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"
53 #include <linux/namei.h>
54 #include <linux/init.h>
55 #include <linux/slab.h>
56 #include <linux/mount.h>
57 #include <linux/mempool.h>
58 #include <linux/writeback.h>
59 #include <linux/kthread.h>
60 #include <linux/freezer.h>
61 #include <linux/parser.h>
63 static const struct super_operations xfs_super_operations;
64 struct bio_set *xfs_ioend_bioset;
66 static struct kset *xfs_kset; /* top-level xfs sysfs dir */
68 static struct xfs_kobj xfs_dbg_kobj; /* global debug sysfs attrs */
72 * Table driven mount option parser.
75 Opt_logbufs, Opt_logbsize, Opt_logdev, Opt_rtdev, Opt_biosize,
76 Opt_wsync, Opt_noalign, Opt_swalloc, Opt_sunit, Opt_swidth, Opt_nouuid,
77 Opt_mtpt, Opt_grpid, Opt_nogrpid, Opt_bsdgroups, Opt_sysvgroups,
78 Opt_allocsize, Opt_norecovery, Opt_barrier, Opt_nobarrier,
79 Opt_inode64, Opt_inode32, Opt_ikeep, Opt_noikeep,
80 Opt_largeio, Opt_nolargeio, Opt_attr2, Opt_noattr2, Opt_filestreams,
81 Opt_quota, Opt_noquota, Opt_usrquota, Opt_grpquota, Opt_prjquota,
82 Opt_uquota, Opt_gquota, Opt_pquota,
83 Opt_uqnoenforce, Opt_gqnoenforce, Opt_pqnoenforce, Opt_qnoenforce,
84 Opt_discard, Opt_nodiscard, Opt_dax, Opt_err,
87 static const match_table_t tokens = {
88 {Opt_logbufs, "logbufs=%u"}, /* number of XFS log buffers */
89 {Opt_logbsize, "logbsize=%s"}, /* size of XFS log buffers */
90 {Opt_logdev, "logdev=%s"}, /* log device */
91 {Opt_rtdev, "rtdev=%s"}, /* realtime I/O device */
92 {Opt_biosize, "biosize=%u"}, /* log2 of preferred buffered io size */
93 {Opt_wsync, "wsync"}, /* safe-mode nfs compatible mount */
94 {Opt_noalign, "noalign"}, /* turn off stripe alignment */
95 {Opt_swalloc, "swalloc"}, /* turn on stripe width allocation */
96 {Opt_sunit, "sunit=%u"}, /* data volume stripe unit */
97 {Opt_swidth, "swidth=%u"}, /* data volume stripe width */
98 {Opt_nouuid, "nouuid"}, /* ignore filesystem UUID */
99 {Opt_mtpt, "mtpt"}, /* filesystem mount point */
100 {Opt_grpid, "grpid"}, /* group-ID from parent directory */
101 {Opt_nogrpid, "nogrpid"}, /* group-ID from current process */
102 {Opt_bsdgroups, "bsdgroups"}, /* group-ID from parent directory */
103 {Opt_sysvgroups,"sysvgroups"}, /* group-ID from current process */
104 {Opt_allocsize, "allocsize=%s"},/* preferred allocation size */
105 {Opt_norecovery,"norecovery"}, /* don't run XFS recovery */
106 {Opt_barrier, "barrier"}, /* use writer barriers for log write and
107 * unwritten extent conversion */
108 {Opt_nobarrier, "nobarrier"}, /* .. disable */
109 {Opt_inode64, "inode64"}, /* inodes can be allocated anywhere */
110 {Opt_inode32, "inode32"}, /* inode allocation limited to
111 * XFS_MAXINUMBER_32 */
112 {Opt_ikeep, "ikeep"}, /* do not free empty inode clusters */
113 {Opt_noikeep, "noikeep"}, /* free empty inode clusters */
114 {Opt_largeio, "largeio"}, /* report large I/O sizes in stat() */
115 {Opt_nolargeio, "nolargeio"}, /* do not report large I/O sizes
117 {Opt_attr2, "attr2"}, /* do use attr2 attribute format */
118 {Opt_noattr2, "noattr2"}, /* do not use attr2 attribute format */
119 {Opt_filestreams,"filestreams"},/* use filestreams allocator */
120 {Opt_quota, "quota"}, /* disk quotas (user) */
121 {Opt_noquota, "noquota"}, /* no quotas */
122 {Opt_usrquota, "usrquota"}, /* user quota enabled */
123 {Opt_grpquota, "grpquota"}, /* group quota enabled */
124 {Opt_prjquota, "prjquota"}, /* project quota enabled */
125 {Opt_uquota, "uquota"}, /* user quota (IRIX variant) */
126 {Opt_gquota, "gquota"}, /* group quota (IRIX variant) */
127 {Opt_pquota, "pquota"}, /* project quota (IRIX variant) */
128 {Opt_uqnoenforce,"uqnoenforce"},/* user quota limit enforcement */
129 {Opt_gqnoenforce,"gqnoenforce"},/* group quota limit enforcement */
130 {Opt_pqnoenforce,"pqnoenforce"},/* project quota limit enforcement */
131 {Opt_qnoenforce, "qnoenforce"}, /* same as uqnoenforce */
132 {Opt_discard, "discard"}, /* Discard unused blocks */
133 {Opt_nodiscard, "nodiscard"}, /* Do not discard unused blocks */
135 {Opt_dax, "dax"}, /* Enable direct access to bdev pages */
141 suffix_kstrtoint(const substring_t *s, unsigned int base, int *res)
143 int last, shift_left_factor = 0, _res;
147 value = match_strdup(s);
151 last = strlen(value) - 1;
152 if (value[last] == 'K' || value[last] == 'k') {
153 shift_left_factor = 10;
156 if (value[last] == 'M' || value[last] == 'm') {
157 shift_left_factor = 20;
160 if (value[last] == 'G' || value[last] == 'g') {
161 shift_left_factor = 30;
165 if (kstrtoint(value, base, &_res))
168 *res = _res << shift_left_factor;
173 * This function fills in xfs_mount_t fields based on mount args.
174 * Note: the superblock has _not_ yet been read in.
176 * Note that this function leaks the various device name allocations on
177 * failure. The caller takes care of them.
179 * *sb is const because this is also used to test options on the remount
180 * path, and we don't want this to have any side effects at remount time.
181 * Today this function does not change *sb, but just to future-proof...
185 struct xfs_mount *mp,
188 const struct super_block *sb = mp->m_super;
190 substring_t args[MAX_OPT_ARGS];
194 __uint8_t iosizelog = 0;
197 * set up the mount name first so all the errors will refer to the
200 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
203 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
206 * Copy binary VFS mount flags we are interested in.
208 if (sb->s_flags & MS_RDONLY)
209 mp->m_flags |= XFS_MOUNT_RDONLY;
210 if (sb->s_flags & MS_DIRSYNC)
211 mp->m_flags |= XFS_MOUNT_DIRSYNC;
212 if (sb->s_flags & MS_SYNCHRONOUS)
213 mp->m_flags |= XFS_MOUNT_WSYNC;
216 * Set some default flags that could be cleared by the mount option
219 mp->m_flags |= XFS_MOUNT_BARRIER;
220 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
223 * These can be overridden by the mount option parsing.
231 while ((p = strsep(&options, ",")) != NULL) {
237 token = match_token(p, tokens, args);
240 if (match_int(args, &mp->m_logbufs))
244 if (suffix_kstrtoint(args, 10, &mp->m_logbsize))
248 mp->m_logname = match_strdup(args);
253 xfs_warn(mp, "%s option not allowed on this system", p);
256 mp->m_rtname = match_strdup(args);
262 if (suffix_kstrtoint(args, 10, &iosize))
264 iosizelog = ffs(iosize) - 1;
268 mp->m_flags |= XFS_MOUNT_GRPID;
272 mp->m_flags &= ~XFS_MOUNT_GRPID;
275 mp->m_flags |= XFS_MOUNT_WSYNC;
278 mp->m_flags |= XFS_MOUNT_NORECOVERY;
281 mp->m_flags |= XFS_MOUNT_NOALIGN;
284 mp->m_flags |= XFS_MOUNT_SWALLOC;
287 if (match_int(args, &dsunit))
291 if (match_int(args, &dswidth))
295 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
298 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
301 mp->m_flags |= XFS_MOUNT_NOUUID;
304 mp->m_flags |= XFS_MOUNT_BARRIER;
307 mp->m_flags &= ~XFS_MOUNT_BARRIER;
310 mp->m_flags |= XFS_MOUNT_IKEEP;
313 mp->m_flags &= ~XFS_MOUNT_IKEEP;
316 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
319 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
322 mp->m_flags |= XFS_MOUNT_ATTR2;
325 mp->m_flags &= ~XFS_MOUNT_ATTR2;
326 mp->m_flags |= XFS_MOUNT_NOATTR2;
328 case Opt_filestreams:
329 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
332 mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT;
333 mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD;
334 mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE;
339 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
343 case Opt_uqnoenforce:
344 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
345 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
349 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
352 case Opt_pqnoenforce:
353 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
354 mp->m_qflags &= ~XFS_PQUOTA_ENFD;
358 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
361 case Opt_gqnoenforce:
362 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
363 mp->m_qflags &= ~XFS_GQUOTA_ENFD;
366 mp->m_flags |= XFS_MOUNT_DISCARD;
369 mp->m_flags &= ~XFS_MOUNT_DISCARD;
373 mp->m_flags |= XFS_MOUNT_DAX;
377 xfs_warn(mp, "unknown mount option [%s].", p);
383 * no recovery flag requires a read-only mount
385 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
386 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
387 xfs_warn(mp, "no-recovery mounts must be read-only.");
391 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
393 "sunit and swidth options incompatible with the noalign option");
397 #ifndef CONFIG_XFS_QUOTA
398 if (XFS_IS_QUOTA_RUNNING(mp)) {
399 xfs_warn(mp, "quota support not available in this kernel.");
404 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
405 xfs_warn(mp, "sunit and swidth must be specified together");
409 if (dsunit && (dswidth % dsunit != 0)) {
411 "stripe width (%d) must be a multiple of the stripe unit (%d)",
417 if (dsunit && !(mp->m_flags & XFS_MOUNT_NOALIGN)) {
419 * At this point the superblock has not been read
420 * in, therefore we do not know the block size.
421 * Before the mount call ends we will convert
424 mp->m_dalign = dsunit;
425 mp->m_swidth = dswidth;
428 if (mp->m_logbufs != -1 &&
429 mp->m_logbufs != 0 &&
430 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
431 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
432 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
433 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
436 if (mp->m_logbsize != -1 &&
437 mp->m_logbsize != 0 &&
438 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
439 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
440 !is_power_of_2(mp->m_logbsize))) {
442 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
448 if (iosizelog > XFS_MAX_IO_LOG ||
449 iosizelog < XFS_MIN_IO_LOG) {
450 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]",
451 iosizelog, XFS_MIN_IO_LOG,
456 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
457 mp->m_readio_log = iosizelog;
458 mp->m_writeio_log = iosizelog;
464 struct proc_xfs_info {
471 struct xfs_mount *mp,
474 static struct proc_xfs_info xfs_info_set[] = {
475 /* the few simple ones we can get from the mount struct */
476 { XFS_MOUNT_IKEEP, ",ikeep" },
477 { XFS_MOUNT_WSYNC, ",wsync" },
478 { XFS_MOUNT_NOALIGN, ",noalign" },
479 { XFS_MOUNT_SWALLOC, ",swalloc" },
480 { XFS_MOUNT_NOUUID, ",nouuid" },
481 { XFS_MOUNT_NORECOVERY, ",norecovery" },
482 { XFS_MOUNT_ATTR2, ",attr2" },
483 { XFS_MOUNT_FILESTREAMS, ",filestreams" },
484 { XFS_MOUNT_GRPID, ",grpid" },
485 { XFS_MOUNT_DISCARD, ",discard" },
486 { XFS_MOUNT_SMALL_INUMS, ",inode32" },
487 { XFS_MOUNT_DAX, ",dax" },
490 static struct proc_xfs_info xfs_info_unset[] = {
491 /* the few simple ones we can get from the mount struct */
492 { XFS_MOUNT_COMPAT_IOSIZE, ",largeio" },
493 { XFS_MOUNT_BARRIER, ",nobarrier" },
494 { XFS_MOUNT_SMALL_INUMS, ",inode64" },
497 struct proc_xfs_info *xfs_infop;
499 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
500 if (mp->m_flags & xfs_infop->flag)
501 seq_puts(m, xfs_infop->str);
503 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
504 if (!(mp->m_flags & xfs_infop->flag))
505 seq_puts(m, xfs_infop->str);
508 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
509 seq_printf(m, ",allocsize=%dk",
510 (int)(1 << mp->m_writeio_log) >> 10);
512 if (mp->m_logbufs > 0)
513 seq_printf(m, ",logbufs=%d", mp->m_logbufs);
514 if (mp->m_logbsize > 0)
515 seq_printf(m, ",logbsize=%dk", mp->m_logbsize >> 10);
518 seq_show_option(m, "logdev", mp->m_logname);
520 seq_show_option(m, "rtdev", mp->m_rtname);
522 if (mp->m_dalign > 0)
523 seq_printf(m, ",sunit=%d",
524 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
525 if (mp->m_swidth > 0)
526 seq_printf(m, ",swidth=%d",
527 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
529 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
530 seq_puts(m, ",usrquota");
531 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
532 seq_puts(m, ",uqnoenforce");
534 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
535 if (mp->m_qflags & XFS_PQUOTA_ENFD)
536 seq_puts(m, ",prjquota");
538 seq_puts(m, ",pqnoenforce");
540 if (mp->m_qflags & XFS_GQUOTA_ACCT) {
541 if (mp->m_qflags & XFS_GQUOTA_ENFD)
542 seq_puts(m, ",grpquota");
544 seq_puts(m, ",gqnoenforce");
547 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
548 seq_puts(m, ",noquota");
554 unsigned int blockshift)
556 unsigned int pagefactor = 1;
557 unsigned int bitshift = BITS_PER_LONG - 1;
559 /* Figure out maximum filesize, on Linux this can depend on
560 * the filesystem blocksize (on 32 bit platforms).
561 * __block_write_begin does this in an [unsigned] long...
562 * page->index << (PAGE_SHIFT - bbits)
563 * So, for page sized blocks (4K on 32 bit platforms),
564 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
565 * (((u64)PAGE_SIZE << (BITS_PER_LONG-1))-1)
566 * but for smaller blocksizes it is less (bbits = log2 bsize).
567 * Note1: get_block_t takes a long (implicit cast from above)
568 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
569 * can optionally convert the [unsigned] long from above into
570 * an [unsigned] long long.
573 #if BITS_PER_LONG == 32
574 # if defined(CONFIG_LBDAF)
575 ASSERT(sizeof(sector_t) == 8);
576 pagefactor = PAGE_SIZE;
577 bitshift = BITS_PER_LONG;
579 pagefactor = PAGE_SIZE >> (PAGE_SHIFT - blockshift);
583 return (((__uint64_t)pagefactor) << bitshift) - 1;
587 * Set parameters for inode allocation heuristics, taking into account
588 * filesystem size and inode32/inode64 mount options; i.e. specifically
589 * whether or not XFS_MOUNT_SMALL_INUMS is set.
591 * Inode allocation patterns are altered only if inode32 is requested
592 * (XFS_MOUNT_SMALL_INUMS), and the filesystem is sufficiently large.
593 * If altered, XFS_MOUNT_32BITINODES is set as well.
595 * An agcount independent of that in the mount structure is provided
596 * because in the growfs case, mp->m_sb.sb_agcount is not yet updated
597 * to the potentially higher ag count.
599 * Returns the maximum AG index which may contain inodes.
603 struct xfs_mount *mp,
604 xfs_agnumber_t agcount)
606 xfs_agnumber_t index;
607 xfs_agnumber_t maxagi = 0;
608 xfs_sb_t *sbp = &mp->m_sb;
609 xfs_agnumber_t max_metadata;
614 * Calculate how much should be reserved for inodes to meet
615 * the max inode percentage. Used only for inode32.
617 if (mp->m_maxicount) {
620 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
622 icount += sbp->sb_agblocks - 1;
623 do_div(icount, sbp->sb_agblocks);
624 max_metadata = icount;
626 max_metadata = agcount;
629 /* Get the last possible inode in the filesystem */
630 agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks - 1, 0);
631 ino = XFS_AGINO_TO_INO(mp, agcount - 1, agino);
634 * If user asked for no more than 32-bit inodes, and the fs is
635 * sufficiently large, set XFS_MOUNT_32BITINODES if we must alter
636 * the allocator to accommodate the request.
638 if ((mp->m_flags & XFS_MOUNT_SMALL_INUMS) && ino > XFS_MAXINUMBER_32)
639 mp->m_flags |= XFS_MOUNT_32BITINODES;
641 mp->m_flags &= ~XFS_MOUNT_32BITINODES;
643 for (index = 0; index < agcount; index++) {
644 struct xfs_perag *pag;
646 ino = XFS_AGINO_TO_INO(mp, index, agino);
648 pag = xfs_perag_get(mp, index);
650 if (mp->m_flags & XFS_MOUNT_32BITINODES) {
651 if (ino > XFS_MAXINUMBER_32) {
652 pag->pagi_inodeok = 0;
653 pag->pagf_metadata = 0;
655 pag->pagi_inodeok = 1;
657 if (index < max_metadata)
658 pag->pagf_metadata = 1;
660 pag->pagf_metadata = 0;
663 pag->pagi_inodeok = 1;
664 pag->pagf_metadata = 0;
670 return (mp->m_flags & XFS_MOUNT_32BITINODES) ? maxagi : agcount;
677 struct block_device **bdevp)
681 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
683 if (IS_ERR(*bdevp)) {
684 error = PTR_ERR(*bdevp);
685 xfs_warn(mp, "Invalid device [%s], error=%d", name, error);
693 struct block_device *bdev)
696 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
700 xfs_blkdev_issue_flush(
701 xfs_buftarg_t *buftarg)
703 blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL);
708 struct xfs_mount *mp)
710 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
711 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
712 xfs_free_buftarg(mp, mp->m_logdev_targp);
713 xfs_blkdev_put(logdev);
715 if (mp->m_rtdev_targp) {
716 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
717 xfs_free_buftarg(mp, mp->m_rtdev_targp);
718 xfs_blkdev_put(rtdev);
720 xfs_free_buftarg(mp, mp->m_ddev_targp);
724 * The file system configurations are:
725 * (1) device (partition) with data and internal log
726 * (2) logical volume with data and log subvolumes.
727 * (3) logical volume with data, log, and realtime subvolumes.
729 * We only have to handle opening the log and realtime volumes here if
730 * they are present. The data subvolume has already been opened by
731 * get_sb_bdev() and is stored in sb->s_bdev.
735 struct xfs_mount *mp)
737 struct block_device *ddev = mp->m_super->s_bdev;
738 struct block_device *logdev = NULL, *rtdev = NULL;
742 * Open real time and log devices - order is important.
745 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
751 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
753 goto out_close_logdev;
755 if (rtdev == ddev || rtdev == logdev) {
757 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
759 goto out_close_rtdev;
764 * Setup xfs_mount buffer target pointers
767 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev);
768 if (!mp->m_ddev_targp)
769 goto out_close_rtdev;
772 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev);
773 if (!mp->m_rtdev_targp)
774 goto out_free_ddev_targ;
777 if (logdev && logdev != ddev) {
778 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev);
779 if (!mp->m_logdev_targp)
780 goto out_free_rtdev_targ;
782 mp->m_logdev_targp = mp->m_ddev_targp;
788 if (mp->m_rtdev_targp)
789 xfs_free_buftarg(mp, mp->m_rtdev_targp);
791 xfs_free_buftarg(mp, mp->m_ddev_targp);
793 xfs_blkdev_put(rtdev);
795 if (logdev && logdev != ddev)
796 xfs_blkdev_put(logdev);
802 * Setup xfs_mount buffer target pointers based on superblock
806 struct xfs_mount *mp)
810 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize);
814 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
815 unsigned int log_sector_size = BBSIZE;
817 if (xfs_sb_version_hassector(&mp->m_sb))
818 log_sector_size = mp->m_sb.sb_logsectsize;
819 error = xfs_setsize_buftarg(mp->m_logdev_targp,
824 if (mp->m_rtdev_targp) {
825 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
826 mp->m_sb.sb_sectsize);
835 xfs_init_mount_workqueues(
836 struct xfs_mount *mp)
838 mp->m_buf_workqueue = alloc_workqueue("xfs-buf/%s",
839 WQ_MEM_RECLAIM|WQ_FREEZABLE, 1, mp->m_fsname);
840 if (!mp->m_buf_workqueue)
843 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s",
844 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
845 if (!mp->m_data_workqueue)
846 goto out_destroy_buf;
848 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
849 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
850 if (!mp->m_unwritten_workqueue)
851 goto out_destroy_data_iodone_queue;
853 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s",
854 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
855 if (!mp->m_cil_workqueue)
856 goto out_destroy_unwritten;
858 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
859 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
860 if (!mp->m_reclaim_workqueue)
861 goto out_destroy_cil;
863 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
864 WQ_MEM_RECLAIM|WQ_FREEZABLE|WQ_HIGHPRI, 0,
866 if (!mp->m_log_workqueue)
867 goto out_destroy_reclaim;
869 mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s",
870 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
871 if (!mp->m_eofblocks_workqueue)
872 goto out_destroy_log;
877 destroy_workqueue(mp->m_log_workqueue);
879 destroy_workqueue(mp->m_reclaim_workqueue);
881 destroy_workqueue(mp->m_cil_workqueue);
882 out_destroy_unwritten:
883 destroy_workqueue(mp->m_unwritten_workqueue);
884 out_destroy_data_iodone_queue:
885 destroy_workqueue(mp->m_data_workqueue);
887 destroy_workqueue(mp->m_buf_workqueue);
893 xfs_destroy_mount_workqueues(
894 struct xfs_mount *mp)
896 destroy_workqueue(mp->m_eofblocks_workqueue);
897 destroy_workqueue(mp->m_log_workqueue);
898 destroy_workqueue(mp->m_reclaim_workqueue);
899 destroy_workqueue(mp->m_cil_workqueue);
900 destroy_workqueue(mp->m_data_workqueue);
901 destroy_workqueue(mp->m_unwritten_workqueue);
902 destroy_workqueue(mp->m_buf_workqueue);
906 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
907 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
908 * for IO to complete so that we effectively throttle multiple callers to the
909 * rate at which IO is completing.
913 struct xfs_mount *mp)
915 struct super_block *sb = mp->m_super;
917 if (down_read_trylock(&sb->s_umount)) {
919 up_read(&sb->s_umount);
923 /* Catch misguided souls that try to use this interface on XFS */
924 STATIC struct inode *
926 struct super_block *sb)
933 * Now that the generic code is guaranteed not to be accessing
934 * the linux inode, we can inactivate and reclaim the inode.
937 xfs_fs_destroy_inode(
940 struct xfs_inode *ip = XFS_I(inode);
942 trace_xfs_destroy_inode(ip);
944 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
945 XFS_STATS_INC(ip->i_mount, vn_rele);
946 XFS_STATS_INC(ip->i_mount, vn_remove);
950 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
951 XFS_STATS_INC(ip->i_mount, vn_reclaim);
954 * We should never get here with one of the reclaim flags already set.
956 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
957 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
960 * We always use background reclaim here because even if the
961 * inode is clean, it still may be under IO and hence we have
962 * to take the flush lock. The background reclaim path handles
963 * this more efficiently than we can here, so simply let background
964 * reclaim tear down all inodes.
966 xfs_inode_set_reclaim_tag(ip);
970 * Slab object creation initialisation for the XFS inode.
971 * This covers only the idempotent fields in the XFS inode;
972 * all other fields need to be initialised on allocation
973 * from the slab. This avoids the need to repeatedly initialise
974 * fields in the xfs inode that left in the initialise state
975 * when freeing the inode.
978 xfs_fs_inode_init_once(
981 struct xfs_inode *ip = inode;
983 memset(ip, 0, sizeof(struct xfs_inode));
986 inode_init_once(VFS_I(ip));
989 atomic_set(&ip->i_pincount, 0);
990 spin_lock_init(&ip->i_flags_lock);
992 mrlock_init(&ip->i_mmaplock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
993 "xfsino", ip->i_ino);
994 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
995 "xfsino", ip->i_ino);
999 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1000 * serialised against cache hits here via the inode->i_lock and igrab() in
1001 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1002 * racing with us, and it avoids needing to grab a spinlock here for every inode
1003 * we drop the final reference on.
1007 struct inode *inode)
1009 struct xfs_inode *ip = XFS_I(inode);
1012 * If this unlinked inode is in the middle of recovery, don't
1013 * drop the inode just yet; log recovery will take care of
1014 * that. See the comment for this inode flag.
1016 if (ip->i_flags & XFS_IRECOVERY) {
1017 ASSERT(ip->i_mount->m_log->l_flags & XLOG_RECOVERY_NEEDED);
1021 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);
1026 struct xfs_mount *mp)
1028 kfree(mp->m_fsname);
1029 kfree(mp->m_rtname);
1030 kfree(mp->m_logname);
1035 struct super_block *sb,
1038 struct xfs_mount *mp = XFS_M(sb);
1041 * Doing anything during the async pass would be counterproductive.
1046 xfs_log_force(mp, XFS_LOG_SYNC);
1049 * The disk must be active because we're syncing.
1050 * We schedule log work now (now that the disk is
1051 * active) instead of later (when it might not be).
1053 flush_delayed_work(&mp->m_log->l_work);
1061 struct dentry *dentry,
1062 struct kstatfs *statp)
1064 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1065 xfs_sb_t *sbp = &mp->m_sb;
1066 struct xfs_inode *ip = XFS_I(d_inode(dentry));
1067 __uint64_t fakeinos, id;
1070 __uint64_t fdblocks;
1074 statp->f_type = XFS_SB_MAGIC;
1075 statp->f_namelen = MAXNAMELEN - 1;
1077 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1078 statp->f_fsid.val[0] = (u32)id;
1079 statp->f_fsid.val[1] = (u32)(id >> 32);
1081 icount = percpu_counter_sum(&mp->m_icount);
1082 ifree = percpu_counter_sum(&mp->m_ifree);
1083 fdblocks = percpu_counter_sum(&mp->m_fdblocks);
1085 spin_lock(&mp->m_sb_lock);
1086 statp->f_bsize = sbp->sb_blocksize;
1087 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1088 statp->f_blocks = sbp->sb_dblocks - lsize;
1089 spin_unlock(&mp->m_sb_lock);
1091 statp->f_bfree = fdblocks - mp->m_alloc_set_aside;
1092 statp->f_bavail = statp->f_bfree;
1094 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1095 statp->f_files = MIN(icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1096 if (mp->m_maxicount)
1097 statp->f_files = min_t(typeof(statp->f_files),
1101 /* If sb_icount overshot maxicount, report actual allocation */
1102 statp->f_files = max_t(typeof(statp->f_files),
1106 /* make sure statp->f_ffree does not underflow */
1107 ffree = statp->f_files - (icount - ifree);
1108 statp->f_ffree = max_t(__int64_t, ffree, 0);
1111 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
1112 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) ==
1113 (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))
1114 xfs_qm_statvfs(ip, statp);
1119 xfs_save_resvblks(struct xfs_mount *mp)
1121 __uint64_t resblks = 0;
1123 mp->m_resblks_save = mp->m_resblks;
1124 xfs_reserve_blocks(mp, &resblks, NULL);
1128 xfs_restore_resvblks(struct xfs_mount *mp)
1132 if (mp->m_resblks_save) {
1133 resblks = mp->m_resblks_save;
1134 mp->m_resblks_save = 0;
1136 resblks = xfs_default_resblks(mp);
1138 xfs_reserve_blocks(mp, &resblks, NULL);
1142 * Trigger writeback of all the dirty metadata in the file system.
1144 * This ensures that the metadata is written to their location on disk rather
1145 * than just existing in transactions in the log. This means after a quiesce
1146 * there is no log replay required to write the inodes to disk - this is the
1147 * primary difference between a sync and a quiesce.
1149 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1150 * it is started again when appropriate.
1154 struct xfs_mount *mp)
1158 /* wait for all modifications to complete */
1159 while (atomic_read(&mp->m_active_trans) > 0)
1162 /* force the log to unpin objects from the now complete transactions */
1163 xfs_log_force(mp, XFS_LOG_SYNC);
1165 /* reclaim inodes to do any IO before the freeze completes */
1166 xfs_reclaim_inodes(mp, 0);
1167 xfs_reclaim_inodes(mp, SYNC_WAIT);
1169 /* Push the superblock and write an unmount record */
1170 error = xfs_log_sbcount(mp);
1172 xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
1173 "Frozen image may not be consistent.");
1175 * Just warn here till VFS can correctly support
1176 * read-only remount without racing.
1178 WARN_ON(atomic_read(&mp->m_active_trans) != 0);
1180 xfs_log_quiesce(mp);
1184 xfs_test_remount_options(
1185 struct super_block *sb,
1186 struct xfs_mount *mp,
1190 struct xfs_mount *tmp_mp;
1192 tmp_mp = kmem_zalloc(sizeof(*tmp_mp), KM_MAYFAIL);
1196 tmp_mp->m_super = sb;
1197 error = xfs_parseargs(tmp_mp, options);
1198 xfs_free_fsname(tmp_mp);
1206 struct super_block *sb,
1210 struct xfs_mount *mp = XFS_M(sb);
1211 xfs_sb_t *sbp = &mp->m_sb;
1212 substring_t args[MAX_OPT_ARGS];
1216 /* First, check for complete junk; i.e. invalid options */
1217 error = xfs_test_remount_options(sb, mp, options);
1221 sync_filesystem(sb);
1222 while ((p = strsep(&options, ",")) != NULL) {
1228 token = match_token(p, tokens, args);
1231 mp->m_flags |= XFS_MOUNT_BARRIER;
1234 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1237 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
1238 mp->m_maxagi = xfs_set_inode_alloc(mp, sbp->sb_agcount);
1241 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
1242 mp->m_maxagi = xfs_set_inode_alloc(mp, sbp->sb_agcount);
1246 * Logically we would return an error here to prevent
1247 * users from believing they might have changed
1248 * mount options using remount which can't be changed.
1250 * But unfortunately mount(8) adds all options from
1251 * mtab and fstab to the mount arguments in some cases
1252 * so we can't blindly reject options, but have to
1253 * check for each specified option if it actually
1254 * differs from the currently set option and only
1255 * reject it if that's the case.
1257 * Until that is implemented we return success for
1258 * every remount request, and silently ignore all
1259 * options that we can't actually change.
1263 "mount option \"%s\" not supported for remount", p);
1272 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1273 if (mp->m_flags & XFS_MOUNT_NORECOVERY) {
1275 "ro->rw transition prohibited on norecovery mount");
1279 if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
1280 xfs_sb_has_ro_compat_feature(sbp,
1281 XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
1283 "ro->rw transition prohibited on unknown (0x%x) ro-compat filesystem",
1284 (sbp->sb_features_ro_compat &
1285 XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
1289 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1292 * If this is the first remount to writeable state we
1293 * might have some superblock changes to update.
1295 if (mp->m_update_sb) {
1296 error = xfs_sync_sb(mp, false);
1298 xfs_warn(mp, "failed to write sb changes");
1301 mp->m_update_sb = false;
1305 * Fill out the reserve pool if it is empty. Use the stashed
1306 * value if it is non-zero, otherwise go with the default.
1308 xfs_restore_resvblks(mp);
1309 xfs_log_work_queue(mp);
1310 xfs_queue_eofblocks(mp);
1314 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1316 * Before we sync the metadata, we need to free up the reserve
1317 * block pool so that the used block count in the superblock on
1318 * disk is correct at the end of the remount. Stash the current
1319 * reserve pool size so that if we get remounted rw, we can
1320 * return it to the same size.
1322 xfs_save_resvblks(mp);
1325 * Cancel background eofb scanning so it cannot race with the
1326 * final log force+buftarg wait and deadlock the remount.
1328 cancel_delayed_work_sync(&mp->m_eofblocks_work);
1330 xfs_quiesce_attr(mp);
1331 mp->m_flags |= XFS_MOUNT_RDONLY;
1338 * Second stage of a freeze. The data is already frozen so we only
1339 * need to take care of the metadata. Once that's done sync the superblock
1340 * to the log to dirty it in case of a crash while frozen. This ensures that we
1341 * will recover the unlinked inode lists on the next mount.
1345 struct super_block *sb)
1347 struct xfs_mount *mp = XFS_M(sb);
1349 xfs_save_resvblks(mp);
1350 xfs_quiesce_attr(mp);
1351 return xfs_sync_sb(mp, true);
1356 struct super_block *sb)
1358 struct xfs_mount *mp = XFS_M(sb);
1360 xfs_restore_resvblks(mp);
1361 xfs_log_work_queue(mp);
1366 xfs_fs_show_options(
1368 struct dentry *root)
1370 return xfs_showargs(XFS_M(root->d_sb), m);
1374 * This function fills in xfs_mount_t fields based on mount args.
1375 * Note: the superblock _has_ now been read in.
1379 struct xfs_mount *mp)
1381 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1383 /* Fail a mount where the logbuf is smaller than the log stripe */
1384 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1385 if (mp->m_logbsize <= 0 &&
1386 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1387 mp->m_logbsize = mp->m_sb.sb_logsunit;
1388 } else if (mp->m_logbsize > 0 &&
1389 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1391 "logbuf size must be greater than or equal to log stripe size");
1395 /* Fail a mount if the logbuf is larger than 32K */
1396 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1398 "logbuf size for version 1 logs must be 16K or 32K");
1404 * V5 filesystems always use attr2 format for attributes.
1406 if (xfs_sb_version_hascrc(&mp->m_sb) &&
1407 (mp->m_flags & XFS_MOUNT_NOATTR2)) {
1408 xfs_warn(mp, "Cannot mount a V5 filesystem as noattr2. "
1409 "attr2 is always enabled for V5 filesystems.");
1414 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1415 * told by noattr2 to turn it off
1417 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1418 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1419 mp->m_flags |= XFS_MOUNT_ATTR2;
1422 * prohibit r/w mounts of read-only filesystems
1424 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1426 "cannot mount a read-only filesystem as read-write");
1430 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
1431 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) &&
1432 !xfs_sb_version_has_pquotino(&mp->m_sb)) {
1434 "Super block does not support project and group quota together");
1442 xfs_init_percpu_counters(
1443 struct xfs_mount *mp)
1447 error = percpu_counter_init(&mp->m_icount, 0, GFP_KERNEL);
1451 error = percpu_counter_init(&mp->m_ifree, 0, GFP_KERNEL);
1455 error = percpu_counter_init(&mp->m_fdblocks, 0, GFP_KERNEL);
1462 percpu_counter_destroy(&mp->m_ifree);
1464 percpu_counter_destroy(&mp->m_icount);
1469 xfs_reinit_percpu_counters(
1470 struct xfs_mount *mp)
1472 percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount);
1473 percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree);
1474 percpu_counter_set(&mp->m_fdblocks, mp->m_sb.sb_fdblocks);
1478 xfs_destroy_percpu_counters(
1479 struct xfs_mount *mp)
1481 percpu_counter_destroy(&mp->m_icount);
1482 percpu_counter_destroy(&mp->m_ifree);
1483 percpu_counter_destroy(&mp->m_fdblocks);
1488 struct super_block *sb,
1493 struct xfs_mount *mp = NULL;
1494 int flags = 0, error = -ENOMEM;
1496 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1500 spin_lock_init(&mp->m_sb_lock);
1501 mutex_init(&mp->m_growlock);
1502 atomic_set(&mp->m_active_trans, 0);
1503 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1504 INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
1505 mp->m_kobj.kobject.kset = xfs_kset;
1510 error = xfs_parseargs(mp, (char *)data);
1512 goto out_free_fsname;
1514 sb_min_blocksize(sb, BBSIZE);
1515 sb->s_xattr = xfs_xattr_handlers;
1516 sb->s_export_op = &xfs_export_operations;
1517 #ifdef CONFIG_XFS_QUOTA
1518 sb->s_qcop = &xfs_quotactl_operations;
1519 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
1521 sb->s_op = &xfs_super_operations;
1524 flags |= XFS_MFSI_QUIET;
1526 error = xfs_open_devices(mp);
1528 goto out_free_fsname;
1530 error = xfs_init_mount_workqueues(mp);
1532 goto out_close_devices;
1534 error = xfs_init_percpu_counters(mp);
1536 goto out_destroy_workqueues;
1538 /* Allocate stats memory before we do operations that might use it */
1539 mp->m_stats.xs_stats = alloc_percpu(struct xfsstats);
1540 if (!mp->m_stats.xs_stats) {
1542 goto out_destroy_counters;
1545 error = xfs_readsb(mp, flags);
1547 goto out_free_stats;
1549 error = xfs_finish_flags(mp);
1553 error = xfs_setup_devices(mp);
1557 error = xfs_filestream_mount(mp);
1562 * we must configure the block size in the superblock before we run the
1563 * full mount process as the mount process can lookup and cache inodes.
1565 sb->s_magic = XFS_SB_MAGIC;
1566 sb->s_blocksize = mp->m_sb.sb_blocksize;
1567 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1568 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1569 sb->s_max_links = XFS_MAXLINK;
1570 sb->s_time_gran = 1;
1571 set_posix_acl_flag(sb);
1573 /* version 5 superblocks support inode version counters. */
1574 if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
1575 sb->s_flags |= MS_I_VERSION;
1577 if (mp->m_flags & XFS_MOUNT_DAX) {
1579 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1581 error = bdev_dax_supported(sb, sb->s_blocksize);
1584 "DAX unsupported by block device. Turning off DAX.");
1585 mp->m_flags &= ~XFS_MOUNT_DAX;
1589 if (xfs_sb_version_hasrmapbt(&mp->m_sb)) {
1590 if (mp->m_sb.sb_rblocks) {
1592 "EXPERIMENTAL reverse mapping btree not compatible with realtime device!");
1594 goto out_filestream_unmount;
1597 "EXPERIMENTAL reverse mapping btree feature enabled. Use at your own risk!");
1600 error = xfs_mountfs(mp);
1602 goto out_filestream_unmount;
1604 root = igrab(VFS_I(mp->m_rootip));
1609 sb->s_root = d_make_root(root);
1617 out_filestream_unmount:
1618 xfs_filestream_unmount(mp);
1622 free_percpu(mp->m_stats.xs_stats);
1623 out_destroy_counters:
1624 xfs_destroy_percpu_counters(mp);
1625 out_destroy_workqueues:
1626 xfs_destroy_mount_workqueues(mp);
1628 xfs_close_devices(mp);
1630 xfs_free_fsname(mp);
1636 xfs_filestream_unmount(mp);
1643 struct super_block *sb)
1645 struct xfs_mount *mp = XFS_M(sb);
1647 xfs_notice(mp, "Unmounting Filesystem");
1648 xfs_filestream_unmount(mp);
1652 free_percpu(mp->m_stats.xs_stats);
1653 xfs_destroy_percpu_counters(mp);
1654 xfs_destroy_mount_workqueues(mp);
1655 xfs_close_devices(mp);
1656 xfs_free_fsname(mp);
1660 STATIC struct dentry *
1662 struct file_system_type *fs_type,
1664 const char *dev_name,
1667 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1671 xfs_fs_nr_cached_objects(
1672 struct super_block *sb,
1673 struct shrink_control *sc)
1675 return xfs_reclaim_inodes_count(XFS_M(sb));
1679 xfs_fs_free_cached_objects(
1680 struct super_block *sb,
1681 struct shrink_control *sc)
1683 return xfs_reclaim_inodes_nr(XFS_M(sb), sc->nr_to_scan);
1686 static const struct super_operations xfs_super_operations = {
1687 .alloc_inode = xfs_fs_alloc_inode,
1688 .destroy_inode = xfs_fs_destroy_inode,
1689 .drop_inode = xfs_fs_drop_inode,
1690 .put_super = xfs_fs_put_super,
1691 .sync_fs = xfs_fs_sync_fs,
1692 .freeze_fs = xfs_fs_freeze,
1693 .unfreeze_fs = xfs_fs_unfreeze,
1694 .statfs = xfs_fs_statfs,
1695 .remount_fs = xfs_fs_remount,
1696 .show_options = xfs_fs_show_options,
1697 .nr_cached_objects = xfs_fs_nr_cached_objects,
1698 .free_cached_objects = xfs_fs_free_cached_objects,
1701 static struct file_system_type xfs_fs_type = {
1702 .owner = THIS_MODULE,
1704 .mount = xfs_fs_mount,
1705 .kill_sb = kill_block_super,
1706 .fs_flags = FS_REQUIRES_DEV,
1708 MODULE_ALIAS_FS("xfs");
1711 xfs_init_zones(void)
1713 xfs_ioend_bioset = bioset_create(4 * MAX_BUF_PER_PAGE,
1714 offsetof(struct xfs_ioend, io_inline_bio));
1715 if (!xfs_ioend_bioset)
1718 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1720 if (!xfs_log_ticket_zone)
1721 goto out_free_ioend_bioset;
1723 xfs_bmap_free_item_zone = kmem_zone_init(
1724 sizeof(struct xfs_extent_free_item),
1725 "xfs_bmap_free_item");
1726 if (!xfs_bmap_free_item_zone)
1727 goto out_destroy_log_ticket_zone;
1729 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1731 if (!xfs_btree_cur_zone)
1732 goto out_destroy_bmap_free_item_zone;
1734 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1736 if (!xfs_da_state_zone)
1737 goto out_destroy_btree_cur_zone;
1739 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1740 if (!xfs_ifork_zone)
1741 goto out_destroy_da_state_zone;
1743 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1744 if (!xfs_trans_zone)
1745 goto out_destroy_ifork_zone;
1747 xfs_log_item_desc_zone =
1748 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1749 "xfs_log_item_desc");
1750 if (!xfs_log_item_desc_zone)
1751 goto out_destroy_trans_zone;
1754 * The size of the zone allocated buf log item is the maximum
1755 * size possible under XFS. This wastes a little bit of memory,
1756 * but it is much faster.
1758 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item),
1760 if (!xfs_buf_item_zone)
1761 goto out_destroy_log_item_desc_zone;
1763 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1764 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1765 sizeof(xfs_extent_t))), "xfs_efd_item");
1767 goto out_destroy_buf_item_zone;
1769 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1770 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1771 sizeof(xfs_extent_t))), "xfs_efi_item");
1773 goto out_destroy_efd_zone;
1776 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1777 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD |
1778 KM_ZONE_ACCOUNT, xfs_fs_inode_init_once);
1779 if (!xfs_inode_zone)
1780 goto out_destroy_efi_zone;
1783 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1784 KM_ZONE_SPREAD, NULL);
1786 goto out_destroy_inode_zone;
1787 xfs_icreate_zone = kmem_zone_init(sizeof(struct xfs_icreate_item),
1789 if (!xfs_icreate_zone)
1790 goto out_destroy_ili_zone;
1792 xfs_rud_zone = kmem_zone_init(sizeof(struct xfs_rud_log_item),
1795 goto out_destroy_icreate_zone;
1797 xfs_rui_zone = kmem_zone_init(
1798 xfs_rui_log_item_sizeof(XFS_RUI_MAX_FAST_EXTENTS),
1801 goto out_destroy_rud_zone;
1803 xfs_cud_zone = kmem_zone_init(sizeof(struct xfs_cud_log_item),
1806 goto out_destroy_rui_zone;
1808 xfs_cui_zone = kmem_zone_init(
1809 xfs_cui_log_item_sizeof(XFS_CUI_MAX_FAST_EXTENTS),
1812 goto out_destroy_cud_zone;
1814 xfs_bud_zone = kmem_zone_init(sizeof(struct xfs_bud_log_item),
1817 goto out_destroy_cui_zone;
1819 xfs_bui_zone = kmem_zone_init(
1820 xfs_bui_log_item_sizeof(XFS_BUI_MAX_FAST_EXTENTS),
1823 goto out_destroy_bud_zone;
1827 out_destroy_bud_zone:
1828 kmem_zone_destroy(xfs_bud_zone);
1829 out_destroy_cui_zone:
1830 kmem_zone_destroy(xfs_cui_zone);
1831 out_destroy_cud_zone:
1832 kmem_zone_destroy(xfs_cud_zone);
1833 out_destroy_rui_zone:
1834 kmem_zone_destroy(xfs_rui_zone);
1835 out_destroy_rud_zone:
1836 kmem_zone_destroy(xfs_rud_zone);
1837 out_destroy_icreate_zone:
1838 kmem_zone_destroy(xfs_icreate_zone);
1839 out_destroy_ili_zone:
1840 kmem_zone_destroy(xfs_ili_zone);
1841 out_destroy_inode_zone:
1842 kmem_zone_destroy(xfs_inode_zone);
1843 out_destroy_efi_zone:
1844 kmem_zone_destroy(xfs_efi_zone);
1845 out_destroy_efd_zone:
1846 kmem_zone_destroy(xfs_efd_zone);
1847 out_destroy_buf_item_zone:
1848 kmem_zone_destroy(xfs_buf_item_zone);
1849 out_destroy_log_item_desc_zone:
1850 kmem_zone_destroy(xfs_log_item_desc_zone);
1851 out_destroy_trans_zone:
1852 kmem_zone_destroy(xfs_trans_zone);
1853 out_destroy_ifork_zone:
1854 kmem_zone_destroy(xfs_ifork_zone);
1855 out_destroy_da_state_zone:
1856 kmem_zone_destroy(xfs_da_state_zone);
1857 out_destroy_btree_cur_zone:
1858 kmem_zone_destroy(xfs_btree_cur_zone);
1859 out_destroy_bmap_free_item_zone:
1860 kmem_zone_destroy(xfs_bmap_free_item_zone);
1861 out_destroy_log_ticket_zone:
1862 kmem_zone_destroy(xfs_log_ticket_zone);
1863 out_free_ioend_bioset:
1864 bioset_free(xfs_ioend_bioset);
1870 xfs_destroy_zones(void)
1873 * Make sure all delayed rcu free are flushed before we
1877 kmem_zone_destroy(xfs_bui_zone);
1878 kmem_zone_destroy(xfs_bud_zone);
1879 kmem_zone_destroy(xfs_cui_zone);
1880 kmem_zone_destroy(xfs_cud_zone);
1881 kmem_zone_destroy(xfs_rui_zone);
1882 kmem_zone_destroy(xfs_rud_zone);
1883 kmem_zone_destroy(xfs_icreate_zone);
1884 kmem_zone_destroy(xfs_ili_zone);
1885 kmem_zone_destroy(xfs_inode_zone);
1886 kmem_zone_destroy(xfs_efi_zone);
1887 kmem_zone_destroy(xfs_efd_zone);
1888 kmem_zone_destroy(xfs_buf_item_zone);
1889 kmem_zone_destroy(xfs_log_item_desc_zone);
1890 kmem_zone_destroy(xfs_trans_zone);
1891 kmem_zone_destroy(xfs_ifork_zone);
1892 kmem_zone_destroy(xfs_da_state_zone);
1893 kmem_zone_destroy(xfs_btree_cur_zone);
1894 kmem_zone_destroy(xfs_bmap_free_item_zone);
1895 kmem_zone_destroy(xfs_log_ticket_zone);
1896 bioset_free(xfs_ioend_bioset);
1900 xfs_init_workqueues(void)
1903 * The allocation workqueue can be used in memory reclaim situations
1904 * (writepage path), and parallelism is only limited by the number of
1905 * AGs in all the filesystems mounted. Hence use the default large
1906 * max_active value for this workqueue.
1908 xfs_alloc_wq = alloc_workqueue("xfsalloc",
1909 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0);
1917 xfs_destroy_workqueues(void)
1919 destroy_workqueue(xfs_alloc_wq);
1927 xfs_check_ondisk_structs();
1929 printk(KERN_INFO XFS_VERSION_STRING " with "
1930 XFS_BUILD_OPTIONS " enabled\n");
1932 xfs_extent_free_init_defer_op();
1933 xfs_rmap_update_init_defer_op();
1934 xfs_refcount_update_init_defer_op();
1935 xfs_bmap_update_init_defer_op();
1939 error = xfs_init_zones();
1943 error = xfs_init_workqueues();
1945 goto out_destroy_zones;
1947 error = xfs_mru_cache_init();
1949 goto out_destroy_wq;
1951 error = xfs_buf_init();
1953 goto out_mru_cache_uninit;
1955 error = xfs_init_procfs();
1957 goto out_buf_terminate;
1959 error = xfs_sysctl_register();
1961 goto out_cleanup_procfs;
1963 xfs_kset = kset_create_and_add("xfs", NULL, fs_kobj);
1966 goto out_sysctl_unregister;
1969 xfsstats.xs_kobj.kobject.kset = xfs_kset;
1971 xfsstats.xs_stats = alloc_percpu(struct xfsstats);
1972 if (!xfsstats.xs_stats) {
1974 goto out_kset_unregister;
1977 error = xfs_sysfs_init(&xfsstats.xs_kobj, &xfs_stats_ktype, NULL,
1980 goto out_free_stats;
1983 xfs_dbg_kobj.kobject.kset = xfs_kset;
1984 error = xfs_sysfs_init(&xfs_dbg_kobj, &xfs_dbg_ktype, NULL, "debug");
1986 goto out_remove_stats_kobj;
1989 error = xfs_qm_init();
1991 goto out_remove_dbg_kobj;
1993 error = register_filesystem(&xfs_fs_type);
2000 out_remove_dbg_kobj:
2002 xfs_sysfs_del(&xfs_dbg_kobj);
2003 out_remove_stats_kobj:
2005 xfs_sysfs_del(&xfsstats.xs_kobj);
2007 free_percpu(xfsstats.xs_stats);
2008 out_kset_unregister:
2009 kset_unregister(xfs_kset);
2010 out_sysctl_unregister:
2011 xfs_sysctl_unregister();
2013 xfs_cleanup_procfs();
2015 xfs_buf_terminate();
2016 out_mru_cache_uninit:
2017 xfs_mru_cache_uninit();
2019 xfs_destroy_workqueues();
2021 xfs_destroy_zones();
2030 unregister_filesystem(&xfs_fs_type);
2032 xfs_sysfs_del(&xfs_dbg_kobj);
2034 xfs_sysfs_del(&xfsstats.xs_kobj);
2035 free_percpu(xfsstats.xs_stats);
2036 kset_unregister(xfs_kset);
2037 xfs_sysctl_unregister();
2038 xfs_cleanup_procfs();
2039 xfs_buf_terminate();
2040 xfs_mru_cache_uninit();
2041 xfs_destroy_workqueues();
2042 xfs_destroy_zones();
2043 xfs_uuid_table_free();
2046 module_init(init_xfs_fs);
2047 module_exit(exit_xfs_fs);
2049 MODULE_AUTHOR("Silicon Graphics, Inc.");
2050 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
2051 MODULE_LICENSE("GPL");