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
22 #include "xfs_trans.h"
26 #include "xfs_alloc.h"
27 #include "xfs_quota.h"
28 #include "xfs_mount.h"
29 #include "xfs_bmap_btree.h"
30 #include "xfs_alloc_btree.h"
31 #include "xfs_ialloc_btree.h"
32 #include "xfs_dinode.h"
33 #include "xfs_inode.h"
34 #include "xfs_btree.h"
35 #include "xfs_ialloc.h"
37 #include "xfs_rtalloc.h"
38 #include "xfs_error.h"
39 #include "xfs_itable.h"
40 #include "xfs_fsops.h"
42 #include "xfs_buf_item.h"
43 #include "xfs_utils.h"
44 #include "xfs_vnodeops.h"
45 #include "xfs_log_priv.h"
46 #include "xfs_trans_priv.h"
47 #include "xfs_filestream.h"
48 #include "xfs_da_btree.h"
49 #include "xfs_extfree_item.h"
50 #include "xfs_mru_cache.h"
51 #include "xfs_inode_item.h"
52 #include "xfs_icache.h"
53 #include "xfs_trace.h"
54 #include "xfs_icreate_item.h"
56 #include <linux/namei.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
59 #include <linux/mount.h>
60 #include <linux/mempool.h>
61 #include <linux/writeback.h>
62 #include <linux/kthread.h>
63 #include <linux/freezer.h>
64 #include <linux/parser.h>
66 static const struct super_operations xfs_super_operations
;
67 static kmem_zone_t
*xfs_ioend_zone
;
68 mempool_t
*xfs_ioend_pool
;
70 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
71 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
72 #define MNTOPT_LOGDEV "logdev" /* log device */
73 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
74 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
75 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
76 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
77 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
78 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
79 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
80 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
81 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
82 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
83 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
84 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
85 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
86 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
87 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
88 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
89 * unwritten extent conversion */
90 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
91 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
92 #define MNTOPT_32BITINODE "inode32" /* inode allocation limited to
93 * XFS_MAXINUMBER_32 */
94 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
95 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
96 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
97 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
99 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
100 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
101 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
102 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
103 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
104 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
105 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
106 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
107 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
108 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
109 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
110 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
111 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
112 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
113 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
114 #define MNTOPT_DELAYLOG "delaylog" /* Delayed logging enabled */
115 #define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed logging disabled */
116 #define MNTOPT_DISCARD "discard" /* Discard unused blocks */
117 #define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */
120 * Table driven mount option parser.
122 * Currently only used for remount, but it will be used for mount
123 * in the future, too.
133 static const match_table_t tokens
= {
134 {Opt_barrier
, "barrier"},
135 {Opt_nobarrier
, "nobarrier"},
136 {Opt_inode64
, "inode64"},
137 {Opt_inode32
, "inode32"},
143 suffix_kstrtoint(char *s
, unsigned int base
, int *res
)
145 int last
, shift_left_factor
= 0, _res
;
148 last
= strlen(value
) - 1;
149 if (value
[last
] == 'K' || value
[last
] == 'k') {
150 shift_left_factor
= 10;
153 if (value
[last
] == 'M' || value
[last
] == 'm') {
154 shift_left_factor
= 20;
157 if (value
[last
] == 'G' || value
[last
] == 'g') {
158 shift_left_factor
= 30;
162 if (kstrtoint(s
, base
, &_res
))
164 *res
= _res
<< shift_left_factor
;
169 * This function fills in xfs_mount_t fields based on mount args.
170 * Note: the superblock has _not_ yet been read in.
172 * Note that this function leaks the various device name allocations on
173 * failure. The caller takes care of them.
177 struct xfs_mount
*mp
,
180 struct super_block
*sb
= mp
->m_super
;
181 char *this_char
, *value
;
185 __uint8_t iosizelog
= 0;
188 * set up the mount name first so all the errors will refer to the
191 mp
->m_fsname
= kstrndup(sb
->s_id
, MAXNAMELEN
, GFP_KERNEL
);
194 mp
->m_fsname_len
= strlen(mp
->m_fsname
) + 1;
197 * Copy binary VFS mount flags we are interested in.
199 if (sb
->s_flags
& MS_RDONLY
)
200 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
201 if (sb
->s_flags
& MS_DIRSYNC
)
202 mp
->m_flags
|= XFS_MOUNT_DIRSYNC
;
203 if (sb
->s_flags
& MS_SYNCHRONOUS
)
204 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
207 * Set some default flags that could be cleared by the mount option
210 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
211 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
213 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
217 * These can be overridden by the mount option parsing.
225 while ((this_char
= strsep(&options
, ",")) != NULL
) {
228 if ((value
= strchr(this_char
, '=')) != NULL
)
231 if (!strcmp(this_char
, MNTOPT_LOGBUFS
)) {
232 if (!value
|| !*value
) {
233 xfs_warn(mp
, "%s option requires an argument",
237 if (kstrtoint(value
, 10, &mp
->m_logbufs
))
239 } else if (!strcmp(this_char
, MNTOPT_LOGBSIZE
)) {
240 if (!value
|| !*value
) {
241 xfs_warn(mp
, "%s option requires an argument",
245 if (suffix_kstrtoint(value
, 10, &mp
->m_logbsize
))
247 } else if (!strcmp(this_char
, MNTOPT_LOGDEV
)) {
248 if (!value
|| !*value
) {
249 xfs_warn(mp
, "%s option requires an argument",
253 mp
->m_logname
= kstrndup(value
, MAXNAMELEN
, GFP_KERNEL
);
256 } else if (!strcmp(this_char
, MNTOPT_MTPT
)) {
257 xfs_warn(mp
, "%s option not allowed on this system",
260 } else if (!strcmp(this_char
, MNTOPT_RTDEV
)) {
261 if (!value
|| !*value
) {
262 xfs_warn(mp
, "%s option requires an argument",
266 mp
->m_rtname
= kstrndup(value
, MAXNAMELEN
, GFP_KERNEL
);
269 } else if (!strcmp(this_char
, MNTOPT_BIOSIZE
)) {
270 if (!value
|| !*value
) {
271 xfs_warn(mp
, "%s option requires an argument",
275 if (kstrtoint(value
, 10, &iosize
))
277 iosizelog
= ffs(iosize
) - 1;
278 } else if (!strcmp(this_char
, MNTOPT_ALLOCSIZE
)) {
279 if (!value
|| !*value
) {
280 xfs_warn(mp
, "%s option requires an argument",
284 if (suffix_kstrtoint(value
, 10, &iosize
))
286 iosizelog
= ffs(iosize
) - 1;
287 } else if (!strcmp(this_char
, MNTOPT_GRPID
) ||
288 !strcmp(this_char
, MNTOPT_BSDGROUPS
)) {
289 mp
->m_flags
|= XFS_MOUNT_GRPID
;
290 } else if (!strcmp(this_char
, MNTOPT_NOGRPID
) ||
291 !strcmp(this_char
, MNTOPT_SYSVGROUPS
)) {
292 mp
->m_flags
&= ~XFS_MOUNT_GRPID
;
293 } else if (!strcmp(this_char
, MNTOPT_WSYNC
)) {
294 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
295 } else if (!strcmp(this_char
, MNTOPT_NORECOVERY
)) {
296 mp
->m_flags
|= XFS_MOUNT_NORECOVERY
;
297 } else if (!strcmp(this_char
, MNTOPT_NOALIGN
)) {
298 mp
->m_flags
|= XFS_MOUNT_NOALIGN
;
299 } else if (!strcmp(this_char
, MNTOPT_SWALLOC
)) {
300 mp
->m_flags
|= XFS_MOUNT_SWALLOC
;
301 } else if (!strcmp(this_char
, MNTOPT_SUNIT
)) {
302 if (!value
|| !*value
) {
303 xfs_warn(mp
, "%s option requires an argument",
307 if (kstrtoint(value
, 10, &dsunit
))
309 } else if (!strcmp(this_char
, MNTOPT_SWIDTH
)) {
310 if (!value
|| !*value
) {
311 xfs_warn(mp
, "%s option requires an argument",
315 if (kstrtoint(value
, 10, &dswidth
))
317 } else if (!strcmp(this_char
, MNTOPT_32BITINODE
)) {
318 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
319 } else if (!strcmp(this_char
, MNTOPT_64BITINODE
)) {
320 mp
->m_flags
&= ~XFS_MOUNT_SMALL_INUMS
;
322 xfs_warn(mp
, "%s option not allowed on this system",
326 } else if (!strcmp(this_char
, MNTOPT_NOUUID
)) {
327 mp
->m_flags
|= XFS_MOUNT_NOUUID
;
328 } else if (!strcmp(this_char
, MNTOPT_BARRIER
)) {
329 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
330 } else if (!strcmp(this_char
, MNTOPT_NOBARRIER
)) {
331 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
332 } else if (!strcmp(this_char
, MNTOPT_IKEEP
)) {
333 mp
->m_flags
|= XFS_MOUNT_IKEEP
;
334 } else if (!strcmp(this_char
, MNTOPT_NOIKEEP
)) {
335 mp
->m_flags
&= ~XFS_MOUNT_IKEEP
;
336 } else if (!strcmp(this_char
, MNTOPT_LARGEIO
)) {
337 mp
->m_flags
&= ~XFS_MOUNT_COMPAT_IOSIZE
;
338 } else if (!strcmp(this_char
, MNTOPT_NOLARGEIO
)) {
339 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
340 } else if (!strcmp(this_char
, MNTOPT_ATTR2
)) {
341 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
342 } else if (!strcmp(this_char
, MNTOPT_NOATTR2
)) {
343 mp
->m_flags
&= ~XFS_MOUNT_ATTR2
;
344 mp
->m_flags
|= XFS_MOUNT_NOATTR2
;
345 } else if (!strcmp(this_char
, MNTOPT_FILESTREAM
)) {
346 mp
->m_flags
|= XFS_MOUNT_FILESTREAMS
;
347 } else if (!strcmp(this_char
, MNTOPT_NOQUOTA
)) {
348 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACCT
;
349 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ENFD
;
350 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACTIVE
;
351 } else if (!strcmp(this_char
, MNTOPT_QUOTA
) ||
352 !strcmp(this_char
, MNTOPT_UQUOTA
) ||
353 !strcmp(this_char
, MNTOPT_USRQUOTA
)) {
354 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
|
356 } else if (!strcmp(this_char
, MNTOPT_QUOTANOENF
) ||
357 !strcmp(this_char
, MNTOPT_UQUOTANOENF
)) {
358 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
);
359 mp
->m_qflags
&= ~XFS_UQUOTA_ENFD
;
360 } else if (!strcmp(this_char
, MNTOPT_PQUOTA
) ||
361 !strcmp(this_char
, MNTOPT_PRJQUOTA
)) {
362 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
|
364 } else if (!strcmp(this_char
, MNTOPT_PQUOTANOENF
)) {
365 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
);
366 mp
->m_qflags
&= ~XFS_PQUOTA_ENFD
;
367 } else if (!strcmp(this_char
, MNTOPT_GQUOTA
) ||
368 !strcmp(this_char
, MNTOPT_GRPQUOTA
)) {
369 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
|
371 } else if (!strcmp(this_char
, MNTOPT_GQUOTANOENF
)) {
372 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
);
373 mp
->m_qflags
&= ~XFS_GQUOTA_ENFD
;
374 } else if (!strcmp(this_char
, MNTOPT_DELAYLOG
)) {
376 "delaylog is the default now, option is deprecated.");
377 } else if (!strcmp(this_char
, MNTOPT_NODELAYLOG
)) {
379 "nodelaylog support has been removed, option is deprecated.");
380 } else if (!strcmp(this_char
, MNTOPT_DISCARD
)) {
381 mp
->m_flags
|= XFS_MOUNT_DISCARD
;
382 } else if (!strcmp(this_char
, MNTOPT_NODISCARD
)) {
383 mp
->m_flags
&= ~XFS_MOUNT_DISCARD
;
384 } else if (!strcmp(this_char
, "ihashsize")) {
386 "ihashsize no longer used, option is deprecated.");
387 } else if (!strcmp(this_char
, "osyncisdsync")) {
389 "osyncisdsync has no effect, option is deprecated.");
390 } else if (!strcmp(this_char
, "osyncisosync")) {
392 "osyncisosync has no effect, option is deprecated.");
393 } else if (!strcmp(this_char
, "irixsgid")) {
395 "irixsgid is now a sysctl(2) variable, option is deprecated.");
397 xfs_warn(mp
, "unknown mount option [%s].", this_char
);
403 * no recovery flag requires a read-only mount
405 if ((mp
->m_flags
& XFS_MOUNT_NORECOVERY
) &&
406 !(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
407 xfs_warn(mp
, "no-recovery mounts must be read-only.");
411 if ((mp
->m_flags
& XFS_MOUNT_NOALIGN
) && (dsunit
|| dswidth
)) {
413 "sunit and swidth options incompatible with the noalign option");
417 #ifndef CONFIG_XFS_QUOTA
418 if (XFS_IS_QUOTA_RUNNING(mp
)) {
419 xfs_warn(mp
, "quota support not available in this kernel.");
424 if ((mp
->m_qflags
& (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
)) &&
425 (mp
->m_qflags
& (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
))) {
426 xfs_warn(mp
, "cannot mount with both project and group quota");
430 if ((dsunit
&& !dswidth
) || (!dsunit
&& dswidth
)) {
431 xfs_warn(mp
, "sunit and swidth must be specified together");
435 if (dsunit
&& (dswidth
% dsunit
!= 0)) {
437 "stripe width (%d) must be a multiple of the stripe unit (%d)",
443 if (dsunit
&& !(mp
->m_flags
& XFS_MOUNT_NOALIGN
)) {
445 * At this point the superblock has not been read
446 * in, therefore we do not know the block size.
447 * Before the mount call ends we will convert
450 mp
->m_dalign
= dsunit
;
451 mp
->m_swidth
= dswidth
;
454 if (mp
->m_logbufs
!= -1 &&
455 mp
->m_logbufs
!= 0 &&
456 (mp
->m_logbufs
< XLOG_MIN_ICLOGS
||
457 mp
->m_logbufs
> XLOG_MAX_ICLOGS
)) {
458 xfs_warn(mp
, "invalid logbufs value: %d [not %d-%d]",
459 mp
->m_logbufs
, XLOG_MIN_ICLOGS
, XLOG_MAX_ICLOGS
);
460 return XFS_ERROR(EINVAL
);
462 if (mp
->m_logbsize
!= -1 &&
463 mp
->m_logbsize
!= 0 &&
464 (mp
->m_logbsize
< XLOG_MIN_RECORD_BSIZE
||
465 mp
->m_logbsize
> XLOG_MAX_RECORD_BSIZE
||
466 !is_power_of_2(mp
->m_logbsize
))) {
468 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
470 return XFS_ERROR(EINVAL
);
474 if (iosizelog
> XFS_MAX_IO_LOG
||
475 iosizelog
< XFS_MIN_IO_LOG
) {
476 xfs_warn(mp
, "invalid log iosize: %d [not %d-%d]",
477 iosizelog
, XFS_MIN_IO_LOG
,
479 return XFS_ERROR(EINVAL
);
482 mp
->m_flags
|= XFS_MOUNT_DFLT_IOSIZE
;
483 mp
->m_readio_log
= iosizelog
;
484 mp
->m_writeio_log
= iosizelog
;
490 struct proc_xfs_info
{
497 struct xfs_mount
*mp
,
500 static struct proc_xfs_info xfs_info_set
[] = {
501 /* the few simple ones we can get from the mount struct */
502 { XFS_MOUNT_IKEEP
, "," MNTOPT_IKEEP
},
503 { XFS_MOUNT_WSYNC
, "," MNTOPT_WSYNC
},
504 { XFS_MOUNT_NOALIGN
, "," MNTOPT_NOALIGN
},
505 { XFS_MOUNT_SWALLOC
, "," MNTOPT_SWALLOC
},
506 { XFS_MOUNT_NOUUID
, "," MNTOPT_NOUUID
},
507 { XFS_MOUNT_NORECOVERY
, "," MNTOPT_NORECOVERY
},
508 { XFS_MOUNT_ATTR2
, "," MNTOPT_ATTR2
},
509 { XFS_MOUNT_FILESTREAMS
, "," MNTOPT_FILESTREAM
},
510 { XFS_MOUNT_GRPID
, "," MNTOPT_GRPID
},
511 { XFS_MOUNT_DISCARD
, "," MNTOPT_DISCARD
},
512 { XFS_MOUNT_SMALL_INUMS
, "," MNTOPT_32BITINODE
},
515 static struct proc_xfs_info xfs_info_unset
[] = {
516 /* the few simple ones we can get from the mount struct */
517 { XFS_MOUNT_COMPAT_IOSIZE
, "," MNTOPT_LARGEIO
},
518 { XFS_MOUNT_BARRIER
, "," MNTOPT_NOBARRIER
},
519 { XFS_MOUNT_SMALL_INUMS
, "," MNTOPT_64BITINODE
},
522 struct proc_xfs_info
*xfs_infop
;
524 for (xfs_infop
= xfs_info_set
; xfs_infop
->flag
; xfs_infop
++) {
525 if (mp
->m_flags
& xfs_infop
->flag
)
526 seq_puts(m
, xfs_infop
->str
);
528 for (xfs_infop
= xfs_info_unset
; xfs_infop
->flag
; xfs_infop
++) {
529 if (!(mp
->m_flags
& xfs_infop
->flag
))
530 seq_puts(m
, xfs_infop
->str
);
533 if (mp
->m_flags
& XFS_MOUNT_DFLT_IOSIZE
)
534 seq_printf(m
, "," MNTOPT_ALLOCSIZE
"=%dk",
535 (int)(1 << mp
->m_writeio_log
) >> 10);
537 if (mp
->m_logbufs
> 0)
538 seq_printf(m
, "," MNTOPT_LOGBUFS
"=%d", mp
->m_logbufs
);
539 if (mp
->m_logbsize
> 0)
540 seq_printf(m
, "," MNTOPT_LOGBSIZE
"=%dk", mp
->m_logbsize
>> 10);
543 seq_printf(m
, "," MNTOPT_LOGDEV
"=%s", mp
->m_logname
);
545 seq_printf(m
, "," MNTOPT_RTDEV
"=%s", mp
->m_rtname
);
547 if (mp
->m_dalign
> 0)
548 seq_printf(m
, "," MNTOPT_SUNIT
"=%d",
549 (int)XFS_FSB_TO_BB(mp
, mp
->m_dalign
));
550 if (mp
->m_swidth
> 0)
551 seq_printf(m
, "," MNTOPT_SWIDTH
"=%d",
552 (int)XFS_FSB_TO_BB(mp
, mp
->m_swidth
));
554 if (mp
->m_qflags
& (XFS_UQUOTA_ACCT
|XFS_UQUOTA_ENFD
))
555 seq_puts(m
, "," MNTOPT_USRQUOTA
);
556 else if (mp
->m_qflags
& XFS_UQUOTA_ACCT
)
557 seq_puts(m
, "," MNTOPT_UQUOTANOENF
);
559 /* Either project or group quotas can be active, not both */
561 if (mp
->m_qflags
& XFS_PQUOTA_ACCT
) {
562 if (mp
->m_qflags
& XFS_PQUOTA_ENFD
)
563 seq_puts(m
, "," MNTOPT_PRJQUOTA
);
565 seq_puts(m
, "," MNTOPT_PQUOTANOENF
);
566 } else if (mp
->m_qflags
& XFS_GQUOTA_ACCT
) {
567 if (mp
->m_qflags
& XFS_GQUOTA_ENFD
)
568 seq_puts(m
, "," MNTOPT_GRPQUOTA
);
570 seq_puts(m
, "," MNTOPT_GQUOTANOENF
);
573 if (!(mp
->m_qflags
& XFS_ALL_QUOTA_ACCT
))
574 seq_puts(m
, "," MNTOPT_NOQUOTA
);
580 unsigned int blockshift
)
582 unsigned int pagefactor
= 1;
583 unsigned int bitshift
= BITS_PER_LONG
- 1;
585 /* Figure out maximum filesize, on Linux this can depend on
586 * the filesystem blocksize (on 32 bit platforms).
587 * __block_write_begin does this in an [unsigned] long...
588 * page->index << (PAGE_CACHE_SHIFT - bbits)
589 * So, for page sized blocks (4K on 32 bit platforms),
590 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
591 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
592 * but for smaller blocksizes it is less (bbits = log2 bsize).
593 * Note1: get_block_t takes a long (implicit cast from above)
594 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
595 * can optionally convert the [unsigned] long from above into
596 * an [unsigned] long long.
599 #if BITS_PER_LONG == 32
600 # if defined(CONFIG_LBDAF)
601 ASSERT(sizeof(sector_t
) == 8);
602 pagefactor
= PAGE_CACHE_SIZE
;
603 bitshift
= BITS_PER_LONG
;
605 pagefactor
= PAGE_CACHE_SIZE
>> (PAGE_CACHE_SHIFT
- blockshift
);
609 return (((__uint64_t
)pagefactor
) << bitshift
) - 1;
613 xfs_set_inode32(struct xfs_mount
*mp
)
615 xfs_agnumber_t index
= 0;
616 xfs_agnumber_t maxagi
= 0;
617 xfs_sb_t
*sbp
= &mp
->m_sb
;
618 xfs_agnumber_t max_metadata
;
619 xfs_agino_t agino
= XFS_OFFBNO_TO_AGINO(mp
, sbp
->sb_agblocks
-1, 0);
620 xfs_ino_t ino
= XFS_AGINO_TO_INO(mp
, sbp
->sb_agcount
-1, agino
);
623 /* Calculate how much should be reserved for inodes to meet
624 * the max inode percentage.
626 if (mp
->m_maxicount
) {
629 icount
= sbp
->sb_dblocks
* sbp
->sb_imax_pct
;
631 icount
+= sbp
->sb_agblocks
- 1;
632 do_div(icount
, sbp
->sb_agblocks
);
633 max_metadata
= icount
;
635 max_metadata
= sbp
->sb_agcount
;
638 for (index
= 0; index
< sbp
->sb_agcount
; index
++) {
639 ino
= XFS_AGINO_TO_INO(mp
, index
, agino
);
641 if (ino
> XFS_MAXINUMBER_32
) {
642 pag
= xfs_perag_get(mp
, index
);
643 pag
->pagi_inodeok
= 0;
644 pag
->pagf_metadata
= 0;
649 pag
= xfs_perag_get(mp
, index
);
650 pag
->pagi_inodeok
= 1;
652 if (index
< max_metadata
)
653 pag
->pagf_metadata
= 1;
656 mp
->m_flags
|= (XFS_MOUNT_32BITINODES
|
657 XFS_MOUNT_SMALL_INUMS
);
663 xfs_set_inode64(struct xfs_mount
*mp
)
665 xfs_agnumber_t index
= 0;
667 for (index
= 0; index
< mp
->m_sb
.sb_agcount
; index
++) {
668 struct xfs_perag
*pag
;
670 pag
= xfs_perag_get(mp
, index
);
671 pag
->pagi_inodeok
= 1;
672 pag
->pagf_metadata
= 0;
676 /* There is no need for lock protection on m_flags,
677 * the rw_semaphore of the VFS superblock is locked
678 * during mount/umount/remount operations, so this is
679 * enough to avoid concurency on the m_flags field
681 mp
->m_flags
&= ~(XFS_MOUNT_32BITINODES
|
682 XFS_MOUNT_SMALL_INUMS
);
690 struct block_device
**bdevp
)
694 *bdevp
= blkdev_get_by_path(name
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
,
696 if (IS_ERR(*bdevp
)) {
697 error
= PTR_ERR(*bdevp
);
698 xfs_warn(mp
, "Invalid device [%s], error=%d\n", name
, error
);
706 struct block_device
*bdev
)
709 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
713 xfs_blkdev_issue_flush(
714 xfs_buftarg_t
*buftarg
)
716 blkdev_issue_flush(buftarg
->bt_bdev
, GFP_NOFS
, NULL
);
721 struct xfs_mount
*mp
)
723 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
724 struct block_device
*logdev
= mp
->m_logdev_targp
->bt_bdev
;
725 xfs_free_buftarg(mp
, mp
->m_logdev_targp
);
726 xfs_blkdev_put(logdev
);
728 if (mp
->m_rtdev_targp
) {
729 struct block_device
*rtdev
= mp
->m_rtdev_targp
->bt_bdev
;
730 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
731 xfs_blkdev_put(rtdev
);
733 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
737 * The file system configurations are:
738 * (1) device (partition) with data and internal log
739 * (2) logical volume with data and log subvolumes.
740 * (3) logical volume with data, log, and realtime subvolumes.
742 * We only have to handle opening the log and realtime volumes here if
743 * they are present. The data subvolume has already been opened by
744 * get_sb_bdev() and is stored in sb->s_bdev.
748 struct xfs_mount
*mp
)
750 struct block_device
*ddev
= mp
->m_super
->s_bdev
;
751 struct block_device
*logdev
= NULL
, *rtdev
= NULL
;
755 * Open real time and log devices - order is important.
758 error
= xfs_blkdev_get(mp
, mp
->m_logname
, &logdev
);
764 error
= xfs_blkdev_get(mp
, mp
->m_rtname
, &rtdev
);
766 goto out_close_logdev
;
768 if (rtdev
== ddev
|| rtdev
== logdev
) {
770 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
772 goto out_close_rtdev
;
777 * Setup xfs_mount buffer target pointers
780 mp
->m_ddev_targp
= xfs_alloc_buftarg(mp
, ddev
, 0, mp
->m_fsname
);
781 if (!mp
->m_ddev_targp
)
782 goto out_close_rtdev
;
785 mp
->m_rtdev_targp
= xfs_alloc_buftarg(mp
, rtdev
, 1,
787 if (!mp
->m_rtdev_targp
)
788 goto out_free_ddev_targ
;
791 if (logdev
&& logdev
!= ddev
) {
792 mp
->m_logdev_targp
= xfs_alloc_buftarg(mp
, logdev
, 1,
794 if (!mp
->m_logdev_targp
)
795 goto out_free_rtdev_targ
;
797 mp
->m_logdev_targp
= mp
->m_ddev_targp
;
803 if (mp
->m_rtdev_targp
)
804 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
806 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
809 xfs_blkdev_put(rtdev
);
811 if (logdev
&& logdev
!= ddev
)
812 xfs_blkdev_put(logdev
);
818 * Setup xfs_mount buffer target pointers based on superblock
822 struct xfs_mount
*mp
)
826 error
= xfs_setsize_buftarg(mp
->m_ddev_targp
, mp
->m_sb
.sb_blocksize
,
827 mp
->m_sb
.sb_sectsize
);
831 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
832 unsigned int log_sector_size
= BBSIZE
;
834 if (xfs_sb_version_hassector(&mp
->m_sb
))
835 log_sector_size
= mp
->m_sb
.sb_logsectsize
;
836 error
= xfs_setsize_buftarg(mp
->m_logdev_targp
,
837 mp
->m_sb
.sb_blocksize
,
842 if (mp
->m_rtdev_targp
) {
843 error
= xfs_setsize_buftarg(mp
->m_rtdev_targp
,
844 mp
->m_sb
.sb_blocksize
,
845 mp
->m_sb
.sb_sectsize
);
854 xfs_init_mount_workqueues(
855 struct xfs_mount
*mp
)
857 mp
->m_data_workqueue
= alloc_workqueue("xfs-data/%s",
858 WQ_MEM_RECLAIM
, 0, mp
->m_fsname
);
859 if (!mp
->m_data_workqueue
)
862 mp
->m_unwritten_workqueue
= alloc_workqueue("xfs-conv/%s",
863 WQ_MEM_RECLAIM
, 0, mp
->m_fsname
);
864 if (!mp
->m_unwritten_workqueue
)
865 goto out_destroy_data_iodone_queue
;
867 mp
->m_cil_workqueue
= alloc_workqueue("xfs-cil/%s",
868 WQ_MEM_RECLAIM
, 0, mp
->m_fsname
);
869 if (!mp
->m_cil_workqueue
)
870 goto out_destroy_unwritten
;
872 mp
->m_reclaim_workqueue
= alloc_workqueue("xfs-reclaim/%s",
873 WQ_NON_REENTRANT
, 0, mp
->m_fsname
);
874 if (!mp
->m_reclaim_workqueue
)
875 goto out_destroy_cil
;
877 mp
->m_log_workqueue
= alloc_workqueue("xfs-log/%s",
878 WQ_NON_REENTRANT
, 0, mp
->m_fsname
);
879 if (!mp
->m_log_workqueue
)
880 goto out_destroy_reclaim
;
882 mp
->m_eofblocks_workqueue
= alloc_workqueue("xfs-eofblocks/%s",
883 WQ_NON_REENTRANT
, 0, mp
->m_fsname
);
884 if (!mp
->m_eofblocks_workqueue
)
885 goto out_destroy_log
;
890 destroy_workqueue(mp
->m_log_workqueue
);
892 destroy_workqueue(mp
->m_reclaim_workqueue
);
894 destroy_workqueue(mp
->m_cil_workqueue
);
895 out_destroy_unwritten
:
896 destroy_workqueue(mp
->m_unwritten_workqueue
);
897 out_destroy_data_iodone_queue
:
898 destroy_workqueue(mp
->m_data_workqueue
);
904 xfs_destroy_mount_workqueues(
905 struct xfs_mount
*mp
)
907 destroy_workqueue(mp
->m_eofblocks_workqueue
);
908 destroy_workqueue(mp
->m_log_workqueue
);
909 destroy_workqueue(mp
->m_reclaim_workqueue
);
910 destroy_workqueue(mp
->m_cil_workqueue
);
911 destroy_workqueue(mp
->m_data_workqueue
);
912 destroy_workqueue(mp
->m_unwritten_workqueue
);
916 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
917 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
918 * for IO to complete so that we effectively throttle multiple callers to the
919 * rate at which IO is completing.
923 struct xfs_mount
*mp
)
925 struct super_block
*sb
= mp
->m_super
;
927 if (down_read_trylock(&sb
->s_umount
)) {
929 up_read(&sb
->s_umount
);
933 /* Catch misguided souls that try to use this interface on XFS */
934 STATIC
struct inode
*
936 struct super_block
*sb
)
943 * Now that the generic code is guaranteed not to be accessing
944 * the linux inode, we can reclaim the inode.
947 xfs_fs_destroy_inode(
950 struct xfs_inode
*ip
= XFS_I(inode
);
952 trace_xfs_destroy_inode(ip
);
954 XFS_STATS_INC(vn_reclaim
);
956 /* bad inode, get out here ASAP */
957 if (is_bad_inode(inode
))
960 ASSERT(XFS_FORCED_SHUTDOWN(ip
->i_mount
) || ip
->i_delayed_blks
== 0);
963 * We should never get here with one of the reclaim flags already set.
965 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIMABLE
));
966 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIM
));
969 * We always use background reclaim here because even if the
970 * inode is clean, it still may be under IO and hence we have
971 * to take the flush lock. The background reclaim path handles
972 * this more efficiently than we can here, so simply let background
973 * reclaim tear down all inodes.
976 xfs_inode_set_reclaim_tag(ip
);
980 * Slab object creation initialisation for the XFS inode.
981 * This covers only the idempotent fields in the XFS inode;
982 * all other fields need to be initialised on allocation
983 * from the slab. This avoids the need to repeatedly initialise
984 * fields in the xfs inode that left in the initialise state
985 * when freeing the inode.
988 xfs_fs_inode_init_once(
991 struct xfs_inode
*ip
= inode
;
993 memset(ip
, 0, sizeof(struct xfs_inode
));
996 inode_init_once(VFS_I(ip
));
999 atomic_set(&ip
->i_pincount
, 0);
1000 spin_lock_init(&ip
->i_flags_lock
);
1002 mrlock_init(&ip
->i_lock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
1003 "xfsino", ip
->i_ino
);
1008 struct inode
*inode
)
1010 xfs_inode_t
*ip
= XFS_I(inode
);
1012 ASSERT(!rwsem_is_locked(&ip
->i_iolock
.mr_lock
));
1014 trace_xfs_evict_inode(ip
);
1016 truncate_inode_pages(&inode
->i_data
, 0);
1018 XFS_STATS_INC(vn_rele
);
1019 XFS_STATS_INC(vn_remove
);
1020 XFS_STATS_DEC(vn_active
);
1026 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1027 * serialised against cache hits here via the inode->i_lock and igrab() in
1028 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1029 * racing with us, and it avoids needing to grab a spinlock here for every inode
1030 * we drop the final reference on.
1034 struct inode
*inode
)
1036 struct xfs_inode
*ip
= XFS_I(inode
);
1038 return generic_drop_inode(inode
) || (ip
->i_flags
& XFS_IDONTCACHE
);
1043 struct xfs_mount
*mp
)
1045 kfree(mp
->m_fsname
);
1046 kfree(mp
->m_rtname
);
1047 kfree(mp
->m_logname
);
1052 struct super_block
*sb
)
1054 struct xfs_mount
*mp
= XFS_M(sb
);
1056 xfs_filestream_unmount(mp
);
1060 xfs_icsb_destroy_counters(mp
);
1061 xfs_destroy_mount_workqueues(mp
);
1062 xfs_close_devices(mp
);
1063 xfs_free_fsname(mp
);
1069 struct super_block
*sb
,
1072 struct xfs_mount
*mp
= XFS_M(sb
);
1075 * Doing anything during the async pass would be counterproductive.
1080 xfs_log_force(mp
, XFS_LOG_SYNC
);
1083 * The disk must be active because we're syncing.
1084 * We schedule log work now (now that the disk is
1085 * active) instead of later (when it might not be).
1087 flush_delayed_work(&mp
->m_log
->l_work
);
1095 struct dentry
*dentry
,
1096 struct kstatfs
*statp
)
1098 struct xfs_mount
*mp
= XFS_M(dentry
->d_sb
);
1099 xfs_sb_t
*sbp
= &mp
->m_sb
;
1100 struct xfs_inode
*ip
= XFS_I(dentry
->d_inode
);
1101 __uint64_t fakeinos
, id
;
1105 statp
->f_type
= XFS_SB_MAGIC
;
1106 statp
->f_namelen
= MAXNAMELEN
- 1;
1108 id
= huge_encode_dev(mp
->m_ddev_targp
->bt_dev
);
1109 statp
->f_fsid
.val
[0] = (u32
)id
;
1110 statp
->f_fsid
.val
[1] = (u32
)(id
>> 32);
1112 xfs_icsb_sync_counters(mp
, XFS_ICSB_LAZY_COUNT
);
1114 spin_lock(&mp
->m_sb_lock
);
1115 statp
->f_bsize
= sbp
->sb_blocksize
;
1116 lsize
= sbp
->sb_logstart
? sbp
->sb_logblocks
: 0;
1117 statp
->f_blocks
= sbp
->sb_dblocks
- lsize
;
1118 statp
->f_bfree
= statp
->f_bavail
=
1119 sbp
->sb_fdblocks
- XFS_ALLOC_SET_ASIDE(mp
);
1120 fakeinos
= statp
->f_bfree
<< sbp
->sb_inopblog
;
1122 MIN(sbp
->sb_icount
+ fakeinos
, (__uint64_t
)XFS_MAXINUMBER
);
1123 if (mp
->m_maxicount
)
1124 statp
->f_files
= min_t(typeof(statp
->f_files
),
1128 /* make sure statp->f_ffree does not underflow */
1129 ffree
= statp
->f_files
- (sbp
->sb_icount
- sbp
->sb_ifree
);
1130 statp
->f_ffree
= max_t(__int64_t
, ffree
, 0);
1132 spin_unlock(&mp
->m_sb_lock
);
1134 if ((ip
->i_d
.di_flags
& XFS_DIFLAG_PROJINHERIT
) &&
1135 ((mp
->m_qflags
& (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))) ==
1136 (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))
1137 xfs_qm_statvfs(ip
, statp
);
1142 xfs_save_resvblks(struct xfs_mount
*mp
)
1144 __uint64_t resblks
= 0;
1146 mp
->m_resblks_save
= mp
->m_resblks
;
1147 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1151 xfs_restore_resvblks(struct xfs_mount
*mp
)
1155 if (mp
->m_resblks_save
) {
1156 resblks
= mp
->m_resblks_save
;
1157 mp
->m_resblks_save
= 0;
1159 resblks
= xfs_default_resblks(mp
);
1161 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1165 * Trigger writeback of all the dirty metadata in the file system.
1167 * This ensures that the metadata is written to their location on disk rather
1168 * than just existing in transactions in the log. This means after a quiesce
1169 * there is no log replay required to write the inodes to disk - this is the
1170 * primary difference between a sync and a quiesce.
1172 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1173 * it is started again when appropriate.
1177 struct xfs_mount
*mp
)
1181 /* wait for all modifications to complete */
1182 while (atomic_read(&mp
->m_active_trans
) > 0)
1185 /* force the log to unpin objects from the now complete transactions */
1186 xfs_log_force(mp
, XFS_LOG_SYNC
);
1188 /* reclaim inodes to do any IO before the freeze completes */
1189 xfs_reclaim_inodes(mp
, 0);
1190 xfs_reclaim_inodes(mp
, SYNC_WAIT
);
1192 /* Push the superblock and write an unmount record */
1193 error
= xfs_log_sbcount(mp
);
1195 xfs_warn(mp
, "xfs_attr_quiesce: failed to log sb changes. "
1196 "Frozen image may not be consistent.");
1198 * Just warn here till VFS can correctly support
1199 * read-only remount without racing.
1201 WARN_ON(atomic_read(&mp
->m_active_trans
) != 0);
1203 xfs_log_quiesce(mp
);
1208 struct super_block
*sb
,
1212 struct xfs_mount
*mp
= XFS_M(sb
);
1213 substring_t args
[MAX_OPT_ARGS
];
1217 while ((p
= strsep(&options
, ",")) != NULL
) {
1223 token
= match_token(p
, tokens
, args
);
1226 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
1229 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
1232 mp
->m_maxagi
= xfs_set_inode64(mp
);
1235 mp
->m_maxagi
= xfs_set_inode32(mp
);
1239 * Logically we would return an error here to prevent
1240 * users from believing they might have changed
1241 * mount options using remount which can't be changed.
1243 * But unfortunately mount(8) adds all options from
1244 * mtab and fstab to the mount arguments in some cases
1245 * so we can't blindly reject options, but have to
1246 * check for each specified option if it actually
1247 * differs from the currently set option and only
1248 * reject it if that's the case.
1250 * Until that is implemented we return success for
1251 * every remount request, and silently ignore all
1252 * options that we can't actually change.
1256 "mount option \"%s\" not supported for remount\n", p
);
1265 if ((mp
->m_flags
& XFS_MOUNT_RDONLY
) && !(*flags
& MS_RDONLY
)) {
1266 mp
->m_flags
&= ~XFS_MOUNT_RDONLY
;
1269 * If this is the first remount to writeable state we
1270 * might have some superblock changes to update.
1272 if (mp
->m_update_flags
) {
1273 error
= xfs_mount_log_sb(mp
, mp
->m_update_flags
);
1275 xfs_warn(mp
, "failed to write sb changes");
1278 mp
->m_update_flags
= 0;
1282 * Fill out the reserve pool if it is empty. Use the stashed
1283 * value if it is non-zero, otherwise go with the default.
1285 xfs_restore_resvblks(mp
);
1286 xfs_log_work_queue(mp
);
1290 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
) && (*flags
& MS_RDONLY
)) {
1292 * Before we sync the metadata, we need to free up the reserve
1293 * block pool so that the used block count in the superblock on
1294 * disk is correct at the end of the remount. Stash the current
1295 * reserve pool size so that if we get remounted rw, we can
1296 * return it to the same size.
1298 xfs_save_resvblks(mp
);
1299 xfs_quiesce_attr(mp
);
1300 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
1307 * Second stage of a freeze. The data is already frozen so we only
1308 * need to take care of the metadata. Once that's done write a dummy
1309 * record to dirty the log in case of a crash while frozen.
1313 struct super_block
*sb
)
1315 struct xfs_mount
*mp
= XFS_M(sb
);
1317 xfs_save_resvblks(mp
);
1318 xfs_quiesce_attr(mp
);
1319 return -xfs_fs_log_dummy(mp
);
1324 struct super_block
*sb
)
1326 struct xfs_mount
*mp
= XFS_M(sb
);
1328 xfs_restore_resvblks(mp
);
1329 xfs_log_work_queue(mp
);
1334 xfs_fs_show_options(
1336 struct dentry
*root
)
1338 return -xfs_showargs(XFS_M(root
->d_sb
), m
);
1342 * This function fills in xfs_mount_t fields based on mount args.
1343 * Note: the superblock _has_ now been read in.
1347 struct xfs_mount
*mp
)
1349 int ronly
= (mp
->m_flags
& XFS_MOUNT_RDONLY
);
1351 /* Fail a mount where the logbuf is smaller than the log stripe */
1352 if (xfs_sb_version_haslogv2(&mp
->m_sb
)) {
1353 if (mp
->m_logbsize
<= 0 &&
1354 mp
->m_sb
.sb_logsunit
> XLOG_BIG_RECORD_BSIZE
) {
1355 mp
->m_logbsize
= mp
->m_sb
.sb_logsunit
;
1356 } else if (mp
->m_logbsize
> 0 &&
1357 mp
->m_logbsize
< mp
->m_sb
.sb_logsunit
) {
1359 "logbuf size must be greater than or equal to log stripe size");
1360 return XFS_ERROR(EINVAL
);
1363 /* Fail a mount if the logbuf is larger than 32K */
1364 if (mp
->m_logbsize
> XLOG_BIG_RECORD_BSIZE
) {
1366 "logbuf size for version 1 logs must be 16K or 32K");
1367 return XFS_ERROR(EINVAL
);
1372 * V5 filesystems always use attr2 format for attributes.
1374 if (xfs_sb_version_hascrc(&mp
->m_sb
) &&
1375 (mp
->m_flags
& XFS_MOUNT_NOATTR2
)) {
1377 "Cannot mount a V5 filesystem as %s. %s is always enabled for V5 filesystems.",
1378 MNTOPT_NOATTR2
, MNTOPT_ATTR2
);
1379 return XFS_ERROR(EINVAL
);
1383 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1384 * told by noattr2 to turn it off
1386 if (xfs_sb_version_hasattr2(&mp
->m_sb
) &&
1387 !(mp
->m_flags
& XFS_MOUNT_NOATTR2
))
1388 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
1391 * prohibit r/w mounts of read-only filesystems
1393 if ((mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) && !ronly
) {
1395 "cannot mount a read-only filesystem as read-write");
1396 return XFS_ERROR(EROFS
);
1404 struct super_block
*sb
,
1409 struct xfs_mount
*mp
= NULL
;
1410 int flags
= 0, error
= ENOMEM
;
1412 mp
= kzalloc(sizeof(struct xfs_mount
), GFP_KERNEL
);
1416 spin_lock_init(&mp
->m_sb_lock
);
1417 mutex_init(&mp
->m_growlock
);
1418 atomic_set(&mp
->m_active_trans
, 0);
1419 INIT_DELAYED_WORK(&mp
->m_reclaim_work
, xfs_reclaim_worker
);
1420 INIT_DELAYED_WORK(&mp
->m_eofblocks_work
, xfs_eofblocks_worker
);
1425 error
= xfs_parseargs(mp
, (char *)data
);
1427 goto out_free_fsname
;
1429 sb_min_blocksize(sb
, BBSIZE
);
1430 sb
->s_xattr
= xfs_xattr_handlers
;
1431 sb
->s_export_op
= &xfs_export_operations
;
1432 #ifdef CONFIG_XFS_QUOTA
1433 sb
->s_qcop
= &xfs_quotactl_operations
;
1435 sb
->s_op
= &xfs_super_operations
;
1438 flags
|= XFS_MFSI_QUIET
;
1440 error
= xfs_open_devices(mp
);
1442 goto out_free_fsname
;
1444 error
= xfs_init_mount_workqueues(mp
);
1446 goto out_close_devices
;
1448 error
= xfs_icsb_init_counters(mp
);
1450 goto out_destroy_workqueues
;
1452 error
= xfs_readsb(mp
, flags
);
1454 goto out_destroy_counters
;
1456 error
= xfs_finish_flags(mp
);
1460 error
= xfs_setup_devices(mp
);
1464 error
= xfs_filestream_mount(mp
);
1469 * we must configure the block size in the superblock before we run the
1470 * full mount process as the mount process can lookup and cache inodes.
1472 sb
->s_magic
= XFS_SB_MAGIC
;
1473 sb
->s_blocksize
= mp
->m_sb
.sb_blocksize
;
1474 sb
->s_blocksize_bits
= ffs(sb
->s_blocksize
) - 1;
1475 sb
->s_maxbytes
= xfs_max_file_offset(sb
->s_blocksize_bits
);
1476 sb
->s_max_links
= XFS_MAXLINK
;
1477 sb
->s_time_gran
= 1;
1478 set_posix_acl_flag(sb
);
1480 /* version 5 superblocks support inode version counters. */
1481 if (XFS_SB_VERSION_NUM(&mp
->m_sb
) == XFS_SB_VERSION_5
)
1482 sb
->s_flags
|= MS_I_VERSION
;
1484 error
= xfs_mountfs(mp
);
1486 goto out_filestream_unmount
;
1488 root
= igrab(VFS_I(mp
->m_rootip
));
1493 if (is_bad_inode(root
)) {
1497 sb
->s_root
= d_make_root(root
);
1505 out_filestream_unmount
:
1506 xfs_filestream_unmount(mp
);
1509 out_destroy_counters
:
1510 xfs_icsb_destroy_counters(mp
);
1511 out_destroy_workqueues
:
1512 xfs_destroy_mount_workqueues(mp
);
1514 xfs_close_devices(mp
);
1516 xfs_free_fsname(mp
);
1522 xfs_filestream_unmount(mp
);
1527 STATIC
struct dentry
*
1529 struct file_system_type
*fs_type
,
1531 const char *dev_name
,
1534 return mount_bdev(fs_type
, flags
, dev_name
, data
, xfs_fs_fill_super
);
1538 xfs_fs_nr_cached_objects(
1539 struct super_block
*sb
)
1541 return xfs_reclaim_inodes_count(XFS_M(sb
));
1545 xfs_fs_free_cached_objects(
1546 struct super_block
*sb
,
1549 xfs_reclaim_inodes_nr(XFS_M(sb
), nr_to_scan
);
1552 static const struct super_operations xfs_super_operations
= {
1553 .alloc_inode
= xfs_fs_alloc_inode
,
1554 .destroy_inode
= xfs_fs_destroy_inode
,
1555 .evict_inode
= xfs_fs_evict_inode
,
1556 .drop_inode
= xfs_fs_drop_inode
,
1557 .put_super
= xfs_fs_put_super
,
1558 .sync_fs
= xfs_fs_sync_fs
,
1559 .freeze_fs
= xfs_fs_freeze
,
1560 .unfreeze_fs
= xfs_fs_unfreeze
,
1561 .statfs
= xfs_fs_statfs
,
1562 .remount_fs
= xfs_fs_remount
,
1563 .show_options
= xfs_fs_show_options
,
1564 .nr_cached_objects
= xfs_fs_nr_cached_objects
,
1565 .free_cached_objects
= xfs_fs_free_cached_objects
,
1568 static struct file_system_type xfs_fs_type
= {
1569 .owner
= THIS_MODULE
,
1571 .mount
= xfs_fs_mount
,
1572 .kill_sb
= kill_block_super
,
1573 .fs_flags
= FS_REQUIRES_DEV
,
1575 MODULE_ALIAS_FS("xfs");
1578 xfs_init_zones(void)
1581 xfs_ioend_zone
= kmem_zone_init(sizeof(xfs_ioend_t
), "xfs_ioend");
1582 if (!xfs_ioend_zone
)
1585 xfs_ioend_pool
= mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE
,
1587 if (!xfs_ioend_pool
)
1588 goto out_destroy_ioend_zone
;
1590 xfs_log_ticket_zone
= kmem_zone_init(sizeof(xlog_ticket_t
),
1592 if (!xfs_log_ticket_zone
)
1593 goto out_destroy_ioend_pool
;
1595 xfs_bmap_free_item_zone
= kmem_zone_init(sizeof(xfs_bmap_free_item_t
),
1596 "xfs_bmap_free_item");
1597 if (!xfs_bmap_free_item_zone
)
1598 goto out_destroy_log_ticket_zone
;
1600 xfs_btree_cur_zone
= kmem_zone_init(sizeof(xfs_btree_cur_t
),
1602 if (!xfs_btree_cur_zone
)
1603 goto out_destroy_bmap_free_item_zone
;
1605 xfs_da_state_zone
= kmem_zone_init(sizeof(xfs_da_state_t
),
1607 if (!xfs_da_state_zone
)
1608 goto out_destroy_btree_cur_zone
;
1610 xfs_ifork_zone
= kmem_zone_init(sizeof(xfs_ifork_t
), "xfs_ifork");
1611 if (!xfs_ifork_zone
)
1612 goto out_destroy_da_state_zone
;
1614 xfs_trans_zone
= kmem_zone_init(sizeof(xfs_trans_t
), "xfs_trans");
1615 if (!xfs_trans_zone
)
1616 goto out_destroy_ifork_zone
;
1618 xfs_log_item_desc_zone
=
1619 kmem_zone_init(sizeof(struct xfs_log_item_desc
),
1620 "xfs_log_item_desc");
1621 if (!xfs_log_item_desc_zone
)
1622 goto out_destroy_trans_zone
;
1625 * The size of the zone allocated buf log item is the maximum
1626 * size possible under XFS. This wastes a little bit of memory,
1627 * but it is much faster.
1629 xfs_buf_item_zone
= kmem_zone_init(sizeof(struct xfs_buf_log_item
),
1631 if (!xfs_buf_item_zone
)
1632 goto out_destroy_log_item_desc_zone
;
1634 xfs_efd_zone
= kmem_zone_init((sizeof(xfs_efd_log_item_t
) +
1635 ((XFS_EFD_MAX_FAST_EXTENTS
- 1) *
1636 sizeof(xfs_extent_t
))), "xfs_efd_item");
1638 goto out_destroy_buf_item_zone
;
1640 xfs_efi_zone
= kmem_zone_init((sizeof(xfs_efi_log_item_t
) +
1641 ((XFS_EFI_MAX_FAST_EXTENTS
- 1) *
1642 sizeof(xfs_extent_t
))), "xfs_efi_item");
1644 goto out_destroy_efd_zone
;
1647 kmem_zone_init_flags(sizeof(xfs_inode_t
), "xfs_inode",
1648 KM_ZONE_HWALIGN
| KM_ZONE_RECLAIM
| KM_ZONE_SPREAD
,
1649 xfs_fs_inode_init_once
);
1650 if (!xfs_inode_zone
)
1651 goto out_destroy_efi_zone
;
1654 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t
), "xfs_ili",
1655 KM_ZONE_SPREAD
, NULL
);
1657 goto out_destroy_inode_zone
;
1658 xfs_icreate_zone
= kmem_zone_init(sizeof(struct xfs_icreate_item
),
1660 if (!xfs_icreate_zone
)
1661 goto out_destroy_ili_zone
;
1665 out_destroy_ili_zone
:
1666 kmem_zone_destroy(xfs_ili_zone
);
1667 out_destroy_inode_zone
:
1668 kmem_zone_destroy(xfs_inode_zone
);
1669 out_destroy_efi_zone
:
1670 kmem_zone_destroy(xfs_efi_zone
);
1671 out_destroy_efd_zone
:
1672 kmem_zone_destroy(xfs_efd_zone
);
1673 out_destroy_buf_item_zone
:
1674 kmem_zone_destroy(xfs_buf_item_zone
);
1675 out_destroy_log_item_desc_zone
:
1676 kmem_zone_destroy(xfs_log_item_desc_zone
);
1677 out_destroy_trans_zone
:
1678 kmem_zone_destroy(xfs_trans_zone
);
1679 out_destroy_ifork_zone
:
1680 kmem_zone_destroy(xfs_ifork_zone
);
1681 out_destroy_da_state_zone
:
1682 kmem_zone_destroy(xfs_da_state_zone
);
1683 out_destroy_btree_cur_zone
:
1684 kmem_zone_destroy(xfs_btree_cur_zone
);
1685 out_destroy_bmap_free_item_zone
:
1686 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1687 out_destroy_log_ticket_zone
:
1688 kmem_zone_destroy(xfs_log_ticket_zone
);
1689 out_destroy_ioend_pool
:
1690 mempool_destroy(xfs_ioend_pool
);
1691 out_destroy_ioend_zone
:
1692 kmem_zone_destroy(xfs_ioend_zone
);
1698 xfs_destroy_zones(void)
1701 * Make sure all delayed rcu free are flushed before we
1705 kmem_zone_destroy(xfs_icreate_zone
);
1706 kmem_zone_destroy(xfs_ili_zone
);
1707 kmem_zone_destroy(xfs_inode_zone
);
1708 kmem_zone_destroy(xfs_efi_zone
);
1709 kmem_zone_destroy(xfs_efd_zone
);
1710 kmem_zone_destroy(xfs_buf_item_zone
);
1711 kmem_zone_destroy(xfs_log_item_desc_zone
);
1712 kmem_zone_destroy(xfs_trans_zone
);
1713 kmem_zone_destroy(xfs_ifork_zone
);
1714 kmem_zone_destroy(xfs_da_state_zone
);
1715 kmem_zone_destroy(xfs_btree_cur_zone
);
1716 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1717 kmem_zone_destroy(xfs_log_ticket_zone
);
1718 mempool_destroy(xfs_ioend_pool
);
1719 kmem_zone_destroy(xfs_ioend_zone
);
1724 xfs_init_workqueues(void)
1727 * The allocation workqueue can be used in memory reclaim situations
1728 * (writepage path), and parallelism is only limited by the number of
1729 * AGs in all the filesystems mounted. Hence use the default large
1730 * max_active value for this workqueue.
1732 xfs_alloc_wq
= alloc_workqueue("xfsalloc", WQ_MEM_RECLAIM
, 0);
1740 xfs_destroy_workqueues(void)
1742 destroy_workqueue(xfs_alloc_wq
);
1750 printk(KERN_INFO XFS_VERSION_STRING
" with "
1751 XFS_BUILD_OPTIONS
" enabled\n");
1755 error
= xfs_init_zones();
1759 error
= xfs_init_workqueues();
1761 goto out_destroy_zones
;
1763 error
= xfs_mru_cache_init();
1765 goto out_destroy_wq
;
1767 error
= xfs_filestream_init();
1769 goto out_mru_cache_uninit
;
1771 error
= xfs_buf_init();
1773 goto out_filestream_uninit
;
1775 error
= xfs_init_procfs();
1777 goto out_buf_terminate
;
1779 error
= xfs_sysctl_register();
1781 goto out_cleanup_procfs
;
1783 error
= xfs_qm_init();
1785 goto out_sysctl_unregister
;
1787 error
= register_filesystem(&xfs_fs_type
);
1794 out_sysctl_unregister
:
1795 xfs_sysctl_unregister();
1797 xfs_cleanup_procfs();
1799 xfs_buf_terminate();
1800 out_filestream_uninit
:
1801 xfs_filestream_uninit();
1802 out_mru_cache_uninit
:
1803 xfs_mru_cache_uninit();
1805 xfs_destroy_workqueues();
1807 xfs_destroy_zones();
1816 unregister_filesystem(&xfs_fs_type
);
1817 xfs_sysctl_unregister();
1818 xfs_cleanup_procfs();
1819 xfs_buf_terminate();
1820 xfs_filestream_uninit();
1821 xfs_mru_cache_uninit();
1822 xfs_destroy_workqueues();
1823 xfs_destroy_zones();
1826 module_init(init_xfs_fs
);
1827 module_exit(exit_xfs_fs
);
1829 MODULE_AUTHOR("Silicon Graphics, Inc.");
1830 MODULE_DESCRIPTION(XFS_VERSION_STRING
" with " XFS_BUILD_OPTIONS
" enabled");
1831 MODULE_LICENSE("GPL");