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"
49 #include "xfs_rmap_item.h"
51 #include <linux/namei.h>
52 #include <linux/init.h>
53 #include <linux/slab.h>
54 #include <linux/mount.h>
55 #include <linux/mempool.h>
56 #include <linux/writeback.h>
57 #include <linux/kthread.h>
58 #include <linux/freezer.h>
59 #include <linux/parser.h>
61 static const struct super_operations xfs_super_operations
;
62 struct bio_set
*xfs_ioend_bioset
;
64 static struct kset
*xfs_kset
; /* top-level xfs sysfs dir */
66 static struct xfs_kobj xfs_dbg_kobj
; /* global debug sysfs attrs */
70 * Table driven mount option parser.
73 Opt_logbufs
, Opt_logbsize
, Opt_logdev
, Opt_rtdev
, Opt_biosize
,
74 Opt_wsync
, Opt_noalign
, Opt_swalloc
, Opt_sunit
, Opt_swidth
, Opt_nouuid
,
75 Opt_mtpt
, Opt_grpid
, Opt_nogrpid
, Opt_bsdgroups
, Opt_sysvgroups
,
76 Opt_allocsize
, Opt_norecovery
, Opt_barrier
, Opt_nobarrier
,
77 Opt_inode64
, Opt_inode32
, Opt_ikeep
, Opt_noikeep
,
78 Opt_largeio
, Opt_nolargeio
, Opt_attr2
, Opt_noattr2
, Opt_filestreams
,
79 Opt_quota
, Opt_noquota
, Opt_usrquota
, Opt_grpquota
, Opt_prjquota
,
80 Opt_uquota
, Opt_gquota
, Opt_pquota
,
81 Opt_uqnoenforce
, Opt_gqnoenforce
, Opt_pqnoenforce
, Opt_qnoenforce
,
82 Opt_discard
, Opt_nodiscard
, Opt_dax
, Opt_err
,
85 static const match_table_t tokens
= {
86 {Opt_logbufs
, "logbufs=%u"}, /* number of XFS log buffers */
87 {Opt_logbsize
, "logbsize=%s"}, /* size of XFS log buffers */
88 {Opt_logdev
, "logdev=%s"}, /* log device */
89 {Opt_rtdev
, "rtdev=%s"}, /* realtime I/O device */
90 {Opt_biosize
, "biosize=%u"}, /* log2 of preferred buffered io size */
91 {Opt_wsync
, "wsync"}, /* safe-mode nfs compatible mount */
92 {Opt_noalign
, "noalign"}, /* turn off stripe alignment */
93 {Opt_swalloc
, "swalloc"}, /* turn on stripe width allocation */
94 {Opt_sunit
, "sunit=%u"}, /* data volume stripe unit */
95 {Opt_swidth
, "swidth=%u"}, /* data volume stripe width */
96 {Opt_nouuid
, "nouuid"}, /* ignore filesystem UUID */
97 {Opt_mtpt
, "mtpt"}, /* filesystem mount point */
98 {Opt_grpid
, "grpid"}, /* group-ID from parent directory */
99 {Opt_nogrpid
, "nogrpid"}, /* group-ID from current process */
100 {Opt_bsdgroups
, "bsdgroups"}, /* group-ID from parent directory */
101 {Opt_sysvgroups
,"sysvgroups"}, /* group-ID from current process */
102 {Opt_allocsize
, "allocsize=%s"},/* preferred allocation size */
103 {Opt_norecovery
,"norecovery"}, /* don't run XFS recovery */
104 {Opt_barrier
, "barrier"}, /* use writer barriers for log write and
105 * unwritten extent conversion */
106 {Opt_nobarrier
, "nobarrier"}, /* .. disable */
107 {Opt_inode64
, "inode64"}, /* inodes can be allocated anywhere */
108 {Opt_inode32
, "inode32"}, /* inode allocation limited to
109 * XFS_MAXINUMBER_32 */
110 {Opt_ikeep
, "ikeep"}, /* do not free empty inode clusters */
111 {Opt_noikeep
, "noikeep"}, /* free empty inode clusters */
112 {Opt_largeio
, "largeio"}, /* report large I/O sizes in stat() */
113 {Opt_nolargeio
, "nolargeio"}, /* do not report large I/O sizes
115 {Opt_attr2
, "attr2"}, /* do use attr2 attribute format */
116 {Opt_noattr2
, "noattr2"}, /* do not use attr2 attribute format */
117 {Opt_filestreams
,"filestreams"},/* use filestreams allocator */
118 {Opt_quota
, "quota"}, /* disk quotas (user) */
119 {Opt_noquota
, "noquota"}, /* no quotas */
120 {Opt_usrquota
, "usrquota"}, /* user quota enabled */
121 {Opt_grpquota
, "grpquota"}, /* group quota enabled */
122 {Opt_prjquota
, "prjquota"}, /* project quota enabled */
123 {Opt_uquota
, "uquota"}, /* user quota (IRIX variant) */
124 {Opt_gquota
, "gquota"}, /* group quota (IRIX variant) */
125 {Opt_pquota
, "pquota"}, /* project quota (IRIX variant) */
126 {Opt_uqnoenforce
,"uqnoenforce"},/* user quota limit enforcement */
127 {Opt_gqnoenforce
,"gqnoenforce"},/* group quota limit enforcement */
128 {Opt_pqnoenforce
,"pqnoenforce"},/* project quota limit enforcement */
129 {Opt_qnoenforce
, "qnoenforce"}, /* same as uqnoenforce */
130 {Opt_discard
, "discard"}, /* Discard unused blocks */
131 {Opt_nodiscard
, "nodiscard"}, /* Do not discard unused blocks */
133 {Opt_dax
, "dax"}, /* Enable direct access to bdev pages */
139 suffix_kstrtoint(const substring_t
*s
, unsigned int base
, int *res
)
141 int last
, shift_left_factor
= 0, _res
;
145 value
= match_strdup(s
);
149 last
= strlen(value
) - 1;
150 if (value
[last
] == 'K' || value
[last
] == 'k') {
151 shift_left_factor
= 10;
154 if (value
[last
] == 'M' || value
[last
] == 'm') {
155 shift_left_factor
= 20;
158 if (value
[last
] == 'G' || value
[last
] == 'g') {
159 shift_left_factor
= 30;
163 if (kstrtoint(value
, base
, &_res
))
166 *res
= _res
<< shift_left_factor
;
171 * This function fills in xfs_mount_t fields based on mount args.
172 * Note: the superblock has _not_ yet been read in.
174 * Note that this function leaks the various device name allocations on
175 * failure. The caller takes care of them.
177 * *sb is const because this is also used to test options on the remount
178 * path, and we don't want this to have any side effects at remount time.
179 * Today this function does not change *sb, but just to future-proof...
183 struct xfs_mount
*mp
,
186 const struct super_block
*sb
= mp
->m_super
;
188 substring_t args
[MAX_OPT_ARGS
];
192 __uint8_t iosizelog
= 0;
195 * set up the mount name first so all the errors will refer to the
198 mp
->m_fsname
= kstrndup(sb
->s_id
, MAXNAMELEN
, GFP_KERNEL
);
201 mp
->m_fsname_len
= strlen(mp
->m_fsname
) + 1;
204 * Copy binary VFS mount flags we are interested in.
206 if (sb
->s_flags
& MS_RDONLY
)
207 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
208 if (sb
->s_flags
& MS_DIRSYNC
)
209 mp
->m_flags
|= XFS_MOUNT_DIRSYNC
;
210 if (sb
->s_flags
& MS_SYNCHRONOUS
)
211 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
214 * Set some default flags that could be cleared by the mount option
217 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
218 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
221 * These can be overridden by the mount option parsing.
229 while ((p
= strsep(&options
, ",")) != NULL
) {
235 token
= match_token(p
, tokens
, args
);
238 if (match_int(args
, &mp
->m_logbufs
))
242 if (suffix_kstrtoint(args
, 10, &mp
->m_logbsize
))
246 mp
->m_logname
= match_strdup(args
);
251 xfs_warn(mp
, "%s option not allowed on this system", p
);
254 mp
->m_rtname
= match_strdup(args
);
260 if (suffix_kstrtoint(args
, 10, &iosize
))
262 iosizelog
= ffs(iosize
) - 1;
266 mp
->m_flags
|= XFS_MOUNT_GRPID
;
270 mp
->m_flags
&= ~XFS_MOUNT_GRPID
;
273 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
276 mp
->m_flags
|= XFS_MOUNT_NORECOVERY
;
279 mp
->m_flags
|= XFS_MOUNT_NOALIGN
;
282 mp
->m_flags
|= XFS_MOUNT_SWALLOC
;
285 if (match_int(args
, &dsunit
))
289 if (match_int(args
, &dswidth
))
293 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
296 mp
->m_flags
&= ~XFS_MOUNT_SMALL_INUMS
;
299 mp
->m_flags
|= XFS_MOUNT_NOUUID
;
302 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
305 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
308 mp
->m_flags
|= XFS_MOUNT_IKEEP
;
311 mp
->m_flags
&= ~XFS_MOUNT_IKEEP
;
314 mp
->m_flags
&= ~XFS_MOUNT_COMPAT_IOSIZE
;
317 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
320 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
323 mp
->m_flags
&= ~XFS_MOUNT_ATTR2
;
324 mp
->m_flags
|= XFS_MOUNT_NOATTR2
;
326 case Opt_filestreams
:
327 mp
->m_flags
|= XFS_MOUNT_FILESTREAMS
;
330 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACCT
;
331 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ENFD
;
332 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACTIVE
;
337 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
|
341 case Opt_uqnoenforce
:
342 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
);
343 mp
->m_qflags
&= ~XFS_UQUOTA_ENFD
;
347 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
|
350 case Opt_pqnoenforce
:
351 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
);
352 mp
->m_qflags
&= ~XFS_PQUOTA_ENFD
;
356 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
|
359 case Opt_gqnoenforce
:
360 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
);
361 mp
->m_qflags
&= ~XFS_GQUOTA_ENFD
;
364 mp
->m_flags
|= XFS_MOUNT_DISCARD
;
367 mp
->m_flags
&= ~XFS_MOUNT_DISCARD
;
371 mp
->m_flags
|= XFS_MOUNT_DAX
;
375 xfs_warn(mp
, "unknown mount option [%s].", p
);
381 * no recovery flag requires a read-only mount
383 if ((mp
->m_flags
& XFS_MOUNT_NORECOVERY
) &&
384 !(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
385 xfs_warn(mp
, "no-recovery mounts must be read-only.");
389 if ((mp
->m_flags
& XFS_MOUNT_NOALIGN
) && (dsunit
|| dswidth
)) {
391 "sunit and swidth options incompatible with the noalign option");
395 #ifndef CONFIG_XFS_QUOTA
396 if (XFS_IS_QUOTA_RUNNING(mp
)) {
397 xfs_warn(mp
, "quota support not available in this kernel.");
402 if ((dsunit
&& !dswidth
) || (!dsunit
&& dswidth
)) {
403 xfs_warn(mp
, "sunit and swidth must be specified together");
407 if (dsunit
&& (dswidth
% dsunit
!= 0)) {
409 "stripe width (%d) must be a multiple of the stripe unit (%d)",
415 if (dsunit
&& !(mp
->m_flags
& XFS_MOUNT_NOALIGN
)) {
417 * At this point the superblock has not been read
418 * in, therefore we do not know the block size.
419 * Before the mount call ends we will convert
422 mp
->m_dalign
= dsunit
;
423 mp
->m_swidth
= dswidth
;
426 if (mp
->m_logbufs
!= -1 &&
427 mp
->m_logbufs
!= 0 &&
428 (mp
->m_logbufs
< XLOG_MIN_ICLOGS
||
429 mp
->m_logbufs
> XLOG_MAX_ICLOGS
)) {
430 xfs_warn(mp
, "invalid logbufs value: %d [not %d-%d]",
431 mp
->m_logbufs
, XLOG_MIN_ICLOGS
, XLOG_MAX_ICLOGS
);
434 if (mp
->m_logbsize
!= -1 &&
435 mp
->m_logbsize
!= 0 &&
436 (mp
->m_logbsize
< XLOG_MIN_RECORD_BSIZE
||
437 mp
->m_logbsize
> XLOG_MAX_RECORD_BSIZE
||
438 !is_power_of_2(mp
->m_logbsize
))) {
440 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
446 if (iosizelog
> XFS_MAX_IO_LOG
||
447 iosizelog
< XFS_MIN_IO_LOG
) {
448 xfs_warn(mp
, "invalid log iosize: %d [not %d-%d]",
449 iosizelog
, XFS_MIN_IO_LOG
,
454 mp
->m_flags
|= XFS_MOUNT_DFLT_IOSIZE
;
455 mp
->m_readio_log
= iosizelog
;
456 mp
->m_writeio_log
= iosizelog
;
462 struct proc_xfs_info
{
469 struct xfs_mount
*mp
,
472 static struct proc_xfs_info xfs_info_set
[] = {
473 /* the few simple ones we can get from the mount struct */
474 { XFS_MOUNT_IKEEP
, ",ikeep" },
475 { XFS_MOUNT_WSYNC
, ",wsync" },
476 { XFS_MOUNT_NOALIGN
, ",noalign" },
477 { XFS_MOUNT_SWALLOC
, ",swalloc" },
478 { XFS_MOUNT_NOUUID
, ",nouuid" },
479 { XFS_MOUNT_NORECOVERY
, ",norecovery" },
480 { XFS_MOUNT_ATTR2
, ",attr2" },
481 { XFS_MOUNT_FILESTREAMS
, ",filestreams" },
482 { XFS_MOUNT_GRPID
, ",grpid" },
483 { XFS_MOUNT_DISCARD
, ",discard" },
484 { XFS_MOUNT_SMALL_INUMS
, ",inode32" },
485 { XFS_MOUNT_DAX
, ",dax" },
488 static struct proc_xfs_info xfs_info_unset
[] = {
489 /* the few simple ones we can get from the mount struct */
490 { XFS_MOUNT_COMPAT_IOSIZE
, ",largeio" },
491 { XFS_MOUNT_BARRIER
, ",nobarrier" },
492 { XFS_MOUNT_SMALL_INUMS
, ",inode64" },
495 struct proc_xfs_info
*xfs_infop
;
497 for (xfs_infop
= xfs_info_set
; xfs_infop
->flag
; xfs_infop
++) {
498 if (mp
->m_flags
& xfs_infop
->flag
)
499 seq_puts(m
, xfs_infop
->str
);
501 for (xfs_infop
= xfs_info_unset
; xfs_infop
->flag
; xfs_infop
++) {
502 if (!(mp
->m_flags
& xfs_infop
->flag
))
503 seq_puts(m
, xfs_infop
->str
);
506 if (mp
->m_flags
& XFS_MOUNT_DFLT_IOSIZE
)
507 seq_printf(m
, ",allocsize=%dk",
508 (int)(1 << mp
->m_writeio_log
) >> 10);
510 if (mp
->m_logbufs
> 0)
511 seq_printf(m
, ",logbufs=%d", mp
->m_logbufs
);
512 if (mp
->m_logbsize
> 0)
513 seq_printf(m
, ",logbsize=%dk", mp
->m_logbsize
>> 10);
516 seq_show_option(m
, "logdev", mp
->m_logname
);
518 seq_show_option(m
, "rtdev", mp
->m_rtname
);
520 if (mp
->m_dalign
> 0)
521 seq_printf(m
, ",sunit=%d",
522 (int)XFS_FSB_TO_BB(mp
, mp
->m_dalign
));
523 if (mp
->m_swidth
> 0)
524 seq_printf(m
, ",swidth=%d",
525 (int)XFS_FSB_TO_BB(mp
, mp
->m_swidth
));
527 if (mp
->m_qflags
& (XFS_UQUOTA_ACCT
|XFS_UQUOTA_ENFD
))
528 seq_puts(m
, ",usrquota");
529 else if (mp
->m_qflags
& XFS_UQUOTA_ACCT
)
530 seq_puts(m
, ",uqnoenforce");
532 if (mp
->m_qflags
& XFS_PQUOTA_ACCT
) {
533 if (mp
->m_qflags
& XFS_PQUOTA_ENFD
)
534 seq_puts(m
, ",prjquota");
536 seq_puts(m
, ",pqnoenforce");
538 if (mp
->m_qflags
& XFS_GQUOTA_ACCT
) {
539 if (mp
->m_qflags
& XFS_GQUOTA_ENFD
)
540 seq_puts(m
, ",grpquota");
542 seq_puts(m
, ",gqnoenforce");
545 if (!(mp
->m_qflags
& XFS_ALL_QUOTA_ACCT
))
546 seq_puts(m
, ",noquota");
552 unsigned int blockshift
)
554 unsigned int pagefactor
= 1;
555 unsigned int bitshift
= BITS_PER_LONG
- 1;
557 /* Figure out maximum filesize, on Linux this can depend on
558 * the filesystem blocksize (on 32 bit platforms).
559 * __block_write_begin does this in an [unsigned] long...
560 * page->index << (PAGE_SHIFT - bbits)
561 * So, for page sized blocks (4K on 32 bit platforms),
562 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
563 * (((u64)PAGE_SIZE << (BITS_PER_LONG-1))-1)
564 * but for smaller blocksizes it is less (bbits = log2 bsize).
565 * Note1: get_block_t takes a long (implicit cast from above)
566 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
567 * can optionally convert the [unsigned] long from above into
568 * an [unsigned] long long.
571 #if BITS_PER_LONG == 32
572 # if defined(CONFIG_LBDAF)
573 ASSERT(sizeof(sector_t
) == 8);
574 pagefactor
= PAGE_SIZE
;
575 bitshift
= BITS_PER_LONG
;
577 pagefactor
= PAGE_SIZE
>> (PAGE_SHIFT
- blockshift
);
581 return (((__uint64_t
)pagefactor
) << bitshift
) - 1;
585 * Set parameters for inode allocation heuristics, taking into account
586 * filesystem size and inode32/inode64 mount options; i.e. specifically
587 * whether or not XFS_MOUNT_SMALL_INUMS is set.
589 * Inode allocation patterns are altered only if inode32 is requested
590 * (XFS_MOUNT_SMALL_INUMS), and the filesystem is sufficiently large.
591 * If altered, XFS_MOUNT_32BITINODES is set as well.
593 * An agcount independent of that in the mount structure is provided
594 * because in the growfs case, mp->m_sb.sb_agcount is not yet updated
595 * to the potentially higher ag count.
597 * Returns the maximum AG index which may contain inodes.
601 struct xfs_mount
*mp
,
602 xfs_agnumber_t agcount
)
604 xfs_agnumber_t index
;
605 xfs_agnumber_t maxagi
= 0;
606 xfs_sb_t
*sbp
= &mp
->m_sb
;
607 xfs_agnumber_t max_metadata
;
612 * Calculate how much should be reserved for inodes to meet
613 * the max inode percentage. Used only for inode32.
615 if (mp
->m_maxicount
) {
618 icount
= sbp
->sb_dblocks
* sbp
->sb_imax_pct
;
620 icount
+= sbp
->sb_agblocks
- 1;
621 do_div(icount
, sbp
->sb_agblocks
);
622 max_metadata
= icount
;
624 max_metadata
= agcount
;
627 /* Get the last possible inode in the filesystem */
628 agino
= XFS_OFFBNO_TO_AGINO(mp
, sbp
->sb_agblocks
- 1, 0);
629 ino
= XFS_AGINO_TO_INO(mp
, agcount
- 1, agino
);
632 * If user asked for no more than 32-bit inodes, and the fs is
633 * sufficiently large, set XFS_MOUNT_32BITINODES if we must alter
634 * the allocator to accommodate the request.
636 if ((mp
->m_flags
& XFS_MOUNT_SMALL_INUMS
) && ino
> XFS_MAXINUMBER_32
)
637 mp
->m_flags
|= XFS_MOUNT_32BITINODES
;
639 mp
->m_flags
&= ~XFS_MOUNT_32BITINODES
;
641 for (index
= 0; index
< agcount
; index
++) {
642 struct xfs_perag
*pag
;
644 ino
= XFS_AGINO_TO_INO(mp
, index
, agino
);
646 pag
= xfs_perag_get(mp
, index
);
648 if (mp
->m_flags
& XFS_MOUNT_32BITINODES
) {
649 if (ino
> XFS_MAXINUMBER_32
) {
650 pag
->pagi_inodeok
= 0;
651 pag
->pagf_metadata
= 0;
653 pag
->pagi_inodeok
= 1;
655 if (index
< max_metadata
)
656 pag
->pagf_metadata
= 1;
658 pag
->pagf_metadata
= 0;
661 pag
->pagi_inodeok
= 1;
662 pag
->pagf_metadata
= 0;
668 return (mp
->m_flags
& XFS_MOUNT_32BITINODES
) ? maxagi
: agcount
;
675 struct block_device
**bdevp
)
679 *bdevp
= blkdev_get_by_path(name
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
,
681 if (IS_ERR(*bdevp
)) {
682 error
= PTR_ERR(*bdevp
);
683 xfs_warn(mp
, "Invalid device [%s], error=%d", name
, error
);
691 struct block_device
*bdev
)
694 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
698 xfs_blkdev_issue_flush(
699 xfs_buftarg_t
*buftarg
)
701 blkdev_issue_flush(buftarg
->bt_bdev
, GFP_NOFS
, NULL
);
706 struct xfs_mount
*mp
)
708 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
709 struct block_device
*logdev
= mp
->m_logdev_targp
->bt_bdev
;
710 xfs_free_buftarg(mp
, mp
->m_logdev_targp
);
711 xfs_blkdev_put(logdev
);
713 if (mp
->m_rtdev_targp
) {
714 struct block_device
*rtdev
= mp
->m_rtdev_targp
->bt_bdev
;
715 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
716 xfs_blkdev_put(rtdev
);
718 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
722 * The file system configurations are:
723 * (1) device (partition) with data and internal log
724 * (2) logical volume with data and log subvolumes.
725 * (3) logical volume with data, log, and realtime subvolumes.
727 * We only have to handle opening the log and realtime volumes here if
728 * they are present. The data subvolume has already been opened by
729 * get_sb_bdev() and is stored in sb->s_bdev.
733 struct xfs_mount
*mp
)
735 struct block_device
*ddev
= mp
->m_super
->s_bdev
;
736 struct block_device
*logdev
= NULL
, *rtdev
= NULL
;
740 * Open real time and log devices - order is important.
743 error
= xfs_blkdev_get(mp
, mp
->m_logname
, &logdev
);
749 error
= xfs_blkdev_get(mp
, mp
->m_rtname
, &rtdev
);
751 goto out_close_logdev
;
753 if (rtdev
== ddev
|| rtdev
== logdev
) {
755 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
757 goto out_close_rtdev
;
762 * Setup xfs_mount buffer target pointers
765 mp
->m_ddev_targp
= xfs_alloc_buftarg(mp
, ddev
);
766 if (!mp
->m_ddev_targp
)
767 goto out_close_rtdev
;
770 mp
->m_rtdev_targp
= xfs_alloc_buftarg(mp
, rtdev
);
771 if (!mp
->m_rtdev_targp
)
772 goto out_free_ddev_targ
;
775 if (logdev
&& logdev
!= ddev
) {
776 mp
->m_logdev_targp
= xfs_alloc_buftarg(mp
, logdev
);
777 if (!mp
->m_logdev_targp
)
778 goto out_free_rtdev_targ
;
780 mp
->m_logdev_targp
= mp
->m_ddev_targp
;
786 if (mp
->m_rtdev_targp
)
787 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
789 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
791 xfs_blkdev_put(rtdev
);
793 if (logdev
&& logdev
!= ddev
)
794 xfs_blkdev_put(logdev
);
800 * Setup xfs_mount buffer target pointers based on superblock
804 struct xfs_mount
*mp
)
808 error
= xfs_setsize_buftarg(mp
->m_ddev_targp
, mp
->m_sb
.sb_sectsize
);
812 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
813 unsigned int log_sector_size
= BBSIZE
;
815 if (xfs_sb_version_hassector(&mp
->m_sb
))
816 log_sector_size
= mp
->m_sb
.sb_logsectsize
;
817 error
= xfs_setsize_buftarg(mp
->m_logdev_targp
,
822 if (mp
->m_rtdev_targp
) {
823 error
= xfs_setsize_buftarg(mp
->m_rtdev_targp
,
824 mp
->m_sb
.sb_sectsize
);
833 xfs_init_mount_workqueues(
834 struct xfs_mount
*mp
)
836 mp
->m_buf_workqueue
= alloc_workqueue("xfs-buf/%s",
837 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 1, mp
->m_fsname
);
838 if (!mp
->m_buf_workqueue
)
841 mp
->m_data_workqueue
= alloc_workqueue("xfs-data/%s",
842 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
843 if (!mp
->m_data_workqueue
)
844 goto out_destroy_buf
;
846 mp
->m_unwritten_workqueue
= alloc_workqueue("xfs-conv/%s",
847 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
848 if (!mp
->m_unwritten_workqueue
)
849 goto out_destroy_data_iodone_queue
;
851 mp
->m_cil_workqueue
= alloc_workqueue("xfs-cil/%s",
852 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
853 if (!mp
->m_cil_workqueue
)
854 goto out_destroy_unwritten
;
856 mp
->m_reclaim_workqueue
= alloc_workqueue("xfs-reclaim/%s",
857 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
858 if (!mp
->m_reclaim_workqueue
)
859 goto out_destroy_cil
;
861 mp
->m_log_workqueue
= alloc_workqueue("xfs-log/%s",
862 WQ_MEM_RECLAIM
|WQ_FREEZABLE
|WQ_HIGHPRI
, 0,
864 if (!mp
->m_log_workqueue
)
865 goto out_destroy_reclaim
;
867 mp
->m_eofblocks_workqueue
= alloc_workqueue("xfs-eofblocks/%s",
868 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
869 if (!mp
->m_eofblocks_workqueue
)
870 goto out_destroy_log
;
875 destroy_workqueue(mp
->m_log_workqueue
);
877 destroy_workqueue(mp
->m_reclaim_workqueue
);
879 destroy_workqueue(mp
->m_cil_workqueue
);
880 out_destroy_unwritten
:
881 destroy_workqueue(mp
->m_unwritten_workqueue
);
882 out_destroy_data_iodone_queue
:
883 destroy_workqueue(mp
->m_data_workqueue
);
885 destroy_workqueue(mp
->m_buf_workqueue
);
891 xfs_destroy_mount_workqueues(
892 struct xfs_mount
*mp
)
894 destroy_workqueue(mp
->m_eofblocks_workqueue
);
895 destroy_workqueue(mp
->m_log_workqueue
);
896 destroy_workqueue(mp
->m_reclaim_workqueue
);
897 destroy_workqueue(mp
->m_cil_workqueue
);
898 destroy_workqueue(mp
->m_data_workqueue
);
899 destroy_workqueue(mp
->m_unwritten_workqueue
);
900 destroy_workqueue(mp
->m_buf_workqueue
);
904 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
905 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
906 * for IO to complete so that we effectively throttle multiple callers to the
907 * rate at which IO is completing.
911 struct xfs_mount
*mp
)
913 struct super_block
*sb
= mp
->m_super
;
915 if (down_read_trylock(&sb
->s_umount
)) {
917 up_read(&sb
->s_umount
);
921 /* Catch misguided souls that try to use this interface on XFS */
922 STATIC
struct inode
*
924 struct super_block
*sb
)
931 * Now that the generic code is guaranteed not to be accessing
932 * the linux inode, we can inactivate and reclaim the inode.
935 xfs_fs_destroy_inode(
938 struct xfs_inode
*ip
= XFS_I(inode
);
940 trace_xfs_destroy_inode(ip
);
942 ASSERT(!rwsem_is_locked(&ip
->i_iolock
.mr_lock
));
943 XFS_STATS_INC(ip
->i_mount
, vn_rele
);
944 XFS_STATS_INC(ip
->i_mount
, vn_remove
);
948 ASSERT(XFS_FORCED_SHUTDOWN(ip
->i_mount
) || ip
->i_delayed_blks
== 0);
949 XFS_STATS_INC(ip
->i_mount
, vn_reclaim
);
952 * We should never get here with one of the reclaim flags already set.
954 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIMABLE
));
955 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIM
));
958 * We always use background reclaim here because even if the
959 * inode is clean, it still may be under IO and hence we have
960 * to take the flush lock. The background reclaim path handles
961 * this more efficiently than we can here, so simply let background
962 * reclaim tear down all inodes.
964 xfs_inode_set_reclaim_tag(ip
);
968 * Slab object creation initialisation for the XFS inode.
969 * This covers only the idempotent fields in the XFS inode;
970 * all other fields need to be initialised on allocation
971 * from the slab. This avoids the need to repeatedly initialise
972 * fields in the xfs inode that left in the initialise state
973 * when freeing the inode.
976 xfs_fs_inode_init_once(
979 struct xfs_inode
*ip
= inode
;
981 memset(ip
, 0, sizeof(struct xfs_inode
));
984 inode_init_once(VFS_I(ip
));
987 atomic_set(&ip
->i_pincount
, 0);
988 spin_lock_init(&ip
->i_flags_lock
);
990 mrlock_init(&ip
->i_mmaplock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
991 "xfsino", ip
->i_ino
);
992 mrlock_init(&ip
->i_lock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
993 "xfsino", ip
->i_ino
);
997 * We do an unlocked check for XFS_IDONTCACHE here because we are already
998 * serialised against cache hits here via the inode->i_lock and igrab() in
999 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1000 * racing with us, and it avoids needing to grab a spinlock here for every inode
1001 * we drop the final reference on.
1005 struct inode
*inode
)
1007 struct xfs_inode
*ip
= XFS_I(inode
);
1009 return generic_drop_inode(inode
) || (ip
->i_flags
& XFS_IDONTCACHE
);
1014 struct xfs_mount
*mp
)
1016 kfree(mp
->m_fsname
);
1017 kfree(mp
->m_rtname
);
1018 kfree(mp
->m_logname
);
1023 struct super_block
*sb
,
1026 struct xfs_mount
*mp
= XFS_M(sb
);
1029 * Doing anything during the async pass would be counterproductive.
1034 xfs_log_force(mp
, XFS_LOG_SYNC
);
1037 * The disk must be active because we're syncing.
1038 * We schedule log work now (now that the disk is
1039 * active) instead of later (when it might not be).
1041 flush_delayed_work(&mp
->m_log
->l_work
);
1049 struct dentry
*dentry
,
1050 struct kstatfs
*statp
)
1052 struct xfs_mount
*mp
= XFS_M(dentry
->d_sb
);
1053 xfs_sb_t
*sbp
= &mp
->m_sb
;
1054 struct xfs_inode
*ip
= XFS_I(d_inode(dentry
));
1055 __uint64_t fakeinos
, id
;
1058 __uint64_t fdblocks
;
1062 statp
->f_type
= XFS_SB_MAGIC
;
1063 statp
->f_namelen
= MAXNAMELEN
- 1;
1065 id
= huge_encode_dev(mp
->m_ddev_targp
->bt_dev
);
1066 statp
->f_fsid
.val
[0] = (u32
)id
;
1067 statp
->f_fsid
.val
[1] = (u32
)(id
>> 32);
1069 icount
= percpu_counter_sum(&mp
->m_icount
);
1070 ifree
= percpu_counter_sum(&mp
->m_ifree
);
1071 fdblocks
= percpu_counter_sum(&mp
->m_fdblocks
);
1073 spin_lock(&mp
->m_sb_lock
);
1074 statp
->f_bsize
= sbp
->sb_blocksize
;
1075 lsize
= sbp
->sb_logstart
? sbp
->sb_logblocks
: 0;
1076 statp
->f_blocks
= sbp
->sb_dblocks
- lsize
;
1077 spin_unlock(&mp
->m_sb_lock
);
1079 statp
->f_bfree
= fdblocks
- mp
->m_alloc_set_aside
;
1080 statp
->f_bavail
= statp
->f_bfree
;
1082 fakeinos
= statp
->f_bfree
<< sbp
->sb_inopblog
;
1083 statp
->f_files
= MIN(icount
+ fakeinos
, (__uint64_t
)XFS_MAXINUMBER
);
1084 if (mp
->m_maxicount
)
1085 statp
->f_files
= min_t(typeof(statp
->f_files
),
1089 /* If sb_icount overshot maxicount, report actual allocation */
1090 statp
->f_files
= max_t(typeof(statp
->f_files
),
1094 /* make sure statp->f_ffree does not underflow */
1095 ffree
= statp
->f_files
- (icount
- ifree
);
1096 statp
->f_ffree
= max_t(__int64_t
, ffree
, 0);
1099 if ((ip
->i_d
.di_flags
& XFS_DIFLAG_PROJINHERIT
) &&
1100 ((mp
->m_qflags
& (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))) ==
1101 (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))
1102 xfs_qm_statvfs(ip
, statp
);
1107 xfs_save_resvblks(struct xfs_mount
*mp
)
1109 __uint64_t resblks
= 0;
1111 mp
->m_resblks_save
= mp
->m_resblks
;
1112 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1116 xfs_restore_resvblks(struct xfs_mount
*mp
)
1120 if (mp
->m_resblks_save
) {
1121 resblks
= mp
->m_resblks_save
;
1122 mp
->m_resblks_save
= 0;
1124 resblks
= xfs_default_resblks(mp
);
1126 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1130 * Trigger writeback of all the dirty metadata in the file system.
1132 * This ensures that the metadata is written to their location on disk rather
1133 * than just existing in transactions in the log. This means after a quiesce
1134 * there is no log replay required to write the inodes to disk - this is the
1135 * primary difference between a sync and a quiesce.
1137 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1138 * it is started again when appropriate.
1142 struct xfs_mount
*mp
)
1146 /* wait for all modifications to complete */
1147 while (atomic_read(&mp
->m_active_trans
) > 0)
1150 /* force the log to unpin objects from the now complete transactions */
1151 xfs_log_force(mp
, XFS_LOG_SYNC
);
1153 /* reclaim inodes to do any IO before the freeze completes */
1154 xfs_reclaim_inodes(mp
, 0);
1155 xfs_reclaim_inodes(mp
, SYNC_WAIT
);
1157 /* Push the superblock and write an unmount record */
1158 error
= xfs_log_sbcount(mp
);
1160 xfs_warn(mp
, "xfs_attr_quiesce: failed to log sb changes. "
1161 "Frozen image may not be consistent.");
1163 * Just warn here till VFS can correctly support
1164 * read-only remount without racing.
1166 WARN_ON(atomic_read(&mp
->m_active_trans
) != 0);
1168 xfs_log_quiesce(mp
);
1172 xfs_test_remount_options(
1173 struct super_block
*sb
,
1174 struct xfs_mount
*mp
,
1178 struct xfs_mount
*tmp_mp
;
1180 tmp_mp
= kmem_zalloc(sizeof(*tmp_mp
), KM_MAYFAIL
);
1184 tmp_mp
->m_super
= sb
;
1185 error
= xfs_parseargs(tmp_mp
, options
);
1186 xfs_free_fsname(tmp_mp
);
1194 struct super_block
*sb
,
1198 struct xfs_mount
*mp
= XFS_M(sb
);
1199 xfs_sb_t
*sbp
= &mp
->m_sb
;
1200 substring_t args
[MAX_OPT_ARGS
];
1204 /* First, check for complete junk; i.e. invalid options */
1205 error
= xfs_test_remount_options(sb
, mp
, options
);
1209 sync_filesystem(sb
);
1210 while ((p
= strsep(&options
, ",")) != NULL
) {
1216 token
= match_token(p
, tokens
, args
);
1219 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
1222 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
1225 mp
->m_flags
&= ~XFS_MOUNT_SMALL_INUMS
;
1226 mp
->m_maxagi
= xfs_set_inode_alloc(mp
, sbp
->sb_agcount
);
1229 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
1230 mp
->m_maxagi
= xfs_set_inode_alloc(mp
, sbp
->sb_agcount
);
1234 * Logically we would return an error here to prevent
1235 * users from believing they might have changed
1236 * mount options using remount which can't be changed.
1238 * But unfortunately mount(8) adds all options from
1239 * mtab and fstab to the mount arguments in some cases
1240 * so we can't blindly reject options, but have to
1241 * check for each specified option if it actually
1242 * differs from the currently set option and only
1243 * reject it if that's the case.
1245 * Until that is implemented we return success for
1246 * every remount request, and silently ignore all
1247 * options that we can't actually change.
1251 "mount option \"%s\" not supported for remount", p
);
1260 if ((mp
->m_flags
& XFS_MOUNT_RDONLY
) && !(*flags
& MS_RDONLY
)) {
1261 if (mp
->m_flags
& XFS_MOUNT_NORECOVERY
) {
1263 "ro->rw transition prohibited on norecovery mount");
1267 if (XFS_SB_VERSION_NUM(sbp
) == XFS_SB_VERSION_5
&&
1268 xfs_sb_has_ro_compat_feature(sbp
,
1269 XFS_SB_FEAT_RO_COMPAT_UNKNOWN
)) {
1271 "ro->rw transition prohibited on unknown (0x%x) ro-compat filesystem",
1272 (sbp
->sb_features_ro_compat
&
1273 XFS_SB_FEAT_RO_COMPAT_UNKNOWN
));
1277 mp
->m_flags
&= ~XFS_MOUNT_RDONLY
;
1280 * If this is the first remount to writeable state we
1281 * might have some superblock changes to update.
1283 if (mp
->m_update_sb
) {
1284 error
= xfs_sync_sb(mp
, false);
1286 xfs_warn(mp
, "failed to write sb changes");
1289 mp
->m_update_sb
= false;
1293 * Fill out the reserve pool if it is empty. Use the stashed
1294 * value if it is non-zero, otherwise go with the default.
1296 xfs_restore_resvblks(mp
);
1297 xfs_log_work_queue(mp
);
1298 xfs_queue_eofblocks(mp
);
1302 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
) && (*flags
& MS_RDONLY
)) {
1304 * Before we sync the metadata, we need to free up the reserve
1305 * block pool so that the used block count in the superblock on
1306 * disk is correct at the end of the remount. Stash the current
1307 * reserve pool size so that if we get remounted rw, we can
1308 * return it to the same size.
1310 xfs_save_resvblks(mp
);
1313 * Cancel background eofb scanning so it cannot race with the
1314 * final log force+buftarg wait and deadlock the remount.
1316 cancel_delayed_work_sync(&mp
->m_eofblocks_work
);
1318 xfs_quiesce_attr(mp
);
1319 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
1326 * Second stage of a freeze. The data is already frozen so we only
1327 * need to take care of the metadata. Once that's done sync the superblock
1328 * to the log to dirty it in case of a crash while frozen. This ensures that we
1329 * will recover the unlinked inode lists on the next mount.
1333 struct super_block
*sb
)
1335 struct xfs_mount
*mp
= XFS_M(sb
);
1337 xfs_save_resvblks(mp
);
1338 xfs_quiesce_attr(mp
);
1339 return xfs_sync_sb(mp
, true);
1344 struct super_block
*sb
)
1346 struct xfs_mount
*mp
= XFS_M(sb
);
1348 xfs_restore_resvblks(mp
);
1349 xfs_log_work_queue(mp
);
1354 xfs_fs_show_options(
1356 struct dentry
*root
)
1358 return xfs_showargs(XFS_M(root
->d_sb
), m
);
1362 * This function fills in xfs_mount_t fields based on mount args.
1363 * Note: the superblock _has_ now been read in.
1367 struct xfs_mount
*mp
)
1369 int ronly
= (mp
->m_flags
& XFS_MOUNT_RDONLY
);
1371 /* Fail a mount where the logbuf is smaller than the log stripe */
1372 if (xfs_sb_version_haslogv2(&mp
->m_sb
)) {
1373 if (mp
->m_logbsize
<= 0 &&
1374 mp
->m_sb
.sb_logsunit
> XLOG_BIG_RECORD_BSIZE
) {
1375 mp
->m_logbsize
= mp
->m_sb
.sb_logsunit
;
1376 } else if (mp
->m_logbsize
> 0 &&
1377 mp
->m_logbsize
< mp
->m_sb
.sb_logsunit
) {
1379 "logbuf size must be greater than or equal to log stripe size");
1383 /* Fail a mount if the logbuf is larger than 32K */
1384 if (mp
->m_logbsize
> XLOG_BIG_RECORD_BSIZE
) {
1386 "logbuf size for version 1 logs must be 16K or 32K");
1392 * V5 filesystems always use attr2 format for attributes.
1394 if (xfs_sb_version_hascrc(&mp
->m_sb
) &&
1395 (mp
->m_flags
& XFS_MOUNT_NOATTR2
)) {
1396 xfs_warn(mp
, "Cannot mount a V5 filesystem as noattr2. "
1397 "attr2 is always enabled for V5 filesystems.");
1402 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1403 * told by noattr2 to turn it off
1405 if (xfs_sb_version_hasattr2(&mp
->m_sb
) &&
1406 !(mp
->m_flags
& XFS_MOUNT_NOATTR2
))
1407 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
1410 * prohibit r/w mounts of read-only filesystems
1412 if ((mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) && !ronly
) {
1414 "cannot mount a read-only filesystem as read-write");
1418 if ((mp
->m_qflags
& (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
)) &&
1419 (mp
->m_qflags
& (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
)) &&
1420 !xfs_sb_version_has_pquotino(&mp
->m_sb
)) {
1422 "Super block does not support project and group quota together");
1430 xfs_init_percpu_counters(
1431 struct xfs_mount
*mp
)
1435 error
= percpu_counter_init(&mp
->m_icount
, 0, GFP_KERNEL
);
1439 error
= percpu_counter_init(&mp
->m_ifree
, 0, GFP_KERNEL
);
1443 error
= percpu_counter_init(&mp
->m_fdblocks
, 0, GFP_KERNEL
);
1450 percpu_counter_destroy(&mp
->m_ifree
);
1452 percpu_counter_destroy(&mp
->m_icount
);
1457 xfs_reinit_percpu_counters(
1458 struct xfs_mount
*mp
)
1460 percpu_counter_set(&mp
->m_icount
, mp
->m_sb
.sb_icount
);
1461 percpu_counter_set(&mp
->m_ifree
, mp
->m_sb
.sb_ifree
);
1462 percpu_counter_set(&mp
->m_fdblocks
, mp
->m_sb
.sb_fdblocks
);
1466 xfs_destroy_percpu_counters(
1467 struct xfs_mount
*mp
)
1469 percpu_counter_destroy(&mp
->m_icount
);
1470 percpu_counter_destroy(&mp
->m_ifree
);
1471 percpu_counter_destroy(&mp
->m_fdblocks
);
1476 struct super_block
*sb
,
1481 struct xfs_mount
*mp
= NULL
;
1482 int flags
= 0, error
= -ENOMEM
;
1484 mp
= kzalloc(sizeof(struct xfs_mount
), GFP_KERNEL
);
1488 spin_lock_init(&mp
->m_sb_lock
);
1489 mutex_init(&mp
->m_growlock
);
1490 atomic_set(&mp
->m_active_trans
, 0);
1491 INIT_DELAYED_WORK(&mp
->m_reclaim_work
, xfs_reclaim_worker
);
1492 INIT_DELAYED_WORK(&mp
->m_eofblocks_work
, xfs_eofblocks_worker
);
1493 mp
->m_kobj
.kobject
.kset
= xfs_kset
;
1498 error
= xfs_parseargs(mp
, (char *)data
);
1500 goto out_free_fsname
;
1502 sb_min_blocksize(sb
, BBSIZE
);
1503 sb
->s_xattr
= xfs_xattr_handlers
;
1504 sb
->s_export_op
= &xfs_export_operations
;
1505 #ifdef CONFIG_XFS_QUOTA
1506 sb
->s_qcop
= &xfs_quotactl_operations
;
1507 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
1509 sb
->s_op
= &xfs_super_operations
;
1512 flags
|= XFS_MFSI_QUIET
;
1514 error
= xfs_open_devices(mp
);
1516 goto out_free_fsname
;
1518 error
= xfs_init_mount_workqueues(mp
);
1520 goto out_close_devices
;
1522 error
= xfs_init_percpu_counters(mp
);
1524 goto out_destroy_workqueues
;
1526 /* Allocate stats memory before we do operations that might use it */
1527 mp
->m_stats
.xs_stats
= alloc_percpu(struct xfsstats
);
1528 if (!mp
->m_stats
.xs_stats
) {
1530 goto out_destroy_counters
;
1533 error
= xfs_readsb(mp
, flags
);
1535 goto out_free_stats
;
1537 error
= xfs_finish_flags(mp
);
1541 error
= xfs_setup_devices(mp
);
1545 error
= xfs_filestream_mount(mp
);
1550 * we must configure the block size in the superblock before we run the
1551 * full mount process as the mount process can lookup and cache inodes.
1553 sb
->s_magic
= XFS_SB_MAGIC
;
1554 sb
->s_blocksize
= mp
->m_sb
.sb_blocksize
;
1555 sb
->s_blocksize_bits
= ffs(sb
->s_blocksize
) - 1;
1556 sb
->s_maxbytes
= xfs_max_file_offset(sb
->s_blocksize_bits
);
1557 sb
->s_max_links
= XFS_MAXLINK
;
1558 sb
->s_time_gran
= 1;
1559 set_posix_acl_flag(sb
);
1561 /* version 5 superblocks support inode version counters. */
1562 if (XFS_SB_VERSION_NUM(&mp
->m_sb
) == XFS_SB_VERSION_5
)
1563 sb
->s_flags
|= MS_I_VERSION
;
1565 if (mp
->m_flags
& XFS_MOUNT_DAX
) {
1567 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1569 error
= bdev_dax_supported(sb
, sb
->s_blocksize
);
1572 "DAX unsupported by block device. Turning off DAX.");
1573 mp
->m_flags
&= ~XFS_MOUNT_DAX
;
1577 if (xfs_sb_version_hasrmapbt(&mp
->m_sb
))
1579 "EXPERIMENTAL reverse mapping btree feature enabled. Use at your own risk!");
1581 error
= xfs_mountfs(mp
);
1583 goto out_filestream_unmount
;
1585 root
= igrab(VFS_I(mp
->m_rootip
));
1590 sb
->s_root
= d_make_root(root
);
1598 out_filestream_unmount
:
1599 xfs_filestream_unmount(mp
);
1603 free_percpu(mp
->m_stats
.xs_stats
);
1604 out_destroy_counters
:
1605 xfs_destroy_percpu_counters(mp
);
1606 out_destroy_workqueues
:
1607 xfs_destroy_mount_workqueues(mp
);
1609 xfs_close_devices(mp
);
1611 xfs_free_fsname(mp
);
1617 xfs_filestream_unmount(mp
);
1624 struct super_block
*sb
)
1626 struct xfs_mount
*mp
= XFS_M(sb
);
1628 xfs_notice(mp
, "Unmounting Filesystem");
1629 xfs_filestream_unmount(mp
);
1633 free_percpu(mp
->m_stats
.xs_stats
);
1634 xfs_destroy_percpu_counters(mp
);
1635 xfs_destroy_mount_workqueues(mp
);
1636 xfs_close_devices(mp
);
1637 xfs_free_fsname(mp
);
1641 STATIC
struct dentry
*
1643 struct file_system_type
*fs_type
,
1645 const char *dev_name
,
1648 return mount_bdev(fs_type
, flags
, dev_name
, data
, xfs_fs_fill_super
);
1652 xfs_fs_nr_cached_objects(
1653 struct super_block
*sb
,
1654 struct shrink_control
*sc
)
1656 return xfs_reclaim_inodes_count(XFS_M(sb
));
1660 xfs_fs_free_cached_objects(
1661 struct super_block
*sb
,
1662 struct shrink_control
*sc
)
1664 return xfs_reclaim_inodes_nr(XFS_M(sb
), sc
->nr_to_scan
);
1667 static const struct super_operations xfs_super_operations
= {
1668 .alloc_inode
= xfs_fs_alloc_inode
,
1669 .destroy_inode
= xfs_fs_destroy_inode
,
1670 .drop_inode
= xfs_fs_drop_inode
,
1671 .put_super
= xfs_fs_put_super
,
1672 .sync_fs
= xfs_fs_sync_fs
,
1673 .freeze_fs
= xfs_fs_freeze
,
1674 .unfreeze_fs
= xfs_fs_unfreeze
,
1675 .statfs
= xfs_fs_statfs
,
1676 .remount_fs
= xfs_fs_remount
,
1677 .show_options
= xfs_fs_show_options
,
1678 .nr_cached_objects
= xfs_fs_nr_cached_objects
,
1679 .free_cached_objects
= xfs_fs_free_cached_objects
,
1682 static struct file_system_type xfs_fs_type
= {
1683 .owner
= THIS_MODULE
,
1685 .mount
= xfs_fs_mount
,
1686 .kill_sb
= kill_block_super
,
1687 .fs_flags
= FS_REQUIRES_DEV
,
1689 MODULE_ALIAS_FS("xfs");
1692 xfs_init_zones(void)
1694 xfs_ioend_bioset
= bioset_create(4 * MAX_BUF_PER_PAGE
,
1695 offsetof(struct xfs_ioend
, io_inline_bio
));
1696 if (!xfs_ioend_bioset
)
1699 xfs_log_ticket_zone
= kmem_zone_init(sizeof(xlog_ticket_t
),
1701 if (!xfs_log_ticket_zone
)
1702 goto out_free_ioend_bioset
;
1704 xfs_bmap_free_item_zone
= kmem_zone_init(
1705 sizeof(struct xfs_extent_free_item
),
1706 "xfs_bmap_free_item");
1707 if (!xfs_bmap_free_item_zone
)
1708 goto out_destroy_log_ticket_zone
;
1710 xfs_btree_cur_zone
= kmem_zone_init(sizeof(xfs_btree_cur_t
),
1712 if (!xfs_btree_cur_zone
)
1713 goto out_destroy_bmap_free_item_zone
;
1715 xfs_da_state_zone
= kmem_zone_init(sizeof(xfs_da_state_t
),
1717 if (!xfs_da_state_zone
)
1718 goto out_destroy_btree_cur_zone
;
1720 xfs_ifork_zone
= kmem_zone_init(sizeof(xfs_ifork_t
), "xfs_ifork");
1721 if (!xfs_ifork_zone
)
1722 goto out_destroy_da_state_zone
;
1724 xfs_trans_zone
= kmem_zone_init(sizeof(xfs_trans_t
), "xfs_trans");
1725 if (!xfs_trans_zone
)
1726 goto out_destroy_ifork_zone
;
1728 xfs_log_item_desc_zone
=
1729 kmem_zone_init(sizeof(struct xfs_log_item_desc
),
1730 "xfs_log_item_desc");
1731 if (!xfs_log_item_desc_zone
)
1732 goto out_destroy_trans_zone
;
1735 * The size of the zone allocated buf log item is the maximum
1736 * size possible under XFS. This wastes a little bit of memory,
1737 * but it is much faster.
1739 xfs_buf_item_zone
= kmem_zone_init(sizeof(struct xfs_buf_log_item
),
1741 if (!xfs_buf_item_zone
)
1742 goto out_destroy_log_item_desc_zone
;
1744 xfs_efd_zone
= kmem_zone_init((sizeof(xfs_efd_log_item_t
) +
1745 ((XFS_EFD_MAX_FAST_EXTENTS
- 1) *
1746 sizeof(xfs_extent_t
))), "xfs_efd_item");
1748 goto out_destroy_buf_item_zone
;
1750 xfs_efi_zone
= kmem_zone_init((sizeof(xfs_efi_log_item_t
) +
1751 ((XFS_EFI_MAX_FAST_EXTENTS
- 1) *
1752 sizeof(xfs_extent_t
))), "xfs_efi_item");
1754 goto out_destroy_efd_zone
;
1757 kmem_zone_init_flags(sizeof(xfs_inode_t
), "xfs_inode",
1758 KM_ZONE_HWALIGN
| KM_ZONE_RECLAIM
| KM_ZONE_SPREAD
|
1759 KM_ZONE_ACCOUNT
, xfs_fs_inode_init_once
);
1760 if (!xfs_inode_zone
)
1761 goto out_destroy_efi_zone
;
1764 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t
), "xfs_ili",
1765 KM_ZONE_SPREAD
, NULL
);
1767 goto out_destroy_inode_zone
;
1768 xfs_icreate_zone
= kmem_zone_init(sizeof(struct xfs_icreate_item
),
1770 if (!xfs_icreate_zone
)
1771 goto out_destroy_ili_zone
;
1773 xfs_rud_zone
= kmem_zone_init(sizeof(struct xfs_rud_log_item
),
1776 goto out_destroy_icreate_zone
;
1778 xfs_rui_zone
= kmem_zone_init((sizeof(struct xfs_rui_log_item
) +
1779 ((XFS_RUI_MAX_FAST_EXTENTS
- 1) *
1780 sizeof(struct xfs_map_extent
))),
1783 goto out_destroy_rud_zone
;
1787 out_destroy_rud_zone
:
1788 kmem_zone_destroy(xfs_rud_zone
);
1789 out_destroy_icreate_zone
:
1790 kmem_zone_destroy(xfs_icreate_zone
);
1791 out_destroy_ili_zone
:
1792 kmem_zone_destroy(xfs_ili_zone
);
1793 out_destroy_inode_zone
:
1794 kmem_zone_destroy(xfs_inode_zone
);
1795 out_destroy_efi_zone
:
1796 kmem_zone_destroy(xfs_efi_zone
);
1797 out_destroy_efd_zone
:
1798 kmem_zone_destroy(xfs_efd_zone
);
1799 out_destroy_buf_item_zone
:
1800 kmem_zone_destroy(xfs_buf_item_zone
);
1801 out_destroy_log_item_desc_zone
:
1802 kmem_zone_destroy(xfs_log_item_desc_zone
);
1803 out_destroy_trans_zone
:
1804 kmem_zone_destroy(xfs_trans_zone
);
1805 out_destroy_ifork_zone
:
1806 kmem_zone_destroy(xfs_ifork_zone
);
1807 out_destroy_da_state_zone
:
1808 kmem_zone_destroy(xfs_da_state_zone
);
1809 out_destroy_btree_cur_zone
:
1810 kmem_zone_destroy(xfs_btree_cur_zone
);
1811 out_destroy_bmap_free_item_zone
:
1812 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1813 out_destroy_log_ticket_zone
:
1814 kmem_zone_destroy(xfs_log_ticket_zone
);
1815 out_free_ioend_bioset
:
1816 bioset_free(xfs_ioend_bioset
);
1822 xfs_destroy_zones(void)
1825 * Make sure all delayed rcu free are flushed before we
1829 kmem_zone_destroy(xfs_rui_zone
);
1830 kmem_zone_destroy(xfs_rud_zone
);
1831 kmem_zone_destroy(xfs_icreate_zone
);
1832 kmem_zone_destroy(xfs_ili_zone
);
1833 kmem_zone_destroy(xfs_inode_zone
);
1834 kmem_zone_destroy(xfs_efi_zone
);
1835 kmem_zone_destroy(xfs_efd_zone
);
1836 kmem_zone_destroy(xfs_buf_item_zone
);
1837 kmem_zone_destroy(xfs_log_item_desc_zone
);
1838 kmem_zone_destroy(xfs_trans_zone
);
1839 kmem_zone_destroy(xfs_ifork_zone
);
1840 kmem_zone_destroy(xfs_da_state_zone
);
1841 kmem_zone_destroy(xfs_btree_cur_zone
);
1842 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1843 kmem_zone_destroy(xfs_log_ticket_zone
);
1844 bioset_free(xfs_ioend_bioset
);
1848 xfs_init_workqueues(void)
1851 * The allocation workqueue can be used in memory reclaim situations
1852 * (writepage path), and parallelism is only limited by the number of
1853 * AGs in all the filesystems mounted. Hence use the default large
1854 * max_active value for this workqueue.
1856 xfs_alloc_wq
= alloc_workqueue("xfsalloc",
1857 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0);
1865 xfs_destroy_workqueues(void)
1867 destroy_workqueue(xfs_alloc_wq
);
1875 xfs_check_ondisk_structs();
1877 printk(KERN_INFO XFS_VERSION_STRING
" with "
1878 XFS_BUILD_OPTIONS
" enabled\n");
1880 xfs_extent_free_init_defer_op();
1881 xfs_rmap_update_init_defer_op();
1885 error
= xfs_init_zones();
1889 error
= xfs_init_workqueues();
1891 goto out_destroy_zones
;
1893 error
= xfs_mru_cache_init();
1895 goto out_destroy_wq
;
1897 error
= xfs_buf_init();
1899 goto out_mru_cache_uninit
;
1901 error
= xfs_init_procfs();
1903 goto out_buf_terminate
;
1905 error
= xfs_sysctl_register();
1907 goto out_cleanup_procfs
;
1909 xfs_kset
= kset_create_and_add("xfs", NULL
, fs_kobj
);
1912 goto out_sysctl_unregister
;
1915 xfsstats
.xs_kobj
.kobject
.kset
= xfs_kset
;
1917 xfsstats
.xs_stats
= alloc_percpu(struct xfsstats
);
1918 if (!xfsstats
.xs_stats
) {
1920 goto out_kset_unregister
;
1923 error
= xfs_sysfs_init(&xfsstats
.xs_kobj
, &xfs_stats_ktype
, NULL
,
1926 goto out_free_stats
;
1929 xfs_dbg_kobj
.kobject
.kset
= xfs_kset
;
1930 error
= xfs_sysfs_init(&xfs_dbg_kobj
, &xfs_dbg_ktype
, NULL
, "debug");
1932 goto out_remove_stats_kobj
;
1935 error
= xfs_qm_init();
1937 goto out_remove_dbg_kobj
;
1939 error
= register_filesystem(&xfs_fs_type
);
1946 out_remove_dbg_kobj
:
1948 xfs_sysfs_del(&xfs_dbg_kobj
);
1949 out_remove_stats_kobj
:
1951 xfs_sysfs_del(&xfsstats
.xs_kobj
);
1953 free_percpu(xfsstats
.xs_stats
);
1954 out_kset_unregister
:
1955 kset_unregister(xfs_kset
);
1956 out_sysctl_unregister
:
1957 xfs_sysctl_unregister();
1959 xfs_cleanup_procfs();
1961 xfs_buf_terminate();
1962 out_mru_cache_uninit
:
1963 xfs_mru_cache_uninit();
1965 xfs_destroy_workqueues();
1967 xfs_destroy_zones();
1976 unregister_filesystem(&xfs_fs_type
);
1978 xfs_sysfs_del(&xfs_dbg_kobj
);
1980 xfs_sysfs_del(&xfsstats
.xs_kobj
);
1981 free_percpu(xfsstats
.xs_stats
);
1982 kset_unregister(xfs_kset
);
1983 xfs_sysctl_unregister();
1984 xfs_cleanup_procfs();
1985 xfs_buf_terminate();
1986 xfs_mru_cache_uninit();
1987 xfs_destroy_workqueues();
1988 xfs_destroy_zones();
1989 xfs_uuid_table_free();
1992 module_init(init_xfs_fs
);
1993 module_exit(exit_xfs_fs
);
1995 MODULE_AUTHOR("Silicon Graphics, Inc.");
1996 MODULE_DESCRIPTION(XFS_VERSION_STRING
" with " XFS_BUILD_OPTIONS
" enabled");
1997 MODULE_LICENSE("GPL");