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/init.h>
56 #include <linux/slab.h>
57 #include <linux/mount.h>
58 #include <linux/mempool.h>
59 #include <linux/writeback.h>
60 #include <linux/kthread.h>
61 #include <linux/freezer.h>
62 #include <linux/parser.h>
64 static const struct super_operations xfs_super_operations;
65 struct bio_set *xfs_ioend_bioset;
67 static struct kset *xfs_kset; /* top-level xfs sysfs dir */
69 static struct xfs_kobj xfs_dbg_kobj; /* global debug sysfs attrs */
73 * Table driven mount option parser.
76 Opt_logbufs, Opt_logbsize, Opt_logdev, Opt_rtdev, Opt_biosize,
77 Opt_wsync, Opt_noalign, Opt_swalloc, Opt_sunit, Opt_swidth, Opt_nouuid,
78 Opt_mtpt, Opt_grpid, Opt_nogrpid, Opt_bsdgroups, Opt_sysvgroups,
79 Opt_allocsize, Opt_norecovery, Opt_barrier, Opt_nobarrier,
80 Opt_inode64, Opt_inode32, Opt_ikeep, Opt_noikeep,
81 Opt_largeio, Opt_nolargeio, Opt_attr2, Opt_noattr2, Opt_filestreams,
82 Opt_quota, Opt_noquota, Opt_usrquota, Opt_grpquota, Opt_prjquota,
83 Opt_uquota, Opt_gquota, Opt_pquota,
84 Opt_uqnoenforce, Opt_gqnoenforce, Opt_pqnoenforce, Opt_qnoenforce,
85 Opt_discard, Opt_nodiscard, Opt_dax, Opt_err,
88 static const match_table_t tokens = {
89 {Opt_logbufs, "logbufs=%u"}, /* number of XFS log buffers */
90 {Opt_logbsize, "logbsize=%s"}, /* size of XFS log buffers */
91 {Opt_logdev, "logdev=%s"}, /* log device */
92 {Opt_rtdev, "rtdev=%s"}, /* realtime I/O device */
93 {Opt_biosize, "biosize=%u"}, /* log2 of preferred buffered io size */
94 {Opt_wsync, "wsync"}, /* safe-mode nfs compatible mount */
95 {Opt_noalign, "noalign"}, /* turn off stripe alignment */
96 {Opt_swalloc, "swalloc"}, /* turn on stripe width allocation */
97 {Opt_sunit, "sunit=%u"}, /* data volume stripe unit */
98 {Opt_swidth, "swidth=%u"}, /* data volume stripe width */
99 {Opt_nouuid, "nouuid"}, /* ignore filesystem UUID */
100 {Opt_mtpt, "mtpt"}, /* filesystem mount point */
101 {Opt_grpid, "grpid"}, /* group-ID from parent directory */
102 {Opt_nogrpid, "nogrpid"}, /* group-ID from current process */
103 {Opt_bsdgroups, "bsdgroups"}, /* group-ID from parent directory */
104 {Opt_sysvgroups,"sysvgroups"}, /* group-ID from current process */
105 {Opt_allocsize, "allocsize=%s"},/* preferred allocation size */
106 {Opt_norecovery,"norecovery"}, /* don't run XFS recovery */
107 {Opt_barrier, "barrier"}, /* use writer barriers for log write and
108 * unwritten extent conversion */
109 {Opt_nobarrier, "nobarrier"}, /* .. disable */
110 {Opt_inode64, "inode64"}, /* inodes can be allocated anywhere */
111 {Opt_inode32, "inode32"}, /* inode allocation limited to
112 * XFS_MAXINUMBER_32 */
113 {Opt_ikeep, "ikeep"}, /* do not free empty inode clusters */
114 {Opt_noikeep, "noikeep"}, /* free empty inode clusters */
115 {Opt_largeio, "largeio"}, /* report large I/O sizes in stat() */
116 {Opt_nolargeio, "nolargeio"}, /* do not report large I/O sizes
118 {Opt_attr2, "attr2"}, /* do use attr2 attribute format */
119 {Opt_noattr2, "noattr2"}, /* do not use attr2 attribute format */
120 {Opt_filestreams,"filestreams"},/* use filestreams allocator */
121 {Opt_quota, "quota"}, /* disk quotas (user) */
122 {Opt_noquota, "noquota"}, /* no quotas */
123 {Opt_usrquota, "usrquota"}, /* user quota enabled */
124 {Opt_grpquota, "grpquota"}, /* group quota enabled */
125 {Opt_prjquota, "prjquota"}, /* project quota enabled */
126 {Opt_uquota, "uquota"}, /* user quota (IRIX variant) */
127 {Opt_gquota, "gquota"}, /* group quota (IRIX variant) */
128 {Opt_pquota, "pquota"}, /* project quota (IRIX variant) */
129 {Opt_uqnoenforce,"uqnoenforce"},/* user quota limit enforcement */
130 {Opt_gqnoenforce,"gqnoenforce"},/* group quota limit enforcement */
131 {Opt_pqnoenforce,"pqnoenforce"},/* project quota limit enforcement */
132 {Opt_qnoenforce, "qnoenforce"}, /* same as uqnoenforce */
133 {Opt_discard, "discard"}, /* Discard unused blocks */
134 {Opt_nodiscard, "nodiscard"}, /* Do not discard unused blocks */
136 {Opt_dax, "dax"}, /* Enable direct access to bdev pages */
142 suffix_kstrtoint(const substring_t *s, unsigned int base, int *res)
144 int last, shift_left_factor = 0, _res;
148 value = match_strdup(s);
152 last = strlen(value) - 1;
153 if (value[last] == 'K' || value[last] == 'k') {
154 shift_left_factor = 10;
157 if (value[last] == 'M' || value[last] == 'm') {
158 shift_left_factor = 20;
161 if (value[last] == 'G' || value[last] == 'g') {
162 shift_left_factor = 30;
166 if (kstrtoint(value, base, &_res))
169 *res = _res << shift_left_factor;
174 * This function fills in xfs_mount_t fields based on mount args.
175 * Note: the superblock has _not_ yet been read in.
177 * Note that this function leaks the various device name allocations on
178 * failure. The caller takes care of them.
180 * *sb is const because this is also used to test options on the remount
181 * path, and we don't want this to have any side effects at remount time.
182 * Today this function does not change *sb, but just to future-proof...
186 struct xfs_mount *mp,
189 const struct super_block *sb = mp->m_super;
191 substring_t args[MAX_OPT_ARGS];
195 __uint8_t iosizelog = 0;
198 * set up the mount name first so all the errors will refer to the
201 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
204 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
207 * Copy binary VFS mount flags we are interested in.
209 if (sb->s_flags & MS_RDONLY)
210 mp->m_flags |= XFS_MOUNT_RDONLY;
211 if (sb->s_flags & MS_DIRSYNC)
212 mp->m_flags |= XFS_MOUNT_DIRSYNC;
213 if (sb->s_flags & MS_SYNCHRONOUS)
214 mp->m_flags |= XFS_MOUNT_WSYNC;
217 * Set some default flags that could be cleared by the mount option
220 mp->m_flags |= XFS_MOUNT_BARRIER;
221 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
224 * These can be overridden by the mount option parsing.
232 while ((p = strsep(&options, ",")) != NULL) {
238 token = match_token(p, tokens, args);
241 if (match_int(args, &mp->m_logbufs))
245 if (suffix_kstrtoint(args, 10, &mp->m_logbsize))
249 mp->m_logname = match_strdup(args);
254 xfs_warn(mp, "%s option not allowed on this system", p);
257 mp->m_rtname = match_strdup(args);
263 if (suffix_kstrtoint(args, 10, &iosize))
265 iosizelog = ffs(iosize) - 1;
269 mp->m_flags |= XFS_MOUNT_GRPID;
273 mp->m_flags &= ~XFS_MOUNT_GRPID;
276 mp->m_flags |= XFS_MOUNT_WSYNC;
279 mp->m_flags |= XFS_MOUNT_NORECOVERY;
282 mp->m_flags |= XFS_MOUNT_NOALIGN;
285 mp->m_flags |= XFS_MOUNT_SWALLOC;
288 if (match_int(args, &dsunit))
292 if (match_int(args, &dswidth))
296 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
299 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
302 mp->m_flags |= XFS_MOUNT_NOUUID;
305 mp->m_flags |= XFS_MOUNT_BARRIER;
308 mp->m_flags &= ~XFS_MOUNT_BARRIER;
311 mp->m_flags |= XFS_MOUNT_IKEEP;
314 mp->m_flags &= ~XFS_MOUNT_IKEEP;
317 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
320 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
323 mp->m_flags |= XFS_MOUNT_ATTR2;
326 mp->m_flags &= ~XFS_MOUNT_ATTR2;
327 mp->m_flags |= XFS_MOUNT_NOATTR2;
329 case Opt_filestreams:
330 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
333 mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT;
334 mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD;
335 mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE;
340 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
344 case Opt_uqnoenforce:
345 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
346 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
350 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
353 case Opt_pqnoenforce:
354 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
355 mp->m_qflags &= ~XFS_PQUOTA_ENFD;
359 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
362 case Opt_gqnoenforce:
363 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
364 mp->m_qflags &= ~XFS_GQUOTA_ENFD;
367 mp->m_flags |= XFS_MOUNT_DISCARD;
370 mp->m_flags &= ~XFS_MOUNT_DISCARD;
374 mp->m_flags |= XFS_MOUNT_DAX;
378 xfs_warn(mp, "unknown mount option [%s].", p);
384 * no recovery flag requires a read-only mount
386 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
387 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
388 xfs_warn(mp, "no-recovery mounts must be read-only.");
392 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
394 "sunit and swidth options incompatible with the noalign option");
398 #ifndef CONFIG_XFS_QUOTA
399 if (XFS_IS_QUOTA_RUNNING(mp)) {
400 xfs_warn(mp, "quota support not available in this kernel.");
405 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
406 xfs_warn(mp, "sunit and swidth must be specified together");
410 if (dsunit && (dswidth % dsunit != 0)) {
412 "stripe width (%d) must be a multiple of the stripe unit (%d)",
418 if (dsunit && !(mp->m_flags & XFS_MOUNT_NOALIGN)) {
420 * At this point the superblock has not been read
421 * in, therefore we do not know the block size.
422 * Before the mount call ends we will convert
425 mp->m_dalign = dsunit;
426 mp->m_swidth = dswidth;
429 if (mp->m_logbufs != -1 &&
430 mp->m_logbufs != 0 &&
431 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
432 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
433 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
434 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
437 if (mp->m_logbsize != -1 &&
438 mp->m_logbsize != 0 &&
439 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
440 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
441 !is_power_of_2(mp->m_logbsize))) {
443 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
449 if (iosizelog > XFS_MAX_IO_LOG ||
450 iosizelog < XFS_MIN_IO_LOG) {
451 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]",
452 iosizelog, XFS_MIN_IO_LOG,
457 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
458 mp->m_readio_log = iosizelog;
459 mp->m_writeio_log = iosizelog;
465 struct proc_xfs_info {
472 struct xfs_mount *mp,
475 static struct proc_xfs_info xfs_info_set[] = {
476 /* the few simple ones we can get from the mount struct */
477 { XFS_MOUNT_IKEEP, ",ikeep" },
478 { XFS_MOUNT_WSYNC, ",wsync" },
479 { XFS_MOUNT_NOALIGN, ",noalign" },
480 { XFS_MOUNT_SWALLOC, ",swalloc" },
481 { XFS_MOUNT_NOUUID, ",nouuid" },
482 { XFS_MOUNT_NORECOVERY, ",norecovery" },
483 { XFS_MOUNT_ATTR2, ",attr2" },
484 { XFS_MOUNT_FILESTREAMS, ",filestreams" },
485 { XFS_MOUNT_GRPID, ",grpid" },
486 { XFS_MOUNT_DISCARD, ",discard" },
487 { XFS_MOUNT_SMALL_INUMS, ",inode32" },
488 { XFS_MOUNT_DAX, ",dax" },
491 static struct proc_xfs_info xfs_info_unset[] = {
492 /* the few simple ones we can get from the mount struct */
493 { XFS_MOUNT_COMPAT_IOSIZE, ",largeio" },
494 { XFS_MOUNT_BARRIER, ",nobarrier" },
495 { XFS_MOUNT_SMALL_INUMS, ",inode64" },
498 struct proc_xfs_info *xfs_infop;
500 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
501 if (mp->m_flags & xfs_infop->flag)
502 seq_puts(m, xfs_infop->str);
504 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
505 if (!(mp->m_flags & xfs_infop->flag))
506 seq_puts(m, xfs_infop->str);
509 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
510 seq_printf(m, ",allocsize=%dk",
511 (int)(1 << mp->m_writeio_log) >> 10);
513 if (mp->m_logbufs > 0)
514 seq_printf(m, ",logbufs=%d", mp->m_logbufs);
515 if (mp->m_logbsize > 0)
516 seq_printf(m, ",logbsize=%dk", mp->m_logbsize >> 10);
519 seq_show_option(m, "logdev", mp->m_logname);
521 seq_show_option(m, "rtdev", mp->m_rtname);
523 if (mp->m_dalign > 0)
524 seq_printf(m, ",sunit=%d",
525 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
526 if (mp->m_swidth > 0)
527 seq_printf(m, ",swidth=%d",
528 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
530 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
531 seq_puts(m, ",usrquota");
532 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
533 seq_puts(m, ",uqnoenforce");
535 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
536 if (mp->m_qflags & XFS_PQUOTA_ENFD)
537 seq_puts(m, ",prjquota");
539 seq_puts(m, ",pqnoenforce");
541 if (mp->m_qflags & XFS_GQUOTA_ACCT) {
542 if (mp->m_qflags & XFS_GQUOTA_ENFD)
543 seq_puts(m, ",grpquota");
545 seq_puts(m, ",gqnoenforce");
548 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
549 seq_puts(m, ",noquota");
555 unsigned int blockshift)
557 unsigned int pagefactor = 1;
558 unsigned int bitshift = BITS_PER_LONG - 1;
560 /* Figure out maximum filesize, on Linux this can depend on
561 * the filesystem blocksize (on 32 bit platforms).
562 * __block_write_begin does this in an [unsigned] long...
563 * page->index << (PAGE_SHIFT - bbits)
564 * So, for page sized blocks (4K on 32 bit platforms),
565 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
566 * (((u64)PAGE_SIZE << (BITS_PER_LONG-1))-1)
567 * but for smaller blocksizes it is less (bbits = log2 bsize).
568 * Note1: get_block_t takes a long (implicit cast from above)
569 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
570 * can optionally convert the [unsigned] long from above into
571 * an [unsigned] long long.
574 #if BITS_PER_LONG == 32
575 # if defined(CONFIG_LBDAF)
576 ASSERT(sizeof(sector_t) == 8);
577 pagefactor = PAGE_SIZE;
578 bitshift = BITS_PER_LONG;
580 pagefactor = PAGE_SIZE >> (PAGE_SHIFT - blockshift);
584 return (((__uint64_t)pagefactor) << bitshift) - 1;
588 * Set parameters for inode allocation heuristics, taking into account
589 * filesystem size and inode32/inode64 mount options; i.e. specifically
590 * whether or not XFS_MOUNT_SMALL_INUMS is set.
592 * Inode allocation patterns are altered only if inode32 is requested
593 * (XFS_MOUNT_SMALL_INUMS), and the filesystem is sufficiently large.
594 * If altered, XFS_MOUNT_32BITINODES is set as well.
596 * An agcount independent of that in the mount structure is provided
597 * because in the growfs case, mp->m_sb.sb_agcount is not yet updated
598 * to the potentially higher ag count.
600 * Returns the maximum AG index which may contain inodes.
604 struct xfs_mount *mp,
605 xfs_agnumber_t agcount)
607 xfs_agnumber_t index;
608 xfs_agnumber_t maxagi = 0;
609 xfs_sb_t *sbp = &mp->m_sb;
610 xfs_agnumber_t max_metadata;
615 * Calculate how much should be reserved for inodes to meet
616 * the max inode percentage. Used only for inode32.
618 if (mp->m_maxicount) {
621 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
623 icount += sbp->sb_agblocks - 1;
624 do_div(icount, sbp->sb_agblocks);
625 max_metadata = icount;
627 max_metadata = agcount;
630 /* Get the last possible inode in the filesystem */
631 agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks - 1, 0);
632 ino = XFS_AGINO_TO_INO(mp, agcount - 1, agino);
635 * If user asked for no more than 32-bit inodes, and the fs is
636 * sufficiently large, set XFS_MOUNT_32BITINODES if we must alter
637 * the allocator to accommodate the request.
639 if ((mp->m_flags & XFS_MOUNT_SMALL_INUMS) && ino > XFS_MAXINUMBER_32)
640 mp->m_flags |= XFS_MOUNT_32BITINODES;
642 mp->m_flags &= ~XFS_MOUNT_32BITINODES;
644 for (index = 0; index < agcount; index++) {
645 struct xfs_perag *pag;
647 ino = XFS_AGINO_TO_INO(mp, index, agino);
649 pag = xfs_perag_get(mp, index);
651 if (mp->m_flags & XFS_MOUNT_32BITINODES) {
652 if (ino > XFS_MAXINUMBER_32) {
653 pag->pagi_inodeok = 0;
654 pag->pagf_metadata = 0;
656 pag->pagi_inodeok = 1;
658 if (index < max_metadata)
659 pag->pagf_metadata = 1;
661 pag->pagf_metadata = 0;
664 pag->pagi_inodeok = 1;
665 pag->pagf_metadata = 0;
671 return (mp->m_flags & XFS_MOUNT_32BITINODES) ? maxagi : agcount;
678 struct block_device **bdevp)
682 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
684 if (IS_ERR(*bdevp)) {
685 error = PTR_ERR(*bdevp);
686 xfs_warn(mp, "Invalid device [%s], error=%d", name, error);
694 struct block_device *bdev)
697 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
701 xfs_blkdev_issue_flush(
702 xfs_buftarg_t *buftarg)
704 blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL);
709 struct xfs_mount *mp)
711 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
712 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
713 xfs_free_buftarg(mp, mp->m_logdev_targp);
714 xfs_blkdev_put(logdev);
716 if (mp->m_rtdev_targp) {
717 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
718 xfs_free_buftarg(mp, mp->m_rtdev_targp);
719 xfs_blkdev_put(rtdev);
721 xfs_free_buftarg(mp, mp->m_ddev_targp);
725 * The file system configurations are:
726 * (1) device (partition) with data and internal log
727 * (2) logical volume with data and log subvolumes.
728 * (3) logical volume with data, log, and realtime subvolumes.
730 * We only have to handle opening the log and realtime volumes here if
731 * they are present. The data subvolume has already been opened by
732 * get_sb_bdev() and is stored in sb->s_bdev.
736 struct xfs_mount *mp)
738 struct block_device *ddev = mp->m_super->s_bdev;
739 struct block_device *logdev = NULL, *rtdev = NULL;
743 * Open real time and log devices - order is important.
746 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
752 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
754 goto out_close_logdev;
756 if (rtdev == ddev || rtdev == logdev) {
758 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
760 goto out_close_rtdev;
765 * Setup xfs_mount buffer target pointers
768 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev);
769 if (!mp->m_ddev_targp)
770 goto out_close_rtdev;
773 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev);
774 if (!mp->m_rtdev_targp)
775 goto out_free_ddev_targ;
778 if (logdev && logdev != ddev) {
779 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev);
780 if (!mp->m_logdev_targp)
781 goto out_free_rtdev_targ;
783 mp->m_logdev_targp = mp->m_ddev_targp;
789 if (mp->m_rtdev_targp)
790 xfs_free_buftarg(mp, mp->m_rtdev_targp);
792 xfs_free_buftarg(mp, mp->m_ddev_targp);
794 xfs_blkdev_put(rtdev);
796 if (logdev && logdev != ddev)
797 xfs_blkdev_put(logdev);
803 * Setup xfs_mount buffer target pointers based on superblock
807 struct xfs_mount *mp)
811 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize);
815 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
816 unsigned int log_sector_size = BBSIZE;
818 if (xfs_sb_version_hassector(&mp->m_sb))
819 log_sector_size = mp->m_sb.sb_logsectsize;
820 error = xfs_setsize_buftarg(mp->m_logdev_targp,
825 if (mp->m_rtdev_targp) {
826 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
827 mp->m_sb.sb_sectsize);
836 xfs_init_mount_workqueues(
837 struct xfs_mount *mp)
839 mp->m_buf_workqueue = alloc_workqueue("xfs-buf/%s",
840 WQ_MEM_RECLAIM|WQ_FREEZABLE, 1, mp->m_fsname);
841 if (!mp->m_buf_workqueue)
844 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s",
845 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
846 if (!mp->m_data_workqueue)
847 goto out_destroy_buf;
849 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
850 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
851 if (!mp->m_unwritten_workqueue)
852 goto out_destroy_data_iodone_queue;
854 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s",
855 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
856 if (!mp->m_cil_workqueue)
857 goto out_destroy_unwritten;
859 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
860 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
861 if (!mp->m_reclaim_workqueue)
862 goto out_destroy_cil;
864 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
865 WQ_MEM_RECLAIM|WQ_FREEZABLE|WQ_HIGHPRI, 0,
867 if (!mp->m_log_workqueue)
868 goto out_destroy_reclaim;
870 mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s",
871 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
872 if (!mp->m_eofblocks_workqueue)
873 goto out_destroy_log;
878 destroy_workqueue(mp->m_log_workqueue);
880 destroy_workqueue(mp->m_reclaim_workqueue);
882 destroy_workqueue(mp->m_cil_workqueue);
883 out_destroy_unwritten:
884 destroy_workqueue(mp->m_unwritten_workqueue);
885 out_destroy_data_iodone_queue:
886 destroy_workqueue(mp->m_data_workqueue);
888 destroy_workqueue(mp->m_buf_workqueue);
894 xfs_destroy_mount_workqueues(
895 struct xfs_mount *mp)
897 destroy_workqueue(mp->m_eofblocks_workqueue);
898 destroy_workqueue(mp->m_log_workqueue);
899 destroy_workqueue(mp->m_reclaim_workqueue);
900 destroy_workqueue(mp->m_cil_workqueue);
901 destroy_workqueue(mp->m_data_workqueue);
902 destroy_workqueue(mp->m_unwritten_workqueue);
903 destroy_workqueue(mp->m_buf_workqueue);
907 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
908 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
909 * for IO to complete so that we effectively throttle multiple callers to the
910 * rate at which IO is completing.
914 struct xfs_mount *mp)
916 struct super_block *sb = mp->m_super;
918 if (down_read_trylock(&sb->s_umount)) {
920 up_read(&sb->s_umount);
924 /* Catch misguided souls that try to use this interface on XFS */
925 STATIC struct inode *
927 struct super_block *sb)
934 * Now that the generic code is guaranteed not to be accessing
935 * the linux inode, we can inactivate and reclaim the inode.
938 xfs_fs_destroy_inode(
941 struct xfs_inode *ip = XFS_I(inode);
944 trace_xfs_destroy_inode(ip);
946 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
947 XFS_STATS_INC(ip->i_mount, vn_rele);
948 XFS_STATS_INC(ip->i_mount, vn_remove);
950 if (xfs_is_reflink_inode(ip)) {
951 error = xfs_reflink_cancel_cow_range(ip, 0, NULLFILEOFF);
952 if (error && !XFS_FORCED_SHUTDOWN(ip->i_mount))
953 xfs_warn(ip->i_mount,
954 "Error %d while evicting CoW blocks for inode %llu.",
960 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
961 XFS_STATS_INC(ip->i_mount, vn_reclaim);
964 * We should never get here with one of the reclaim flags already set.
966 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
967 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
970 * We always use background reclaim here because even if the
971 * inode is clean, it still may be under IO and hence we have
972 * to take the flush lock. The background reclaim path handles
973 * this more efficiently than we can here, so simply let background
974 * reclaim tear down all inodes.
976 xfs_inode_set_reclaim_tag(ip);
980 * Slab object creation initialisation for the XFS inode.
981 * This covers only the idempotent fields in the XFS inode;
982 * all other fields need to be initialised on allocation
983 * from the slab. This avoids the need to repeatedly initialise
984 * fields in the xfs inode that left in the initialise state
985 * when freeing the inode.
988 xfs_fs_inode_init_once(
991 struct xfs_inode *ip = inode;
993 memset(ip, 0, sizeof(struct xfs_inode));
996 inode_init_once(VFS_I(ip));
999 atomic_set(&ip->i_pincount, 0);
1000 spin_lock_init(&ip->i_flags_lock);
1002 mrlock_init(&ip->i_mmaplock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
1003 "xfsino", ip->i_ino);
1004 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
1005 "xfsino", ip->i_ino);
1009 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1010 * serialised against cache hits here via the inode->i_lock and igrab() in
1011 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1012 * racing with us, and it avoids needing to grab a spinlock here for every inode
1013 * we drop the final reference on.
1017 struct inode *inode)
1019 struct xfs_inode *ip = XFS_I(inode);
1022 * If this unlinked inode is in the middle of recovery, don't
1023 * drop the inode just yet; log recovery will take care of
1024 * that. See the comment for this inode flag.
1026 if (ip->i_flags & XFS_IRECOVERY) {
1027 ASSERT(ip->i_mount->m_log->l_flags & XLOG_RECOVERY_NEEDED);
1031 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);
1036 struct xfs_mount *mp)
1038 kfree(mp->m_fsname);
1039 kfree(mp->m_rtname);
1040 kfree(mp->m_logname);
1045 struct super_block *sb,
1048 struct xfs_mount *mp = XFS_M(sb);
1051 * Doing anything during the async pass would be counterproductive.
1056 xfs_log_force(mp, XFS_LOG_SYNC);
1059 * The disk must be active because we're syncing.
1060 * We schedule log work now (now that the disk is
1061 * active) instead of later (when it might not be).
1063 flush_delayed_work(&mp->m_log->l_work);
1071 struct dentry *dentry,
1072 struct kstatfs *statp)
1074 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1075 xfs_sb_t *sbp = &mp->m_sb;
1076 struct xfs_inode *ip = XFS_I(d_inode(dentry));
1077 __uint64_t fakeinos, id;
1080 __uint64_t fdblocks;
1084 statp->f_type = XFS_SB_MAGIC;
1085 statp->f_namelen = MAXNAMELEN - 1;
1087 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1088 statp->f_fsid.val[0] = (u32)id;
1089 statp->f_fsid.val[1] = (u32)(id >> 32);
1091 icount = percpu_counter_sum(&mp->m_icount);
1092 ifree = percpu_counter_sum(&mp->m_ifree);
1093 fdblocks = percpu_counter_sum(&mp->m_fdblocks);
1095 spin_lock(&mp->m_sb_lock);
1096 statp->f_bsize = sbp->sb_blocksize;
1097 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1098 statp->f_blocks = sbp->sb_dblocks - lsize;
1099 spin_unlock(&mp->m_sb_lock);
1101 statp->f_bfree = fdblocks - mp->m_alloc_set_aside;
1102 statp->f_bavail = statp->f_bfree;
1104 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1105 statp->f_files = MIN(icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1106 if (mp->m_maxicount)
1107 statp->f_files = min_t(typeof(statp->f_files),
1111 /* If sb_icount overshot maxicount, report actual allocation */
1112 statp->f_files = max_t(typeof(statp->f_files),
1116 /* make sure statp->f_ffree does not underflow */
1117 ffree = statp->f_files - (icount - ifree);
1118 statp->f_ffree = max_t(__int64_t, ffree, 0);
1121 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
1122 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) ==
1123 (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))
1124 xfs_qm_statvfs(ip, statp);
1129 xfs_save_resvblks(struct xfs_mount *mp)
1131 __uint64_t resblks = 0;
1133 mp->m_resblks_save = mp->m_resblks;
1134 xfs_reserve_blocks(mp, &resblks, NULL);
1138 xfs_restore_resvblks(struct xfs_mount *mp)
1142 if (mp->m_resblks_save) {
1143 resblks = mp->m_resblks_save;
1144 mp->m_resblks_save = 0;
1146 resblks = xfs_default_resblks(mp);
1148 xfs_reserve_blocks(mp, &resblks, NULL);
1152 * Trigger writeback of all the dirty metadata in the file system.
1154 * This ensures that the metadata is written to their location on disk rather
1155 * than just existing in transactions in the log. This means after a quiesce
1156 * there is no log replay required to write the inodes to disk - this is the
1157 * primary difference between a sync and a quiesce.
1159 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1160 * it is started again when appropriate.
1164 struct xfs_mount *mp)
1168 /* wait for all modifications to complete */
1169 while (atomic_read(&mp->m_active_trans) > 0)
1172 /* force the log to unpin objects from the now complete transactions */
1173 xfs_log_force(mp, XFS_LOG_SYNC);
1175 /* reclaim inodes to do any IO before the freeze completes */
1176 xfs_reclaim_inodes(mp, 0);
1177 xfs_reclaim_inodes(mp, SYNC_WAIT);
1179 /* Push the superblock and write an unmount record */
1180 error = xfs_log_sbcount(mp);
1182 xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
1183 "Frozen image may not be consistent.");
1185 * Just warn here till VFS can correctly support
1186 * read-only remount without racing.
1188 WARN_ON(atomic_read(&mp->m_active_trans) != 0);
1190 xfs_log_quiesce(mp);
1194 xfs_test_remount_options(
1195 struct super_block *sb,
1196 struct xfs_mount *mp,
1200 struct xfs_mount *tmp_mp;
1202 tmp_mp = kmem_zalloc(sizeof(*tmp_mp), KM_MAYFAIL);
1206 tmp_mp->m_super = sb;
1207 error = xfs_parseargs(tmp_mp, options);
1208 xfs_free_fsname(tmp_mp);
1216 struct super_block *sb,
1220 struct xfs_mount *mp = XFS_M(sb);
1221 xfs_sb_t *sbp = &mp->m_sb;
1222 substring_t args[MAX_OPT_ARGS];
1226 /* First, check for complete junk; i.e. invalid options */
1227 error = xfs_test_remount_options(sb, mp, options);
1231 sync_filesystem(sb);
1232 while ((p = strsep(&options, ",")) != NULL) {
1238 token = match_token(p, tokens, args);
1241 mp->m_flags |= XFS_MOUNT_BARRIER;
1244 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1247 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
1248 mp->m_maxagi = xfs_set_inode_alloc(mp, sbp->sb_agcount);
1251 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
1252 mp->m_maxagi = xfs_set_inode_alloc(mp, sbp->sb_agcount);
1256 * Logically we would return an error here to prevent
1257 * users from believing they might have changed
1258 * mount options using remount which can't be changed.
1260 * But unfortunately mount(8) adds all options from
1261 * mtab and fstab to the mount arguments in some cases
1262 * so we can't blindly reject options, but have to
1263 * check for each specified option if it actually
1264 * differs from the currently set option and only
1265 * reject it if that's the case.
1267 * Until that is implemented we return success for
1268 * every remount request, and silently ignore all
1269 * options that we can't actually change.
1273 "mount option \"%s\" not supported for remount", p);
1282 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1283 if (mp->m_flags & XFS_MOUNT_NORECOVERY) {
1285 "ro->rw transition prohibited on norecovery mount");
1289 if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
1290 xfs_sb_has_ro_compat_feature(sbp,
1291 XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
1293 "ro->rw transition prohibited on unknown (0x%x) ro-compat filesystem",
1294 (sbp->sb_features_ro_compat &
1295 XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
1299 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1302 * If this is the first remount to writeable state we
1303 * might have some superblock changes to update.
1305 if (mp->m_update_sb) {
1306 error = xfs_sync_sb(mp, false);
1308 xfs_warn(mp, "failed to write sb changes");
1311 mp->m_update_sb = false;
1315 * Fill out the reserve pool if it is empty. Use the stashed
1316 * value if it is non-zero, otherwise go with the default.
1318 xfs_restore_resvblks(mp);
1319 xfs_log_work_queue(mp);
1320 xfs_queue_eofblocks(mp);
1322 /* Recover any CoW blocks that never got remapped. */
1323 error = xfs_reflink_recover_cow(mp);
1326 "Error %d recovering leftover CoW allocations.", error);
1327 xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
1331 /* Create the per-AG metadata reservation pool .*/
1332 error = xfs_fs_reserve_ag_blocks(mp);
1333 if (error && error != -ENOSPC)
1338 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1339 /* Free the per-AG metadata reservation pool. */
1340 error = xfs_fs_unreserve_ag_blocks(mp);
1342 xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
1347 * Before we sync the metadata, we need to free up the reserve
1348 * block pool so that the used block count in the superblock on
1349 * disk is correct at the end of the remount. Stash the current
1350 * reserve pool size so that if we get remounted rw, we can
1351 * return it to the same size.
1353 xfs_save_resvblks(mp);
1356 * Cancel background eofb scanning so it cannot race with the
1357 * final log force+buftarg wait and deadlock the remount.
1359 cancel_delayed_work_sync(&mp->m_eofblocks_work);
1361 xfs_quiesce_attr(mp);
1362 mp->m_flags |= XFS_MOUNT_RDONLY;
1369 * Second stage of a freeze. The data is already frozen so we only
1370 * need to take care of the metadata. Once that's done sync the superblock
1371 * to the log to dirty it in case of a crash while frozen. This ensures that we
1372 * will recover the unlinked inode lists on the next mount.
1376 struct super_block *sb)
1378 struct xfs_mount *mp = XFS_M(sb);
1380 xfs_save_resvblks(mp);
1381 xfs_quiesce_attr(mp);
1382 return xfs_sync_sb(mp, true);
1387 struct super_block *sb)
1389 struct xfs_mount *mp = XFS_M(sb);
1391 xfs_restore_resvblks(mp);
1392 xfs_log_work_queue(mp);
1397 xfs_fs_show_options(
1399 struct dentry *root)
1401 return xfs_showargs(XFS_M(root->d_sb), m);
1405 * This function fills in xfs_mount_t fields based on mount args.
1406 * Note: the superblock _has_ now been read in.
1410 struct xfs_mount *mp)
1412 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1414 /* Fail a mount where the logbuf is smaller than the log stripe */
1415 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1416 if (mp->m_logbsize <= 0 &&
1417 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1418 mp->m_logbsize = mp->m_sb.sb_logsunit;
1419 } else if (mp->m_logbsize > 0 &&
1420 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1422 "logbuf size must be greater than or equal to log stripe size");
1426 /* Fail a mount if the logbuf is larger than 32K */
1427 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1429 "logbuf size for version 1 logs must be 16K or 32K");
1435 * V5 filesystems always use attr2 format for attributes.
1437 if (xfs_sb_version_hascrc(&mp->m_sb) &&
1438 (mp->m_flags & XFS_MOUNT_NOATTR2)) {
1439 xfs_warn(mp, "Cannot mount a V5 filesystem as noattr2. "
1440 "attr2 is always enabled for V5 filesystems.");
1445 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1446 * told by noattr2 to turn it off
1448 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1449 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1450 mp->m_flags |= XFS_MOUNT_ATTR2;
1453 * prohibit r/w mounts of read-only filesystems
1455 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1457 "cannot mount a read-only filesystem as read-write");
1461 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
1462 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) &&
1463 !xfs_sb_version_has_pquotino(&mp->m_sb)) {
1465 "Super block does not support project and group quota together");
1473 xfs_init_percpu_counters(
1474 struct xfs_mount *mp)
1478 error = percpu_counter_init(&mp->m_icount, 0, GFP_KERNEL);
1482 error = percpu_counter_init(&mp->m_ifree, 0, GFP_KERNEL);
1486 error = percpu_counter_init(&mp->m_fdblocks, 0, GFP_KERNEL);
1493 percpu_counter_destroy(&mp->m_ifree);
1495 percpu_counter_destroy(&mp->m_icount);
1500 xfs_reinit_percpu_counters(
1501 struct xfs_mount *mp)
1503 percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount);
1504 percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree);
1505 percpu_counter_set(&mp->m_fdblocks, mp->m_sb.sb_fdblocks);
1509 xfs_destroy_percpu_counters(
1510 struct xfs_mount *mp)
1512 percpu_counter_destroy(&mp->m_icount);
1513 percpu_counter_destroy(&mp->m_ifree);
1514 percpu_counter_destroy(&mp->m_fdblocks);
1519 struct super_block *sb,
1524 struct xfs_mount *mp = NULL;
1525 int flags = 0, error = -ENOMEM;
1527 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1531 spin_lock_init(&mp->m_sb_lock);
1532 mutex_init(&mp->m_growlock);
1533 atomic_set(&mp->m_active_trans, 0);
1534 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1535 INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
1536 INIT_DELAYED_WORK(&mp->m_cowblocks_work, xfs_cowblocks_worker);
1537 mp->m_kobj.kobject.kset = xfs_kset;
1542 error = xfs_parseargs(mp, (char *)data);
1544 goto out_free_fsname;
1546 sb_min_blocksize(sb, BBSIZE);
1547 sb->s_xattr = xfs_xattr_handlers;
1548 sb->s_export_op = &xfs_export_operations;
1549 #ifdef CONFIG_XFS_QUOTA
1550 sb->s_qcop = &xfs_quotactl_operations;
1551 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
1553 sb->s_op = &xfs_super_operations;
1556 flags |= XFS_MFSI_QUIET;
1558 error = xfs_open_devices(mp);
1560 goto out_free_fsname;
1562 error = xfs_init_mount_workqueues(mp);
1564 goto out_close_devices;
1566 error = xfs_init_percpu_counters(mp);
1568 goto out_destroy_workqueues;
1570 /* Allocate stats memory before we do operations that might use it */
1571 mp->m_stats.xs_stats = alloc_percpu(struct xfsstats);
1572 if (!mp->m_stats.xs_stats) {
1574 goto out_destroy_counters;
1577 error = xfs_readsb(mp, flags);
1579 goto out_free_stats;
1581 error = xfs_finish_flags(mp);
1585 error = xfs_setup_devices(mp);
1589 error = xfs_filestream_mount(mp);
1594 * we must configure the block size in the superblock before we run the
1595 * full mount process as the mount process can lookup and cache inodes.
1597 sb->s_magic = XFS_SB_MAGIC;
1598 sb->s_blocksize = mp->m_sb.sb_blocksize;
1599 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1600 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1601 sb->s_max_links = XFS_MAXLINK;
1602 sb->s_time_gran = 1;
1603 set_posix_acl_flag(sb);
1605 /* version 5 superblocks support inode version counters. */
1606 if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
1607 sb->s_flags |= MS_I_VERSION;
1609 if (mp->m_flags & XFS_MOUNT_DAX) {
1611 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1613 error = bdev_dax_supported(sb, sb->s_blocksize);
1616 "DAX unsupported by block device. Turning off DAX.");
1617 mp->m_flags &= ~XFS_MOUNT_DAX;
1619 if (xfs_sb_version_hasreflink(&mp->m_sb))
1621 "DAX and reflink have not been tested together!");
1624 if (xfs_sb_version_hasrmapbt(&mp->m_sb)) {
1625 if (mp->m_sb.sb_rblocks) {
1627 "EXPERIMENTAL reverse mapping btree not compatible with realtime device!");
1629 goto out_filestream_unmount;
1632 "EXPERIMENTAL reverse mapping btree feature enabled. Use at your own risk!");
1635 if (xfs_sb_version_hasreflink(&mp->m_sb))
1637 "EXPERIMENTAL reflink feature enabled. Use at your own risk!");
1639 error = xfs_mountfs(mp);
1641 goto out_filestream_unmount;
1643 root = igrab(VFS_I(mp->m_rootip));
1648 sb->s_root = d_make_root(root);
1656 out_filestream_unmount:
1657 xfs_filestream_unmount(mp);
1661 free_percpu(mp->m_stats.xs_stats);
1662 out_destroy_counters:
1663 xfs_destroy_percpu_counters(mp);
1664 out_destroy_workqueues:
1665 xfs_destroy_mount_workqueues(mp);
1667 xfs_close_devices(mp);
1669 xfs_free_fsname(mp);
1675 xfs_filestream_unmount(mp);
1682 struct super_block *sb)
1684 struct xfs_mount *mp = XFS_M(sb);
1686 xfs_notice(mp, "Unmounting Filesystem");
1687 xfs_filestream_unmount(mp);
1691 free_percpu(mp->m_stats.xs_stats);
1692 xfs_destroy_percpu_counters(mp);
1693 xfs_destroy_mount_workqueues(mp);
1694 xfs_close_devices(mp);
1695 xfs_free_fsname(mp);
1699 STATIC struct dentry *
1701 struct file_system_type *fs_type,
1703 const char *dev_name,
1706 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1710 xfs_fs_nr_cached_objects(
1711 struct super_block *sb,
1712 struct shrink_control *sc)
1714 return xfs_reclaim_inodes_count(XFS_M(sb));
1718 xfs_fs_free_cached_objects(
1719 struct super_block *sb,
1720 struct shrink_control *sc)
1722 return xfs_reclaim_inodes_nr(XFS_M(sb), sc->nr_to_scan);
1725 static const struct super_operations xfs_super_operations = {
1726 .alloc_inode = xfs_fs_alloc_inode,
1727 .destroy_inode = xfs_fs_destroy_inode,
1728 .drop_inode = xfs_fs_drop_inode,
1729 .put_super = xfs_fs_put_super,
1730 .sync_fs = xfs_fs_sync_fs,
1731 .freeze_fs = xfs_fs_freeze,
1732 .unfreeze_fs = xfs_fs_unfreeze,
1733 .statfs = xfs_fs_statfs,
1734 .remount_fs = xfs_fs_remount,
1735 .show_options = xfs_fs_show_options,
1736 .nr_cached_objects = xfs_fs_nr_cached_objects,
1737 .free_cached_objects = xfs_fs_free_cached_objects,
1740 static struct file_system_type xfs_fs_type = {
1741 .owner = THIS_MODULE,
1743 .mount = xfs_fs_mount,
1744 .kill_sb = kill_block_super,
1745 .fs_flags = FS_REQUIRES_DEV,
1747 MODULE_ALIAS_FS("xfs");
1750 xfs_init_zones(void)
1752 xfs_ioend_bioset = bioset_create(4 * MAX_BUF_PER_PAGE,
1753 offsetof(struct xfs_ioend, io_inline_bio));
1754 if (!xfs_ioend_bioset)
1757 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1759 if (!xfs_log_ticket_zone)
1760 goto out_free_ioend_bioset;
1762 xfs_bmap_free_item_zone = kmem_zone_init(
1763 sizeof(struct xfs_extent_free_item),
1764 "xfs_bmap_free_item");
1765 if (!xfs_bmap_free_item_zone)
1766 goto out_destroy_log_ticket_zone;
1768 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1770 if (!xfs_btree_cur_zone)
1771 goto out_destroy_bmap_free_item_zone;
1773 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1775 if (!xfs_da_state_zone)
1776 goto out_destroy_btree_cur_zone;
1778 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1779 if (!xfs_ifork_zone)
1780 goto out_destroy_da_state_zone;
1782 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1783 if (!xfs_trans_zone)
1784 goto out_destroy_ifork_zone;
1786 xfs_log_item_desc_zone =
1787 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1788 "xfs_log_item_desc");
1789 if (!xfs_log_item_desc_zone)
1790 goto out_destroy_trans_zone;
1793 * The size of the zone allocated buf log item is the maximum
1794 * size possible under XFS. This wastes a little bit of memory,
1795 * but it is much faster.
1797 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item),
1799 if (!xfs_buf_item_zone)
1800 goto out_destroy_log_item_desc_zone;
1802 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1803 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1804 sizeof(xfs_extent_t))), "xfs_efd_item");
1806 goto out_destroy_buf_item_zone;
1808 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1809 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1810 sizeof(xfs_extent_t))), "xfs_efi_item");
1812 goto out_destroy_efd_zone;
1815 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1816 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD |
1817 KM_ZONE_ACCOUNT, xfs_fs_inode_init_once);
1818 if (!xfs_inode_zone)
1819 goto out_destroy_efi_zone;
1822 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1823 KM_ZONE_SPREAD, NULL);
1825 goto out_destroy_inode_zone;
1826 xfs_icreate_zone = kmem_zone_init(sizeof(struct xfs_icreate_item),
1828 if (!xfs_icreate_zone)
1829 goto out_destroy_ili_zone;
1831 xfs_rud_zone = kmem_zone_init(sizeof(struct xfs_rud_log_item),
1834 goto out_destroy_icreate_zone;
1836 xfs_rui_zone = kmem_zone_init(
1837 xfs_rui_log_item_sizeof(XFS_RUI_MAX_FAST_EXTENTS),
1840 goto out_destroy_rud_zone;
1842 xfs_cud_zone = kmem_zone_init(sizeof(struct xfs_cud_log_item),
1845 goto out_destroy_rui_zone;
1847 xfs_cui_zone = kmem_zone_init(
1848 xfs_cui_log_item_sizeof(XFS_CUI_MAX_FAST_EXTENTS),
1851 goto out_destroy_cud_zone;
1853 xfs_bud_zone = kmem_zone_init(sizeof(struct xfs_bud_log_item),
1856 goto out_destroy_cui_zone;
1858 xfs_bui_zone = kmem_zone_init(
1859 xfs_bui_log_item_sizeof(XFS_BUI_MAX_FAST_EXTENTS),
1862 goto out_destroy_bud_zone;
1866 out_destroy_bud_zone:
1867 kmem_zone_destroy(xfs_bud_zone);
1868 out_destroy_cui_zone:
1869 kmem_zone_destroy(xfs_cui_zone);
1870 out_destroy_cud_zone:
1871 kmem_zone_destroy(xfs_cud_zone);
1872 out_destroy_rui_zone:
1873 kmem_zone_destroy(xfs_rui_zone);
1874 out_destroy_rud_zone:
1875 kmem_zone_destroy(xfs_rud_zone);
1876 out_destroy_icreate_zone:
1877 kmem_zone_destroy(xfs_icreate_zone);
1878 out_destroy_ili_zone:
1879 kmem_zone_destroy(xfs_ili_zone);
1880 out_destroy_inode_zone:
1881 kmem_zone_destroy(xfs_inode_zone);
1882 out_destroy_efi_zone:
1883 kmem_zone_destroy(xfs_efi_zone);
1884 out_destroy_efd_zone:
1885 kmem_zone_destroy(xfs_efd_zone);
1886 out_destroy_buf_item_zone:
1887 kmem_zone_destroy(xfs_buf_item_zone);
1888 out_destroy_log_item_desc_zone:
1889 kmem_zone_destroy(xfs_log_item_desc_zone);
1890 out_destroy_trans_zone:
1891 kmem_zone_destroy(xfs_trans_zone);
1892 out_destroy_ifork_zone:
1893 kmem_zone_destroy(xfs_ifork_zone);
1894 out_destroy_da_state_zone:
1895 kmem_zone_destroy(xfs_da_state_zone);
1896 out_destroy_btree_cur_zone:
1897 kmem_zone_destroy(xfs_btree_cur_zone);
1898 out_destroy_bmap_free_item_zone:
1899 kmem_zone_destroy(xfs_bmap_free_item_zone);
1900 out_destroy_log_ticket_zone:
1901 kmem_zone_destroy(xfs_log_ticket_zone);
1902 out_free_ioend_bioset:
1903 bioset_free(xfs_ioend_bioset);
1909 xfs_destroy_zones(void)
1912 * Make sure all delayed rcu free are flushed before we
1916 kmem_zone_destroy(xfs_bui_zone);
1917 kmem_zone_destroy(xfs_bud_zone);
1918 kmem_zone_destroy(xfs_cui_zone);
1919 kmem_zone_destroy(xfs_cud_zone);
1920 kmem_zone_destroy(xfs_rui_zone);
1921 kmem_zone_destroy(xfs_rud_zone);
1922 kmem_zone_destroy(xfs_icreate_zone);
1923 kmem_zone_destroy(xfs_ili_zone);
1924 kmem_zone_destroy(xfs_inode_zone);
1925 kmem_zone_destroy(xfs_efi_zone);
1926 kmem_zone_destroy(xfs_efd_zone);
1927 kmem_zone_destroy(xfs_buf_item_zone);
1928 kmem_zone_destroy(xfs_log_item_desc_zone);
1929 kmem_zone_destroy(xfs_trans_zone);
1930 kmem_zone_destroy(xfs_ifork_zone);
1931 kmem_zone_destroy(xfs_da_state_zone);
1932 kmem_zone_destroy(xfs_btree_cur_zone);
1933 kmem_zone_destroy(xfs_bmap_free_item_zone);
1934 kmem_zone_destroy(xfs_log_ticket_zone);
1935 bioset_free(xfs_ioend_bioset);
1939 xfs_init_workqueues(void)
1942 * The allocation workqueue can be used in memory reclaim situations
1943 * (writepage path), and parallelism is only limited by the number of
1944 * AGs in all the filesystems mounted. Hence use the default large
1945 * max_active value for this workqueue.
1947 xfs_alloc_wq = alloc_workqueue("xfsalloc",
1948 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0);
1956 xfs_destroy_workqueues(void)
1958 destroy_workqueue(xfs_alloc_wq);
1966 xfs_check_ondisk_structs();
1968 printk(KERN_INFO XFS_VERSION_STRING " with "
1969 XFS_BUILD_OPTIONS " enabled\n");
1971 xfs_extent_free_init_defer_op();
1972 xfs_rmap_update_init_defer_op();
1973 xfs_refcount_update_init_defer_op();
1974 xfs_bmap_update_init_defer_op();
1978 error = xfs_init_zones();
1982 error = xfs_init_workqueues();
1984 goto out_destroy_zones;
1986 error = xfs_mru_cache_init();
1988 goto out_destroy_wq;
1990 error = xfs_buf_init();
1992 goto out_mru_cache_uninit;
1994 error = xfs_init_procfs();
1996 goto out_buf_terminate;
1998 error = xfs_sysctl_register();
2000 goto out_cleanup_procfs;
2002 xfs_kset = kset_create_and_add("xfs", NULL, fs_kobj);
2005 goto out_sysctl_unregister;
2008 xfsstats.xs_kobj.kobject.kset = xfs_kset;
2010 xfsstats.xs_stats = alloc_percpu(struct xfsstats);
2011 if (!xfsstats.xs_stats) {
2013 goto out_kset_unregister;
2016 error = xfs_sysfs_init(&xfsstats.xs_kobj, &xfs_stats_ktype, NULL,
2019 goto out_free_stats;
2022 xfs_dbg_kobj.kobject.kset = xfs_kset;
2023 error = xfs_sysfs_init(&xfs_dbg_kobj, &xfs_dbg_ktype, NULL, "debug");
2025 goto out_remove_stats_kobj;
2028 error = xfs_qm_init();
2030 goto out_remove_dbg_kobj;
2032 error = register_filesystem(&xfs_fs_type);
2039 out_remove_dbg_kobj:
2041 xfs_sysfs_del(&xfs_dbg_kobj);
2042 out_remove_stats_kobj:
2044 xfs_sysfs_del(&xfsstats.xs_kobj);
2046 free_percpu(xfsstats.xs_stats);
2047 out_kset_unregister:
2048 kset_unregister(xfs_kset);
2049 out_sysctl_unregister:
2050 xfs_sysctl_unregister();
2052 xfs_cleanup_procfs();
2054 xfs_buf_terminate();
2055 out_mru_cache_uninit:
2056 xfs_mru_cache_uninit();
2058 xfs_destroy_workqueues();
2060 xfs_destroy_zones();
2069 unregister_filesystem(&xfs_fs_type);
2071 xfs_sysfs_del(&xfs_dbg_kobj);
2073 xfs_sysfs_del(&xfsstats.xs_kobj);
2074 free_percpu(xfsstats.xs_stats);
2075 kset_unregister(xfs_kset);
2076 xfs_sysctl_unregister();
2077 xfs_cleanup_procfs();
2078 xfs_buf_terminate();
2079 xfs_mru_cache_uninit();
2080 xfs_destroy_workqueues();
2081 xfs_destroy_zones();
2082 xfs_uuid_table_free();
2085 module_init(init_xfs_fs);
2086 module_exit(exit_xfs_fs);
2088 MODULE_AUTHOR("Silicon Graphics, Inc.");
2089 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
2090 MODULE_LICENSE("GPL");