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"
49 #include <linux/namei.h>
50 #include <linux/init.h>
51 #include <linux/slab.h>
52 #include <linux/mount.h>
53 #include <linux/mempool.h>
54 #include <linux/writeback.h>
55 #include <linux/kthread.h>
56 #include <linux/freezer.h>
57 #include <linux/parser.h>
59 static const struct super_operations xfs_super_operations;
60 static kmem_zone_t *xfs_ioend_zone;
61 mempool_t *xfs_ioend_pool;
63 static struct kset *xfs_kset; /* top-level xfs sysfs dir */
65 static struct xfs_kobj xfs_dbg_kobj; /* global debug sysfs attrs */
69 * Table driven mount option parser.
72 Opt_logbufs, Opt_logbsize, Opt_logdev, Opt_rtdev, Opt_biosize,
73 Opt_wsync, Opt_noalign, Opt_swalloc, Opt_sunit, Opt_swidth, Opt_nouuid,
74 Opt_mtpt, Opt_grpid, Opt_nogrpid, Opt_bsdgroups, Opt_sysvgroups,
75 Opt_allocsize, Opt_norecovery, Opt_barrier, Opt_nobarrier,
76 Opt_inode64, Opt_inode32, Opt_ikeep, Opt_noikeep,
77 Opt_largeio, Opt_nolargeio, Opt_attr2, Opt_noattr2, Opt_filestreams,
78 Opt_quota, Opt_noquota, Opt_usrquota, Opt_grpquota, Opt_prjquota,
79 Opt_uquota, Opt_gquota, Opt_pquota,
80 Opt_uqnoenforce, Opt_gqnoenforce, Opt_pqnoenforce, Opt_qnoenforce,
81 Opt_discard, Opt_nodiscard, Opt_dax, Opt_err,
84 static const match_table_t tokens = {
85 {Opt_logbufs, "logbufs=%u"}, /* number of XFS log buffers */
86 {Opt_logbsize, "logbsize=%s"}, /* size of XFS log buffers */
87 {Opt_logdev, "logdev=%s"}, /* log device */
88 {Opt_rtdev, "rtdev=%s"}, /* realtime I/O device */
89 {Opt_biosize, "biosize=%u"}, /* log2 of preferred buffered io size */
90 {Opt_wsync, "wsync"}, /* safe-mode nfs compatible mount */
91 {Opt_noalign, "noalign"}, /* turn off stripe alignment */
92 {Opt_swalloc, "swalloc"}, /* turn on stripe width allocation */
93 {Opt_sunit, "sunit=%u"}, /* data volume stripe unit */
94 {Opt_swidth, "swidth=%u"}, /* data volume stripe width */
95 {Opt_nouuid, "nouuid"}, /* ignore filesystem UUID */
96 {Opt_mtpt, "mtpt"}, /* filesystem mount point */
97 {Opt_grpid, "grpid"}, /* group-ID from parent directory */
98 {Opt_nogrpid, "nogrpid"}, /* group-ID from current process */
99 {Opt_bsdgroups, "bsdgroups"}, /* group-ID from parent directory */
100 {Opt_sysvgroups,"sysvgroups"}, /* group-ID from current process */
101 {Opt_allocsize, "allocsize=%s"},/* preferred allocation size */
102 {Opt_norecovery,"norecovery"}, /* don't run XFS recovery */
103 {Opt_barrier, "barrier"}, /* use writer barriers for log write and
104 * unwritten extent conversion */
105 {Opt_nobarrier, "nobarrier"}, /* .. disable */
106 {Opt_inode64, "inode64"}, /* inodes can be allocated anywhere */
107 {Opt_inode32, "inode32"}, /* inode allocation limited to
108 * XFS_MAXINUMBER_32 */
109 {Opt_ikeep, "ikeep"}, /* do not free empty inode clusters */
110 {Opt_noikeep, "noikeep"}, /* free empty inode clusters */
111 {Opt_largeio, "largeio"}, /* report large I/O sizes in stat() */
112 {Opt_nolargeio, "nolargeio"}, /* do not report large I/O sizes
114 {Opt_attr2, "attr2"}, /* do use attr2 attribute format */
115 {Opt_noattr2, "noattr2"}, /* do not use attr2 attribute format */
116 {Opt_filestreams,"filestreams"},/* use filestreams allocator */
117 {Opt_quota, "quota"}, /* disk quotas (user) */
118 {Opt_noquota, "noquota"}, /* no quotas */
119 {Opt_usrquota, "usrquota"}, /* user quota enabled */
120 {Opt_grpquota, "grpquota"}, /* group quota enabled */
121 {Opt_prjquota, "prjquota"}, /* project quota enabled */
122 {Opt_uquota, "uquota"}, /* user quota (IRIX variant) */
123 {Opt_gquota, "gquota"}, /* group quota (IRIX variant) */
124 {Opt_pquota, "pquota"}, /* project quota (IRIX variant) */
125 {Opt_uqnoenforce,"uqnoenforce"},/* user quota limit enforcement */
126 {Opt_gqnoenforce,"gqnoenforce"},/* group quota limit enforcement */
127 {Opt_pqnoenforce,"pqnoenforce"},/* project quota limit enforcement */
128 {Opt_qnoenforce, "qnoenforce"}, /* same as uqnoenforce */
129 {Opt_discard, "discard"}, /* Discard unused blocks */
130 {Opt_nodiscard, "nodiscard"}, /* Do not discard unused blocks */
132 {Opt_dax, "dax"}, /* Enable direct access to bdev pages */
138 suffix_kstrtoint(const substring_t *s, unsigned int base, int *res)
140 int last, shift_left_factor = 0, _res;
144 value = match_strdup(s);
148 last = strlen(value) - 1;
149 if (value[last] == 'K' || value[last] == 'k') {
150 shift_left_factor = 10;
153 if (value[last] == 'M' || value[last] == 'm') {
154 shift_left_factor = 20;
157 if (value[last] == 'G' || value[last] == 'g') {
158 shift_left_factor = 30;
162 if (kstrtoint(value, base, &_res))
165 *res = _res << shift_left_factor;
170 * This function fills in xfs_mount_t fields based on mount args.
171 * Note: the superblock has _not_ yet been read in.
173 * Note that this function leaks the various device name allocations on
174 * failure. The caller takes care of them.
176 * *sb is const because this is also used to test options on the remount
177 * path, and we don't want this to have any side effects at remount time.
178 * Today this function does not change *sb, but just to future-proof...
182 struct xfs_mount *mp,
185 const struct super_block *sb = mp->m_super;
187 substring_t args[MAX_OPT_ARGS];
191 __uint8_t iosizelog = 0;
194 * set up the mount name first so all the errors will refer to the
197 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
200 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
203 * Copy binary VFS mount flags we are interested in.
205 if (sb->s_flags & MS_RDONLY)
206 mp->m_flags |= XFS_MOUNT_RDONLY;
207 if (sb->s_flags & MS_DIRSYNC)
208 mp->m_flags |= XFS_MOUNT_DIRSYNC;
209 if (sb->s_flags & MS_SYNCHRONOUS)
210 mp->m_flags |= XFS_MOUNT_WSYNC;
213 * Set some default flags that could be cleared by the mount option
216 mp->m_flags |= XFS_MOUNT_BARRIER;
217 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
220 * These can be overridden by the mount option parsing.
228 while ((p = strsep(&options, ",")) != NULL) {
234 token = match_token(p, tokens, args);
237 if (match_int(args, &mp->m_logbufs))
241 if (suffix_kstrtoint(args, 10, &mp->m_logbsize))
245 mp->m_logname = match_strdup(args);
250 xfs_warn(mp, "%s option not allowed on this system", p);
253 mp->m_rtname = match_strdup(args);
259 if (suffix_kstrtoint(args, 10, &iosize))
261 iosizelog = ffs(iosize) - 1;
265 mp->m_flags |= XFS_MOUNT_GRPID;
269 mp->m_flags &= ~XFS_MOUNT_GRPID;
272 mp->m_flags |= XFS_MOUNT_WSYNC;
275 mp->m_flags |= XFS_MOUNT_NORECOVERY;
278 mp->m_flags |= XFS_MOUNT_NOALIGN;
281 mp->m_flags |= XFS_MOUNT_SWALLOC;
284 if (match_int(args, &dsunit))
288 if (match_int(args, &dswidth))
292 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
295 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
298 mp->m_flags |= XFS_MOUNT_NOUUID;
301 mp->m_flags |= XFS_MOUNT_BARRIER;
304 mp->m_flags &= ~XFS_MOUNT_BARRIER;
307 mp->m_flags |= XFS_MOUNT_IKEEP;
310 mp->m_flags &= ~XFS_MOUNT_IKEEP;
313 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
316 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
319 mp->m_flags |= XFS_MOUNT_ATTR2;
322 mp->m_flags &= ~XFS_MOUNT_ATTR2;
323 mp->m_flags |= XFS_MOUNT_NOATTR2;
325 case Opt_filestreams:
326 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
329 mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT;
330 mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD;
331 mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE;
336 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
340 case Opt_uqnoenforce:
341 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
342 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
346 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
349 case Opt_pqnoenforce:
350 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
351 mp->m_qflags &= ~XFS_PQUOTA_ENFD;
354 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
357 case Opt_gqnoenforce:
358 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
359 mp->m_qflags &= ~XFS_GQUOTA_ENFD;
362 mp->m_flags |= XFS_MOUNT_DISCARD;
365 mp->m_flags &= ~XFS_MOUNT_DISCARD;
369 mp->m_flags |= XFS_MOUNT_DAX;
373 xfs_warn(mp, "unknown mount option [%s].", p);
379 * no recovery flag requires a read-only mount
381 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
382 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
383 xfs_warn(mp, "no-recovery mounts must be read-only.");
387 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
389 "sunit and swidth options incompatible with the noalign option");
393 #ifndef CONFIG_XFS_QUOTA
394 if (XFS_IS_QUOTA_RUNNING(mp)) {
395 xfs_warn(mp, "quota support not available in this kernel.");
400 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
401 xfs_warn(mp, "sunit and swidth must be specified together");
405 if (dsunit && (dswidth % dsunit != 0)) {
407 "stripe width (%d) must be a multiple of the stripe unit (%d)",
413 if (dsunit && !(mp->m_flags & XFS_MOUNT_NOALIGN)) {
415 * At this point the superblock has not been read
416 * in, therefore we do not know the block size.
417 * Before the mount call ends we will convert
420 mp->m_dalign = dsunit;
421 mp->m_swidth = dswidth;
424 if (mp->m_logbufs != -1 &&
425 mp->m_logbufs != 0 &&
426 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
427 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
428 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
429 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
432 if (mp->m_logbsize != -1 &&
433 mp->m_logbsize != 0 &&
434 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
435 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
436 !is_power_of_2(mp->m_logbsize))) {
438 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
444 if (iosizelog > XFS_MAX_IO_LOG ||
445 iosizelog < XFS_MIN_IO_LOG) {
446 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]",
447 iosizelog, XFS_MIN_IO_LOG,
452 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
453 mp->m_readio_log = iosizelog;
454 mp->m_writeio_log = iosizelog;
460 struct proc_xfs_info {
467 struct xfs_mount *mp,
470 static struct proc_xfs_info xfs_info_set[] = {
471 /* the few simple ones we can get from the mount struct */
472 { XFS_MOUNT_IKEEP, ",ikeep" },
473 { XFS_MOUNT_WSYNC, ",wsync" },
474 { XFS_MOUNT_NOALIGN, ",noalign" },
475 { XFS_MOUNT_SWALLOC, ",swalloc" },
476 { XFS_MOUNT_NOUUID, ",nouuid" },
477 { XFS_MOUNT_NORECOVERY, ",norecovery" },
478 { XFS_MOUNT_ATTR2, ",attr2" },
479 { XFS_MOUNT_FILESTREAMS, ",filestreams" },
480 { XFS_MOUNT_GRPID, ",grpid" },
481 { XFS_MOUNT_DISCARD, ",discard" },
482 { XFS_MOUNT_SMALL_INUMS, ",inode32" },
483 { XFS_MOUNT_DAX, ",dax" },
486 static struct proc_xfs_info xfs_info_unset[] = {
487 /* the few simple ones we can get from the mount struct */
488 { XFS_MOUNT_COMPAT_IOSIZE, ",largeio" },
489 { XFS_MOUNT_BARRIER, ",nobarrier" },
490 { XFS_MOUNT_SMALL_INUMS, ",inode64" },
493 struct proc_xfs_info *xfs_infop;
495 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
496 if (mp->m_flags & xfs_infop->flag)
497 seq_puts(m, xfs_infop->str);
499 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
500 if (!(mp->m_flags & xfs_infop->flag))
501 seq_puts(m, xfs_infop->str);
504 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
505 seq_printf(m, ",allocsize=%dk",
506 (int)(1 << mp->m_writeio_log) >> 10);
508 if (mp->m_logbufs > 0)
509 seq_printf(m, ",logbufs=%d", mp->m_logbufs);
510 if (mp->m_logbsize > 0)
511 seq_printf(m, ",logbsize=%dk", mp->m_logbsize >> 10);
514 seq_show_option(m, "logdev", mp->m_logname);
516 seq_show_option(m, "rtdev", mp->m_rtname);
518 if (mp->m_dalign > 0)
519 seq_printf(m, ",sunit=%d",
520 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
521 if (mp->m_swidth > 0)
522 seq_printf(m, ",swidth=%d",
523 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
525 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
526 seq_puts(m, ",usrquota");
527 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
528 seq_puts(m, ",uqnoenforce");
530 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
531 if (mp->m_qflags & XFS_PQUOTA_ENFD)
532 seq_puts(m, ",prjquota");
534 seq_puts(m, ",pqnoenforce");
536 if (mp->m_qflags & XFS_GQUOTA_ACCT) {
537 if (mp->m_qflags & XFS_GQUOTA_ENFD)
538 seq_puts(m, ",grpquota");
540 seq_puts(m, ",gqnoenforce");
543 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
544 seq_puts(m, ",noquota");
550 unsigned int blockshift)
552 unsigned int pagefactor = 1;
553 unsigned int bitshift = BITS_PER_LONG - 1;
555 /* Figure out maximum filesize, on Linux this can depend on
556 * the filesystem blocksize (on 32 bit platforms).
557 * __block_write_begin does this in an [unsigned] long...
558 * page->index << (PAGE_CACHE_SHIFT - bbits)
559 * So, for page sized blocks (4K on 32 bit platforms),
560 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
561 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
562 * but for smaller blocksizes it is less (bbits = log2 bsize).
563 * Note1: get_block_t takes a long (implicit cast from above)
564 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
565 * can optionally convert the [unsigned] long from above into
566 * an [unsigned] long long.
569 #if BITS_PER_LONG == 32
570 # if defined(CONFIG_LBDAF)
571 ASSERT(sizeof(sector_t) == 8);
572 pagefactor = PAGE_CACHE_SIZE;
573 bitshift = BITS_PER_LONG;
575 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
579 return (((__uint64_t)pagefactor) << bitshift) - 1;
583 * xfs_set_inode32() and xfs_set_inode64() are passed an agcount
584 * because in the growfs case, mp->m_sb.sb_agcount is not updated
585 * yet to the potentially higher ag count.
588 xfs_set_inode32(struct xfs_mount *mp, xfs_agnumber_t agcount)
590 xfs_agnumber_t index = 0;
591 xfs_agnumber_t maxagi = 0;
592 xfs_sb_t *sbp = &mp->m_sb;
593 xfs_agnumber_t max_metadata;
598 /* Calculate how much should be reserved for inodes to meet
599 * the max inode percentage.
601 if (mp->m_maxicount) {
604 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
606 icount += sbp->sb_agblocks - 1;
607 do_div(icount, sbp->sb_agblocks);
608 max_metadata = icount;
610 max_metadata = agcount;
613 agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks - 1, 0);
615 for (index = 0; index < agcount; index++) {
616 ino = XFS_AGINO_TO_INO(mp, index, agino);
618 if (ino > XFS_MAXINUMBER_32) {
619 pag = xfs_perag_get(mp, index);
620 pag->pagi_inodeok = 0;
621 pag->pagf_metadata = 0;
626 pag = xfs_perag_get(mp, index);
627 pag->pagi_inodeok = 1;
629 if (index < max_metadata)
630 pag->pagf_metadata = 1;
633 mp->m_flags |= (XFS_MOUNT_32BITINODES |
634 XFS_MOUNT_SMALL_INUMS);
640 xfs_set_inode64(struct xfs_mount *mp, xfs_agnumber_t agcount)
642 xfs_agnumber_t index = 0;
644 for (index = 0; index < agcount; index++) {
645 struct xfs_perag *pag;
647 pag = xfs_perag_get(mp, index);
648 pag->pagi_inodeok = 1;
649 pag->pagf_metadata = 0;
653 /* There is no need for lock protection on m_flags,
654 * the rw_semaphore of the VFS superblock is locked
655 * during mount/umount/remount operations, so this is
656 * enough to avoid concurency on the m_flags field
658 mp->m_flags &= ~(XFS_MOUNT_32BITINODES |
659 XFS_MOUNT_SMALL_INUMS);
667 struct block_device **bdevp)
671 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
673 if (IS_ERR(*bdevp)) {
674 error = PTR_ERR(*bdevp);
675 xfs_warn(mp, "Invalid device [%s], error=%d", name, error);
683 struct block_device *bdev)
686 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
690 xfs_blkdev_issue_flush(
691 xfs_buftarg_t *buftarg)
693 blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL);
698 struct xfs_mount *mp)
700 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
701 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
702 xfs_free_buftarg(mp, mp->m_logdev_targp);
703 xfs_blkdev_put(logdev);
705 if (mp->m_rtdev_targp) {
706 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
707 xfs_free_buftarg(mp, mp->m_rtdev_targp);
708 xfs_blkdev_put(rtdev);
710 xfs_free_buftarg(mp, mp->m_ddev_targp);
714 * The file system configurations are:
715 * (1) device (partition) with data and internal log
716 * (2) logical volume with data and log subvolumes.
717 * (3) logical volume with data, log, and realtime subvolumes.
719 * We only have to handle opening the log and realtime volumes here if
720 * they are present. The data subvolume has already been opened by
721 * get_sb_bdev() and is stored in sb->s_bdev.
725 struct xfs_mount *mp)
727 struct block_device *ddev = mp->m_super->s_bdev;
728 struct block_device *logdev = NULL, *rtdev = NULL;
732 * Open real time and log devices - order is important.
735 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
741 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
743 goto out_close_logdev;
745 if (rtdev == ddev || rtdev == logdev) {
747 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
749 goto out_close_rtdev;
754 * Setup xfs_mount buffer target pointers
757 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev);
758 if (!mp->m_ddev_targp)
759 goto out_close_rtdev;
762 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev);
763 if (!mp->m_rtdev_targp)
764 goto out_free_ddev_targ;
767 if (logdev && logdev != ddev) {
768 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev);
769 if (!mp->m_logdev_targp)
770 goto out_free_rtdev_targ;
772 mp->m_logdev_targp = mp->m_ddev_targp;
778 if (mp->m_rtdev_targp)
779 xfs_free_buftarg(mp, mp->m_rtdev_targp);
781 xfs_free_buftarg(mp, mp->m_ddev_targp);
783 xfs_blkdev_put(rtdev);
785 if (logdev && logdev != ddev)
786 xfs_blkdev_put(logdev);
792 * Setup xfs_mount buffer target pointers based on superblock
796 struct xfs_mount *mp)
800 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize);
804 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
805 unsigned int log_sector_size = BBSIZE;
807 if (xfs_sb_version_hassector(&mp->m_sb))
808 log_sector_size = mp->m_sb.sb_logsectsize;
809 error = xfs_setsize_buftarg(mp->m_logdev_targp,
814 if (mp->m_rtdev_targp) {
815 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
816 mp->m_sb.sb_sectsize);
825 xfs_init_mount_workqueues(
826 struct xfs_mount *mp)
828 mp->m_buf_workqueue = alloc_workqueue("xfs-buf/%s",
829 WQ_MEM_RECLAIM|WQ_FREEZABLE, 1, mp->m_fsname);
830 if (!mp->m_buf_workqueue)
833 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s",
834 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
835 if (!mp->m_data_workqueue)
836 goto out_destroy_buf;
838 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
839 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
840 if (!mp->m_unwritten_workqueue)
841 goto out_destroy_data_iodone_queue;
843 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s",
844 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
845 if (!mp->m_cil_workqueue)
846 goto out_destroy_unwritten;
848 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
849 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
850 if (!mp->m_reclaim_workqueue)
851 goto out_destroy_cil;
853 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
854 WQ_MEM_RECLAIM|WQ_FREEZABLE|WQ_HIGHPRI, 0,
856 if (!mp->m_log_workqueue)
857 goto out_destroy_reclaim;
859 mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s",
860 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
861 if (!mp->m_eofblocks_workqueue)
862 goto out_destroy_log;
867 destroy_workqueue(mp->m_log_workqueue);
869 destroy_workqueue(mp->m_reclaim_workqueue);
871 destroy_workqueue(mp->m_cil_workqueue);
872 out_destroy_unwritten:
873 destroy_workqueue(mp->m_unwritten_workqueue);
874 out_destroy_data_iodone_queue:
875 destroy_workqueue(mp->m_data_workqueue);
877 destroy_workqueue(mp->m_buf_workqueue);
883 xfs_destroy_mount_workqueues(
884 struct xfs_mount *mp)
886 destroy_workqueue(mp->m_eofblocks_workqueue);
887 destroy_workqueue(mp->m_log_workqueue);
888 destroy_workqueue(mp->m_reclaim_workqueue);
889 destroy_workqueue(mp->m_cil_workqueue);
890 destroy_workqueue(mp->m_data_workqueue);
891 destroy_workqueue(mp->m_unwritten_workqueue);
892 destroy_workqueue(mp->m_buf_workqueue);
896 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
897 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
898 * for IO to complete so that we effectively throttle multiple callers to the
899 * rate at which IO is completing.
903 struct xfs_mount *mp)
905 struct super_block *sb = mp->m_super;
907 if (down_read_trylock(&sb->s_umount)) {
909 up_read(&sb->s_umount);
913 /* Catch misguided souls that try to use this interface on XFS */
914 STATIC struct inode *
916 struct super_block *sb)
923 * Now that the generic code is guaranteed not to be accessing
924 * the linux inode, we can reclaim the inode.
927 xfs_fs_destroy_inode(
930 struct xfs_inode *ip = XFS_I(inode);
932 trace_xfs_destroy_inode(ip);
934 XFS_STATS_INC(ip->i_mount, vn_reclaim);
936 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
939 * We should never get here with one of the reclaim flags already set.
941 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
942 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
945 * We always use background reclaim here because even if the
946 * inode is clean, it still may be under IO and hence we have
947 * to take the flush lock. The background reclaim path handles
948 * this more efficiently than we can here, so simply let background
949 * reclaim tear down all inodes.
951 xfs_inode_set_reclaim_tag(ip);
955 * Slab object creation initialisation for the XFS inode.
956 * This covers only the idempotent fields in the XFS inode;
957 * all other fields need to be initialised on allocation
958 * from the slab. This avoids the need to repeatedly initialise
959 * fields in the xfs inode that left in the initialise state
960 * when freeing the inode.
963 xfs_fs_inode_init_once(
966 struct xfs_inode *ip = inode;
968 memset(ip, 0, sizeof(struct xfs_inode));
971 inode_init_once(VFS_I(ip));
974 atomic_set(&ip->i_pincount, 0);
975 spin_lock_init(&ip->i_flags_lock);
977 mrlock_init(&ip->i_mmaplock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
978 "xfsino", ip->i_ino);
979 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
980 "xfsino", ip->i_ino);
987 xfs_inode_t *ip = XFS_I(inode);
989 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
991 trace_xfs_evict_inode(ip);
993 truncate_inode_pages_final(&inode->i_data);
995 XFS_STATS_INC(ip->i_mount, vn_rele);
996 XFS_STATS_INC(ip->i_mount, vn_remove);
1002 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1003 * serialised against cache hits here via the inode->i_lock and igrab() in
1004 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1005 * racing with us, and it avoids needing to grab a spinlock here for every inode
1006 * we drop the final reference on.
1010 struct inode *inode)
1012 struct xfs_inode *ip = XFS_I(inode);
1014 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);
1019 struct xfs_mount *mp)
1021 kfree(mp->m_fsname);
1022 kfree(mp->m_rtname);
1023 kfree(mp->m_logname);
1028 struct super_block *sb,
1031 struct xfs_mount *mp = XFS_M(sb);
1034 * Doing anything during the async pass would be counterproductive.
1039 xfs_log_force(mp, XFS_LOG_SYNC);
1042 * The disk must be active because we're syncing.
1043 * We schedule log work now (now that the disk is
1044 * active) instead of later (when it might not be).
1046 flush_delayed_work(&mp->m_log->l_work);
1054 struct dentry *dentry,
1055 struct kstatfs *statp)
1057 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1058 xfs_sb_t *sbp = &mp->m_sb;
1059 struct xfs_inode *ip = XFS_I(d_inode(dentry));
1060 __uint64_t fakeinos, id;
1063 __uint64_t fdblocks;
1067 statp->f_type = XFS_SB_MAGIC;
1068 statp->f_namelen = MAXNAMELEN - 1;
1070 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1071 statp->f_fsid.val[0] = (u32)id;
1072 statp->f_fsid.val[1] = (u32)(id >> 32);
1074 icount = percpu_counter_sum(&mp->m_icount);
1075 ifree = percpu_counter_sum(&mp->m_ifree);
1076 fdblocks = percpu_counter_sum(&mp->m_fdblocks);
1078 spin_lock(&mp->m_sb_lock);
1079 statp->f_bsize = sbp->sb_blocksize;
1080 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1081 statp->f_blocks = sbp->sb_dblocks - lsize;
1082 spin_unlock(&mp->m_sb_lock);
1084 statp->f_bfree = fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1085 statp->f_bavail = statp->f_bfree;
1087 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1088 statp->f_files = MIN(icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1089 if (mp->m_maxicount)
1090 statp->f_files = min_t(typeof(statp->f_files),
1094 /* If sb_icount overshot maxicount, report actual allocation */
1095 statp->f_files = max_t(typeof(statp->f_files),
1099 /* make sure statp->f_ffree does not underflow */
1100 ffree = statp->f_files - (icount - ifree);
1101 statp->f_ffree = max_t(__int64_t, ffree, 0);
1104 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
1105 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) ==
1106 (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))
1107 xfs_qm_statvfs(ip, statp);
1112 xfs_save_resvblks(struct xfs_mount *mp)
1114 __uint64_t resblks = 0;
1116 mp->m_resblks_save = mp->m_resblks;
1117 xfs_reserve_blocks(mp, &resblks, NULL);
1121 xfs_restore_resvblks(struct xfs_mount *mp)
1125 if (mp->m_resblks_save) {
1126 resblks = mp->m_resblks_save;
1127 mp->m_resblks_save = 0;
1129 resblks = xfs_default_resblks(mp);
1131 xfs_reserve_blocks(mp, &resblks, NULL);
1135 * Trigger writeback of all the dirty metadata in the file system.
1137 * This ensures that the metadata is written to their location on disk rather
1138 * than just existing in transactions in the log. This means after a quiesce
1139 * there is no log replay required to write the inodes to disk - this is the
1140 * primary difference between a sync and a quiesce.
1142 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1143 * it is started again when appropriate.
1147 struct xfs_mount *mp)
1151 /* wait for all modifications to complete */
1152 while (atomic_read(&mp->m_active_trans) > 0)
1155 /* force the log to unpin objects from the now complete transactions */
1156 xfs_log_force(mp, XFS_LOG_SYNC);
1158 /* reclaim inodes to do any IO before the freeze completes */
1159 xfs_reclaim_inodes(mp, 0);
1160 xfs_reclaim_inodes(mp, SYNC_WAIT);
1162 /* Push the superblock and write an unmount record */
1163 error = xfs_log_sbcount(mp);
1165 xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
1166 "Frozen image may not be consistent.");
1168 * Just warn here till VFS can correctly support
1169 * read-only remount without racing.
1171 WARN_ON(atomic_read(&mp->m_active_trans) != 0);
1173 xfs_log_quiesce(mp);
1177 xfs_test_remount_options(
1178 struct super_block *sb,
1179 struct xfs_mount *mp,
1183 struct xfs_mount *tmp_mp;
1185 tmp_mp = kmem_zalloc(sizeof(*tmp_mp), KM_MAYFAIL);
1189 tmp_mp->m_super = sb;
1190 error = xfs_parseargs(tmp_mp, options);
1191 xfs_free_fsname(tmp_mp);
1199 struct super_block *sb,
1203 struct xfs_mount *mp = XFS_M(sb);
1204 xfs_sb_t *sbp = &mp->m_sb;
1205 substring_t args[MAX_OPT_ARGS];
1209 /* First, check for complete junk; i.e. invalid options */
1210 error = xfs_test_remount_options(sb, mp, options);
1214 sync_filesystem(sb);
1215 while ((p = strsep(&options, ",")) != NULL) {
1221 token = match_token(p, tokens, args);
1224 mp->m_flags |= XFS_MOUNT_BARRIER;
1227 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1230 mp->m_maxagi = xfs_set_inode64(mp, sbp->sb_agcount);
1233 mp->m_maxagi = xfs_set_inode32(mp, sbp->sb_agcount);
1237 * Logically we would return an error here to prevent
1238 * users from believing they might have changed
1239 * mount options using remount which can't be changed.
1241 * But unfortunately mount(8) adds all options from
1242 * mtab and fstab to the mount arguments in some cases
1243 * so we can't blindly reject options, but have to
1244 * check for each specified option if it actually
1245 * differs from the currently set option and only
1246 * reject it if that's the case.
1248 * Until that is implemented we return success for
1249 * every remount request, and silently ignore all
1250 * options that we can't actually change.
1254 "mount option \"%s\" not supported for remount", p);
1263 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1264 if (mp->m_flags & XFS_MOUNT_NORECOVERY) {
1266 "ro->rw transition prohibited on norecovery mount");
1270 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1273 * If this is the first remount to writeable state we
1274 * might have some superblock changes to update.
1276 if (mp->m_update_sb) {
1277 error = xfs_sync_sb(mp, false);
1279 xfs_warn(mp, "failed to write sb changes");
1282 mp->m_update_sb = false;
1286 * Fill out the reserve pool if it is empty. Use the stashed
1287 * value if it is non-zero, otherwise go with the default.
1289 xfs_restore_resvblks(mp);
1290 xfs_log_work_queue(mp);
1294 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1296 * Before we sync the metadata, we need to free up the reserve
1297 * block pool so that the used block count in the superblock on
1298 * disk is correct at the end of the remount. Stash the current
1299 * reserve pool size so that if we get remounted rw, we can
1300 * return it to the same size.
1302 xfs_save_resvblks(mp);
1303 xfs_quiesce_attr(mp);
1304 mp->m_flags |= XFS_MOUNT_RDONLY;
1311 * Second stage of a freeze. The data is already frozen so we only
1312 * need to take care of the metadata. Once that's done sync the superblock
1313 * to the log to dirty it in case of a crash while frozen. This ensures that we
1314 * will recover the unlinked inode lists on the next mount.
1318 struct super_block *sb)
1320 struct xfs_mount *mp = XFS_M(sb);
1322 xfs_save_resvblks(mp);
1323 xfs_quiesce_attr(mp);
1324 return xfs_sync_sb(mp, true);
1329 struct super_block *sb)
1331 struct xfs_mount *mp = XFS_M(sb);
1333 xfs_restore_resvblks(mp);
1334 xfs_log_work_queue(mp);
1339 xfs_fs_show_options(
1341 struct dentry *root)
1343 return xfs_showargs(XFS_M(root->d_sb), m);
1347 * This function fills in xfs_mount_t fields based on mount args.
1348 * Note: the superblock _has_ now been read in.
1352 struct xfs_mount *mp)
1354 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1356 /* Fail a mount where the logbuf is smaller than the log stripe */
1357 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1358 if (mp->m_logbsize <= 0 &&
1359 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1360 mp->m_logbsize = mp->m_sb.sb_logsunit;
1361 } else if (mp->m_logbsize > 0 &&
1362 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1364 "logbuf size must be greater than or equal to log stripe size");
1368 /* Fail a mount if the logbuf is larger than 32K */
1369 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1371 "logbuf size for version 1 logs must be 16K or 32K");
1377 * V5 filesystems always use attr2 format for attributes.
1379 if (xfs_sb_version_hascrc(&mp->m_sb) &&
1380 (mp->m_flags & XFS_MOUNT_NOATTR2)) {
1381 xfs_warn(mp, "Cannot mount a V5 filesystem as noattr2. "
1382 "attr2 is always enabled for V5 filesystems.");
1387 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1388 * told by noattr2 to turn it off
1390 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1391 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1392 mp->m_flags |= XFS_MOUNT_ATTR2;
1395 * prohibit r/w mounts of read-only filesystems
1397 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1399 "cannot mount a read-only filesystem as read-write");
1403 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
1404 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) &&
1405 !xfs_sb_version_has_pquotino(&mp->m_sb)) {
1407 "Super block does not support project and group quota together");
1415 xfs_init_percpu_counters(
1416 struct xfs_mount *mp)
1420 error = percpu_counter_init(&mp->m_icount, 0, GFP_KERNEL);
1424 error = percpu_counter_init(&mp->m_ifree, 0, GFP_KERNEL);
1428 error = percpu_counter_init(&mp->m_fdblocks, 0, GFP_KERNEL);
1435 percpu_counter_destroy(&mp->m_ifree);
1437 percpu_counter_destroy(&mp->m_icount);
1442 xfs_reinit_percpu_counters(
1443 struct xfs_mount *mp)
1445 percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount);
1446 percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree);
1447 percpu_counter_set(&mp->m_fdblocks, mp->m_sb.sb_fdblocks);
1451 xfs_destroy_percpu_counters(
1452 struct xfs_mount *mp)
1454 percpu_counter_destroy(&mp->m_icount);
1455 percpu_counter_destroy(&mp->m_ifree);
1456 percpu_counter_destroy(&mp->m_fdblocks);
1461 struct super_block *sb,
1466 struct xfs_mount *mp = NULL;
1467 int flags = 0, error = -ENOMEM;
1469 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1473 spin_lock_init(&mp->m_sb_lock);
1474 mutex_init(&mp->m_growlock);
1475 atomic_set(&mp->m_active_trans, 0);
1476 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1477 INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
1478 mp->m_kobj.kobject.kset = xfs_kset;
1483 error = xfs_parseargs(mp, (char *)data);
1485 goto out_free_fsname;
1487 sb_min_blocksize(sb, BBSIZE);
1488 sb->s_xattr = xfs_xattr_handlers;
1489 sb->s_export_op = &xfs_export_operations;
1490 #ifdef CONFIG_XFS_QUOTA
1491 sb->s_qcop = &xfs_quotactl_operations;
1492 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
1494 sb->s_op = &xfs_super_operations;
1497 flags |= XFS_MFSI_QUIET;
1499 error = xfs_open_devices(mp);
1501 goto out_free_fsname;
1503 error = xfs_init_mount_workqueues(mp);
1505 goto out_close_devices;
1507 error = xfs_init_percpu_counters(mp);
1509 goto out_destroy_workqueues;
1511 /* Allocate stats memory before we do operations that might use it */
1512 mp->m_stats.xs_stats = alloc_percpu(struct xfsstats);
1513 if (!mp->m_stats.xs_stats) {
1515 goto out_destroy_counters;
1518 error = xfs_readsb(mp, flags);
1520 goto out_free_stats;
1522 error = xfs_finish_flags(mp);
1526 error = xfs_setup_devices(mp);
1530 error = xfs_filestream_mount(mp);
1535 * we must configure the block size in the superblock before we run the
1536 * full mount process as the mount process can lookup and cache inodes.
1538 sb->s_magic = XFS_SB_MAGIC;
1539 sb->s_blocksize = mp->m_sb.sb_blocksize;
1540 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1541 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1542 sb->s_max_links = XFS_MAXLINK;
1543 sb->s_time_gran = 1;
1544 set_posix_acl_flag(sb);
1546 /* version 5 superblocks support inode version counters. */
1547 if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
1548 sb->s_flags |= MS_I_VERSION;
1550 if (mp->m_flags & XFS_MOUNT_DAX) {
1552 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1553 if (sb->s_blocksize != PAGE_SIZE) {
1555 "Filesystem block size invalid for DAX Turning DAX off.");
1556 mp->m_flags &= ~XFS_MOUNT_DAX;
1557 } else if (!sb->s_bdev->bd_disk->fops->direct_access) {
1559 "Block device does not support DAX Turning DAX off.");
1560 mp->m_flags &= ~XFS_MOUNT_DAX;
1564 if (xfs_sb_version_hassparseinodes(&mp->m_sb))
1566 "EXPERIMENTAL sparse inode feature enabled. Use at your own risk!");
1568 error = xfs_mountfs(mp);
1570 goto out_filestream_unmount;
1572 root = igrab(VFS_I(mp->m_rootip));
1577 sb->s_root = d_make_root(root);
1585 out_filestream_unmount:
1586 xfs_filestream_unmount(mp);
1590 free_percpu(mp->m_stats.xs_stats);
1591 out_destroy_counters:
1592 xfs_destroy_percpu_counters(mp);
1593 out_destroy_workqueues:
1594 xfs_destroy_mount_workqueues(mp);
1596 xfs_close_devices(mp);
1598 xfs_free_fsname(mp);
1604 xfs_filestream_unmount(mp);
1611 struct super_block *sb)
1613 struct xfs_mount *mp = XFS_M(sb);
1615 xfs_notice(mp, "Unmounting Filesystem");
1616 xfs_filestream_unmount(mp);
1620 free_percpu(mp->m_stats.xs_stats);
1621 xfs_destroy_percpu_counters(mp);
1622 xfs_destroy_mount_workqueues(mp);
1623 xfs_close_devices(mp);
1624 xfs_free_fsname(mp);
1628 STATIC struct dentry *
1630 struct file_system_type *fs_type,
1632 const char *dev_name,
1635 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1639 xfs_fs_nr_cached_objects(
1640 struct super_block *sb,
1641 struct shrink_control *sc)
1643 return xfs_reclaim_inodes_count(XFS_M(sb));
1647 xfs_fs_free_cached_objects(
1648 struct super_block *sb,
1649 struct shrink_control *sc)
1651 return xfs_reclaim_inodes_nr(XFS_M(sb), sc->nr_to_scan);
1654 static const struct super_operations xfs_super_operations = {
1655 .alloc_inode = xfs_fs_alloc_inode,
1656 .destroy_inode = xfs_fs_destroy_inode,
1657 .evict_inode = xfs_fs_evict_inode,
1658 .drop_inode = xfs_fs_drop_inode,
1659 .put_super = xfs_fs_put_super,
1660 .sync_fs = xfs_fs_sync_fs,
1661 .freeze_fs = xfs_fs_freeze,
1662 .unfreeze_fs = xfs_fs_unfreeze,
1663 .statfs = xfs_fs_statfs,
1664 .remount_fs = xfs_fs_remount,
1665 .show_options = xfs_fs_show_options,
1666 .nr_cached_objects = xfs_fs_nr_cached_objects,
1667 .free_cached_objects = xfs_fs_free_cached_objects,
1670 static struct file_system_type xfs_fs_type = {
1671 .owner = THIS_MODULE,
1673 .mount = xfs_fs_mount,
1674 .kill_sb = kill_block_super,
1675 .fs_flags = FS_REQUIRES_DEV,
1677 MODULE_ALIAS_FS("xfs");
1680 xfs_init_zones(void)
1683 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1684 if (!xfs_ioend_zone)
1687 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1689 if (!xfs_ioend_pool)
1690 goto out_destroy_ioend_zone;
1692 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1694 if (!xfs_log_ticket_zone)
1695 goto out_destroy_ioend_pool;
1697 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1698 "xfs_bmap_free_item");
1699 if (!xfs_bmap_free_item_zone)
1700 goto out_destroy_log_ticket_zone;
1702 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1704 if (!xfs_btree_cur_zone)
1705 goto out_destroy_bmap_free_item_zone;
1707 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1709 if (!xfs_da_state_zone)
1710 goto out_destroy_btree_cur_zone;
1712 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1713 if (!xfs_ifork_zone)
1714 goto out_destroy_da_state_zone;
1716 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1717 if (!xfs_trans_zone)
1718 goto out_destroy_ifork_zone;
1720 xfs_log_item_desc_zone =
1721 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1722 "xfs_log_item_desc");
1723 if (!xfs_log_item_desc_zone)
1724 goto out_destroy_trans_zone;
1727 * The size of the zone allocated buf log item is the maximum
1728 * size possible under XFS. This wastes a little bit of memory,
1729 * but it is much faster.
1731 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item),
1733 if (!xfs_buf_item_zone)
1734 goto out_destroy_log_item_desc_zone;
1736 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1737 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1738 sizeof(xfs_extent_t))), "xfs_efd_item");
1740 goto out_destroy_buf_item_zone;
1742 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1743 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1744 sizeof(xfs_extent_t))), "xfs_efi_item");
1746 goto out_destroy_efd_zone;
1749 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1750 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD |
1751 KM_ZONE_ACCOUNT, xfs_fs_inode_init_once);
1752 if (!xfs_inode_zone)
1753 goto out_destroy_efi_zone;
1756 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1757 KM_ZONE_SPREAD, NULL);
1759 goto out_destroy_inode_zone;
1760 xfs_icreate_zone = kmem_zone_init(sizeof(struct xfs_icreate_item),
1762 if (!xfs_icreate_zone)
1763 goto out_destroy_ili_zone;
1767 out_destroy_ili_zone:
1768 kmem_zone_destroy(xfs_ili_zone);
1769 out_destroy_inode_zone:
1770 kmem_zone_destroy(xfs_inode_zone);
1771 out_destroy_efi_zone:
1772 kmem_zone_destroy(xfs_efi_zone);
1773 out_destroy_efd_zone:
1774 kmem_zone_destroy(xfs_efd_zone);
1775 out_destroy_buf_item_zone:
1776 kmem_zone_destroy(xfs_buf_item_zone);
1777 out_destroy_log_item_desc_zone:
1778 kmem_zone_destroy(xfs_log_item_desc_zone);
1779 out_destroy_trans_zone:
1780 kmem_zone_destroy(xfs_trans_zone);
1781 out_destroy_ifork_zone:
1782 kmem_zone_destroy(xfs_ifork_zone);
1783 out_destroy_da_state_zone:
1784 kmem_zone_destroy(xfs_da_state_zone);
1785 out_destroy_btree_cur_zone:
1786 kmem_zone_destroy(xfs_btree_cur_zone);
1787 out_destroy_bmap_free_item_zone:
1788 kmem_zone_destroy(xfs_bmap_free_item_zone);
1789 out_destroy_log_ticket_zone:
1790 kmem_zone_destroy(xfs_log_ticket_zone);
1791 out_destroy_ioend_pool:
1792 mempool_destroy(xfs_ioend_pool);
1793 out_destroy_ioend_zone:
1794 kmem_zone_destroy(xfs_ioend_zone);
1800 xfs_destroy_zones(void)
1803 * Make sure all delayed rcu free are flushed before we
1807 kmem_zone_destroy(xfs_icreate_zone);
1808 kmem_zone_destroy(xfs_ili_zone);
1809 kmem_zone_destroy(xfs_inode_zone);
1810 kmem_zone_destroy(xfs_efi_zone);
1811 kmem_zone_destroy(xfs_efd_zone);
1812 kmem_zone_destroy(xfs_buf_item_zone);
1813 kmem_zone_destroy(xfs_log_item_desc_zone);
1814 kmem_zone_destroy(xfs_trans_zone);
1815 kmem_zone_destroy(xfs_ifork_zone);
1816 kmem_zone_destroy(xfs_da_state_zone);
1817 kmem_zone_destroy(xfs_btree_cur_zone);
1818 kmem_zone_destroy(xfs_bmap_free_item_zone);
1819 kmem_zone_destroy(xfs_log_ticket_zone);
1820 mempool_destroy(xfs_ioend_pool);
1821 kmem_zone_destroy(xfs_ioend_zone);
1826 xfs_init_workqueues(void)
1829 * The allocation workqueue can be used in memory reclaim situations
1830 * (writepage path), and parallelism is only limited by the number of
1831 * AGs in all the filesystems mounted. Hence use the default large
1832 * max_active value for this workqueue.
1834 xfs_alloc_wq = alloc_workqueue("xfsalloc",
1835 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0);
1843 xfs_destroy_workqueues(void)
1845 destroy_workqueue(xfs_alloc_wq);
1853 printk(KERN_INFO XFS_VERSION_STRING " with "
1854 XFS_BUILD_OPTIONS " enabled\n");
1858 error = xfs_init_zones();
1862 error = xfs_init_workqueues();
1864 goto out_destroy_zones;
1866 error = xfs_mru_cache_init();
1868 goto out_destroy_wq;
1870 error = xfs_buf_init();
1872 goto out_mru_cache_uninit;
1874 error = xfs_init_procfs();
1876 goto out_buf_terminate;
1878 error = xfs_sysctl_register();
1880 goto out_cleanup_procfs;
1882 xfs_kset = kset_create_and_add("xfs", NULL, fs_kobj);
1885 goto out_sysctl_unregister;
1888 xfsstats.xs_kobj.kobject.kset = xfs_kset;
1890 xfsstats.xs_stats = alloc_percpu(struct xfsstats);
1891 if (!xfsstats.xs_stats) {
1893 goto out_kset_unregister;
1896 error = xfs_sysfs_init(&xfsstats.xs_kobj, &xfs_stats_ktype, NULL,
1899 goto out_free_stats;
1902 xfs_dbg_kobj.kobject.kset = xfs_kset;
1903 error = xfs_sysfs_init(&xfs_dbg_kobj, &xfs_dbg_ktype, NULL, "debug");
1905 goto out_remove_stats_kobj;
1908 error = xfs_qm_init();
1910 goto out_remove_dbg_kobj;
1912 error = register_filesystem(&xfs_fs_type);
1919 out_remove_dbg_kobj:
1921 xfs_sysfs_del(&xfs_dbg_kobj);
1922 out_remove_stats_kobj:
1924 xfs_sysfs_del(&xfsstats.xs_kobj);
1926 free_percpu(xfsstats.xs_stats);
1927 out_kset_unregister:
1928 kset_unregister(xfs_kset);
1929 out_sysctl_unregister:
1930 xfs_sysctl_unregister();
1932 xfs_cleanup_procfs();
1934 xfs_buf_terminate();
1935 out_mru_cache_uninit:
1936 xfs_mru_cache_uninit();
1938 xfs_destroy_workqueues();
1940 xfs_destroy_zones();
1949 unregister_filesystem(&xfs_fs_type);
1951 xfs_sysfs_del(&xfs_dbg_kobj);
1953 xfs_sysfs_del(&xfsstats.xs_kobj);
1954 free_percpu(xfsstats.xs_stats);
1955 kset_unregister(xfs_kset);
1956 xfs_sysctl_unregister();
1957 xfs_cleanup_procfs();
1958 xfs_buf_terminate();
1959 xfs_mru_cache_uninit();
1960 xfs_destroy_workqueues();
1961 xfs_destroy_zones();
1962 xfs_uuid_table_free();
1965 module_init(init_xfs_fs);
1966 module_exit(exit_xfs_fs);
1968 MODULE_AUTHOR("Silicon Graphics, Inc.");
1969 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1970 MODULE_LICENSE("GPL");