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_dmapi.h"
28 #include "xfs_quota.h"
29 #include "xfs_mount.h"
30 #include "xfs_bmap_btree.h"
31 #include "xfs_alloc_btree.h"
32 #include "xfs_ialloc_btree.h"
33 #include "xfs_dir2_sf.h"
34 #include "xfs_attr_sf.h"
35 #include "xfs_dinode.h"
36 #include "xfs_inode.h"
37 #include "xfs_btree.h"
38 #include "xfs_btree_trace.h"
39 #include "xfs_ialloc.h"
41 #include "xfs_rtalloc.h"
42 #include "xfs_error.h"
43 #include "xfs_itable.h"
44 #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_dir2_trace.h"
57 #include "xfs_extfree_item.h"
58 #include "xfs_mru_cache.h"
59 #include "xfs_inode_item.h"
62 #include <linux/namei.h>
63 #include <linux/init.h>
64 #include <linux/mount.h>
65 #include <linux/mempool.h>
66 #include <linux/writeback.h>
67 #include <linux/kthread.h>
68 #include <linux/freezer.h>
69 #include <linux/parser.h>
71 static struct super_operations xfs_super_operations
;
72 static kmem_zone_t
*xfs_ioend_zone
;
73 mempool_t
*xfs_ioend_pool
;
75 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
76 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
77 #define MNTOPT_LOGDEV "logdev" /* log device */
78 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
79 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
80 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
81 #define MNTOPT_INO64 "ino64" /* force inodes into 64-bit range */
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_INO64
)) {
295 mp
->m_flags
|= XFS_MOUNT_INO64
;
296 mp
->m_inoadd
= XFS_INO64_OFFSET
;
299 "XFS: %s option not allowed on this system",
303 } else if (!strcmp(this_char
, MNTOPT_NOALIGN
)) {
304 mp
->m_flags
|= XFS_MOUNT_NOALIGN
;
305 } else if (!strcmp(this_char
, MNTOPT_SWALLOC
)) {
306 mp
->m_flags
|= XFS_MOUNT_SWALLOC
;
307 } else if (!strcmp(this_char
, MNTOPT_SUNIT
)) {
308 if (!value
|| !*value
) {
310 "XFS: %s option requires an argument",
314 dsunit
= simple_strtoul(value
, &eov
, 10);
315 } else if (!strcmp(this_char
, MNTOPT_SWIDTH
)) {
316 if (!value
|| !*value
) {
318 "XFS: %s option requires an argument",
322 dswidth
= simple_strtoul(value
, &eov
, 10);
323 } else if (!strcmp(this_char
, MNTOPT_64BITINODE
)) {
324 mp
->m_flags
&= ~XFS_MOUNT_SMALL_INUMS
;
327 "XFS: %s option not allowed on this system",
331 } else if (!strcmp(this_char
, MNTOPT_NOUUID
)) {
332 mp
->m_flags
|= XFS_MOUNT_NOUUID
;
333 } else if (!strcmp(this_char
, MNTOPT_BARRIER
)) {
334 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
335 } else if (!strcmp(this_char
, MNTOPT_NOBARRIER
)) {
336 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
337 } else if (!strcmp(this_char
, MNTOPT_IKEEP
)) {
338 mp
->m_flags
|= XFS_MOUNT_IKEEP
;
339 } else if (!strcmp(this_char
, MNTOPT_NOIKEEP
)) {
340 dmapi_implies_ikeep
= 0;
341 mp
->m_flags
&= ~XFS_MOUNT_IKEEP
;
342 } else if (!strcmp(this_char
, MNTOPT_LARGEIO
)) {
343 mp
->m_flags
&= ~XFS_MOUNT_COMPAT_IOSIZE
;
344 } else if (!strcmp(this_char
, MNTOPT_NOLARGEIO
)) {
345 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
346 } else if (!strcmp(this_char
, MNTOPT_ATTR2
)) {
347 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
348 } else if (!strcmp(this_char
, MNTOPT_NOATTR2
)) {
349 mp
->m_flags
&= ~XFS_MOUNT_ATTR2
;
350 mp
->m_flags
|= XFS_MOUNT_NOATTR2
;
351 } else if (!strcmp(this_char
, MNTOPT_FILESTREAM
)) {
352 mp
->m_flags
|= XFS_MOUNT_FILESTREAMS
;
353 } else if (!strcmp(this_char
, MNTOPT_NOQUOTA
)) {
354 mp
->m_qflags
&= ~(XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
|
355 XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
|
356 XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
|
357 XFS_UQUOTA_ENFD
| XFS_OQUOTA_ENFD
);
358 } else if (!strcmp(this_char
, MNTOPT_QUOTA
) ||
359 !strcmp(this_char
, MNTOPT_UQUOTA
) ||
360 !strcmp(this_char
, MNTOPT_USRQUOTA
)) {
361 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
|
363 } else if (!strcmp(this_char
, MNTOPT_QUOTANOENF
) ||
364 !strcmp(this_char
, MNTOPT_UQUOTANOENF
)) {
365 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
);
366 mp
->m_qflags
&= ~XFS_UQUOTA_ENFD
;
367 } else if (!strcmp(this_char
, MNTOPT_PQUOTA
) ||
368 !strcmp(this_char
, MNTOPT_PRJQUOTA
)) {
369 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
|
371 } else if (!strcmp(this_char
, MNTOPT_PQUOTANOENF
)) {
372 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
);
373 mp
->m_qflags
&= ~XFS_OQUOTA_ENFD
;
374 } else if (!strcmp(this_char
, MNTOPT_GQUOTA
) ||
375 !strcmp(this_char
, MNTOPT_GRPQUOTA
)) {
376 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
|
378 } else if (!strcmp(this_char
, MNTOPT_GQUOTANOENF
)) {
379 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
);
380 mp
->m_qflags
&= ~XFS_OQUOTA_ENFD
;
381 } else if (!strcmp(this_char
, MNTOPT_DMAPI
)) {
382 mp
->m_flags
|= XFS_MOUNT_DMAPI
;
383 } else if (!strcmp(this_char
, MNTOPT_XDSM
)) {
384 mp
->m_flags
|= XFS_MOUNT_DMAPI
;
385 } else if (!strcmp(this_char
, MNTOPT_DMI
)) {
386 mp
->m_flags
|= XFS_MOUNT_DMAPI
;
387 } else if (!strcmp(this_char
, "ihashsize")) {
389 "XFS: ihashsize no longer used, option is deprecated.");
390 } else if (!strcmp(this_char
, "osyncisdsync")) {
391 /* no-op, this is now the default */
393 "XFS: osyncisdsync is now the default, option is deprecated.");
394 } else if (!strcmp(this_char
, "irixsgid")) {
396 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
399 "XFS: unknown mount option [%s].", this_char
);
405 * no recovery flag requires a read-only mount
407 if ((mp
->m_flags
& XFS_MOUNT_NORECOVERY
) &&
408 !(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
409 cmn_err(CE_WARN
, "XFS: no-recovery mounts must be read-only.");
413 if ((mp
->m_flags
& XFS_MOUNT_NOALIGN
) && (dsunit
|| dswidth
)) {
415 "XFS: sunit and swidth options incompatible with the noalign option");
419 if ((mp
->m_qflags
& (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
)) &&
420 (mp
->m_qflags
& (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
))) {
422 "XFS: cannot mount with both project and group quota");
426 if ((mp
->m_flags
& XFS_MOUNT_DMAPI
) && (!*mtpt
|| *mtpt
[0] == '\0')) {
427 printk("XFS: %s option needs the mount point option as well\n",
432 if ((dsunit
&& !dswidth
) || (!dsunit
&& dswidth
)) {
434 "XFS: sunit and swidth must be specified together");
438 if (dsunit
&& (dswidth
% dsunit
!= 0)) {
440 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
446 * Applications using DMI filesystems often expect the
447 * inode generation number to be monotonically increasing.
448 * If we delete inode chunks we break this assumption, so
449 * keep unused inode chunks on disk for DMI filesystems
450 * until we come up with a better solution.
451 * Note that if "ikeep" or "noikeep" mount options are
452 * supplied, then they are honored.
454 if ((mp
->m_flags
& XFS_MOUNT_DMAPI
) && dmapi_implies_ikeep
)
455 mp
->m_flags
|= XFS_MOUNT_IKEEP
;
458 if (!(mp
->m_flags
& XFS_MOUNT_NOALIGN
)) {
460 * At this point the superblock has not been read
461 * in, therefore we do not know the block size.
462 * Before the mount call ends we will convert
466 mp
->m_dalign
= dsunit
;
467 mp
->m_flags
|= XFS_MOUNT_RETERR
;
471 mp
->m_swidth
= dswidth
;
474 if (mp
->m_logbufs
!= -1 &&
475 mp
->m_logbufs
!= 0 &&
476 (mp
->m_logbufs
< XLOG_MIN_ICLOGS
||
477 mp
->m_logbufs
> XLOG_MAX_ICLOGS
)) {
479 "XFS: invalid logbufs value: %d [not %d-%d]",
480 mp
->m_logbufs
, XLOG_MIN_ICLOGS
, XLOG_MAX_ICLOGS
);
481 return XFS_ERROR(EINVAL
);
483 if (mp
->m_logbsize
!= -1 &&
484 mp
->m_logbsize
!= 0 &&
485 (mp
->m_logbsize
< XLOG_MIN_RECORD_BSIZE
||
486 mp
->m_logbsize
> XLOG_MAX_RECORD_BSIZE
||
487 !is_power_of_2(mp
->m_logbsize
))) {
489 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
491 return XFS_ERROR(EINVAL
);
494 mp
->m_fsname
= kstrndup(sb
->s_id
, MAXNAMELEN
, GFP_KERNEL
);
497 mp
->m_fsname_len
= strlen(mp
->m_fsname
) + 1;
500 if (iosizelog
> XFS_MAX_IO_LOG
||
501 iosizelog
< XFS_MIN_IO_LOG
) {
503 "XFS: invalid log iosize: %d [not %d-%d]",
504 iosizelog
, XFS_MIN_IO_LOG
,
506 return XFS_ERROR(EINVAL
);
509 mp
->m_flags
|= XFS_MOUNT_DFLT_IOSIZE
;
510 mp
->m_readio_log
= iosizelog
;
511 mp
->m_writeio_log
= iosizelog
;
517 struct proc_xfs_info
{
524 struct xfs_mount
*mp
,
527 static struct proc_xfs_info xfs_info_set
[] = {
528 /* the few simple ones we can get from the mount struct */
529 { XFS_MOUNT_IKEEP
, "," MNTOPT_IKEEP
},
530 { XFS_MOUNT_WSYNC
, "," MNTOPT_WSYNC
},
531 { XFS_MOUNT_INO64
, "," MNTOPT_INO64
},
532 { XFS_MOUNT_NOALIGN
, "," MNTOPT_NOALIGN
},
533 { XFS_MOUNT_SWALLOC
, "," MNTOPT_SWALLOC
},
534 { XFS_MOUNT_NOUUID
, "," MNTOPT_NOUUID
},
535 { XFS_MOUNT_NORECOVERY
, "," MNTOPT_NORECOVERY
},
536 { XFS_MOUNT_OSYNCISOSYNC
, "," MNTOPT_OSYNCISOSYNC
},
537 { XFS_MOUNT_ATTR2
, "," MNTOPT_ATTR2
},
538 { XFS_MOUNT_FILESTREAMS
, "," MNTOPT_FILESTREAM
},
539 { XFS_MOUNT_DMAPI
, "," MNTOPT_DMAPI
},
540 { XFS_MOUNT_GRPID
, "," MNTOPT_GRPID
},
543 static struct proc_xfs_info xfs_info_unset
[] = {
544 /* the few simple ones we can get from the mount struct */
545 { XFS_MOUNT_COMPAT_IOSIZE
, "," MNTOPT_LARGEIO
},
546 { XFS_MOUNT_BARRIER
, "," MNTOPT_NOBARRIER
},
547 { XFS_MOUNT_SMALL_INUMS
, "," MNTOPT_64BITINODE
},
550 struct proc_xfs_info
*xfs_infop
;
552 for (xfs_infop
= xfs_info_set
; xfs_infop
->flag
; xfs_infop
++) {
553 if (mp
->m_flags
& xfs_infop
->flag
)
554 seq_puts(m
, xfs_infop
->str
);
556 for (xfs_infop
= xfs_info_unset
; xfs_infop
->flag
; xfs_infop
++) {
557 if (!(mp
->m_flags
& xfs_infop
->flag
))
558 seq_puts(m
, xfs_infop
->str
);
561 if (mp
->m_flags
& XFS_MOUNT_DFLT_IOSIZE
)
562 seq_printf(m
, "," MNTOPT_ALLOCSIZE
"=%dk",
563 (int)(1 << mp
->m_writeio_log
) >> 10);
565 if (mp
->m_logbufs
> 0)
566 seq_printf(m
, "," MNTOPT_LOGBUFS
"=%d", mp
->m_logbufs
);
567 if (mp
->m_logbsize
> 0)
568 seq_printf(m
, "," MNTOPT_LOGBSIZE
"=%dk", mp
->m_logbsize
>> 10);
571 seq_printf(m
, "," MNTOPT_LOGDEV
"=%s", mp
->m_logname
);
573 seq_printf(m
, "," MNTOPT_RTDEV
"=%s", mp
->m_rtname
);
575 if (mp
->m_dalign
> 0)
576 seq_printf(m
, "," MNTOPT_SUNIT
"=%d",
577 (int)XFS_FSB_TO_BB(mp
, mp
->m_dalign
));
578 if (mp
->m_swidth
> 0)
579 seq_printf(m
, "," MNTOPT_SWIDTH
"=%d",
580 (int)XFS_FSB_TO_BB(mp
, mp
->m_swidth
));
582 if (mp
->m_qflags
& (XFS_UQUOTA_ACCT
|XFS_UQUOTA_ENFD
))
583 seq_puts(m
, "," MNTOPT_USRQUOTA
);
584 else if (mp
->m_qflags
& XFS_UQUOTA_ACCT
)
585 seq_puts(m
, "," MNTOPT_UQUOTANOENF
);
587 if (mp
->m_qflags
& (XFS_PQUOTA_ACCT
|XFS_OQUOTA_ENFD
))
588 seq_puts(m
, "," MNTOPT_PRJQUOTA
);
589 else if (mp
->m_qflags
& XFS_PQUOTA_ACCT
)
590 seq_puts(m
, "," MNTOPT_PQUOTANOENF
);
592 if (mp
->m_qflags
& (XFS_GQUOTA_ACCT
|XFS_OQUOTA_ENFD
))
593 seq_puts(m
, "," MNTOPT_GRPQUOTA
);
594 else if (mp
->m_qflags
& XFS_GQUOTA_ACCT
)
595 seq_puts(m
, "," MNTOPT_GQUOTANOENF
);
597 if (!(mp
->m_qflags
& XFS_ALL_QUOTA_ACCT
))
598 seq_puts(m
, "," MNTOPT_NOQUOTA
);
604 unsigned int blockshift
)
606 unsigned int pagefactor
= 1;
607 unsigned int bitshift
= BITS_PER_LONG
- 1;
609 /* Figure out maximum filesize, on Linux this can depend on
610 * the filesystem blocksize (on 32 bit platforms).
611 * __block_prepare_write does this in an [unsigned] long...
612 * page->index << (PAGE_CACHE_SHIFT - bbits)
613 * So, for page sized blocks (4K on 32 bit platforms),
614 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
615 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
616 * but for smaller blocksizes it is less (bbits = log2 bsize).
617 * Note1: get_block_t takes a long (implicit cast from above)
618 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
619 * can optionally convert the [unsigned] long from above into
620 * an [unsigned] long long.
623 #if BITS_PER_LONG == 32
624 # if defined(CONFIG_LBD)
625 ASSERT(sizeof(sector_t
) == 8);
626 pagefactor
= PAGE_CACHE_SIZE
;
627 bitshift
= BITS_PER_LONG
;
629 pagefactor
= PAGE_CACHE_SIZE
>> (PAGE_CACHE_SHIFT
- blockshift
);
633 return (((__uint64_t
)pagefactor
) << bitshift
) - 1;
640 struct block_device
**bdevp
)
644 *bdevp
= open_bdev_exclusive(name
, FMODE_READ
|FMODE_WRITE
, mp
);
645 if (IS_ERR(*bdevp
)) {
646 error
= PTR_ERR(*bdevp
);
647 printk("XFS: Invalid device [%s], error=%d\n", name
, error
);
655 struct block_device
*bdev
)
658 close_bdev_exclusive(bdev
, FMODE_READ
|FMODE_WRITE
);
662 * Try to write out the superblock using barriers.
668 xfs_buf_t
*sbp
= xfs_getsb(mp
, 0);
673 XFS_BUF_UNDELAYWRITE(sbp
);
675 XFS_BUF_UNASYNC(sbp
);
676 XFS_BUF_ORDERED(sbp
);
679 error
= xfs_iowait(sbp
);
682 * Clear all the flags we set and possible error state in the
683 * buffer. We only did the write to try out whether barriers
684 * worked and shouldn't leave any traces in the superblock
688 XFS_BUF_ERROR(sbp
, 0);
689 XFS_BUF_UNORDERED(sbp
);
696 xfs_mountfs_check_barriers(xfs_mount_t
*mp
)
700 if (mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
701 xfs_fs_cmn_err(CE_NOTE
, mp
,
702 "Disabling barriers, not supported with external log device");
703 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
707 if (xfs_readonly_buftarg(mp
->m_ddev_targp
)) {
708 xfs_fs_cmn_err(CE_NOTE
, mp
,
709 "Disabling barriers, underlying device is readonly");
710 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
714 error
= xfs_barrier_test(mp
);
716 xfs_fs_cmn_err(CE_NOTE
, mp
,
717 "Disabling barriers, trial barrier write failed");
718 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
724 xfs_blkdev_issue_flush(
725 xfs_buftarg_t
*buftarg
)
727 blkdev_issue_flush(buftarg
->bt_bdev
, NULL
);
732 struct xfs_mount
*mp
)
734 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
735 struct block_device
*logdev
= mp
->m_logdev_targp
->bt_bdev
;
736 xfs_free_buftarg(mp
, mp
->m_logdev_targp
);
737 xfs_blkdev_put(logdev
);
739 if (mp
->m_rtdev_targp
) {
740 struct block_device
*rtdev
= mp
->m_rtdev_targp
->bt_bdev
;
741 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
742 xfs_blkdev_put(rtdev
);
744 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
748 * The file system configurations are:
749 * (1) device (partition) with data and internal log
750 * (2) logical volume with data and log subvolumes.
751 * (3) logical volume with data, log, and realtime subvolumes.
753 * We only have to handle opening the log and realtime volumes here if
754 * they are present. The data subvolume has already been opened by
755 * get_sb_bdev() and is stored in sb->s_bdev.
759 struct xfs_mount
*mp
)
761 struct block_device
*ddev
= mp
->m_super
->s_bdev
;
762 struct block_device
*logdev
= NULL
, *rtdev
= NULL
;
766 * Open real time and log devices - order is important.
769 error
= xfs_blkdev_get(mp
, mp
->m_logname
, &logdev
);
775 error
= xfs_blkdev_get(mp
, mp
->m_rtname
, &rtdev
);
777 goto out_close_logdev
;
779 if (rtdev
== ddev
|| rtdev
== logdev
) {
781 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
783 goto out_close_rtdev
;
788 * Setup xfs_mount buffer target pointers
791 mp
->m_ddev_targp
= xfs_alloc_buftarg(ddev
, 0);
792 if (!mp
->m_ddev_targp
)
793 goto out_close_rtdev
;
796 mp
->m_rtdev_targp
= xfs_alloc_buftarg(rtdev
, 1);
797 if (!mp
->m_rtdev_targp
)
798 goto out_free_ddev_targ
;
801 if (logdev
&& logdev
!= ddev
) {
802 mp
->m_logdev_targp
= xfs_alloc_buftarg(logdev
, 1);
803 if (!mp
->m_logdev_targp
)
804 goto out_free_rtdev_targ
;
806 mp
->m_logdev_targp
= mp
->m_ddev_targp
;
812 if (mp
->m_rtdev_targp
)
813 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
815 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
818 xfs_blkdev_put(rtdev
);
820 if (logdev
&& logdev
!= ddev
)
821 xfs_blkdev_put(logdev
);
827 * Setup xfs_mount buffer target pointers based on superblock
831 struct xfs_mount
*mp
)
835 error
= xfs_setsize_buftarg(mp
->m_ddev_targp
, mp
->m_sb
.sb_blocksize
,
836 mp
->m_sb
.sb_sectsize
);
840 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
841 unsigned int log_sector_size
= BBSIZE
;
843 if (xfs_sb_version_hassector(&mp
->m_sb
))
844 log_sector_size
= mp
->m_sb
.sb_logsectsize
;
845 error
= xfs_setsize_buftarg(mp
->m_logdev_targp
,
846 mp
->m_sb
.sb_blocksize
,
851 if (mp
->m_rtdev_targp
) {
852 error
= xfs_setsize_buftarg(mp
->m_rtdev_targp
,
853 mp
->m_sb
.sb_blocksize
,
854 mp
->m_sb
.sb_sectsize
);
863 * XFS AIL push thread support
867 struct xfs_ail
*ailp
,
868 xfs_lsn_t threshold_lsn
)
870 ailp
->xa_target
= threshold_lsn
;
871 wake_up_process(ailp
->xa_task
);
878 struct xfs_ail
*ailp
= data
;
879 xfs_lsn_t last_pushed_lsn
= 0;
882 while (!kthread_should_stop()) {
884 schedule_timeout_interruptible(msecs_to_jiffies(tout
));
890 ASSERT(ailp
->xa_mount
->m_log
);
891 if (XFS_FORCED_SHUTDOWN(ailp
->xa_mount
))
894 tout
= xfsaild_push(ailp
, &last_pushed_lsn
);
902 struct xfs_ail
*ailp
)
905 ailp
->xa_task
= kthread_run(xfsaild
, ailp
, "xfsaild");
906 if (IS_ERR(ailp
->xa_task
))
907 return -PTR_ERR(ailp
->xa_task
);
913 struct xfs_ail
*ailp
)
915 kthread_stop(ailp
->xa_task
);
919 /* Catch misguided souls that try to use this interface on XFS */
920 STATIC
struct inode
*
922 struct super_block
*sb
)
929 * Now that the generic code is guaranteed not to be accessing
930 * the linux inode, we can reclaim the inode.
933 xfs_fs_destroy_inode(
936 xfs_inode_t
*ip
= XFS_I(inode
);
938 XFS_STATS_INC(vn_reclaim
);
940 panic("%s: cannot reclaim 0x%p\n", __func__
, inode
);
944 * Slab object creation initialisation for the XFS inode.
945 * This covers only the idempotent fields in the XFS inode;
946 * all other fields need to be initialised on allocation
947 * from the slab. This avoids the need to repeatedly intialise
948 * fields in the xfs inode that left in the initialise state
949 * when freeing the inode.
952 xfs_fs_inode_init_once(
955 struct xfs_inode
*ip
= inode
;
957 memset(ip
, 0, sizeof(struct xfs_inode
));
960 inode_init_once(VFS_I(ip
));
963 atomic_set(&ip
->i_iocount
, 0);
964 atomic_set(&ip
->i_pincount
, 0);
965 spin_lock_init(&ip
->i_flags_lock
);
966 init_waitqueue_head(&ip
->i_ipin_wait
);
968 * Because we want to use a counting completion, complete
969 * the flush completion once to allow a single access to
970 * the flush completion without blocking.
972 init_completion(&ip
->i_flush
);
973 complete(&ip
->i_flush
);
975 mrlock_init(&ip
->i_lock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
976 "xfsino", ip
->i_ino
);
977 mrlock_init(&ip
->i_iolock
, MRLOCK_BARRIER
, "xfsio", ip
->i_ino
);
981 * Attempt to flush the inode, this will actually fail
982 * if the inode is pinned, but we dirty the inode again
983 * at the point when it is unpinned after a log write,
984 * since this is when the inode itself becomes flushable.
991 struct xfs_inode
*ip
= XFS_I(inode
);
992 struct xfs_mount
*mp
= ip
->i_mount
;
995 xfs_itrace_entry(ip
);
997 if (XFS_FORCED_SHUTDOWN(mp
))
998 return XFS_ERROR(EIO
);
1001 error
= xfs_wait_on_pages(ip
, 0, -1);
1007 * Bypass inodes which have already been cleaned by
1008 * the inode flush clustering code inside xfs_iflush
1010 if (xfs_inode_clean(ip
))
1014 * We make this non-blocking if the inode is contended, return
1015 * EAGAIN to indicate to the caller that they did not succeed.
1016 * This prevents the flush path from blocking on inodes inside
1017 * another operation right now, they get caught later by xfs_sync.
1020 xfs_ilock(ip
, XFS_ILOCK_SHARED
);
1023 error
= xfs_iflush(ip
, XFS_IFLUSH_SYNC
);
1026 if (!xfs_ilock_nowait(ip
, XFS_ILOCK_SHARED
))
1028 if (xfs_ipincount(ip
) || !xfs_iflock_nowait(ip
))
1031 error
= xfs_iflush(ip
, XFS_IFLUSH_ASYNC_NOBLOCK
);
1035 xfs_iunlock(ip
, XFS_ILOCK_SHARED
);
1038 * if we failed to write out the inode then mark
1039 * it dirty again so we'll try again later.
1042 xfs_mark_inode_dirty_sync(ip
);
1048 struct inode
*inode
)
1050 xfs_inode_t
*ip
= XFS_I(inode
);
1052 xfs_itrace_entry(ip
);
1053 XFS_STATS_INC(vn_rele
);
1054 XFS_STATS_INC(vn_remove
);
1055 XFS_STATS_DEC(vn_active
);
1062 struct xfs_mount
*mp
)
1064 kfree(mp
->m_fsname
);
1065 kfree(mp
->m_rtname
);
1066 kfree(mp
->m_logname
);
1071 struct super_block
*sb
)
1073 struct xfs_mount
*mp
= XFS_M(sb
);
1074 struct xfs_inode
*rip
= mp
->m_rootip
;
1075 int unmount_event_flags
= 0;
1078 xfs_sync_inodes(mp
, SYNC_ATTR
|SYNC_DELWRI
);
1081 if (mp
->m_flags
& XFS_MOUNT_DMAPI
) {
1082 unmount_event_flags
=
1083 (mp
->m_dmevmask
& (1 << DM_EVENT_UNMOUNT
)) ?
1084 0 : DM_FLAGS_UNWANTED
;
1086 * Ignore error from dmapi here, first unmount is not allowed
1087 * to fail anyway, and second we wouldn't want to fail a
1088 * unmount because of dmapi.
1090 XFS_SEND_PREUNMOUNT(mp
, rip
, DM_RIGHT_NULL
, rip
, DM_RIGHT_NULL
,
1091 NULL
, NULL
, 0, 0, unmount_event_flags
);
1096 * Blow away any referenced inode in the filestreams cache.
1097 * This can and will cause log traffic as inodes go inactive
1100 xfs_filestream_unmount(mp
);
1102 XFS_bflush(mp
->m_ddev_targp
);
1104 if (mp
->m_flags
& XFS_MOUNT_DMAPI
) {
1105 XFS_SEND_UNMOUNT(mp
, rip
, DM_RIGHT_NULL
, 0, 0,
1106 unmount_event_flags
);
1111 xfs_icsb_destroy_counters(mp
);
1112 xfs_close_devices(mp
);
1115 xfs_free_fsname(mp
);
1121 struct super_block
*sb
)
1123 if (!(sb
->s_flags
& MS_RDONLY
))
1124 xfs_sync_fsdata(XFS_M(sb
), 0);
1130 struct super_block
*sb
,
1133 struct xfs_mount
*mp
= XFS_M(sb
);
1137 * Treat a sync operation like a freeze. This is to work
1138 * around a race in sync_inodes() which works in two phases
1139 * - an asynchronous flush, which can write out an inode
1140 * without waiting for file size updates to complete, and a
1141 * synchronous flush, which wont do anything because the
1142 * async flush removed the inode's dirty flag. Also
1143 * sync_inodes() will not see any files that just have
1144 * outstanding transactions to be flushed because we don't
1145 * dirty the Linux inode until after the transaction I/O
1148 if (wait
|| unlikely(sb
->s_frozen
== SB_FREEZE_WRITE
))
1149 error
= xfs_quiesce_data(mp
);
1151 error
= xfs_sync_fsdata(mp
, 0);
1154 if (unlikely(laptop_mode
)) {
1155 int prev_sync_seq
= mp
->m_sync_seq
;
1158 * The disk must be active because we're syncing.
1159 * We schedule xfssyncd now (now that the disk is
1160 * active) instead of later (when it might not be).
1162 wake_up_process(mp
->m_sync_task
);
1164 * We have to wait for the sync iteration to complete.
1165 * If we don't, the disk activity caused by the sync
1166 * will come after the sync is completed, and that
1167 * triggers another sync from laptop mode.
1169 wait_event(mp
->m_wait_single_sync_task
,
1170 mp
->m_sync_seq
!= prev_sync_seq
);
1178 struct dentry
*dentry
,
1179 struct kstatfs
*statp
)
1181 struct xfs_mount
*mp
= XFS_M(dentry
->d_sb
);
1182 xfs_sb_t
*sbp
= &mp
->m_sb
;
1183 __uint64_t fakeinos
, id
;
1186 statp
->f_type
= XFS_SB_MAGIC
;
1187 statp
->f_namelen
= MAXNAMELEN
- 1;
1189 id
= huge_encode_dev(mp
->m_ddev_targp
->bt_dev
);
1190 statp
->f_fsid
.val
[0] = (u32
)id
;
1191 statp
->f_fsid
.val
[1] = (u32
)(id
>> 32);
1193 xfs_icsb_sync_counters(mp
, XFS_ICSB_LAZY_COUNT
);
1195 spin_lock(&mp
->m_sb_lock
);
1196 statp
->f_bsize
= sbp
->sb_blocksize
;
1197 lsize
= sbp
->sb_logstart
? sbp
->sb_logblocks
: 0;
1198 statp
->f_blocks
= sbp
->sb_dblocks
- lsize
;
1199 statp
->f_bfree
= statp
->f_bavail
=
1200 sbp
->sb_fdblocks
- XFS_ALLOC_SET_ASIDE(mp
);
1201 fakeinos
= statp
->f_bfree
<< sbp
->sb_inopblog
;
1203 fakeinos
+= mp
->m_inoadd
;
1206 MIN(sbp
->sb_icount
+ fakeinos
, (__uint64_t
)XFS_MAXINUMBER
);
1207 if (mp
->m_maxicount
)
1211 statp
->f_files
= min_t(typeof(statp
->f_files
),
1214 statp
->f_ffree
= statp
->f_files
- (sbp
->sb_icount
- sbp
->sb_ifree
);
1215 spin_unlock(&mp
->m_sb_lock
);
1217 XFS_QM_DQSTATVFS(XFS_I(dentry
->d_inode
), statp
);
1223 struct super_block
*sb
,
1227 struct xfs_mount
*mp
= XFS_M(sb
);
1228 substring_t args
[MAX_OPT_ARGS
];
1232 while ((p
= strsep(&options
, ",")) != NULL
) {
1238 token
= match_token(p
, tokens
, args
);
1241 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
1244 * Test if barriers are actually working if we can,
1245 * else delay this check until the filesystem is
1248 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
))
1249 xfs_mountfs_check_barriers(mp
);
1252 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
1256 * Logically we would return an error here to prevent
1257 * users from believing they might have changed
1258 * mount options using remount which can't be changed.
1260 * But unfortunately mount(8) adds all options from
1261 * mtab and fstab to the mount arguments in some cases
1262 * so we can't blindly reject options, but have to
1263 * check for each specified option if it actually
1264 * differs from the currently set option and only
1265 * reject it if that's the case.
1267 * Until that is implemented we return success for
1268 * every remount request, and silently ignore all
1269 * options that we can't actually change.
1273 "XFS: mount option \"%s\" not supported for remount\n", p
);
1282 if ((mp
->m_flags
& XFS_MOUNT_RDONLY
) && !(*flags
& MS_RDONLY
)) {
1283 mp
->m_flags
&= ~XFS_MOUNT_RDONLY
;
1284 if (mp
->m_flags
& XFS_MOUNT_BARRIER
)
1285 xfs_mountfs_check_barriers(mp
);
1288 * If this is the first remount to writeable state we
1289 * might have some superblock changes to update.
1291 if (mp
->m_update_flags
) {
1292 error
= xfs_mount_log_sb(mp
, mp
->m_update_flags
);
1295 "XFS: failed to write sb changes");
1298 mp
->m_update_flags
= 0;
1303 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
) && (*flags
& MS_RDONLY
)) {
1304 xfs_quiesce_data(mp
);
1305 xfs_quiesce_attr(mp
);
1306 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
1313 * Second stage of a freeze. The data is already frozen so we only
1314 * need to take care of the metadata. Once that's done write a dummy
1315 * record to dirty the log in case of a crash while frozen.
1319 struct super_block
*sb
)
1321 struct xfs_mount
*mp
= XFS_M(sb
);
1323 xfs_quiesce_attr(mp
);
1324 return -xfs_fs_log_dummy(mp
);
1328 xfs_fs_show_options(
1330 struct vfsmount
*mnt
)
1332 return -xfs_showargs(XFS_M(mnt
->mnt_sb
), m
);
1336 * This function fills in xfs_mount_t fields based on mount args.
1337 * Note: the superblock _has_ now been read in.
1341 struct xfs_mount
*mp
)
1343 int ronly
= (mp
->m_flags
& XFS_MOUNT_RDONLY
);
1345 /* Fail a mount where the logbuf is smaller than the log stripe */
1346 if (xfs_sb_version_haslogv2(&mp
->m_sb
)) {
1347 if (mp
->m_logbsize
<= 0 &&
1348 mp
->m_sb
.sb_logsunit
> XLOG_BIG_RECORD_BSIZE
) {
1349 mp
->m_logbsize
= mp
->m_sb
.sb_logsunit
;
1350 } else if (mp
->m_logbsize
> 0 &&
1351 mp
->m_logbsize
< mp
->m_sb
.sb_logsunit
) {
1353 "XFS: logbuf size must be greater than or equal to log stripe size");
1354 return XFS_ERROR(EINVAL
);
1357 /* Fail a mount if the logbuf is larger than 32K */
1358 if (mp
->m_logbsize
> XLOG_BIG_RECORD_BSIZE
) {
1360 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1361 return XFS_ERROR(EINVAL
);
1366 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1367 * told by noattr2 to turn it off
1369 if (xfs_sb_version_hasattr2(&mp
->m_sb
) &&
1370 !(mp
->m_flags
& XFS_MOUNT_NOATTR2
))
1371 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
1374 * prohibit r/w mounts of read-only filesystems
1376 if ((mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) && !ronly
) {
1378 "XFS: cannot mount a read-only filesystem as read-write");
1379 return XFS_ERROR(EROFS
);
1387 struct super_block
*sb
,
1392 struct xfs_mount
*mp
= NULL
;
1393 int flags
= 0, error
= ENOMEM
;
1396 mp
= kzalloc(sizeof(struct xfs_mount
), GFP_KERNEL
);
1400 spin_lock_init(&mp
->m_sb_lock
);
1401 mutex_init(&mp
->m_growlock
);
1402 atomic_set(&mp
->m_active_trans
, 0);
1403 INIT_LIST_HEAD(&mp
->m_sync_list
);
1404 spin_lock_init(&mp
->m_sync_lock
);
1405 init_waitqueue_head(&mp
->m_wait_single_sync_task
);
1410 error
= xfs_parseargs(mp
, (char *)data
, &mtpt
);
1412 goto out_free_fsname
;
1414 sb_min_blocksize(sb
, BBSIZE
);
1415 sb
->s_xattr
= xfs_xattr_handlers
;
1416 sb
->s_export_op
= &xfs_export_operations
;
1417 #ifdef CONFIG_XFS_QUOTA
1418 sb
->s_qcop
= &xfs_quotactl_operations
;
1420 sb
->s_op
= &xfs_super_operations
;
1422 error
= xfs_dmops_get(mp
);
1424 goto out_free_fsname
;
1425 error
= xfs_qmops_get(mp
);
1430 flags
|= XFS_MFSI_QUIET
;
1432 error
= xfs_open_devices(mp
);
1436 if (xfs_icsb_init_counters(mp
))
1437 mp
->m_flags
|= XFS_MOUNT_NO_PERCPU_SB
;
1439 error
= xfs_readsb(mp
, flags
);
1441 goto out_destroy_counters
;
1443 error
= xfs_finish_flags(mp
);
1447 error
= xfs_setup_devices(mp
);
1451 if (mp
->m_flags
& XFS_MOUNT_BARRIER
)
1452 xfs_mountfs_check_barriers(mp
);
1454 error
= xfs_filestream_mount(mp
);
1458 error
= xfs_mountfs(mp
);
1460 goto out_filestream_unmount
;
1462 XFS_SEND_MOUNT(mp
, DM_RIGHT_NULL
, mtpt
, mp
->m_fsname
);
1465 sb
->s_magic
= XFS_SB_MAGIC
;
1466 sb
->s_blocksize
= mp
->m_sb
.sb_blocksize
;
1467 sb
->s_blocksize_bits
= ffs(sb
->s_blocksize
) - 1;
1468 sb
->s_maxbytes
= xfs_max_file_offset(sb
->s_blocksize_bits
);
1469 sb
->s_time_gran
= 1;
1470 set_posix_acl_flag(sb
);
1472 root
= igrab(VFS_I(mp
->m_rootip
));
1477 if (is_bad_inode(root
)) {
1481 sb
->s_root
= d_alloc_root(root
);
1487 error
= xfs_syncd_init(mp
);
1493 xfs_itrace_exit(XFS_I(sb
->s_root
->d_inode
));
1496 out_filestream_unmount
:
1497 xfs_filestream_unmount(mp
);
1500 out_destroy_counters
:
1501 xfs_icsb_destroy_counters(mp
);
1502 xfs_close_devices(mp
);
1508 xfs_free_fsname(mp
);
1524 * Blow away any referenced inode in the filestreams cache.
1525 * This can and will cause log traffic as inodes go inactive
1528 xfs_filestream_unmount(mp
);
1530 XFS_bflush(mp
->m_ddev_targp
);
1538 struct file_system_type
*fs_type
,
1540 const char *dev_name
,
1542 struct vfsmount
*mnt
)
1544 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, xfs_fs_fill_super
,
1548 static struct super_operations xfs_super_operations
= {
1549 .alloc_inode
= xfs_fs_alloc_inode
,
1550 .destroy_inode
= xfs_fs_destroy_inode
,
1551 .write_inode
= xfs_fs_write_inode
,
1552 .clear_inode
= xfs_fs_clear_inode
,
1553 .put_super
= xfs_fs_put_super
,
1554 .write_super
= xfs_fs_write_super
,
1555 .sync_fs
= xfs_fs_sync_super
,
1556 .freeze_fs
= xfs_fs_freeze
,
1557 .statfs
= xfs_fs_statfs
,
1558 .remount_fs
= xfs_fs_remount
,
1559 .show_options
= xfs_fs_show_options
,
1562 static struct file_system_type xfs_fs_type
= {
1563 .owner
= THIS_MODULE
,
1565 .get_sb
= xfs_fs_get_sb
,
1566 .kill_sb
= kill_block_super
,
1567 .fs_flags
= FS_REQUIRES_DEV
,
1571 xfs_alloc_trace_bufs(void)
1573 #ifdef XFS_ALLOC_TRACE
1574 xfs_alloc_trace_buf
= ktrace_alloc(XFS_ALLOC_TRACE_SIZE
, KM_MAYFAIL
);
1575 if (!xfs_alloc_trace_buf
)
1578 #ifdef XFS_BMAP_TRACE
1579 xfs_bmap_trace_buf
= ktrace_alloc(XFS_BMAP_TRACE_SIZE
, KM_MAYFAIL
);
1580 if (!xfs_bmap_trace_buf
)
1581 goto out_free_alloc_trace
;
1583 #ifdef XFS_BTREE_TRACE
1584 xfs_allocbt_trace_buf
= ktrace_alloc(XFS_ALLOCBT_TRACE_SIZE
,
1586 if (!xfs_allocbt_trace_buf
)
1587 goto out_free_bmap_trace
;
1589 xfs_inobt_trace_buf
= ktrace_alloc(XFS_INOBT_TRACE_SIZE
, KM_MAYFAIL
);
1590 if (!xfs_inobt_trace_buf
)
1591 goto out_free_allocbt_trace
;
1593 xfs_bmbt_trace_buf
= ktrace_alloc(XFS_BMBT_TRACE_SIZE
, KM_MAYFAIL
);
1594 if (!xfs_bmbt_trace_buf
)
1595 goto out_free_inobt_trace
;
1597 #ifdef XFS_ATTR_TRACE
1598 xfs_attr_trace_buf
= ktrace_alloc(XFS_ATTR_TRACE_SIZE
, KM_MAYFAIL
);
1599 if (!xfs_attr_trace_buf
)
1600 goto out_free_bmbt_trace
;
1602 #ifdef XFS_DIR2_TRACE
1603 xfs_dir2_trace_buf
= ktrace_alloc(XFS_DIR2_GTRACE_SIZE
, KM_MAYFAIL
);
1604 if (!xfs_dir2_trace_buf
)
1605 goto out_free_attr_trace
;
1610 #ifdef XFS_DIR2_TRACE
1611 out_free_attr_trace
:
1613 #ifdef XFS_ATTR_TRACE
1614 ktrace_free(xfs_attr_trace_buf
);
1615 out_free_bmbt_trace
:
1617 #ifdef XFS_BTREE_TRACE
1618 ktrace_free(xfs_bmbt_trace_buf
);
1619 out_free_inobt_trace
:
1620 ktrace_free(xfs_inobt_trace_buf
);
1621 out_free_allocbt_trace
:
1622 ktrace_free(xfs_allocbt_trace_buf
);
1623 out_free_bmap_trace
:
1625 #ifdef XFS_BMAP_TRACE
1626 ktrace_free(xfs_bmap_trace_buf
);
1627 out_free_alloc_trace
:
1629 #ifdef XFS_ALLOC_TRACE
1630 ktrace_free(xfs_alloc_trace_buf
);
1637 xfs_free_trace_bufs(void)
1639 #ifdef XFS_DIR2_TRACE
1640 ktrace_free(xfs_dir2_trace_buf
);
1642 #ifdef XFS_ATTR_TRACE
1643 ktrace_free(xfs_attr_trace_buf
);
1645 #ifdef XFS_BTREE_TRACE
1646 ktrace_free(xfs_bmbt_trace_buf
);
1647 ktrace_free(xfs_inobt_trace_buf
);
1648 ktrace_free(xfs_allocbt_trace_buf
);
1650 #ifdef XFS_BMAP_TRACE
1651 ktrace_free(xfs_bmap_trace_buf
);
1653 #ifdef XFS_ALLOC_TRACE
1654 ktrace_free(xfs_alloc_trace_buf
);
1659 xfs_init_zones(void)
1662 xfs_ioend_zone
= kmem_zone_init(sizeof(xfs_ioend_t
), "xfs_ioend");
1663 if (!xfs_ioend_zone
)
1666 xfs_ioend_pool
= mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE
,
1668 if (!xfs_ioend_pool
)
1669 goto out_destroy_ioend_zone
;
1671 xfs_log_ticket_zone
= kmem_zone_init(sizeof(xlog_ticket_t
),
1673 if (!xfs_log_ticket_zone
)
1674 goto out_destroy_ioend_pool
;
1676 xfs_bmap_free_item_zone
= kmem_zone_init(sizeof(xfs_bmap_free_item_t
),
1677 "xfs_bmap_free_item");
1678 if (!xfs_bmap_free_item_zone
)
1679 goto out_destroy_log_ticket_zone
;
1681 xfs_btree_cur_zone
= kmem_zone_init(sizeof(xfs_btree_cur_t
),
1683 if (!xfs_btree_cur_zone
)
1684 goto out_destroy_bmap_free_item_zone
;
1686 xfs_da_state_zone
= kmem_zone_init(sizeof(xfs_da_state_t
),
1688 if (!xfs_da_state_zone
)
1689 goto out_destroy_btree_cur_zone
;
1691 xfs_dabuf_zone
= kmem_zone_init(sizeof(xfs_dabuf_t
), "xfs_dabuf");
1692 if (!xfs_dabuf_zone
)
1693 goto out_destroy_da_state_zone
;
1695 xfs_ifork_zone
= kmem_zone_init(sizeof(xfs_ifork_t
), "xfs_ifork");
1696 if (!xfs_ifork_zone
)
1697 goto out_destroy_dabuf_zone
;
1699 xfs_trans_zone
= kmem_zone_init(sizeof(xfs_trans_t
), "xfs_trans");
1700 if (!xfs_trans_zone
)
1701 goto out_destroy_ifork_zone
;
1704 * The size of the zone allocated buf log item is the maximum
1705 * size possible under XFS. This wastes a little bit of memory,
1706 * but it is much faster.
1708 xfs_buf_item_zone
= kmem_zone_init((sizeof(xfs_buf_log_item_t
) +
1709 (((XFS_MAX_BLOCKSIZE
/ XFS_BLI_CHUNK
) /
1710 NBWORD
) * sizeof(int))), "xfs_buf_item");
1711 if (!xfs_buf_item_zone
)
1712 goto out_destroy_trans_zone
;
1714 xfs_efd_zone
= kmem_zone_init((sizeof(xfs_efd_log_item_t
) +
1715 ((XFS_EFD_MAX_FAST_EXTENTS
- 1) *
1716 sizeof(xfs_extent_t
))), "xfs_efd_item");
1718 goto out_destroy_buf_item_zone
;
1720 xfs_efi_zone
= kmem_zone_init((sizeof(xfs_efi_log_item_t
) +
1721 ((XFS_EFI_MAX_FAST_EXTENTS
- 1) *
1722 sizeof(xfs_extent_t
))), "xfs_efi_item");
1724 goto out_destroy_efd_zone
;
1727 kmem_zone_init_flags(sizeof(xfs_inode_t
), "xfs_inode",
1728 KM_ZONE_HWALIGN
| KM_ZONE_RECLAIM
| KM_ZONE_SPREAD
,
1729 xfs_fs_inode_init_once
);
1730 if (!xfs_inode_zone
)
1731 goto out_destroy_efi_zone
;
1734 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t
), "xfs_ili",
1735 KM_ZONE_SPREAD
, NULL
);
1737 goto out_destroy_inode_zone
;
1739 #ifdef CONFIG_XFS_POSIX_ACL
1740 xfs_acl_zone
= kmem_zone_init(sizeof(xfs_acl_t
), "xfs_acl");
1742 goto out_destroy_ili_zone
;
1747 #ifdef CONFIG_XFS_POSIX_ACL
1748 out_destroy_ili_zone
:
1750 kmem_zone_destroy(xfs_ili_zone
);
1751 out_destroy_inode_zone
:
1752 kmem_zone_destroy(xfs_inode_zone
);
1753 out_destroy_efi_zone
:
1754 kmem_zone_destroy(xfs_efi_zone
);
1755 out_destroy_efd_zone
:
1756 kmem_zone_destroy(xfs_efd_zone
);
1757 out_destroy_buf_item_zone
:
1758 kmem_zone_destroy(xfs_buf_item_zone
);
1759 out_destroy_trans_zone
:
1760 kmem_zone_destroy(xfs_trans_zone
);
1761 out_destroy_ifork_zone
:
1762 kmem_zone_destroy(xfs_ifork_zone
);
1763 out_destroy_dabuf_zone
:
1764 kmem_zone_destroy(xfs_dabuf_zone
);
1765 out_destroy_da_state_zone
:
1766 kmem_zone_destroy(xfs_da_state_zone
);
1767 out_destroy_btree_cur_zone
:
1768 kmem_zone_destroy(xfs_btree_cur_zone
);
1769 out_destroy_bmap_free_item_zone
:
1770 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1771 out_destroy_log_ticket_zone
:
1772 kmem_zone_destroy(xfs_log_ticket_zone
);
1773 out_destroy_ioend_pool
:
1774 mempool_destroy(xfs_ioend_pool
);
1775 out_destroy_ioend_zone
:
1776 kmem_zone_destroy(xfs_ioend_zone
);
1782 xfs_destroy_zones(void)
1784 #ifdef CONFIG_XFS_POSIX_ACL
1785 kmem_zone_destroy(xfs_acl_zone
);
1787 kmem_zone_destroy(xfs_ili_zone
);
1788 kmem_zone_destroy(xfs_inode_zone
);
1789 kmem_zone_destroy(xfs_efi_zone
);
1790 kmem_zone_destroy(xfs_efd_zone
);
1791 kmem_zone_destroy(xfs_buf_item_zone
);
1792 kmem_zone_destroy(xfs_trans_zone
);
1793 kmem_zone_destroy(xfs_ifork_zone
);
1794 kmem_zone_destroy(xfs_dabuf_zone
);
1795 kmem_zone_destroy(xfs_da_state_zone
);
1796 kmem_zone_destroy(xfs_btree_cur_zone
);
1797 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1798 kmem_zone_destroy(xfs_log_ticket_zone
);
1799 mempool_destroy(xfs_ioend_pool
);
1800 kmem_zone_destroy(xfs_ioend_zone
);
1809 printk(KERN_INFO XFS_VERSION_STRING
" with "
1810 XFS_BUILD_OPTIONS
" enabled\n");
1816 error
= xfs_init_zones();
1820 error
= xfs_alloc_trace_bufs();
1822 goto out_destroy_zones
;
1824 error
= xfs_mru_cache_init();
1826 goto out_free_trace_buffers
;
1828 error
= xfs_filestream_init();
1830 goto out_mru_cache_uninit
;
1832 error
= xfs_buf_init();
1834 goto out_filestream_uninit
;
1836 error
= xfs_init_procfs();
1838 goto out_buf_terminate
;
1840 error
= xfs_sysctl_register();
1842 goto out_cleanup_procfs
;
1846 error
= register_filesystem(&xfs_fs_type
);
1848 goto out_sysctl_unregister
;
1851 out_sysctl_unregister
:
1852 xfs_sysctl_unregister();
1854 xfs_cleanup_procfs();
1856 xfs_buf_terminate();
1857 out_filestream_uninit
:
1858 xfs_filestream_uninit();
1859 out_mru_cache_uninit
:
1860 xfs_mru_cache_uninit();
1861 out_free_trace_buffers
:
1862 xfs_free_trace_bufs();
1864 xfs_destroy_zones();
1873 unregister_filesystem(&xfs_fs_type
);
1874 xfs_sysctl_unregister();
1875 xfs_cleanup_procfs();
1876 xfs_buf_terminate();
1877 xfs_filestream_uninit();
1878 xfs_mru_cache_uninit();
1879 xfs_free_trace_bufs();
1880 xfs_destroy_zones();
1884 module_init(init_xfs_fs
);
1885 module_exit(exit_xfs_fs
);
1887 MODULE_AUTHOR("Silicon Graphics, Inc.");
1888 MODULE_DESCRIPTION(XFS_VERSION_STRING
" with " XFS_BUILD_OPTIONS
" enabled");
1889 MODULE_LICENSE("GPL");