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
23 #include "xfs_trans.h"
27 #include "xfs_alloc.h"
28 #include "xfs_dmapi.h"
29 #include "xfs_quota.h"
30 #include "xfs_mount.h"
31 #include "xfs_bmap_btree.h"
32 #include "xfs_alloc_btree.h"
33 #include "xfs_ialloc_btree.h"
34 #include "xfs_dir2_sf.h"
35 #include "xfs_attr_sf.h"
36 #include "xfs_dinode.h"
37 #include "xfs_inode.h"
38 #include "xfs_btree.h"
39 #include "xfs_btree_trace.h"
40 #include "xfs_ialloc.h"
42 #include "xfs_rtalloc.h"
43 #include "xfs_error.h"
44 #include "xfs_itable.h"
45 #include "xfs_fsops.h"
48 #include "xfs_buf_item.h"
49 #include "xfs_utils.h"
50 #include "xfs_vnodeops.h"
51 #include "xfs_version.h"
52 #include "xfs_log_priv.h"
53 #include "xfs_trans_priv.h"
54 #include "xfs_filestream.h"
55 #include "xfs_da_btree.h"
56 #include "xfs_extfree_item.h"
57 #include "xfs_mru_cache.h"
58 #include "xfs_inode_item.h"
60 #include "xfs_trace.h"
62 #include <linux/namei.h>
63 #include <linux/init.h>
64 #include <linux/slab.h>
65 #include <linux/mount.h>
66 #include <linux/mempool.h>
67 #include <linux/writeback.h>
68 #include <linux/kthread.h>
69 #include <linux/freezer.h>
70 #include <linux/parser.h>
72 static const struct super_operations xfs_super_operations
;
73 static kmem_zone_t
*xfs_ioend_zone
;
74 mempool_t
*xfs_ioend_pool
;
76 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
77 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
78 #define MNTOPT_LOGDEV "logdev" /* log device */
79 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
80 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
81 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
82 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
83 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
84 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
85 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
86 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
87 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
88 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
89 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
90 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
91 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
92 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
93 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
94 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
95 * unwritten extent conversion */
96 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
97 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
98 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
99 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
100 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
101 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
102 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
104 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
105 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
106 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
107 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
108 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
109 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
110 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
111 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
112 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
113 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
114 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
115 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
116 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
117 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
118 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
119 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
120 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
121 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
124 * Table driven mount option parser.
126 * Currently only used for remount, but it will be used for mount
127 * in the future, too.
130 Opt_barrier
, Opt_nobarrier
, Opt_err
133 static const match_table_t tokens
= {
134 {Opt_barrier
, "barrier"},
135 {Opt_nobarrier
, "nobarrier"},
141 suffix_strtoul(char *s
, char **endp
, unsigned int base
)
143 int last
, shift_left_factor
= 0;
146 last
= strlen(value
) - 1;
147 if (value
[last
] == 'K' || value
[last
] == 'k') {
148 shift_left_factor
= 10;
151 if (value
[last
] == 'M' || value
[last
] == 'm') {
152 shift_left_factor
= 20;
155 if (value
[last
] == 'G' || value
[last
] == 'g') {
156 shift_left_factor
= 30;
160 return simple_strtoul((const char *)s
, endp
, base
) << shift_left_factor
;
164 * This function fills in xfs_mount_t fields based on mount args.
165 * Note: the superblock has _not_ yet been read in.
167 * Note that this function leaks the various device name allocations on
168 * failure. The caller takes care of them.
172 struct xfs_mount
*mp
,
176 struct super_block
*sb
= mp
->m_super
;
177 char *this_char
, *value
, *eov
;
181 int dmapi_implies_ikeep
= 1;
182 __uint8_t iosizelog
= 0;
185 * Copy binary VFS mount flags we are interested in.
187 if (sb
->s_flags
& MS_RDONLY
)
188 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
189 if (sb
->s_flags
& MS_DIRSYNC
)
190 mp
->m_flags
|= XFS_MOUNT_DIRSYNC
;
191 if (sb
->s_flags
& MS_SYNCHRONOUS
)
192 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
195 * Set some default flags that could be cleared by the mount option
198 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
199 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
200 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
203 * These can be overridden by the mount option parsing.
211 while ((this_char
= strsep(&options
, ",")) != NULL
) {
214 if ((value
= strchr(this_char
, '=')) != NULL
)
217 if (!strcmp(this_char
, MNTOPT_LOGBUFS
)) {
218 if (!value
|| !*value
) {
220 "XFS: %s option requires an argument",
224 mp
->m_logbufs
= simple_strtoul(value
, &eov
, 10);
225 } else if (!strcmp(this_char
, MNTOPT_LOGBSIZE
)) {
226 if (!value
|| !*value
) {
228 "XFS: %s option requires an argument",
232 mp
->m_logbsize
= suffix_strtoul(value
, &eov
, 10);
233 } else if (!strcmp(this_char
, MNTOPT_LOGDEV
)) {
234 if (!value
|| !*value
) {
236 "XFS: %s option requires an argument",
240 mp
->m_logname
= kstrndup(value
, MAXNAMELEN
, GFP_KERNEL
);
243 } else if (!strcmp(this_char
, MNTOPT_MTPT
)) {
244 if (!value
|| !*value
) {
246 "XFS: %s option requires an argument",
250 *mtpt
= kstrndup(value
, MAXNAMELEN
, GFP_KERNEL
);
253 } else if (!strcmp(this_char
, MNTOPT_RTDEV
)) {
254 if (!value
|| !*value
) {
256 "XFS: %s option requires an argument",
260 mp
->m_rtname
= kstrndup(value
, MAXNAMELEN
, GFP_KERNEL
);
263 } else if (!strcmp(this_char
, MNTOPT_BIOSIZE
)) {
264 if (!value
|| !*value
) {
266 "XFS: %s option requires an argument",
270 iosize
= simple_strtoul(value
, &eov
, 10);
271 iosizelog
= ffs(iosize
) - 1;
272 } else if (!strcmp(this_char
, MNTOPT_ALLOCSIZE
)) {
273 if (!value
|| !*value
) {
275 "XFS: %s option requires an argument",
279 iosize
= suffix_strtoul(value
, &eov
, 10);
280 iosizelog
= ffs(iosize
) - 1;
281 } else if (!strcmp(this_char
, MNTOPT_GRPID
) ||
282 !strcmp(this_char
, MNTOPT_BSDGROUPS
)) {
283 mp
->m_flags
|= XFS_MOUNT_GRPID
;
284 } else if (!strcmp(this_char
, MNTOPT_NOGRPID
) ||
285 !strcmp(this_char
, MNTOPT_SYSVGROUPS
)) {
286 mp
->m_flags
&= ~XFS_MOUNT_GRPID
;
287 } else if (!strcmp(this_char
, MNTOPT_WSYNC
)) {
288 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
289 } else if (!strcmp(this_char
, MNTOPT_OSYNCISOSYNC
)) {
290 mp
->m_flags
|= XFS_MOUNT_OSYNCISOSYNC
;
291 } else if (!strcmp(this_char
, MNTOPT_NORECOVERY
)) {
292 mp
->m_flags
|= XFS_MOUNT_NORECOVERY
;
293 } else if (!strcmp(this_char
, MNTOPT_NOALIGN
)) {
294 mp
->m_flags
|= XFS_MOUNT_NOALIGN
;
295 } else if (!strcmp(this_char
, MNTOPT_SWALLOC
)) {
296 mp
->m_flags
|= XFS_MOUNT_SWALLOC
;
297 } else if (!strcmp(this_char
, MNTOPT_SUNIT
)) {
298 if (!value
|| !*value
) {
300 "XFS: %s option requires an argument",
304 dsunit
= simple_strtoul(value
, &eov
, 10);
305 } else if (!strcmp(this_char
, MNTOPT_SWIDTH
)) {
306 if (!value
|| !*value
) {
308 "XFS: %s option requires an argument",
312 dswidth
= simple_strtoul(value
, &eov
, 10);
313 } else if (!strcmp(this_char
, MNTOPT_64BITINODE
)) {
314 mp
->m_flags
&= ~XFS_MOUNT_SMALL_INUMS
;
317 "XFS: %s option not allowed on this system",
321 } else if (!strcmp(this_char
, MNTOPT_NOUUID
)) {
322 mp
->m_flags
|= XFS_MOUNT_NOUUID
;
323 } else if (!strcmp(this_char
, MNTOPT_BARRIER
)) {
324 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
325 } else if (!strcmp(this_char
, MNTOPT_NOBARRIER
)) {
326 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
327 } else if (!strcmp(this_char
, MNTOPT_IKEEP
)) {
328 mp
->m_flags
|= XFS_MOUNT_IKEEP
;
329 } else if (!strcmp(this_char
, MNTOPT_NOIKEEP
)) {
330 dmapi_implies_ikeep
= 0;
331 mp
->m_flags
&= ~XFS_MOUNT_IKEEP
;
332 } else if (!strcmp(this_char
, MNTOPT_LARGEIO
)) {
333 mp
->m_flags
&= ~XFS_MOUNT_COMPAT_IOSIZE
;
334 } else if (!strcmp(this_char
, MNTOPT_NOLARGEIO
)) {
335 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
336 } else if (!strcmp(this_char
, MNTOPT_ATTR2
)) {
337 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
338 } else if (!strcmp(this_char
, MNTOPT_NOATTR2
)) {
339 mp
->m_flags
&= ~XFS_MOUNT_ATTR2
;
340 mp
->m_flags
|= XFS_MOUNT_NOATTR2
;
341 } else if (!strcmp(this_char
, MNTOPT_FILESTREAM
)) {
342 mp
->m_flags
|= XFS_MOUNT_FILESTREAMS
;
343 } else if (!strcmp(this_char
, MNTOPT_NOQUOTA
)) {
344 mp
->m_qflags
&= ~(XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
|
345 XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
|
346 XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
|
347 XFS_UQUOTA_ENFD
| XFS_OQUOTA_ENFD
);
348 } else if (!strcmp(this_char
, MNTOPT_QUOTA
) ||
349 !strcmp(this_char
, MNTOPT_UQUOTA
) ||
350 !strcmp(this_char
, MNTOPT_USRQUOTA
)) {
351 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
|
353 } else if (!strcmp(this_char
, MNTOPT_QUOTANOENF
) ||
354 !strcmp(this_char
, MNTOPT_UQUOTANOENF
)) {
355 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
);
356 mp
->m_qflags
&= ~XFS_UQUOTA_ENFD
;
357 } else if (!strcmp(this_char
, MNTOPT_PQUOTA
) ||
358 !strcmp(this_char
, MNTOPT_PRJQUOTA
)) {
359 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
|
361 } else if (!strcmp(this_char
, MNTOPT_PQUOTANOENF
)) {
362 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
);
363 mp
->m_qflags
&= ~XFS_OQUOTA_ENFD
;
364 } else if (!strcmp(this_char
, MNTOPT_GQUOTA
) ||
365 !strcmp(this_char
, MNTOPT_GRPQUOTA
)) {
366 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
|
368 } else if (!strcmp(this_char
, MNTOPT_GQUOTANOENF
)) {
369 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
);
370 mp
->m_qflags
&= ~XFS_OQUOTA_ENFD
;
371 } else if (!strcmp(this_char
, MNTOPT_DMAPI
)) {
372 mp
->m_flags
|= XFS_MOUNT_DMAPI
;
373 } else if (!strcmp(this_char
, MNTOPT_XDSM
)) {
374 mp
->m_flags
|= XFS_MOUNT_DMAPI
;
375 } else if (!strcmp(this_char
, MNTOPT_DMI
)) {
376 mp
->m_flags
|= XFS_MOUNT_DMAPI
;
377 } else if (!strcmp(this_char
, "ihashsize")) {
379 "XFS: ihashsize no longer used, option is deprecated.");
380 } else if (!strcmp(this_char
, "osyncisdsync")) {
381 /* no-op, this is now the default */
383 "XFS: osyncisdsync is now the default, option is deprecated.");
384 } else if (!strcmp(this_char
, "irixsgid")) {
386 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
389 "XFS: unknown mount option [%s].", this_char
);
395 * no recovery flag requires a read-only mount
397 if ((mp
->m_flags
& XFS_MOUNT_NORECOVERY
) &&
398 !(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
399 cmn_err(CE_WARN
, "XFS: no-recovery mounts must be read-only.");
403 if ((mp
->m_flags
& XFS_MOUNT_NOALIGN
) && (dsunit
|| dswidth
)) {
405 "XFS: sunit and swidth options incompatible with the noalign option");
409 #ifndef CONFIG_XFS_QUOTA
410 if (XFS_IS_QUOTA_RUNNING(mp
)) {
412 "XFS: quota support not available in this kernel.");
417 if ((mp
->m_qflags
& (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
)) &&
418 (mp
->m_qflags
& (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
))) {
420 "XFS: cannot mount with both project and group quota");
424 if ((mp
->m_flags
& XFS_MOUNT_DMAPI
) && (!*mtpt
|| *mtpt
[0] == '\0')) {
425 printk("XFS: %s option needs the mount point option as well\n",
430 if ((dsunit
&& !dswidth
) || (!dsunit
&& dswidth
)) {
432 "XFS: sunit and swidth must be specified together");
436 if (dsunit
&& (dswidth
% dsunit
!= 0)) {
438 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
444 * Applications using DMI filesystems often expect the
445 * inode generation number to be monotonically increasing.
446 * If we delete inode chunks we break this assumption, so
447 * keep unused inode chunks on disk for DMI filesystems
448 * until we come up with a better solution.
449 * Note that if "ikeep" or "noikeep" mount options are
450 * supplied, then they are honored.
452 if ((mp
->m_flags
& XFS_MOUNT_DMAPI
) && dmapi_implies_ikeep
)
453 mp
->m_flags
|= XFS_MOUNT_IKEEP
;
456 if (!(mp
->m_flags
& XFS_MOUNT_NOALIGN
)) {
458 * At this point the superblock has not been read
459 * in, therefore we do not know the block size.
460 * Before the mount call ends we will convert
464 mp
->m_dalign
= dsunit
;
465 mp
->m_flags
|= XFS_MOUNT_RETERR
;
469 mp
->m_swidth
= dswidth
;
472 if (mp
->m_logbufs
!= -1 &&
473 mp
->m_logbufs
!= 0 &&
474 (mp
->m_logbufs
< XLOG_MIN_ICLOGS
||
475 mp
->m_logbufs
> XLOG_MAX_ICLOGS
)) {
477 "XFS: invalid logbufs value: %d [not %d-%d]",
478 mp
->m_logbufs
, XLOG_MIN_ICLOGS
, XLOG_MAX_ICLOGS
);
479 return XFS_ERROR(EINVAL
);
481 if (mp
->m_logbsize
!= -1 &&
482 mp
->m_logbsize
!= 0 &&
483 (mp
->m_logbsize
< XLOG_MIN_RECORD_BSIZE
||
484 mp
->m_logbsize
> XLOG_MAX_RECORD_BSIZE
||
485 !is_power_of_2(mp
->m_logbsize
))) {
487 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
489 return XFS_ERROR(EINVAL
);
492 mp
->m_fsname
= kstrndup(sb
->s_id
, MAXNAMELEN
, GFP_KERNEL
);
495 mp
->m_fsname_len
= strlen(mp
->m_fsname
) + 1;
498 if (iosizelog
> XFS_MAX_IO_LOG
||
499 iosizelog
< XFS_MIN_IO_LOG
) {
501 "XFS: invalid log iosize: %d [not %d-%d]",
502 iosizelog
, XFS_MIN_IO_LOG
,
504 return XFS_ERROR(EINVAL
);
507 mp
->m_flags
|= XFS_MOUNT_DFLT_IOSIZE
;
508 mp
->m_readio_log
= iosizelog
;
509 mp
->m_writeio_log
= iosizelog
;
515 struct proc_xfs_info
{
522 struct xfs_mount
*mp
,
525 static struct proc_xfs_info xfs_info_set
[] = {
526 /* the few simple ones we can get from the mount struct */
527 { XFS_MOUNT_IKEEP
, "," MNTOPT_IKEEP
},
528 { XFS_MOUNT_WSYNC
, "," MNTOPT_WSYNC
},
529 { XFS_MOUNT_NOALIGN
, "," MNTOPT_NOALIGN
},
530 { XFS_MOUNT_SWALLOC
, "," MNTOPT_SWALLOC
},
531 { XFS_MOUNT_NOUUID
, "," MNTOPT_NOUUID
},
532 { XFS_MOUNT_NORECOVERY
, "," MNTOPT_NORECOVERY
},
533 { XFS_MOUNT_OSYNCISOSYNC
, "," MNTOPT_OSYNCISOSYNC
},
534 { XFS_MOUNT_ATTR2
, "," MNTOPT_ATTR2
},
535 { XFS_MOUNT_FILESTREAMS
, "," MNTOPT_FILESTREAM
},
536 { XFS_MOUNT_DMAPI
, "," MNTOPT_DMAPI
},
537 { XFS_MOUNT_GRPID
, "," MNTOPT_GRPID
},
540 static struct proc_xfs_info xfs_info_unset
[] = {
541 /* the few simple ones we can get from the mount struct */
542 { XFS_MOUNT_COMPAT_IOSIZE
, "," MNTOPT_LARGEIO
},
543 { XFS_MOUNT_BARRIER
, "," MNTOPT_NOBARRIER
},
544 { XFS_MOUNT_SMALL_INUMS
, "," MNTOPT_64BITINODE
},
547 struct proc_xfs_info
*xfs_infop
;
549 for (xfs_infop
= xfs_info_set
; xfs_infop
->flag
; xfs_infop
++) {
550 if (mp
->m_flags
& xfs_infop
->flag
)
551 seq_puts(m
, xfs_infop
->str
);
553 for (xfs_infop
= xfs_info_unset
; xfs_infop
->flag
; xfs_infop
++) {
554 if (!(mp
->m_flags
& xfs_infop
->flag
))
555 seq_puts(m
, xfs_infop
->str
);
558 if (mp
->m_flags
& XFS_MOUNT_DFLT_IOSIZE
)
559 seq_printf(m
, "," MNTOPT_ALLOCSIZE
"=%dk",
560 (int)(1 << mp
->m_writeio_log
) >> 10);
562 if (mp
->m_logbufs
> 0)
563 seq_printf(m
, "," MNTOPT_LOGBUFS
"=%d", mp
->m_logbufs
);
564 if (mp
->m_logbsize
> 0)
565 seq_printf(m
, "," MNTOPT_LOGBSIZE
"=%dk", mp
->m_logbsize
>> 10);
568 seq_printf(m
, "," MNTOPT_LOGDEV
"=%s", mp
->m_logname
);
570 seq_printf(m
, "," MNTOPT_RTDEV
"=%s", mp
->m_rtname
);
572 if (mp
->m_dalign
> 0)
573 seq_printf(m
, "," MNTOPT_SUNIT
"=%d",
574 (int)XFS_FSB_TO_BB(mp
, mp
->m_dalign
));
575 if (mp
->m_swidth
> 0)
576 seq_printf(m
, "," MNTOPT_SWIDTH
"=%d",
577 (int)XFS_FSB_TO_BB(mp
, mp
->m_swidth
));
579 if (mp
->m_qflags
& (XFS_UQUOTA_ACCT
|XFS_UQUOTA_ENFD
))
580 seq_puts(m
, "," MNTOPT_USRQUOTA
);
581 else if (mp
->m_qflags
& XFS_UQUOTA_ACCT
)
582 seq_puts(m
, "," MNTOPT_UQUOTANOENF
);
584 /* Either project or group quotas can be active, not both */
586 if (mp
->m_qflags
& XFS_PQUOTA_ACCT
) {
587 if (mp
->m_qflags
& XFS_OQUOTA_ENFD
)
588 seq_puts(m
, "," MNTOPT_PRJQUOTA
);
590 seq_puts(m
, "," MNTOPT_PQUOTANOENF
);
591 } else if (mp
->m_qflags
& XFS_GQUOTA_ACCT
) {
592 if (mp
->m_qflags
& XFS_OQUOTA_ENFD
)
593 seq_puts(m
, "," MNTOPT_GRPQUOTA
);
595 seq_puts(m
, "," MNTOPT_GQUOTANOENF
);
598 if (!(mp
->m_qflags
& XFS_ALL_QUOTA_ACCT
))
599 seq_puts(m
, "," MNTOPT_NOQUOTA
);
605 unsigned int blockshift
)
607 unsigned int pagefactor
= 1;
608 unsigned int bitshift
= BITS_PER_LONG
- 1;
610 /* Figure out maximum filesize, on Linux this can depend on
611 * the filesystem blocksize (on 32 bit platforms).
612 * __block_prepare_write does this in an [unsigned] long...
613 * page->index << (PAGE_CACHE_SHIFT - bbits)
614 * So, for page sized blocks (4K on 32 bit platforms),
615 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
616 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
617 * but for smaller blocksizes it is less (bbits = log2 bsize).
618 * Note1: get_block_t takes a long (implicit cast from above)
619 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
620 * can optionally convert the [unsigned] long from above into
621 * an [unsigned] long long.
624 #if BITS_PER_LONG == 32
625 # if defined(CONFIG_LBDAF)
626 ASSERT(sizeof(sector_t
) == 8);
627 pagefactor
= PAGE_CACHE_SIZE
;
628 bitshift
= BITS_PER_LONG
;
630 pagefactor
= PAGE_CACHE_SIZE
>> (PAGE_CACHE_SHIFT
- blockshift
);
634 return (((__uint64_t
)pagefactor
) << bitshift
) - 1;
641 struct block_device
**bdevp
)
645 *bdevp
= open_bdev_exclusive(name
, FMODE_READ
|FMODE_WRITE
, mp
);
646 if (IS_ERR(*bdevp
)) {
647 error
= PTR_ERR(*bdevp
);
648 printk("XFS: Invalid device [%s], error=%d\n", name
, error
);
656 struct block_device
*bdev
)
659 close_bdev_exclusive(bdev
, FMODE_READ
|FMODE_WRITE
);
663 * Try to write out the superblock using barriers.
669 xfs_buf_t
*sbp
= xfs_getsb(mp
, 0);
674 XFS_BUF_UNDELAYWRITE(sbp
);
676 XFS_BUF_UNASYNC(sbp
);
677 XFS_BUF_ORDERED(sbp
);
680 error
= xfs_iowait(sbp
);
683 * Clear all the flags we set and possible error state in the
684 * buffer. We only did the write to try out whether barriers
685 * worked and shouldn't leave any traces in the superblock
689 XFS_BUF_ERROR(sbp
, 0);
690 XFS_BUF_UNORDERED(sbp
);
697 xfs_mountfs_check_barriers(xfs_mount_t
*mp
)
701 if (mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
702 xfs_fs_cmn_err(CE_NOTE
, mp
,
703 "Disabling barriers, not supported with external log device");
704 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
708 if (xfs_readonly_buftarg(mp
->m_ddev_targp
)) {
709 xfs_fs_cmn_err(CE_NOTE
, mp
,
710 "Disabling barriers, underlying device is readonly");
711 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
715 error
= xfs_barrier_test(mp
);
717 xfs_fs_cmn_err(CE_NOTE
, mp
,
718 "Disabling barriers, trial barrier write failed");
719 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
725 xfs_blkdev_issue_flush(
726 xfs_buftarg_t
*buftarg
)
728 blkdev_issue_flush(buftarg
->bt_bdev
, GFP_KERNEL
, NULL
,
734 struct xfs_mount
*mp
)
736 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
737 struct block_device
*logdev
= mp
->m_logdev_targp
->bt_bdev
;
738 xfs_free_buftarg(mp
, mp
->m_logdev_targp
);
739 xfs_blkdev_put(logdev
);
741 if (mp
->m_rtdev_targp
) {
742 struct block_device
*rtdev
= mp
->m_rtdev_targp
->bt_bdev
;
743 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
744 xfs_blkdev_put(rtdev
);
746 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
750 * The file system configurations are:
751 * (1) device (partition) with data and internal log
752 * (2) logical volume with data and log subvolumes.
753 * (3) logical volume with data, log, and realtime subvolumes.
755 * We only have to handle opening the log and realtime volumes here if
756 * they are present. The data subvolume has already been opened by
757 * get_sb_bdev() and is stored in sb->s_bdev.
761 struct xfs_mount
*mp
)
763 struct block_device
*ddev
= mp
->m_super
->s_bdev
;
764 struct block_device
*logdev
= NULL
, *rtdev
= NULL
;
768 * Open real time and log devices - order is important.
771 error
= xfs_blkdev_get(mp
, mp
->m_logname
, &logdev
);
777 error
= xfs_blkdev_get(mp
, mp
->m_rtname
, &rtdev
);
779 goto out_close_logdev
;
781 if (rtdev
== ddev
|| rtdev
== logdev
) {
783 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
785 goto out_close_rtdev
;
790 * Setup xfs_mount buffer target pointers
793 mp
->m_ddev_targp
= xfs_alloc_buftarg(ddev
, 0, mp
->m_fsname
);
794 if (!mp
->m_ddev_targp
)
795 goto out_close_rtdev
;
798 mp
->m_rtdev_targp
= xfs_alloc_buftarg(rtdev
, 1, mp
->m_fsname
);
799 if (!mp
->m_rtdev_targp
)
800 goto out_free_ddev_targ
;
803 if (logdev
&& logdev
!= ddev
) {
804 mp
->m_logdev_targp
= xfs_alloc_buftarg(logdev
, 1, mp
->m_fsname
);
805 if (!mp
->m_logdev_targp
)
806 goto out_free_rtdev_targ
;
808 mp
->m_logdev_targp
= mp
->m_ddev_targp
;
814 if (mp
->m_rtdev_targp
)
815 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
817 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
820 xfs_blkdev_put(rtdev
);
822 if (logdev
&& logdev
!= ddev
)
823 xfs_blkdev_put(logdev
);
829 * Setup xfs_mount buffer target pointers based on superblock
833 struct xfs_mount
*mp
)
837 error
= xfs_setsize_buftarg(mp
->m_ddev_targp
, mp
->m_sb
.sb_blocksize
,
838 mp
->m_sb
.sb_sectsize
);
842 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
843 unsigned int log_sector_size
= BBSIZE
;
845 if (xfs_sb_version_hassector(&mp
->m_sb
))
846 log_sector_size
= mp
->m_sb
.sb_logsectsize
;
847 error
= xfs_setsize_buftarg(mp
->m_logdev_targp
,
848 mp
->m_sb
.sb_blocksize
,
853 if (mp
->m_rtdev_targp
) {
854 error
= xfs_setsize_buftarg(mp
->m_rtdev_targp
,
855 mp
->m_sb
.sb_blocksize
,
856 mp
->m_sb
.sb_sectsize
);
865 * XFS AIL push thread support
869 struct xfs_ail
*ailp
,
870 xfs_lsn_t threshold_lsn
)
872 ailp
->xa_target
= threshold_lsn
;
873 wake_up_process(ailp
->xa_task
);
880 struct xfs_ail
*ailp
= data
;
881 xfs_lsn_t last_pushed_lsn
= 0;
882 long tout
= 0; /* milliseconds */
884 while (!kthread_should_stop()) {
885 schedule_timeout_interruptible(tout
?
886 msecs_to_jiffies(tout
) : MAX_SCHEDULE_TIMEOUT
);
891 ASSERT(ailp
->xa_mount
->m_log
);
892 if (XFS_FORCED_SHUTDOWN(ailp
->xa_mount
))
895 tout
= xfsaild_push(ailp
, &last_pushed_lsn
);
903 struct xfs_ail
*ailp
)
906 ailp
->xa_task
= kthread_run(xfsaild
, ailp
, "xfsaild/%s",
907 ailp
->xa_mount
->m_fsname
);
908 if (IS_ERR(ailp
->xa_task
))
909 return -PTR_ERR(ailp
->xa_task
);
915 struct xfs_ail
*ailp
)
917 kthread_stop(ailp
->xa_task
);
921 /* Catch misguided souls that try to use this interface on XFS */
922 STATIC
struct inode
*
924 struct super_block
*sb
)
931 * Now that the generic code is guaranteed not to be accessing
932 * the linux inode, we can reclaim the inode.
935 xfs_fs_destroy_inode(
938 struct xfs_inode
*ip
= XFS_I(inode
);
940 xfs_itrace_entry(ip
);
942 XFS_STATS_INC(vn_reclaim
);
944 /* bad inode, get out here ASAP */
945 if (is_bad_inode(inode
))
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.
966 xfs_inode_set_reclaim_tag(ip
);
970 * Slab object creation initialisation for the XFS inode.
971 * This covers only the idempotent fields in the XFS inode;
972 * all other fields need to be initialised on allocation
973 * from the slab. This avoids the need to repeatedly intialise
974 * fields in the xfs inode that left in the initialise state
975 * when freeing the inode.
978 xfs_fs_inode_init_once(
981 struct xfs_inode
*ip
= inode
;
983 memset(ip
, 0, sizeof(struct xfs_inode
));
986 inode_init_once(VFS_I(ip
));
989 atomic_set(&ip
->i_iocount
, 0);
990 atomic_set(&ip
->i_pincount
, 0);
991 spin_lock_init(&ip
->i_flags_lock
);
992 init_waitqueue_head(&ip
->i_ipin_wait
);
994 * Because we want to use a counting completion, complete
995 * the flush completion once to allow a single access to
996 * the flush completion without blocking.
998 init_completion(&ip
->i_flush
);
999 complete(&ip
->i_flush
);
1001 mrlock_init(&ip
->i_lock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
1002 "xfsino", ip
->i_ino
);
1006 * Dirty the XFS inode when mark_inode_dirty_sync() is called so that
1007 * we catch unlogged VFS level updates to the inode. Care must be taken
1008 * here - the transaction code calls mark_inode_dirty_sync() to mark the
1009 * VFS inode dirty in a transaction and clears the i_update_core field;
1010 * it must clear the field after calling mark_inode_dirty_sync() to
1011 * correctly indicate that the dirty state has been propagated into the
1014 * We need the barrier() to maintain correct ordering between unlogged
1015 * updates and the transaction commit code that clears the i_update_core
1016 * field. This requires all updates to be completed before marking the
1021 struct inode
*inode
)
1024 XFS_I(inode
)->i_update_core
= 1;
1029 struct xfs_inode
*ip
)
1031 struct xfs_mount
*mp
= ip
->i_mount
;
1032 struct xfs_trans
*tp
;
1035 xfs_iunlock(ip
, XFS_ILOCK_SHARED
);
1036 tp
= xfs_trans_alloc(mp
, XFS_TRANS_FSYNC_TS
);
1037 error
= xfs_trans_reserve(tp
, 0, XFS_FSYNC_TS_LOG_RES(mp
), 0, 0, 0);
1040 xfs_trans_cancel(tp
, 0);
1041 /* we need to return with the lock hold shared */
1042 xfs_ilock(ip
, XFS_ILOCK_SHARED
);
1046 xfs_ilock(ip
, XFS_ILOCK_EXCL
);
1049 * Note - it's possible that we might have pushed ourselves out of the
1050 * way during trans_reserve which would flush the inode. But there's
1051 * no guarantee that the inode buffer has actually gone out yet (it's
1052 * delwri). Plus the buffer could be pinned anyway if it's part of
1053 * an inode in another recent transaction. So we play it safe and
1054 * fire off the transaction anyway.
1056 xfs_trans_ijoin(tp
, ip
, XFS_ILOCK_EXCL
);
1057 xfs_trans_ihold(tp
, ip
);
1058 xfs_trans_log_inode(tp
, ip
, XFS_ILOG_CORE
);
1059 xfs_trans_set_sync(tp
);
1060 error
= xfs_trans_commit(tp
, 0);
1061 xfs_ilock_demote(ip
, XFS_ILOCK_EXCL
);
1068 struct inode
*inode
,
1069 struct writeback_control
*wbc
)
1071 struct xfs_inode
*ip
= XFS_I(inode
);
1072 struct xfs_mount
*mp
= ip
->i_mount
;
1075 xfs_itrace_entry(ip
);
1077 if (XFS_FORCED_SHUTDOWN(mp
))
1078 return XFS_ERROR(EIO
);
1080 if (wbc
->sync_mode
== WB_SYNC_ALL
) {
1082 * Make sure the inode has hit stable storage. By using the
1083 * log and the fsync transactions we reduce the IOs we have
1084 * to do here from two (log and inode) to just the log.
1086 * Note: We still need to do a delwri write of the inode after
1087 * this to flush it to the backing buffer so that bulkstat
1088 * works properly if this is the first time the inode has been
1089 * written. Because we hold the ilock atomically over the
1090 * transaction commit and the inode flush we are guaranteed
1091 * that the inode is not pinned when it returns. If the flush
1092 * lock is already held, then the inode has already been
1093 * flushed once and we don't need to flush it again. Hence
1094 * the code will only flush the inode if it isn't already
1098 xfs_ilock(ip
, XFS_ILOCK_SHARED
);
1099 if (ip
->i_update_core
) {
1100 error
= xfs_log_inode(ip
);
1106 * We make this non-blocking if the inode is contended, return
1107 * EAGAIN to indicate to the caller that they did not succeed.
1108 * This prevents the flush path from blocking on inodes inside
1109 * another operation right now, they get caught later by xfs_sync.
1111 if (!xfs_ilock_nowait(ip
, XFS_ILOCK_SHARED
))
1115 if (xfs_ipincount(ip
) || !xfs_iflock_nowait(ip
))
1119 * Now we have the flush lock and the inode is not pinned, we can check
1120 * if the inode is really clean as we know that there are no pending
1121 * transaction completions, it is not waiting on the delayed write
1122 * queue and there is no IO in progress.
1124 if (xfs_inode_clean(ip
)) {
1129 error
= xfs_iflush(ip
, 0);
1132 xfs_iunlock(ip
, XFS_ILOCK_SHARED
);
1135 * if we failed to write out the inode then mark
1136 * it dirty again so we'll try again later.
1139 xfs_mark_inode_dirty_sync(ip
);
1145 struct inode
*inode
)
1147 xfs_inode_t
*ip
= XFS_I(inode
);
1149 xfs_itrace_entry(ip
);
1150 XFS_STATS_INC(vn_rele
);
1151 XFS_STATS_INC(vn_remove
);
1152 XFS_STATS_DEC(vn_active
);
1155 * The iolock is used by the file system to coordinate reads,
1156 * writes, and block truncates. Up to this point the lock
1157 * protected concurrent accesses by users of the inode. But
1158 * from here forward we're doing some final processing of the
1159 * inode because we're done with it, and although we reuse the
1160 * iolock for protection it is really a distinct lock class
1161 * (in the lockdep sense) from before. To keep lockdep happy
1162 * (and basically indicate what we are doing), we explicitly
1163 * re-init the iolock here.
1165 ASSERT(!rwsem_is_locked(&ip
->i_iolock
.mr_lock
));
1166 mrlock_init(&ip
->i_iolock
, MRLOCK_BARRIER
, "xfsio", ip
->i_ino
);
1173 struct xfs_mount
*mp
)
1175 kfree(mp
->m_fsname
);
1176 kfree(mp
->m_rtname
);
1177 kfree(mp
->m_logname
);
1182 struct super_block
*sb
)
1184 struct xfs_mount
*mp
= XFS_M(sb
);
1188 if (!(sb
->s_flags
& MS_RDONLY
)) {
1190 * XXX(hch): this should be SYNC_WAIT.
1192 * Or more likely not needed at all because the VFS is already
1193 * calling ->sync_fs after shutting down all filestem
1194 * operations and just before calling ->put_super.
1196 xfs_sync_data(mp
, 0);
1197 xfs_sync_attr(mp
, 0);
1200 XFS_SEND_PREUNMOUNT(mp
);
1203 * Blow away any referenced inode in the filestreams cache.
1204 * This can and will cause log traffic as inodes go inactive
1207 xfs_filestream_unmount(mp
);
1209 XFS_bflush(mp
->m_ddev_targp
);
1211 XFS_SEND_UNMOUNT(mp
);
1215 xfs_inode_shrinker_unregister(mp
);
1216 xfs_icsb_destroy_counters(mp
);
1217 xfs_close_devices(mp
);
1219 xfs_free_fsname(mp
);
1225 struct super_block
*sb
,
1228 struct xfs_mount
*mp
= XFS_M(sb
);
1232 * Not much we can do for the first async pass. Writing out the
1233 * superblock would be counter-productive as we are going to redirty
1234 * when writing out other data and metadata (and writing out a single
1235 * block is quite fast anyway).
1237 * Try to asynchronously kick off quota syncing at least.
1240 xfs_qm_sync(mp
, SYNC_TRYLOCK
);
1244 error
= xfs_quiesce_data(mp
);
1249 int prev_sync_seq
= mp
->m_sync_seq
;
1252 * The disk must be active because we're syncing.
1253 * We schedule xfssyncd now (now that the disk is
1254 * active) instead of later (when it might not be).
1256 wake_up_process(mp
->m_sync_task
);
1258 * We have to wait for the sync iteration to complete.
1259 * If we don't, the disk activity caused by the sync
1260 * will come after the sync is completed, and that
1261 * triggers another sync from laptop mode.
1263 wait_event(mp
->m_wait_single_sync_task
,
1264 mp
->m_sync_seq
!= prev_sync_seq
);
1272 struct dentry
*dentry
,
1273 struct kstatfs
*statp
)
1275 struct xfs_mount
*mp
= XFS_M(dentry
->d_sb
);
1276 xfs_sb_t
*sbp
= &mp
->m_sb
;
1277 struct xfs_inode
*ip
= XFS_I(dentry
->d_inode
);
1278 __uint64_t fakeinos
, id
;
1281 statp
->f_type
= XFS_SB_MAGIC
;
1282 statp
->f_namelen
= MAXNAMELEN
- 1;
1284 id
= huge_encode_dev(mp
->m_ddev_targp
->bt_dev
);
1285 statp
->f_fsid
.val
[0] = (u32
)id
;
1286 statp
->f_fsid
.val
[1] = (u32
)(id
>> 32);
1288 xfs_icsb_sync_counters(mp
, XFS_ICSB_LAZY_COUNT
);
1290 spin_lock(&mp
->m_sb_lock
);
1291 statp
->f_bsize
= sbp
->sb_blocksize
;
1292 lsize
= sbp
->sb_logstart
? sbp
->sb_logblocks
: 0;
1293 statp
->f_blocks
= sbp
->sb_dblocks
- lsize
;
1294 statp
->f_bfree
= statp
->f_bavail
=
1295 sbp
->sb_fdblocks
- XFS_ALLOC_SET_ASIDE(mp
);
1296 fakeinos
= statp
->f_bfree
<< sbp
->sb_inopblog
;
1298 MIN(sbp
->sb_icount
+ fakeinos
, (__uint64_t
)XFS_MAXINUMBER
);
1299 if (mp
->m_maxicount
)
1300 statp
->f_files
= min_t(typeof(statp
->f_files
),
1303 statp
->f_ffree
= statp
->f_files
- (sbp
->sb_icount
- sbp
->sb_ifree
);
1304 spin_unlock(&mp
->m_sb_lock
);
1306 if ((ip
->i_d
.di_flags
& XFS_DIFLAG_PROJINHERIT
) ||
1307 ((mp
->m_qflags
& (XFS_PQUOTA_ACCT
|XFS_OQUOTA_ENFD
))) ==
1308 (XFS_PQUOTA_ACCT
|XFS_OQUOTA_ENFD
))
1309 xfs_qm_statvfs(ip
, statp
);
1314 xfs_save_resvblks(struct xfs_mount
*mp
)
1316 __uint64_t resblks
= 0;
1318 mp
->m_resblks_save
= mp
->m_resblks
;
1319 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1323 xfs_restore_resvblks(struct xfs_mount
*mp
)
1327 if (mp
->m_resblks_save
) {
1328 resblks
= mp
->m_resblks_save
;
1329 mp
->m_resblks_save
= 0;
1331 resblks
= xfs_default_resblks(mp
);
1333 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1338 struct super_block
*sb
,
1342 struct xfs_mount
*mp
= XFS_M(sb
);
1343 substring_t args
[MAX_OPT_ARGS
];
1347 while ((p
= strsep(&options
, ",")) != NULL
) {
1353 token
= match_token(p
, tokens
, args
);
1356 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
1359 * Test if barriers are actually working if we can,
1360 * else delay this check until the filesystem is
1363 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
))
1364 xfs_mountfs_check_barriers(mp
);
1367 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
1371 * Logically we would return an error here to prevent
1372 * users from believing they might have changed
1373 * mount options using remount which can't be changed.
1375 * But unfortunately mount(8) adds all options from
1376 * mtab and fstab to the mount arguments in some cases
1377 * so we can't blindly reject options, but have to
1378 * check for each specified option if it actually
1379 * differs from the currently set option and only
1380 * reject it if that's the case.
1382 * Until that is implemented we return success for
1383 * every remount request, and silently ignore all
1384 * options that we can't actually change.
1388 "XFS: mount option \"%s\" not supported for remount\n", p
);
1397 if ((mp
->m_flags
& XFS_MOUNT_RDONLY
) && !(*flags
& MS_RDONLY
)) {
1398 mp
->m_flags
&= ~XFS_MOUNT_RDONLY
;
1399 if (mp
->m_flags
& XFS_MOUNT_BARRIER
)
1400 xfs_mountfs_check_barriers(mp
);
1403 * If this is the first remount to writeable state we
1404 * might have some superblock changes to update.
1406 if (mp
->m_update_flags
) {
1407 error
= xfs_mount_log_sb(mp
, mp
->m_update_flags
);
1410 "XFS: failed to write sb changes");
1413 mp
->m_update_flags
= 0;
1417 * Fill out the reserve pool if it is empty. Use the stashed
1418 * value if it is non-zero, otherwise go with the default.
1420 xfs_restore_resvblks(mp
);
1424 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
) && (*flags
& MS_RDONLY
)) {
1426 * After we have synced the data but before we sync the
1427 * metadata, we need to free up the reserve block pool so that
1428 * the used block count in the superblock on disk is correct at
1429 * the end of the remount. Stash the current reserve pool size
1430 * so that if we get remounted rw, we can return it to the same
1434 xfs_quiesce_data(mp
);
1435 xfs_save_resvblks(mp
);
1436 xfs_quiesce_attr(mp
);
1437 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
1444 * Second stage of a freeze. The data is already frozen so we only
1445 * need to take care of the metadata. Once that's done write a dummy
1446 * record to dirty the log in case of a crash while frozen.
1450 struct super_block
*sb
)
1452 struct xfs_mount
*mp
= XFS_M(sb
);
1454 xfs_save_resvblks(mp
);
1455 xfs_quiesce_attr(mp
);
1456 return -xfs_fs_log_dummy(mp
);
1461 struct super_block
*sb
)
1463 struct xfs_mount
*mp
= XFS_M(sb
);
1465 xfs_restore_resvblks(mp
);
1470 xfs_fs_show_options(
1472 struct vfsmount
*mnt
)
1474 return -xfs_showargs(XFS_M(mnt
->mnt_sb
), m
);
1478 * This function fills in xfs_mount_t fields based on mount args.
1479 * Note: the superblock _has_ now been read in.
1483 struct xfs_mount
*mp
)
1485 int ronly
= (mp
->m_flags
& XFS_MOUNT_RDONLY
);
1487 /* Fail a mount where the logbuf is smaller than the log stripe */
1488 if (xfs_sb_version_haslogv2(&mp
->m_sb
)) {
1489 if (mp
->m_logbsize
<= 0 &&
1490 mp
->m_sb
.sb_logsunit
> XLOG_BIG_RECORD_BSIZE
) {
1491 mp
->m_logbsize
= mp
->m_sb
.sb_logsunit
;
1492 } else if (mp
->m_logbsize
> 0 &&
1493 mp
->m_logbsize
< mp
->m_sb
.sb_logsunit
) {
1495 "XFS: logbuf size must be greater than or equal to log stripe size");
1496 return XFS_ERROR(EINVAL
);
1499 /* Fail a mount if the logbuf is larger than 32K */
1500 if (mp
->m_logbsize
> XLOG_BIG_RECORD_BSIZE
) {
1502 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1503 return XFS_ERROR(EINVAL
);
1508 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1509 * told by noattr2 to turn it off
1511 if (xfs_sb_version_hasattr2(&mp
->m_sb
) &&
1512 !(mp
->m_flags
& XFS_MOUNT_NOATTR2
))
1513 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
1516 * prohibit r/w mounts of read-only filesystems
1518 if ((mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) && !ronly
) {
1520 "XFS: cannot mount a read-only filesystem as read-write");
1521 return XFS_ERROR(EROFS
);
1529 struct super_block
*sb
,
1534 struct xfs_mount
*mp
= NULL
;
1535 int flags
= 0, error
= ENOMEM
;
1538 mp
= kzalloc(sizeof(struct xfs_mount
), GFP_KERNEL
);
1542 spin_lock_init(&mp
->m_sb_lock
);
1543 mutex_init(&mp
->m_growlock
);
1544 atomic_set(&mp
->m_active_trans
, 0);
1545 INIT_LIST_HEAD(&mp
->m_sync_list
);
1546 spin_lock_init(&mp
->m_sync_lock
);
1547 init_waitqueue_head(&mp
->m_wait_single_sync_task
);
1552 error
= xfs_parseargs(mp
, (char *)data
, &mtpt
);
1554 goto out_free_fsname
;
1556 sb_min_blocksize(sb
, BBSIZE
);
1557 sb
->s_xattr
= xfs_xattr_handlers
;
1558 sb
->s_export_op
= &xfs_export_operations
;
1559 #ifdef CONFIG_XFS_QUOTA
1560 sb
->s_qcop
= &xfs_quotactl_operations
;
1562 sb
->s_op
= &xfs_super_operations
;
1564 error
= xfs_dmops_get(mp
);
1566 goto out_free_fsname
;
1569 flags
|= XFS_MFSI_QUIET
;
1571 error
= xfs_open_devices(mp
);
1575 if (xfs_icsb_init_counters(mp
))
1576 mp
->m_flags
|= XFS_MOUNT_NO_PERCPU_SB
;
1578 error
= xfs_readsb(mp
, flags
);
1580 goto out_destroy_counters
;
1582 error
= xfs_finish_flags(mp
);
1586 error
= xfs_setup_devices(mp
);
1590 if (mp
->m_flags
& XFS_MOUNT_BARRIER
)
1591 xfs_mountfs_check_barriers(mp
);
1593 error
= xfs_filestream_mount(mp
);
1597 error
= xfs_mountfs(mp
);
1599 goto out_filestream_unmount
;
1601 XFS_SEND_MOUNT(mp
, DM_RIGHT_NULL
, mtpt
, mp
->m_fsname
);
1603 sb
->s_magic
= XFS_SB_MAGIC
;
1604 sb
->s_blocksize
= mp
->m_sb
.sb_blocksize
;
1605 sb
->s_blocksize_bits
= ffs(sb
->s_blocksize
) - 1;
1606 sb
->s_maxbytes
= xfs_max_file_offset(sb
->s_blocksize_bits
);
1607 sb
->s_time_gran
= 1;
1608 set_posix_acl_flag(sb
);
1610 root
= igrab(VFS_I(mp
->m_rootip
));
1615 if (is_bad_inode(root
)) {
1619 sb
->s_root
= d_alloc_root(root
);
1625 error
= xfs_syncd_init(mp
);
1629 xfs_inode_shrinker_register(mp
);
1634 out_filestream_unmount
:
1635 xfs_filestream_unmount(mp
);
1638 out_destroy_counters
:
1639 xfs_icsb_destroy_counters(mp
);
1640 xfs_close_devices(mp
);
1644 xfs_free_fsname(mp
);
1660 * Blow away any referenced inode in the filestreams cache.
1661 * This can and will cause log traffic as inodes go inactive
1664 xfs_filestream_unmount(mp
);
1666 XFS_bflush(mp
->m_ddev_targp
);
1674 struct file_system_type
*fs_type
,
1676 const char *dev_name
,
1678 struct vfsmount
*mnt
)
1680 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, xfs_fs_fill_super
,
1684 static const struct super_operations xfs_super_operations
= {
1685 .alloc_inode
= xfs_fs_alloc_inode
,
1686 .destroy_inode
= xfs_fs_destroy_inode
,
1687 .dirty_inode
= xfs_fs_dirty_inode
,
1688 .write_inode
= xfs_fs_write_inode
,
1689 .clear_inode
= xfs_fs_clear_inode
,
1690 .put_super
= xfs_fs_put_super
,
1691 .sync_fs
= xfs_fs_sync_fs
,
1692 .freeze_fs
= xfs_fs_freeze
,
1693 .unfreeze_fs
= xfs_fs_unfreeze
,
1694 .statfs
= xfs_fs_statfs
,
1695 .remount_fs
= xfs_fs_remount
,
1696 .show_options
= xfs_fs_show_options
,
1699 static struct file_system_type xfs_fs_type
= {
1700 .owner
= THIS_MODULE
,
1702 .get_sb
= xfs_fs_get_sb
,
1703 .kill_sb
= kill_block_super
,
1704 .fs_flags
= FS_REQUIRES_DEV
,
1708 xfs_init_zones(void)
1711 xfs_ioend_zone
= kmem_zone_init(sizeof(xfs_ioend_t
), "xfs_ioend");
1712 if (!xfs_ioend_zone
)
1715 xfs_ioend_pool
= mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE
,
1717 if (!xfs_ioend_pool
)
1718 goto out_destroy_ioend_zone
;
1720 xfs_log_ticket_zone
= kmem_zone_init(sizeof(xlog_ticket_t
),
1722 if (!xfs_log_ticket_zone
)
1723 goto out_destroy_ioend_pool
;
1725 xfs_bmap_free_item_zone
= kmem_zone_init(sizeof(xfs_bmap_free_item_t
),
1726 "xfs_bmap_free_item");
1727 if (!xfs_bmap_free_item_zone
)
1728 goto out_destroy_log_ticket_zone
;
1730 xfs_btree_cur_zone
= kmem_zone_init(sizeof(xfs_btree_cur_t
),
1732 if (!xfs_btree_cur_zone
)
1733 goto out_destroy_bmap_free_item_zone
;
1735 xfs_da_state_zone
= kmem_zone_init(sizeof(xfs_da_state_t
),
1737 if (!xfs_da_state_zone
)
1738 goto out_destroy_btree_cur_zone
;
1740 xfs_dabuf_zone
= kmem_zone_init(sizeof(xfs_dabuf_t
), "xfs_dabuf");
1741 if (!xfs_dabuf_zone
)
1742 goto out_destroy_da_state_zone
;
1744 xfs_ifork_zone
= kmem_zone_init(sizeof(xfs_ifork_t
), "xfs_ifork");
1745 if (!xfs_ifork_zone
)
1746 goto out_destroy_dabuf_zone
;
1748 xfs_trans_zone
= kmem_zone_init(sizeof(xfs_trans_t
), "xfs_trans");
1749 if (!xfs_trans_zone
)
1750 goto out_destroy_ifork_zone
;
1753 * The size of the zone allocated buf log item is the maximum
1754 * size possible under XFS. This wastes a little bit of memory,
1755 * but it is much faster.
1757 xfs_buf_item_zone
= kmem_zone_init((sizeof(xfs_buf_log_item_t
) +
1758 (((XFS_MAX_BLOCKSIZE
/ XFS_BLI_CHUNK
) /
1759 NBWORD
) * sizeof(int))), "xfs_buf_item");
1760 if (!xfs_buf_item_zone
)
1761 goto out_destroy_trans_zone
;
1763 xfs_efd_zone
= kmem_zone_init((sizeof(xfs_efd_log_item_t
) +
1764 ((XFS_EFD_MAX_FAST_EXTENTS
- 1) *
1765 sizeof(xfs_extent_t
))), "xfs_efd_item");
1767 goto out_destroy_buf_item_zone
;
1769 xfs_efi_zone
= kmem_zone_init((sizeof(xfs_efi_log_item_t
) +
1770 ((XFS_EFI_MAX_FAST_EXTENTS
- 1) *
1771 sizeof(xfs_extent_t
))), "xfs_efi_item");
1773 goto out_destroy_efd_zone
;
1776 kmem_zone_init_flags(sizeof(xfs_inode_t
), "xfs_inode",
1777 KM_ZONE_HWALIGN
| KM_ZONE_RECLAIM
| KM_ZONE_SPREAD
,
1778 xfs_fs_inode_init_once
);
1779 if (!xfs_inode_zone
)
1780 goto out_destroy_efi_zone
;
1783 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t
), "xfs_ili",
1784 KM_ZONE_SPREAD
, NULL
);
1786 goto out_destroy_inode_zone
;
1790 out_destroy_inode_zone
:
1791 kmem_zone_destroy(xfs_inode_zone
);
1792 out_destroy_efi_zone
:
1793 kmem_zone_destroy(xfs_efi_zone
);
1794 out_destroy_efd_zone
:
1795 kmem_zone_destroy(xfs_efd_zone
);
1796 out_destroy_buf_item_zone
:
1797 kmem_zone_destroy(xfs_buf_item_zone
);
1798 out_destroy_trans_zone
:
1799 kmem_zone_destroy(xfs_trans_zone
);
1800 out_destroy_ifork_zone
:
1801 kmem_zone_destroy(xfs_ifork_zone
);
1802 out_destroy_dabuf_zone
:
1803 kmem_zone_destroy(xfs_dabuf_zone
);
1804 out_destroy_da_state_zone
:
1805 kmem_zone_destroy(xfs_da_state_zone
);
1806 out_destroy_btree_cur_zone
:
1807 kmem_zone_destroy(xfs_btree_cur_zone
);
1808 out_destroy_bmap_free_item_zone
:
1809 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1810 out_destroy_log_ticket_zone
:
1811 kmem_zone_destroy(xfs_log_ticket_zone
);
1812 out_destroy_ioend_pool
:
1813 mempool_destroy(xfs_ioend_pool
);
1814 out_destroy_ioend_zone
:
1815 kmem_zone_destroy(xfs_ioend_zone
);
1821 xfs_destroy_zones(void)
1823 kmem_zone_destroy(xfs_ili_zone
);
1824 kmem_zone_destroy(xfs_inode_zone
);
1825 kmem_zone_destroy(xfs_efi_zone
);
1826 kmem_zone_destroy(xfs_efd_zone
);
1827 kmem_zone_destroy(xfs_buf_item_zone
);
1828 kmem_zone_destroy(xfs_trans_zone
);
1829 kmem_zone_destroy(xfs_ifork_zone
);
1830 kmem_zone_destroy(xfs_dabuf_zone
);
1831 kmem_zone_destroy(xfs_da_state_zone
);
1832 kmem_zone_destroy(xfs_btree_cur_zone
);
1833 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1834 kmem_zone_destroy(xfs_log_ticket_zone
);
1835 mempool_destroy(xfs_ioend_pool
);
1836 kmem_zone_destroy(xfs_ioend_zone
);
1845 printk(KERN_INFO XFS_VERSION_STRING
" with "
1846 XFS_BUILD_OPTIONS
" enabled\n");
1851 error
= xfs_init_zones();
1855 error
= xfs_mru_cache_init();
1857 goto out_destroy_zones
;
1859 error
= xfs_filestream_init();
1861 goto out_mru_cache_uninit
;
1863 error
= xfs_buf_init();
1865 goto out_filestream_uninit
;
1867 error
= xfs_init_procfs();
1869 goto out_buf_terminate
;
1871 error
= xfs_sysctl_register();
1873 goto out_cleanup_procfs
;
1876 xfs_inode_shrinker_init();
1878 error
= register_filesystem(&xfs_fs_type
);
1880 goto out_sysctl_unregister
;
1883 out_sysctl_unregister
:
1884 xfs_sysctl_unregister();
1886 xfs_cleanup_procfs();
1888 xfs_buf_terminate();
1889 out_filestream_uninit
:
1890 xfs_filestream_uninit();
1891 out_mru_cache_uninit
:
1892 xfs_mru_cache_uninit();
1894 xfs_destroy_zones();
1903 unregister_filesystem(&xfs_fs_type
);
1904 xfs_inode_shrinker_destroy();
1905 xfs_sysctl_unregister();
1906 xfs_cleanup_procfs();
1907 xfs_buf_terminate();
1908 xfs_filestream_uninit();
1909 xfs_mru_cache_uninit();
1910 xfs_destroy_zones();
1913 module_init(init_xfs_fs
);
1914 module_exit(exit_xfs_fs
);
1916 MODULE_AUTHOR("Silicon Graphics, Inc.");
1917 MODULE_DESCRIPTION(XFS_VERSION_STRING
" with " XFS_BUILD_OPTIONS
" enabled");
1918 MODULE_LICENSE("GPL");