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"
48 #include "xfs_ondisk.h"
50 #include <linux/namei.h>
51 #include <linux/init.h>
52 #include <linux/slab.h>
53 #include <linux/mount.h>
54 #include <linux/mempool.h>
55 #include <linux/writeback.h>
56 #include <linux/kthread.h>
57 #include <linux/freezer.h>
58 #include <linux/parser.h>
60 static const struct super_operations xfs_super_operations
;
61 struct bio_set
*xfs_ioend_bioset
;
63 static struct kset
*xfs_kset
; /* top-level xfs sysfs dir */
65 static struct xfs_kobj xfs_dbg_kobj
; /* global debug sysfs attrs */
69 * Table driven mount option parser.
72 Opt_logbufs
, Opt_logbsize
, Opt_logdev
, Opt_rtdev
, Opt_biosize
,
73 Opt_wsync
, Opt_noalign
, Opt_swalloc
, Opt_sunit
, Opt_swidth
, Opt_nouuid
,
74 Opt_mtpt
, Opt_grpid
, Opt_nogrpid
, Opt_bsdgroups
, Opt_sysvgroups
,
75 Opt_allocsize
, Opt_norecovery
, Opt_barrier
, Opt_nobarrier
,
76 Opt_inode64
, Opt_inode32
, Opt_ikeep
, Opt_noikeep
,
77 Opt_largeio
, Opt_nolargeio
, Opt_attr2
, Opt_noattr2
, Opt_filestreams
,
78 Opt_quota
, Opt_noquota
, Opt_usrquota
, Opt_grpquota
, Opt_prjquota
,
79 Opt_uquota
, Opt_gquota
, Opt_pquota
,
80 Opt_uqnoenforce
, Opt_gqnoenforce
, Opt_pqnoenforce
, Opt_qnoenforce
,
81 Opt_discard
, Opt_nodiscard
, Opt_dax
, Opt_err
,
84 static const match_table_t tokens
= {
85 {Opt_logbufs
, "logbufs=%u"}, /* number of XFS log buffers */
86 {Opt_logbsize
, "logbsize=%s"}, /* size of XFS log buffers */
87 {Opt_logdev
, "logdev=%s"}, /* log device */
88 {Opt_rtdev
, "rtdev=%s"}, /* realtime I/O device */
89 {Opt_biosize
, "biosize=%u"}, /* log2 of preferred buffered io size */
90 {Opt_wsync
, "wsync"}, /* safe-mode nfs compatible mount */
91 {Opt_noalign
, "noalign"}, /* turn off stripe alignment */
92 {Opt_swalloc
, "swalloc"}, /* turn on stripe width allocation */
93 {Opt_sunit
, "sunit=%u"}, /* data volume stripe unit */
94 {Opt_swidth
, "swidth=%u"}, /* data volume stripe width */
95 {Opt_nouuid
, "nouuid"}, /* ignore filesystem UUID */
96 {Opt_mtpt
, "mtpt"}, /* filesystem mount point */
97 {Opt_grpid
, "grpid"}, /* group-ID from parent directory */
98 {Opt_nogrpid
, "nogrpid"}, /* group-ID from current process */
99 {Opt_bsdgroups
, "bsdgroups"}, /* group-ID from parent directory */
100 {Opt_sysvgroups
,"sysvgroups"}, /* group-ID from current process */
101 {Opt_allocsize
, "allocsize=%s"},/* preferred allocation size */
102 {Opt_norecovery
,"norecovery"}, /* don't run XFS recovery */
103 {Opt_barrier
, "barrier"}, /* use writer barriers for log write and
104 * unwritten extent conversion */
105 {Opt_nobarrier
, "nobarrier"}, /* .. disable */
106 {Opt_inode64
, "inode64"}, /* inodes can be allocated anywhere */
107 {Opt_inode32
, "inode32"}, /* inode allocation limited to
108 * XFS_MAXINUMBER_32 */
109 {Opt_ikeep
, "ikeep"}, /* do not free empty inode clusters */
110 {Opt_noikeep
, "noikeep"}, /* free empty inode clusters */
111 {Opt_largeio
, "largeio"}, /* report large I/O sizes in stat() */
112 {Opt_nolargeio
, "nolargeio"}, /* do not report large I/O sizes
114 {Opt_attr2
, "attr2"}, /* do use attr2 attribute format */
115 {Opt_noattr2
, "noattr2"}, /* do not use attr2 attribute format */
116 {Opt_filestreams
,"filestreams"},/* use filestreams allocator */
117 {Opt_quota
, "quota"}, /* disk quotas (user) */
118 {Opt_noquota
, "noquota"}, /* no quotas */
119 {Opt_usrquota
, "usrquota"}, /* user quota enabled */
120 {Opt_grpquota
, "grpquota"}, /* group quota enabled */
121 {Opt_prjquota
, "prjquota"}, /* project quota enabled */
122 {Opt_uquota
, "uquota"}, /* user quota (IRIX variant) */
123 {Opt_gquota
, "gquota"}, /* group quota (IRIX variant) */
124 {Opt_pquota
, "pquota"}, /* project quota (IRIX variant) */
125 {Opt_uqnoenforce
,"uqnoenforce"},/* user quota limit enforcement */
126 {Opt_gqnoenforce
,"gqnoenforce"},/* group quota limit enforcement */
127 {Opt_pqnoenforce
,"pqnoenforce"},/* project quota limit enforcement */
128 {Opt_qnoenforce
, "qnoenforce"}, /* same as uqnoenforce */
129 {Opt_discard
, "discard"}, /* Discard unused blocks */
130 {Opt_nodiscard
, "nodiscard"}, /* Do not discard unused blocks */
132 {Opt_dax
, "dax"}, /* Enable direct access to bdev pages */
138 suffix_kstrtoint(const substring_t
*s
, unsigned int base
, int *res
)
140 int last
, shift_left_factor
= 0, _res
;
144 value
= match_strdup(s
);
148 last
= strlen(value
) - 1;
149 if (value
[last
] == 'K' || value
[last
] == 'k') {
150 shift_left_factor
= 10;
153 if (value
[last
] == 'M' || value
[last
] == 'm') {
154 shift_left_factor
= 20;
157 if (value
[last
] == 'G' || value
[last
] == 'g') {
158 shift_left_factor
= 30;
162 if (kstrtoint(value
, base
, &_res
))
165 *res
= _res
<< shift_left_factor
;
170 * This function fills in xfs_mount_t fields based on mount args.
171 * Note: the superblock has _not_ yet been read in.
173 * Note that this function leaks the various device name allocations on
174 * failure. The caller takes care of them.
176 * *sb is const because this is also used to test options on the remount
177 * path, and we don't want this to have any side effects at remount time.
178 * Today this function does not change *sb, but just to future-proof...
182 struct xfs_mount
*mp
,
185 const struct super_block
*sb
= mp
->m_super
;
187 substring_t args
[MAX_OPT_ARGS
];
191 __uint8_t iosizelog
= 0;
194 * set up the mount name first so all the errors will refer to the
197 mp
->m_fsname
= kstrndup(sb
->s_id
, MAXNAMELEN
, GFP_KERNEL
);
200 mp
->m_fsname_len
= strlen(mp
->m_fsname
) + 1;
203 * Copy binary VFS mount flags we are interested in.
205 if (sb
->s_flags
& MS_RDONLY
)
206 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
207 if (sb
->s_flags
& MS_DIRSYNC
)
208 mp
->m_flags
|= XFS_MOUNT_DIRSYNC
;
209 if (sb
->s_flags
& MS_SYNCHRONOUS
)
210 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
213 * Set some default flags that could be cleared by the mount option
216 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
217 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
220 * These can be overridden by the mount option parsing.
228 while ((p
= strsep(&options
, ",")) != NULL
) {
234 token
= match_token(p
, tokens
, args
);
237 if (match_int(args
, &mp
->m_logbufs
))
241 if (suffix_kstrtoint(args
, 10, &mp
->m_logbsize
))
245 mp
->m_logname
= match_strdup(args
);
250 xfs_warn(mp
, "%s option not allowed on this system", p
);
253 mp
->m_rtname
= match_strdup(args
);
259 if (suffix_kstrtoint(args
, 10, &iosize
))
261 iosizelog
= ffs(iosize
) - 1;
265 mp
->m_flags
|= XFS_MOUNT_GRPID
;
269 mp
->m_flags
&= ~XFS_MOUNT_GRPID
;
272 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
275 mp
->m_flags
|= XFS_MOUNT_NORECOVERY
;
278 mp
->m_flags
|= XFS_MOUNT_NOALIGN
;
281 mp
->m_flags
|= XFS_MOUNT_SWALLOC
;
284 if (match_int(args
, &dsunit
))
288 if (match_int(args
, &dswidth
))
292 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
295 mp
->m_flags
&= ~XFS_MOUNT_SMALL_INUMS
;
298 mp
->m_flags
|= XFS_MOUNT_NOUUID
;
301 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
304 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
307 mp
->m_flags
|= XFS_MOUNT_IKEEP
;
310 mp
->m_flags
&= ~XFS_MOUNT_IKEEP
;
313 mp
->m_flags
&= ~XFS_MOUNT_COMPAT_IOSIZE
;
316 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
319 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
322 mp
->m_flags
&= ~XFS_MOUNT_ATTR2
;
323 mp
->m_flags
|= XFS_MOUNT_NOATTR2
;
325 case Opt_filestreams
:
326 mp
->m_flags
|= XFS_MOUNT_FILESTREAMS
;
329 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACCT
;
330 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ENFD
;
331 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACTIVE
;
336 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
|
340 case Opt_uqnoenforce
:
341 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
);
342 mp
->m_qflags
&= ~XFS_UQUOTA_ENFD
;
346 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
|
349 case Opt_pqnoenforce
:
350 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
);
351 mp
->m_qflags
&= ~XFS_PQUOTA_ENFD
;
355 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
|
358 case Opt_gqnoenforce
:
359 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
);
360 mp
->m_qflags
&= ~XFS_GQUOTA_ENFD
;
363 mp
->m_flags
|= XFS_MOUNT_DISCARD
;
366 mp
->m_flags
&= ~XFS_MOUNT_DISCARD
;
370 mp
->m_flags
|= XFS_MOUNT_DAX
;
374 xfs_warn(mp
, "unknown mount option [%s].", p
);
380 * no recovery flag requires a read-only mount
382 if ((mp
->m_flags
& XFS_MOUNT_NORECOVERY
) &&
383 !(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
384 xfs_warn(mp
, "no-recovery mounts must be read-only.");
388 if ((mp
->m_flags
& XFS_MOUNT_NOALIGN
) && (dsunit
|| dswidth
)) {
390 "sunit and swidth options incompatible with the noalign option");
394 #ifndef CONFIG_XFS_QUOTA
395 if (XFS_IS_QUOTA_RUNNING(mp
)) {
396 xfs_warn(mp
, "quota support not available in this kernel.");
401 if ((dsunit
&& !dswidth
) || (!dsunit
&& dswidth
)) {
402 xfs_warn(mp
, "sunit and swidth must be specified together");
406 if (dsunit
&& (dswidth
% dsunit
!= 0)) {
408 "stripe width (%d) must be a multiple of the stripe unit (%d)",
414 if (dsunit
&& !(mp
->m_flags
& XFS_MOUNT_NOALIGN
)) {
416 * At this point the superblock has not been read
417 * in, therefore we do not know the block size.
418 * Before the mount call ends we will convert
421 mp
->m_dalign
= dsunit
;
422 mp
->m_swidth
= dswidth
;
425 if (mp
->m_logbufs
!= -1 &&
426 mp
->m_logbufs
!= 0 &&
427 (mp
->m_logbufs
< XLOG_MIN_ICLOGS
||
428 mp
->m_logbufs
> XLOG_MAX_ICLOGS
)) {
429 xfs_warn(mp
, "invalid logbufs value: %d [not %d-%d]",
430 mp
->m_logbufs
, XLOG_MIN_ICLOGS
, XLOG_MAX_ICLOGS
);
433 if (mp
->m_logbsize
!= -1 &&
434 mp
->m_logbsize
!= 0 &&
435 (mp
->m_logbsize
< XLOG_MIN_RECORD_BSIZE
||
436 mp
->m_logbsize
> XLOG_MAX_RECORD_BSIZE
||
437 !is_power_of_2(mp
->m_logbsize
))) {
439 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
445 if (iosizelog
> XFS_MAX_IO_LOG
||
446 iosizelog
< XFS_MIN_IO_LOG
) {
447 xfs_warn(mp
, "invalid log iosize: %d [not %d-%d]",
448 iosizelog
, XFS_MIN_IO_LOG
,
453 mp
->m_flags
|= XFS_MOUNT_DFLT_IOSIZE
;
454 mp
->m_readio_log
= iosizelog
;
455 mp
->m_writeio_log
= iosizelog
;
461 struct proc_xfs_info
{
468 struct xfs_mount
*mp
,
471 static struct proc_xfs_info xfs_info_set
[] = {
472 /* the few simple ones we can get from the mount struct */
473 { XFS_MOUNT_IKEEP
, ",ikeep" },
474 { XFS_MOUNT_WSYNC
, ",wsync" },
475 { XFS_MOUNT_NOALIGN
, ",noalign" },
476 { XFS_MOUNT_SWALLOC
, ",swalloc" },
477 { XFS_MOUNT_NOUUID
, ",nouuid" },
478 { XFS_MOUNT_NORECOVERY
, ",norecovery" },
479 { XFS_MOUNT_ATTR2
, ",attr2" },
480 { XFS_MOUNT_FILESTREAMS
, ",filestreams" },
481 { XFS_MOUNT_GRPID
, ",grpid" },
482 { XFS_MOUNT_DISCARD
, ",discard" },
483 { XFS_MOUNT_SMALL_INUMS
, ",inode32" },
484 { XFS_MOUNT_DAX
, ",dax" },
487 static struct proc_xfs_info xfs_info_unset
[] = {
488 /* the few simple ones we can get from the mount struct */
489 { XFS_MOUNT_COMPAT_IOSIZE
, ",largeio" },
490 { XFS_MOUNT_BARRIER
, ",nobarrier" },
491 { XFS_MOUNT_SMALL_INUMS
, ",inode64" },
494 struct proc_xfs_info
*xfs_infop
;
496 for (xfs_infop
= xfs_info_set
; xfs_infop
->flag
; xfs_infop
++) {
497 if (mp
->m_flags
& xfs_infop
->flag
)
498 seq_puts(m
, xfs_infop
->str
);
500 for (xfs_infop
= xfs_info_unset
; xfs_infop
->flag
; xfs_infop
++) {
501 if (!(mp
->m_flags
& xfs_infop
->flag
))
502 seq_puts(m
, xfs_infop
->str
);
505 if (mp
->m_flags
& XFS_MOUNT_DFLT_IOSIZE
)
506 seq_printf(m
, ",allocsize=%dk",
507 (int)(1 << mp
->m_writeio_log
) >> 10);
509 if (mp
->m_logbufs
> 0)
510 seq_printf(m
, ",logbufs=%d", mp
->m_logbufs
);
511 if (mp
->m_logbsize
> 0)
512 seq_printf(m
, ",logbsize=%dk", mp
->m_logbsize
>> 10);
515 seq_show_option(m
, "logdev", mp
->m_logname
);
517 seq_show_option(m
, "rtdev", mp
->m_rtname
);
519 if (mp
->m_dalign
> 0)
520 seq_printf(m
, ",sunit=%d",
521 (int)XFS_FSB_TO_BB(mp
, mp
->m_dalign
));
522 if (mp
->m_swidth
> 0)
523 seq_printf(m
, ",swidth=%d",
524 (int)XFS_FSB_TO_BB(mp
, mp
->m_swidth
));
526 if (mp
->m_qflags
& (XFS_UQUOTA_ACCT
|XFS_UQUOTA_ENFD
))
527 seq_puts(m
, ",usrquota");
528 else if (mp
->m_qflags
& XFS_UQUOTA_ACCT
)
529 seq_puts(m
, ",uqnoenforce");
531 if (mp
->m_qflags
& XFS_PQUOTA_ACCT
) {
532 if (mp
->m_qflags
& XFS_PQUOTA_ENFD
)
533 seq_puts(m
, ",prjquota");
535 seq_puts(m
, ",pqnoenforce");
537 if (mp
->m_qflags
& XFS_GQUOTA_ACCT
) {
538 if (mp
->m_qflags
& XFS_GQUOTA_ENFD
)
539 seq_puts(m
, ",grpquota");
541 seq_puts(m
, ",gqnoenforce");
544 if (!(mp
->m_qflags
& XFS_ALL_QUOTA_ACCT
))
545 seq_puts(m
, ",noquota");
551 unsigned int blockshift
)
553 unsigned int pagefactor
= 1;
554 unsigned int bitshift
= BITS_PER_LONG
- 1;
556 /* Figure out maximum filesize, on Linux this can depend on
557 * the filesystem blocksize (on 32 bit platforms).
558 * __block_write_begin does this in an [unsigned] long...
559 * page->index << (PAGE_SHIFT - bbits)
560 * So, for page sized blocks (4K on 32 bit platforms),
561 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
562 * (((u64)PAGE_SIZE << (BITS_PER_LONG-1))-1)
563 * but for smaller blocksizes it is less (bbits = log2 bsize).
564 * Note1: get_block_t takes a long (implicit cast from above)
565 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
566 * can optionally convert the [unsigned] long from above into
567 * an [unsigned] long long.
570 #if BITS_PER_LONG == 32
571 # if defined(CONFIG_LBDAF)
572 ASSERT(sizeof(sector_t
) == 8);
573 pagefactor
= PAGE_SIZE
;
574 bitshift
= BITS_PER_LONG
;
576 pagefactor
= PAGE_SIZE
>> (PAGE_SHIFT
- blockshift
);
580 return (((__uint64_t
)pagefactor
) << bitshift
) - 1;
584 * Set parameters for inode allocation heuristics, taking into account
585 * filesystem size and inode32/inode64 mount options; i.e. specifically
586 * whether or not XFS_MOUNT_SMALL_INUMS is set.
588 * Inode allocation patterns are altered only if inode32 is requested
589 * (XFS_MOUNT_SMALL_INUMS), and the filesystem is sufficiently large.
590 * If altered, XFS_MOUNT_32BITINODES is set as well.
592 * An agcount independent of that in the mount structure is provided
593 * because in the growfs case, mp->m_sb.sb_agcount is not yet updated
594 * to the potentially higher ag count.
596 * Returns the maximum AG index which may contain inodes.
600 struct xfs_mount
*mp
,
601 xfs_agnumber_t agcount
)
603 xfs_agnumber_t index
;
604 xfs_agnumber_t maxagi
= 0;
605 xfs_sb_t
*sbp
= &mp
->m_sb
;
606 xfs_agnumber_t max_metadata
;
611 * Calculate how much should be reserved for inodes to meet
612 * the max inode percentage. Used only for inode32.
614 if (mp
->m_maxicount
) {
617 icount
= sbp
->sb_dblocks
* sbp
->sb_imax_pct
;
619 icount
+= sbp
->sb_agblocks
- 1;
620 do_div(icount
, sbp
->sb_agblocks
);
621 max_metadata
= icount
;
623 max_metadata
= agcount
;
626 /* Get the last possible inode in the filesystem */
627 agino
= XFS_OFFBNO_TO_AGINO(mp
, sbp
->sb_agblocks
- 1, 0);
628 ino
= XFS_AGINO_TO_INO(mp
, agcount
- 1, agino
);
631 * If user asked for no more than 32-bit inodes, and the fs is
632 * sufficiently large, set XFS_MOUNT_32BITINODES if we must alter
633 * the allocator to accommodate the request.
635 if ((mp
->m_flags
& XFS_MOUNT_SMALL_INUMS
) && ino
> XFS_MAXINUMBER_32
)
636 mp
->m_flags
|= XFS_MOUNT_32BITINODES
;
638 mp
->m_flags
&= ~XFS_MOUNT_32BITINODES
;
640 for (index
= 0; index
< agcount
; index
++) {
641 struct xfs_perag
*pag
;
643 ino
= XFS_AGINO_TO_INO(mp
, index
, agino
);
645 pag
= xfs_perag_get(mp
, index
);
647 if (mp
->m_flags
& XFS_MOUNT_32BITINODES
) {
648 if (ino
> XFS_MAXINUMBER_32
) {
649 pag
->pagi_inodeok
= 0;
650 pag
->pagf_metadata
= 0;
652 pag
->pagi_inodeok
= 1;
654 if (index
< max_metadata
)
655 pag
->pagf_metadata
= 1;
657 pag
->pagf_metadata
= 0;
660 pag
->pagi_inodeok
= 1;
661 pag
->pagf_metadata
= 0;
667 return (mp
->m_flags
& XFS_MOUNT_32BITINODES
) ? maxagi
: agcount
;
674 struct block_device
**bdevp
)
678 *bdevp
= blkdev_get_by_path(name
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
,
680 if (IS_ERR(*bdevp
)) {
681 error
= PTR_ERR(*bdevp
);
682 xfs_warn(mp
, "Invalid device [%s], error=%d", name
, error
);
690 struct block_device
*bdev
)
693 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
697 xfs_blkdev_issue_flush(
698 xfs_buftarg_t
*buftarg
)
700 blkdev_issue_flush(buftarg
->bt_bdev
, GFP_NOFS
, NULL
);
705 struct xfs_mount
*mp
)
707 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
708 struct block_device
*logdev
= mp
->m_logdev_targp
->bt_bdev
;
709 xfs_free_buftarg(mp
, mp
->m_logdev_targp
);
710 xfs_blkdev_put(logdev
);
712 if (mp
->m_rtdev_targp
) {
713 struct block_device
*rtdev
= mp
->m_rtdev_targp
->bt_bdev
;
714 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
715 xfs_blkdev_put(rtdev
);
717 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
721 * The file system configurations are:
722 * (1) device (partition) with data and internal log
723 * (2) logical volume with data and log subvolumes.
724 * (3) logical volume with data, log, and realtime subvolumes.
726 * We only have to handle opening the log and realtime volumes here if
727 * they are present. The data subvolume has already been opened by
728 * get_sb_bdev() and is stored in sb->s_bdev.
732 struct xfs_mount
*mp
)
734 struct block_device
*ddev
= mp
->m_super
->s_bdev
;
735 struct block_device
*logdev
= NULL
, *rtdev
= NULL
;
739 * Open real time and log devices - order is important.
742 error
= xfs_blkdev_get(mp
, mp
->m_logname
, &logdev
);
748 error
= xfs_blkdev_get(mp
, mp
->m_rtname
, &rtdev
);
750 goto out_close_logdev
;
752 if (rtdev
== ddev
|| rtdev
== logdev
) {
754 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
756 goto out_close_rtdev
;
761 * Setup xfs_mount buffer target pointers
764 mp
->m_ddev_targp
= xfs_alloc_buftarg(mp
, ddev
);
765 if (!mp
->m_ddev_targp
)
766 goto out_close_rtdev
;
769 mp
->m_rtdev_targp
= xfs_alloc_buftarg(mp
, rtdev
);
770 if (!mp
->m_rtdev_targp
)
771 goto out_free_ddev_targ
;
774 if (logdev
&& logdev
!= ddev
) {
775 mp
->m_logdev_targp
= xfs_alloc_buftarg(mp
, logdev
);
776 if (!mp
->m_logdev_targp
)
777 goto out_free_rtdev_targ
;
779 mp
->m_logdev_targp
= mp
->m_ddev_targp
;
785 if (mp
->m_rtdev_targp
)
786 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
788 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
790 xfs_blkdev_put(rtdev
);
792 if (logdev
&& logdev
!= ddev
)
793 xfs_blkdev_put(logdev
);
799 * Setup xfs_mount buffer target pointers based on superblock
803 struct xfs_mount
*mp
)
807 error
= xfs_setsize_buftarg(mp
->m_ddev_targp
, mp
->m_sb
.sb_sectsize
);
811 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
812 unsigned int log_sector_size
= BBSIZE
;
814 if (xfs_sb_version_hassector(&mp
->m_sb
))
815 log_sector_size
= mp
->m_sb
.sb_logsectsize
;
816 error
= xfs_setsize_buftarg(mp
->m_logdev_targp
,
821 if (mp
->m_rtdev_targp
) {
822 error
= xfs_setsize_buftarg(mp
->m_rtdev_targp
,
823 mp
->m_sb
.sb_sectsize
);
832 xfs_init_mount_workqueues(
833 struct xfs_mount
*mp
)
835 mp
->m_buf_workqueue
= alloc_workqueue("xfs-buf/%s",
836 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 1, mp
->m_fsname
);
837 if (!mp
->m_buf_workqueue
)
840 mp
->m_data_workqueue
= alloc_workqueue("xfs-data/%s",
841 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
842 if (!mp
->m_data_workqueue
)
843 goto out_destroy_buf
;
845 mp
->m_unwritten_workqueue
= alloc_workqueue("xfs-conv/%s",
846 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
847 if (!mp
->m_unwritten_workqueue
)
848 goto out_destroy_data_iodone_queue
;
850 mp
->m_cil_workqueue
= alloc_workqueue("xfs-cil/%s",
851 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
852 if (!mp
->m_cil_workqueue
)
853 goto out_destroy_unwritten
;
855 mp
->m_reclaim_workqueue
= alloc_workqueue("xfs-reclaim/%s",
856 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
857 if (!mp
->m_reclaim_workqueue
)
858 goto out_destroy_cil
;
860 mp
->m_log_workqueue
= alloc_workqueue("xfs-log/%s",
861 WQ_MEM_RECLAIM
|WQ_FREEZABLE
|WQ_HIGHPRI
, 0,
863 if (!mp
->m_log_workqueue
)
864 goto out_destroy_reclaim
;
866 mp
->m_eofblocks_workqueue
= alloc_workqueue("xfs-eofblocks/%s",
867 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
868 if (!mp
->m_eofblocks_workqueue
)
869 goto out_destroy_log
;
874 destroy_workqueue(mp
->m_log_workqueue
);
876 destroy_workqueue(mp
->m_reclaim_workqueue
);
878 destroy_workqueue(mp
->m_cil_workqueue
);
879 out_destroy_unwritten
:
880 destroy_workqueue(mp
->m_unwritten_workqueue
);
881 out_destroy_data_iodone_queue
:
882 destroy_workqueue(mp
->m_data_workqueue
);
884 destroy_workqueue(mp
->m_buf_workqueue
);
890 xfs_destroy_mount_workqueues(
891 struct xfs_mount
*mp
)
893 destroy_workqueue(mp
->m_eofblocks_workqueue
);
894 destroy_workqueue(mp
->m_log_workqueue
);
895 destroy_workqueue(mp
->m_reclaim_workqueue
);
896 destroy_workqueue(mp
->m_cil_workqueue
);
897 destroy_workqueue(mp
->m_data_workqueue
);
898 destroy_workqueue(mp
->m_unwritten_workqueue
);
899 destroy_workqueue(mp
->m_buf_workqueue
);
903 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
904 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
905 * for IO to complete so that we effectively throttle multiple callers to the
906 * rate at which IO is completing.
910 struct xfs_mount
*mp
)
912 struct super_block
*sb
= mp
->m_super
;
914 if (down_read_trylock(&sb
->s_umount
)) {
916 up_read(&sb
->s_umount
);
920 /* Catch misguided souls that try to use this interface on XFS */
921 STATIC
struct inode
*
923 struct super_block
*sb
)
930 * Now that the generic code is guaranteed not to be accessing
931 * the linux inode, we can inactivate and reclaim the inode.
934 xfs_fs_destroy_inode(
937 struct xfs_inode
*ip
= XFS_I(inode
);
939 trace_xfs_destroy_inode(ip
);
941 ASSERT(!rwsem_is_locked(&ip
->i_iolock
.mr_lock
));
942 XFS_STATS_INC(ip
->i_mount
, vn_rele
);
943 XFS_STATS_INC(ip
->i_mount
, vn_remove
);
947 ASSERT(XFS_FORCED_SHUTDOWN(ip
->i_mount
) || ip
->i_delayed_blks
== 0);
948 XFS_STATS_INC(ip
->i_mount
, vn_reclaim
);
951 * We should never get here with one of the reclaim flags already set.
953 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIMABLE
));
954 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIM
));
957 * We always use background reclaim here because even if the
958 * inode is clean, it still may be under IO and hence we have
959 * to take the flush lock. The background reclaim path handles
960 * this more efficiently than we can here, so simply let background
961 * reclaim tear down all inodes.
963 xfs_inode_set_reclaim_tag(ip
);
967 * Slab object creation initialisation for the XFS inode.
968 * This covers only the idempotent fields in the XFS inode;
969 * all other fields need to be initialised on allocation
970 * from the slab. This avoids the need to repeatedly initialise
971 * fields in the xfs inode that left in the initialise state
972 * when freeing the inode.
975 xfs_fs_inode_init_once(
978 struct xfs_inode
*ip
= inode
;
980 memset(ip
, 0, sizeof(struct xfs_inode
));
983 inode_init_once(VFS_I(ip
));
986 atomic_set(&ip
->i_pincount
, 0);
987 spin_lock_init(&ip
->i_flags_lock
);
989 mrlock_init(&ip
->i_mmaplock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
990 "xfsino", ip
->i_ino
);
991 mrlock_init(&ip
->i_lock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
992 "xfsino", ip
->i_ino
);
996 * We do an unlocked check for XFS_IDONTCACHE here because we are already
997 * serialised against cache hits here via the inode->i_lock and igrab() in
998 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
999 * racing with us, and it avoids needing to grab a spinlock here for every inode
1000 * we drop the final reference on.
1004 struct inode
*inode
)
1006 struct xfs_inode
*ip
= XFS_I(inode
);
1008 return generic_drop_inode(inode
) || (ip
->i_flags
& XFS_IDONTCACHE
);
1013 struct xfs_mount
*mp
)
1015 kfree(mp
->m_fsname
);
1016 kfree(mp
->m_rtname
);
1017 kfree(mp
->m_logname
);
1022 struct super_block
*sb
,
1025 struct xfs_mount
*mp
= XFS_M(sb
);
1028 * Doing anything during the async pass would be counterproductive.
1033 xfs_log_force(mp
, XFS_LOG_SYNC
);
1036 * The disk must be active because we're syncing.
1037 * We schedule log work now (now that the disk is
1038 * active) instead of later (when it might not be).
1040 flush_delayed_work(&mp
->m_log
->l_work
);
1048 struct dentry
*dentry
,
1049 struct kstatfs
*statp
)
1051 struct xfs_mount
*mp
= XFS_M(dentry
->d_sb
);
1052 xfs_sb_t
*sbp
= &mp
->m_sb
;
1053 struct xfs_inode
*ip
= XFS_I(d_inode(dentry
));
1054 __uint64_t fakeinos
, id
;
1057 __uint64_t fdblocks
;
1061 statp
->f_type
= XFS_SB_MAGIC
;
1062 statp
->f_namelen
= MAXNAMELEN
- 1;
1064 id
= huge_encode_dev(mp
->m_ddev_targp
->bt_dev
);
1065 statp
->f_fsid
.val
[0] = (u32
)id
;
1066 statp
->f_fsid
.val
[1] = (u32
)(id
>> 32);
1068 icount
= percpu_counter_sum(&mp
->m_icount
);
1069 ifree
= percpu_counter_sum(&mp
->m_ifree
);
1070 fdblocks
= percpu_counter_sum(&mp
->m_fdblocks
);
1072 spin_lock(&mp
->m_sb_lock
);
1073 statp
->f_bsize
= sbp
->sb_blocksize
;
1074 lsize
= sbp
->sb_logstart
? sbp
->sb_logblocks
: 0;
1075 statp
->f_blocks
= sbp
->sb_dblocks
- lsize
;
1076 spin_unlock(&mp
->m_sb_lock
);
1078 statp
->f_bfree
= fdblocks
- XFS_ALLOC_SET_ASIDE(mp
);
1079 statp
->f_bavail
= statp
->f_bfree
;
1081 fakeinos
= statp
->f_bfree
<< sbp
->sb_inopblog
;
1082 statp
->f_files
= MIN(icount
+ fakeinos
, (__uint64_t
)XFS_MAXINUMBER
);
1083 if (mp
->m_maxicount
)
1084 statp
->f_files
= min_t(typeof(statp
->f_files
),
1088 /* If sb_icount overshot maxicount, report actual allocation */
1089 statp
->f_files
= max_t(typeof(statp
->f_files
),
1093 /* make sure statp->f_ffree does not underflow */
1094 ffree
= statp
->f_files
- (icount
- ifree
);
1095 statp
->f_ffree
= max_t(__int64_t
, ffree
, 0);
1098 if ((ip
->i_d
.di_flags
& XFS_DIFLAG_PROJINHERIT
) &&
1099 ((mp
->m_qflags
& (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))) ==
1100 (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))
1101 xfs_qm_statvfs(ip
, statp
);
1106 xfs_save_resvblks(struct xfs_mount
*mp
)
1108 __uint64_t resblks
= 0;
1110 mp
->m_resblks_save
= mp
->m_resblks
;
1111 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1115 xfs_restore_resvblks(struct xfs_mount
*mp
)
1119 if (mp
->m_resblks_save
) {
1120 resblks
= mp
->m_resblks_save
;
1121 mp
->m_resblks_save
= 0;
1123 resblks
= xfs_default_resblks(mp
);
1125 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1129 * Trigger writeback of all the dirty metadata in the file system.
1131 * This ensures that the metadata is written to their location on disk rather
1132 * than just existing in transactions in the log. This means after a quiesce
1133 * there is no log replay required to write the inodes to disk - this is the
1134 * primary difference between a sync and a quiesce.
1136 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1137 * it is started again when appropriate.
1141 struct xfs_mount
*mp
)
1145 /* wait for all modifications to complete */
1146 while (atomic_read(&mp
->m_active_trans
) > 0)
1149 /* force the log to unpin objects from the now complete transactions */
1150 xfs_log_force(mp
, XFS_LOG_SYNC
);
1152 /* reclaim inodes to do any IO before the freeze completes */
1153 xfs_reclaim_inodes(mp
, 0);
1154 xfs_reclaim_inodes(mp
, SYNC_WAIT
);
1156 /* Push the superblock and write an unmount record */
1157 error
= xfs_log_sbcount(mp
);
1159 xfs_warn(mp
, "xfs_attr_quiesce: failed to log sb changes. "
1160 "Frozen image may not be consistent.");
1162 * Just warn here till VFS can correctly support
1163 * read-only remount without racing.
1165 WARN_ON(atomic_read(&mp
->m_active_trans
) != 0);
1167 xfs_log_quiesce(mp
);
1171 xfs_test_remount_options(
1172 struct super_block
*sb
,
1173 struct xfs_mount
*mp
,
1177 struct xfs_mount
*tmp_mp
;
1179 tmp_mp
= kmem_zalloc(sizeof(*tmp_mp
), KM_MAYFAIL
);
1183 tmp_mp
->m_super
= sb
;
1184 error
= xfs_parseargs(tmp_mp
, options
);
1185 xfs_free_fsname(tmp_mp
);
1193 struct super_block
*sb
,
1197 struct xfs_mount
*mp
= XFS_M(sb
);
1198 xfs_sb_t
*sbp
= &mp
->m_sb
;
1199 substring_t args
[MAX_OPT_ARGS
];
1203 /* First, check for complete junk; i.e. invalid options */
1204 error
= xfs_test_remount_options(sb
, mp
, options
);
1208 sync_filesystem(sb
);
1209 while ((p
= strsep(&options
, ",")) != NULL
) {
1215 token
= match_token(p
, tokens
, args
);
1218 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
1221 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
1224 mp
->m_flags
&= ~XFS_MOUNT_SMALL_INUMS
;
1225 mp
->m_maxagi
= xfs_set_inode_alloc(mp
, sbp
->sb_agcount
);
1228 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
1229 mp
->m_maxagi
= xfs_set_inode_alloc(mp
, sbp
->sb_agcount
);
1233 * Logically we would return an error here to prevent
1234 * users from believing they might have changed
1235 * mount options using remount which can't be changed.
1237 * But unfortunately mount(8) adds all options from
1238 * mtab and fstab to the mount arguments in some cases
1239 * so we can't blindly reject options, but have to
1240 * check for each specified option if it actually
1241 * differs from the currently set option and only
1242 * reject it if that's the case.
1244 * Until that is implemented we return success for
1245 * every remount request, and silently ignore all
1246 * options that we can't actually change.
1250 "mount option \"%s\" not supported for remount", p
);
1259 if ((mp
->m_flags
& XFS_MOUNT_RDONLY
) && !(*flags
& MS_RDONLY
)) {
1260 if (mp
->m_flags
& XFS_MOUNT_NORECOVERY
) {
1262 "ro->rw transition prohibited on norecovery mount");
1266 if (XFS_SB_VERSION_NUM(sbp
) == XFS_SB_VERSION_5
&&
1267 xfs_sb_has_ro_compat_feature(sbp
,
1268 XFS_SB_FEAT_RO_COMPAT_UNKNOWN
)) {
1270 "ro->rw transition prohibited on unknown (0x%x) ro-compat filesystem",
1271 (sbp
->sb_features_ro_compat
&
1272 XFS_SB_FEAT_RO_COMPAT_UNKNOWN
));
1276 mp
->m_flags
&= ~XFS_MOUNT_RDONLY
;
1279 * If this is the first remount to writeable state we
1280 * might have some superblock changes to update.
1282 if (mp
->m_update_sb
) {
1283 error
= xfs_sync_sb(mp
, false);
1285 xfs_warn(mp
, "failed to write sb changes");
1288 mp
->m_update_sb
= false;
1292 * Fill out the reserve pool if it is empty. Use the stashed
1293 * value if it is non-zero, otherwise go with the default.
1295 xfs_restore_resvblks(mp
);
1296 xfs_log_work_queue(mp
);
1300 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
) && (*flags
& MS_RDONLY
)) {
1302 * Before we sync the metadata, we need to free up the reserve
1303 * block pool so that the used block count in the superblock on
1304 * disk is correct at the end of the remount. Stash the current
1305 * reserve pool size so that if we get remounted rw, we can
1306 * return it to the same size.
1308 xfs_save_resvblks(mp
);
1309 xfs_quiesce_attr(mp
);
1310 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
1317 * Second stage of a freeze. The data is already frozen so we only
1318 * need to take care of the metadata. Once that's done sync the superblock
1319 * to the log to dirty it in case of a crash while frozen. This ensures that we
1320 * will recover the unlinked inode lists on the next mount.
1324 struct super_block
*sb
)
1326 struct xfs_mount
*mp
= XFS_M(sb
);
1328 xfs_save_resvblks(mp
);
1329 xfs_quiesce_attr(mp
);
1330 return xfs_sync_sb(mp
, true);
1335 struct super_block
*sb
)
1337 struct xfs_mount
*mp
= XFS_M(sb
);
1339 xfs_restore_resvblks(mp
);
1340 xfs_log_work_queue(mp
);
1345 xfs_fs_show_options(
1347 struct dentry
*root
)
1349 return xfs_showargs(XFS_M(root
->d_sb
), m
);
1353 * This function fills in xfs_mount_t fields based on mount args.
1354 * Note: the superblock _has_ now been read in.
1358 struct xfs_mount
*mp
)
1360 int ronly
= (mp
->m_flags
& XFS_MOUNT_RDONLY
);
1362 /* Fail a mount where the logbuf is smaller than the log stripe */
1363 if (xfs_sb_version_haslogv2(&mp
->m_sb
)) {
1364 if (mp
->m_logbsize
<= 0 &&
1365 mp
->m_sb
.sb_logsunit
> XLOG_BIG_RECORD_BSIZE
) {
1366 mp
->m_logbsize
= mp
->m_sb
.sb_logsunit
;
1367 } else if (mp
->m_logbsize
> 0 &&
1368 mp
->m_logbsize
< mp
->m_sb
.sb_logsunit
) {
1370 "logbuf size must be greater than or equal to log stripe size");
1374 /* Fail a mount if the logbuf is larger than 32K */
1375 if (mp
->m_logbsize
> XLOG_BIG_RECORD_BSIZE
) {
1377 "logbuf size for version 1 logs must be 16K or 32K");
1383 * V5 filesystems always use attr2 format for attributes.
1385 if (xfs_sb_version_hascrc(&mp
->m_sb
) &&
1386 (mp
->m_flags
& XFS_MOUNT_NOATTR2
)) {
1387 xfs_warn(mp
, "Cannot mount a V5 filesystem as noattr2. "
1388 "attr2 is always enabled for V5 filesystems.");
1393 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1394 * told by noattr2 to turn it off
1396 if (xfs_sb_version_hasattr2(&mp
->m_sb
) &&
1397 !(mp
->m_flags
& XFS_MOUNT_NOATTR2
))
1398 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
1401 * prohibit r/w mounts of read-only filesystems
1403 if ((mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) && !ronly
) {
1405 "cannot mount a read-only filesystem as read-write");
1409 if ((mp
->m_qflags
& (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
)) &&
1410 (mp
->m_qflags
& (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
)) &&
1411 !xfs_sb_version_has_pquotino(&mp
->m_sb
)) {
1413 "Super block does not support project and group quota together");
1421 xfs_init_percpu_counters(
1422 struct xfs_mount
*mp
)
1426 error
= percpu_counter_init(&mp
->m_icount
, 0, GFP_KERNEL
);
1430 error
= percpu_counter_init(&mp
->m_ifree
, 0, GFP_KERNEL
);
1434 error
= percpu_counter_init(&mp
->m_fdblocks
, 0, GFP_KERNEL
);
1441 percpu_counter_destroy(&mp
->m_ifree
);
1443 percpu_counter_destroy(&mp
->m_icount
);
1448 xfs_reinit_percpu_counters(
1449 struct xfs_mount
*mp
)
1451 percpu_counter_set(&mp
->m_icount
, mp
->m_sb
.sb_icount
);
1452 percpu_counter_set(&mp
->m_ifree
, mp
->m_sb
.sb_ifree
);
1453 percpu_counter_set(&mp
->m_fdblocks
, mp
->m_sb
.sb_fdblocks
);
1457 xfs_destroy_percpu_counters(
1458 struct xfs_mount
*mp
)
1460 percpu_counter_destroy(&mp
->m_icount
);
1461 percpu_counter_destroy(&mp
->m_ifree
);
1462 percpu_counter_destroy(&mp
->m_fdblocks
);
1467 struct super_block
*sb
,
1472 struct xfs_mount
*mp
= NULL
;
1473 int flags
= 0, error
= -ENOMEM
;
1475 mp
= kzalloc(sizeof(struct xfs_mount
), GFP_KERNEL
);
1479 spin_lock_init(&mp
->m_sb_lock
);
1480 mutex_init(&mp
->m_growlock
);
1481 atomic_set(&mp
->m_active_trans
, 0);
1482 INIT_DELAYED_WORK(&mp
->m_reclaim_work
, xfs_reclaim_worker
);
1483 INIT_DELAYED_WORK(&mp
->m_eofblocks_work
, xfs_eofblocks_worker
);
1484 mp
->m_kobj
.kobject
.kset
= xfs_kset
;
1489 error
= xfs_parseargs(mp
, (char *)data
);
1491 goto out_free_fsname
;
1493 sb_min_blocksize(sb
, BBSIZE
);
1494 sb
->s_xattr
= xfs_xattr_handlers
;
1495 sb
->s_export_op
= &xfs_export_operations
;
1496 #ifdef CONFIG_XFS_QUOTA
1497 sb
->s_qcop
= &xfs_quotactl_operations
;
1498 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
1500 sb
->s_op
= &xfs_super_operations
;
1503 flags
|= XFS_MFSI_QUIET
;
1505 error
= xfs_open_devices(mp
);
1507 goto out_free_fsname
;
1509 error
= xfs_init_mount_workqueues(mp
);
1511 goto out_close_devices
;
1513 error
= xfs_init_percpu_counters(mp
);
1515 goto out_destroy_workqueues
;
1517 /* Allocate stats memory before we do operations that might use it */
1518 mp
->m_stats
.xs_stats
= alloc_percpu(struct xfsstats
);
1519 if (!mp
->m_stats
.xs_stats
) {
1521 goto out_destroy_counters
;
1524 error
= xfs_readsb(mp
, flags
);
1526 goto out_free_stats
;
1528 error
= xfs_finish_flags(mp
);
1532 error
= xfs_setup_devices(mp
);
1536 error
= xfs_filestream_mount(mp
);
1541 * we must configure the block size in the superblock before we run the
1542 * full mount process as the mount process can lookup and cache inodes.
1544 sb
->s_magic
= XFS_SB_MAGIC
;
1545 sb
->s_blocksize
= mp
->m_sb
.sb_blocksize
;
1546 sb
->s_blocksize_bits
= ffs(sb
->s_blocksize
) - 1;
1547 sb
->s_maxbytes
= xfs_max_file_offset(sb
->s_blocksize_bits
);
1548 sb
->s_max_links
= XFS_MAXLINK
;
1549 sb
->s_time_gran
= 1;
1550 set_posix_acl_flag(sb
);
1552 /* version 5 superblocks support inode version counters. */
1553 if (XFS_SB_VERSION_NUM(&mp
->m_sb
) == XFS_SB_VERSION_5
)
1554 sb
->s_flags
|= MS_I_VERSION
;
1556 if (mp
->m_flags
& XFS_MOUNT_DAX
) {
1558 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1560 error
= bdev_dax_supported(sb
, sb
->s_blocksize
);
1563 "DAX unsupported by block device. Turning off DAX.");
1564 mp
->m_flags
&= ~XFS_MOUNT_DAX
;
1568 if (xfs_sb_version_hassparseinodes(&mp
->m_sb
))
1570 "EXPERIMENTAL sparse inode feature enabled. Use at your own risk!");
1572 error
= xfs_mountfs(mp
);
1574 goto out_filestream_unmount
;
1576 root
= igrab(VFS_I(mp
->m_rootip
));
1581 sb
->s_root
= d_make_root(root
);
1589 out_filestream_unmount
:
1590 xfs_filestream_unmount(mp
);
1594 free_percpu(mp
->m_stats
.xs_stats
);
1595 out_destroy_counters
:
1596 xfs_destroy_percpu_counters(mp
);
1597 out_destroy_workqueues
:
1598 xfs_destroy_mount_workqueues(mp
);
1600 xfs_close_devices(mp
);
1602 xfs_free_fsname(mp
);
1608 xfs_filestream_unmount(mp
);
1615 struct super_block
*sb
)
1617 struct xfs_mount
*mp
= XFS_M(sb
);
1619 xfs_notice(mp
, "Unmounting Filesystem");
1620 xfs_filestream_unmount(mp
);
1624 free_percpu(mp
->m_stats
.xs_stats
);
1625 xfs_destroy_percpu_counters(mp
);
1626 xfs_destroy_mount_workqueues(mp
);
1627 xfs_close_devices(mp
);
1628 xfs_free_fsname(mp
);
1632 STATIC
struct dentry
*
1634 struct file_system_type
*fs_type
,
1636 const char *dev_name
,
1639 return mount_bdev(fs_type
, flags
, dev_name
, data
, xfs_fs_fill_super
);
1643 xfs_fs_nr_cached_objects(
1644 struct super_block
*sb
,
1645 struct shrink_control
*sc
)
1647 return xfs_reclaim_inodes_count(XFS_M(sb
));
1651 xfs_fs_free_cached_objects(
1652 struct super_block
*sb
,
1653 struct shrink_control
*sc
)
1655 return xfs_reclaim_inodes_nr(XFS_M(sb
), sc
->nr_to_scan
);
1658 static const struct super_operations xfs_super_operations
= {
1659 .alloc_inode
= xfs_fs_alloc_inode
,
1660 .destroy_inode
= xfs_fs_destroy_inode
,
1661 .drop_inode
= xfs_fs_drop_inode
,
1662 .put_super
= xfs_fs_put_super
,
1663 .sync_fs
= xfs_fs_sync_fs
,
1664 .freeze_fs
= xfs_fs_freeze
,
1665 .unfreeze_fs
= xfs_fs_unfreeze
,
1666 .statfs
= xfs_fs_statfs
,
1667 .remount_fs
= xfs_fs_remount
,
1668 .show_options
= xfs_fs_show_options
,
1669 .nr_cached_objects
= xfs_fs_nr_cached_objects
,
1670 .free_cached_objects
= xfs_fs_free_cached_objects
,
1673 static struct file_system_type xfs_fs_type
= {
1674 .owner
= THIS_MODULE
,
1676 .mount
= xfs_fs_mount
,
1677 .kill_sb
= kill_block_super
,
1678 .fs_flags
= FS_REQUIRES_DEV
,
1680 MODULE_ALIAS_FS("xfs");
1683 xfs_init_zones(void)
1685 xfs_ioend_bioset
= bioset_create(4 * MAX_BUF_PER_PAGE
,
1686 offsetof(struct xfs_ioend
, io_inline_bio
));
1687 if (!xfs_ioend_bioset
)
1690 xfs_log_ticket_zone
= kmem_zone_init(sizeof(xlog_ticket_t
),
1692 if (!xfs_log_ticket_zone
)
1693 goto out_free_ioend_bioset
;
1695 xfs_bmap_free_item_zone
= kmem_zone_init(sizeof(xfs_bmap_free_item_t
),
1696 "xfs_bmap_free_item");
1697 if (!xfs_bmap_free_item_zone
)
1698 goto out_destroy_log_ticket_zone
;
1700 xfs_btree_cur_zone
= kmem_zone_init(sizeof(xfs_btree_cur_t
),
1702 if (!xfs_btree_cur_zone
)
1703 goto out_destroy_bmap_free_item_zone
;
1705 xfs_da_state_zone
= kmem_zone_init(sizeof(xfs_da_state_t
),
1707 if (!xfs_da_state_zone
)
1708 goto out_destroy_btree_cur_zone
;
1710 xfs_ifork_zone
= kmem_zone_init(sizeof(xfs_ifork_t
), "xfs_ifork");
1711 if (!xfs_ifork_zone
)
1712 goto out_destroy_da_state_zone
;
1714 xfs_trans_zone
= kmem_zone_init(sizeof(xfs_trans_t
), "xfs_trans");
1715 if (!xfs_trans_zone
)
1716 goto out_destroy_ifork_zone
;
1718 xfs_log_item_desc_zone
=
1719 kmem_zone_init(sizeof(struct xfs_log_item_desc
),
1720 "xfs_log_item_desc");
1721 if (!xfs_log_item_desc_zone
)
1722 goto out_destroy_trans_zone
;
1725 * The size of the zone allocated buf log item is the maximum
1726 * size possible under XFS. This wastes a little bit of memory,
1727 * but it is much faster.
1729 xfs_buf_item_zone
= kmem_zone_init(sizeof(struct xfs_buf_log_item
),
1731 if (!xfs_buf_item_zone
)
1732 goto out_destroy_log_item_desc_zone
;
1734 xfs_efd_zone
= kmem_zone_init((sizeof(xfs_efd_log_item_t
) +
1735 ((XFS_EFD_MAX_FAST_EXTENTS
- 1) *
1736 sizeof(xfs_extent_t
))), "xfs_efd_item");
1738 goto out_destroy_buf_item_zone
;
1740 xfs_efi_zone
= kmem_zone_init((sizeof(xfs_efi_log_item_t
) +
1741 ((XFS_EFI_MAX_FAST_EXTENTS
- 1) *
1742 sizeof(xfs_extent_t
))), "xfs_efi_item");
1744 goto out_destroy_efd_zone
;
1747 kmem_zone_init_flags(sizeof(xfs_inode_t
), "xfs_inode",
1748 KM_ZONE_HWALIGN
| KM_ZONE_RECLAIM
| KM_ZONE_SPREAD
|
1749 KM_ZONE_ACCOUNT
, xfs_fs_inode_init_once
);
1750 if (!xfs_inode_zone
)
1751 goto out_destroy_efi_zone
;
1754 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t
), "xfs_ili",
1755 KM_ZONE_SPREAD
, NULL
);
1757 goto out_destroy_inode_zone
;
1758 xfs_icreate_zone
= kmem_zone_init(sizeof(struct xfs_icreate_item
),
1760 if (!xfs_icreate_zone
)
1761 goto out_destroy_ili_zone
;
1765 out_destroy_ili_zone
:
1766 kmem_zone_destroy(xfs_ili_zone
);
1767 out_destroy_inode_zone
:
1768 kmem_zone_destroy(xfs_inode_zone
);
1769 out_destroy_efi_zone
:
1770 kmem_zone_destroy(xfs_efi_zone
);
1771 out_destroy_efd_zone
:
1772 kmem_zone_destroy(xfs_efd_zone
);
1773 out_destroy_buf_item_zone
:
1774 kmem_zone_destroy(xfs_buf_item_zone
);
1775 out_destroy_log_item_desc_zone
:
1776 kmem_zone_destroy(xfs_log_item_desc_zone
);
1777 out_destroy_trans_zone
:
1778 kmem_zone_destroy(xfs_trans_zone
);
1779 out_destroy_ifork_zone
:
1780 kmem_zone_destroy(xfs_ifork_zone
);
1781 out_destroy_da_state_zone
:
1782 kmem_zone_destroy(xfs_da_state_zone
);
1783 out_destroy_btree_cur_zone
:
1784 kmem_zone_destroy(xfs_btree_cur_zone
);
1785 out_destroy_bmap_free_item_zone
:
1786 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1787 out_destroy_log_ticket_zone
:
1788 kmem_zone_destroy(xfs_log_ticket_zone
);
1789 out_free_ioend_bioset
:
1790 bioset_free(xfs_ioend_bioset
);
1796 xfs_destroy_zones(void)
1799 * Make sure all delayed rcu free are flushed before we
1803 kmem_zone_destroy(xfs_icreate_zone
);
1804 kmem_zone_destroy(xfs_ili_zone
);
1805 kmem_zone_destroy(xfs_inode_zone
);
1806 kmem_zone_destroy(xfs_efi_zone
);
1807 kmem_zone_destroy(xfs_efd_zone
);
1808 kmem_zone_destroy(xfs_buf_item_zone
);
1809 kmem_zone_destroy(xfs_log_item_desc_zone
);
1810 kmem_zone_destroy(xfs_trans_zone
);
1811 kmem_zone_destroy(xfs_ifork_zone
);
1812 kmem_zone_destroy(xfs_da_state_zone
);
1813 kmem_zone_destroy(xfs_btree_cur_zone
);
1814 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1815 kmem_zone_destroy(xfs_log_ticket_zone
);
1816 bioset_free(xfs_ioend_bioset
);
1820 xfs_init_workqueues(void)
1823 * The allocation workqueue can be used in memory reclaim situations
1824 * (writepage path), and parallelism is only limited by the number of
1825 * AGs in all the filesystems mounted. Hence use the default large
1826 * max_active value for this workqueue.
1828 xfs_alloc_wq
= alloc_workqueue("xfsalloc",
1829 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0);
1837 xfs_destroy_workqueues(void)
1839 destroy_workqueue(xfs_alloc_wq
);
1847 xfs_check_ondisk_structs();
1849 printk(KERN_INFO XFS_VERSION_STRING
" with "
1850 XFS_BUILD_OPTIONS
" enabled\n");
1854 error
= xfs_init_zones();
1858 error
= xfs_init_workqueues();
1860 goto out_destroy_zones
;
1862 error
= xfs_mru_cache_init();
1864 goto out_destroy_wq
;
1866 error
= xfs_buf_init();
1868 goto out_mru_cache_uninit
;
1870 error
= xfs_init_procfs();
1872 goto out_buf_terminate
;
1874 error
= xfs_sysctl_register();
1876 goto out_cleanup_procfs
;
1878 xfs_kset
= kset_create_and_add("xfs", NULL
, fs_kobj
);
1881 goto out_sysctl_unregister
;
1884 xfsstats
.xs_kobj
.kobject
.kset
= xfs_kset
;
1886 xfsstats
.xs_stats
= alloc_percpu(struct xfsstats
);
1887 if (!xfsstats
.xs_stats
) {
1889 goto out_kset_unregister
;
1892 error
= xfs_sysfs_init(&xfsstats
.xs_kobj
, &xfs_stats_ktype
, NULL
,
1895 goto out_free_stats
;
1898 xfs_dbg_kobj
.kobject
.kset
= xfs_kset
;
1899 error
= xfs_sysfs_init(&xfs_dbg_kobj
, &xfs_dbg_ktype
, NULL
, "debug");
1901 goto out_remove_stats_kobj
;
1904 error
= xfs_qm_init();
1906 goto out_remove_dbg_kobj
;
1908 error
= register_filesystem(&xfs_fs_type
);
1915 out_remove_dbg_kobj
:
1917 xfs_sysfs_del(&xfs_dbg_kobj
);
1918 out_remove_stats_kobj
:
1920 xfs_sysfs_del(&xfsstats
.xs_kobj
);
1922 free_percpu(xfsstats
.xs_stats
);
1923 out_kset_unregister
:
1924 kset_unregister(xfs_kset
);
1925 out_sysctl_unregister
:
1926 xfs_sysctl_unregister();
1928 xfs_cleanup_procfs();
1930 xfs_buf_terminate();
1931 out_mru_cache_uninit
:
1932 xfs_mru_cache_uninit();
1934 xfs_destroy_workqueues();
1936 xfs_destroy_zones();
1945 unregister_filesystem(&xfs_fs_type
);
1947 xfs_sysfs_del(&xfs_dbg_kobj
);
1949 xfs_sysfs_del(&xfsstats
.xs_kobj
);
1950 free_percpu(xfsstats
.xs_stats
);
1951 kset_unregister(xfs_kset
);
1952 xfs_sysctl_unregister();
1953 xfs_cleanup_procfs();
1954 xfs_buf_terminate();
1955 xfs_mru_cache_uninit();
1956 xfs_destroy_workqueues();
1957 xfs_destroy_zones();
1958 xfs_uuid_table_free();
1961 module_init(init_xfs_fs
);
1962 module_exit(exit_xfs_fs
);
1964 MODULE_AUTHOR("Silicon Graphics, Inc.");
1965 MODULE_DESCRIPTION(XFS_VERSION_STRING
" with " XFS_BUILD_OPTIONS
" enabled");
1966 MODULE_LICENSE("GPL");