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"
49 #include "xfs_buf_item.h"
50 #include "xfs_utils.h"
51 #include "xfs_vnodeops.h"
52 #include "xfs_vfsops.h"
53 #include "xfs_version.h"
54 #include "xfs_log_priv.h"
55 #include "xfs_trans_priv.h"
56 #include "xfs_filestream.h"
57 #include "xfs_da_btree.h"
58 #include "xfs_dir2_trace.h"
59 #include "xfs_extfree_item.h"
60 #include "xfs_mru_cache.h"
61 #include "xfs_inode_item.h"
64 #include <linux/namei.h>
65 #include <linux/init.h>
66 #include <linux/mount.h>
67 #include <linux/mempool.h>
68 #include <linux/writeback.h>
69 #include <linux/kthread.h>
70 #include <linux/freezer.h>
71 #include <linux/parser.h>
73 static struct quotactl_ops xfs_quotactl_operations
;
74 static struct super_operations xfs_super_operations
;
75 static kmem_zone_t
*xfs_ioend_zone
;
76 mempool_t
*xfs_ioend_pool
;
78 STATIC
struct xfs_mount_args
*
80 struct super_block
*sb
,
83 struct xfs_mount_args
*args
;
85 args
= kzalloc(sizeof(struct xfs_mount_args
), GFP_KERNEL
);
89 args
->logbufs
= args
->logbufsize
= -1;
90 strncpy(args
->fsname
, sb
->s_id
, MAXNAMELEN
);
92 /* Copy the already-parsed mount(2) flags we're interested in */
93 if (sb
->s_flags
& MS_DIRSYNC
)
94 args
->flags
|= XFSMNT_DIRSYNC
;
95 if (sb
->s_flags
& MS_SYNCHRONOUS
)
96 args
->flags
|= XFSMNT_WSYNC
;
98 args
->flags
|= XFSMNT_QUIET
;
99 args
->flags
|= XFSMNT_32BITINODES
;
104 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
105 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
106 #define MNTOPT_LOGDEV "logdev" /* log device */
107 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
108 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
109 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
110 #define MNTOPT_INO64 "ino64" /* force inodes into 64-bit range */
111 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
112 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
113 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
114 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
115 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
116 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
117 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
118 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
119 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
120 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
121 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
122 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
123 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
124 * unwritten extent conversion */
125 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
126 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
127 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
128 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
129 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
130 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
131 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
133 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
134 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
135 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
136 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
137 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
138 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
139 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
140 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
141 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
142 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
143 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
144 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
145 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
146 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
147 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
148 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
149 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
150 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
153 * Table driven mount option parser.
155 * Currently only used for remount, but it will be used for mount
156 * in the future, too.
159 Opt_barrier
, Opt_nobarrier
, Opt_err
162 static const match_table_t tokens
= {
163 {Opt_barrier
, "barrier"},
164 {Opt_nobarrier
, "nobarrier"},
170 suffix_strtoul(char *s
, char **endp
, unsigned int base
)
172 int last
, shift_left_factor
= 0;
175 last
= strlen(value
) - 1;
176 if (value
[last
] == 'K' || value
[last
] == 'k') {
177 shift_left_factor
= 10;
180 if (value
[last
] == 'M' || value
[last
] == 'm') {
181 shift_left_factor
= 20;
184 if (value
[last
] == 'G' || value
[last
] == 'g') {
185 shift_left_factor
= 30;
189 return simple_strtoul((const char *)s
, endp
, base
) << shift_left_factor
;
194 struct xfs_mount
*mp
,
196 struct xfs_mount_args
*args
,
199 char *this_char
, *value
, *eov
;
200 int dsunit
, dswidth
, vol_dsunit
, vol_dswidth
;
202 int dmapi_implies_ikeep
= 1;
204 args
->flags
|= XFSMNT_BARRIER
;
205 args
->flags2
|= XFSMNT2_COMPAT_IOSIZE
;
210 iosize
= dsunit
= dswidth
= vol_dsunit
= vol_dswidth
= 0;
212 while ((this_char
= strsep(&options
, ",")) != NULL
) {
215 if ((value
= strchr(this_char
, '=')) != NULL
)
218 if (!strcmp(this_char
, MNTOPT_LOGBUFS
)) {
219 if (!value
|| !*value
) {
221 "XFS: %s option requires an argument",
225 args
->logbufs
= simple_strtoul(value
, &eov
, 10);
226 } else if (!strcmp(this_char
, MNTOPT_LOGBSIZE
)) {
227 if (!value
|| !*value
) {
229 "XFS: %s option requires an argument",
233 args
->logbufsize
= suffix_strtoul(value
, &eov
, 10);
234 } else if (!strcmp(this_char
, MNTOPT_LOGDEV
)) {
235 if (!value
|| !*value
) {
237 "XFS: %s option requires an argument",
241 strncpy(args
->logname
, value
, MAXNAMELEN
);
242 } else if (!strcmp(this_char
, MNTOPT_MTPT
)) {
243 if (!value
|| !*value
) {
245 "XFS: %s option requires an argument",
249 strncpy(args
->mtpt
, value
, MAXNAMELEN
);
250 } else if (!strcmp(this_char
, MNTOPT_RTDEV
)) {
251 if (!value
|| !*value
) {
253 "XFS: %s option requires an argument",
257 strncpy(args
->rtname
, value
, MAXNAMELEN
);
258 } else if (!strcmp(this_char
, MNTOPT_BIOSIZE
)) {
259 if (!value
|| !*value
) {
261 "XFS: %s option requires an argument",
265 iosize
= simple_strtoul(value
, &eov
, 10);
266 args
->flags
|= XFSMNT_IOSIZE
;
267 args
->iosizelog
= (uint8_t) iosize
;
268 } else if (!strcmp(this_char
, MNTOPT_ALLOCSIZE
)) {
269 if (!value
|| !*value
) {
271 "XFS: %s option requires an argument",
275 iosize
= suffix_strtoul(value
, &eov
, 10);
276 args
->flags
|= XFSMNT_IOSIZE
;
277 args
->iosizelog
= ffs(iosize
) - 1;
278 } else if (!strcmp(this_char
, MNTOPT_GRPID
) ||
279 !strcmp(this_char
, MNTOPT_BSDGROUPS
)) {
280 mp
->m_flags
|= XFS_MOUNT_GRPID
;
281 } else if (!strcmp(this_char
, MNTOPT_NOGRPID
) ||
282 !strcmp(this_char
, MNTOPT_SYSVGROUPS
)) {
283 mp
->m_flags
&= ~XFS_MOUNT_GRPID
;
284 } else if (!strcmp(this_char
, MNTOPT_WSYNC
)) {
285 args
->flags
|= XFSMNT_WSYNC
;
286 } else if (!strcmp(this_char
, MNTOPT_OSYNCISOSYNC
)) {
287 args
->flags
|= XFSMNT_OSYNCISOSYNC
;
288 } else if (!strcmp(this_char
, MNTOPT_NORECOVERY
)) {
289 args
->flags
|= XFSMNT_NORECOVERY
;
290 } else if (!strcmp(this_char
, MNTOPT_INO64
)) {
291 args
->flags
|= XFSMNT_INO64
;
294 "XFS: %s option not allowed on this system",
298 } else if (!strcmp(this_char
, MNTOPT_NOALIGN
)) {
299 args
->flags
|= XFSMNT_NOALIGN
;
300 } else if (!strcmp(this_char
, MNTOPT_SWALLOC
)) {
301 args
->flags
|= XFSMNT_SWALLOC
;
302 } else if (!strcmp(this_char
, MNTOPT_SUNIT
)) {
303 if (!value
|| !*value
) {
305 "XFS: %s option requires an argument",
309 dsunit
= simple_strtoul(value
, &eov
, 10);
310 } else if (!strcmp(this_char
, MNTOPT_SWIDTH
)) {
311 if (!value
|| !*value
) {
313 "XFS: %s option requires an argument",
317 dswidth
= simple_strtoul(value
, &eov
, 10);
318 } else if (!strcmp(this_char
, MNTOPT_64BITINODE
)) {
319 args
->flags
&= ~XFSMNT_32BITINODES
;
322 "XFS: %s option not allowed on this system",
326 } else if (!strcmp(this_char
, MNTOPT_NOUUID
)) {
327 args
->flags
|= XFSMNT_NOUUID
;
328 } else if (!strcmp(this_char
, MNTOPT_BARRIER
)) {
329 args
->flags
|= XFSMNT_BARRIER
;
330 } else if (!strcmp(this_char
, MNTOPT_NOBARRIER
)) {
331 args
->flags
&= ~XFSMNT_BARRIER
;
332 } else if (!strcmp(this_char
, MNTOPT_IKEEP
)) {
333 args
->flags
|= XFSMNT_IKEEP
;
334 } else if (!strcmp(this_char
, MNTOPT_NOIKEEP
)) {
335 dmapi_implies_ikeep
= 0;
336 args
->flags
&= ~XFSMNT_IKEEP
;
337 } else if (!strcmp(this_char
, MNTOPT_LARGEIO
)) {
338 args
->flags2
&= ~XFSMNT2_COMPAT_IOSIZE
;
339 } else if (!strcmp(this_char
, MNTOPT_NOLARGEIO
)) {
340 args
->flags2
|= XFSMNT2_COMPAT_IOSIZE
;
341 } else if (!strcmp(this_char
, MNTOPT_ATTR2
)) {
342 args
->flags
|= XFSMNT_ATTR2
;
343 } else if (!strcmp(this_char
, MNTOPT_NOATTR2
)) {
344 args
->flags
&= ~XFSMNT_ATTR2
;
345 args
->flags
|= XFSMNT_NOATTR2
;
346 } else if (!strcmp(this_char
, MNTOPT_FILESTREAM
)) {
347 args
->flags2
|= XFSMNT2_FILESTREAMS
;
348 } else if (!strcmp(this_char
, MNTOPT_NOQUOTA
)) {
349 args
->flags
&= ~(XFSMNT_UQUOTAENF
|XFSMNT_UQUOTA
);
350 args
->flags
&= ~(XFSMNT_GQUOTAENF
|XFSMNT_GQUOTA
);
351 } else if (!strcmp(this_char
, MNTOPT_QUOTA
) ||
352 !strcmp(this_char
, MNTOPT_UQUOTA
) ||
353 !strcmp(this_char
, MNTOPT_USRQUOTA
)) {
354 args
->flags
|= XFSMNT_UQUOTA
| XFSMNT_UQUOTAENF
;
355 } else if (!strcmp(this_char
, MNTOPT_QUOTANOENF
) ||
356 !strcmp(this_char
, MNTOPT_UQUOTANOENF
)) {
357 args
->flags
|= XFSMNT_UQUOTA
;
358 args
->flags
&= ~XFSMNT_UQUOTAENF
;
359 } else if (!strcmp(this_char
, MNTOPT_PQUOTA
) ||
360 !strcmp(this_char
, MNTOPT_PRJQUOTA
)) {
361 args
->flags
|= XFSMNT_PQUOTA
| XFSMNT_PQUOTAENF
;
362 } else if (!strcmp(this_char
, MNTOPT_PQUOTANOENF
)) {
363 args
->flags
|= XFSMNT_PQUOTA
;
364 args
->flags
&= ~XFSMNT_PQUOTAENF
;
365 } else if (!strcmp(this_char
, MNTOPT_GQUOTA
) ||
366 !strcmp(this_char
, MNTOPT_GRPQUOTA
)) {
367 args
->flags
|= XFSMNT_GQUOTA
| XFSMNT_GQUOTAENF
;
368 } else if (!strcmp(this_char
, MNTOPT_GQUOTANOENF
)) {
369 args
->flags
|= XFSMNT_GQUOTA
;
370 args
->flags
&= ~XFSMNT_GQUOTAENF
;
371 } else if (!strcmp(this_char
, MNTOPT_DMAPI
)) {
372 args
->flags
|= XFSMNT_DMAPI
;
373 } else if (!strcmp(this_char
, MNTOPT_XDSM
)) {
374 args
->flags
|= XFSMNT_DMAPI
;
375 } else if (!strcmp(this_char
, MNTOPT_DMI
)) {
376 args
->flags
|= XFSMNT_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
);
394 if (args
->flags
& XFSMNT_NORECOVERY
) {
395 if ((mp
->m_flags
& XFS_MOUNT_RDONLY
) == 0) {
397 "XFS: no-recovery mounts must be read-only.");
402 if ((args
->flags
& XFSMNT_NOALIGN
) && (dsunit
|| dswidth
)) {
404 "XFS: sunit and swidth options incompatible with the noalign option");
408 if ((args
->flags
& XFSMNT_GQUOTA
) && (args
->flags
& XFSMNT_PQUOTA
)) {
410 "XFS: cannot mount with both project and group quota");
414 if ((args
->flags
& XFSMNT_DMAPI
) && *args
->mtpt
== '\0') {
415 printk("XFS: %s option needs the mount point option as well\n",
420 if ((dsunit
&& !dswidth
) || (!dsunit
&& dswidth
)) {
422 "XFS: sunit and swidth must be specified together");
426 if (dsunit
&& (dswidth
% dsunit
!= 0)) {
428 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
434 * Applications using DMI filesystems often expect the
435 * inode generation number to be monotonically increasing.
436 * If we delete inode chunks we break this assumption, so
437 * keep unused inode chunks on disk for DMI filesystems
438 * until we come up with a better solution.
439 * Note that if "ikeep" or "noikeep" mount options are
440 * supplied, then they are honored.
442 if ((args
->flags
& XFSMNT_DMAPI
) && dmapi_implies_ikeep
)
443 args
->flags
|= XFSMNT_IKEEP
;
445 if ((args
->flags
& XFSMNT_NOALIGN
) != XFSMNT_NOALIGN
) {
447 args
->sunit
= dsunit
;
448 args
->flags
|= XFSMNT_RETERR
;
450 args
->sunit
= vol_dsunit
;
452 dswidth
? (args
->swidth
= dswidth
) :
453 (args
->swidth
= vol_dswidth
);
455 args
->sunit
= args
->swidth
= 0;
459 if (args
->flags
& XFSMNT_32BITINODES
)
460 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
462 args
->flags
|= XFSMNT_FLAGS2
;
466 struct proc_xfs_info
{
473 struct xfs_mount
*mp
,
476 static struct proc_xfs_info xfs_info_set
[] = {
477 /* the few simple ones we can get from the mount struct */
478 { XFS_MOUNT_IKEEP
, "," MNTOPT_IKEEP
},
479 { XFS_MOUNT_WSYNC
, "," MNTOPT_WSYNC
},
480 { XFS_MOUNT_INO64
, "," MNTOPT_INO64
},
481 { XFS_MOUNT_NOALIGN
, "," MNTOPT_NOALIGN
},
482 { XFS_MOUNT_SWALLOC
, "," MNTOPT_SWALLOC
},
483 { XFS_MOUNT_NOUUID
, "," MNTOPT_NOUUID
},
484 { XFS_MOUNT_NORECOVERY
, "," MNTOPT_NORECOVERY
},
485 { XFS_MOUNT_OSYNCISOSYNC
, "," MNTOPT_OSYNCISOSYNC
},
486 { XFS_MOUNT_ATTR2
, "," MNTOPT_ATTR2
},
487 { XFS_MOUNT_FILESTREAMS
, "," MNTOPT_FILESTREAM
},
488 { XFS_MOUNT_DMAPI
, "," MNTOPT_DMAPI
},
489 { XFS_MOUNT_GRPID
, "," MNTOPT_GRPID
},
492 static struct proc_xfs_info xfs_info_unset
[] = {
493 /* the few simple ones we can get from the mount struct */
494 { XFS_MOUNT_COMPAT_IOSIZE
, "," MNTOPT_LARGEIO
},
495 { XFS_MOUNT_BARRIER
, "," MNTOPT_NOBARRIER
},
496 { XFS_MOUNT_SMALL_INUMS
, "," MNTOPT_64BITINODE
},
499 struct proc_xfs_info
*xfs_infop
;
501 for (xfs_infop
= xfs_info_set
; xfs_infop
->flag
; xfs_infop
++) {
502 if (mp
->m_flags
& xfs_infop
->flag
)
503 seq_puts(m
, xfs_infop
->str
);
505 for (xfs_infop
= xfs_info_unset
; xfs_infop
->flag
; xfs_infop
++) {
506 if (!(mp
->m_flags
& xfs_infop
->flag
))
507 seq_puts(m
, xfs_infop
->str
);
510 if (mp
->m_flags
& XFS_MOUNT_DFLT_IOSIZE
)
511 seq_printf(m
, "," MNTOPT_ALLOCSIZE
"=%dk",
512 (int)(1 << mp
->m_writeio_log
) >> 10);
514 if (mp
->m_logbufs
> 0)
515 seq_printf(m
, "," MNTOPT_LOGBUFS
"=%d", mp
->m_logbufs
);
516 if (mp
->m_logbsize
> 0)
517 seq_printf(m
, "," MNTOPT_LOGBSIZE
"=%dk", mp
->m_logbsize
>> 10);
520 seq_printf(m
, "," MNTOPT_LOGDEV
"=%s", mp
->m_logname
);
522 seq_printf(m
, "," MNTOPT_RTDEV
"=%s", mp
->m_rtname
);
524 if (mp
->m_dalign
> 0)
525 seq_printf(m
, "," MNTOPT_SUNIT
"=%d",
526 (int)XFS_FSB_TO_BB(mp
, mp
->m_dalign
));
527 if (mp
->m_swidth
> 0)
528 seq_printf(m
, "," MNTOPT_SWIDTH
"=%d",
529 (int)XFS_FSB_TO_BB(mp
, mp
->m_swidth
));
531 if (mp
->m_qflags
& (XFS_UQUOTA_ACCT
|XFS_UQUOTA_ENFD
))
532 seq_puts(m
, "," MNTOPT_USRQUOTA
);
533 else if (mp
->m_qflags
& XFS_UQUOTA_ACCT
)
534 seq_puts(m
, "," MNTOPT_UQUOTANOENF
);
536 if (mp
->m_qflags
& (XFS_PQUOTA_ACCT
|XFS_OQUOTA_ENFD
))
537 seq_puts(m
, "," MNTOPT_PRJQUOTA
);
538 else if (mp
->m_qflags
& XFS_PQUOTA_ACCT
)
539 seq_puts(m
, "," MNTOPT_PQUOTANOENF
);
541 if (mp
->m_qflags
& (XFS_GQUOTA_ACCT
|XFS_OQUOTA_ENFD
))
542 seq_puts(m
, "," MNTOPT_GRPQUOTA
);
543 else if (mp
->m_qflags
& XFS_GQUOTA_ACCT
)
544 seq_puts(m
, "," MNTOPT_GQUOTANOENF
);
546 if (!(mp
->m_qflags
& XFS_ALL_QUOTA_ACCT
))
547 seq_puts(m
, "," MNTOPT_NOQUOTA
);
553 unsigned int blockshift
)
555 unsigned int pagefactor
= 1;
556 unsigned int bitshift
= BITS_PER_LONG
- 1;
558 /* Figure out maximum filesize, on Linux this can depend on
559 * the filesystem blocksize (on 32 bit platforms).
560 * __block_prepare_write does this in an [unsigned] long...
561 * page->index << (PAGE_CACHE_SHIFT - bbits)
562 * So, for page sized blocks (4K on 32 bit platforms),
563 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
564 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
565 * but for smaller blocksizes it is less (bbits = log2 bsize).
566 * Note1: get_block_t takes a long (implicit cast from above)
567 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
568 * can optionally convert the [unsigned] long from above into
569 * an [unsigned] long long.
572 #if BITS_PER_LONG == 32
573 # if defined(CONFIG_LBD)
574 ASSERT(sizeof(sector_t
) == 8);
575 pagefactor
= PAGE_CACHE_SIZE
;
576 bitshift
= BITS_PER_LONG
;
578 pagefactor
= PAGE_CACHE_SIZE
>> (PAGE_CACHE_SHIFT
- blockshift
);
582 return (((__uint64_t
)pagefactor
) << bitshift
) - 1;
589 struct block_device
**bdevp
)
593 *bdevp
= open_bdev_exclusive(name
, FMODE_READ
|FMODE_WRITE
, mp
);
594 if (IS_ERR(*bdevp
)) {
595 error
= PTR_ERR(*bdevp
);
596 printk("XFS: Invalid device [%s], error=%d\n", name
, error
);
604 struct block_device
*bdev
)
607 close_bdev_exclusive(bdev
, FMODE_READ
|FMODE_WRITE
);
611 * Try to write out the superblock using barriers.
617 xfs_buf_t
*sbp
= xfs_getsb(mp
, 0);
622 XFS_BUF_UNDELAYWRITE(sbp
);
624 XFS_BUF_UNASYNC(sbp
);
625 XFS_BUF_ORDERED(sbp
);
628 error
= xfs_iowait(sbp
);
631 * Clear all the flags we set and possible error state in the
632 * buffer. We only did the write to try out whether barriers
633 * worked and shouldn't leave any traces in the superblock
637 XFS_BUF_ERROR(sbp
, 0);
638 XFS_BUF_UNORDERED(sbp
);
645 xfs_mountfs_check_barriers(xfs_mount_t
*mp
)
649 if (mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
650 xfs_fs_cmn_err(CE_NOTE
, mp
,
651 "Disabling barriers, not supported with external log device");
652 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
656 if (xfs_readonly_buftarg(mp
->m_ddev_targp
)) {
657 xfs_fs_cmn_err(CE_NOTE
, mp
,
658 "Disabling barriers, underlying device is readonly");
659 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
663 error
= xfs_barrier_test(mp
);
665 xfs_fs_cmn_err(CE_NOTE
, mp
,
666 "Disabling barriers, trial barrier write failed");
667 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
673 xfs_blkdev_issue_flush(
674 xfs_buftarg_t
*buftarg
)
676 blkdev_issue_flush(buftarg
->bt_bdev
, NULL
);
681 struct xfs_mount
*mp
)
683 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
684 struct block_device
*logdev
= mp
->m_logdev_targp
->bt_bdev
;
685 xfs_free_buftarg(mp
->m_logdev_targp
);
686 xfs_blkdev_put(logdev
);
688 if (mp
->m_rtdev_targp
) {
689 struct block_device
*rtdev
= mp
->m_rtdev_targp
->bt_bdev
;
690 xfs_free_buftarg(mp
->m_rtdev_targp
);
691 xfs_blkdev_put(rtdev
);
693 xfs_free_buftarg(mp
->m_ddev_targp
);
697 * The file system configurations are:
698 * (1) device (partition) with data and internal log
699 * (2) logical volume with data and log subvolumes.
700 * (3) logical volume with data, log, and realtime subvolumes.
702 * We only have to handle opening the log and realtime volumes here if
703 * they are present. The data subvolume has already been opened by
704 * get_sb_bdev() and is stored in sb->s_bdev.
708 struct xfs_mount
*mp
,
709 struct xfs_mount_args
*args
)
711 struct block_device
*ddev
= mp
->m_super
->s_bdev
;
712 struct block_device
*logdev
= NULL
, *rtdev
= NULL
;
716 * Open real time and log devices - order is important.
718 if (args
->logname
[0]) {
719 error
= xfs_blkdev_get(mp
, args
->logname
, &logdev
);
724 if (args
->rtname
[0]) {
725 error
= xfs_blkdev_get(mp
, args
->rtname
, &rtdev
);
727 goto out_close_logdev
;
729 if (rtdev
== ddev
|| rtdev
== logdev
) {
731 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
733 goto out_close_rtdev
;
738 * Setup xfs_mount buffer target pointers
741 mp
->m_ddev_targp
= xfs_alloc_buftarg(ddev
, 0);
742 if (!mp
->m_ddev_targp
)
743 goto out_close_rtdev
;
746 mp
->m_rtdev_targp
= xfs_alloc_buftarg(rtdev
, 1);
747 if (!mp
->m_rtdev_targp
)
748 goto out_free_ddev_targ
;
751 if (logdev
&& logdev
!= ddev
) {
752 mp
->m_logdev_targp
= xfs_alloc_buftarg(logdev
, 1);
753 if (!mp
->m_logdev_targp
)
754 goto out_free_rtdev_targ
;
756 mp
->m_logdev_targp
= mp
->m_ddev_targp
;
762 if (mp
->m_rtdev_targp
)
763 xfs_free_buftarg(mp
->m_rtdev_targp
);
765 xfs_free_buftarg(mp
->m_ddev_targp
);
768 xfs_blkdev_put(rtdev
);
770 if (logdev
&& logdev
!= ddev
)
771 xfs_blkdev_put(logdev
);
777 * Setup xfs_mount buffer target pointers based on superblock
781 struct xfs_mount
*mp
)
785 error
= xfs_setsize_buftarg(mp
->m_ddev_targp
, mp
->m_sb
.sb_blocksize
,
786 mp
->m_sb
.sb_sectsize
);
790 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
791 unsigned int log_sector_size
= BBSIZE
;
793 if (xfs_sb_version_hassector(&mp
->m_sb
))
794 log_sector_size
= mp
->m_sb
.sb_logsectsize
;
795 error
= xfs_setsize_buftarg(mp
->m_logdev_targp
,
796 mp
->m_sb
.sb_blocksize
,
801 if (mp
->m_rtdev_targp
) {
802 error
= xfs_setsize_buftarg(mp
->m_rtdev_targp
,
803 mp
->m_sb
.sb_blocksize
,
804 mp
->m_sb
.sb_sectsize
);
813 * XFS AIL push thread support
818 xfs_lsn_t threshold_lsn
)
820 mp
->m_ail
.xa_target
= threshold_lsn
;
821 wake_up_process(mp
->m_ail
.xa_task
);
828 xfs_mount_t
*mp
= (xfs_mount_t
*)data
;
829 xfs_lsn_t last_pushed_lsn
= 0;
832 while (!kthread_should_stop()) {
834 schedule_timeout_interruptible(msecs_to_jiffies(tout
));
841 if (XFS_FORCED_SHUTDOWN(mp
))
844 tout
= xfsaild_push(mp
, &last_pushed_lsn
);
854 mp
->m_ail
.xa_target
= 0;
855 mp
->m_ail
.xa_task
= kthread_run(xfsaild
, mp
, "xfsaild");
856 if (IS_ERR(mp
->m_ail
.xa_task
))
857 return -PTR_ERR(mp
->m_ail
.xa_task
);
865 kthread_stop(mp
->m_ail
.xa_task
);
869 /* Catch misguided souls that try to use this interface on XFS */
870 STATIC
struct inode
*
872 struct super_block
*sb
)
879 * Now that the generic code is guaranteed not to be accessing
880 * the linux inode, we can reclaim the inode.
883 xfs_fs_destroy_inode(
886 xfs_inode_t
*ip
= XFS_I(inode
);
888 XFS_STATS_INC(vn_reclaim
);
890 panic("%s: cannot reclaim 0x%p\n", __func__
, inode
);
894 * Slab object creation initialisation for the XFS inode.
895 * This covers only the idempotent fields in the XFS inode;
896 * all other fields need to be initialised on allocation
897 * from the slab. This avoids the need to repeatedly intialise
898 * fields in the xfs inode that left in the initialise state
899 * when freeing the inode.
902 xfs_fs_inode_init_once(
905 struct xfs_inode
*ip
= inode
;
907 memset(ip
, 0, sizeof(struct xfs_inode
));
910 inode_init_once(VFS_I(ip
));
913 atomic_set(&ip
->i_iocount
, 0);
914 atomic_set(&ip
->i_pincount
, 0);
915 spin_lock_init(&ip
->i_flags_lock
);
916 init_waitqueue_head(&ip
->i_ipin_wait
);
918 * Because we want to use a counting completion, complete
919 * the flush completion once to allow a single access to
920 * the flush completion without blocking.
922 init_completion(&ip
->i_flush
);
923 complete(&ip
->i_flush
);
925 mrlock_init(&ip
->i_lock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
926 "xfsino", ip
->i_ino
);
927 mrlock_init(&ip
->i_iolock
, MRLOCK_BARRIER
, "xfsio", ip
->i_ino
);
931 * Attempt to flush the inode, this will actually fail
932 * if the inode is pinned, but we dirty the inode again
933 * at the point when it is unpinned after a log write,
934 * since this is when the inode itself becomes flushable.
944 xfs_itrace_entry(XFS_I(inode
));
946 filemap_fdatawait(inode
->i_mapping
);
949 error
= xfs_inode_flush(XFS_I(inode
), flags
);
951 * if we failed to write out the inode then mark
952 * it dirty again so we'll try again later.
955 xfs_mark_inode_dirty_sync(XFS_I(inode
));
964 xfs_inode_t
*ip
= XFS_I(inode
);
966 xfs_itrace_entry(ip
);
967 XFS_STATS_INC(vn_rele
);
968 XFS_STATS_INC(vn_remove
);
969 XFS_STATS_DEC(vn_active
);
972 xfs_iflags_clear(ip
, XFS_IMODIFIED
);
977 struct xfs_mount
*mp
)
981 kfree(mp
->m_logname
);
986 struct super_block
*sb
)
988 struct xfs_mount
*mp
= XFS_M(sb
);
989 struct xfs_inode
*rip
= mp
->m_rootip
;
990 int unmount_event_flags
= 0;
994 xfs_sync_inodes(mp
, SYNC_ATTR
|SYNC_DELWRI
);
997 if (mp
->m_flags
& XFS_MOUNT_DMAPI
) {
998 unmount_event_flags
=
999 (mp
->m_dmevmask
& (1 << DM_EVENT_UNMOUNT
)) ?
1000 0 : DM_FLAGS_UNWANTED
;
1002 * Ignore error from dmapi here, first unmount is not allowed
1003 * to fail anyway, and second we wouldn't want to fail a
1004 * unmount because of dmapi.
1006 XFS_SEND_PREUNMOUNT(mp
, rip
, DM_RIGHT_NULL
, rip
, DM_RIGHT_NULL
,
1007 NULL
, NULL
, 0, 0, unmount_event_flags
);
1012 * Blow away any referenced inode in the filestreams cache.
1013 * This can and will cause log traffic as inodes go inactive
1016 xfs_filestream_unmount(mp
);
1018 XFS_bflush(mp
->m_ddev_targp
);
1019 error
= xfs_unmount_flush(mp
, 0);
1022 if (mp
->m_flags
& XFS_MOUNT_DMAPI
) {
1023 XFS_SEND_UNMOUNT(mp
, rip
, DM_RIGHT_NULL
, 0, 0,
1024 unmount_event_flags
);
1029 xfs_icsb_destroy_counters(mp
);
1030 xfs_close_devices(mp
);
1033 xfs_free_fsname(mp
);
1039 struct super_block
*sb
)
1041 if (!(sb
->s_flags
& MS_RDONLY
))
1042 xfs_sync_fsdata(XFS_M(sb
), 0);
1048 struct super_block
*sb
,
1051 struct xfs_mount
*mp
= XFS_M(sb
);
1055 * Treat a sync operation like a freeze. This is to work
1056 * around a race in sync_inodes() which works in two phases
1057 * - an asynchronous flush, which can write out an inode
1058 * without waiting for file size updates to complete, and a
1059 * synchronous flush, which wont do anything because the
1060 * async flush removed the inode's dirty flag. Also
1061 * sync_inodes() will not see any files that just have
1062 * outstanding transactions to be flushed because we don't
1063 * dirty the Linux inode until after the transaction I/O
1066 if (wait
|| unlikely(sb
->s_frozen
== SB_FREEZE_WRITE
))
1067 error
= xfs_quiesce_data(mp
);
1069 error
= xfs_sync_fsdata(mp
, 0);
1072 if (unlikely(laptop_mode
)) {
1073 int prev_sync_seq
= mp
->m_sync_seq
;
1076 * The disk must be active because we're syncing.
1077 * We schedule xfssyncd now (now that the disk is
1078 * active) instead of later (when it might not be).
1080 wake_up_process(mp
->m_sync_task
);
1082 * We have to wait for the sync iteration to complete.
1083 * If we don't, the disk activity caused by the sync
1084 * will come after the sync is completed, and that
1085 * triggers another sync from laptop mode.
1087 wait_event(mp
->m_wait_single_sync_task
,
1088 mp
->m_sync_seq
!= prev_sync_seq
);
1096 struct dentry
*dentry
,
1097 struct kstatfs
*statp
)
1099 struct xfs_mount
*mp
= XFS_M(dentry
->d_sb
);
1100 xfs_sb_t
*sbp
= &mp
->m_sb
;
1101 __uint64_t fakeinos
, id
;
1104 statp
->f_type
= XFS_SB_MAGIC
;
1105 statp
->f_namelen
= MAXNAMELEN
- 1;
1107 id
= huge_encode_dev(mp
->m_ddev_targp
->bt_dev
);
1108 statp
->f_fsid
.val
[0] = (u32
)id
;
1109 statp
->f_fsid
.val
[1] = (u32
)(id
>> 32);
1111 xfs_icsb_sync_counters(mp
, XFS_ICSB_LAZY_COUNT
);
1113 spin_lock(&mp
->m_sb_lock
);
1114 statp
->f_bsize
= sbp
->sb_blocksize
;
1115 lsize
= sbp
->sb_logstart
? sbp
->sb_logblocks
: 0;
1116 statp
->f_blocks
= sbp
->sb_dblocks
- lsize
;
1117 statp
->f_bfree
= statp
->f_bavail
=
1118 sbp
->sb_fdblocks
- XFS_ALLOC_SET_ASIDE(mp
);
1119 fakeinos
= statp
->f_bfree
<< sbp
->sb_inopblog
;
1121 fakeinos
+= mp
->m_inoadd
;
1124 MIN(sbp
->sb_icount
+ fakeinos
, (__uint64_t
)XFS_MAXINUMBER
);
1125 if (mp
->m_maxicount
)
1129 statp
->f_files
= min_t(typeof(statp
->f_files
),
1132 statp
->f_ffree
= statp
->f_files
- (sbp
->sb_icount
- sbp
->sb_ifree
);
1133 spin_unlock(&mp
->m_sb_lock
);
1135 XFS_QM_DQSTATVFS(XFS_I(dentry
->d_inode
), statp
);
1141 struct super_block
*sb
,
1145 struct xfs_mount
*mp
= XFS_M(sb
);
1146 substring_t args
[MAX_OPT_ARGS
];
1149 while ((p
= strsep(&options
, ",")) != NULL
) {
1155 token
= match_token(p
, tokens
, args
);
1158 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
1161 * Test if barriers are actually working if we can,
1162 * else delay this check until the filesystem is
1165 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
))
1166 xfs_mountfs_check_barriers(mp
);
1169 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
1173 * Logically we would return an error here to prevent
1174 * users from believing they might have changed
1175 * mount options using remount which can't be changed.
1177 * But unfortunately mount(8) adds all options from
1178 * mtab and fstab to the mount arguments in some cases
1179 * so we can't blindly reject options, but have to
1180 * check for each specified option if it actually
1181 * differs from the currently set option and only
1182 * reject it if that's the case.
1184 * Until that is implemented we return success for
1185 * every remount request, and silently ignore all
1186 * options that we can't actually change.
1190 "XFS: mount option \"%s\" not supported for remount\n", p
);
1199 if ((mp
->m_flags
& XFS_MOUNT_RDONLY
) && !(*flags
& MS_RDONLY
)) {
1200 mp
->m_flags
&= ~XFS_MOUNT_RDONLY
;
1201 if (mp
->m_flags
& XFS_MOUNT_BARRIER
)
1202 xfs_mountfs_check_barriers(mp
);
1206 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
) && (*flags
& MS_RDONLY
)) {
1207 xfs_quiesce_data(mp
);
1208 xfs_quiesce_attr(mp
);
1209 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
1216 * Second stage of a freeze. The data is already frozen so we only
1217 * need to take care of the metadata. Once that's done write a dummy
1218 * record to dirty the log in case of a crash while frozen.
1222 struct super_block
*sb
)
1224 struct xfs_mount
*mp
= XFS_M(sb
);
1226 xfs_quiesce_attr(mp
);
1227 xfs_fs_log_dummy(mp
);
1231 xfs_fs_show_options(
1233 struct vfsmount
*mnt
)
1235 return -xfs_showargs(XFS_M(mnt
->mnt_sb
), m
);
1240 struct super_block
*sb
,
1243 return -XFS_QM_QUOTACTL(XFS_M(sb
), Q_XQUOTASYNC
, 0, NULL
);
1248 struct super_block
*sb
,
1249 struct fs_quota_stat
*fqs
)
1251 return -XFS_QM_QUOTACTL(XFS_M(sb
), Q_XGETQSTAT
, 0, (caddr_t
)fqs
);
1256 struct super_block
*sb
,
1260 return -XFS_QM_QUOTACTL(XFS_M(sb
), op
, 0, (caddr_t
)&flags
);
1265 struct super_block
*sb
,
1268 struct fs_disk_quota
*fdq
)
1270 return -XFS_QM_QUOTACTL(XFS_M(sb
),
1271 (type
== USRQUOTA
) ? Q_XGETQUOTA
:
1272 ((type
== GRPQUOTA
) ? Q_XGETGQUOTA
:
1273 Q_XGETPQUOTA
), id
, (caddr_t
)fdq
);
1278 struct super_block
*sb
,
1281 struct fs_disk_quota
*fdq
)
1283 return -XFS_QM_QUOTACTL(XFS_M(sb
),
1284 (type
== USRQUOTA
) ? Q_XSETQLIM
:
1285 ((type
== GRPQUOTA
) ? Q_XSETGQLIM
:
1286 Q_XSETPQLIM
), id
, (caddr_t
)fdq
);
1290 * This function fills in xfs_mount_t fields based on mount args.
1291 * Note: the superblock has _not_ yet been read in.
1295 struct xfs_mount_args
*ap
,
1296 struct xfs_mount
*mp
)
1300 /* Values are in BBs */
1301 if ((ap
->flags
& XFSMNT_NOALIGN
) != XFSMNT_NOALIGN
) {
1303 * At this point the superblock has not been read
1304 * in, therefore we do not know the block size.
1305 * Before the mount call ends we will convert
1308 mp
->m_dalign
= ap
->sunit
;
1309 mp
->m_swidth
= ap
->swidth
;
1312 if (ap
->logbufs
!= -1 &&
1314 (ap
->logbufs
< XLOG_MIN_ICLOGS
||
1315 ap
->logbufs
> XLOG_MAX_ICLOGS
)) {
1317 "XFS: invalid logbufs value: %d [not %d-%d]",
1318 ap
->logbufs
, XLOG_MIN_ICLOGS
, XLOG_MAX_ICLOGS
);
1319 return XFS_ERROR(EINVAL
);
1321 mp
->m_logbufs
= ap
->logbufs
;
1322 if (ap
->logbufsize
!= -1 &&
1323 ap
->logbufsize
!= 0 &&
1324 (ap
->logbufsize
< XLOG_MIN_RECORD_BSIZE
||
1325 ap
->logbufsize
> XLOG_MAX_RECORD_BSIZE
||
1326 !is_power_of_2(ap
->logbufsize
))) {
1328 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
1330 return XFS_ERROR(EINVAL
);
1335 mp
->m_logbsize
= ap
->logbufsize
;
1336 mp
->m_fsname_len
= strlen(ap
->fsname
) + 1;
1338 mp
->m_fsname
= kstrdup(ap
->fsname
, GFP_KERNEL
);
1342 if (ap
->rtname
[0]) {
1343 mp
->m_rtname
= kstrdup(ap
->rtname
, GFP_KERNEL
);
1345 goto out_free_fsname
;
1349 if (ap
->logname
[0]) {
1350 mp
->m_logname
= kstrdup(ap
->logname
, GFP_KERNEL
);
1352 goto out_free_rtname
;
1355 if (ap
->flags
& XFSMNT_WSYNC
)
1356 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
1358 if (ap
->flags
& XFSMNT_INO64
) {
1359 mp
->m_flags
|= XFS_MOUNT_INO64
;
1360 mp
->m_inoadd
= XFS_INO64_OFFSET
;
1363 if (ap
->flags
& XFSMNT_RETERR
)
1364 mp
->m_flags
|= XFS_MOUNT_RETERR
;
1365 if (ap
->flags
& XFSMNT_NOALIGN
)
1366 mp
->m_flags
|= XFS_MOUNT_NOALIGN
;
1367 if (ap
->flags
& XFSMNT_SWALLOC
)
1368 mp
->m_flags
|= XFS_MOUNT_SWALLOC
;
1369 if (ap
->flags
& XFSMNT_OSYNCISOSYNC
)
1370 mp
->m_flags
|= XFS_MOUNT_OSYNCISOSYNC
;
1371 if (ap
->flags
& XFSMNT_32BITINODES
)
1372 mp
->m_flags
|= XFS_MOUNT_32BITINODES
;
1374 if (ap
->flags
& XFSMNT_IOSIZE
) {
1375 if (ap
->iosizelog
> XFS_MAX_IO_LOG
||
1376 ap
->iosizelog
< XFS_MIN_IO_LOG
) {
1378 "XFS: invalid log iosize: %d [not %d-%d]",
1379 ap
->iosizelog
, XFS_MIN_IO_LOG
,
1381 return XFS_ERROR(EINVAL
);
1384 mp
->m_flags
|= XFS_MOUNT_DFLT_IOSIZE
;
1385 mp
->m_readio_log
= mp
->m_writeio_log
= ap
->iosizelog
;
1388 if (ap
->flags
& XFSMNT_IKEEP
)
1389 mp
->m_flags
|= XFS_MOUNT_IKEEP
;
1390 if (ap
->flags
& XFSMNT_DIRSYNC
)
1391 mp
->m_flags
|= XFS_MOUNT_DIRSYNC
;
1392 if (ap
->flags
& XFSMNT_ATTR2
)
1393 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
1394 if (ap
->flags
& XFSMNT_NOATTR2
)
1395 mp
->m_flags
|= XFS_MOUNT_NOATTR2
;
1397 if (ap
->flags2
& XFSMNT2_COMPAT_IOSIZE
)
1398 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
1401 * no recovery flag requires a read-only mount
1403 if (ap
->flags
& XFSMNT_NORECOVERY
) {
1404 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
1406 "XFS: tried to mount a FS read-write without recovery!");
1407 return XFS_ERROR(EINVAL
);
1409 mp
->m_flags
|= XFS_MOUNT_NORECOVERY
;
1412 if (ap
->flags
& XFSMNT_NOUUID
)
1413 mp
->m_flags
|= XFS_MOUNT_NOUUID
;
1414 if (ap
->flags
& XFSMNT_BARRIER
)
1415 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
1417 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
1419 if (ap
->flags2
& XFSMNT2_FILESTREAMS
)
1420 mp
->m_flags
|= XFS_MOUNT_FILESTREAMS
;
1422 if (ap
->flags
& XFSMNT_DMAPI
)
1423 mp
->m_flags
|= XFS_MOUNT_DMAPI
;
1428 kfree(mp
->m_rtname
);
1430 kfree(mp
->m_fsname
);
1436 * This function fills in xfs_mount_t fields based on mount args.
1437 * Note: the superblock _has_ now been read in.
1441 struct xfs_mount_args
*ap
,
1442 struct xfs_mount
*mp
)
1444 int ronly
= (mp
->m_flags
& XFS_MOUNT_RDONLY
);
1446 /* Fail a mount where the logbuf is smaller then the log stripe */
1447 if (xfs_sb_version_haslogv2(&mp
->m_sb
)) {
1448 if ((ap
->logbufsize
<= 0) &&
1449 (mp
->m_sb
.sb_logsunit
> XLOG_BIG_RECORD_BSIZE
)) {
1450 mp
->m_logbsize
= mp
->m_sb
.sb_logsunit
;
1451 } else if (ap
->logbufsize
> 0 &&
1452 ap
->logbufsize
< mp
->m_sb
.sb_logsunit
) {
1454 "XFS: logbuf size must be greater than or equal to log stripe size");
1455 return XFS_ERROR(EINVAL
);
1458 /* Fail a mount if the logbuf is larger than 32K */
1459 if (ap
->logbufsize
> XLOG_BIG_RECORD_BSIZE
) {
1461 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1462 return XFS_ERROR(EINVAL
);
1467 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1468 * told by noattr2 to turn it off
1470 if (xfs_sb_version_hasattr2(&mp
->m_sb
) &&
1471 !(ap
->flags
& XFSMNT_NOATTR2
))
1472 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
1475 * prohibit r/w mounts of read-only filesystems
1477 if ((mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) && !ronly
) {
1479 "XFS: cannot mount a read-only filesystem as read-write");
1480 return XFS_ERROR(EROFS
);
1484 * check for shared mount.
1486 if (ap
->flags
& XFSMNT_SHARED
) {
1487 if (!xfs_sb_version_hasshared(&mp
->m_sb
))
1488 return XFS_ERROR(EINVAL
);
1491 * For IRIX 6.5, shared mounts must have the shared
1492 * version bit set, have the persistent readonly
1493 * field set, must be version 0 and can only be mounted
1496 if (!ronly
|| !(mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) ||
1497 (mp
->m_sb
.sb_shared_vn
!= 0))
1498 return XFS_ERROR(EINVAL
);
1500 mp
->m_flags
|= XFS_MOUNT_SHARED
;
1503 * Shared XFS V0 can't deal with DMI. Return EINVAL.
1505 if (mp
->m_sb
.sb_shared_vn
== 0 && (ap
->flags
& XFSMNT_DMAPI
))
1506 return XFS_ERROR(EINVAL
);
1509 if (ap
->flags
& XFSMNT_UQUOTA
) {
1510 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
);
1511 if (ap
->flags
& XFSMNT_UQUOTAENF
)
1512 mp
->m_qflags
|= XFS_UQUOTA_ENFD
;
1515 if (ap
->flags
& XFSMNT_GQUOTA
) {
1516 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
);
1517 if (ap
->flags
& XFSMNT_GQUOTAENF
)
1518 mp
->m_qflags
|= XFS_OQUOTA_ENFD
;
1519 } else if (ap
->flags
& XFSMNT_PQUOTA
) {
1520 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
);
1521 if (ap
->flags
& XFSMNT_PQUOTAENF
)
1522 mp
->m_qflags
|= XFS_OQUOTA_ENFD
;
1530 struct super_block
*sb
,
1535 struct xfs_mount
*mp
= NULL
;
1536 struct xfs_mount_args
*args
;
1537 int flags
= 0, error
= ENOMEM
;
1539 args
= xfs_args_allocate(sb
, silent
);
1543 mp
= kzalloc(sizeof(struct xfs_mount
), GFP_KERNEL
);
1547 spin_lock_init(&mp
->m_sb_lock
);
1548 mutex_init(&mp
->m_growlock
);
1549 atomic_set(&mp
->m_active_trans
, 0);
1550 INIT_LIST_HEAD(&mp
->m_sync_list
);
1551 spin_lock_init(&mp
->m_sync_lock
);
1552 init_waitqueue_head(&mp
->m_wait_single_sync_task
);
1557 if (sb
->s_flags
& MS_RDONLY
)
1558 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
1560 error
= xfs_parseargs(mp
, (char *)data
, args
, 0);
1564 sb_min_blocksize(sb
, BBSIZE
);
1565 sb
->s_xattr
= xfs_xattr_handlers
;
1566 sb
->s_export_op
= &xfs_export_operations
;
1567 sb
->s_qcop
= &xfs_quotactl_operations
;
1568 sb
->s_op
= &xfs_super_operations
;
1570 error
= xfs_dmops_get(mp
, args
);
1573 error
= xfs_qmops_get(mp
, args
);
1577 if (args
->flags
& XFSMNT_QUIET
)
1578 flags
|= XFS_MFSI_QUIET
;
1580 error
= xfs_open_devices(mp
, args
);
1584 if (xfs_icsb_init_counters(mp
))
1585 mp
->m_flags
|= XFS_MOUNT_NO_PERCPU_SB
;
1588 * Setup flags based on mount(2) options and then the superblock
1590 error
= xfs_start_flags(args
, mp
);
1592 goto out_free_fsname
;
1593 error
= xfs_readsb(mp
, flags
);
1595 goto out_free_fsname
;
1596 error
= xfs_finish_flags(args
, mp
);
1600 error
= xfs_setup_devices(mp
);
1604 if (mp
->m_flags
& XFS_MOUNT_BARRIER
)
1605 xfs_mountfs_check_barriers(mp
);
1607 error
= xfs_filestream_mount(mp
);
1611 error
= xfs_mountfs(mp
);
1613 goto out_filestream_unmount
;
1615 XFS_SEND_MOUNT(mp
, DM_RIGHT_NULL
, args
->mtpt
, args
->fsname
);
1618 sb
->s_magic
= XFS_SB_MAGIC
;
1619 sb
->s_blocksize
= mp
->m_sb
.sb_blocksize
;
1620 sb
->s_blocksize_bits
= ffs(sb
->s_blocksize
) - 1;
1621 sb
->s_maxbytes
= xfs_max_file_offset(sb
->s_blocksize_bits
);
1622 sb
->s_time_gran
= 1;
1623 set_posix_acl_flag(sb
);
1625 root
= igrab(VFS_I(mp
->m_rootip
));
1630 if (is_bad_inode(root
)) {
1634 sb
->s_root
= d_alloc_root(root
);
1640 error
= xfs_syncd_init(mp
);
1644 xfs_itrace_exit(XFS_I(sb
->s_root
->d_inode
));
1649 out_filestream_unmount
:
1650 xfs_filestream_unmount(mp
);
1654 xfs_free_fsname(mp
);
1655 xfs_icsb_destroy_counters(mp
);
1656 xfs_close_devices(mp
);
1677 * Blow away any referenced inode in the filestreams cache.
1678 * This can and will cause log traffic as inodes go inactive
1681 xfs_filestream_unmount(mp
);
1683 XFS_bflush(mp
->m_ddev_targp
);
1684 error
= xfs_unmount_flush(mp
, 0);
1693 struct file_system_type
*fs_type
,
1695 const char *dev_name
,
1697 struct vfsmount
*mnt
)
1699 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, xfs_fs_fill_super
,
1703 static struct super_operations xfs_super_operations
= {
1704 .alloc_inode
= xfs_fs_alloc_inode
,
1705 .destroy_inode
= xfs_fs_destroy_inode
,
1706 .write_inode
= xfs_fs_write_inode
,
1707 .clear_inode
= xfs_fs_clear_inode
,
1708 .put_super
= xfs_fs_put_super
,
1709 .write_super
= xfs_fs_write_super
,
1710 .sync_fs
= xfs_fs_sync_super
,
1711 .write_super_lockfs
= xfs_fs_lockfs
,
1712 .statfs
= xfs_fs_statfs
,
1713 .remount_fs
= xfs_fs_remount
,
1714 .show_options
= xfs_fs_show_options
,
1717 static struct quotactl_ops xfs_quotactl_operations
= {
1718 .quota_sync
= xfs_fs_quotasync
,
1719 .get_xstate
= xfs_fs_getxstate
,
1720 .set_xstate
= xfs_fs_setxstate
,
1721 .get_xquota
= xfs_fs_getxquota
,
1722 .set_xquota
= xfs_fs_setxquota
,
1725 static struct file_system_type xfs_fs_type
= {
1726 .owner
= THIS_MODULE
,
1728 .get_sb
= xfs_fs_get_sb
,
1729 .kill_sb
= kill_block_super
,
1730 .fs_flags
= FS_REQUIRES_DEV
,
1734 xfs_alloc_trace_bufs(void)
1736 #ifdef XFS_ALLOC_TRACE
1737 xfs_alloc_trace_buf
= ktrace_alloc(XFS_ALLOC_TRACE_SIZE
, KM_MAYFAIL
);
1738 if (!xfs_alloc_trace_buf
)
1741 #ifdef XFS_BMAP_TRACE
1742 xfs_bmap_trace_buf
= ktrace_alloc(XFS_BMAP_TRACE_SIZE
, KM_MAYFAIL
);
1743 if (!xfs_bmap_trace_buf
)
1744 goto out_free_alloc_trace
;
1746 #ifdef XFS_BTREE_TRACE
1747 xfs_allocbt_trace_buf
= ktrace_alloc(XFS_ALLOCBT_TRACE_SIZE
,
1749 if (!xfs_allocbt_trace_buf
)
1750 goto out_free_bmap_trace
;
1752 xfs_inobt_trace_buf
= ktrace_alloc(XFS_INOBT_TRACE_SIZE
, KM_MAYFAIL
);
1753 if (!xfs_inobt_trace_buf
)
1754 goto out_free_allocbt_trace
;
1756 xfs_bmbt_trace_buf
= ktrace_alloc(XFS_BMBT_TRACE_SIZE
, KM_MAYFAIL
);
1757 if (!xfs_bmbt_trace_buf
)
1758 goto out_free_inobt_trace
;
1760 #ifdef XFS_ATTR_TRACE
1761 xfs_attr_trace_buf
= ktrace_alloc(XFS_ATTR_TRACE_SIZE
, KM_MAYFAIL
);
1762 if (!xfs_attr_trace_buf
)
1763 goto out_free_bmbt_trace
;
1765 #ifdef XFS_DIR2_TRACE
1766 xfs_dir2_trace_buf
= ktrace_alloc(XFS_DIR2_GTRACE_SIZE
, KM_MAYFAIL
);
1767 if (!xfs_dir2_trace_buf
)
1768 goto out_free_attr_trace
;
1773 #ifdef XFS_DIR2_TRACE
1774 out_free_attr_trace
:
1776 #ifdef XFS_ATTR_TRACE
1777 ktrace_free(xfs_attr_trace_buf
);
1778 out_free_bmbt_trace
:
1780 #ifdef XFS_BTREE_TRACE
1781 ktrace_free(xfs_bmbt_trace_buf
);
1782 out_free_inobt_trace
:
1783 ktrace_free(xfs_inobt_trace_buf
);
1784 out_free_allocbt_trace
:
1785 ktrace_free(xfs_allocbt_trace_buf
);
1786 out_free_bmap_trace
:
1788 #ifdef XFS_BMAP_TRACE
1789 ktrace_free(xfs_bmap_trace_buf
);
1790 out_free_alloc_trace
:
1792 #ifdef XFS_ALLOC_TRACE
1793 ktrace_free(xfs_alloc_trace_buf
);
1800 xfs_free_trace_bufs(void)
1802 #ifdef XFS_DIR2_TRACE
1803 ktrace_free(xfs_dir2_trace_buf
);
1805 #ifdef XFS_ATTR_TRACE
1806 ktrace_free(xfs_attr_trace_buf
);
1808 #ifdef XFS_BTREE_TRACE
1809 ktrace_free(xfs_bmbt_trace_buf
);
1810 ktrace_free(xfs_inobt_trace_buf
);
1811 ktrace_free(xfs_allocbt_trace_buf
);
1813 #ifdef XFS_BMAP_TRACE
1814 ktrace_free(xfs_bmap_trace_buf
);
1816 #ifdef XFS_ALLOC_TRACE
1817 ktrace_free(xfs_alloc_trace_buf
);
1822 xfs_init_zones(void)
1825 xfs_ioend_zone
= kmem_zone_init(sizeof(xfs_ioend_t
), "xfs_ioend");
1826 if (!xfs_ioend_zone
)
1829 xfs_ioend_pool
= mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE
,
1831 if (!xfs_ioend_pool
)
1832 goto out_destroy_ioend_zone
;
1834 xfs_log_ticket_zone
= kmem_zone_init(sizeof(xlog_ticket_t
),
1836 if (!xfs_log_ticket_zone
)
1837 goto out_destroy_ioend_pool
;
1839 xfs_bmap_free_item_zone
= kmem_zone_init(sizeof(xfs_bmap_free_item_t
),
1840 "xfs_bmap_free_item");
1841 if (!xfs_bmap_free_item_zone
)
1842 goto out_destroy_log_ticket_zone
;
1844 xfs_btree_cur_zone
= kmem_zone_init(sizeof(xfs_btree_cur_t
),
1846 if (!xfs_btree_cur_zone
)
1847 goto out_destroy_bmap_free_item_zone
;
1849 xfs_da_state_zone
= kmem_zone_init(sizeof(xfs_da_state_t
),
1851 if (!xfs_da_state_zone
)
1852 goto out_destroy_btree_cur_zone
;
1854 xfs_dabuf_zone
= kmem_zone_init(sizeof(xfs_dabuf_t
), "xfs_dabuf");
1855 if (!xfs_dabuf_zone
)
1856 goto out_destroy_da_state_zone
;
1858 xfs_ifork_zone
= kmem_zone_init(sizeof(xfs_ifork_t
), "xfs_ifork");
1859 if (!xfs_ifork_zone
)
1860 goto out_destroy_dabuf_zone
;
1862 xfs_trans_zone
= kmem_zone_init(sizeof(xfs_trans_t
), "xfs_trans");
1863 if (!xfs_trans_zone
)
1864 goto out_destroy_ifork_zone
;
1867 * The size of the zone allocated buf log item is the maximum
1868 * size possible under XFS. This wastes a little bit of memory,
1869 * but it is much faster.
1871 xfs_buf_item_zone
= kmem_zone_init((sizeof(xfs_buf_log_item_t
) +
1872 (((XFS_MAX_BLOCKSIZE
/ XFS_BLI_CHUNK
) /
1873 NBWORD
) * sizeof(int))), "xfs_buf_item");
1874 if (!xfs_buf_item_zone
)
1875 goto out_destroy_trans_zone
;
1877 xfs_efd_zone
= kmem_zone_init((sizeof(xfs_efd_log_item_t
) +
1878 ((XFS_EFD_MAX_FAST_EXTENTS
- 1) *
1879 sizeof(xfs_extent_t
))), "xfs_efd_item");
1881 goto out_destroy_buf_item_zone
;
1883 xfs_efi_zone
= kmem_zone_init((sizeof(xfs_efi_log_item_t
) +
1884 ((XFS_EFI_MAX_FAST_EXTENTS
- 1) *
1885 sizeof(xfs_extent_t
))), "xfs_efi_item");
1887 goto out_destroy_efd_zone
;
1890 kmem_zone_init_flags(sizeof(xfs_inode_t
), "xfs_inode",
1891 KM_ZONE_HWALIGN
| KM_ZONE_RECLAIM
| KM_ZONE_SPREAD
,
1892 xfs_fs_inode_init_once
);
1893 if (!xfs_inode_zone
)
1894 goto out_destroy_efi_zone
;
1897 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t
), "xfs_ili",
1898 KM_ZONE_SPREAD
, NULL
);
1900 goto out_destroy_inode_zone
;
1902 #ifdef CONFIG_XFS_POSIX_ACL
1903 xfs_acl_zone
= kmem_zone_init(sizeof(xfs_acl_t
), "xfs_acl");
1905 goto out_destroy_ili_zone
;
1910 #ifdef CONFIG_XFS_POSIX_ACL
1911 out_destroy_ili_zone
:
1913 kmem_zone_destroy(xfs_ili_zone
);
1914 out_destroy_inode_zone
:
1915 kmem_zone_destroy(xfs_inode_zone
);
1916 out_destroy_efi_zone
:
1917 kmem_zone_destroy(xfs_efi_zone
);
1918 out_destroy_efd_zone
:
1919 kmem_zone_destroy(xfs_efd_zone
);
1920 out_destroy_buf_item_zone
:
1921 kmem_zone_destroy(xfs_buf_item_zone
);
1922 out_destroy_trans_zone
:
1923 kmem_zone_destroy(xfs_trans_zone
);
1924 out_destroy_ifork_zone
:
1925 kmem_zone_destroy(xfs_ifork_zone
);
1926 out_destroy_dabuf_zone
:
1927 kmem_zone_destroy(xfs_dabuf_zone
);
1928 out_destroy_da_state_zone
:
1929 kmem_zone_destroy(xfs_da_state_zone
);
1930 out_destroy_btree_cur_zone
:
1931 kmem_zone_destroy(xfs_btree_cur_zone
);
1932 out_destroy_bmap_free_item_zone
:
1933 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1934 out_destroy_log_ticket_zone
:
1935 kmem_zone_destroy(xfs_log_ticket_zone
);
1936 out_destroy_ioend_pool
:
1937 mempool_destroy(xfs_ioend_pool
);
1938 out_destroy_ioend_zone
:
1939 kmem_zone_destroy(xfs_ioend_zone
);
1945 xfs_destroy_zones(void)
1947 #ifdef CONFIG_XFS_POSIX_ACL
1948 kmem_zone_destroy(xfs_acl_zone
);
1950 kmem_zone_destroy(xfs_ili_zone
);
1951 kmem_zone_destroy(xfs_inode_zone
);
1952 kmem_zone_destroy(xfs_efi_zone
);
1953 kmem_zone_destroy(xfs_efd_zone
);
1954 kmem_zone_destroy(xfs_buf_item_zone
);
1955 kmem_zone_destroy(xfs_trans_zone
);
1956 kmem_zone_destroy(xfs_ifork_zone
);
1957 kmem_zone_destroy(xfs_dabuf_zone
);
1958 kmem_zone_destroy(xfs_da_state_zone
);
1959 kmem_zone_destroy(xfs_btree_cur_zone
);
1960 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1961 kmem_zone_destroy(xfs_log_ticket_zone
);
1962 mempool_destroy(xfs_ioend_pool
);
1963 kmem_zone_destroy(xfs_ioend_zone
);
1971 static char message
[] __initdata
= KERN_INFO \
1972 XFS_VERSION_STRING
" with " XFS_BUILD_OPTIONS
" enabled\n";
1980 error
= xfs_init_zones();
1984 error
= xfs_alloc_trace_bufs();
1986 goto out_destroy_zones
;
1988 error
= xfs_mru_cache_init();
1990 goto out_free_trace_buffers
;
1992 error
= xfs_filestream_init();
1994 goto out_mru_cache_uninit
;
1996 error
= xfs_buf_init();
1998 goto out_filestream_uninit
;
2000 error
= xfs_init_procfs();
2002 goto out_buf_terminate
;
2004 error
= xfs_sysctl_register();
2006 goto out_cleanup_procfs
;
2010 error
= register_filesystem(&xfs_fs_type
);
2012 goto out_sysctl_unregister
;
2015 out_sysctl_unregister
:
2016 xfs_sysctl_unregister();
2018 xfs_cleanup_procfs();
2020 xfs_buf_terminate();
2021 out_filestream_uninit
:
2022 xfs_filestream_uninit();
2023 out_mru_cache_uninit
:
2024 xfs_mru_cache_uninit();
2025 out_free_trace_buffers
:
2026 xfs_free_trace_bufs();
2028 xfs_destroy_zones();
2037 unregister_filesystem(&xfs_fs_type
);
2038 xfs_sysctl_unregister();
2039 xfs_cleanup_procfs();
2040 xfs_buf_terminate();
2041 xfs_filestream_uninit();
2042 xfs_mru_cache_uninit();
2043 xfs_free_trace_bufs();
2044 xfs_destroy_zones();
2048 module_init(init_xfs_fs
);
2049 module_exit(exit_xfs_fs
);
2051 MODULE_AUTHOR("Silicon Graphics, Inc.");
2052 MODULE_DESCRIPTION(XFS_VERSION_STRING
" with " XFS_BUILD_OPTIONS
" enabled");
2053 MODULE_LICENSE("GPL");