1477 /* version 5 superblocks support inode version counters. */
1478 if (XFS_SB_VERSION_NUM(&mp
->m_sb
) == XFS_SB_VERSION_5
)
1479 sb
->s_flags
|= MS_I_VERSION
;
1481 error
= xfs_mountfs(mp
);
1483 goto out_filestream_unmount
;
1485 root
= igrab(VFS_I(mp
->m_rootip
));
1490 sb
->s_root
= d_make_root(root
);
1498 out_filestream_unmount
:
1499 xfs_filestream_unmount(mp
);
1502 out_destroy_counters
:
1503 xfs_icsb_destroy_counters(mp
);
1504 out_destroy_workqueues
:
1505 xfs_destroy_mount_workqueues(mp
);
1507 xfs_close_devices(mp
);
1509 xfs_free_fsname(mp
);
1515 xfs_filestream_unmount(mp
);
1520 STATIC
struct dentry
*
1522 struct file_system_type
*fs_type
,
1524 const char *dev_name
,
1527 return mount_bdev(fs_type
, flags
, dev_name
, data
, xfs_fs_fill_super
);
1531 xfs_fs_nr_cached_objects(
1532 struct super_block
*sb
,
1535 return xfs_reclaim_inodes_count(XFS_M(sb
));
1539 xfs_fs_free_cached_objects(
1540 struct super_block
*sb
,
1544 return xfs_reclaim_inodes_nr(XFS_M(sb
), nr_to_scan
);
1547 static const struct super_operations xfs_super_operations
= {
1548 .alloc_inode
= xfs_fs_alloc_inode
,
1549 .destroy_inode
= xfs_fs_destroy_inode
,
1550 .evict_inode
= xfs_fs_evict_inode
,
1551 .drop_inode
= xfs_fs_drop_inode
,
1552 .put_super
= xfs_fs_put_super
,
1553 .sync_fs
= xfs_fs_sync_fs
,
1554 .freeze_fs
= xfs_fs_freeze
,
1555 .unfreeze_fs
= xfs_fs_unfreeze
,
1556 .statfs
= xfs_fs_statfs
,
1557 .remount_fs
= xfs_fs_remount
,
1558 .show_options
= xfs_fs_show_options
,
1559 .nr_cached_objects
= xfs_fs_nr_cached_objects
,
1560 .free_cached_objects
= xfs_fs_free_cached_objects
,
1563 static struct file_system_type xfs_fs_type
= {
1564 .owner
= THIS_MODULE
,
1566 .mount
= xfs_fs_mount
,
1567 .kill_sb
= kill_block_super
,
1568 .fs_flags
= FS_REQUIRES_DEV
,
1570 MODULE_ALIAS_FS("xfs");
1573 xfs_init_zones(void)
1576 xfs_ioend_zone
= kmem_zone_init(sizeof(xfs_ioend_t
), "xfs_ioend");
1577 if (!xfs_ioend_zone
)
1580 xfs_ioend_pool
= mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE
,
1582 if (!xfs_ioend_pool
)
1583 goto out_destroy_ioend_zone
;
1585 xfs_log_ticket_zone
= kmem_zone_init(sizeof(xlog_ticket_t
),
1587 if (!xfs_log_ticket_zone
)
1588 goto out_destroy_ioend_pool
;
1590 xfs_bmap_free_item_zone
= kmem_zone_init(sizeof(xfs_bmap_free_item_t
),
1591 "xfs_bmap_free_item");
1592 if (!xfs_bmap_free_item_zone
)
1593 goto out_destroy_log_ticket_zone
;
1595 xfs_btree_cur_zone
= kmem_zone_init(sizeof(xfs_btree_cur_t
),
1597 if (!xfs_btree_cur_zone
)
1598 goto out_destroy_bmap_free_item_zone
;
1600 xfs_da_state_zone
= kmem_zone_init(sizeof(xfs_da_state_t
),
1602 if (!xfs_da_state_zone
)
1603 goto out_destroy_btree_cur_zone
;
1605 xfs_ifork_zone
= kmem_zone_init(sizeof(xfs_ifork_t
), "xfs_ifork");
1606 if (!xfs_ifork_zone
)
1607 goto out_destroy_da_state_zone
;
1609 xfs_trans_zone
= kmem_zone_init(sizeof(xfs_trans_t
), "xfs_trans");
1610 if (!xfs_trans_zone
)
1611 goto out_destroy_ifork_zone
;
1613 xfs_log_item_desc_zone
=
1614 kmem_zone_init(sizeof(struct xfs_log_item_desc
),
1615 "xfs_log_item_desc");
1616 if (!xfs_log_item_desc_zone
)
1617 goto out_destroy_trans_zone
;
1620 * The size of the zone allocated buf log item is the maximum
1621 * size possible under XFS. This wastes a little bit of memory,
1622 * but it is much faster.
1624 xfs_buf_item_zone
= kmem_zone_init(sizeof(struct xfs_buf_log_item
),
1626 if (!xfs_buf_item_zone
)
1627 goto out_destroy_log_item_desc_zone
;
1629 xfs_efd_zone
= kmem_zone_init((sizeof(xfs_efd_log_item_t
) +
1630 ((XFS_EFD_MAX_FAST_EXTENTS
- 1) *
1631 sizeof(xfs_extent_t
))), "xfs_efd_item");
1633 goto out_destroy_buf_item_zone
;
1635 xfs_efi_zone
= kmem_zone_init((sizeof(xfs_efi_log_item_t
) +
1636 ((XFS_EFI_MAX_FAST_EXTENTS
- 1) *
1637 sizeof(xfs_extent_t
))), "xfs_efi_item");
1639 goto out_destroy_efd_zone
;
1642 kmem_zone_init_flags(sizeof(xfs_inode_t
), "xfs_inode",
1643 KM_ZONE_HWALIGN
| KM_ZONE_RECLAIM
| KM_ZONE_SPREAD
,
1644 xfs_fs_inode_init_once
);
1645 if (!xfs_inode_zone
)
1646 goto out_destroy_efi_zone
;
1649 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t
), "xfs_ili",
1650 KM_ZONE_SPREAD
, NULL
);
1652 goto out_destroy_inode_zone
;
1653 xfs_icreate_zone
= kmem_zone_init(sizeof(struct xfs_icreate_item
),
1655 if (!xfs_icreate_zone
)
1656 goto out_destroy_ili_zone
;
1660 out_destroy_ili_zone
:
1661 kmem_zone_destroy(xfs_ili_zone
);
1662 out_destroy_inode_zone
:
1663 kmem_zone_destroy(xfs_inode_zone
);
1664 out_destroy_efi_zone
:
1665 kmem_zone_destroy(xfs_efi_zone
);
1666 out_destroy_efd_zone
:
1667 kmem_zone_destroy(xfs_efd_zone
);
1668 out_destroy_buf_item_zone
:
1669 kmem_zone_destroy(xfs_buf_item_zone
);
1670 out_destroy_log_item_desc_zone
:
1671 kmem_zone_destroy(xfs_log_item_desc_zone
);
1672 out_destroy_trans_zone
:
1673 kmem_zone_destroy(xfs_trans_zone
);
1674 out_destroy_ifork_zone
:
1675 kmem_zone_destroy(xfs_ifork_zone
);
1676 out_destroy_da_state_zone
:
1677 kmem_zone_destroy(xfs_da_state_zone
);
1678 out_destroy_btree_cur_zone
:
1679 kmem_zone_destroy(xfs_btree_cur_zone
);
1680 out_destroy_bmap_free_item_zone
:
1681 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1682 out_destroy_log_ticket_zone
:
1683 kmem_zone_destroy(xfs_log_ticket_zone
);
1684 out_destroy_ioend_pool
:
1685 mempool_destroy(xfs_ioend_pool
);
1686 out_destroy_ioend_zone
:
1687 kmem_zone_destroy(xfs_ioend_zone
);
1693 xfs_destroy_zones(void)
1696 * Make sure all delayed rcu free are flushed before we
1700 kmem_zone_destroy(xfs_icreate_zone
);
1701 kmem_zone_destroy(xfs_ili_zone
);
1702 kmem_zone_destroy(xfs_inode_zone
);
1703 kmem_zone_destroy(xfs_efi_zone
);
1704 kmem_zone_destroy(xfs_efd_zone
);
1705 kmem_zone_destroy(xfs_buf_item_zone
);
1706 kmem_zone_destroy(xfs_log_item_desc_zone
);
1707 kmem_zone_destroy(xfs_trans_zone
);
1708 kmem_zone_destroy(xfs_ifork_zone
);
1709 kmem_zone_destroy(xfs_da_state_zone
);
1710 kmem_zone_destroy(xfs_btree_cur_zone
);
1711 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1712 kmem_zone_destroy(xfs_log_ticket_zone
);
1713 mempool_destroy(xfs_ioend_pool
);
1714 kmem_zone_destroy(xfs_ioend_zone
);
1719 xfs_init_workqueues(void)
1722 * The allocation workqueue can be used in memory reclaim situations
1723 * (writepage path), and parallelism is only limited by the number of
1724 * AGs in all the filesystems mounted. Hence use the default large
1725 * max_active value for this workqueue.
1727 xfs_alloc_wq
= alloc_workqueue("xfsalloc", WQ_MEM_RECLAIM
, 0);
1735 xfs_destroy_workqueues(void)
1737 destroy_workqueue(xfs_alloc_wq
);
1745 printk(KERN_INFO XFS_VERSION_STRING
" with "
1746 XFS_BUILD_OPTIONS
" enabled\n");
1750 error
= xfs_init_zones();
1754 error
= xfs_init_workqueues();
1756 goto out_destroy_zones
;
1758 error
= xfs_mru_cache_init();
1760 goto out_destroy_wq
;
1762 error
= xfs_filestream_init();
1764 goto out_mru_cache_uninit
;
1766 error
= xfs_buf_init();
1768 goto out_filestream_uninit
;
1770 error
= xfs_init_procfs();
1772 goto out_buf_terminate
;
1774 error
= xfs_sysctl_register();
1776 goto out_cleanup_procfs
;
1778 error
= xfs_qm_init();
1780 goto out_sysctl_unregister
;
1782 error
= register_filesystem(&xfs_fs_type
);
1789 out_sysctl_unregister
:
1790 xfs_sysctl_unregister();
1792 xfs_cleanup_procfs();
1794 xfs_buf_terminate();
1795 out_filestream_uninit
:
1796 xfs_filestream_uninit();
1797 out_mru_cache_uninit
:
1798 xfs_mru_cache_uninit();
1800 xfs_destroy_workqueues();
1802 xfs_destroy_zones();
1811 unregister_filesystem(&xfs_fs_type
);
1812 xfs_sysctl_unregister();
1813 xfs_cleanup_procfs();
1814 xfs_buf_terminate();
1815 xfs_filestream_uninit();
1816 xfs_mru_cache_uninit();
1817 xfs_destroy_workqueues();
1818 xfs_destroy_zones();
1821 module_init(init_xfs_fs
);
1822 module_exit(exit_xfs_fs
);
1824 MODULE_AUTHOR("Silicon Graphics, Inc.");
1825 MODULE_DESCRIPTION(XFS_VERSION_STRING
" with " XFS_BUILD_OPTIONS
" enabled");
1826 MODULE_LICENSE("GPL");