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"
53 #include "xfs_trace.h"
55 #include <linux/namei.h>
56 #include <linux/init.h>
57 #include <linux/slab.h>
58 #include <linux/mount.h>
59 #include <linux/mempool.h>
60 #include <linux/writeback.h>
61 #include <linux/kthread.h>
62 #include <linux/freezer.h>
63 #include <linux/parser.h>
65 static const struct super_operations xfs_super_operations
;
66 static kmem_zone_t
*xfs_ioend_zone
;
67 mempool_t
*xfs_ioend_pool
;
69 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
70 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
71 #define MNTOPT_LOGDEV "logdev" /* log device */
72 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
73 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
74 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
75 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
76 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
77 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
78 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
79 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
80 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
81 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
82 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
83 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
84 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
85 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
86 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
87 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
88 * unwritten extent conversion */
89 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
90 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
91 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
92 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
93 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
94 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
96 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
97 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
98 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
99 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
100 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
101 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
102 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
103 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
104 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
105 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
106 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
107 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
108 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
109 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
110 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
111 #define MNTOPT_DELAYLOG "delaylog" /* Delayed logging enabled */
112 #define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed logging disabled */
113 #define MNTOPT_DISCARD "discard" /* Discard unused blocks */
114 #define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */
117 * Table driven mount option parser.
119 * Currently only used for remount, but it will be used for mount
120 * in the future, too.
123 Opt_barrier
, Opt_nobarrier
, Opt_err
126 static const match_table_t tokens
= {
127 {Opt_barrier
, "barrier"},
128 {Opt_nobarrier
, "nobarrier"},
134 suffix_strtoul(char *s
, char **endp
, unsigned int base
)
136 int last
, shift_left_factor
= 0;
139 last
= strlen(value
) - 1;
140 if (value
[last
] == 'K' || value
[last
] == 'k') {
141 shift_left_factor
= 10;
144 if (value
[last
] == 'M' || value
[last
] == 'm') {
145 shift_left_factor
= 20;
148 if (value
[last
] == 'G' || value
[last
] == 'g') {
149 shift_left_factor
= 30;
153 return simple_strtoul((const char *)s
, endp
, base
) << shift_left_factor
;
157 * This function fills in xfs_mount_t fields based on mount args.
158 * Note: the superblock has _not_ yet been read in.
160 * Note that this function leaks the various device name allocations on
161 * failure. The caller takes care of them.
165 struct xfs_mount
*mp
,
168 struct super_block
*sb
= mp
->m_super
;
169 char *this_char
, *value
, *eov
;
173 __uint8_t iosizelog
= 0;
176 * set up the mount name first so all the errors will refer to the
179 mp
->m_fsname
= kstrndup(sb
->s_id
, MAXNAMELEN
, GFP_KERNEL
);
182 mp
->m_fsname_len
= strlen(mp
->m_fsname
) + 1;
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
) {
219 xfs_warn(mp
, "%s option requires an argument",
223 mp
->m_logbufs
= simple_strtoul(value
, &eov
, 10);
224 } else if (!strcmp(this_char
, MNTOPT_LOGBSIZE
)) {
225 if (!value
|| !*value
) {
226 xfs_warn(mp
, "%s option requires an argument",
230 mp
->m_logbsize
= suffix_strtoul(value
, &eov
, 10);
231 } else if (!strcmp(this_char
, MNTOPT_LOGDEV
)) {
232 if (!value
|| !*value
) {
233 xfs_warn(mp
, "%s option requires an argument",
237 mp
->m_logname
= kstrndup(value
, MAXNAMELEN
, GFP_KERNEL
);
240 } else if (!strcmp(this_char
, MNTOPT_MTPT
)) {
241 xfs_warn(mp
, "%s option not allowed on this system",
244 } else if (!strcmp(this_char
, MNTOPT_RTDEV
)) {
245 if (!value
|| !*value
) {
246 xfs_warn(mp
, "%s option requires an argument",
250 mp
->m_rtname
= kstrndup(value
, MAXNAMELEN
, GFP_KERNEL
);
253 } else if (!strcmp(this_char
, MNTOPT_BIOSIZE
)) {
254 if (!value
|| !*value
) {
255 xfs_warn(mp
, "%s option requires an argument",
259 iosize
= simple_strtoul(value
, &eov
, 10);
260 iosizelog
= ffs(iosize
) - 1;
261 } else if (!strcmp(this_char
, MNTOPT_ALLOCSIZE
)) {
262 if (!value
|| !*value
) {
263 xfs_warn(mp
, "%s option requires an argument",
267 iosize
= suffix_strtoul(value
, &eov
, 10);
268 iosizelog
= ffs(iosize
) - 1;
269 } else if (!strcmp(this_char
, MNTOPT_GRPID
) ||
270 !strcmp(this_char
, MNTOPT_BSDGROUPS
)) {
271 mp
->m_flags
|= XFS_MOUNT_GRPID
;
272 } else if (!strcmp(this_char
, MNTOPT_NOGRPID
) ||
273 !strcmp(this_char
, MNTOPT_SYSVGROUPS
)) {
274 mp
->m_flags
&= ~XFS_MOUNT_GRPID
;
275 } else if (!strcmp(this_char
, MNTOPT_WSYNC
)) {
276 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
277 } else if (!strcmp(this_char
, MNTOPT_NORECOVERY
)) {
278 mp
->m_flags
|= XFS_MOUNT_NORECOVERY
;
279 } else if (!strcmp(this_char
, MNTOPT_NOALIGN
)) {
280 mp
->m_flags
|= XFS_MOUNT_NOALIGN
;
281 } else if (!strcmp(this_char
, MNTOPT_SWALLOC
)) {
282 mp
->m_flags
|= XFS_MOUNT_SWALLOC
;
283 } else if (!strcmp(this_char
, MNTOPT_SUNIT
)) {
284 if (!value
|| !*value
) {
285 xfs_warn(mp
, "%s option requires an argument",
289 dsunit
= simple_strtoul(value
, &eov
, 10);
290 } else if (!strcmp(this_char
, MNTOPT_SWIDTH
)) {
291 if (!value
|| !*value
) {
292 xfs_warn(mp
, "%s option requires an argument",
296 dswidth
= simple_strtoul(value
, &eov
, 10);
297 } else if (!strcmp(this_char
, MNTOPT_64BITINODE
)) {
298 mp
->m_flags
&= ~XFS_MOUNT_SMALL_INUMS
;
300 xfs_warn(mp
, "%s option not allowed on this system",
304 } else if (!strcmp(this_char
, MNTOPT_NOUUID
)) {
305 mp
->m_flags
|= XFS_MOUNT_NOUUID
;
306 } else if (!strcmp(this_char
, MNTOPT_BARRIER
)) {
307 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
308 } else if (!strcmp(this_char
, MNTOPT_NOBARRIER
)) {
309 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
310 } else if (!strcmp(this_char
, MNTOPT_IKEEP
)) {
311 mp
->m_flags
|= XFS_MOUNT_IKEEP
;
312 } else if (!strcmp(this_char
, MNTOPT_NOIKEEP
)) {
313 mp
->m_flags
&= ~XFS_MOUNT_IKEEP
;
314 } else if (!strcmp(this_char
, MNTOPT_LARGEIO
)) {
315 mp
->m_flags
&= ~XFS_MOUNT_COMPAT_IOSIZE
;
316 } else if (!strcmp(this_char
, MNTOPT_NOLARGEIO
)) {
317 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
318 } else if (!strcmp(this_char
, MNTOPT_ATTR2
)) {
319 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
320 } else if (!strcmp(this_char
, MNTOPT_NOATTR2
)) {
321 mp
->m_flags
&= ~XFS_MOUNT_ATTR2
;
322 mp
->m_flags
|= XFS_MOUNT_NOATTR2
;
323 } else if (!strcmp(this_char
, MNTOPT_FILESTREAM
)) {
324 mp
->m_flags
|= XFS_MOUNT_FILESTREAMS
;
325 } else if (!strcmp(this_char
, MNTOPT_NOQUOTA
)) {
326 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACCT
;
327 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ENFD
;
328 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACTIVE
;
329 } else if (!strcmp(this_char
, MNTOPT_QUOTA
) ||
330 !strcmp(this_char
, MNTOPT_UQUOTA
) ||
331 !strcmp(this_char
, MNTOPT_USRQUOTA
)) {
332 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
|
334 } else if (!strcmp(this_char
, MNTOPT_QUOTANOENF
) ||
335 !strcmp(this_char
, MNTOPT_UQUOTANOENF
)) {
336 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
);
337 mp
->m_qflags
&= ~XFS_UQUOTA_ENFD
;
338 } else if (!strcmp(this_char
, MNTOPT_PQUOTA
) ||
339 !strcmp(this_char
, MNTOPT_PRJQUOTA
)) {
340 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
|
342 } else if (!strcmp(this_char
, MNTOPT_PQUOTANOENF
)) {
343 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
);
344 mp
->m_qflags
&= ~XFS_OQUOTA_ENFD
;
345 } else if (!strcmp(this_char
, MNTOPT_GQUOTA
) ||
346 !strcmp(this_char
, MNTOPT_GRPQUOTA
)) {
347 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
|
349 } else if (!strcmp(this_char
, MNTOPT_GQUOTANOENF
)) {
350 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
);
351 mp
->m_qflags
&= ~XFS_OQUOTA_ENFD
;
352 } else if (!strcmp(this_char
, MNTOPT_DELAYLOG
)) {
354 "delaylog is the default now, option is deprecated.");
355 } else if (!strcmp(this_char
, MNTOPT_NODELAYLOG
)) {
357 "nodelaylog support has been removed, option is deprecated.");
358 } else if (!strcmp(this_char
, MNTOPT_DISCARD
)) {
359 mp
->m_flags
|= XFS_MOUNT_DISCARD
;
360 } else if (!strcmp(this_char
, MNTOPT_NODISCARD
)) {
361 mp
->m_flags
&= ~XFS_MOUNT_DISCARD
;
362 } else if (!strcmp(this_char
, "ihashsize")) {
364 "ihashsize no longer used, option is deprecated.");
365 } else if (!strcmp(this_char
, "osyncisdsync")) {
367 "osyncisdsync has no effect, option is deprecated.");
368 } else if (!strcmp(this_char
, "osyncisosync")) {
370 "osyncisosync has no effect, option is deprecated.");
371 } else if (!strcmp(this_char
, "irixsgid")) {
373 "irixsgid is now a sysctl(2) variable, option is deprecated.");
375 xfs_warn(mp
, "unknown mount option [%s].", this_char
);
381 * no recovery flag requires a read-only mount
383 if ((mp
->m_flags
& XFS_MOUNT_NORECOVERY
) &&
384 !(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
385 xfs_warn(mp
, "no-recovery mounts must be read-only.");
389 if ((mp
->m_flags
& XFS_MOUNT_NOALIGN
) && (dsunit
|| dswidth
)) {
391 "sunit and swidth options incompatible with the noalign option");
395 #ifndef CONFIG_XFS_QUOTA
396 if (XFS_IS_QUOTA_RUNNING(mp
)) {
397 xfs_warn(mp
, "quota support not available in this kernel.");
402 if ((mp
->m_qflags
& (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
)) &&
403 (mp
->m_qflags
& (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
))) {
404 xfs_warn(mp
, "cannot mount with both project and group quota");
408 if ((dsunit
&& !dswidth
) || (!dsunit
&& dswidth
)) {
409 xfs_warn(mp
, "sunit and swidth must be specified together");
413 if (dsunit
&& (dswidth
% dsunit
!= 0)) {
415 "stripe width (%d) must be a multiple of the stripe unit (%d)",
421 if (!(mp
->m_flags
& XFS_MOUNT_NOALIGN
)) {
423 * At this point the superblock has not been read
424 * in, therefore we do not know the block size.
425 * Before the mount call ends we will convert
429 mp
->m_dalign
= dsunit
;
430 mp
->m_flags
|= XFS_MOUNT_RETERR
;
434 mp
->m_swidth
= dswidth
;
437 if (mp
->m_logbufs
!= -1 &&
438 mp
->m_logbufs
!= 0 &&
439 (mp
->m_logbufs
< XLOG_MIN_ICLOGS
||
440 mp
->m_logbufs
> XLOG_MAX_ICLOGS
)) {
441 xfs_warn(mp
, "invalid logbufs value: %d [not %d-%d]",
442 mp
->m_logbufs
, XLOG_MIN_ICLOGS
, XLOG_MAX_ICLOGS
);
443 return XFS_ERROR(EINVAL
);
445 if (mp
->m_logbsize
!= -1 &&
446 mp
->m_logbsize
!= 0 &&
447 (mp
->m_logbsize
< XLOG_MIN_RECORD_BSIZE
||
448 mp
->m_logbsize
> XLOG_MAX_RECORD_BSIZE
||
449 !is_power_of_2(mp
->m_logbsize
))) {
451 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
453 return XFS_ERROR(EINVAL
);
457 if (iosizelog
> XFS_MAX_IO_LOG
||
458 iosizelog
< XFS_MIN_IO_LOG
) {
459 xfs_warn(mp
, "invalid log iosize: %d [not %d-%d]",
460 iosizelog
, XFS_MIN_IO_LOG
,
462 return XFS_ERROR(EINVAL
);
465 mp
->m_flags
|= XFS_MOUNT_DFLT_IOSIZE
;
466 mp
->m_readio_log
= iosizelog
;
467 mp
->m_writeio_log
= iosizelog
;
473 struct proc_xfs_info
{
480 struct xfs_mount
*mp
,
483 static struct proc_xfs_info xfs_info_set
[] = {
484 /* the few simple ones we can get from the mount struct */
485 { XFS_MOUNT_IKEEP
, "," MNTOPT_IKEEP
},
486 { XFS_MOUNT_WSYNC
, "," MNTOPT_WSYNC
},
487 { XFS_MOUNT_NOALIGN
, "," MNTOPT_NOALIGN
},
488 { XFS_MOUNT_SWALLOC
, "," MNTOPT_SWALLOC
},
489 { XFS_MOUNT_NOUUID
, "," MNTOPT_NOUUID
},
490 { XFS_MOUNT_NORECOVERY
, "," MNTOPT_NORECOVERY
},
491 { XFS_MOUNT_ATTR2
, "," MNTOPT_ATTR2
},
492 { XFS_MOUNT_FILESTREAMS
, "," MNTOPT_FILESTREAM
},
493 { XFS_MOUNT_GRPID
, "," MNTOPT_GRPID
},
494 { XFS_MOUNT_DISCARD
, "," MNTOPT_DISCARD
},
497 static struct proc_xfs_info xfs_info_unset
[] = {
498 /* the few simple ones we can get from the mount struct */
499 { XFS_MOUNT_COMPAT_IOSIZE
, "," MNTOPT_LARGEIO
},
500 { XFS_MOUNT_BARRIER
, "," MNTOPT_NOBARRIER
},
501 { XFS_MOUNT_SMALL_INUMS
, "," MNTOPT_64BITINODE
},
504 struct proc_xfs_info
*xfs_infop
;
506 for (xfs_infop
= xfs_info_set
; xfs_infop
->flag
; xfs_infop
++) {
507 if (mp
->m_flags
& xfs_infop
->flag
)
508 seq_puts(m
, xfs_infop
->str
);
510 for (xfs_infop
= xfs_info_unset
; xfs_infop
->flag
; xfs_infop
++) {
511 if (!(mp
->m_flags
& xfs_infop
->flag
))
512 seq_puts(m
, xfs_infop
->str
);
515 if (mp
->m_flags
& XFS_MOUNT_DFLT_IOSIZE
)
516 seq_printf(m
, "," MNTOPT_ALLOCSIZE
"=%dk",
517 (int)(1 << mp
->m_writeio_log
) >> 10);
519 if (mp
->m_logbufs
> 0)
520 seq_printf(m
, "," MNTOPT_LOGBUFS
"=%d", mp
->m_logbufs
);
521 if (mp
->m_logbsize
> 0)
522 seq_printf(m
, "," MNTOPT_LOGBSIZE
"=%dk", mp
->m_logbsize
>> 10);
525 seq_printf(m
, "," MNTOPT_LOGDEV
"=%s", mp
->m_logname
);
527 seq_printf(m
, "," MNTOPT_RTDEV
"=%s", mp
->m_rtname
);
529 if (mp
->m_dalign
> 0)
530 seq_printf(m
, "," MNTOPT_SUNIT
"=%d",
531 (int)XFS_FSB_TO_BB(mp
, mp
->m_dalign
));
532 if (mp
->m_swidth
> 0)
533 seq_printf(m
, "," MNTOPT_SWIDTH
"=%d",
534 (int)XFS_FSB_TO_BB(mp
, mp
->m_swidth
));
536 if (mp
->m_qflags
& (XFS_UQUOTA_ACCT
|XFS_UQUOTA_ENFD
))
537 seq_puts(m
, "," MNTOPT_USRQUOTA
);
538 else if (mp
->m_qflags
& XFS_UQUOTA_ACCT
)
539 seq_puts(m
, "," MNTOPT_UQUOTANOENF
);
541 /* Either project or group quotas can be active, not both */
543 if (mp
->m_qflags
& XFS_PQUOTA_ACCT
) {
544 if (mp
->m_qflags
& XFS_OQUOTA_ENFD
)
545 seq_puts(m
, "," MNTOPT_PRJQUOTA
);
547 seq_puts(m
, "," MNTOPT_PQUOTANOENF
);
548 } else if (mp
->m_qflags
& XFS_GQUOTA_ACCT
) {
549 if (mp
->m_qflags
& XFS_OQUOTA_ENFD
)
550 seq_puts(m
, "," MNTOPT_GRPQUOTA
);
552 seq_puts(m
, "," MNTOPT_GQUOTANOENF
);
555 if (!(mp
->m_qflags
& XFS_ALL_QUOTA_ACCT
))
556 seq_puts(m
, "," MNTOPT_NOQUOTA
);
562 unsigned int blockshift
)
564 unsigned int pagefactor
= 1;
565 unsigned int bitshift
= BITS_PER_LONG
- 1;
567 /* Figure out maximum filesize, on Linux this can depend on
568 * the filesystem blocksize (on 32 bit platforms).
569 * __block_write_begin does this in an [unsigned] long...
570 * page->index << (PAGE_CACHE_SHIFT - bbits)
571 * So, for page sized blocks (4K on 32 bit platforms),
572 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
573 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
574 * but for smaller blocksizes it is less (bbits = log2 bsize).
575 * Note1: get_block_t takes a long (implicit cast from above)
576 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
577 * can optionally convert the [unsigned] long from above into
578 * an [unsigned] long long.
581 #if BITS_PER_LONG == 32
582 # if defined(CONFIG_LBDAF)
583 ASSERT(sizeof(sector_t
) == 8);
584 pagefactor
= PAGE_CACHE_SIZE
;
585 bitshift
= BITS_PER_LONG
;
587 pagefactor
= PAGE_CACHE_SIZE
>> (PAGE_CACHE_SHIFT
- blockshift
);
591 return (((__uint64_t
)pagefactor
) << bitshift
) - 1;
598 struct block_device
**bdevp
)
602 *bdevp
= blkdev_get_by_path(name
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
,
604 if (IS_ERR(*bdevp
)) {
605 error
= PTR_ERR(*bdevp
);
606 xfs_warn(mp
, "Invalid device [%s], error=%d\n", name
, error
);
614 struct block_device
*bdev
)
617 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
621 xfs_blkdev_issue_flush(
622 xfs_buftarg_t
*buftarg
)
624 blkdev_issue_flush(buftarg
->bt_bdev
, GFP_NOFS
, NULL
);
629 struct xfs_mount
*mp
)
631 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
632 struct block_device
*logdev
= mp
->m_logdev_targp
->bt_bdev
;
633 xfs_free_buftarg(mp
, mp
->m_logdev_targp
);
634 xfs_blkdev_put(logdev
);
636 if (mp
->m_rtdev_targp
) {
637 struct block_device
*rtdev
= mp
->m_rtdev_targp
->bt_bdev
;
638 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
639 xfs_blkdev_put(rtdev
);
641 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
645 * The file system configurations are:
646 * (1) device (partition) with data and internal log
647 * (2) logical volume with data and log subvolumes.
648 * (3) logical volume with data, log, and realtime subvolumes.
650 * We only have to handle opening the log and realtime volumes here if
651 * they are present. The data subvolume has already been opened by
652 * get_sb_bdev() and is stored in sb->s_bdev.
656 struct xfs_mount
*mp
)
658 struct block_device
*ddev
= mp
->m_super
->s_bdev
;
659 struct block_device
*logdev
= NULL
, *rtdev
= NULL
;
663 * Open real time and log devices - order is important.
666 error
= xfs_blkdev_get(mp
, mp
->m_logname
, &logdev
);
672 error
= xfs_blkdev_get(mp
, mp
->m_rtname
, &rtdev
);
674 goto out_close_logdev
;
676 if (rtdev
== ddev
|| rtdev
== logdev
) {
678 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
680 goto out_close_rtdev
;
685 * Setup xfs_mount buffer target pointers
688 mp
->m_ddev_targp
= xfs_alloc_buftarg(mp
, ddev
, 0, mp
->m_fsname
);
689 if (!mp
->m_ddev_targp
)
690 goto out_close_rtdev
;
693 mp
->m_rtdev_targp
= xfs_alloc_buftarg(mp
, rtdev
, 1,
695 if (!mp
->m_rtdev_targp
)
696 goto out_free_ddev_targ
;
699 if (logdev
&& logdev
!= ddev
) {
700 mp
->m_logdev_targp
= xfs_alloc_buftarg(mp
, logdev
, 1,
702 if (!mp
->m_logdev_targp
)
703 goto out_free_rtdev_targ
;
705 mp
->m_logdev_targp
= mp
->m_ddev_targp
;
711 if (mp
->m_rtdev_targp
)
712 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
714 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
717 xfs_blkdev_put(rtdev
);
719 if (logdev
&& logdev
!= ddev
)
720 xfs_blkdev_put(logdev
);
726 * Setup xfs_mount buffer target pointers based on superblock
730 struct xfs_mount
*mp
)
734 error
= xfs_setsize_buftarg(mp
->m_ddev_targp
, mp
->m_sb
.sb_blocksize
,
735 mp
->m_sb
.sb_sectsize
);
739 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
740 unsigned int log_sector_size
= BBSIZE
;
742 if (xfs_sb_version_hassector(&mp
->m_sb
))
743 log_sector_size
= mp
->m_sb
.sb_logsectsize
;
744 error
= xfs_setsize_buftarg(mp
->m_logdev_targp
,
745 mp
->m_sb
.sb_blocksize
,
750 if (mp
->m_rtdev_targp
) {
751 error
= xfs_setsize_buftarg(mp
->m_rtdev_targp
,
752 mp
->m_sb
.sb_blocksize
,
753 mp
->m_sb
.sb_sectsize
);
762 xfs_init_mount_workqueues(
763 struct xfs_mount
*mp
)
765 mp
->m_data_workqueue
= alloc_workqueue("xfs-data/%s",
766 WQ_MEM_RECLAIM
, 0, mp
->m_fsname
);
767 if (!mp
->m_data_workqueue
)
770 mp
->m_unwritten_workqueue
= alloc_workqueue("xfs-conv/%s",
771 WQ_MEM_RECLAIM
, 0, mp
->m_fsname
);
772 if (!mp
->m_unwritten_workqueue
)
773 goto out_destroy_data_iodone_queue
;
775 mp
->m_cil_workqueue
= alloc_workqueue("xfs-cil/%s",
776 WQ_MEM_RECLAIM
, 0, mp
->m_fsname
);
777 if (!mp
->m_cil_workqueue
)
778 goto out_destroy_unwritten
;
781 out_destroy_unwritten
:
782 destroy_workqueue(mp
->m_unwritten_workqueue
);
783 out_destroy_data_iodone_queue
:
784 destroy_workqueue(mp
->m_data_workqueue
);
790 xfs_destroy_mount_workqueues(
791 struct xfs_mount
*mp
)
793 destroy_workqueue(mp
->m_cil_workqueue
);
794 destroy_workqueue(mp
->m_data_workqueue
);
795 destroy_workqueue(mp
->m_unwritten_workqueue
);
798 /* Catch misguided souls that try to use this interface on XFS */
799 STATIC
struct inode
*
801 struct super_block
*sb
)
808 * Now that the generic code is guaranteed not to be accessing
809 * the linux inode, we can reclaim the inode.
812 xfs_fs_destroy_inode(
815 struct xfs_inode
*ip
= XFS_I(inode
);
817 trace_xfs_destroy_inode(ip
);
819 XFS_STATS_INC(vn_reclaim
);
821 /* bad inode, get out here ASAP */
822 if (is_bad_inode(inode
))
825 ASSERT(XFS_FORCED_SHUTDOWN(ip
->i_mount
) || ip
->i_delayed_blks
== 0);
828 * We should never get here with one of the reclaim flags already set.
830 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIMABLE
));
831 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIM
));
834 * We always use background reclaim here because even if the
835 * inode is clean, it still may be under IO and hence we have
836 * to take the flush lock. The background reclaim path handles
837 * this more efficiently than we can here, so simply let background
838 * reclaim tear down all inodes.
841 xfs_inode_set_reclaim_tag(ip
);
845 * Slab object creation initialisation for the XFS inode.
846 * This covers only the idempotent fields in the XFS inode;
847 * all other fields need to be initialised on allocation
848 * from the slab. This avoids the need to repeatedly initialise
849 * fields in the xfs inode that left in the initialise state
850 * when freeing the inode.
853 xfs_fs_inode_init_once(
856 struct xfs_inode
*ip
= inode
;
858 memset(ip
, 0, sizeof(struct xfs_inode
));
861 inode_init_once(VFS_I(ip
));
864 atomic_set(&ip
->i_pincount
, 0);
865 spin_lock_init(&ip
->i_flags_lock
);
867 mrlock_init(&ip
->i_lock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
868 "xfsino", ip
->i_ino
);
872 * This is called by the VFS when dirtying inode metadata. This can happen
873 * for a few reasons, but we only care about timestamp updates, given that
874 * we handled the rest ourselves. In theory no other calls should happen,
875 * but for example generic_write_end() keeps dirtying the inode after
876 * updating i_size. Thus we check that the flags are exactly I_DIRTY_SYNC,
877 * and skip this call otherwise.
879 * We'll hopefull get a different method just for updating timestamps soon,
880 * at which point this hack can go away, and maybe we'll also get real
881 * error handling here.
888 struct xfs_inode
*ip
= XFS_I(inode
);
889 struct xfs_mount
*mp
= ip
->i_mount
;
890 struct xfs_trans
*tp
;
893 if (flags
!= I_DIRTY_SYNC
)
896 trace_xfs_dirty_inode(ip
);
898 tp
= xfs_trans_alloc(mp
, XFS_TRANS_FSYNC_TS
);
899 error
= xfs_trans_reserve(tp
, 0, XFS_FSYNC_TS_LOG_RES(mp
), 0, 0, 0);
901 xfs_trans_cancel(tp
, 0);
904 xfs_ilock(ip
, XFS_ILOCK_EXCL
);
906 * Grab all the latest timestamps from the Linux inode.
908 ip
->i_d
.di_atime
.t_sec
= (__int32_t
)inode
->i_atime
.tv_sec
;
909 ip
->i_d
.di_atime
.t_nsec
= (__int32_t
)inode
->i_atime
.tv_nsec
;
910 ip
->i_d
.di_ctime
.t_sec
= (__int32_t
)inode
->i_ctime
.tv_sec
;
911 ip
->i_d
.di_ctime
.t_nsec
= (__int32_t
)inode
->i_ctime
.tv_nsec
;
912 ip
->i_d
.di_mtime
.t_sec
= (__int32_t
)inode
->i_mtime
.tv_sec
;
913 ip
->i_d
.di_mtime
.t_nsec
= (__int32_t
)inode
->i_mtime
.tv_nsec
;
915 xfs_trans_ijoin(tp
, ip
, XFS_ILOCK_EXCL
);
916 xfs_trans_log_inode(tp
, ip
, XFS_ILOG_TIMESTAMP
);
917 error
= xfs_trans_commit(tp
, 0);
923 xfs_warn(mp
, "failed to update timestamps for inode 0x%llx", ip
->i_ino
);
930 xfs_inode_t
*ip
= XFS_I(inode
);
932 trace_xfs_evict_inode(ip
);
934 truncate_inode_pages(&inode
->i_data
, 0);
935 end_writeback(inode
);
936 XFS_STATS_INC(vn_rele
);
937 XFS_STATS_INC(vn_remove
);
938 XFS_STATS_DEC(vn_active
);
941 * The iolock is used by the file system to coordinate reads,
942 * writes, and block truncates. Up to this point the lock
943 * protected concurrent accesses by users of the inode. But
944 * from here forward we're doing some final processing of the
945 * inode because we're done with it, and although we reuse the
946 * iolock for protection it is really a distinct lock class
947 * (in the lockdep sense) from before. To keep lockdep happy
948 * (and basically indicate what we are doing), we explicitly
949 * re-init the iolock here.
951 ASSERT(!rwsem_is_locked(&ip
->i_iolock
.mr_lock
));
952 mrlock_init(&ip
->i_iolock
, MRLOCK_BARRIER
, "xfsio", ip
->i_ino
);
953 lockdep_set_class_and_name(&ip
->i_iolock
.mr_lock
,
954 &xfs_iolock_reclaimable
, "xfs_iolock_reclaimable");
960 * We do an unlocked check for XFS_IDONTCACHE here because we are already
961 * serialised against cache hits here via the inode->i_lock and igrab() in
962 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
963 * racing with us, and it avoids needing to grab a spinlock here for every inode
964 * we drop the final reference on.
970 struct xfs_inode
*ip
= XFS_I(inode
);
972 return generic_drop_inode(inode
) || (ip
->i_flags
& XFS_IDONTCACHE
);
977 struct xfs_mount
*mp
)
981 kfree(mp
->m_logname
);
986 struct super_block
*sb
)
988 struct xfs_mount
*mp
= XFS_M(sb
);
990 xfs_filestream_unmount(mp
);
994 xfs_icsb_destroy_counters(mp
);
995 xfs_destroy_mount_workqueues(mp
);
996 xfs_close_devices(mp
);
1003 struct super_block
*sb
,
1006 struct xfs_mount
*mp
= XFS_M(sb
);
1010 * Doing anything during the async pass would be counterproductive.
1015 error
= xfs_quiesce_data(mp
);
1021 * The disk must be active because we're syncing.
1022 * We schedule xfssyncd now (now that the disk is
1023 * active) instead of later (when it might not be).
1025 flush_delayed_work_sync(&mp
->m_sync_work
);
1033 struct dentry
*dentry
,
1034 struct kstatfs
*statp
)
1036 struct xfs_mount
*mp
= XFS_M(dentry
->d_sb
);
1037 xfs_sb_t
*sbp
= &mp
->m_sb
;
1038 struct xfs_inode
*ip
= XFS_I(dentry
->d_inode
);
1039 __uint64_t fakeinos
, id
;
1043 statp
->f_type
= XFS_SB_MAGIC
;
1044 statp
->f_namelen
= MAXNAMELEN
- 1;
1046 id
= huge_encode_dev(mp
->m_ddev_targp
->bt_dev
);
1047 statp
->f_fsid
.val
[0] = (u32
)id
;
1048 statp
->f_fsid
.val
[1] = (u32
)(id
>> 32);
1050 xfs_icsb_sync_counters(mp
, XFS_ICSB_LAZY_COUNT
);
1052 spin_lock(&mp
->m_sb_lock
);
1053 statp
->f_bsize
= sbp
->sb_blocksize
;
1054 lsize
= sbp
->sb_logstart
? sbp
->sb_logblocks
: 0;
1055 statp
->f_blocks
= sbp
->sb_dblocks
- lsize
;
1056 statp
->f_bfree
= statp
->f_bavail
=
1057 sbp
->sb_fdblocks
- XFS_ALLOC_SET_ASIDE(mp
);
1058 fakeinos
= statp
->f_bfree
<< sbp
->sb_inopblog
;
1060 MIN(sbp
->sb_icount
+ fakeinos
, (__uint64_t
)XFS_MAXINUMBER
);
1061 if (mp
->m_maxicount
)
1062 statp
->f_files
= min_t(typeof(statp
->f_files
),
1066 /* make sure statp->f_ffree does not underflow */
1067 ffree
= statp
->f_files
- (sbp
->sb_icount
- sbp
->sb_ifree
);
1068 statp
->f_ffree
= max_t(__int64_t
, ffree
, 0);
1070 spin_unlock(&mp
->m_sb_lock
);
1072 if ((ip
->i_d
.di_flags
& XFS_DIFLAG_PROJINHERIT
) &&
1073 ((mp
->m_qflags
& (XFS_PQUOTA_ACCT
|XFS_OQUOTA_ENFD
))) ==
1074 (XFS_PQUOTA_ACCT
|XFS_OQUOTA_ENFD
))
1075 xfs_qm_statvfs(ip
, statp
);
1080 xfs_save_resvblks(struct xfs_mount
*mp
)
1082 __uint64_t resblks
= 0;
1084 mp
->m_resblks_save
= mp
->m_resblks
;
1085 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1089 xfs_restore_resvblks(struct xfs_mount
*mp
)
1093 if (mp
->m_resblks_save
) {
1094 resblks
= mp
->m_resblks_save
;
1095 mp
->m_resblks_save
= 0;
1097 resblks
= xfs_default_resblks(mp
);
1099 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1104 struct super_block
*sb
,
1108 struct xfs_mount
*mp
= XFS_M(sb
);
1109 substring_t args
[MAX_OPT_ARGS
];
1113 while ((p
= strsep(&options
, ",")) != NULL
) {
1119 token
= match_token(p
, tokens
, args
);
1122 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
1125 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
1129 * Logically we would return an error here to prevent
1130 * users from believing they might have changed
1131 * mount options using remount which can't be changed.
1133 * But unfortunately mount(8) adds all options from
1134 * mtab and fstab to the mount arguments in some cases
1135 * so we can't blindly reject options, but have to
1136 * check for each specified option if it actually
1137 * differs from the currently set option and only
1138 * reject it if that's the case.
1140 * Until that is implemented we return success for
1141 * every remount request, and silently ignore all
1142 * options that we can't actually change.
1146 "mount option \"%s\" not supported for remount\n", p
);
1155 if ((mp
->m_flags
& XFS_MOUNT_RDONLY
) && !(*flags
& MS_RDONLY
)) {
1156 mp
->m_flags
&= ~XFS_MOUNT_RDONLY
;
1159 * If this is the first remount to writeable state we
1160 * might have some superblock changes to update.
1162 if (mp
->m_update_flags
) {
1163 error
= xfs_mount_log_sb(mp
, mp
->m_update_flags
);
1165 xfs_warn(mp
, "failed to write sb changes");
1168 mp
->m_update_flags
= 0;
1172 * Fill out the reserve pool if it is empty. Use the stashed
1173 * value if it is non-zero, otherwise go with the default.
1175 xfs_restore_resvblks(mp
);
1179 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
) && (*flags
& MS_RDONLY
)) {
1181 * After we have synced the data but before we sync the
1182 * metadata, we need to free up the reserve block pool so that
1183 * the used block count in the superblock on disk is correct at
1184 * the end of the remount. Stash the current reserve pool size
1185 * so that if we get remounted rw, we can return it to the same
1189 xfs_quiesce_data(mp
);
1190 xfs_save_resvblks(mp
);
1191 xfs_quiesce_attr(mp
);
1192 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
1199 * Second stage of a freeze. The data is already frozen so we only
1200 * need to take care of the metadata. Once that's done write a dummy
1201 * record to dirty the log in case of a crash while frozen.
1205 struct super_block
*sb
)
1207 struct xfs_mount
*mp
= XFS_M(sb
);
1209 xfs_save_resvblks(mp
);
1210 xfs_quiesce_attr(mp
);
1211 return -xfs_fs_log_dummy(mp
);
1216 struct super_block
*sb
)
1218 struct xfs_mount
*mp
= XFS_M(sb
);
1220 xfs_restore_resvblks(mp
);
1225 xfs_fs_show_options(
1227 struct dentry
*root
)
1229 return -xfs_showargs(XFS_M(root
->d_sb
), m
);
1233 * This function fills in xfs_mount_t fields based on mount args.
1234 * Note: the superblock _has_ now been read in.
1238 struct xfs_mount
*mp
)
1240 int ronly
= (mp
->m_flags
& XFS_MOUNT_RDONLY
);
1242 /* Fail a mount where the logbuf is smaller than the log stripe */
1243 if (xfs_sb_version_haslogv2(&mp
->m_sb
)) {
1244 if (mp
->m_logbsize
<= 0 &&
1245 mp
->m_sb
.sb_logsunit
> XLOG_BIG_RECORD_BSIZE
) {
1246 mp
->m_logbsize
= mp
->m_sb
.sb_logsunit
;
1247 } else if (mp
->m_logbsize
> 0 &&
1248 mp
->m_logbsize
< mp
->m_sb
.sb_logsunit
) {
1250 "logbuf size must be greater than or equal to log stripe size");
1251 return XFS_ERROR(EINVAL
);
1254 /* Fail a mount if the logbuf is larger than 32K */
1255 if (mp
->m_logbsize
> XLOG_BIG_RECORD_BSIZE
) {
1257 "logbuf size for version 1 logs must be 16K or 32K");
1258 return XFS_ERROR(EINVAL
);
1263 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1264 * told by noattr2 to turn it off
1266 if (xfs_sb_version_hasattr2(&mp
->m_sb
) &&
1267 !(mp
->m_flags
& XFS_MOUNT_NOATTR2
))
1268 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
1271 * prohibit r/w mounts of read-only filesystems
1273 if ((mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) && !ronly
) {
1275 "cannot mount a read-only filesystem as read-write");
1276 return XFS_ERROR(EROFS
);
1284 struct super_block
*sb
,
1289 struct xfs_mount
*mp
= NULL
;
1290 int flags
= 0, error
= ENOMEM
;
1292 mp
= kzalloc(sizeof(struct xfs_mount
), GFP_KERNEL
);
1296 spin_lock_init(&mp
->m_sb_lock
);
1297 mutex_init(&mp
->m_growlock
);
1298 atomic_set(&mp
->m_active_trans
, 0);
1303 error
= xfs_parseargs(mp
, (char *)data
);
1305 goto out_free_fsname
;
1307 sb_min_blocksize(sb
, BBSIZE
);
1308 sb
->s_xattr
= xfs_xattr_handlers
;
1309 sb
->s_export_op
= &xfs_export_operations
;
1310 #ifdef CONFIG_XFS_QUOTA
1311 sb
->s_qcop
= &xfs_quotactl_operations
;
1313 sb
->s_op
= &xfs_super_operations
;
1316 flags
|= XFS_MFSI_QUIET
;
1318 error
= xfs_open_devices(mp
);
1320 goto out_free_fsname
;
1322 error
= xfs_init_mount_workqueues(mp
);
1324 goto out_close_devices
;
1326 error
= xfs_icsb_init_counters(mp
);
1328 goto out_destroy_workqueues
;
1330 error
= xfs_readsb(mp
, flags
);
1332 goto out_destroy_counters
;
1334 error
= xfs_finish_flags(mp
);
1338 error
= xfs_setup_devices(mp
);
1342 error
= xfs_filestream_mount(mp
);
1347 * we must configure the block size in the superblock before we run the
1348 * full mount process as the mount process can lookup and cache inodes.
1349 * For the same reason we must also initialise the syncd and register
1350 * the inode cache shrinker so that inodes can be reclaimed during
1351 * operations like a quotacheck that iterate all inodes in the
1354 sb
->s_magic
= XFS_SB_MAGIC
;
1355 sb
->s_blocksize
= mp
->m_sb
.sb_blocksize
;
1356 sb
->s_blocksize_bits
= ffs(sb
->s_blocksize
) - 1;
1357 sb
->s_maxbytes
= xfs_max_file_offset(sb
->s_blocksize_bits
);
1358 sb
->s_max_links
= XFS_MAXLINK
;
1359 sb
->s_time_gran
= 1;
1360 set_posix_acl_flag(sb
);
1362 error
= xfs_syncd_init(mp
);
1364 goto out_filestream_unmount
;
1366 error
= xfs_mountfs(mp
);
1368 goto out_syncd_stop
;
1370 root
= igrab(VFS_I(mp
->m_rootip
));
1375 if (is_bad_inode(root
)) {
1379 sb
->s_root
= d_make_root(root
);
1388 out_filestream_unmount
:
1389 xfs_filestream_unmount(mp
);
1392 out_destroy_counters
:
1393 xfs_icsb_destroy_counters(mp
);
1394 out_destroy_workqueues
:
1395 xfs_destroy_mount_workqueues(mp
);
1397 xfs_close_devices(mp
);
1399 xfs_free_fsname(mp
);
1405 xfs_filestream_unmount(mp
);
1411 STATIC
struct dentry
*
1413 struct file_system_type
*fs_type
,
1415 const char *dev_name
,
1418 return mount_bdev(fs_type
, flags
, dev_name
, data
, xfs_fs_fill_super
);
1422 xfs_fs_nr_cached_objects(
1423 struct super_block
*sb
)
1425 return xfs_reclaim_inodes_count(XFS_M(sb
));
1429 xfs_fs_free_cached_objects(
1430 struct super_block
*sb
,
1433 xfs_reclaim_inodes_nr(XFS_M(sb
), nr_to_scan
);
1436 static const struct super_operations xfs_super_operations
= {
1437 .alloc_inode
= xfs_fs_alloc_inode
,
1438 .destroy_inode
= xfs_fs_destroy_inode
,
1439 .dirty_inode
= xfs_fs_dirty_inode
,
1440 .evict_inode
= xfs_fs_evict_inode
,
1441 .drop_inode
= xfs_fs_drop_inode
,
1442 .put_super
= xfs_fs_put_super
,
1443 .sync_fs
= xfs_fs_sync_fs
,
1444 .freeze_fs
= xfs_fs_freeze
,
1445 .unfreeze_fs
= xfs_fs_unfreeze
,
1446 .statfs
= xfs_fs_statfs
,
1447 .remount_fs
= xfs_fs_remount
,
1448 .show_options
= xfs_fs_show_options
,
1449 .nr_cached_objects
= xfs_fs_nr_cached_objects
,
1450 .free_cached_objects
= xfs_fs_free_cached_objects
,
1453 static struct file_system_type xfs_fs_type
= {
1454 .owner
= THIS_MODULE
,
1456 .mount
= xfs_fs_mount
,
1457 .kill_sb
= kill_block_super
,
1458 .fs_flags
= FS_REQUIRES_DEV
,
1462 xfs_init_zones(void)
1465 xfs_ioend_zone
= kmem_zone_init(sizeof(xfs_ioend_t
), "xfs_ioend");
1466 if (!xfs_ioend_zone
)
1469 xfs_ioend_pool
= mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE
,
1471 if (!xfs_ioend_pool
)
1472 goto out_destroy_ioend_zone
;
1474 xfs_log_ticket_zone
= kmem_zone_init(sizeof(xlog_ticket_t
),
1476 if (!xfs_log_ticket_zone
)
1477 goto out_destroy_ioend_pool
;
1479 xfs_bmap_free_item_zone
= kmem_zone_init(sizeof(xfs_bmap_free_item_t
),
1480 "xfs_bmap_free_item");
1481 if (!xfs_bmap_free_item_zone
)
1482 goto out_destroy_log_ticket_zone
;
1484 xfs_btree_cur_zone
= kmem_zone_init(sizeof(xfs_btree_cur_t
),
1486 if (!xfs_btree_cur_zone
)
1487 goto out_destroy_bmap_free_item_zone
;
1489 xfs_da_state_zone
= kmem_zone_init(sizeof(xfs_da_state_t
),
1491 if (!xfs_da_state_zone
)
1492 goto out_destroy_btree_cur_zone
;
1494 xfs_dabuf_zone
= kmem_zone_init(sizeof(xfs_dabuf_t
), "xfs_dabuf");
1495 if (!xfs_dabuf_zone
)
1496 goto out_destroy_da_state_zone
;
1498 xfs_ifork_zone
= kmem_zone_init(sizeof(xfs_ifork_t
), "xfs_ifork");
1499 if (!xfs_ifork_zone
)
1500 goto out_destroy_dabuf_zone
;
1502 xfs_trans_zone
= kmem_zone_init(sizeof(xfs_trans_t
), "xfs_trans");
1503 if (!xfs_trans_zone
)
1504 goto out_destroy_ifork_zone
;
1506 xfs_log_item_desc_zone
=
1507 kmem_zone_init(sizeof(struct xfs_log_item_desc
),
1508 "xfs_log_item_desc");
1509 if (!xfs_log_item_desc_zone
)
1510 goto out_destroy_trans_zone
;
1513 * The size of the zone allocated buf log item is the maximum
1514 * size possible under XFS. This wastes a little bit of memory,
1515 * but it is much faster.
1517 xfs_buf_item_zone
= kmem_zone_init((sizeof(xfs_buf_log_item_t
) +
1518 (((XFS_MAX_BLOCKSIZE
/ XFS_BLF_CHUNK
) /
1519 NBWORD
) * sizeof(int))), "xfs_buf_item");
1520 if (!xfs_buf_item_zone
)
1521 goto out_destroy_log_item_desc_zone
;
1523 xfs_efd_zone
= kmem_zone_init((sizeof(xfs_efd_log_item_t
) +
1524 ((XFS_EFD_MAX_FAST_EXTENTS
- 1) *
1525 sizeof(xfs_extent_t
))), "xfs_efd_item");
1527 goto out_destroy_buf_item_zone
;
1529 xfs_efi_zone
= kmem_zone_init((sizeof(xfs_efi_log_item_t
) +
1530 ((XFS_EFI_MAX_FAST_EXTENTS
- 1) *
1531 sizeof(xfs_extent_t
))), "xfs_efi_item");
1533 goto out_destroy_efd_zone
;
1536 kmem_zone_init_flags(sizeof(xfs_inode_t
), "xfs_inode",
1537 KM_ZONE_HWALIGN
| KM_ZONE_RECLAIM
| KM_ZONE_SPREAD
,
1538 xfs_fs_inode_init_once
);
1539 if (!xfs_inode_zone
)
1540 goto out_destroy_efi_zone
;
1543 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t
), "xfs_ili",
1544 KM_ZONE_SPREAD
, NULL
);
1546 goto out_destroy_inode_zone
;
1550 out_destroy_inode_zone
:
1551 kmem_zone_destroy(xfs_inode_zone
);
1552 out_destroy_efi_zone
:
1553 kmem_zone_destroy(xfs_efi_zone
);
1554 out_destroy_efd_zone
:
1555 kmem_zone_destroy(xfs_efd_zone
);
1556 out_destroy_buf_item_zone
:
1557 kmem_zone_destroy(xfs_buf_item_zone
);
1558 out_destroy_log_item_desc_zone
:
1559 kmem_zone_destroy(xfs_log_item_desc_zone
);
1560 out_destroy_trans_zone
:
1561 kmem_zone_destroy(xfs_trans_zone
);
1562 out_destroy_ifork_zone
:
1563 kmem_zone_destroy(xfs_ifork_zone
);
1564 out_destroy_dabuf_zone
:
1565 kmem_zone_destroy(xfs_dabuf_zone
);
1566 out_destroy_da_state_zone
:
1567 kmem_zone_destroy(xfs_da_state_zone
);
1568 out_destroy_btree_cur_zone
:
1569 kmem_zone_destroy(xfs_btree_cur_zone
);
1570 out_destroy_bmap_free_item_zone
:
1571 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1572 out_destroy_log_ticket_zone
:
1573 kmem_zone_destroy(xfs_log_ticket_zone
);
1574 out_destroy_ioend_pool
:
1575 mempool_destroy(xfs_ioend_pool
);
1576 out_destroy_ioend_zone
:
1577 kmem_zone_destroy(xfs_ioend_zone
);
1583 xfs_destroy_zones(void)
1585 kmem_zone_destroy(xfs_ili_zone
);
1586 kmem_zone_destroy(xfs_inode_zone
);
1587 kmem_zone_destroy(xfs_efi_zone
);
1588 kmem_zone_destroy(xfs_efd_zone
);
1589 kmem_zone_destroy(xfs_buf_item_zone
);
1590 kmem_zone_destroy(xfs_log_item_desc_zone
);
1591 kmem_zone_destroy(xfs_trans_zone
);
1592 kmem_zone_destroy(xfs_ifork_zone
);
1593 kmem_zone_destroy(xfs_dabuf_zone
);
1594 kmem_zone_destroy(xfs_da_state_zone
);
1595 kmem_zone_destroy(xfs_btree_cur_zone
);
1596 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1597 kmem_zone_destroy(xfs_log_ticket_zone
);
1598 mempool_destroy(xfs_ioend_pool
);
1599 kmem_zone_destroy(xfs_ioend_zone
);
1604 xfs_init_workqueues(void)
1607 * We never want to the same work item to run twice, reclaiming inodes
1608 * or idling the log is not going to get any faster by multiple CPUs
1609 * competing for ressources. Use the default large max_active value
1610 * so that even lots of filesystems can perform these task in parallel.
1612 xfs_syncd_wq
= alloc_workqueue("xfssyncd", WQ_NON_REENTRANT
, 0);
1617 * The allocation workqueue can be used in memory reclaim situations
1618 * (writepage path), and parallelism is only limited by the number of
1619 * AGs in all the filesystems mounted. Hence use the default large
1620 * max_active value for this workqueue.
1622 xfs_alloc_wq
= alloc_workqueue("xfsalloc", WQ_MEM_RECLAIM
, 0);
1624 goto out_destroy_syncd
;
1629 destroy_workqueue(xfs_syncd_wq
);
1634 xfs_destroy_workqueues(void)
1636 destroy_workqueue(xfs_alloc_wq
);
1637 destroy_workqueue(xfs_syncd_wq
);
1645 printk(KERN_INFO XFS_VERSION_STRING
" with "
1646 XFS_BUILD_OPTIONS
" enabled\n");
1650 error
= xfs_init_zones();
1654 error
= xfs_init_workqueues();
1656 goto out_destroy_zones
;
1658 error
= xfs_mru_cache_init();
1660 goto out_destroy_wq
;
1662 error
= xfs_filestream_init();
1664 goto out_mru_cache_uninit
;
1666 error
= xfs_buf_init();
1668 goto out_filestream_uninit
;
1670 error
= xfs_init_procfs();
1672 goto out_buf_terminate
;
1674 error
= xfs_sysctl_register();
1676 goto out_cleanup_procfs
;
1678 error
= xfs_qm_init();
1680 goto out_sysctl_unregister
;
1682 error
= register_filesystem(&xfs_fs_type
);
1689 out_sysctl_unregister
:
1690 xfs_sysctl_unregister();
1692 xfs_cleanup_procfs();
1694 xfs_buf_terminate();
1695 out_filestream_uninit
:
1696 xfs_filestream_uninit();
1697 out_mru_cache_uninit
:
1698 xfs_mru_cache_uninit();
1700 xfs_destroy_workqueues();
1702 xfs_destroy_zones();
1711 unregister_filesystem(&xfs_fs_type
);
1712 xfs_sysctl_unregister();
1713 xfs_cleanup_procfs();
1714 xfs_buf_terminate();
1715 xfs_filestream_uninit();
1716 xfs_mru_cache_uninit();
1717 xfs_destroy_workqueues();
1718 xfs_destroy_zones();
1721 module_init(init_xfs_fs
);
1722 module_exit(exit_xfs_fs
);
1724 MODULE_AUTHOR("Silicon Graphics, Inc.");
1725 MODULE_DESCRIPTION(XFS_VERSION_STRING
" with " XFS_BUILD_OPTIONS
" enabled");
1726 MODULE_LICENSE("GPL");