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
20 #include "xfs_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
25 #include "xfs_mount.h"
26 #include "xfs_da_format.h"
27 #include "xfs_inode.h"
28 #include "xfs_btree.h"
30 #include "xfs_alloc.h"
31 #include "xfs_error.h"
32 #include "xfs_fsops.h"
33 #include "xfs_trans.h"
34 #include "xfs_buf_item.h"
36 #include "xfs_log_priv.h"
37 #include "xfs_da_btree.h"
39 #include "xfs_extfree_item.h"
40 #include "xfs_mru_cache.h"
41 #include "xfs_inode_item.h"
42 #include "xfs_icache.h"
43 #include "xfs_trace.h"
44 #include "xfs_icreate_item.h"
45 #include "xfs_filestream.h"
46 #include "xfs_quota.h"
47 #include "xfs_sysfs.h"
49 #include <linux/namei.h>
50 #include <linux/init.h>
51 #include <linux/slab.h>
52 #include <linux/mount.h>
53 #include <linux/mempool.h>
54 #include <linux/writeback.h>
55 #include <linux/kthread.h>
56 #include <linux/freezer.h>
57 #include <linux/parser.h>
59 static const struct super_operations xfs_super_operations
;
60 static kmem_zone_t
*xfs_ioend_zone
;
61 mempool_t
*xfs_ioend_pool
;
63 static struct kset
*xfs_kset
; /* top-level xfs sysfs dir */
64 static struct xfs_kobj xfs_stats_kobj
; /* global stats sysfs attrs */
66 static struct xfs_kobj xfs_dbg_kobj
; /* global debug sysfs attrs */
69 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
70 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
71 #define MNTOPT_LOGDEV "logdev" /* log device */
72 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
73 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
74 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
75 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
76 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
77 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
78 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
79 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
80 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
81 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
82 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
83 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
84 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
85 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
86 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
87 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
88 * unwritten extent conversion */
89 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
90 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
91 #define MNTOPT_32BITINODE "inode32" /* inode allocation limited to
92 * XFS_MAXINUMBER_32 */
93 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
94 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
95 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
96 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
98 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
99 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
100 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
101 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
102 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
103 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
104 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
105 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
106 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
107 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
108 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
109 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
110 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
111 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
112 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
113 #define MNTOPT_DISCARD "discard" /* Discard unused blocks */
114 #define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */
116 #define MNTOPT_DAX "dax" /* Enable direct access to bdev pages */
119 * Table driven mount option parser.
121 * Currently only used for remount, but it will be used for mount
122 * in the future, too.
132 static const match_table_t tokens
= {
133 {Opt_barrier
, "barrier"},
134 {Opt_nobarrier
, "nobarrier"},
135 {Opt_inode64
, "inode64"},
136 {Opt_inode32
, "inode32"},
142 suffix_kstrtoint(char *s
, unsigned int base
, int *res
)
144 int last
, shift_left_factor
= 0, _res
;
147 last
= strlen(value
) - 1;
148 if (value
[last
] == 'K' || value
[last
] == 'k') {
149 shift_left_factor
= 10;
152 if (value
[last
] == 'M' || value
[last
] == 'm') {
153 shift_left_factor
= 20;
156 if (value
[last
] == 'G' || value
[last
] == 'g') {
157 shift_left_factor
= 30;
161 if (kstrtoint(s
, base
, &_res
))
163 *res
= _res
<< shift_left_factor
;
168 * This function fills in xfs_mount_t fields based on mount args.
169 * Note: the superblock has _not_ yet been read in.
171 * Note that this function leaks the various device name allocations on
172 * failure. The caller takes care of them.
176 struct xfs_mount
*mp
,
179 struct super_block
*sb
= mp
->m_super
;
180 char *this_char
, *value
;
184 __uint8_t iosizelog
= 0;
187 * set up the mount name first so all the errors will refer to the
190 mp
->m_fsname
= kstrndup(sb
->s_id
, MAXNAMELEN
, GFP_KERNEL
);
193 mp
->m_fsname_len
= strlen(mp
->m_fsname
) + 1;
196 * Copy binary VFS mount flags we are interested in.
198 if (sb
->s_flags
& MS_RDONLY
)
199 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
200 if (sb
->s_flags
& MS_DIRSYNC
)
201 mp
->m_flags
|= XFS_MOUNT_DIRSYNC
;
202 if (sb
->s_flags
& MS_SYNCHRONOUS
)
203 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
206 * Set some default flags that could be cleared by the mount option
209 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
210 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
213 * These can be overridden by the mount option parsing.
221 while ((this_char
= strsep(&options
, ",")) != NULL
) {
224 if ((value
= strchr(this_char
, '=')) != NULL
)
227 if (!strcmp(this_char
, MNTOPT_LOGBUFS
)) {
228 if (!value
|| !*value
) {
229 xfs_warn(mp
, "%s option requires an argument",
233 if (kstrtoint(value
, 10, &mp
->m_logbufs
))
235 } else if (!strcmp(this_char
, MNTOPT_LOGBSIZE
)) {
236 if (!value
|| !*value
) {
237 xfs_warn(mp
, "%s option requires an argument",
241 if (suffix_kstrtoint(value
, 10, &mp
->m_logbsize
))
243 } else if (!strcmp(this_char
, MNTOPT_LOGDEV
)) {
244 if (!value
|| !*value
) {
245 xfs_warn(mp
, "%s option requires an argument",
249 mp
->m_logname
= kstrndup(value
, MAXNAMELEN
, GFP_KERNEL
);
252 } else if (!strcmp(this_char
, MNTOPT_MTPT
)) {
253 xfs_warn(mp
, "%s option not allowed on this system",
256 } else if (!strcmp(this_char
, MNTOPT_RTDEV
)) {
257 if (!value
|| !*value
) {
258 xfs_warn(mp
, "%s option requires an argument",
262 mp
->m_rtname
= kstrndup(value
, MAXNAMELEN
, GFP_KERNEL
);
265 } else if (!strcmp(this_char
, MNTOPT_ALLOCSIZE
) ||
266 !strcmp(this_char
, MNTOPT_BIOSIZE
)) {
267 if (!value
|| !*value
) {
268 xfs_warn(mp
, "%s option requires an argument",
272 if (suffix_kstrtoint(value
, 10, &iosize
))
274 iosizelog
= ffs(iosize
) - 1;
275 } else if (!strcmp(this_char
, MNTOPT_GRPID
) ||
276 !strcmp(this_char
, MNTOPT_BSDGROUPS
)) {
277 mp
->m_flags
|= XFS_MOUNT_GRPID
;
278 } else if (!strcmp(this_char
, MNTOPT_NOGRPID
) ||
279 !strcmp(this_char
, MNTOPT_SYSVGROUPS
)) {
280 mp
->m_flags
&= ~XFS_MOUNT_GRPID
;
281 } else if (!strcmp(this_char
, MNTOPT_WSYNC
)) {
282 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
283 } else if (!strcmp(this_char
, MNTOPT_NORECOVERY
)) {
284 mp
->m_flags
|= XFS_MOUNT_NORECOVERY
;
285 } else if (!strcmp(this_char
, MNTOPT_NOALIGN
)) {
286 mp
->m_flags
|= XFS_MOUNT_NOALIGN
;
287 } else if (!strcmp(this_char
, MNTOPT_SWALLOC
)) {
288 mp
->m_flags
|= XFS_MOUNT_SWALLOC
;
289 } else if (!strcmp(this_char
, MNTOPT_SUNIT
)) {
290 if (!value
|| !*value
) {
291 xfs_warn(mp
, "%s option requires an argument",
295 if (kstrtoint(value
, 10, &dsunit
))
297 } else if (!strcmp(this_char
, MNTOPT_SWIDTH
)) {
298 if (!value
|| !*value
) {
299 xfs_warn(mp
, "%s option requires an argument",
303 if (kstrtoint(value
, 10, &dswidth
))
305 } else if (!strcmp(this_char
, MNTOPT_32BITINODE
)) {
306 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
307 } else if (!strcmp(this_char
, MNTOPT_64BITINODE
)) {
308 mp
->m_flags
&= ~XFS_MOUNT_SMALL_INUMS
;
309 } else if (!strcmp(this_char
, MNTOPT_NOUUID
)) {
310 mp
->m_flags
|= XFS_MOUNT_NOUUID
;
311 } else if (!strcmp(this_char
, MNTOPT_BARRIER
)) {
312 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
313 } else if (!strcmp(this_char
, MNTOPT_NOBARRIER
)) {
314 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
315 } else if (!strcmp(this_char
, MNTOPT_IKEEP
)) {
316 mp
->m_flags
|= XFS_MOUNT_IKEEP
;
317 } else if (!strcmp(this_char
, MNTOPT_NOIKEEP
)) {
318 mp
->m_flags
&= ~XFS_MOUNT_IKEEP
;
319 } else if (!strcmp(this_char
, MNTOPT_LARGEIO
)) {
320 mp
->m_flags
&= ~XFS_MOUNT_COMPAT_IOSIZE
;
321 } else if (!strcmp(this_char
, MNTOPT_NOLARGEIO
)) {
322 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
323 } else if (!strcmp(this_char
, MNTOPT_ATTR2
)) {
324 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
325 } else if (!strcmp(this_char
, MNTOPT_NOATTR2
)) {
326 mp
->m_flags
&= ~XFS_MOUNT_ATTR2
;
327 mp
->m_flags
|= XFS_MOUNT_NOATTR2
;
328 } else if (!strcmp(this_char
, MNTOPT_FILESTREAM
)) {
329 mp
->m_flags
|= XFS_MOUNT_FILESTREAMS
;
330 } else if (!strcmp(this_char
, MNTOPT_NOQUOTA
)) {
331 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACCT
;
332 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ENFD
;
333 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACTIVE
;
334 } else if (!strcmp(this_char
, MNTOPT_QUOTA
) ||
335 !strcmp(this_char
, MNTOPT_UQUOTA
) ||
336 !strcmp(this_char
, MNTOPT_USRQUOTA
)) {
337 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
|
339 } else if (!strcmp(this_char
, MNTOPT_QUOTANOENF
) ||
340 !strcmp(this_char
, MNTOPT_UQUOTANOENF
)) {
341 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
);
342 mp
->m_qflags
&= ~XFS_UQUOTA_ENFD
;
343 } else if (!strcmp(this_char
, MNTOPT_PQUOTA
) ||
344 !strcmp(this_char
, MNTOPT_PRJQUOTA
)) {
345 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
|
347 } else if (!strcmp(this_char
, MNTOPT_PQUOTANOENF
)) {
348 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
);
349 mp
->m_qflags
&= ~XFS_PQUOTA_ENFD
;
350 } else if (!strcmp(this_char
, MNTOPT_GQUOTA
) ||
351 !strcmp(this_char
, MNTOPT_GRPQUOTA
)) {
352 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
|
354 } else if (!strcmp(this_char
, MNTOPT_GQUOTANOENF
)) {
355 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
);
356 mp
->m_qflags
&= ~XFS_GQUOTA_ENFD
;
357 } else if (!strcmp(this_char
, MNTOPT_DISCARD
)) {
358 mp
->m_flags
|= XFS_MOUNT_DISCARD
;
359 } else if (!strcmp(this_char
, MNTOPT_NODISCARD
)) {
360 mp
->m_flags
&= ~XFS_MOUNT_DISCARD
;
362 } else if (!strcmp(this_char
, MNTOPT_DAX
)) {
363 mp
->m_flags
|= XFS_MOUNT_DAX
;
366 xfs_warn(mp
, "unknown mount option [%s].", this_char
);
372 * no recovery flag requires a read-only mount
374 if ((mp
->m_flags
& XFS_MOUNT_NORECOVERY
) &&
375 !(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
376 xfs_warn(mp
, "no-recovery mounts must be read-only.");
380 if ((mp
->m_flags
& XFS_MOUNT_NOALIGN
) && (dsunit
|| dswidth
)) {
382 "sunit and swidth options incompatible with the noalign option");
386 #ifndef CONFIG_XFS_QUOTA
387 if (XFS_IS_QUOTA_RUNNING(mp
)) {
388 xfs_warn(mp
, "quota support not available in this kernel.");
393 if ((dsunit
&& !dswidth
) || (!dsunit
&& dswidth
)) {
394 xfs_warn(mp
, "sunit and swidth must be specified together");
398 if (dsunit
&& (dswidth
% dsunit
!= 0)) {
400 "stripe width (%d) must be a multiple of the stripe unit (%d)",
406 if (dsunit
&& !(mp
->m_flags
& XFS_MOUNT_NOALIGN
)) {
408 * At this point the superblock has not been read
409 * in, therefore we do not know the block size.
410 * Before the mount call ends we will convert
413 mp
->m_dalign
= dsunit
;
414 mp
->m_swidth
= dswidth
;
417 if (mp
->m_logbufs
!= -1 &&
418 mp
->m_logbufs
!= 0 &&
419 (mp
->m_logbufs
< XLOG_MIN_ICLOGS
||
420 mp
->m_logbufs
> XLOG_MAX_ICLOGS
)) {
421 xfs_warn(mp
, "invalid logbufs value: %d [not %d-%d]",
422 mp
->m_logbufs
, XLOG_MIN_ICLOGS
, XLOG_MAX_ICLOGS
);
425 if (mp
->m_logbsize
!= -1 &&
426 mp
->m_logbsize
!= 0 &&
427 (mp
->m_logbsize
< XLOG_MIN_RECORD_BSIZE
||
428 mp
->m_logbsize
> XLOG_MAX_RECORD_BSIZE
||
429 !is_power_of_2(mp
->m_logbsize
))) {
431 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
437 if (iosizelog
> XFS_MAX_IO_LOG
||
438 iosizelog
< XFS_MIN_IO_LOG
) {
439 xfs_warn(mp
, "invalid log iosize: %d [not %d-%d]",
440 iosizelog
, XFS_MIN_IO_LOG
,
445 mp
->m_flags
|= XFS_MOUNT_DFLT_IOSIZE
;
446 mp
->m_readio_log
= iosizelog
;
447 mp
->m_writeio_log
= iosizelog
;
453 struct proc_xfs_info
{
460 struct xfs_mount
*mp
,
463 static struct proc_xfs_info xfs_info_set
[] = {
464 /* the few simple ones we can get from the mount struct */
465 { XFS_MOUNT_IKEEP
, "," MNTOPT_IKEEP
},
466 { XFS_MOUNT_WSYNC
, "," MNTOPT_WSYNC
},
467 { XFS_MOUNT_NOALIGN
, "," MNTOPT_NOALIGN
},
468 { XFS_MOUNT_SWALLOC
, "," MNTOPT_SWALLOC
},
469 { XFS_MOUNT_NOUUID
, "," MNTOPT_NOUUID
},
470 { XFS_MOUNT_NORECOVERY
, "," MNTOPT_NORECOVERY
},
471 { XFS_MOUNT_ATTR2
, "," MNTOPT_ATTR2
},
472 { XFS_MOUNT_FILESTREAMS
, "," MNTOPT_FILESTREAM
},
473 { XFS_MOUNT_GRPID
, "," MNTOPT_GRPID
},
474 { XFS_MOUNT_DISCARD
, "," MNTOPT_DISCARD
},
475 { XFS_MOUNT_SMALL_INUMS
, "," MNTOPT_32BITINODE
},
476 { XFS_MOUNT_DAX
, "," MNTOPT_DAX
},
479 static struct proc_xfs_info xfs_info_unset
[] = {
480 /* the few simple ones we can get from the mount struct */
481 { XFS_MOUNT_COMPAT_IOSIZE
, "," MNTOPT_LARGEIO
},
482 { XFS_MOUNT_BARRIER
, "," MNTOPT_NOBARRIER
},
483 { XFS_MOUNT_SMALL_INUMS
, "," MNTOPT_64BITINODE
},
486 struct proc_xfs_info
*xfs_infop
;
488 for (xfs_infop
= xfs_info_set
; xfs_infop
->flag
; xfs_infop
++) {
489 if (mp
->m_flags
& xfs_infop
->flag
)
490 seq_puts(m
, xfs_infop
->str
);
492 for (xfs_infop
= xfs_info_unset
; xfs_infop
->flag
; xfs_infop
++) {
493 if (!(mp
->m_flags
& xfs_infop
->flag
))
494 seq_puts(m
, xfs_infop
->str
);
497 if (mp
->m_flags
& XFS_MOUNT_DFLT_IOSIZE
)
498 seq_printf(m
, "," MNTOPT_ALLOCSIZE
"=%dk",
499 (int)(1 << mp
->m_writeio_log
) >> 10);
501 if (mp
->m_logbufs
> 0)
502 seq_printf(m
, "," MNTOPT_LOGBUFS
"=%d", mp
->m_logbufs
);
503 if (mp
->m_logbsize
> 0)
504 seq_printf(m
, "," MNTOPT_LOGBSIZE
"=%dk", mp
->m_logbsize
>> 10);
507 seq_show_option(m
, MNTOPT_LOGDEV
, mp
->m_logname
);
509 seq_show_option(m
, MNTOPT_RTDEV
, mp
->m_rtname
);
511 if (mp
->m_dalign
> 0)
512 seq_printf(m
, "," MNTOPT_SUNIT
"=%d",
513 (int)XFS_FSB_TO_BB(mp
, mp
->m_dalign
));
514 if (mp
->m_swidth
> 0)
515 seq_printf(m
, "," MNTOPT_SWIDTH
"=%d",
516 (int)XFS_FSB_TO_BB(mp
, mp
->m_swidth
));
518 if (mp
->m_qflags
& (XFS_UQUOTA_ACCT
|XFS_UQUOTA_ENFD
))
519 seq_puts(m
, "," MNTOPT_USRQUOTA
);
520 else if (mp
->m_qflags
& XFS_UQUOTA_ACCT
)
521 seq_puts(m
, "," MNTOPT_UQUOTANOENF
);
523 if (mp
->m_qflags
& XFS_PQUOTA_ACCT
) {
524 if (mp
->m_qflags
& XFS_PQUOTA_ENFD
)
525 seq_puts(m
, "," MNTOPT_PRJQUOTA
);
527 seq_puts(m
, "," MNTOPT_PQUOTANOENF
);
529 if (mp
->m_qflags
& XFS_GQUOTA_ACCT
) {
530 if (mp
->m_qflags
& XFS_GQUOTA_ENFD
)
531 seq_puts(m
, "," MNTOPT_GRPQUOTA
);
533 seq_puts(m
, "," MNTOPT_GQUOTANOENF
);
536 if (!(mp
->m_qflags
& XFS_ALL_QUOTA_ACCT
))
537 seq_puts(m
, "," MNTOPT_NOQUOTA
);
543 unsigned int blockshift
)
545 unsigned int pagefactor
= 1;
546 unsigned int bitshift
= BITS_PER_LONG
- 1;
548 /* Figure out maximum filesize, on Linux this can depend on
549 * the filesystem blocksize (on 32 bit platforms).
550 * __block_write_begin does this in an [unsigned] long...
551 * page->index << (PAGE_CACHE_SHIFT - bbits)
552 * So, for page sized blocks (4K on 32 bit platforms),
553 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
554 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
555 * but for smaller blocksizes it is less (bbits = log2 bsize).
556 * Note1: get_block_t takes a long (implicit cast from above)
557 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
558 * can optionally convert the [unsigned] long from above into
559 * an [unsigned] long long.
562 #if BITS_PER_LONG == 32
563 # if defined(CONFIG_LBDAF)
564 ASSERT(sizeof(sector_t
) == 8);
565 pagefactor
= PAGE_CACHE_SIZE
;
566 bitshift
= BITS_PER_LONG
;
568 pagefactor
= PAGE_CACHE_SIZE
>> (PAGE_CACHE_SHIFT
- blockshift
);
572 return (((__uint64_t
)pagefactor
) << bitshift
) - 1;
576 * xfs_set_inode32() and xfs_set_inode64() are passed an agcount
577 * because in the growfs case, mp->m_sb.sb_agcount is not updated
578 * yet to the potentially higher ag count.
581 xfs_set_inode32(struct xfs_mount
*mp
, xfs_agnumber_t agcount
)
583 xfs_agnumber_t index
= 0;
584 xfs_agnumber_t maxagi
= 0;
585 xfs_sb_t
*sbp
= &mp
->m_sb
;
586 xfs_agnumber_t max_metadata
;
591 /* Calculate how much should be reserved for inodes to meet
592 * the max inode percentage.
594 if (mp
->m_maxicount
) {
597 icount
= sbp
->sb_dblocks
* sbp
->sb_imax_pct
;
599 icount
+= sbp
->sb_agblocks
- 1;
600 do_div(icount
, sbp
->sb_agblocks
);
601 max_metadata
= icount
;
603 max_metadata
= agcount
;
606 agino
= XFS_OFFBNO_TO_AGINO(mp
, sbp
->sb_agblocks
- 1, 0);
608 for (index
= 0; index
< agcount
; index
++) {
609 ino
= XFS_AGINO_TO_INO(mp
, index
, agino
);
611 if (ino
> XFS_MAXINUMBER_32
) {
612 pag
= xfs_perag_get(mp
, index
);
613 pag
->pagi_inodeok
= 0;
614 pag
->pagf_metadata
= 0;
619 pag
= xfs_perag_get(mp
, index
);
620 pag
->pagi_inodeok
= 1;
622 if (index
< max_metadata
)
623 pag
->pagf_metadata
= 1;
626 mp
->m_flags
|= (XFS_MOUNT_32BITINODES
|
627 XFS_MOUNT_SMALL_INUMS
);
633 xfs_set_inode64(struct xfs_mount
*mp
, xfs_agnumber_t agcount
)
635 xfs_agnumber_t index
= 0;
637 for (index
= 0; index
< agcount
; index
++) {
638 struct xfs_perag
*pag
;
640 pag
= xfs_perag_get(mp
, index
);
641 pag
->pagi_inodeok
= 1;
642 pag
->pagf_metadata
= 0;
646 /* There is no need for lock protection on m_flags,
647 * the rw_semaphore of the VFS superblock is locked
648 * during mount/umount/remount operations, so this is
649 * enough to avoid concurency on the m_flags field
651 mp
->m_flags
&= ~(XFS_MOUNT_32BITINODES
|
652 XFS_MOUNT_SMALL_INUMS
);
660 struct block_device
**bdevp
)
664 *bdevp
= blkdev_get_by_path(name
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
,
666 if (IS_ERR(*bdevp
)) {
667 error
= PTR_ERR(*bdevp
);
668 xfs_warn(mp
, "Invalid device [%s], error=%d", name
, error
);
676 struct block_device
*bdev
)
679 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
683 xfs_blkdev_issue_flush(
684 xfs_buftarg_t
*buftarg
)
686 blkdev_issue_flush(buftarg
->bt_bdev
, GFP_NOFS
, NULL
);
691 struct xfs_mount
*mp
)
693 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
694 struct block_device
*logdev
= mp
->m_logdev_targp
->bt_bdev
;
695 xfs_free_buftarg(mp
, mp
->m_logdev_targp
);
696 xfs_blkdev_put(logdev
);
698 if (mp
->m_rtdev_targp
) {
699 struct block_device
*rtdev
= mp
->m_rtdev_targp
->bt_bdev
;
700 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
701 xfs_blkdev_put(rtdev
);
703 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
707 * The file system configurations are:
708 * (1) device (partition) with data and internal log
709 * (2) logical volume with data and log subvolumes.
710 * (3) logical volume with data, log, and realtime subvolumes.
712 * We only have to handle opening the log and realtime volumes here if
713 * they are present. The data subvolume has already been opened by
714 * get_sb_bdev() and is stored in sb->s_bdev.
718 struct xfs_mount
*mp
)
720 struct block_device
*ddev
= mp
->m_super
->s_bdev
;
721 struct block_device
*logdev
= NULL
, *rtdev
= NULL
;
725 * Open real time and log devices - order is important.
728 error
= xfs_blkdev_get(mp
, mp
->m_logname
, &logdev
);
734 error
= xfs_blkdev_get(mp
, mp
->m_rtname
, &rtdev
);
736 goto out_close_logdev
;
738 if (rtdev
== ddev
|| rtdev
== logdev
) {
740 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
742 goto out_close_rtdev
;
747 * Setup xfs_mount buffer target pointers
750 mp
->m_ddev_targp
= xfs_alloc_buftarg(mp
, ddev
);
751 if (!mp
->m_ddev_targp
)
752 goto out_close_rtdev
;
755 mp
->m_rtdev_targp
= xfs_alloc_buftarg(mp
, rtdev
);
756 if (!mp
->m_rtdev_targp
)
757 goto out_free_ddev_targ
;
760 if (logdev
&& logdev
!= ddev
) {
761 mp
->m_logdev_targp
= xfs_alloc_buftarg(mp
, logdev
);
762 if (!mp
->m_logdev_targp
)
763 goto out_free_rtdev_targ
;
765 mp
->m_logdev_targp
= mp
->m_ddev_targp
;
771 if (mp
->m_rtdev_targp
)
772 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
774 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
776 xfs_blkdev_put(rtdev
);
778 if (logdev
&& logdev
!= ddev
)
779 xfs_blkdev_put(logdev
);
785 * Setup xfs_mount buffer target pointers based on superblock
789 struct xfs_mount
*mp
)
793 error
= xfs_setsize_buftarg(mp
->m_ddev_targp
, mp
->m_sb
.sb_sectsize
);
797 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
798 unsigned int log_sector_size
= BBSIZE
;
800 if (xfs_sb_version_hassector(&mp
->m_sb
))
801 log_sector_size
= mp
->m_sb
.sb_logsectsize
;
802 error
= xfs_setsize_buftarg(mp
->m_logdev_targp
,
807 if (mp
->m_rtdev_targp
) {
808 error
= xfs_setsize_buftarg(mp
->m_rtdev_targp
,
809 mp
->m_sb
.sb_sectsize
);
818 xfs_init_mount_workqueues(
819 struct xfs_mount
*mp
)
821 mp
->m_buf_workqueue
= alloc_workqueue("xfs-buf/%s",
822 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 1, mp
->m_fsname
);
823 if (!mp
->m_buf_workqueue
)
826 mp
->m_data_workqueue
= alloc_workqueue("xfs-data/%s",
827 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
828 if (!mp
->m_data_workqueue
)
829 goto out_destroy_buf
;
831 mp
->m_unwritten_workqueue
= alloc_workqueue("xfs-conv/%s",
832 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
833 if (!mp
->m_unwritten_workqueue
)
834 goto out_destroy_data_iodone_queue
;
836 mp
->m_cil_workqueue
= alloc_workqueue("xfs-cil/%s",
837 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
838 if (!mp
->m_cil_workqueue
)
839 goto out_destroy_unwritten
;
841 mp
->m_reclaim_workqueue
= alloc_workqueue("xfs-reclaim/%s",
842 WQ_FREEZABLE
, 0, mp
->m_fsname
);
843 if (!mp
->m_reclaim_workqueue
)
844 goto out_destroy_cil
;
846 mp
->m_log_workqueue
= alloc_workqueue("xfs-log/%s",
847 WQ_FREEZABLE
|WQ_HIGHPRI
, 0, mp
->m_fsname
);
848 if (!mp
->m_log_workqueue
)
849 goto out_destroy_reclaim
;
851 mp
->m_eofblocks_workqueue
= alloc_workqueue("xfs-eofblocks/%s",
852 WQ_FREEZABLE
, 0, mp
->m_fsname
);
853 if (!mp
->m_eofblocks_workqueue
)
854 goto out_destroy_log
;
859 destroy_workqueue(mp
->m_log_workqueue
);
861 destroy_workqueue(mp
->m_reclaim_workqueue
);
863 destroy_workqueue(mp
->m_cil_workqueue
);
864 out_destroy_unwritten
:
865 destroy_workqueue(mp
->m_unwritten_workqueue
);
866 out_destroy_data_iodone_queue
:
867 destroy_workqueue(mp
->m_data_workqueue
);
869 destroy_workqueue(mp
->m_buf_workqueue
);
875 xfs_destroy_mount_workqueues(
876 struct xfs_mount
*mp
)
878 destroy_workqueue(mp
->m_eofblocks_workqueue
);
879 destroy_workqueue(mp
->m_log_workqueue
);
880 destroy_workqueue(mp
->m_reclaim_workqueue
);
881 destroy_workqueue(mp
->m_cil_workqueue
);
882 destroy_workqueue(mp
->m_data_workqueue
);
883 destroy_workqueue(mp
->m_unwritten_workqueue
);
884 destroy_workqueue(mp
->m_buf_workqueue
);
888 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
889 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
890 * for IO to complete so that we effectively throttle multiple callers to the
891 * rate at which IO is completing.
895 struct xfs_mount
*mp
)
897 struct super_block
*sb
= mp
->m_super
;
899 if (down_read_trylock(&sb
->s_umount
)) {
901 up_read(&sb
->s_umount
);
905 /* Catch misguided souls that try to use this interface on XFS */
906 STATIC
struct inode
*
908 struct super_block
*sb
)
915 * Now that the generic code is guaranteed not to be accessing
916 * the linux inode, we can reclaim the inode.
919 xfs_fs_destroy_inode(
922 struct xfs_inode
*ip
= XFS_I(inode
);
924 trace_xfs_destroy_inode(ip
);
926 XFS_STATS_INC(vn_reclaim
);
928 ASSERT(XFS_FORCED_SHUTDOWN(ip
->i_mount
) || ip
->i_delayed_blks
== 0);
931 * We should never get here with one of the reclaim flags already set.
933 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIMABLE
));
934 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIM
));
937 * We always use background reclaim here because even if the
938 * inode is clean, it still may be under IO and hence we have
939 * to take the flush lock. The background reclaim path handles
940 * this more efficiently than we can here, so simply let background
941 * reclaim tear down all inodes.
943 xfs_inode_set_reclaim_tag(ip
);
947 * Slab object creation initialisation for the XFS inode.
948 * This covers only the idempotent fields in the XFS inode;
949 * all other fields need to be initialised on allocation
950 * from the slab. This avoids the need to repeatedly initialise
951 * fields in the xfs inode that left in the initialise state
952 * when freeing the inode.
955 xfs_fs_inode_init_once(
958 struct xfs_inode
*ip
= inode
;
960 memset(ip
, 0, sizeof(struct xfs_inode
));
963 inode_init_once(VFS_I(ip
));
966 atomic_set(&ip
->i_pincount
, 0);
967 spin_lock_init(&ip
->i_flags_lock
);
969 mrlock_init(&ip
->i_mmaplock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
970 "xfsino", ip
->i_ino
);
971 mrlock_init(&ip
->i_lock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
972 "xfsino", ip
->i_ino
);
979 xfs_inode_t
*ip
= XFS_I(inode
);
981 ASSERT(!rwsem_is_locked(&ip
->i_iolock
.mr_lock
));
983 trace_xfs_evict_inode(ip
);
985 truncate_inode_pages_final(&inode
->i_data
);
987 XFS_STATS_INC(vn_rele
);
988 XFS_STATS_INC(vn_remove
);
994 * We do an unlocked check for XFS_IDONTCACHE here because we are already
995 * serialised against cache hits here via the inode->i_lock and igrab() in
996 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
997 * racing with us, and it avoids needing to grab a spinlock here for every inode
998 * we drop the final reference on.
1002 struct inode
*inode
)
1004 struct xfs_inode
*ip
= XFS_I(inode
);
1006 return generic_drop_inode(inode
) || (ip
->i_flags
& XFS_IDONTCACHE
);
1011 struct xfs_mount
*mp
)
1013 kfree(mp
->m_fsname
);
1014 kfree(mp
->m_rtname
);
1015 kfree(mp
->m_logname
);
1020 struct super_block
*sb
,
1023 struct xfs_mount
*mp
= XFS_M(sb
);
1026 * Doing anything during the async pass would be counterproductive.
1031 xfs_log_force(mp
, XFS_LOG_SYNC
);
1034 * The disk must be active because we're syncing.
1035 * We schedule log work now (now that the disk is
1036 * active) instead of later (when it might not be).
1038 flush_delayed_work(&mp
->m_log
->l_work
);
1046 struct dentry
*dentry
,
1047 struct kstatfs
*statp
)
1049 struct xfs_mount
*mp
= XFS_M(dentry
->d_sb
);
1050 xfs_sb_t
*sbp
= &mp
->m_sb
;
1051 struct xfs_inode
*ip
= XFS_I(d_inode(dentry
));
1052 __uint64_t fakeinos
, id
;
1055 __uint64_t fdblocks
;
1059 statp
->f_type
= XFS_SB_MAGIC
;
1060 statp
->f_namelen
= MAXNAMELEN
- 1;
1062 id
= huge_encode_dev(mp
->m_ddev_targp
->bt_dev
);
1063 statp
->f_fsid
.val
[0] = (u32
)id
;
1064 statp
->f_fsid
.val
[1] = (u32
)(id
>> 32);
1066 icount
= percpu_counter_sum(&mp
->m_icount
);
1067 ifree
= percpu_counter_sum(&mp
->m_ifree
);
1068 fdblocks
= percpu_counter_sum(&mp
->m_fdblocks
);
1070 spin_lock(&mp
->m_sb_lock
);
1071 statp
->f_bsize
= sbp
->sb_blocksize
;
1072 lsize
= sbp
->sb_logstart
? sbp
->sb_logblocks
: 0;
1073 statp
->f_blocks
= sbp
->sb_dblocks
- lsize
;
1074 spin_unlock(&mp
->m_sb_lock
);
1076 statp
->f_bfree
= fdblocks
- XFS_ALLOC_SET_ASIDE(mp
);
1077 statp
->f_bavail
= statp
->f_bfree
;
1079 fakeinos
= statp
->f_bfree
<< sbp
->sb_inopblog
;
1080 statp
->f_files
= MIN(icount
+ fakeinos
, (__uint64_t
)XFS_MAXINUMBER
);
1081 if (mp
->m_maxicount
)
1082 statp
->f_files
= min_t(typeof(statp
->f_files
),
1086 /* If sb_icount overshot maxicount, report actual allocation */
1087 statp
->f_files
= max_t(typeof(statp
->f_files
),
1091 /* make sure statp->f_ffree does not underflow */
1092 ffree
= statp
->f_files
- (icount
- ifree
);
1093 statp
->f_ffree
= max_t(__int64_t
, ffree
, 0);
1096 if ((ip
->i_d
.di_flags
& XFS_DIFLAG_PROJINHERIT
) &&
1097 ((mp
->m_qflags
& (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))) ==
1098 (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))
1099 xfs_qm_statvfs(ip
, statp
);
1104 xfs_save_resvblks(struct xfs_mount
*mp
)
1106 __uint64_t resblks
= 0;
1108 mp
->m_resblks_save
= mp
->m_resblks
;
1109 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1113 xfs_restore_resvblks(struct xfs_mount
*mp
)
1117 if (mp
->m_resblks_save
) {
1118 resblks
= mp
->m_resblks_save
;
1119 mp
->m_resblks_save
= 0;
1121 resblks
= xfs_default_resblks(mp
);
1123 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1127 * Trigger writeback of all the dirty metadata in the file system.
1129 * This ensures that the metadata is written to their location on disk rather
1130 * than just existing in transactions in the log. This means after a quiesce
1131 * there is no log replay required to write the inodes to disk - this is the
1132 * primary difference between a sync and a quiesce.
1134 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1135 * it is started again when appropriate.
1139 struct xfs_mount
*mp
)
1143 /* wait for all modifications to complete */
1144 while (atomic_read(&mp
->m_active_trans
) > 0)
1147 /* force the log to unpin objects from the now complete transactions */
1148 xfs_log_force(mp
, XFS_LOG_SYNC
);
1150 /* reclaim inodes to do any IO before the freeze completes */
1151 xfs_reclaim_inodes(mp
, 0);
1152 xfs_reclaim_inodes(mp
, SYNC_WAIT
);
1154 /* Push the superblock and write an unmount record */
1155 error
= xfs_log_sbcount(mp
);
1157 xfs_warn(mp
, "xfs_attr_quiesce: failed to log sb changes. "
1158 "Frozen image may not be consistent.");
1160 * Just warn here till VFS can correctly support
1161 * read-only remount without racing.
1163 WARN_ON(atomic_read(&mp
->m_active_trans
) != 0);
1165 xfs_log_quiesce(mp
);
1170 struct super_block
*sb
,
1174 struct xfs_mount
*mp
= XFS_M(sb
);
1175 xfs_sb_t
*sbp
= &mp
->m_sb
;
1176 substring_t args
[MAX_OPT_ARGS
];
1180 sync_filesystem(sb
);
1181 while ((p
= strsep(&options
, ",")) != NULL
) {
1187 token
= match_token(p
, tokens
, args
);
1190 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
1193 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
1196 mp
->m_maxagi
= xfs_set_inode64(mp
, sbp
->sb_agcount
);
1199 mp
->m_maxagi
= xfs_set_inode32(mp
, sbp
->sb_agcount
);
1203 * Logically we would return an error here to prevent
1204 * users from believing they might have changed
1205 * mount options using remount which can't be changed.
1207 * But unfortunately mount(8) adds all options from
1208 * mtab and fstab to the mount arguments in some cases
1209 * so we can't blindly reject options, but have to
1210 * check for each specified option if it actually
1211 * differs from the currently set option and only
1212 * reject it if that's the case.
1214 * Until that is implemented we return success for
1215 * every remount request, and silently ignore all
1216 * options that we can't actually change.
1220 "mount option \"%s\" not supported for remount", p
);
1229 if ((mp
->m_flags
& XFS_MOUNT_RDONLY
) && !(*flags
& MS_RDONLY
)) {
1230 if (mp
->m_flags
& XFS_MOUNT_NORECOVERY
) {
1232 "ro->rw transition prohibited on norecovery mount");
1236 mp
->m_flags
&= ~XFS_MOUNT_RDONLY
;
1239 * If this is the first remount to writeable state we
1240 * might have some superblock changes to update.
1242 if (mp
->m_update_sb
) {
1243 error
= xfs_sync_sb(mp
, false);
1245 xfs_warn(mp
, "failed to write sb changes");
1248 mp
->m_update_sb
= false;
1252 * Fill out the reserve pool if it is empty. Use the stashed
1253 * value if it is non-zero, otherwise go with the default.
1255 xfs_restore_resvblks(mp
);
1256 xfs_log_work_queue(mp
);
1260 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
) && (*flags
& MS_RDONLY
)) {
1262 * Before we sync the metadata, we need to free up the reserve
1263 * block pool so that the used block count in the superblock on
1264 * disk is correct at the end of the remount. Stash the current
1265 * reserve pool size so that if we get remounted rw, we can
1266 * return it to the same size.
1268 xfs_save_resvblks(mp
);
1269 xfs_quiesce_attr(mp
);
1270 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
1277 * Second stage of a freeze. The data is already frozen so we only
1278 * need to take care of the metadata. Once that's done sync the superblock
1279 * to the log to dirty it in case of a crash while frozen. This ensures that we
1280 * will recover the unlinked inode lists on the next mount.
1284 struct super_block
*sb
)
1286 struct xfs_mount
*mp
= XFS_M(sb
);
1288 xfs_save_resvblks(mp
);
1289 xfs_quiesce_attr(mp
);
1290 return xfs_sync_sb(mp
, true);
1295 struct super_block
*sb
)
1297 struct xfs_mount
*mp
= XFS_M(sb
);
1299 xfs_restore_resvblks(mp
);
1300 xfs_log_work_queue(mp
);
1305 xfs_fs_show_options(
1307 struct dentry
*root
)
1309 return xfs_showargs(XFS_M(root
->d_sb
), m
);
1313 * This function fills in xfs_mount_t fields based on mount args.
1314 * Note: the superblock _has_ now been read in.
1318 struct xfs_mount
*mp
)
1320 int ronly
= (mp
->m_flags
& XFS_MOUNT_RDONLY
);
1322 /* Fail a mount where the logbuf is smaller than the log stripe */
1323 if (xfs_sb_version_haslogv2(&mp
->m_sb
)) {
1324 if (mp
->m_logbsize
<= 0 &&
1325 mp
->m_sb
.sb_logsunit
> XLOG_BIG_RECORD_BSIZE
) {
1326 mp
->m_logbsize
= mp
->m_sb
.sb_logsunit
;
1327 } else if (mp
->m_logbsize
> 0 &&
1328 mp
->m_logbsize
< mp
->m_sb
.sb_logsunit
) {
1330 "logbuf size must be greater than or equal to log stripe size");
1334 /* Fail a mount if the logbuf is larger than 32K */
1335 if (mp
->m_logbsize
> XLOG_BIG_RECORD_BSIZE
) {
1337 "logbuf size for version 1 logs must be 16K or 32K");
1343 * V5 filesystems always use attr2 format for attributes.
1345 if (xfs_sb_version_hascrc(&mp
->m_sb
) &&
1346 (mp
->m_flags
& XFS_MOUNT_NOATTR2
)) {
1348 "Cannot mount a V5 filesystem as %s. %s is always enabled for V5 filesystems.",
1349 MNTOPT_NOATTR2
, MNTOPT_ATTR2
);
1354 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1355 * told by noattr2 to turn it off
1357 if (xfs_sb_version_hasattr2(&mp
->m_sb
) &&
1358 !(mp
->m_flags
& XFS_MOUNT_NOATTR2
))
1359 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
1362 * prohibit r/w mounts of read-only filesystems
1364 if ((mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) && !ronly
) {
1366 "cannot mount a read-only filesystem as read-write");
1370 if ((mp
->m_qflags
& (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
)) &&
1371 (mp
->m_qflags
& (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
)) &&
1372 !xfs_sb_version_has_pquotino(&mp
->m_sb
)) {
1374 "Super block does not support project and group quota together");
1382 xfs_init_percpu_counters(
1383 struct xfs_mount
*mp
)
1387 error
= percpu_counter_init(&mp
->m_icount
, 0, GFP_KERNEL
);
1391 error
= percpu_counter_init(&mp
->m_ifree
, 0, GFP_KERNEL
);
1395 error
= percpu_counter_init(&mp
->m_fdblocks
, 0, GFP_KERNEL
);
1402 percpu_counter_destroy(&mp
->m_ifree
);
1404 percpu_counter_destroy(&mp
->m_icount
);
1409 xfs_reinit_percpu_counters(
1410 struct xfs_mount
*mp
)
1412 percpu_counter_set(&mp
->m_icount
, mp
->m_sb
.sb_icount
);
1413 percpu_counter_set(&mp
->m_ifree
, mp
->m_sb
.sb_ifree
);
1414 percpu_counter_set(&mp
->m_fdblocks
, mp
->m_sb
.sb_fdblocks
);
1418 xfs_destroy_percpu_counters(
1419 struct xfs_mount
*mp
)
1421 percpu_counter_destroy(&mp
->m_icount
);
1422 percpu_counter_destroy(&mp
->m_ifree
);
1423 percpu_counter_destroy(&mp
->m_fdblocks
);
1428 struct super_block
*sb
,
1433 struct xfs_mount
*mp
= NULL
;
1434 int flags
= 0, error
= -ENOMEM
;
1436 mp
= kzalloc(sizeof(struct xfs_mount
), GFP_KERNEL
);
1440 spin_lock_init(&mp
->m_sb_lock
);
1441 mutex_init(&mp
->m_growlock
);
1442 atomic_set(&mp
->m_active_trans
, 0);
1443 INIT_DELAYED_WORK(&mp
->m_reclaim_work
, xfs_reclaim_worker
);
1444 INIT_DELAYED_WORK(&mp
->m_eofblocks_work
, xfs_eofblocks_worker
);
1445 mp
->m_kobj
.kobject
.kset
= xfs_kset
;
1450 error
= xfs_parseargs(mp
, (char *)data
);
1452 goto out_free_fsname
;
1454 sb_min_blocksize(sb
, BBSIZE
);
1455 sb
->s_xattr
= xfs_xattr_handlers
;
1456 sb
->s_export_op
= &xfs_export_operations
;
1457 #ifdef CONFIG_XFS_QUOTA
1458 sb
->s_qcop
= &xfs_quotactl_operations
;
1459 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
1461 sb
->s_op
= &xfs_super_operations
;
1464 flags
|= XFS_MFSI_QUIET
;
1466 error
= xfs_open_devices(mp
);
1468 goto out_free_fsname
;
1470 error
= xfs_init_mount_workqueues(mp
);
1472 goto out_close_devices
;
1474 error
= xfs_init_percpu_counters(mp
);
1476 goto out_destroy_workqueues
;
1478 error
= xfs_readsb(mp
, flags
);
1480 goto out_destroy_counters
;
1482 error
= xfs_finish_flags(mp
);
1486 error
= xfs_setup_devices(mp
);
1490 error
= xfs_filestream_mount(mp
);
1495 * we must configure the block size in the superblock before we run the
1496 * full mount process as the mount process can lookup and cache inodes.
1498 sb
->s_magic
= XFS_SB_MAGIC
;
1499 sb
->s_blocksize
= mp
->m_sb
.sb_blocksize
;
1500 sb
->s_blocksize_bits
= ffs(sb
->s_blocksize
) - 1;
1501 sb
->s_maxbytes
= xfs_max_file_offset(sb
->s_blocksize_bits
);
1502 sb
->s_max_links
= XFS_MAXLINK
;
1503 sb
->s_time_gran
= 1;
1504 set_posix_acl_flag(sb
);
1506 /* version 5 superblocks support inode version counters. */
1507 if (XFS_SB_VERSION_NUM(&mp
->m_sb
) == XFS_SB_VERSION_5
)
1508 sb
->s_flags
|= MS_I_VERSION
;
1510 if (mp
->m_flags
& XFS_MOUNT_DAX
) {
1512 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1513 if (sb
->s_blocksize
!= PAGE_SIZE
) {
1515 "Filesystem block size invalid for DAX Turning DAX off.");
1516 mp
->m_flags
&= ~XFS_MOUNT_DAX
;
1517 } else if (!sb
->s_bdev
->bd_disk
->fops
->direct_access
) {
1519 "Block device does not support DAX Turning DAX off.");
1520 mp
->m_flags
&= ~XFS_MOUNT_DAX
;
1524 if (xfs_sb_version_hassparseinodes(&mp
->m_sb
))
1526 "EXPERIMENTAL sparse inode feature enabled. Use at your own risk!");
1528 error
= xfs_mountfs(mp
);
1530 goto out_filestream_unmount
;
1532 root
= igrab(VFS_I(mp
->m_rootip
));
1537 sb
->s_root
= d_make_root(root
);
1545 out_filestream_unmount
:
1546 xfs_filestream_unmount(mp
);
1549 out_destroy_counters
:
1550 xfs_destroy_percpu_counters(mp
);
1551 out_destroy_workqueues
:
1552 xfs_destroy_mount_workqueues(mp
);
1554 xfs_close_devices(mp
);
1556 xfs_free_fsname(mp
);
1562 xfs_filestream_unmount(mp
);
1569 struct super_block
*sb
)
1571 struct xfs_mount
*mp
= XFS_M(sb
);
1573 xfs_notice(mp
, "Unmounting Filesystem");
1574 xfs_filestream_unmount(mp
);
1578 xfs_destroy_percpu_counters(mp
);
1579 xfs_destroy_mount_workqueues(mp
);
1580 xfs_close_devices(mp
);
1581 xfs_free_fsname(mp
);
1585 STATIC
struct dentry
*
1587 struct file_system_type
*fs_type
,
1589 const char *dev_name
,
1592 return mount_bdev(fs_type
, flags
, dev_name
, data
, xfs_fs_fill_super
);
1596 xfs_fs_nr_cached_objects(
1597 struct super_block
*sb
,
1598 struct shrink_control
*sc
)
1600 return xfs_reclaim_inodes_count(XFS_M(sb
));
1604 xfs_fs_free_cached_objects(
1605 struct super_block
*sb
,
1606 struct shrink_control
*sc
)
1608 return xfs_reclaim_inodes_nr(XFS_M(sb
), sc
->nr_to_scan
);
1611 static const struct super_operations xfs_super_operations
= {
1612 .alloc_inode
= xfs_fs_alloc_inode
,
1613 .destroy_inode
= xfs_fs_destroy_inode
,
1614 .evict_inode
= xfs_fs_evict_inode
,
1615 .drop_inode
= xfs_fs_drop_inode
,
1616 .put_super
= xfs_fs_put_super
,
1617 .sync_fs
= xfs_fs_sync_fs
,
1618 .freeze_fs
= xfs_fs_freeze
,
1619 .unfreeze_fs
= xfs_fs_unfreeze
,
1620 .statfs
= xfs_fs_statfs
,
1621 .remount_fs
= xfs_fs_remount
,
1622 .show_options
= xfs_fs_show_options
,
1623 .nr_cached_objects
= xfs_fs_nr_cached_objects
,
1624 .free_cached_objects
= xfs_fs_free_cached_objects
,
1627 static struct file_system_type xfs_fs_type
= {
1628 .owner
= THIS_MODULE
,
1630 .mount
= xfs_fs_mount
,
1631 .kill_sb
= kill_block_super
,
1632 .fs_flags
= FS_REQUIRES_DEV
,
1634 MODULE_ALIAS_FS("xfs");
1637 xfs_init_zones(void)
1640 xfs_ioend_zone
= kmem_zone_init(sizeof(xfs_ioend_t
), "xfs_ioend");
1641 if (!xfs_ioend_zone
)
1644 xfs_ioend_pool
= mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE
,
1646 if (!xfs_ioend_pool
)
1647 goto out_destroy_ioend_zone
;
1649 xfs_log_ticket_zone
= kmem_zone_init(sizeof(xlog_ticket_t
),
1651 if (!xfs_log_ticket_zone
)
1652 goto out_destroy_ioend_pool
;
1654 xfs_bmap_free_item_zone
= kmem_zone_init(sizeof(xfs_bmap_free_item_t
),
1655 "xfs_bmap_free_item");
1656 if (!xfs_bmap_free_item_zone
)
1657 goto out_destroy_log_ticket_zone
;
1659 xfs_btree_cur_zone
= kmem_zone_init(sizeof(xfs_btree_cur_t
),
1661 if (!xfs_btree_cur_zone
)
1662 goto out_destroy_bmap_free_item_zone
;
1664 xfs_da_state_zone
= kmem_zone_init(sizeof(xfs_da_state_t
),
1666 if (!xfs_da_state_zone
)
1667 goto out_destroy_btree_cur_zone
;
1669 xfs_ifork_zone
= kmem_zone_init(sizeof(xfs_ifork_t
), "xfs_ifork");
1670 if (!xfs_ifork_zone
)
1671 goto out_destroy_da_state_zone
;
1673 xfs_trans_zone
= kmem_zone_init(sizeof(xfs_trans_t
), "xfs_trans");
1674 if (!xfs_trans_zone
)
1675 goto out_destroy_ifork_zone
;
1677 xfs_log_item_desc_zone
=
1678 kmem_zone_init(sizeof(struct xfs_log_item_desc
),
1679 "xfs_log_item_desc");
1680 if (!xfs_log_item_desc_zone
)
1681 goto out_destroy_trans_zone
;
1684 * The size of the zone allocated buf log item is the maximum
1685 * size possible under XFS. This wastes a little bit of memory,
1686 * but it is much faster.
1688 xfs_buf_item_zone
= kmem_zone_init(sizeof(struct xfs_buf_log_item
),
1690 if (!xfs_buf_item_zone
)
1691 goto out_destroy_log_item_desc_zone
;
1693 xfs_efd_zone
= kmem_zone_init((sizeof(xfs_efd_log_item_t
) +
1694 ((XFS_EFD_MAX_FAST_EXTENTS
- 1) *
1695 sizeof(xfs_extent_t
))), "xfs_efd_item");
1697 goto out_destroy_buf_item_zone
;
1699 xfs_efi_zone
= kmem_zone_init((sizeof(xfs_efi_log_item_t
) +
1700 ((XFS_EFI_MAX_FAST_EXTENTS
- 1) *
1701 sizeof(xfs_extent_t
))), "xfs_efi_item");
1703 goto out_destroy_efd_zone
;
1706 kmem_zone_init_flags(sizeof(xfs_inode_t
), "xfs_inode",
1707 KM_ZONE_HWALIGN
| KM_ZONE_RECLAIM
| KM_ZONE_SPREAD
,
1708 xfs_fs_inode_init_once
);
1709 if (!xfs_inode_zone
)
1710 goto out_destroy_efi_zone
;
1713 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t
), "xfs_ili",
1714 KM_ZONE_SPREAD
, NULL
);
1716 goto out_destroy_inode_zone
;
1717 xfs_icreate_zone
= kmem_zone_init(sizeof(struct xfs_icreate_item
),
1719 if (!xfs_icreate_zone
)
1720 goto out_destroy_ili_zone
;
1724 out_destroy_ili_zone
:
1725 kmem_zone_destroy(xfs_ili_zone
);
1726 out_destroy_inode_zone
:
1727 kmem_zone_destroy(xfs_inode_zone
);
1728 out_destroy_efi_zone
:
1729 kmem_zone_destroy(xfs_efi_zone
);
1730 out_destroy_efd_zone
:
1731 kmem_zone_destroy(xfs_efd_zone
);
1732 out_destroy_buf_item_zone
:
1733 kmem_zone_destroy(xfs_buf_item_zone
);
1734 out_destroy_log_item_desc_zone
:
1735 kmem_zone_destroy(xfs_log_item_desc_zone
);
1736 out_destroy_trans_zone
:
1737 kmem_zone_destroy(xfs_trans_zone
);
1738 out_destroy_ifork_zone
:
1739 kmem_zone_destroy(xfs_ifork_zone
);
1740 out_destroy_da_state_zone
:
1741 kmem_zone_destroy(xfs_da_state_zone
);
1742 out_destroy_btree_cur_zone
:
1743 kmem_zone_destroy(xfs_btree_cur_zone
);
1744 out_destroy_bmap_free_item_zone
:
1745 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1746 out_destroy_log_ticket_zone
:
1747 kmem_zone_destroy(xfs_log_ticket_zone
);
1748 out_destroy_ioend_pool
:
1749 mempool_destroy(xfs_ioend_pool
);
1750 out_destroy_ioend_zone
:
1751 kmem_zone_destroy(xfs_ioend_zone
);
1757 xfs_destroy_zones(void)
1760 * Make sure all delayed rcu free are flushed before we
1764 kmem_zone_destroy(xfs_icreate_zone
);
1765 kmem_zone_destroy(xfs_ili_zone
);
1766 kmem_zone_destroy(xfs_inode_zone
);
1767 kmem_zone_destroy(xfs_efi_zone
);
1768 kmem_zone_destroy(xfs_efd_zone
);
1769 kmem_zone_destroy(xfs_buf_item_zone
);
1770 kmem_zone_destroy(xfs_log_item_desc_zone
);
1771 kmem_zone_destroy(xfs_trans_zone
);
1772 kmem_zone_destroy(xfs_ifork_zone
);
1773 kmem_zone_destroy(xfs_da_state_zone
);
1774 kmem_zone_destroy(xfs_btree_cur_zone
);
1775 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1776 kmem_zone_destroy(xfs_log_ticket_zone
);
1777 mempool_destroy(xfs_ioend_pool
);
1778 kmem_zone_destroy(xfs_ioend_zone
);
1783 xfs_init_workqueues(void)
1786 * The allocation workqueue can be used in memory reclaim situations
1787 * (writepage path), and parallelism is only limited by the number of
1788 * AGs in all the filesystems mounted. Hence use the default large
1789 * max_active value for this workqueue.
1791 xfs_alloc_wq
= alloc_workqueue("xfsalloc",
1792 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0);
1800 xfs_destroy_workqueues(void)
1802 destroy_workqueue(xfs_alloc_wq
);
1810 printk(KERN_INFO XFS_VERSION_STRING
" with "
1811 XFS_BUILD_OPTIONS
" enabled\n");
1815 error
= xfs_init_zones();
1819 error
= xfs_init_workqueues();
1821 goto out_destroy_zones
;
1823 error
= xfs_mru_cache_init();
1825 goto out_destroy_wq
;
1827 error
= xfs_buf_init();
1829 goto out_mru_cache_uninit
;
1831 error
= xfs_init_procfs();
1833 goto out_buf_terminate
;
1835 error
= xfs_sysctl_register();
1837 goto out_cleanup_procfs
;
1839 xfs_kset
= kset_create_and_add("xfs", NULL
, fs_kobj
);
1842 goto out_sysctl_unregister
;
1845 xfs_stats_kobj
.kobject
.kset
= xfs_kset
;
1846 error
= xfs_sysfs_init(&xfs_stats_kobj
, &xfs_stats_ktype
, NULL
,
1849 goto out_kset_unregister
;
1852 xfs_dbg_kobj
.kobject
.kset
= xfs_kset
;
1853 error
= xfs_sysfs_init(&xfs_dbg_kobj
, &xfs_dbg_ktype
, NULL
, "debug");
1855 goto out_remove_stats_kobj
;
1858 error
= xfs_qm_init();
1860 goto out_remove_dbg_kobj
;
1862 error
= register_filesystem(&xfs_fs_type
);
1869 out_remove_dbg_kobj
:
1871 xfs_sysfs_del(&xfs_dbg_kobj
);
1872 out_remove_stats_kobj
:
1874 xfs_sysfs_del(&xfs_stats_kobj
);
1875 out_kset_unregister
:
1876 kset_unregister(xfs_kset
);
1877 out_sysctl_unregister
:
1878 xfs_sysctl_unregister();
1880 xfs_cleanup_procfs();
1882 xfs_buf_terminate();
1883 out_mru_cache_uninit
:
1884 xfs_mru_cache_uninit();
1886 xfs_destroy_workqueues();
1888 xfs_destroy_zones();
1897 unregister_filesystem(&xfs_fs_type
);
1899 xfs_sysfs_del(&xfs_dbg_kobj
);
1901 xfs_sysfs_del(&xfs_stats_kobj
);
1902 kset_unregister(xfs_kset
);
1903 xfs_sysctl_unregister();
1904 xfs_cleanup_procfs();
1905 xfs_buf_terminate();
1906 xfs_mru_cache_uninit();
1907 xfs_destroy_workqueues();
1908 xfs_destroy_zones();
1911 module_init(init_xfs_fs
);
1912 module_exit(exit_xfs_fs
);
1914 MODULE_AUTHOR("Silicon Graphics, Inc.");
1915 MODULE_DESCRIPTION(XFS_VERSION_STRING
" with " XFS_BUILD_OPTIONS
" enabled");
1916 MODULE_LICENSE("GPL");