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
23 #include "xfs_trans.h"
27 #include "xfs_alloc.h"
28 #include "xfs_quota.h"
29 #include "xfs_mount.h"
30 #include "xfs_bmap_btree.h"
31 #include "xfs_alloc_btree.h"
32 #include "xfs_ialloc_btree.h"
33 #include "xfs_dinode.h"
34 #include "xfs_inode.h"
35 #include "xfs_btree.h"
36 #include "xfs_btree_trace.h"
37 #include "xfs_ialloc.h"
39 #include "xfs_rtalloc.h"
40 #include "xfs_error.h"
41 #include "xfs_itable.h"
42 #include "xfs_fsops.h"
44 #include "xfs_buf_item.h"
45 #include "xfs_utils.h"
46 #include "xfs_vnodeops.h"
47 #include "xfs_version.h"
48 #include "xfs_log_priv.h"
49 #include "xfs_trans_priv.h"
50 #include "xfs_filestream.h"
51 #include "xfs_da_btree.h"
52 #include "xfs_extfree_item.h"
53 #include "xfs_mru_cache.h"
54 #include "xfs_inode_item.h"
56 #include "xfs_trace.h"
58 #include <linux/namei.h>
59 #include <linux/init.h>
60 #include <linux/slab.h>
61 #include <linux/mount.h>
62 #include <linux/mempool.h>
63 #include <linux/writeback.h>
64 #include <linux/kthread.h>
65 #include <linux/freezer.h>
66 #include <linux/parser.h>
68 static const struct super_operations xfs_super_operations;
69 static kmem_zone_t *xfs_ioend_zone;
70 mempool_t *xfs_ioend_pool;
72 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
73 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
74 #define MNTOPT_LOGDEV "logdev" /* log device */
75 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
76 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
77 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
78 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
79 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
80 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
81 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
82 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
83 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
84 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
85 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
86 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
87 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
88 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
89 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
90 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
91 * unwritten extent conversion */
92 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
93 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
94 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
95 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
96 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
97 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
99 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
100 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
101 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
102 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
103 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
104 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
105 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
106 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
107 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
108 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
109 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
110 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
111 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
112 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
113 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
114 #define MNTOPT_DELAYLOG "delaylog" /* Delayed loging enabled */
115 #define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed loging disabled */
118 * Table driven mount option parser.
120 * Currently only used for remount, but it will be used for mount
121 * in the future, too.
124 Opt_barrier, Opt_nobarrier, Opt_err
127 static const match_table_t tokens = {
128 {Opt_barrier, "barrier"},
129 {Opt_nobarrier, "nobarrier"},
135 suffix_strtoul(char *s, char **endp, unsigned int base)
137 int last, shift_left_factor = 0;
140 last = strlen(value) - 1;
141 if (value[last] == 'K' || value[last] == 'k') {
142 shift_left_factor = 10;
145 if (value[last] == 'M' || value[last] == 'm') {
146 shift_left_factor = 20;
149 if (value[last] == 'G' || value[last] == 'g') {
150 shift_left_factor = 30;
154 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
158 * This function fills in xfs_mount_t fields based on mount args.
159 * Note: the superblock has _not_ yet been read in.
161 * Note that this function leaks the various device name allocations on
162 * failure. The caller takes care of them.
166 struct xfs_mount *mp,
169 struct super_block *sb = mp->m_super;
170 char *this_char, *value, *eov;
174 __uint8_t iosizelog = 0;
177 * Copy binary VFS mount flags we are interested in.
179 if (sb->s_flags & MS_RDONLY)
180 mp->m_flags |= XFS_MOUNT_RDONLY;
181 if (sb->s_flags & MS_DIRSYNC)
182 mp->m_flags |= XFS_MOUNT_DIRSYNC;
183 if (sb->s_flags & MS_SYNCHRONOUS)
184 mp->m_flags |= XFS_MOUNT_WSYNC;
187 * Set some default flags that could be cleared by the mount option
190 mp->m_flags |= XFS_MOUNT_BARRIER;
191 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
192 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
195 * These can be overridden by the mount option parsing.
203 while ((this_char = strsep(&options, ",")) != NULL) {
206 if ((value = strchr(this_char, '=')) != NULL)
209 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
210 if (!value || !*value) {
212 "XFS: %s option requires an argument",
216 mp->m_logbufs = simple_strtoul(value, &eov, 10);
217 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
218 if (!value || !*value) {
220 "XFS: %s option requires an argument",
224 mp->m_logbsize = suffix_strtoul(value, &eov, 10);
225 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
226 if (!value || !*value) {
228 "XFS: %s option requires an argument",
232 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
235 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
237 "XFS: %s option not allowed on this system",
240 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
241 if (!value || !*value) {
243 "XFS: %s option requires an argument",
247 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
250 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
251 if (!value || !*value) {
253 "XFS: %s option requires an argument",
257 iosize = simple_strtoul(value, &eov, 10);
258 iosizelog = ffs(iosize) - 1;
259 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
260 if (!value || !*value) {
262 "XFS: %s option requires an argument",
266 iosize = suffix_strtoul(value, &eov, 10);
267 iosizelog = ffs(iosize) - 1;
268 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
269 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
270 mp->m_flags |= XFS_MOUNT_GRPID;
271 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
272 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
273 mp->m_flags &= ~XFS_MOUNT_GRPID;
274 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
275 mp->m_flags |= XFS_MOUNT_WSYNC;
276 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
277 mp->m_flags |= XFS_MOUNT_NORECOVERY;
278 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
279 mp->m_flags |= XFS_MOUNT_NOALIGN;
280 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
281 mp->m_flags |= XFS_MOUNT_SWALLOC;
282 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
283 if (!value || !*value) {
285 "XFS: %s option requires an argument",
289 dsunit = simple_strtoul(value, &eov, 10);
290 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
291 if (!value || !*value) {
293 "XFS: %s option requires an argument",
297 dswidth = simple_strtoul(value, &eov, 10);
298 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
299 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
302 "XFS: %s option not allowed on this system",
306 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
307 mp->m_flags |= XFS_MOUNT_NOUUID;
308 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
309 mp->m_flags |= XFS_MOUNT_BARRIER;
310 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
311 mp->m_flags &= ~XFS_MOUNT_BARRIER;
312 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
313 mp->m_flags |= XFS_MOUNT_IKEEP;
314 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
315 mp->m_flags &= ~XFS_MOUNT_IKEEP;
316 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
317 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
318 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
319 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
320 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
321 mp->m_flags |= XFS_MOUNT_ATTR2;
322 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
323 mp->m_flags &= ~XFS_MOUNT_ATTR2;
324 mp->m_flags |= XFS_MOUNT_NOATTR2;
325 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
326 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
327 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
328 mp->m_qflags &= ~(XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
329 XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
330 XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
331 XFS_UQUOTA_ENFD | XFS_OQUOTA_ENFD);
332 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
333 !strcmp(this_char, MNTOPT_UQUOTA) ||
334 !strcmp(this_char, MNTOPT_USRQUOTA)) {
335 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
337 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
338 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
339 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
340 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
341 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
342 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
343 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
345 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
346 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
347 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
348 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
349 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
350 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
352 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
353 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
354 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
355 } else if (!strcmp(this_char, MNTOPT_DELAYLOG)) {
356 mp->m_flags |= XFS_MOUNT_DELAYLOG;
358 "Enabling EXPERIMENTAL delayed logging feature "
359 "- use at your own risk.\n");
360 } else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) {
361 mp->m_flags &= ~XFS_MOUNT_DELAYLOG;
362 } else if (!strcmp(this_char, "ihashsize")) {
364 "XFS: ihashsize no longer used, option is deprecated.");
365 } else if (!strcmp(this_char, "osyncisdsync")) {
367 "XFS: osyncisdsync has no effect, option is deprecated.");
368 } else if (!strcmp(this_char, "osyncisosync")) {
370 "XFS: osyncisosync has no effect, option is deprecated.");
371 } else if (!strcmp(this_char, "irixsgid")) {
373 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
376 "XFS: unknown mount option [%s].", this_char);
382 * no recovery flag requires a read-only mount
384 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
385 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
386 cmn_err(CE_WARN, "XFS: no-recovery mounts must be read-only.");
390 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
392 "XFS: sunit and swidth options incompatible with the noalign option");
396 #ifndef CONFIG_XFS_QUOTA
397 if (XFS_IS_QUOTA_RUNNING(mp)) {
399 "XFS: quota support not available in this kernel.");
404 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
405 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE))) {
407 "XFS: cannot mount with both project and group quota");
411 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
413 "XFS: sunit and swidth must be specified together");
417 if (dsunit && (dswidth % dsunit != 0)) {
419 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
425 if (!(mp->m_flags & XFS_MOUNT_NOALIGN)) {
427 * At this point the superblock has not been read
428 * in, therefore we do not know the block size.
429 * Before the mount call ends we will convert
433 mp->m_dalign = dsunit;
434 mp->m_flags |= XFS_MOUNT_RETERR;
438 mp->m_swidth = dswidth;
441 if (mp->m_logbufs != -1 &&
442 mp->m_logbufs != 0 &&
443 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
444 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
446 "XFS: invalid logbufs value: %d [not %d-%d]",
447 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
448 return XFS_ERROR(EINVAL);
450 if (mp->m_logbsize != -1 &&
451 mp->m_logbsize != 0 &&
452 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
453 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
454 !is_power_of_2(mp->m_logbsize))) {
456 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
458 return XFS_ERROR(EINVAL);
461 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
464 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
467 if (iosizelog > XFS_MAX_IO_LOG ||
468 iosizelog < XFS_MIN_IO_LOG) {
470 "XFS: invalid log iosize: %d [not %d-%d]",
471 iosizelog, XFS_MIN_IO_LOG,
473 return XFS_ERROR(EINVAL);
476 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
477 mp->m_readio_log = iosizelog;
478 mp->m_writeio_log = iosizelog;
484 struct proc_xfs_info {
491 struct xfs_mount *mp,
494 static struct proc_xfs_info xfs_info_set[] = {
495 /* the few simple ones we can get from the mount struct */
496 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
497 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
498 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
499 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
500 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
501 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
502 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
503 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
504 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
505 { XFS_MOUNT_DELAYLOG, "," MNTOPT_DELAYLOG },
508 static struct proc_xfs_info xfs_info_unset[] = {
509 /* the few simple ones we can get from the mount struct */
510 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
511 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
512 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
515 struct proc_xfs_info *xfs_infop;
517 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
518 if (mp->m_flags & xfs_infop->flag)
519 seq_puts(m, xfs_infop->str);
521 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
522 if (!(mp->m_flags & xfs_infop->flag))
523 seq_puts(m, xfs_infop->str);
526 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
527 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
528 (int)(1 << mp->m_writeio_log) >> 10);
530 if (mp->m_logbufs > 0)
531 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
532 if (mp->m_logbsize > 0)
533 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
536 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
538 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
540 if (mp->m_dalign > 0)
541 seq_printf(m, "," MNTOPT_SUNIT "=%d",
542 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
543 if (mp->m_swidth > 0)
544 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
545 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
547 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
548 seq_puts(m, "," MNTOPT_USRQUOTA);
549 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
550 seq_puts(m, "," MNTOPT_UQUOTANOENF);
552 /* Either project or group quotas can be active, not both */
554 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
555 if (mp->m_qflags & XFS_OQUOTA_ENFD)
556 seq_puts(m, "," MNTOPT_PRJQUOTA);
558 seq_puts(m, "," MNTOPT_PQUOTANOENF);
559 } else if (mp->m_qflags & XFS_GQUOTA_ACCT) {
560 if (mp->m_qflags & XFS_OQUOTA_ENFD)
561 seq_puts(m, "," MNTOPT_GRPQUOTA);
563 seq_puts(m, "," MNTOPT_GQUOTANOENF);
566 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
567 seq_puts(m, "," MNTOPT_NOQUOTA);
573 unsigned int blockshift)
575 unsigned int pagefactor = 1;
576 unsigned int bitshift = BITS_PER_LONG - 1;
578 /* Figure out maximum filesize, on Linux this can depend on
579 * the filesystem blocksize (on 32 bit platforms).
580 * __block_prepare_write does this in an [unsigned] long...
581 * page->index << (PAGE_CACHE_SHIFT - bbits)
582 * So, for page sized blocks (4K on 32 bit platforms),
583 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
584 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
585 * but for smaller blocksizes it is less (bbits = log2 bsize).
586 * Note1: get_block_t takes a long (implicit cast from above)
587 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
588 * can optionally convert the [unsigned] long from above into
589 * an [unsigned] long long.
592 #if BITS_PER_LONG == 32
593 # if defined(CONFIG_LBDAF)
594 ASSERT(sizeof(sector_t) == 8);
595 pagefactor = PAGE_CACHE_SIZE;
596 bitshift = BITS_PER_LONG;
598 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
602 return (((__uint64_t)pagefactor) << bitshift) - 1;
609 struct block_device **bdevp)
613 *bdevp = open_bdev_exclusive(name, FMODE_READ|FMODE_WRITE, mp);
614 if (IS_ERR(*bdevp)) {
615 error = PTR_ERR(*bdevp);
616 printk("XFS: Invalid device [%s], error=%d\n", name, error);
624 struct block_device *bdev)
627 close_bdev_exclusive(bdev, FMODE_READ|FMODE_WRITE);
631 * Try to write out the superblock using barriers.
637 xfs_buf_t *sbp = xfs_getsb(mp, 0);
642 XFS_BUF_UNDELAYWRITE(sbp);
644 XFS_BUF_UNASYNC(sbp);
645 XFS_BUF_ORDERED(sbp);
648 error = xfs_iowait(sbp);
651 * Clear all the flags we set and possible error state in the
652 * buffer. We only did the write to try out whether barriers
653 * worked and shouldn't leave any traces in the superblock
657 XFS_BUF_ERROR(sbp, 0);
658 XFS_BUF_UNORDERED(sbp);
665 xfs_mountfs_check_barriers(xfs_mount_t *mp)
669 if (mp->m_logdev_targp != mp->m_ddev_targp) {
670 xfs_fs_cmn_err(CE_NOTE, mp,
671 "Disabling barriers, not supported with external log device");
672 mp->m_flags &= ~XFS_MOUNT_BARRIER;
676 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
677 xfs_fs_cmn_err(CE_NOTE, mp,
678 "Disabling barriers, underlying device is readonly");
679 mp->m_flags &= ~XFS_MOUNT_BARRIER;
683 error = xfs_barrier_test(mp);
685 xfs_fs_cmn_err(CE_NOTE, mp,
686 "Disabling barriers, trial barrier write failed");
687 mp->m_flags &= ~XFS_MOUNT_BARRIER;
693 xfs_blkdev_issue_flush(
694 xfs_buftarg_t *buftarg)
696 blkdev_issue_flush(buftarg->bt_bdev, GFP_KERNEL, NULL,
702 struct xfs_mount *mp)
704 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
705 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
706 xfs_free_buftarg(mp, mp->m_logdev_targp);
707 xfs_blkdev_put(logdev);
709 if (mp->m_rtdev_targp) {
710 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
711 xfs_free_buftarg(mp, mp->m_rtdev_targp);
712 xfs_blkdev_put(rtdev);
714 xfs_free_buftarg(mp, mp->m_ddev_targp);
718 * The file system configurations are:
719 * (1) device (partition) with data and internal log
720 * (2) logical volume with data and log subvolumes.
721 * (3) logical volume with data, log, and realtime subvolumes.
723 * We only have to handle opening the log and realtime volumes here if
724 * they are present. The data subvolume has already been opened by
725 * get_sb_bdev() and is stored in sb->s_bdev.
729 struct xfs_mount *mp)
731 struct block_device *ddev = mp->m_super->s_bdev;
732 struct block_device *logdev = NULL, *rtdev = NULL;
736 * Open real time and log devices - order is important.
739 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
745 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
747 goto out_close_logdev;
749 if (rtdev == ddev || rtdev == logdev) {
751 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
753 goto out_close_rtdev;
758 * Setup xfs_mount buffer target pointers
761 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, 0, mp->m_fsname);
762 if (!mp->m_ddev_targp)
763 goto out_close_rtdev;
766 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, 1,
768 if (!mp->m_rtdev_targp)
769 goto out_free_ddev_targ;
772 if (logdev && logdev != ddev) {
773 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, 1,
775 if (!mp->m_logdev_targp)
776 goto out_free_rtdev_targ;
778 mp->m_logdev_targp = mp->m_ddev_targp;
784 if (mp->m_rtdev_targp)
785 xfs_free_buftarg(mp, mp->m_rtdev_targp);
787 xfs_free_buftarg(mp, mp->m_ddev_targp);
790 xfs_blkdev_put(rtdev);
792 if (logdev && logdev != ddev)
793 xfs_blkdev_put(logdev);
799 * Setup xfs_mount buffer target pointers based on superblock
803 struct xfs_mount *mp)
807 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
808 mp->m_sb.sb_sectsize);
812 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
813 unsigned int log_sector_size = BBSIZE;
815 if (xfs_sb_version_hassector(&mp->m_sb))
816 log_sector_size = mp->m_sb.sb_logsectsize;
817 error = xfs_setsize_buftarg(mp->m_logdev_targp,
818 mp->m_sb.sb_blocksize,
823 if (mp->m_rtdev_targp) {
824 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
825 mp->m_sb.sb_blocksize,
826 mp->m_sb.sb_sectsize);
835 * XFS AIL push thread support
839 struct xfs_ail *ailp,
840 xfs_lsn_t threshold_lsn)
842 ailp->xa_target = threshold_lsn;
843 wake_up_process(ailp->xa_task);
850 struct xfs_ail *ailp = data;
851 xfs_lsn_t last_pushed_lsn = 0;
852 long tout = 0; /* milliseconds */
854 while (!kthread_should_stop()) {
855 schedule_timeout_interruptible(tout ?
856 msecs_to_jiffies(tout) : MAX_SCHEDULE_TIMEOUT);
861 ASSERT(ailp->xa_mount->m_log);
862 if (XFS_FORCED_SHUTDOWN(ailp->xa_mount))
865 tout = xfsaild_push(ailp, &last_pushed_lsn);
873 struct xfs_ail *ailp)
876 ailp->xa_task = kthread_run(xfsaild, ailp, "xfsaild/%s",
877 ailp->xa_mount->m_fsname);
878 if (IS_ERR(ailp->xa_task))
879 return -PTR_ERR(ailp->xa_task);
885 struct xfs_ail *ailp)
887 kthread_stop(ailp->xa_task);
891 /* Catch misguided souls that try to use this interface on XFS */
892 STATIC struct inode *
894 struct super_block *sb)
901 * Now that the generic code is guaranteed not to be accessing
902 * the linux inode, we can reclaim the inode.
905 xfs_fs_destroy_inode(
908 struct xfs_inode *ip = XFS_I(inode);
910 trace_xfs_destroy_inode(ip);
912 XFS_STATS_INC(vn_reclaim);
914 /* bad inode, get out here ASAP */
915 if (is_bad_inode(inode))
920 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
923 * We should never get here with one of the reclaim flags already set.
925 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
926 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
929 * We always use background reclaim here because even if the
930 * inode is clean, it still may be under IO and hence we have
931 * to take the flush lock. The background reclaim path handles
932 * this more efficiently than we can here, so simply let background
933 * reclaim tear down all inodes.
936 xfs_inode_set_reclaim_tag(ip);
940 * Slab object creation initialisation for the XFS inode.
941 * This covers only the idempotent fields in the XFS inode;
942 * all other fields need to be initialised on allocation
943 * from the slab. This avoids the need to repeatedly intialise
944 * fields in the xfs inode that left in the initialise state
945 * when freeing the inode.
948 xfs_fs_inode_init_once(
951 struct xfs_inode *ip = inode;
953 memset(ip, 0, sizeof(struct xfs_inode));
956 inode_init_once(VFS_I(ip));
959 atomic_set(&ip->i_iocount, 0);
960 atomic_set(&ip->i_pincount, 0);
961 spin_lock_init(&ip->i_flags_lock);
962 init_waitqueue_head(&ip->i_ipin_wait);
964 * Because we want to use a counting completion, complete
965 * the flush completion once to allow a single access to
966 * the flush completion without blocking.
968 init_completion(&ip->i_flush);
969 complete(&ip->i_flush);
971 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
972 "xfsino", ip->i_ino);
976 * Dirty the XFS inode when mark_inode_dirty_sync() is called so that
977 * we catch unlogged VFS level updates to the inode.
979 * We need the barrier() to maintain correct ordering between unlogged
980 * updates and the transaction commit code that clears the i_update_core
981 * field. This requires all updates to be completed before marking the
989 XFS_I(inode)->i_update_core = 1;
994 struct xfs_inode *ip)
996 struct xfs_mount *mp = ip->i_mount;
997 struct xfs_trans *tp;
1000 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1001 tp = xfs_trans_alloc(mp, XFS_TRANS_FSYNC_TS);
1002 error = xfs_trans_reserve(tp, 0, XFS_FSYNC_TS_LOG_RES(mp), 0, 0, 0);
1005 xfs_trans_cancel(tp, 0);
1006 /* we need to return with the lock hold shared */
1007 xfs_ilock(ip, XFS_ILOCK_SHARED);
1011 xfs_ilock(ip, XFS_ILOCK_EXCL);
1014 * Note - it's possible that we might have pushed ourselves out of the
1015 * way during trans_reserve which would flush the inode. But there's
1016 * no guarantee that the inode buffer has actually gone out yet (it's
1017 * delwri). Plus the buffer could be pinned anyway if it's part of
1018 * an inode in another recent transaction. So we play it safe and
1019 * fire off the transaction anyway.
1021 xfs_trans_ijoin(tp, ip);
1022 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1023 error = xfs_trans_commit(tp, 0);
1024 xfs_ilock_demote(ip, XFS_ILOCK_EXCL);
1031 struct inode *inode,
1032 struct writeback_control *wbc)
1034 struct xfs_inode *ip = XFS_I(inode);
1035 struct xfs_mount *mp = ip->i_mount;
1038 trace_xfs_write_inode(ip);
1040 if (XFS_FORCED_SHUTDOWN(mp))
1041 return XFS_ERROR(EIO);
1043 if (wbc->sync_mode == WB_SYNC_ALL) {
1045 * Make sure the inode has made it it into the log. Instead
1046 * of forcing it all the way to stable storage using a
1047 * synchronous transaction we let the log force inside the
1048 * ->sync_fs call do that for thus, which reduces the number
1049 * of synchronous log foces dramatically.
1052 xfs_ilock(ip, XFS_ILOCK_SHARED);
1053 if (ip->i_update_core) {
1054 error = xfs_log_inode(ip);
1060 * We make this non-blocking if the inode is contended, return
1061 * EAGAIN to indicate to the caller that they did not succeed.
1062 * This prevents the flush path from blocking on inodes inside
1063 * another operation right now, they get caught later by
1066 if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED))
1069 if (xfs_ipincount(ip) || !xfs_iflock_nowait(ip))
1073 * Now we have the flush lock and the inode is not pinned, we
1074 * can check if the inode is really clean as we know that
1075 * there are no pending transaction completions, it is not
1076 * waiting on the delayed write queue and there is no IO in
1079 if (xfs_inode_clean(ip)) {
1084 error = xfs_iflush(ip, 0);
1088 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1091 * if we failed to write out the inode then mark
1092 * it dirty again so we'll try again later.
1095 xfs_mark_inode_dirty_sync(ip);
1101 struct inode *inode)
1103 xfs_inode_t *ip = XFS_I(inode);
1105 trace_xfs_evict_inode(ip);
1107 truncate_inode_pages(&inode->i_data, 0);
1108 end_writeback(inode);
1109 XFS_STATS_INC(vn_rele);
1110 XFS_STATS_INC(vn_remove);
1111 XFS_STATS_DEC(vn_active);
1114 * The iolock is used by the file system to coordinate reads,
1115 * writes, and block truncates. Up to this point the lock
1116 * protected concurrent accesses by users of the inode. But
1117 * from here forward we're doing some final processing of the
1118 * inode because we're done with it, and although we reuse the
1119 * iolock for protection it is really a distinct lock class
1120 * (in the lockdep sense) from before. To keep lockdep happy
1121 * (and basically indicate what we are doing), we explicitly
1122 * re-init the iolock here.
1124 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
1125 mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
1132 struct xfs_mount *mp)
1134 kfree(mp->m_fsname);
1135 kfree(mp->m_rtname);
1136 kfree(mp->m_logname);
1141 struct super_block *sb)
1143 struct xfs_mount *mp = XFS_M(sb);
1146 * Unregister the memory shrinker before we tear down the mount
1147 * structure so we don't have memory reclaim racing with us here.
1149 xfs_inode_shrinker_unregister(mp);
1153 * Blow away any referenced inode in the filestreams cache.
1154 * This can and will cause log traffic as inodes go inactive
1157 xfs_filestream_unmount(mp);
1159 XFS_bflush(mp->m_ddev_targp);
1163 xfs_icsb_destroy_counters(mp);
1164 xfs_close_devices(mp);
1165 xfs_free_fsname(mp);
1171 struct super_block *sb,
1174 struct xfs_mount *mp = XFS_M(sb);
1178 * Not much we can do for the first async pass. Writing out the
1179 * superblock would be counter-productive as we are going to redirty
1180 * when writing out other data and metadata (and writing out a single
1181 * block is quite fast anyway).
1183 * Try to asynchronously kick off quota syncing at least.
1186 xfs_qm_sync(mp, SYNC_TRYLOCK);
1190 error = xfs_quiesce_data(mp);
1195 int prev_sync_seq = mp->m_sync_seq;
1198 * The disk must be active because we're syncing.
1199 * We schedule xfssyncd now (now that the disk is
1200 * active) instead of later (when it might not be).
1202 wake_up_process(mp->m_sync_task);
1204 * We have to wait for the sync iteration to complete.
1205 * If we don't, the disk activity caused by the sync
1206 * will come after the sync is completed, and that
1207 * triggers another sync from laptop mode.
1209 wait_event(mp->m_wait_single_sync_task,
1210 mp->m_sync_seq != prev_sync_seq);
1218 struct dentry *dentry,
1219 struct kstatfs *statp)
1221 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1222 xfs_sb_t *sbp = &mp->m_sb;
1223 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1224 __uint64_t fakeinos, id;
1228 statp->f_type = XFS_SB_MAGIC;
1229 statp->f_namelen = MAXNAMELEN - 1;
1231 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1232 statp->f_fsid.val[0] = (u32)id;
1233 statp->f_fsid.val[1] = (u32)(id >> 32);
1235 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1237 spin_lock(&mp->m_sb_lock);
1238 statp->f_bsize = sbp->sb_blocksize;
1239 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1240 statp->f_blocks = sbp->sb_dblocks - lsize;
1241 statp->f_bfree = statp->f_bavail =
1242 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1243 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1245 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1246 if (mp->m_maxicount)
1247 statp->f_files = min_t(typeof(statp->f_files),
1251 /* make sure statp->f_ffree does not underflow */
1252 ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1253 statp->f_ffree = max_t(__int64_t, ffree, 0);
1255 spin_unlock(&mp->m_sb_lock);
1257 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) ||
1258 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))) ==
1259 (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
1260 xfs_qm_statvfs(ip, statp);
1265 xfs_save_resvblks(struct xfs_mount *mp)
1267 __uint64_t resblks = 0;
1269 mp->m_resblks_save = mp->m_resblks;
1270 xfs_reserve_blocks(mp, &resblks, NULL);
1274 xfs_restore_resvblks(struct xfs_mount *mp)
1278 if (mp->m_resblks_save) {
1279 resblks = mp->m_resblks_save;
1280 mp->m_resblks_save = 0;
1282 resblks = xfs_default_resblks(mp);
1284 xfs_reserve_blocks(mp, &resblks, NULL);
1289 struct super_block *sb,
1293 struct xfs_mount *mp = XFS_M(sb);
1294 substring_t args[MAX_OPT_ARGS];
1298 while ((p = strsep(&options, ",")) != NULL) {
1304 token = match_token(p, tokens, args);
1307 mp->m_flags |= XFS_MOUNT_BARRIER;
1310 * Test if barriers are actually working if we can,
1311 * else delay this check until the filesystem is
1314 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1315 xfs_mountfs_check_barriers(mp);
1318 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1322 * Logically we would return an error here to prevent
1323 * users from believing they might have changed
1324 * mount options using remount which can't be changed.
1326 * But unfortunately mount(8) adds all options from
1327 * mtab and fstab to the mount arguments in some cases
1328 * so we can't blindly reject options, but have to
1329 * check for each specified option if it actually
1330 * differs from the currently set option and only
1331 * reject it if that's the case.
1333 * Until that is implemented we return success for
1334 * every remount request, and silently ignore all
1335 * options that we can't actually change.
1339 "XFS: mount option \"%s\" not supported for remount\n", p);
1348 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1349 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1350 if (mp->m_flags & XFS_MOUNT_BARRIER)
1351 xfs_mountfs_check_barriers(mp);
1354 * If this is the first remount to writeable state we
1355 * might have some superblock changes to update.
1357 if (mp->m_update_flags) {
1358 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1361 "XFS: failed to write sb changes");
1364 mp->m_update_flags = 0;
1368 * Fill out the reserve pool if it is empty. Use the stashed
1369 * value if it is non-zero, otherwise go with the default.
1371 xfs_restore_resvblks(mp);
1375 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1377 * After we have synced the data but before we sync the
1378 * metadata, we need to free up the reserve block pool so that
1379 * the used block count in the superblock on disk is correct at
1380 * the end of the remount. Stash the current reserve pool size
1381 * so that if we get remounted rw, we can return it to the same
1385 xfs_quiesce_data(mp);
1386 xfs_save_resvblks(mp);
1387 xfs_quiesce_attr(mp);
1388 mp->m_flags |= XFS_MOUNT_RDONLY;
1395 * Second stage of a freeze. The data is already frozen so we only
1396 * need to take care of the metadata. Once that's done write a dummy
1397 * record to dirty the log in case of a crash while frozen.
1401 struct super_block *sb)
1403 struct xfs_mount *mp = XFS_M(sb);
1405 xfs_save_resvblks(mp);
1406 xfs_quiesce_attr(mp);
1407 return -xfs_fs_log_dummy(mp, SYNC_WAIT);
1412 struct super_block *sb)
1414 struct xfs_mount *mp = XFS_M(sb);
1416 xfs_restore_resvblks(mp);
1421 xfs_fs_show_options(
1423 struct vfsmount *mnt)
1425 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1429 * This function fills in xfs_mount_t fields based on mount args.
1430 * Note: the superblock _has_ now been read in.
1434 struct xfs_mount *mp)
1436 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1438 /* Fail a mount where the logbuf is smaller than the log stripe */
1439 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1440 if (mp->m_logbsize <= 0 &&
1441 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1442 mp->m_logbsize = mp->m_sb.sb_logsunit;
1443 } else if (mp->m_logbsize > 0 &&
1444 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1446 "XFS: logbuf size must be greater than or equal to log stripe size");
1447 return XFS_ERROR(EINVAL);
1450 /* Fail a mount if the logbuf is larger than 32K */
1451 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1453 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1454 return XFS_ERROR(EINVAL);
1459 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1460 * told by noattr2 to turn it off
1462 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1463 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1464 mp->m_flags |= XFS_MOUNT_ATTR2;
1467 * prohibit r/w mounts of read-only filesystems
1469 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1471 "XFS: cannot mount a read-only filesystem as read-write");
1472 return XFS_ERROR(EROFS);
1480 struct super_block *sb,
1485 struct xfs_mount *mp = NULL;
1486 int flags = 0, error = ENOMEM;
1488 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1492 spin_lock_init(&mp->m_sb_lock);
1493 mutex_init(&mp->m_growlock);
1494 atomic_set(&mp->m_active_trans, 0);
1495 INIT_LIST_HEAD(&mp->m_sync_list);
1496 spin_lock_init(&mp->m_sync_lock);
1497 init_waitqueue_head(&mp->m_wait_single_sync_task);
1502 error = xfs_parseargs(mp, (char *)data);
1504 goto out_free_fsname;
1506 sb_min_blocksize(sb, BBSIZE);
1507 sb->s_xattr = xfs_xattr_handlers;
1508 sb->s_export_op = &xfs_export_operations;
1509 #ifdef CONFIG_XFS_QUOTA
1510 sb->s_qcop = &xfs_quotactl_operations;
1512 sb->s_op = &xfs_super_operations;
1515 flags |= XFS_MFSI_QUIET;
1517 error = xfs_open_devices(mp);
1519 goto out_free_fsname;
1521 if (xfs_icsb_init_counters(mp))
1522 mp->m_flags |= XFS_MOUNT_NO_PERCPU_SB;
1524 error = xfs_readsb(mp, flags);
1526 goto out_destroy_counters;
1528 error = xfs_finish_flags(mp);
1532 error = xfs_setup_devices(mp);
1536 if (mp->m_flags & XFS_MOUNT_BARRIER)
1537 xfs_mountfs_check_barriers(mp);
1539 error = xfs_filestream_mount(mp);
1543 error = xfs_mountfs(mp);
1545 goto out_filestream_unmount;
1547 sb->s_magic = XFS_SB_MAGIC;
1548 sb->s_blocksize = mp->m_sb.sb_blocksize;
1549 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1550 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1551 sb->s_time_gran = 1;
1552 set_posix_acl_flag(sb);
1554 root = igrab(VFS_I(mp->m_rootip));
1559 if (is_bad_inode(root)) {
1563 sb->s_root = d_alloc_root(root);
1569 error = xfs_syncd_init(mp);
1573 xfs_inode_shrinker_register(mp);
1577 out_filestream_unmount:
1578 xfs_filestream_unmount(mp);
1581 out_destroy_counters:
1582 xfs_icsb_destroy_counters(mp);
1583 xfs_close_devices(mp);
1585 xfs_free_fsname(mp);
1600 * Blow away any referenced inode in the filestreams cache.
1601 * This can and will cause log traffic as inodes go inactive
1604 xfs_filestream_unmount(mp);
1606 XFS_bflush(mp->m_ddev_targp);
1614 struct file_system_type *fs_type,
1616 const char *dev_name,
1618 struct vfsmount *mnt)
1620 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1624 static const struct super_operations xfs_super_operations = {
1625 .alloc_inode = xfs_fs_alloc_inode,
1626 .destroy_inode = xfs_fs_destroy_inode,
1627 .dirty_inode = xfs_fs_dirty_inode,
1628 .write_inode = xfs_fs_write_inode,
1629 .evict_inode = xfs_fs_evict_inode,
1630 .put_super = xfs_fs_put_super,
1631 .sync_fs = xfs_fs_sync_fs,
1632 .freeze_fs = xfs_fs_freeze,
1633 .unfreeze_fs = xfs_fs_unfreeze,
1634 .statfs = xfs_fs_statfs,
1635 .remount_fs = xfs_fs_remount,
1636 .show_options = xfs_fs_show_options,
1639 static struct file_system_type xfs_fs_type = {
1640 .owner = THIS_MODULE,
1642 .get_sb = xfs_fs_get_sb,
1643 .kill_sb = kill_block_super,
1644 .fs_flags = FS_REQUIRES_DEV,
1648 xfs_init_zones(void)
1651 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1652 if (!xfs_ioend_zone)
1655 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1657 if (!xfs_ioend_pool)
1658 goto out_destroy_ioend_zone;
1660 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1662 if (!xfs_log_ticket_zone)
1663 goto out_destroy_ioend_pool;
1665 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1666 "xfs_bmap_free_item");
1667 if (!xfs_bmap_free_item_zone)
1668 goto out_destroy_log_ticket_zone;
1670 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1672 if (!xfs_btree_cur_zone)
1673 goto out_destroy_bmap_free_item_zone;
1675 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1677 if (!xfs_da_state_zone)
1678 goto out_destroy_btree_cur_zone;
1680 xfs_dabuf_zone = kmem_zone_init(sizeof(xfs_dabuf_t), "xfs_dabuf");
1681 if (!xfs_dabuf_zone)
1682 goto out_destroy_da_state_zone;
1684 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1685 if (!xfs_ifork_zone)
1686 goto out_destroy_dabuf_zone;
1688 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1689 if (!xfs_trans_zone)
1690 goto out_destroy_ifork_zone;
1692 xfs_log_item_desc_zone =
1693 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1694 "xfs_log_item_desc");
1695 if (!xfs_log_item_desc_zone)
1696 goto out_destroy_trans_zone;
1699 * The size of the zone allocated buf log item is the maximum
1700 * size possible under XFS. This wastes a little bit of memory,
1701 * but it is much faster.
1703 xfs_buf_item_zone = kmem_zone_init((sizeof(xfs_buf_log_item_t) +
1704 (((XFS_MAX_BLOCKSIZE / XFS_BLF_CHUNK) /
1705 NBWORD) * sizeof(int))), "xfs_buf_item");
1706 if (!xfs_buf_item_zone)
1707 goto out_destroy_log_item_desc_zone;
1709 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1710 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1711 sizeof(xfs_extent_t))), "xfs_efd_item");
1713 goto out_destroy_buf_item_zone;
1715 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1716 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1717 sizeof(xfs_extent_t))), "xfs_efi_item");
1719 goto out_destroy_efd_zone;
1722 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1723 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1724 xfs_fs_inode_init_once);
1725 if (!xfs_inode_zone)
1726 goto out_destroy_efi_zone;
1729 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1730 KM_ZONE_SPREAD, NULL);
1732 goto out_destroy_inode_zone;
1736 out_destroy_inode_zone:
1737 kmem_zone_destroy(xfs_inode_zone);
1738 out_destroy_efi_zone:
1739 kmem_zone_destroy(xfs_efi_zone);
1740 out_destroy_efd_zone:
1741 kmem_zone_destroy(xfs_efd_zone);
1742 out_destroy_buf_item_zone:
1743 kmem_zone_destroy(xfs_buf_item_zone);
1744 out_destroy_log_item_desc_zone:
1745 kmem_zone_destroy(xfs_log_item_desc_zone);
1746 out_destroy_trans_zone:
1747 kmem_zone_destroy(xfs_trans_zone);
1748 out_destroy_ifork_zone:
1749 kmem_zone_destroy(xfs_ifork_zone);
1750 out_destroy_dabuf_zone:
1751 kmem_zone_destroy(xfs_dabuf_zone);
1752 out_destroy_da_state_zone:
1753 kmem_zone_destroy(xfs_da_state_zone);
1754 out_destroy_btree_cur_zone:
1755 kmem_zone_destroy(xfs_btree_cur_zone);
1756 out_destroy_bmap_free_item_zone:
1757 kmem_zone_destroy(xfs_bmap_free_item_zone);
1758 out_destroy_log_ticket_zone:
1759 kmem_zone_destroy(xfs_log_ticket_zone);
1760 out_destroy_ioend_pool:
1761 mempool_destroy(xfs_ioend_pool);
1762 out_destroy_ioend_zone:
1763 kmem_zone_destroy(xfs_ioend_zone);
1769 xfs_destroy_zones(void)
1771 kmem_zone_destroy(xfs_ili_zone);
1772 kmem_zone_destroy(xfs_inode_zone);
1773 kmem_zone_destroy(xfs_efi_zone);
1774 kmem_zone_destroy(xfs_efd_zone);
1775 kmem_zone_destroy(xfs_buf_item_zone);
1776 kmem_zone_destroy(xfs_log_item_desc_zone);
1777 kmem_zone_destroy(xfs_trans_zone);
1778 kmem_zone_destroy(xfs_ifork_zone);
1779 kmem_zone_destroy(xfs_dabuf_zone);
1780 kmem_zone_destroy(xfs_da_state_zone);
1781 kmem_zone_destroy(xfs_btree_cur_zone);
1782 kmem_zone_destroy(xfs_bmap_free_item_zone);
1783 kmem_zone_destroy(xfs_log_ticket_zone);
1784 mempool_destroy(xfs_ioend_pool);
1785 kmem_zone_destroy(xfs_ioend_zone);
1794 printk(KERN_INFO XFS_VERSION_STRING " with "
1795 XFS_BUILD_OPTIONS " enabled\n");
1800 error = xfs_init_zones();
1804 error = xfs_mru_cache_init();
1806 goto out_destroy_zones;
1808 error = xfs_filestream_init();
1810 goto out_mru_cache_uninit;
1812 error = xfs_buf_init();
1814 goto out_filestream_uninit;
1816 error = xfs_init_procfs();
1818 goto out_buf_terminate;
1820 error = xfs_sysctl_register();
1822 goto out_cleanup_procfs;
1826 error = register_filesystem(&xfs_fs_type);
1828 goto out_sysctl_unregister;
1831 out_sysctl_unregister:
1832 xfs_sysctl_unregister();
1834 xfs_cleanup_procfs();
1836 xfs_buf_terminate();
1837 out_filestream_uninit:
1838 xfs_filestream_uninit();
1839 out_mru_cache_uninit:
1840 xfs_mru_cache_uninit();
1842 xfs_destroy_zones();
1851 unregister_filesystem(&xfs_fs_type);
1852 xfs_sysctl_unregister();
1853 xfs_cleanup_procfs();
1854 xfs_buf_terminate();
1855 xfs_filestream_uninit();
1856 xfs_mru_cache_uninit();
1857 xfs_destroy_zones();
1860 module_init(init_xfs_fs);
1861 module_exit(exit_xfs_fs);
1863 MODULE_AUTHOR("Silicon Graphics, Inc.");
1864 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1865 MODULE_LICENSE("GPL");