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"
50 #include "xfs_refcount_item.h"
51 #include "xfs_bmap_item.h"
52 #include "xfs_reflink.h"
54 #include <linux/namei.h>
55 #include <linux/dax.h>
56 #include <linux/init.h>
57 #include <linux/slab.h>
58 #include <linux/mount.h>
59 #include <linux/mempool.h>
60 #include <linux/writeback.h>
61 #include <linux/kthread.h>
62 #include <linux/freezer.h>
63 #include <linux/parser.h>
65 static const struct super_operations xfs_super_operations
;
66 struct bio_set
*xfs_ioend_bioset
;
68 static struct kset
*xfs_kset
; /* top-level xfs sysfs dir */
70 static struct xfs_kobj xfs_dbg_kobj
; /* global debug sysfs attrs */
74 * Table driven mount option parser.
77 Opt_logbufs
, Opt_logbsize
, Opt_logdev
, Opt_rtdev
, Opt_biosize
,
78 Opt_wsync
, Opt_noalign
, Opt_swalloc
, Opt_sunit
, Opt_swidth
, Opt_nouuid
,
79 Opt_mtpt
, Opt_grpid
, Opt_nogrpid
, Opt_bsdgroups
, Opt_sysvgroups
,
80 Opt_allocsize
, Opt_norecovery
, Opt_barrier
, Opt_nobarrier
,
81 Opt_inode64
, Opt_inode32
, Opt_ikeep
, Opt_noikeep
,
82 Opt_largeio
, Opt_nolargeio
, Opt_attr2
, Opt_noattr2
, Opt_filestreams
,
83 Opt_quota
, Opt_noquota
, Opt_usrquota
, Opt_grpquota
, Opt_prjquota
,
84 Opt_uquota
, Opt_gquota
, Opt_pquota
,
85 Opt_uqnoenforce
, Opt_gqnoenforce
, Opt_pqnoenforce
, Opt_qnoenforce
,
86 Opt_discard
, Opt_nodiscard
, Opt_dax
, Opt_err
,
89 static const match_table_t tokens
= {
90 {Opt_logbufs
, "logbufs=%u"}, /* number of XFS log buffers */
91 {Opt_logbsize
, "logbsize=%s"}, /* size of XFS log buffers */
92 {Opt_logdev
, "logdev=%s"}, /* log device */
93 {Opt_rtdev
, "rtdev=%s"}, /* realtime I/O device */
94 {Opt_biosize
, "biosize=%u"}, /* log2 of preferred buffered io size */
95 {Opt_wsync
, "wsync"}, /* safe-mode nfs compatible mount */
96 {Opt_noalign
, "noalign"}, /* turn off stripe alignment */
97 {Opt_swalloc
, "swalloc"}, /* turn on stripe width allocation */
98 {Opt_sunit
, "sunit=%u"}, /* data volume stripe unit */
99 {Opt_swidth
, "swidth=%u"}, /* data volume stripe width */
100 {Opt_nouuid
, "nouuid"}, /* ignore filesystem UUID */
101 {Opt_mtpt
, "mtpt"}, /* filesystem mount point */
102 {Opt_grpid
, "grpid"}, /* group-ID from parent directory */
103 {Opt_nogrpid
, "nogrpid"}, /* group-ID from current process */
104 {Opt_bsdgroups
, "bsdgroups"}, /* group-ID from parent directory */
105 {Opt_sysvgroups
,"sysvgroups"}, /* group-ID from current process */
106 {Opt_allocsize
, "allocsize=%s"},/* preferred allocation size */
107 {Opt_norecovery
,"norecovery"}, /* don't run XFS recovery */
108 {Opt_inode64
, "inode64"}, /* inodes can be allocated anywhere */
109 {Opt_inode32
, "inode32"}, /* inode allocation limited to
110 * XFS_MAXINUMBER_32 */
111 {Opt_ikeep
, "ikeep"}, /* do not free empty inode clusters */
112 {Opt_noikeep
, "noikeep"}, /* free empty inode clusters */
113 {Opt_largeio
, "largeio"}, /* report large I/O sizes in stat() */
114 {Opt_nolargeio
, "nolargeio"}, /* do not report large I/O sizes
116 {Opt_attr2
, "attr2"}, /* do use attr2 attribute format */
117 {Opt_noattr2
, "noattr2"}, /* do not use attr2 attribute format */
118 {Opt_filestreams
,"filestreams"},/* use filestreams allocator */
119 {Opt_quota
, "quota"}, /* disk quotas (user) */
120 {Opt_noquota
, "noquota"}, /* no quotas */
121 {Opt_usrquota
, "usrquota"}, /* user quota enabled */
122 {Opt_grpquota
, "grpquota"}, /* group quota enabled */
123 {Opt_prjquota
, "prjquota"}, /* project quota enabled */
124 {Opt_uquota
, "uquota"}, /* user quota (IRIX variant) */
125 {Opt_gquota
, "gquota"}, /* group quota (IRIX variant) */
126 {Opt_pquota
, "pquota"}, /* project quota (IRIX variant) */
127 {Opt_uqnoenforce
,"uqnoenforce"},/* user quota limit enforcement */
128 {Opt_gqnoenforce
,"gqnoenforce"},/* group quota limit enforcement */
129 {Opt_pqnoenforce
,"pqnoenforce"},/* project quota limit enforcement */
130 {Opt_qnoenforce
, "qnoenforce"}, /* same as uqnoenforce */
131 {Opt_discard
, "discard"}, /* Discard unused blocks */
132 {Opt_nodiscard
, "nodiscard"}, /* Do not discard unused blocks */
134 {Opt_dax
, "dax"}, /* Enable direct access to bdev pages */
136 /* Deprecated mount options scheduled for removal */
137 {Opt_barrier
, "barrier"}, /* use writer barriers for log write and
138 * unwritten extent conversion */
139 {Opt_nobarrier
, "nobarrier"}, /* .. disable */
146 suffix_kstrtoint(const substring_t
*s
, unsigned int base
, int *res
)
148 int last
, shift_left_factor
= 0, _res
;
152 value
= match_strdup(s
);
156 last
= strlen(value
) - 1;
157 if (value
[last
] == 'K' || value
[last
] == 'k') {
158 shift_left_factor
= 10;
161 if (value
[last
] == 'M' || value
[last
] == 'm') {
162 shift_left_factor
= 20;
165 if (value
[last
] == 'G' || value
[last
] == 'g') {
166 shift_left_factor
= 30;
170 if (kstrtoint(value
, base
, &_res
))
173 *res
= _res
<< shift_left_factor
;
178 * This function fills in xfs_mount_t fields based on mount args.
179 * Note: the superblock has _not_ yet been read in.
181 * Note that this function leaks the various device name allocations on
182 * failure. The caller takes care of them.
184 * *sb is const because this is also used to test options on the remount
185 * path, and we don't want this to have any side effects at remount time.
186 * Today this function does not change *sb, but just to future-proof...
190 struct xfs_mount
*mp
,
193 const struct super_block
*sb
= mp
->m_super
;
195 substring_t args
[MAX_OPT_ARGS
];
199 uint8_t iosizelog
= 0;
202 * set up the mount name first so all the errors will refer to the
205 mp
->m_fsname
= kstrndup(sb
->s_id
, MAXNAMELEN
, GFP_KERNEL
);
208 mp
->m_fsname_len
= strlen(mp
->m_fsname
) + 1;
211 * Copy binary VFS mount flags we are interested in.
214 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
215 if (sb
->s_flags
& MS_DIRSYNC
)
216 mp
->m_flags
|= XFS_MOUNT_DIRSYNC
;
217 if (sb
->s_flags
& MS_SYNCHRONOUS
)
218 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
221 * Set some default flags that could be cleared by the mount option
224 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
225 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
228 * These can be overridden by the mount option parsing.
236 while ((p
= strsep(&options
, ",")) != NULL
) {
242 token
= match_token(p
, tokens
, args
);
245 if (match_int(args
, &mp
->m_logbufs
))
249 if (suffix_kstrtoint(args
, 10, &mp
->m_logbsize
))
253 mp
->m_logname
= match_strdup(args
);
258 xfs_warn(mp
, "%s option not allowed on this system", p
);
261 mp
->m_rtname
= match_strdup(args
);
267 if (suffix_kstrtoint(args
, 10, &iosize
))
269 iosizelog
= ffs(iosize
) - 1;
273 mp
->m_flags
|= XFS_MOUNT_GRPID
;
277 mp
->m_flags
&= ~XFS_MOUNT_GRPID
;
280 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
283 mp
->m_flags
|= XFS_MOUNT_NORECOVERY
;
286 mp
->m_flags
|= XFS_MOUNT_NOALIGN
;
289 mp
->m_flags
|= XFS_MOUNT_SWALLOC
;
292 if (match_int(args
, &dsunit
))
296 if (match_int(args
, &dswidth
))
300 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
303 mp
->m_flags
&= ~XFS_MOUNT_SMALL_INUMS
;
306 mp
->m_flags
|= XFS_MOUNT_NOUUID
;
309 mp
->m_flags
|= XFS_MOUNT_IKEEP
;
312 mp
->m_flags
&= ~XFS_MOUNT_IKEEP
;
315 mp
->m_flags
&= ~XFS_MOUNT_COMPAT_IOSIZE
;
318 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
321 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
324 mp
->m_flags
&= ~XFS_MOUNT_ATTR2
;
325 mp
->m_flags
|= XFS_MOUNT_NOATTR2
;
327 case Opt_filestreams
:
328 mp
->m_flags
|= XFS_MOUNT_FILESTREAMS
;
331 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACCT
;
332 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ENFD
;
333 mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACTIVE
;
338 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
|
342 case Opt_uqnoenforce
:
343 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
);
344 mp
->m_qflags
&= ~XFS_UQUOTA_ENFD
;
348 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
|
351 case Opt_pqnoenforce
:
352 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
);
353 mp
->m_qflags
&= ~XFS_PQUOTA_ENFD
;
357 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
|
360 case Opt_gqnoenforce
:
361 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
);
362 mp
->m_qflags
&= ~XFS_GQUOTA_ENFD
;
365 mp
->m_flags
|= XFS_MOUNT_DISCARD
;
368 mp
->m_flags
&= ~XFS_MOUNT_DISCARD
;
372 mp
->m_flags
|= XFS_MOUNT_DAX
;
376 xfs_warn(mp
, "%s option is deprecated, ignoring.", p
);
377 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
380 xfs_warn(mp
, "%s option is deprecated, ignoring.", p
);
381 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
384 xfs_warn(mp
, "unknown mount option [%s].", p
);
390 * no recovery flag requires a read-only mount
392 if ((mp
->m_flags
& XFS_MOUNT_NORECOVERY
) &&
393 !(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
394 xfs_warn(mp
, "no-recovery mounts must be read-only.");
398 if ((mp
->m_flags
& XFS_MOUNT_NOALIGN
) && (dsunit
|| dswidth
)) {
400 "sunit and swidth options incompatible with the noalign option");
404 #ifndef CONFIG_XFS_QUOTA
405 if (XFS_IS_QUOTA_RUNNING(mp
)) {
406 xfs_warn(mp
, "quota support not available in this kernel.");
411 if ((dsunit
&& !dswidth
) || (!dsunit
&& dswidth
)) {
412 xfs_warn(mp
, "sunit and swidth must be specified together");
416 if (dsunit
&& (dswidth
% dsunit
!= 0)) {
418 "stripe width (%d) must be a multiple of the stripe unit (%d)",
424 if (dsunit
&& !(mp
->m_flags
& XFS_MOUNT_NOALIGN
)) {
426 * At this point the superblock has not been read
427 * in, therefore we do not know the block size.
428 * Before the mount call ends we will convert
431 mp
->m_dalign
= dsunit
;
432 mp
->m_swidth
= dswidth
;
435 if (mp
->m_logbufs
!= -1 &&
436 mp
->m_logbufs
!= 0 &&
437 (mp
->m_logbufs
< XLOG_MIN_ICLOGS
||
438 mp
->m_logbufs
> XLOG_MAX_ICLOGS
)) {
439 xfs_warn(mp
, "invalid logbufs value: %d [not %d-%d]",
440 mp
->m_logbufs
, XLOG_MIN_ICLOGS
, XLOG_MAX_ICLOGS
);
443 if (mp
->m_logbsize
!= -1 &&
444 mp
->m_logbsize
!= 0 &&
445 (mp
->m_logbsize
< XLOG_MIN_RECORD_BSIZE
||
446 mp
->m_logbsize
> XLOG_MAX_RECORD_BSIZE
||
447 !is_power_of_2(mp
->m_logbsize
))) {
449 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
455 if (iosizelog
> XFS_MAX_IO_LOG
||
456 iosizelog
< XFS_MIN_IO_LOG
) {
457 xfs_warn(mp
, "invalid log iosize: %d [not %d-%d]",
458 iosizelog
, XFS_MIN_IO_LOG
,
463 mp
->m_flags
|= XFS_MOUNT_DFLT_IOSIZE
;
464 mp
->m_readio_log
= iosizelog
;
465 mp
->m_writeio_log
= iosizelog
;
471 struct proc_xfs_info
{
478 struct xfs_mount
*mp
,
481 static struct proc_xfs_info xfs_info_set
[] = {
482 /* the few simple ones we can get from the mount struct */
483 { XFS_MOUNT_IKEEP
, ",ikeep" },
484 { XFS_MOUNT_WSYNC
, ",wsync" },
485 { XFS_MOUNT_NOALIGN
, ",noalign" },
486 { XFS_MOUNT_SWALLOC
, ",swalloc" },
487 { XFS_MOUNT_NOUUID
, ",nouuid" },
488 { XFS_MOUNT_NORECOVERY
, ",norecovery" },
489 { XFS_MOUNT_ATTR2
, ",attr2" },
490 { XFS_MOUNT_FILESTREAMS
, ",filestreams" },
491 { XFS_MOUNT_GRPID
, ",grpid" },
492 { XFS_MOUNT_DISCARD
, ",discard" },
493 { XFS_MOUNT_SMALL_INUMS
, ",inode32" },
494 { XFS_MOUNT_DAX
, ",dax" },
497 static struct proc_xfs_info xfs_info_unset
[] = {
498 /* the few simple ones we can get from the mount struct */
499 { XFS_MOUNT_COMPAT_IOSIZE
, ",largeio" },
500 { XFS_MOUNT_BARRIER
, ",nobarrier" },
501 { XFS_MOUNT_SMALL_INUMS
, ",inode64" },
504 struct proc_xfs_info
*xfs_infop
;
506 for (xfs_infop
= xfs_info_set
; xfs_infop
->flag
; xfs_infop
++) {
507 if (mp
->m_flags
& xfs_infop
->flag
)
508 seq_puts(m
, xfs_infop
->str
);
510 for (xfs_infop
= xfs_info_unset
; xfs_infop
->flag
; xfs_infop
++) {
511 if (!(mp
->m_flags
& xfs_infop
->flag
))
512 seq_puts(m
, xfs_infop
->str
);
515 if (mp
->m_flags
& XFS_MOUNT_DFLT_IOSIZE
)
516 seq_printf(m
, ",allocsize=%dk",
517 (int)(1 << mp
->m_writeio_log
) >> 10);
519 if (mp
->m_logbufs
> 0)
520 seq_printf(m
, ",logbufs=%d", mp
->m_logbufs
);
521 if (mp
->m_logbsize
> 0)
522 seq_printf(m
, ",logbsize=%dk", mp
->m_logbsize
>> 10);
525 seq_show_option(m
, "logdev", mp
->m_logname
);
527 seq_show_option(m
, "rtdev", mp
->m_rtname
);
529 if (mp
->m_dalign
> 0)
530 seq_printf(m
, ",sunit=%d",
531 (int)XFS_FSB_TO_BB(mp
, mp
->m_dalign
));
532 if (mp
->m_swidth
> 0)
533 seq_printf(m
, ",swidth=%d",
534 (int)XFS_FSB_TO_BB(mp
, mp
->m_swidth
));
536 if (mp
->m_qflags
& (XFS_UQUOTA_ACCT
|XFS_UQUOTA_ENFD
))
537 seq_puts(m
, ",usrquota");
538 else if (mp
->m_qflags
& XFS_UQUOTA_ACCT
)
539 seq_puts(m
, ",uqnoenforce");
541 if (mp
->m_qflags
& XFS_PQUOTA_ACCT
) {
542 if (mp
->m_qflags
& XFS_PQUOTA_ENFD
)
543 seq_puts(m
, ",prjquota");
545 seq_puts(m
, ",pqnoenforce");
547 if (mp
->m_qflags
& XFS_GQUOTA_ACCT
) {
548 if (mp
->m_qflags
& XFS_GQUOTA_ENFD
)
549 seq_puts(m
, ",grpquota");
551 seq_puts(m
, ",gqnoenforce");
554 if (!(mp
->m_qflags
& XFS_ALL_QUOTA_ACCT
))
555 seq_puts(m
, ",noquota");
561 unsigned int blockshift
)
563 unsigned int pagefactor
= 1;
564 unsigned int bitshift
= BITS_PER_LONG
- 1;
566 /* Figure out maximum filesize, on Linux this can depend on
567 * the filesystem blocksize (on 32 bit platforms).
568 * __block_write_begin does this in an [unsigned] long...
569 * page->index << (PAGE_SHIFT - bbits)
570 * So, for page sized blocks (4K on 32 bit platforms),
571 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
572 * (((u64)PAGE_SIZE << (BITS_PER_LONG-1))-1)
573 * but for smaller blocksizes it is less (bbits = log2 bsize).
574 * Note1: get_block_t takes a long (implicit cast from above)
575 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
576 * can optionally convert the [unsigned] long from above into
577 * an [unsigned] long long.
580 #if BITS_PER_LONG == 32
581 # if defined(CONFIG_LBDAF)
582 ASSERT(sizeof(sector_t
) == 8);
583 pagefactor
= PAGE_SIZE
;
584 bitshift
= BITS_PER_LONG
;
586 pagefactor
= PAGE_SIZE
>> (PAGE_SHIFT
- blockshift
);
590 return (((uint64_t)pagefactor
) << bitshift
) - 1;
594 * Set parameters for inode allocation heuristics, taking into account
595 * filesystem size and inode32/inode64 mount options; i.e. specifically
596 * whether or not XFS_MOUNT_SMALL_INUMS is set.
598 * Inode allocation patterns are altered only if inode32 is requested
599 * (XFS_MOUNT_SMALL_INUMS), and the filesystem is sufficiently large.
600 * If altered, XFS_MOUNT_32BITINODES is set as well.
602 * An agcount independent of that in the mount structure is provided
603 * because in the growfs case, mp->m_sb.sb_agcount is not yet updated
604 * to the potentially higher ag count.
606 * Returns the maximum AG index which may contain inodes.
610 struct xfs_mount
*mp
,
611 xfs_agnumber_t agcount
)
613 xfs_agnumber_t index
;
614 xfs_agnumber_t maxagi
= 0;
615 xfs_sb_t
*sbp
= &mp
->m_sb
;
616 xfs_agnumber_t max_metadata
;
621 * Calculate how much should be reserved for inodes to meet
622 * the max inode percentage. Used only for inode32.
624 if (mp
->m_maxicount
) {
627 icount
= sbp
->sb_dblocks
* sbp
->sb_imax_pct
;
629 icount
+= sbp
->sb_agblocks
- 1;
630 do_div(icount
, sbp
->sb_agblocks
);
631 max_metadata
= icount
;
633 max_metadata
= agcount
;
636 /* Get the last possible inode in the filesystem */
637 agino
= XFS_OFFBNO_TO_AGINO(mp
, sbp
->sb_agblocks
- 1, 0);
638 ino
= XFS_AGINO_TO_INO(mp
, agcount
- 1, agino
);
641 * If user asked for no more than 32-bit inodes, and the fs is
642 * sufficiently large, set XFS_MOUNT_32BITINODES if we must alter
643 * the allocator to accommodate the request.
645 if ((mp
->m_flags
& XFS_MOUNT_SMALL_INUMS
) && ino
> XFS_MAXINUMBER_32
)
646 mp
->m_flags
|= XFS_MOUNT_32BITINODES
;
648 mp
->m_flags
&= ~XFS_MOUNT_32BITINODES
;
650 for (index
= 0; index
< agcount
; index
++) {
651 struct xfs_perag
*pag
;
653 ino
= XFS_AGINO_TO_INO(mp
, index
, agino
);
655 pag
= xfs_perag_get(mp
, index
);
657 if (mp
->m_flags
& XFS_MOUNT_32BITINODES
) {
658 if (ino
> XFS_MAXINUMBER_32
) {
659 pag
->pagi_inodeok
= 0;
660 pag
->pagf_metadata
= 0;
662 pag
->pagi_inodeok
= 1;
664 if (index
< max_metadata
)
665 pag
->pagf_metadata
= 1;
667 pag
->pagf_metadata
= 0;
670 pag
->pagi_inodeok
= 1;
671 pag
->pagf_metadata
= 0;
677 return (mp
->m_flags
& XFS_MOUNT_32BITINODES
) ? maxagi
: agcount
;
684 struct block_device
**bdevp
)
688 *bdevp
= blkdev_get_by_path(name
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
,
690 if (IS_ERR(*bdevp
)) {
691 error
= PTR_ERR(*bdevp
);
692 xfs_warn(mp
, "Invalid device [%s], error=%d", name
, error
);
700 struct block_device
*bdev
)
703 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
|FMODE_EXCL
);
707 xfs_blkdev_issue_flush(
708 xfs_buftarg_t
*buftarg
)
710 blkdev_issue_flush(buftarg
->bt_bdev
, GFP_NOFS
, NULL
);
715 struct xfs_mount
*mp
)
717 struct dax_device
*dax_ddev
= mp
->m_ddev_targp
->bt_daxdev
;
719 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
720 struct block_device
*logdev
= mp
->m_logdev_targp
->bt_bdev
;
721 struct dax_device
*dax_logdev
= mp
->m_logdev_targp
->bt_daxdev
;
723 xfs_free_buftarg(mp
, mp
->m_logdev_targp
);
724 xfs_blkdev_put(logdev
);
725 fs_put_dax(dax_logdev
);
727 if (mp
->m_rtdev_targp
) {
728 struct block_device
*rtdev
= mp
->m_rtdev_targp
->bt_bdev
;
729 struct dax_device
*dax_rtdev
= mp
->m_rtdev_targp
->bt_daxdev
;
731 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
732 xfs_blkdev_put(rtdev
);
733 fs_put_dax(dax_rtdev
);
735 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
736 fs_put_dax(dax_ddev
);
740 * The file system configurations are:
741 * (1) device (partition) with data and internal log
742 * (2) logical volume with data and log subvolumes.
743 * (3) logical volume with data, log, and realtime subvolumes.
745 * We only have to handle opening the log and realtime volumes here if
746 * they are present. The data subvolume has already been opened by
747 * get_sb_bdev() and is stored in sb->s_bdev.
751 struct xfs_mount
*mp
)
753 struct block_device
*ddev
= mp
->m_super
->s_bdev
;
754 struct dax_device
*dax_ddev
= fs_dax_get_by_bdev(ddev
);
755 struct dax_device
*dax_logdev
= NULL
, *dax_rtdev
= NULL
;
756 struct block_device
*logdev
= NULL
, *rtdev
= NULL
;
760 * Open real time and log devices - order is important.
763 error
= xfs_blkdev_get(mp
, mp
->m_logname
, &logdev
);
766 dax_logdev
= fs_dax_get_by_bdev(logdev
);
770 error
= xfs_blkdev_get(mp
, mp
->m_rtname
, &rtdev
);
772 goto out_close_logdev
;
774 if (rtdev
== ddev
|| rtdev
== logdev
) {
776 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
778 goto out_close_rtdev
;
780 dax_rtdev
= fs_dax_get_by_bdev(rtdev
);
784 * Setup xfs_mount buffer target pointers
787 mp
->m_ddev_targp
= xfs_alloc_buftarg(mp
, ddev
, dax_ddev
);
788 if (!mp
->m_ddev_targp
)
789 goto out_close_rtdev
;
792 mp
->m_rtdev_targp
= xfs_alloc_buftarg(mp
, rtdev
, dax_rtdev
);
793 if (!mp
->m_rtdev_targp
)
794 goto out_free_ddev_targ
;
797 if (logdev
&& logdev
!= ddev
) {
798 mp
->m_logdev_targp
= xfs_alloc_buftarg(mp
, logdev
, dax_logdev
);
799 if (!mp
->m_logdev_targp
)
800 goto out_free_rtdev_targ
;
802 mp
->m_logdev_targp
= mp
->m_ddev_targp
;
808 if (mp
->m_rtdev_targp
)
809 xfs_free_buftarg(mp
, mp
->m_rtdev_targp
);
811 xfs_free_buftarg(mp
, mp
->m_ddev_targp
);
813 xfs_blkdev_put(rtdev
);
814 fs_put_dax(dax_rtdev
);
816 if (logdev
&& logdev
!= ddev
) {
817 xfs_blkdev_put(logdev
);
818 fs_put_dax(dax_logdev
);
821 fs_put_dax(dax_ddev
);
826 * Setup xfs_mount buffer target pointers based on superblock
830 struct xfs_mount
*mp
)
834 error
= xfs_setsize_buftarg(mp
->m_ddev_targp
, mp
->m_sb
.sb_sectsize
);
838 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
839 unsigned int log_sector_size
= BBSIZE
;
841 if (xfs_sb_version_hassector(&mp
->m_sb
))
842 log_sector_size
= mp
->m_sb
.sb_logsectsize
;
843 error
= xfs_setsize_buftarg(mp
->m_logdev_targp
,
848 if (mp
->m_rtdev_targp
) {
849 error
= xfs_setsize_buftarg(mp
->m_rtdev_targp
,
850 mp
->m_sb
.sb_sectsize
);
859 xfs_init_mount_workqueues(
860 struct xfs_mount
*mp
)
862 mp
->m_buf_workqueue
= alloc_workqueue("xfs-buf/%s",
863 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 1, mp
->m_fsname
);
864 if (!mp
->m_buf_workqueue
)
867 mp
->m_data_workqueue
= alloc_workqueue("xfs-data/%s",
868 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
869 if (!mp
->m_data_workqueue
)
870 goto out_destroy_buf
;
872 mp
->m_unwritten_workqueue
= alloc_workqueue("xfs-conv/%s",
873 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
874 if (!mp
->m_unwritten_workqueue
)
875 goto out_destroy_data_iodone_queue
;
877 mp
->m_cil_workqueue
= alloc_workqueue("xfs-cil/%s",
878 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
879 if (!mp
->m_cil_workqueue
)
880 goto out_destroy_unwritten
;
882 mp
->m_reclaim_workqueue
= alloc_workqueue("xfs-reclaim/%s",
883 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
884 if (!mp
->m_reclaim_workqueue
)
885 goto out_destroy_cil
;
887 mp
->m_log_workqueue
= alloc_workqueue("xfs-log/%s",
888 WQ_MEM_RECLAIM
|WQ_FREEZABLE
|WQ_HIGHPRI
, 0,
890 if (!mp
->m_log_workqueue
)
891 goto out_destroy_reclaim
;
893 mp
->m_eofblocks_workqueue
= alloc_workqueue("xfs-eofblocks/%s",
894 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0, mp
->m_fsname
);
895 if (!mp
->m_eofblocks_workqueue
)
896 goto out_destroy_log
;
898 mp
->m_sync_workqueue
= alloc_workqueue("xfs-sync/%s", WQ_FREEZABLE
, 0,
900 if (!mp
->m_sync_workqueue
)
901 goto out_destroy_eofb
;
906 destroy_workqueue(mp
->m_eofblocks_workqueue
);
908 destroy_workqueue(mp
->m_log_workqueue
);
910 destroy_workqueue(mp
->m_reclaim_workqueue
);
912 destroy_workqueue(mp
->m_cil_workqueue
);
913 out_destroy_unwritten
:
914 destroy_workqueue(mp
->m_unwritten_workqueue
);
915 out_destroy_data_iodone_queue
:
916 destroy_workqueue(mp
->m_data_workqueue
);
918 destroy_workqueue(mp
->m_buf_workqueue
);
924 xfs_destroy_mount_workqueues(
925 struct xfs_mount
*mp
)
927 destroy_workqueue(mp
->m_sync_workqueue
);
928 destroy_workqueue(mp
->m_eofblocks_workqueue
);
929 destroy_workqueue(mp
->m_log_workqueue
);
930 destroy_workqueue(mp
->m_reclaim_workqueue
);
931 destroy_workqueue(mp
->m_cil_workqueue
);
932 destroy_workqueue(mp
->m_data_workqueue
);
933 destroy_workqueue(mp
->m_unwritten_workqueue
);
934 destroy_workqueue(mp
->m_buf_workqueue
);
938 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
939 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
940 * for IO to complete so that we effectively throttle multiple callers to the
941 * rate at which IO is completing.
945 struct xfs_mount
*mp
)
947 struct super_block
*sb
= mp
->m_super
;
949 if (down_read_trylock(&sb
->s_umount
)) {
951 up_read(&sb
->s_umount
);
955 /* Catch misguided souls that try to use this interface on XFS */
956 STATIC
struct inode
*
958 struct super_block
*sb
)
965 * Now that the generic code is guaranteed not to be accessing
966 * the linux inode, we can inactivate and reclaim the inode.
969 xfs_fs_destroy_inode(
972 struct xfs_inode
*ip
= XFS_I(inode
);
975 trace_xfs_destroy_inode(ip
);
977 ASSERT(!rwsem_is_locked(&inode
->i_rwsem
));
978 XFS_STATS_INC(ip
->i_mount
, vn_rele
);
979 XFS_STATS_INC(ip
->i_mount
, vn_remove
);
981 if (xfs_is_reflink_inode(ip
)) {
982 error
= xfs_reflink_cancel_cow_range(ip
, 0, NULLFILEOFF
, true);
983 if (error
&& !XFS_FORCED_SHUTDOWN(ip
->i_mount
))
984 xfs_warn(ip
->i_mount
,
985 "Error %d while evicting CoW blocks for inode %llu.",
991 ASSERT(XFS_FORCED_SHUTDOWN(ip
->i_mount
) || ip
->i_delayed_blks
== 0);
992 XFS_STATS_INC(ip
->i_mount
, vn_reclaim
);
995 * We should never get here with one of the reclaim flags already set.
997 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIMABLE
));
998 ASSERT_ALWAYS(!xfs_iflags_test(ip
, XFS_IRECLAIM
));
1001 * We always use background reclaim here because even if the
1002 * inode is clean, it still may be under IO and hence we have
1003 * to take the flush lock. The background reclaim path handles
1004 * this more efficiently than we can here, so simply let background
1005 * reclaim tear down all inodes.
1007 xfs_inode_set_reclaim_tag(ip
);
1011 * Slab object creation initialisation for the XFS inode.
1012 * This covers only the idempotent fields in the XFS inode;
1013 * all other fields need to be initialised on allocation
1014 * from the slab. This avoids the need to repeatedly initialise
1015 * fields in the xfs inode that left in the initialise state
1016 * when freeing the inode.
1019 xfs_fs_inode_init_once(
1022 struct xfs_inode
*ip
= inode
;
1024 memset(ip
, 0, sizeof(struct xfs_inode
));
1027 inode_init_once(VFS_I(ip
));
1030 atomic_set(&ip
->i_pincount
, 0);
1031 spin_lock_init(&ip
->i_flags_lock
);
1033 mrlock_init(&ip
->i_mmaplock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
1034 "xfsino", ip
->i_ino
);
1035 mrlock_init(&ip
->i_lock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
1036 "xfsino", ip
->i_ino
);
1040 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1041 * serialised against cache hits here via the inode->i_lock and igrab() in
1042 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1043 * racing with us, and it avoids needing to grab a spinlock here for every inode
1044 * we drop the final reference on.
1048 struct inode
*inode
)
1050 struct xfs_inode
*ip
= XFS_I(inode
);
1053 * If this unlinked inode is in the middle of recovery, don't
1054 * drop the inode just yet; log recovery will take care of
1055 * that. See the comment for this inode flag.
1057 if (ip
->i_flags
& XFS_IRECOVERY
) {
1058 ASSERT(ip
->i_mount
->m_log
->l_flags
& XLOG_RECOVERY_NEEDED
);
1062 return generic_drop_inode(inode
) || (ip
->i_flags
& XFS_IDONTCACHE
);
1067 struct xfs_mount
*mp
)
1069 kfree(mp
->m_fsname
);
1070 kfree(mp
->m_rtname
);
1071 kfree(mp
->m_logname
);
1076 struct super_block
*sb
,
1079 struct xfs_mount
*mp
= XFS_M(sb
);
1082 * Doing anything during the async pass would be counterproductive.
1087 xfs_log_force(mp
, XFS_LOG_SYNC
);
1090 * The disk must be active because we're syncing.
1091 * We schedule log work now (now that the disk is
1092 * active) instead of later (when it might not be).
1094 flush_delayed_work(&mp
->m_log
->l_work
);
1102 struct dentry
*dentry
,
1103 struct kstatfs
*statp
)
1105 struct xfs_mount
*mp
= XFS_M(dentry
->d_sb
);
1106 xfs_sb_t
*sbp
= &mp
->m_sb
;
1107 struct xfs_inode
*ip
= XFS_I(d_inode(dentry
));
1108 uint64_t fakeinos
, id
;
1115 statp
->f_type
= XFS_SB_MAGIC
;
1116 statp
->f_namelen
= MAXNAMELEN
- 1;
1118 id
= huge_encode_dev(mp
->m_ddev_targp
->bt_dev
);
1119 statp
->f_fsid
.val
[0] = (u32
)id
;
1120 statp
->f_fsid
.val
[1] = (u32
)(id
>> 32);
1122 icount
= percpu_counter_sum(&mp
->m_icount
);
1123 ifree
= percpu_counter_sum(&mp
->m_ifree
);
1124 fdblocks
= percpu_counter_sum(&mp
->m_fdblocks
);
1126 spin_lock(&mp
->m_sb_lock
);
1127 statp
->f_bsize
= sbp
->sb_blocksize
;
1128 lsize
= sbp
->sb_logstart
? sbp
->sb_logblocks
: 0;
1129 statp
->f_blocks
= sbp
->sb_dblocks
- lsize
;
1130 spin_unlock(&mp
->m_sb_lock
);
1132 statp
->f_bfree
= fdblocks
- mp
->m_alloc_set_aside
;
1133 statp
->f_bavail
= statp
->f_bfree
;
1135 fakeinos
= statp
->f_bfree
<< sbp
->sb_inopblog
;
1136 statp
->f_files
= MIN(icount
+ fakeinos
, (uint64_t)XFS_MAXINUMBER
);
1137 if (mp
->m_maxicount
)
1138 statp
->f_files
= min_t(typeof(statp
->f_files
),
1142 /* If sb_icount overshot maxicount, report actual allocation */
1143 statp
->f_files
= max_t(typeof(statp
->f_files
),
1147 /* make sure statp->f_ffree does not underflow */
1148 ffree
= statp
->f_files
- (icount
- ifree
);
1149 statp
->f_ffree
= max_t(int64_t, ffree
, 0);
1152 if ((ip
->i_d
.di_flags
& XFS_DIFLAG_PROJINHERIT
) &&
1153 ((mp
->m_qflags
& (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))) ==
1154 (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))
1155 xfs_qm_statvfs(ip
, statp
);
1160 xfs_save_resvblks(struct xfs_mount
*mp
)
1162 uint64_t resblks
= 0;
1164 mp
->m_resblks_save
= mp
->m_resblks
;
1165 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1169 xfs_restore_resvblks(struct xfs_mount
*mp
)
1173 if (mp
->m_resblks_save
) {
1174 resblks
= mp
->m_resblks_save
;
1175 mp
->m_resblks_save
= 0;
1177 resblks
= xfs_default_resblks(mp
);
1179 xfs_reserve_blocks(mp
, &resblks
, NULL
);
1183 * Trigger writeback of all the dirty metadata in the file system.
1185 * This ensures that the metadata is written to their location on disk rather
1186 * than just existing in transactions in the log. This means after a quiesce
1187 * there is no log replay required to write the inodes to disk - this is the
1188 * primary difference between a sync and a quiesce.
1190 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1191 * it is started again when appropriate.
1195 struct xfs_mount
*mp
)
1199 /* wait for all modifications to complete */
1200 while (atomic_read(&mp
->m_active_trans
) > 0)
1203 /* force the log to unpin objects from the now complete transactions */
1204 xfs_log_force(mp
, XFS_LOG_SYNC
);
1206 /* reclaim inodes to do any IO before the freeze completes */
1207 xfs_reclaim_inodes(mp
, 0);
1208 xfs_reclaim_inodes(mp
, SYNC_WAIT
);
1210 /* Push the superblock and write an unmount record */
1211 error
= xfs_log_sbcount(mp
);
1213 xfs_warn(mp
, "xfs_attr_quiesce: failed to log sb changes. "
1214 "Frozen image may not be consistent.");
1216 * Just warn here till VFS can correctly support
1217 * read-only remount without racing.
1219 WARN_ON(atomic_read(&mp
->m_active_trans
) != 0);
1221 xfs_log_quiesce(mp
);
1225 xfs_test_remount_options(
1226 struct super_block
*sb
,
1227 struct xfs_mount
*mp
,
1231 struct xfs_mount
*tmp_mp
;
1233 tmp_mp
= kmem_zalloc(sizeof(*tmp_mp
), KM_MAYFAIL
);
1237 tmp_mp
->m_super
= sb
;
1238 error
= xfs_parseargs(tmp_mp
, options
);
1239 xfs_free_fsname(tmp_mp
);
1247 struct super_block
*sb
,
1251 struct xfs_mount
*mp
= XFS_M(sb
);
1252 xfs_sb_t
*sbp
= &mp
->m_sb
;
1253 substring_t args
[MAX_OPT_ARGS
];
1257 /* First, check for complete junk; i.e. invalid options */
1258 error
= xfs_test_remount_options(sb
, mp
, options
);
1262 sync_filesystem(sb
);
1263 while ((p
= strsep(&options
, ",")) != NULL
) {
1269 token
= match_token(p
, tokens
, args
);
1272 xfs_warn(mp
, "%s option is deprecated, ignoring.", p
);
1273 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
1276 xfs_warn(mp
, "%s option is deprecated, ignoring.", p
);
1277 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
1280 mp
->m_flags
&= ~XFS_MOUNT_SMALL_INUMS
;
1281 mp
->m_maxagi
= xfs_set_inode_alloc(mp
, sbp
->sb_agcount
);
1284 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
1285 mp
->m_maxagi
= xfs_set_inode_alloc(mp
, sbp
->sb_agcount
);
1289 * Logically we would return an error here to prevent
1290 * users from believing they might have changed
1291 * mount options using remount which can't be changed.
1293 * But unfortunately mount(8) adds all options from
1294 * mtab and fstab to the mount arguments in some cases
1295 * so we can't blindly reject options, but have to
1296 * check for each specified option if it actually
1297 * differs from the currently set option and only
1298 * reject it if that's the case.
1300 * Until that is implemented we return success for
1301 * every remount request, and silently ignore all
1302 * options that we can't actually change.
1306 "mount option \"%s\" not supported for remount", p
);
1315 if ((mp
->m_flags
& XFS_MOUNT_RDONLY
) && !(*flags
& MS_RDONLY
)) {
1316 if (mp
->m_flags
& XFS_MOUNT_NORECOVERY
) {
1318 "ro->rw transition prohibited on norecovery mount");
1322 if (XFS_SB_VERSION_NUM(sbp
) == XFS_SB_VERSION_5
&&
1323 xfs_sb_has_ro_compat_feature(sbp
,
1324 XFS_SB_FEAT_RO_COMPAT_UNKNOWN
)) {
1326 "ro->rw transition prohibited on unknown (0x%x) ro-compat filesystem",
1327 (sbp
->sb_features_ro_compat
&
1328 XFS_SB_FEAT_RO_COMPAT_UNKNOWN
));
1332 mp
->m_flags
&= ~XFS_MOUNT_RDONLY
;
1335 * If this is the first remount to writeable state we
1336 * might have some superblock changes to update.
1338 if (mp
->m_update_sb
) {
1339 error
= xfs_sync_sb(mp
, false);
1341 xfs_warn(mp
, "failed to write sb changes");
1344 mp
->m_update_sb
= false;
1348 * Fill out the reserve pool if it is empty. Use the stashed
1349 * value if it is non-zero, otherwise go with the default.
1351 xfs_restore_resvblks(mp
);
1352 xfs_log_work_queue(mp
);
1353 xfs_queue_eofblocks(mp
);
1355 /* Recover any CoW blocks that never got remapped. */
1356 error
= xfs_reflink_recover_cow(mp
);
1359 "Error %d recovering leftover CoW allocations.", error
);
1360 xfs_force_shutdown(mp
, SHUTDOWN_CORRUPT_INCORE
);
1364 /* Create the per-AG metadata reservation pool .*/
1365 error
= xfs_fs_reserve_ag_blocks(mp
);
1366 if (error
&& error
!= -ENOSPC
)
1371 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
) && (*flags
& MS_RDONLY
)) {
1372 /* Free the per-AG metadata reservation pool. */
1373 error
= xfs_fs_unreserve_ag_blocks(mp
);
1375 xfs_force_shutdown(mp
, SHUTDOWN_CORRUPT_INCORE
);
1380 * Before we sync the metadata, we need to free up the reserve
1381 * block pool so that the used block count in the superblock on
1382 * disk is correct at the end of the remount. Stash the current
1383 * reserve pool size so that if we get remounted rw, we can
1384 * return it to the same size.
1386 xfs_save_resvblks(mp
);
1389 * Cancel background eofb scanning so it cannot race with the
1390 * final log force+buftarg wait and deadlock the remount.
1392 cancel_delayed_work_sync(&mp
->m_eofblocks_work
);
1394 xfs_quiesce_attr(mp
);
1395 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
1402 * Second stage of a freeze. The data is already frozen so we only
1403 * need to take care of the metadata. Once that's done sync the superblock
1404 * to the log to dirty it in case of a crash while frozen. This ensures that we
1405 * will recover the unlinked inode lists on the next mount.
1409 struct super_block
*sb
)
1411 struct xfs_mount
*mp
= XFS_M(sb
);
1413 xfs_save_resvblks(mp
);
1414 xfs_quiesce_attr(mp
);
1415 return xfs_sync_sb(mp
, true);
1420 struct super_block
*sb
)
1422 struct xfs_mount
*mp
= XFS_M(sb
);
1424 xfs_restore_resvblks(mp
);
1425 xfs_log_work_queue(mp
);
1430 xfs_fs_show_options(
1432 struct dentry
*root
)
1434 return xfs_showargs(XFS_M(root
->d_sb
), m
);
1438 * This function fills in xfs_mount_t fields based on mount args.
1439 * Note: the superblock _has_ now been read in.
1443 struct xfs_mount
*mp
)
1445 int ronly
= (mp
->m_flags
& XFS_MOUNT_RDONLY
);
1447 /* Fail a mount where the logbuf is smaller than the log stripe */
1448 if (xfs_sb_version_haslogv2(&mp
->m_sb
)) {
1449 if (mp
->m_logbsize
<= 0 &&
1450 mp
->m_sb
.sb_logsunit
> XLOG_BIG_RECORD_BSIZE
) {
1451 mp
->m_logbsize
= mp
->m_sb
.sb_logsunit
;
1452 } else if (mp
->m_logbsize
> 0 &&
1453 mp
->m_logbsize
< mp
->m_sb
.sb_logsunit
) {
1455 "logbuf size must be greater than or equal to log stripe size");
1459 /* Fail a mount if the logbuf is larger than 32K */
1460 if (mp
->m_logbsize
> XLOG_BIG_RECORD_BSIZE
) {
1462 "logbuf size for version 1 logs must be 16K or 32K");
1468 * V5 filesystems always use attr2 format for attributes.
1470 if (xfs_sb_version_hascrc(&mp
->m_sb
) &&
1471 (mp
->m_flags
& XFS_MOUNT_NOATTR2
)) {
1472 xfs_warn(mp
, "Cannot mount a V5 filesystem as noattr2. "
1473 "attr2 is always enabled for V5 filesystems.");
1478 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1479 * told by noattr2 to turn it off
1481 if (xfs_sb_version_hasattr2(&mp
->m_sb
) &&
1482 !(mp
->m_flags
& XFS_MOUNT_NOATTR2
))
1483 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
1486 * prohibit r/w mounts of read-only filesystems
1488 if ((mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) && !ronly
) {
1490 "cannot mount a read-only filesystem as read-write");
1494 if ((mp
->m_qflags
& (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
)) &&
1495 (mp
->m_qflags
& (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
)) &&
1496 !xfs_sb_version_has_pquotino(&mp
->m_sb
)) {
1498 "Super block does not support project and group quota together");
1506 xfs_init_percpu_counters(
1507 struct xfs_mount
*mp
)
1511 error
= percpu_counter_init(&mp
->m_icount
, 0, GFP_KERNEL
);
1515 error
= percpu_counter_init(&mp
->m_ifree
, 0, GFP_KERNEL
);
1519 error
= percpu_counter_init(&mp
->m_fdblocks
, 0, GFP_KERNEL
);
1526 percpu_counter_destroy(&mp
->m_ifree
);
1528 percpu_counter_destroy(&mp
->m_icount
);
1533 xfs_reinit_percpu_counters(
1534 struct xfs_mount
*mp
)
1536 percpu_counter_set(&mp
->m_icount
, mp
->m_sb
.sb_icount
);
1537 percpu_counter_set(&mp
->m_ifree
, mp
->m_sb
.sb_ifree
);
1538 percpu_counter_set(&mp
->m_fdblocks
, mp
->m_sb
.sb_fdblocks
);
1542 xfs_destroy_percpu_counters(
1543 struct xfs_mount
*mp
)
1545 percpu_counter_destroy(&mp
->m_icount
);
1546 percpu_counter_destroy(&mp
->m_ifree
);
1547 percpu_counter_destroy(&mp
->m_fdblocks
);
1552 struct super_block
*sb
,
1557 struct xfs_mount
*mp
= NULL
;
1558 int flags
= 0, error
= -ENOMEM
;
1560 mp
= kzalloc(sizeof(struct xfs_mount
), GFP_KERNEL
);
1564 spin_lock_init(&mp
->m_sb_lock
);
1565 mutex_init(&mp
->m_growlock
);
1566 atomic_set(&mp
->m_active_trans
, 0);
1567 INIT_DELAYED_WORK(&mp
->m_reclaim_work
, xfs_reclaim_worker
);
1568 INIT_DELAYED_WORK(&mp
->m_eofblocks_work
, xfs_eofblocks_worker
);
1569 INIT_DELAYED_WORK(&mp
->m_cowblocks_work
, xfs_cowblocks_worker
);
1570 mp
->m_kobj
.kobject
.kset
= xfs_kset
;
1575 error
= xfs_parseargs(mp
, (char *)data
);
1577 goto out_free_fsname
;
1579 sb_min_blocksize(sb
, BBSIZE
);
1580 sb
->s_xattr
= xfs_xattr_handlers
;
1581 sb
->s_export_op
= &xfs_export_operations
;
1582 #ifdef CONFIG_XFS_QUOTA
1583 sb
->s_qcop
= &xfs_quotactl_operations
;
1584 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
1586 sb
->s_op
= &xfs_super_operations
;
1589 flags
|= XFS_MFSI_QUIET
;
1591 error
= xfs_open_devices(mp
);
1593 goto out_free_fsname
;
1595 error
= xfs_init_mount_workqueues(mp
);
1597 goto out_close_devices
;
1599 error
= xfs_init_percpu_counters(mp
);
1601 goto out_destroy_workqueues
;
1603 /* Allocate stats memory before we do operations that might use it */
1604 mp
->m_stats
.xs_stats
= alloc_percpu(struct xfsstats
);
1605 if (!mp
->m_stats
.xs_stats
) {
1607 goto out_destroy_counters
;
1610 error
= xfs_readsb(mp
, flags
);
1612 goto out_free_stats
;
1614 error
= xfs_finish_flags(mp
);
1618 error
= xfs_setup_devices(mp
);
1622 error
= xfs_filestream_mount(mp
);
1627 * we must configure the block size in the superblock before we run the
1628 * full mount process as the mount process can lookup and cache inodes.
1630 sb
->s_magic
= XFS_SB_MAGIC
;
1631 sb
->s_blocksize
= mp
->m_sb
.sb_blocksize
;
1632 sb
->s_blocksize_bits
= ffs(sb
->s_blocksize
) - 1;
1633 sb
->s_maxbytes
= xfs_max_file_offset(sb
->s_blocksize_bits
);
1634 sb
->s_max_links
= XFS_MAXLINK
;
1635 sb
->s_time_gran
= 1;
1636 set_posix_acl_flag(sb
);
1638 /* version 5 superblocks support inode version counters. */
1639 if (XFS_SB_VERSION_NUM(&mp
->m_sb
) == XFS_SB_VERSION_5
)
1640 sb
->s_flags
|= SB_I_VERSION
;
1642 if (mp
->m_flags
& XFS_MOUNT_DAX
) {
1644 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1646 error
= bdev_dax_supported(sb
, sb
->s_blocksize
);
1649 "DAX unsupported by block device. Turning off DAX.");
1650 mp
->m_flags
&= ~XFS_MOUNT_DAX
;
1652 if (xfs_sb_version_hasreflink(&mp
->m_sb
))
1654 "DAX and reflink have not been tested together!");
1657 if (mp
->m_flags
& XFS_MOUNT_DISCARD
) {
1658 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
1660 if (!blk_queue_discard(q
)) {
1661 xfs_warn(mp
, "mounting with \"discard\" option, but "
1662 "the device does not support discard");
1663 mp
->m_flags
&= ~XFS_MOUNT_DISCARD
;
1667 if (xfs_sb_version_hasrmapbt(&mp
->m_sb
)) {
1668 if (mp
->m_sb
.sb_rblocks
) {
1670 "EXPERIMENTAL reverse mapping btree not compatible with realtime device!");
1672 goto out_filestream_unmount
;
1675 "EXPERIMENTAL reverse mapping btree feature enabled. Use at your own risk!");
1678 if (xfs_sb_version_hasreflink(&mp
->m_sb
))
1680 "EXPERIMENTAL reflink feature enabled. Use at your own risk!");
1682 error
= xfs_mountfs(mp
);
1684 goto out_filestream_unmount
;
1686 root
= igrab(VFS_I(mp
->m_rootip
));
1691 sb
->s_root
= d_make_root(root
);
1699 out_filestream_unmount
:
1700 xfs_filestream_unmount(mp
);
1704 free_percpu(mp
->m_stats
.xs_stats
);
1705 out_destroy_counters
:
1706 xfs_destroy_percpu_counters(mp
);
1707 out_destroy_workqueues
:
1708 xfs_destroy_mount_workqueues(mp
);
1710 xfs_close_devices(mp
);
1712 xfs_free_fsname(mp
);
1718 xfs_filestream_unmount(mp
);
1725 struct super_block
*sb
)
1727 struct xfs_mount
*mp
= XFS_M(sb
);
1729 xfs_notice(mp
, "Unmounting Filesystem");
1730 xfs_filestream_unmount(mp
);
1734 free_percpu(mp
->m_stats
.xs_stats
);
1735 xfs_destroy_percpu_counters(mp
);
1736 xfs_destroy_mount_workqueues(mp
);
1737 xfs_close_devices(mp
);
1738 xfs_free_fsname(mp
);
1742 STATIC
struct dentry
*
1744 struct file_system_type
*fs_type
,
1746 const char *dev_name
,
1749 return mount_bdev(fs_type
, flags
, dev_name
, data
, xfs_fs_fill_super
);
1753 xfs_fs_nr_cached_objects(
1754 struct super_block
*sb
,
1755 struct shrink_control
*sc
)
1757 return xfs_reclaim_inodes_count(XFS_M(sb
));
1761 xfs_fs_free_cached_objects(
1762 struct super_block
*sb
,
1763 struct shrink_control
*sc
)
1765 return xfs_reclaim_inodes_nr(XFS_M(sb
), sc
->nr_to_scan
);
1768 static const struct super_operations xfs_super_operations
= {
1769 .alloc_inode
= xfs_fs_alloc_inode
,
1770 .destroy_inode
= xfs_fs_destroy_inode
,
1771 .drop_inode
= xfs_fs_drop_inode
,
1772 .put_super
= xfs_fs_put_super
,
1773 .sync_fs
= xfs_fs_sync_fs
,
1774 .freeze_fs
= xfs_fs_freeze
,
1775 .unfreeze_fs
= xfs_fs_unfreeze
,
1776 .statfs
= xfs_fs_statfs
,
1777 .remount_fs
= xfs_fs_remount
,
1778 .show_options
= xfs_fs_show_options
,
1779 .nr_cached_objects
= xfs_fs_nr_cached_objects
,
1780 .free_cached_objects
= xfs_fs_free_cached_objects
,
1783 static struct file_system_type xfs_fs_type
= {
1784 .owner
= THIS_MODULE
,
1786 .mount
= xfs_fs_mount
,
1787 .kill_sb
= kill_block_super
,
1788 .fs_flags
= FS_REQUIRES_DEV
,
1790 MODULE_ALIAS_FS("xfs");
1793 xfs_init_zones(void)
1795 xfs_ioend_bioset
= bioset_create(4 * MAX_BUF_PER_PAGE
,
1796 offsetof(struct xfs_ioend
, io_inline_bio
),
1798 if (!xfs_ioend_bioset
)
1801 xfs_log_ticket_zone
= kmem_zone_init(sizeof(xlog_ticket_t
),
1803 if (!xfs_log_ticket_zone
)
1804 goto out_free_ioend_bioset
;
1806 xfs_bmap_free_item_zone
= kmem_zone_init(
1807 sizeof(struct xfs_extent_free_item
),
1808 "xfs_bmap_free_item");
1809 if (!xfs_bmap_free_item_zone
)
1810 goto out_destroy_log_ticket_zone
;
1812 xfs_btree_cur_zone
= kmem_zone_init(sizeof(xfs_btree_cur_t
),
1814 if (!xfs_btree_cur_zone
)
1815 goto out_destroy_bmap_free_item_zone
;
1817 xfs_da_state_zone
= kmem_zone_init(sizeof(xfs_da_state_t
),
1819 if (!xfs_da_state_zone
)
1820 goto out_destroy_btree_cur_zone
;
1822 xfs_ifork_zone
= kmem_zone_init(sizeof(xfs_ifork_t
), "xfs_ifork");
1823 if (!xfs_ifork_zone
)
1824 goto out_destroy_da_state_zone
;
1826 xfs_trans_zone
= kmem_zone_init(sizeof(xfs_trans_t
), "xfs_trans");
1827 if (!xfs_trans_zone
)
1828 goto out_destroy_ifork_zone
;
1830 xfs_log_item_desc_zone
=
1831 kmem_zone_init(sizeof(struct xfs_log_item_desc
),
1832 "xfs_log_item_desc");
1833 if (!xfs_log_item_desc_zone
)
1834 goto out_destroy_trans_zone
;
1837 * The size of the zone allocated buf log item is the maximum
1838 * size possible under XFS. This wastes a little bit of memory,
1839 * but it is much faster.
1841 xfs_buf_item_zone
= kmem_zone_init(sizeof(struct xfs_buf_log_item
),
1843 if (!xfs_buf_item_zone
)
1844 goto out_destroy_log_item_desc_zone
;
1846 xfs_efd_zone
= kmem_zone_init((sizeof(xfs_efd_log_item_t
) +
1847 ((XFS_EFD_MAX_FAST_EXTENTS
- 1) *
1848 sizeof(xfs_extent_t
))), "xfs_efd_item");
1850 goto out_destroy_buf_item_zone
;
1852 xfs_efi_zone
= kmem_zone_init((sizeof(xfs_efi_log_item_t
) +
1853 ((XFS_EFI_MAX_FAST_EXTENTS
- 1) *
1854 sizeof(xfs_extent_t
))), "xfs_efi_item");
1856 goto out_destroy_efd_zone
;
1859 kmem_zone_init_flags(sizeof(xfs_inode_t
), "xfs_inode",
1860 KM_ZONE_HWALIGN
| KM_ZONE_RECLAIM
| KM_ZONE_SPREAD
|
1861 KM_ZONE_ACCOUNT
, xfs_fs_inode_init_once
);
1862 if (!xfs_inode_zone
)
1863 goto out_destroy_efi_zone
;
1866 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t
), "xfs_ili",
1867 KM_ZONE_SPREAD
, NULL
);
1869 goto out_destroy_inode_zone
;
1870 xfs_icreate_zone
= kmem_zone_init(sizeof(struct xfs_icreate_item
),
1872 if (!xfs_icreate_zone
)
1873 goto out_destroy_ili_zone
;
1875 xfs_rud_zone
= kmem_zone_init(sizeof(struct xfs_rud_log_item
),
1878 goto out_destroy_icreate_zone
;
1880 xfs_rui_zone
= kmem_zone_init(
1881 xfs_rui_log_item_sizeof(XFS_RUI_MAX_FAST_EXTENTS
),
1884 goto out_destroy_rud_zone
;
1886 xfs_cud_zone
= kmem_zone_init(sizeof(struct xfs_cud_log_item
),
1889 goto out_destroy_rui_zone
;
1891 xfs_cui_zone
= kmem_zone_init(
1892 xfs_cui_log_item_sizeof(XFS_CUI_MAX_FAST_EXTENTS
),
1895 goto out_destroy_cud_zone
;
1897 xfs_bud_zone
= kmem_zone_init(sizeof(struct xfs_bud_log_item
),
1900 goto out_destroy_cui_zone
;
1902 xfs_bui_zone
= kmem_zone_init(
1903 xfs_bui_log_item_sizeof(XFS_BUI_MAX_FAST_EXTENTS
),
1906 goto out_destroy_bud_zone
;
1910 out_destroy_bud_zone
:
1911 kmem_zone_destroy(xfs_bud_zone
);
1912 out_destroy_cui_zone
:
1913 kmem_zone_destroy(xfs_cui_zone
);
1914 out_destroy_cud_zone
:
1915 kmem_zone_destroy(xfs_cud_zone
);
1916 out_destroy_rui_zone
:
1917 kmem_zone_destroy(xfs_rui_zone
);
1918 out_destroy_rud_zone
:
1919 kmem_zone_destroy(xfs_rud_zone
);
1920 out_destroy_icreate_zone
:
1921 kmem_zone_destroy(xfs_icreate_zone
);
1922 out_destroy_ili_zone
:
1923 kmem_zone_destroy(xfs_ili_zone
);
1924 out_destroy_inode_zone
:
1925 kmem_zone_destroy(xfs_inode_zone
);
1926 out_destroy_efi_zone
:
1927 kmem_zone_destroy(xfs_efi_zone
);
1928 out_destroy_efd_zone
:
1929 kmem_zone_destroy(xfs_efd_zone
);
1930 out_destroy_buf_item_zone
:
1931 kmem_zone_destroy(xfs_buf_item_zone
);
1932 out_destroy_log_item_desc_zone
:
1933 kmem_zone_destroy(xfs_log_item_desc_zone
);
1934 out_destroy_trans_zone
:
1935 kmem_zone_destroy(xfs_trans_zone
);
1936 out_destroy_ifork_zone
:
1937 kmem_zone_destroy(xfs_ifork_zone
);
1938 out_destroy_da_state_zone
:
1939 kmem_zone_destroy(xfs_da_state_zone
);
1940 out_destroy_btree_cur_zone
:
1941 kmem_zone_destroy(xfs_btree_cur_zone
);
1942 out_destroy_bmap_free_item_zone
:
1943 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1944 out_destroy_log_ticket_zone
:
1945 kmem_zone_destroy(xfs_log_ticket_zone
);
1946 out_free_ioend_bioset
:
1947 bioset_free(xfs_ioend_bioset
);
1953 xfs_destroy_zones(void)
1956 * Make sure all delayed rcu free are flushed before we
1960 kmem_zone_destroy(xfs_bui_zone
);
1961 kmem_zone_destroy(xfs_bud_zone
);
1962 kmem_zone_destroy(xfs_cui_zone
);
1963 kmem_zone_destroy(xfs_cud_zone
);
1964 kmem_zone_destroy(xfs_rui_zone
);
1965 kmem_zone_destroy(xfs_rud_zone
);
1966 kmem_zone_destroy(xfs_icreate_zone
);
1967 kmem_zone_destroy(xfs_ili_zone
);
1968 kmem_zone_destroy(xfs_inode_zone
);
1969 kmem_zone_destroy(xfs_efi_zone
);
1970 kmem_zone_destroy(xfs_efd_zone
);
1971 kmem_zone_destroy(xfs_buf_item_zone
);
1972 kmem_zone_destroy(xfs_log_item_desc_zone
);
1973 kmem_zone_destroy(xfs_trans_zone
);
1974 kmem_zone_destroy(xfs_ifork_zone
);
1975 kmem_zone_destroy(xfs_da_state_zone
);
1976 kmem_zone_destroy(xfs_btree_cur_zone
);
1977 kmem_zone_destroy(xfs_bmap_free_item_zone
);
1978 kmem_zone_destroy(xfs_log_ticket_zone
);
1979 bioset_free(xfs_ioend_bioset
);
1983 xfs_init_workqueues(void)
1986 * The allocation workqueue can be used in memory reclaim situations
1987 * (writepage path), and parallelism is only limited by the number of
1988 * AGs in all the filesystems mounted. Hence use the default large
1989 * max_active value for this workqueue.
1991 xfs_alloc_wq
= alloc_workqueue("xfsalloc",
1992 WQ_MEM_RECLAIM
|WQ_FREEZABLE
, 0);
1996 xfs_discard_wq
= alloc_workqueue("xfsdiscard", WQ_UNBOUND
, 0);
1997 if (!xfs_discard_wq
)
1998 goto out_free_alloc_wq
;
2002 destroy_workqueue(xfs_alloc_wq
);
2007 xfs_destroy_workqueues(void)
2009 destroy_workqueue(xfs_discard_wq
);
2010 destroy_workqueue(xfs_alloc_wq
);
2018 xfs_check_ondisk_structs();
2020 printk(KERN_INFO XFS_VERSION_STRING
" with "
2021 XFS_BUILD_OPTIONS
" enabled\n");
2023 xfs_extent_free_init_defer_op();
2024 xfs_rmap_update_init_defer_op();
2025 xfs_refcount_update_init_defer_op();
2026 xfs_bmap_update_init_defer_op();
2030 error
= xfs_init_zones();
2034 error
= xfs_init_workqueues();
2036 goto out_destroy_zones
;
2038 error
= xfs_mru_cache_init();
2040 goto out_destroy_wq
;
2042 error
= xfs_buf_init();
2044 goto out_mru_cache_uninit
;
2046 error
= xfs_init_procfs();
2048 goto out_buf_terminate
;
2050 error
= xfs_sysctl_register();
2052 goto out_cleanup_procfs
;
2054 xfs_kset
= kset_create_and_add("xfs", NULL
, fs_kobj
);
2057 goto out_sysctl_unregister
;
2060 xfsstats
.xs_kobj
.kobject
.kset
= xfs_kset
;
2062 xfsstats
.xs_stats
= alloc_percpu(struct xfsstats
);
2063 if (!xfsstats
.xs_stats
) {
2065 goto out_kset_unregister
;
2068 error
= xfs_sysfs_init(&xfsstats
.xs_kobj
, &xfs_stats_ktype
, NULL
,
2071 goto out_free_stats
;
2074 xfs_dbg_kobj
.kobject
.kset
= xfs_kset
;
2075 error
= xfs_sysfs_init(&xfs_dbg_kobj
, &xfs_dbg_ktype
, NULL
, "debug");
2077 goto out_remove_stats_kobj
;
2080 error
= xfs_qm_init();
2082 goto out_remove_dbg_kobj
;
2084 error
= register_filesystem(&xfs_fs_type
);
2091 out_remove_dbg_kobj
:
2093 xfs_sysfs_del(&xfs_dbg_kobj
);
2094 out_remove_stats_kobj
:
2096 xfs_sysfs_del(&xfsstats
.xs_kobj
);
2098 free_percpu(xfsstats
.xs_stats
);
2099 out_kset_unregister
:
2100 kset_unregister(xfs_kset
);
2101 out_sysctl_unregister
:
2102 xfs_sysctl_unregister();
2104 xfs_cleanup_procfs();
2106 xfs_buf_terminate();
2107 out_mru_cache_uninit
:
2108 xfs_mru_cache_uninit();
2110 xfs_destroy_workqueues();
2112 xfs_destroy_zones();
2121 unregister_filesystem(&xfs_fs_type
);
2123 xfs_sysfs_del(&xfs_dbg_kobj
);
2125 xfs_sysfs_del(&xfsstats
.xs_kobj
);
2126 free_percpu(xfsstats
.xs_stats
);
2127 kset_unregister(xfs_kset
);
2128 xfs_sysctl_unregister();
2129 xfs_cleanup_procfs();
2130 xfs_buf_terminate();
2131 xfs_mru_cache_uninit();
2132 xfs_destroy_workqueues();
2133 xfs_destroy_zones();
2134 xfs_uuid_table_free();
2137 module_init(init_xfs_fs
);
2138 module_exit(exit_xfs_fs
);
2140 MODULE_AUTHOR("Silicon Graphics, Inc.");
2141 MODULE_DESCRIPTION(XFS_VERSION_STRING
" with " XFS_BUILD_OPTIONS
" enabled");
2142 MODULE_LICENSE("GPL");