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_log_priv.h"
48 #include "xfs_trans_priv.h"
49 #include "xfs_filestream.h"
50 #include "xfs_da_btree.h"
51 #include "xfs_extfree_item.h"
52 #include "xfs_mru_cache.h"
53 #include "xfs_inode_item.h"
55 #include "xfs_trace.h"
57 #include <linux/namei.h>
58 #include <linux/init.h>
59 #include <linux/slab.h>
60 #include <linux/mount.h>
61 #include <linux/mempool.h>
62 #include <linux/writeback.h>
63 #include <linux/kthread.h>
64 #include <linux/freezer.h>
65 #include <linux/parser.h>
67 static const struct super_operations xfs_super_operations;
68 static kmem_zone_t *xfs_ioend_zone;
69 mempool_t *xfs_ioend_pool;
71 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
72 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
73 #define MNTOPT_LOGDEV "logdev" /* log device */
74 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
75 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
76 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
77 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
78 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
79 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
80 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
81 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
82 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
83 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
84 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
85 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
86 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
87 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
88 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
89 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
90 * unwritten extent conversion */
91 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
92 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
93 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
94 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
95 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
96 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
98 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
99 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
100 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
101 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
102 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
103 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
104 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
105 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
106 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
107 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
108 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
109 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
110 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
111 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
112 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
113 #define MNTOPT_DELAYLOG "delaylog" /* Delayed loging enabled */
114 #define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed loging disabled */
117 * Table driven mount option parser.
119 * Currently only used for remount, but it will be used for mount
120 * in the future, too.
123 Opt_barrier, Opt_nobarrier, Opt_err
126 static const match_table_t tokens = {
127 {Opt_barrier, "barrier"},
128 {Opt_nobarrier, "nobarrier"},
134 suffix_strtoul(char *s, char **endp, unsigned int base)
136 int last, shift_left_factor = 0;
139 last = strlen(value) - 1;
140 if (value[last] == 'K' || value[last] == 'k') {
141 shift_left_factor = 10;
144 if (value[last] == 'M' || value[last] == 'm') {
145 shift_left_factor = 20;
148 if (value[last] == 'G' || value[last] == 'g') {
149 shift_left_factor = 30;
153 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
157 * This function fills in xfs_mount_t fields based on mount args.
158 * Note: the superblock has _not_ yet been read in.
160 * Note that this function leaks the various device name allocations on
161 * failure. The caller takes care of them.
165 struct xfs_mount *mp,
168 struct super_block *sb = mp->m_super;
169 char *this_char, *value, *eov;
173 __uint8_t iosizelog = 0;
176 * Copy binary VFS mount flags we are interested in.
178 if (sb->s_flags & MS_RDONLY)
179 mp->m_flags |= XFS_MOUNT_RDONLY;
180 if (sb->s_flags & MS_DIRSYNC)
181 mp->m_flags |= XFS_MOUNT_DIRSYNC;
182 if (sb->s_flags & MS_SYNCHRONOUS)
183 mp->m_flags |= XFS_MOUNT_WSYNC;
186 * Set some default flags that could be cleared by the mount option
189 mp->m_flags |= XFS_MOUNT_BARRIER;
190 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
191 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
194 * These can be overridden by the mount option parsing.
202 while ((this_char = strsep(&options, ",")) != NULL) {
205 if ((value = strchr(this_char, '=')) != NULL)
208 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
209 if (!value || !*value) {
211 "XFS: %s option requires an argument",
215 mp->m_logbufs = simple_strtoul(value, &eov, 10);
216 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
217 if (!value || !*value) {
219 "XFS: %s option requires an argument",
223 mp->m_logbsize = suffix_strtoul(value, &eov, 10);
224 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
225 if (!value || !*value) {
227 "XFS: %s option requires an argument",
231 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
234 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
236 "XFS: %s option not allowed on this system",
239 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
240 if (!value || !*value) {
242 "XFS: %s option requires an argument",
246 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
249 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
250 if (!value || !*value) {
252 "XFS: %s option requires an argument",
256 iosize = simple_strtoul(value, &eov, 10);
257 iosizelog = ffs(iosize) - 1;
258 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
259 if (!value || !*value) {
261 "XFS: %s option requires an argument",
265 iosize = suffix_strtoul(value, &eov, 10);
266 iosizelog = ffs(iosize) - 1;
267 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
268 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
269 mp->m_flags |= XFS_MOUNT_GRPID;
270 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
271 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
272 mp->m_flags &= ~XFS_MOUNT_GRPID;
273 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
274 mp->m_flags |= XFS_MOUNT_WSYNC;
275 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
276 mp->m_flags |= XFS_MOUNT_NORECOVERY;
277 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
278 mp->m_flags |= XFS_MOUNT_NOALIGN;
279 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
280 mp->m_flags |= XFS_MOUNT_SWALLOC;
281 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
282 if (!value || !*value) {
284 "XFS: %s option requires an argument",
288 dsunit = simple_strtoul(value, &eov, 10);
289 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
290 if (!value || !*value) {
292 "XFS: %s option requires an argument",
296 dswidth = simple_strtoul(value, &eov, 10);
297 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
298 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
301 "XFS: %s option not allowed on this system",
305 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
306 mp->m_flags |= XFS_MOUNT_NOUUID;
307 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
308 mp->m_flags |= XFS_MOUNT_BARRIER;
309 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
310 mp->m_flags &= ~XFS_MOUNT_BARRIER;
311 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
312 mp->m_flags |= XFS_MOUNT_IKEEP;
313 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
314 mp->m_flags &= ~XFS_MOUNT_IKEEP;
315 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
316 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
317 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
318 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
319 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
320 mp->m_flags |= XFS_MOUNT_ATTR2;
321 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
322 mp->m_flags &= ~XFS_MOUNT_ATTR2;
323 mp->m_flags |= XFS_MOUNT_NOATTR2;
324 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
325 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
326 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
327 mp->m_qflags &= ~(XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
328 XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
329 XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
330 XFS_UQUOTA_ENFD | XFS_OQUOTA_ENFD);
331 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
332 !strcmp(this_char, MNTOPT_UQUOTA) ||
333 !strcmp(this_char, MNTOPT_USRQUOTA)) {
334 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
336 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
337 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
338 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
339 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
340 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
341 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
342 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
344 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
345 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
346 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
347 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
348 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
349 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
351 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
352 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
353 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
354 } else if (!strcmp(this_char, MNTOPT_DELAYLOG)) {
355 mp->m_flags |= XFS_MOUNT_DELAYLOG;
356 } else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) {
357 mp->m_flags &= ~XFS_MOUNT_DELAYLOG;
358 } else if (!strcmp(this_char, "ihashsize")) {
360 "XFS: ihashsize no longer used, option is deprecated.");
361 } else if (!strcmp(this_char, "osyncisdsync")) {
363 "XFS: osyncisdsync has no effect, option is deprecated.");
364 } else if (!strcmp(this_char, "osyncisosync")) {
366 "XFS: osyncisosync has no effect, option is deprecated.");
367 } else if (!strcmp(this_char, "irixsgid")) {
369 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
372 "XFS: unknown mount option [%s].", this_char);
378 * no recovery flag requires a read-only mount
380 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
381 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
382 cmn_err(CE_WARN, "XFS: no-recovery mounts must be read-only.");
386 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
388 "XFS: sunit and swidth options incompatible with the noalign option");
392 #ifndef CONFIG_XFS_QUOTA
393 if (XFS_IS_QUOTA_RUNNING(mp)) {
395 "XFS: quota support not available in this kernel.");
400 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
401 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE))) {
403 "XFS: cannot mount with both project and group quota");
407 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
409 "XFS: sunit and swidth must be specified together");
413 if (dsunit && (dswidth % dsunit != 0)) {
415 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
421 if (!(mp->m_flags & XFS_MOUNT_NOALIGN)) {
423 * At this point the superblock has not been read
424 * in, therefore we do not know the block size.
425 * Before the mount call ends we will convert
429 mp->m_dalign = dsunit;
430 mp->m_flags |= XFS_MOUNT_RETERR;
434 mp->m_swidth = dswidth;
437 if (mp->m_logbufs != -1 &&
438 mp->m_logbufs != 0 &&
439 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
440 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
442 "XFS: invalid logbufs value: %d [not %d-%d]",
443 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
444 return XFS_ERROR(EINVAL);
446 if (mp->m_logbsize != -1 &&
447 mp->m_logbsize != 0 &&
448 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
449 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
450 !is_power_of_2(mp->m_logbsize))) {
452 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
454 return XFS_ERROR(EINVAL);
457 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
460 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
463 if (iosizelog > XFS_MAX_IO_LOG ||
464 iosizelog < XFS_MIN_IO_LOG) {
466 "XFS: invalid log iosize: %d [not %d-%d]",
467 iosizelog, XFS_MIN_IO_LOG,
469 return XFS_ERROR(EINVAL);
472 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
473 mp->m_readio_log = iosizelog;
474 mp->m_writeio_log = iosizelog;
480 struct proc_xfs_info {
487 struct xfs_mount *mp,
490 static struct proc_xfs_info xfs_info_set[] = {
491 /* the few simple ones we can get from the mount struct */
492 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
493 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
494 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
495 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
496 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
497 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
498 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
499 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
500 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
501 { XFS_MOUNT_DELAYLOG, "," MNTOPT_DELAYLOG },
504 static struct proc_xfs_info xfs_info_unset[] = {
505 /* the few simple ones we can get from the mount struct */
506 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
507 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
508 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
511 struct proc_xfs_info *xfs_infop;
513 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
514 if (mp->m_flags & xfs_infop->flag)
515 seq_puts(m, xfs_infop->str);
517 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
518 if (!(mp->m_flags & xfs_infop->flag))
519 seq_puts(m, xfs_infop->str);
522 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
523 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
524 (int)(1 << mp->m_writeio_log) >> 10);
526 if (mp->m_logbufs > 0)
527 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
528 if (mp->m_logbsize > 0)
529 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
532 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
534 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
536 if (mp->m_dalign > 0)
537 seq_printf(m, "," MNTOPT_SUNIT "=%d",
538 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
539 if (mp->m_swidth > 0)
540 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
541 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
543 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
544 seq_puts(m, "," MNTOPT_USRQUOTA);
545 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
546 seq_puts(m, "," MNTOPT_UQUOTANOENF);
548 /* Either project or group quotas can be active, not both */
550 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
551 if (mp->m_qflags & XFS_OQUOTA_ENFD)
552 seq_puts(m, "," MNTOPT_PRJQUOTA);
554 seq_puts(m, "," MNTOPT_PQUOTANOENF);
555 } else if (mp->m_qflags & XFS_GQUOTA_ACCT) {
556 if (mp->m_qflags & XFS_OQUOTA_ENFD)
557 seq_puts(m, "," MNTOPT_GRPQUOTA);
559 seq_puts(m, "," MNTOPT_GQUOTANOENF);
562 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
563 seq_puts(m, "," MNTOPT_NOQUOTA);
569 unsigned int blockshift)
571 unsigned int pagefactor = 1;
572 unsigned int bitshift = BITS_PER_LONG - 1;
574 /* Figure out maximum filesize, on Linux this can depend on
575 * the filesystem blocksize (on 32 bit platforms).
576 * __block_write_begin does this in an [unsigned] long...
577 * page->index << (PAGE_CACHE_SHIFT - bbits)
578 * So, for page sized blocks (4K on 32 bit platforms),
579 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
580 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
581 * but for smaller blocksizes it is less (bbits = log2 bsize).
582 * Note1: get_block_t takes a long (implicit cast from above)
583 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
584 * can optionally convert the [unsigned] long from above into
585 * an [unsigned] long long.
588 #if BITS_PER_LONG == 32
589 # if defined(CONFIG_LBDAF)
590 ASSERT(sizeof(sector_t) == 8);
591 pagefactor = PAGE_CACHE_SIZE;
592 bitshift = BITS_PER_LONG;
594 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
598 return (((__uint64_t)pagefactor) << bitshift) - 1;
605 struct block_device **bdevp)
609 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
611 if (IS_ERR(*bdevp)) {
612 error = PTR_ERR(*bdevp);
613 printk("XFS: Invalid device [%s], error=%d\n", name, error);
621 struct block_device *bdev)
624 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
628 * Try to write out the superblock using barriers.
634 xfs_buf_t *sbp = xfs_getsb(mp, 0);
639 XFS_BUF_UNDELAYWRITE(sbp);
641 XFS_BUF_UNASYNC(sbp);
642 XFS_BUF_ORDERED(sbp);
645 error = xfs_buf_iowait(sbp);
648 * Clear all the flags we set and possible error state in the
649 * buffer. We only did the write to try out whether barriers
650 * worked and shouldn't leave any traces in the superblock
654 XFS_BUF_ERROR(sbp, 0);
655 XFS_BUF_UNORDERED(sbp);
662 xfs_mountfs_check_barriers(xfs_mount_t *mp)
666 if (mp->m_logdev_targp != mp->m_ddev_targp) {
667 xfs_fs_cmn_err(CE_NOTE, mp,
668 "Disabling barriers, not supported with external log device");
669 mp->m_flags &= ~XFS_MOUNT_BARRIER;
673 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
674 xfs_fs_cmn_err(CE_NOTE, mp,
675 "Disabling barriers, underlying device is readonly");
676 mp->m_flags &= ~XFS_MOUNT_BARRIER;
680 error = xfs_barrier_test(mp);
682 xfs_fs_cmn_err(CE_NOTE, mp,
683 "Disabling barriers, trial barrier write failed");
684 mp->m_flags &= ~XFS_MOUNT_BARRIER;
690 xfs_blkdev_issue_flush(
691 xfs_buftarg_t *buftarg)
693 blkdev_issue_flush(buftarg->bt_bdev, GFP_KERNEL, 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 "XFS: 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, 0, mp->m_fsname);
758 if (!mp->m_ddev_targp)
759 goto out_close_rtdev;
762 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, 1,
764 if (!mp->m_rtdev_targp)
765 goto out_free_ddev_targ;
768 if (logdev && logdev != ddev) {
769 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, 1,
771 if (!mp->m_logdev_targp)
772 goto out_free_rtdev_targ;
774 mp->m_logdev_targp = mp->m_ddev_targp;
780 if (mp->m_rtdev_targp)
781 xfs_free_buftarg(mp, mp->m_rtdev_targp);
783 xfs_free_buftarg(mp, mp->m_ddev_targp);
786 xfs_blkdev_put(rtdev);
788 if (logdev && logdev != ddev)
789 xfs_blkdev_put(logdev);
795 * Setup xfs_mount buffer target pointers based on superblock
799 struct xfs_mount *mp)
803 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
804 mp->m_sb.sb_sectsize);
808 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
809 unsigned int log_sector_size = BBSIZE;
811 if (xfs_sb_version_hassector(&mp->m_sb))
812 log_sector_size = mp->m_sb.sb_logsectsize;
813 error = xfs_setsize_buftarg(mp->m_logdev_targp,
814 mp->m_sb.sb_blocksize,
819 if (mp->m_rtdev_targp) {
820 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
821 mp->m_sb.sb_blocksize,
822 mp->m_sb.sb_sectsize);
831 * XFS AIL push thread support
835 struct xfs_ail *ailp,
836 xfs_lsn_t threshold_lsn)
838 /* only ever move the target forwards */
839 if (XFS_LSN_CMP(threshold_lsn, ailp->xa_target) > 0) {
840 ailp->xa_target = threshold_lsn;
841 wake_up_process(ailp->xa_task);
849 struct xfs_ail *ailp = data;
850 xfs_lsn_t last_pushed_lsn = 0;
851 long tout = 0; /* milliseconds */
853 while (!kthread_should_stop()) {
855 * for short sleeps indicating congestion, don't allow us to
856 * get woken early. Otherwise all we do is bang on the AIL lock
857 * without making progress.
859 if (tout && tout <= 20)
860 __set_current_state(TASK_KILLABLE);
862 __set_current_state(TASK_INTERRUPTIBLE);
863 schedule_timeout(tout ?
864 msecs_to_jiffies(tout) : MAX_SCHEDULE_TIMEOUT);
869 ASSERT(ailp->xa_mount->m_log);
870 if (XFS_FORCED_SHUTDOWN(ailp->xa_mount))
873 tout = xfsaild_push(ailp, &last_pushed_lsn);
881 struct xfs_ail *ailp)
884 ailp->xa_task = kthread_run(xfsaild, ailp, "xfsaild/%s",
885 ailp->xa_mount->m_fsname);
886 if (IS_ERR(ailp->xa_task))
887 return -PTR_ERR(ailp->xa_task);
893 struct xfs_ail *ailp)
895 kthread_stop(ailp->xa_task);
899 /* Catch misguided souls that try to use this interface on XFS */
900 STATIC struct inode *
902 struct super_block *sb)
909 * Now that the generic code is guaranteed not to be accessing
910 * the linux inode, we can reclaim the inode.
913 xfs_fs_destroy_inode(
916 struct xfs_inode *ip = XFS_I(inode);
918 trace_xfs_destroy_inode(ip);
920 XFS_STATS_INC(vn_reclaim);
922 /* bad inode, get out here ASAP */
923 if (is_bad_inode(inode))
928 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
931 * We should never get here with one of the reclaim flags already set.
933 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
934 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
937 * We always use background reclaim here because even if the
938 * inode is clean, it still may be under IO and hence we have
939 * to take the flush lock. The background reclaim path handles
940 * this more efficiently than we can here, so simply let background
941 * reclaim tear down all inodes.
944 xfs_inode_set_reclaim_tag(ip);
948 * Slab object creation initialisation for the XFS inode.
949 * This covers only the idempotent fields in the XFS inode;
950 * all other fields need to be initialised on allocation
951 * from the slab. This avoids the need to repeatedly initialise
952 * fields in the xfs inode that left in the initialise state
953 * when freeing the inode.
956 xfs_fs_inode_init_once(
959 struct xfs_inode *ip = inode;
961 memset(ip, 0, sizeof(struct xfs_inode));
964 inode_init_once(VFS_I(ip));
967 atomic_set(&ip->i_iocount, 0);
968 atomic_set(&ip->i_pincount, 0);
969 spin_lock_init(&ip->i_flags_lock);
970 init_waitqueue_head(&ip->i_ipin_wait);
972 * Because we want to use a counting completion, complete
973 * the flush completion once to allow a single access to
974 * the flush completion without blocking.
976 init_completion(&ip->i_flush);
977 complete(&ip->i_flush);
979 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
980 "xfsino", ip->i_ino);
984 * Dirty the XFS inode when mark_inode_dirty_sync() is called so that
985 * we catch unlogged VFS level updates to the inode.
987 * We need the barrier() to maintain correct ordering between unlogged
988 * updates and the transaction commit code that clears the i_update_core
989 * field. This requires all updates to be completed before marking the
997 XFS_I(inode)->i_update_core = 1;
1002 struct xfs_inode *ip)
1004 struct xfs_mount *mp = ip->i_mount;
1005 struct xfs_trans *tp;
1008 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1009 tp = xfs_trans_alloc(mp, XFS_TRANS_FSYNC_TS);
1010 error = xfs_trans_reserve(tp, 0, XFS_FSYNC_TS_LOG_RES(mp), 0, 0, 0);
1013 xfs_trans_cancel(tp, 0);
1014 /* we need to return with the lock hold shared */
1015 xfs_ilock(ip, XFS_ILOCK_SHARED);
1019 xfs_ilock(ip, XFS_ILOCK_EXCL);
1022 * Note - it's possible that we might have pushed ourselves out of the
1023 * way during trans_reserve which would flush the inode. But there's
1024 * no guarantee that the inode buffer has actually gone out yet (it's
1025 * delwri). Plus the buffer could be pinned anyway if it's part of
1026 * an inode in another recent transaction. So we play it safe and
1027 * fire off the transaction anyway.
1029 xfs_trans_ijoin(tp, ip);
1030 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1031 error = xfs_trans_commit(tp, 0);
1032 xfs_ilock_demote(ip, XFS_ILOCK_EXCL);
1039 struct inode *inode,
1040 struct writeback_control *wbc)
1042 struct xfs_inode *ip = XFS_I(inode);
1043 struct xfs_mount *mp = ip->i_mount;
1046 trace_xfs_write_inode(ip);
1048 if (XFS_FORCED_SHUTDOWN(mp))
1049 return XFS_ERROR(EIO);
1051 if (wbc->sync_mode == WB_SYNC_ALL) {
1053 * Make sure the inode has made it it into the log. Instead
1054 * of forcing it all the way to stable storage using a
1055 * synchronous transaction we let the log force inside the
1056 * ->sync_fs call do that for thus, which reduces the number
1057 * of synchronous log foces dramatically.
1060 xfs_ilock(ip, XFS_ILOCK_SHARED);
1061 if (ip->i_update_core) {
1062 error = xfs_log_inode(ip);
1068 * We make this non-blocking if the inode is contended, return
1069 * EAGAIN to indicate to the caller that they did not succeed.
1070 * This prevents the flush path from blocking on inodes inside
1071 * another operation right now, they get caught later by
1074 if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED))
1077 if (xfs_ipincount(ip) || !xfs_iflock_nowait(ip))
1081 * Now we have the flush lock and the inode is not pinned, we
1082 * can check if the inode is really clean as we know that
1083 * there are no pending transaction completions, it is not
1084 * waiting on the delayed write queue and there is no IO in
1087 if (xfs_inode_clean(ip)) {
1092 error = xfs_iflush(ip, 0);
1096 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1099 * if we failed to write out the inode then mark
1100 * it dirty again so we'll try again later.
1103 xfs_mark_inode_dirty_sync(ip);
1109 struct inode *inode)
1111 xfs_inode_t *ip = XFS_I(inode);
1113 trace_xfs_evict_inode(ip);
1115 truncate_inode_pages(&inode->i_data, 0);
1116 end_writeback(inode);
1117 XFS_STATS_INC(vn_rele);
1118 XFS_STATS_INC(vn_remove);
1119 XFS_STATS_DEC(vn_active);
1122 * The iolock is used by the file system to coordinate reads,
1123 * writes, and block truncates. Up to this point the lock
1124 * protected concurrent accesses by users of the inode. But
1125 * from here forward we're doing some final processing of the
1126 * inode because we're done with it, and although we reuse the
1127 * iolock for protection it is really a distinct lock class
1128 * (in the lockdep sense) from before. To keep lockdep happy
1129 * (and basically indicate what we are doing), we explicitly
1130 * re-init the iolock here.
1132 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
1133 mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
1134 lockdep_set_class_and_name(&ip->i_iolock.mr_lock,
1135 &xfs_iolock_reclaimable, "xfs_iolock_reclaimable");
1142 struct xfs_mount *mp)
1144 kfree(mp->m_fsname);
1145 kfree(mp->m_rtname);
1146 kfree(mp->m_logname);
1151 struct super_block *sb)
1153 struct xfs_mount *mp = XFS_M(sb);
1156 * Unregister the memory shrinker before we tear down the mount
1157 * structure so we don't have memory reclaim racing with us here.
1159 xfs_inode_shrinker_unregister(mp);
1163 * Blow away any referenced inode in the filestreams cache.
1164 * This can and will cause log traffic as inodes go inactive
1167 xfs_filestream_unmount(mp);
1169 XFS_bflush(mp->m_ddev_targp);
1173 xfs_icsb_destroy_counters(mp);
1174 xfs_close_devices(mp);
1175 xfs_free_fsname(mp);
1181 struct super_block *sb,
1184 struct xfs_mount *mp = XFS_M(sb);
1188 * Not much we can do for the first async pass. Writing out the
1189 * superblock would be counter-productive as we are going to redirty
1190 * when writing out other data and metadata (and writing out a single
1191 * block is quite fast anyway).
1193 * Try to asynchronously kick off quota syncing at least.
1196 xfs_qm_sync(mp, SYNC_TRYLOCK);
1200 error = xfs_quiesce_data(mp);
1205 int prev_sync_seq = mp->m_sync_seq;
1208 * The disk must be active because we're syncing.
1209 * We schedule xfssyncd now (now that the disk is
1210 * active) instead of later (when it might not be).
1212 wake_up_process(mp->m_sync_task);
1214 * We have to wait for the sync iteration to complete.
1215 * If we don't, the disk activity caused by the sync
1216 * will come after the sync is completed, and that
1217 * triggers another sync from laptop mode.
1219 wait_event(mp->m_wait_single_sync_task,
1220 mp->m_sync_seq != prev_sync_seq);
1228 struct dentry *dentry,
1229 struct kstatfs *statp)
1231 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1232 xfs_sb_t *sbp = &mp->m_sb;
1233 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1234 __uint64_t fakeinos, id;
1238 statp->f_type = XFS_SB_MAGIC;
1239 statp->f_namelen = MAXNAMELEN - 1;
1241 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1242 statp->f_fsid.val[0] = (u32)id;
1243 statp->f_fsid.val[1] = (u32)(id >> 32);
1245 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1247 spin_lock(&mp->m_sb_lock);
1248 statp->f_bsize = sbp->sb_blocksize;
1249 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1250 statp->f_blocks = sbp->sb_dblocks - lsize;
1251 statp->f_bfree = statp->f_bavail =
1252 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1253 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1255 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1256 if (mp->m_maxicount)
1257 statp->f_files = min_t(typeof(statp->f_files),
1261 /* make sure statp->f_ffree does not underflow */
1262 ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1263 statp->f_ffree = max_t(__int64_t, ffree, 0);
1265 spin_unlock(&mp->m_sb_lock);
1267 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) ||
1268 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))) ==
1269 (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
1270 xfs_qm_statvfs(ip, statp);
1275 xfs_save_resvblks(struct xfs_mount *mp)
1277 __uint64_t resblks = 0;
1279 mp->m_resblks_save = mp->m_resblks;
1280 xfs_reserve_blocks(mp, &resblks, NULL);
1284 xfs_restore_resvblks(struct xfs_mount *mp)
1288 if (mp->m_resblks_save) {
1289 resblks = mp->m_resblks_save;
1290 mp->m_resblks_save = 0;
1292 resblks = xfs_default_resblks(mp);
1294 xfs_reserve_blocks(mp, &resblks, NULL);
1299 struct super_block *sb,
1303 struct xfs_mount *mp = XFS_M(sb);
1304 substring_t args[MAX_OPT_ARGS];
1308 while ((p = strsep(&options, ",")) != NULL) {
1314 token = match_token(p, tokens, args);
1317 mp->m_flags |= XFS_MOUNT_BARRIER;
1320 * Test if barriers are actually working if we can,
1321 * else delay this check until the filesystem is
1324 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1325 xfs_mountfs_check_barriers(mp);
1328 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1332 * Logically we would return an error here to prevent
1333 * users from believing they might have changed
1334 * mount options using remount which can't be changed.
1336 * But unfortunately mount(8) adds all options from
1337 * mtab and fstab to the mount arguments in some cases
1338 * so we can't blindly reject options, but have to
1339 * check for each specified option if it actually
1340 * differs from the currently set option and only
1341 * reject it if that's the case.
1343 * Until that is implemented we return success for
1344 * every remount request, and silently ignore all
1345 * options that we can't actually change.
1349 "XFS: mount option \"%s\" not supported for remount\n", p);
1358 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1359 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1360 if (mp->m_flags & XFS_MOUNT_BARRIER)
1361 xfs_mountfs_check_barriers(mp);
1364 * If this is the first remount to writeable state we
1365 * might have some superblock changes to update.
1367 if (mp->m_update_flags) {
1368 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1371 "XFS: failed to write sb changes");
1374 mp->m_update_flags = 0;
1378 * Fill out the reserve pool if it is empty. Use the stashed
1379 * value if it is non-zero, otherwise go with the default.
1381 xfs_restore_resvblks(mp);
1385 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1387 * After we have synced the data but before we sync the
1388 * metadata, we need to free up the reserve block pool so that
1389 * the used block count in the superblock on disk is correct at
1390 * the end of the remount. Stash the current reserve pool size
1391 * so that if we get remounted rw, we can return it to the same
1395 xfs_quiesce_data(mp);
1396 xfs_save_resvblks(mp);
1397 xfs_quiesce_attr(mp);
1398 mp->m_flags |= XFS_MOUNT_RDONLY;
1405 * Second stage of a freeze. The data is already frozen so we only
1406 * need to take care of the metadata. Once that's done write a dummy
1407 * record to dirty the log in case of a crash while frozen.
1411 struct super_block *sb)
1413 struct xfs_mount *mp = XFS_M(sb);
1415 xfs_save_resvblks(mp);
1416 xfs_quiesce_attr(mp);
1417 return -xfs_fs_log_dummy(mp, SYNC_WAIT);
1422 struct super_block *sb)
1424 struct xfs_mount *mp = XFS_M(sb);
1426 xfs_restore_resvblks(mp);
1431 xfs_fs_show_options(
1433 struct vfsmount *mnt)
1435 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1439 * This function fills in xfs_mount_t fields based on mount args.
1440 * Note: the superblock _has_ now been read in.
1444 struct xfs_mount *mp)
1446 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1448 /* Fail a mount where the logbuf is smaller than the log stripe */
1449 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1450 if (mp->m_logbsize <= 0 &&
1451 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1452 mp->m_logbsize = mp->m_sb.sb_logsunit;
1453 } else if (mp->m_logbsize > 0 &&
1454 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1456 "XFS: logbuf size must be greater than or equal to log stripe size");
1457 return XFS_ERROR(EINVAL);
1460 /* Fail a mount if the logbuf is larger than 32K */
1461 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1463 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1464 return XFS_ERROR(EINVAL);
1469 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1470 * told by noattr2 to turn it off
1472 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1473 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1474 mp->m_flags |= XFS_MOUNT_ATTR2;
1477 * prohibit r/w mounts of read-only filesystems
1479 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1481 "XFS: cannot mount a read-only filesystem as read-write");
1482 return XFS_ERROR(EROFS);
1490 struct super_block *sb,
1495 struct xfs_mount *mp = NULL;
1496 int flags = 0, error = ENOMEM;
1498 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1502 spin_lock_init(&mp->m_sb_lock);
1503 mutex_init(&mp->m_growlock);
1504 atomic_set(&mp->m_active_trans, 0);
1505 INIT_LIST_HEAD(&mp->m_sync_list);
1506 spin_lock_init(&mp->m_sync_lock);
1507 init_waitqueue_head(&mp->m_wait_single_sync_task);
1512 error = xfs_parseargs(mp, (char *)data);
1514 goto out_free_fsname;
1516 sb_min_blocksize(sb, BBSIZE);
1517 sb->s_xattr = xfs_xattr_handlers;
1518 sb->s_export_op = &xfs_export_operations;
1519 #ifdef CONFIG_XFS_QUOTA
1520 sb->s_qcop = &xfs_quotactl_operations;
1522 sb->s_op = &xfs_super_operations;
1525 flags |= XFS_MFSI_QUIET;
1527 error = xfs_open_devices(mp);
1529 goto out_free_fsname;
1531 error = xfs_icsb_init_counters(mp);
1533 goto out_close_devices;
1535 error = xfs_readsb(mp, flags);
1537 goto out_destroy_counters;
1539 error = xfs_finish_flags(mp);
1543 error = xfs_setup_devices(mp);
1547 if (mp->m_flags & XFS_MOUNT_BARRIER)
1548 xfs_mountfs_check_barriers(mp);
1550 error = xfs_filestream_mount(mp);
1554 error = xfs_mountfs(mp);
1556 goto out_filestream_unmount;
1558 sb->s_magic = XFS_SB_MAGIC;
1559 sb->s_blocksize = mp->m_sb.sb_blocksize;
1560 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1561 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1562 sb->s_time_gran = 1;
1563 set_posix_acl_flag(sb);
1565 root = igrab(VFS_I(mp->m_rootip));
1570 if (is_bad_inode(root)) {
1574 sb->s_root = d_alloc_root(root);
1580 error = xfs_syncd_init(mp);
1584 xfs_inode_shrinker_register(mp);
1588 out_filestream_unmount:
1589 xfs_filestream_unmount(mp);
1592 out_destroy_counters:
1593 xfs_icsb_destroy_counters(mp);
1595 xfs_close_devices(mp);
1597 xfs_free_fsname(mp);
1612 * Blow away any referenced inode in the filestreams cache.
1613 * This can and will cause log traffic as inodes go inactive
1616 xfs_filestream_unmount(mp);
1618 XFS_bflush(mp->m_ddev_targp);
1624 STATIC struct dentry *
1626 struct file_system_type *fs_type,
1628 const char *dev_name,
1631 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1634 static const struct super_operations xfs_super_operations = {
1635 .alloc_inode = xfs_fs_alloc_inode,
1636 .destroy_inode = xfs_fs_destroy_inode,
1637 .dirty_inode = xfs_fs_dirty_inode,
1638 .write_inode = xfs_fs_write_inode,
1639 .evict_inode = xfs_fs_evict_inode,
1640 .put_super = xfs_fs_put_super,
1641 .sync_fs = xfs_fs_sync_fs,
1642 .freeze_fs = xfs_fs_freeze,
1643 .unfreeze_fs = xfs_fs_unfreeze,
1644 .statfs = xfs_fs_statfs,
1645 .remount_fs = xfs_fs_remount,
1646 .show_options = xfs_fs_show_options,
1649 static struct file_system_type xfs_fs_type = {
1650 .owner = THIS_MODULE,
1652 .mount = xfs_fs_mount,
1653 .kill_sb = kill_block_super,
1654 .fs_flags = FS_REQUIRES_DEV,
1658 xfs_init_zones(void)
1661 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1662 if (!xfs_ioend_zone)
1665 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1667 if (!xfs_ioend_pool)
1668 goto out_destroy_ioend_zone;
1670 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1672 if (!xfs_log_ticket_zone)
1673 goto out_destroy_ioend_pool;
1675 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1676 "xfs_bmap_free_item");
1677 if (!xfs_bmap_free_item_zone)
1678 goto out_destroy_log_ticket_zone;
1680 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1682 if (!xfs_btree_cur_zone)
1683 goto out_destroy_bmap_free_item_zone;
1685 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1687 if (!xfs_da_state_zone)
1688 goto out_destroy_btree_cur_zone;
1690 xfs_dabuf_zone = kmem_zone_init(sizeof(xfs_dabuf_t), "xfs_dabuf");
1691 if (!xfs_dabuf_zone)
1692 goto out_destroy_da_state_zone;
1694 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1695 if (!xfs_ifork_zone)
1696 goto out_destroy_dabuf_zone;
1698 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1699 if (!xfs_trans_zone)
1700 goto out_destroy_ifork_zone;
1702 xfs_log_item_desc_zone =
1703 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1704 "xfs_log_item_desc");
1705 if (!xfs_log_item_desc_zone)
1706 goto out_destroy_trans_zone;
1709 * The size of the zone allocated buf log item is the maximum
1710 * size possible under XFS. This wastes a little bit of memory,
1711 * but it is much faster.
1713 xfs_buf_item_zone = kmem_zone_init((sizeof(xfs_buf_log_item_t) +
1714 (((XFS_MAX_BLOCKSIZE / XFS_BLF_CHUNK) /
1715 NBWORD) * sizeof(int))), "xfs_buf_item");
1716 if (!xfs_buf_item_zone)
1717 goto out_destroy_log_item_desc_zone;
1719 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1720 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1721 sizeof(xfs_extent_t))), "xfs_efd_item");
1723 goto out_destroy_buf_item_zone;
1725 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1726 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1727 sizeof(xfs_extent_t))), "xfs_efi_item");
1729 goto out_destroy_efd_zone;
1732 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1733 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1734 xfs_fs_inode_init_once);
1735 if (!xfs_inode_zone)
1736 goto out_destroy_efi_zone;
1739 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1740 KM_ZONE_SPREAD, NULL);
1742 goto out_destroy_inode_zone;
1746 out_destroy_inode_zone:
1747 kmem_zone_destroy(xfs_inode_zone);
1748 out_destroy_efi_zone:
1749 kmem_zone_destroy(xfs_efi_zone);
1750 out_destroy_efd_zone:
1751 kmem_zone_destroy(xfs_efd_zone);
1752 out_destroy_buf_item_zone:
1753 kmem_zone_destroy(xfs_buf_item_zone);
1754 out_destroy_log_item_desc_zone:
1755 kmem_zone_destroy(xfs_log_item_desc_zone);
1756 out_destroy_trans_zone:
1757 kmem_zone_destroy(xfs_trans_zone);
1758 out_destroy_ifork_zone:
1759 kmem_zone_destroy(xfs_ifork_zone);
1760 out_destroy_dabuf_zone:
1761 kmem_zone_destroy(xfs_dabuf_zone);
1762 out_destroy_da_state_zone:
1763 kmem_zone_destroy(xfs_da_state_zone);
1764 out_destroy_btree_cur_zone:
1765 kmem_zone_destroy(xfs_btree_cur_zone);
1766 out_destroy_bmap_free_item_zone:
1767 kmem_zone_destroy(xfs_bmap_free_item_zone);
1768 out_destroy_log_ticket_zone:
1769 kmem_zone_destroy(xfs_log_ticket_zone);
1770 out_destroy_ioend_pool:
1771 mempool_destroy(xfs_ioend_pool);
1772 out_destroy_ioend_zone:
1773 kmem_zone_destroy(xfs_ioend_zone);
1779 xfs_destroy_zones(void)
1781 kmem_zone_destroy(xfs_ili_zone);
1782 kmem_zone_destroy(xfs_inode_zone);
1783 kmem_zone_destroy(xfs_efi_zone);
1784 kmem_zone_destroy(xfs_efd_zone);
1785 kmem_zone_destroy(xfs_buf_item_zone);
1786 kmem_zone_destroy(xfs_log_item_desc_zone);
1787 kmem_zone_destroy(xfs_trans_zone);
1788 kmem_zone_destroy(xfs_ifork_zone);
1789 kmem_zone_destroy(xfs_dabuf_zone);
1790 kmem_zone_destroy(xfs_da_state_zone);
1791 kmem_zone_destroy(xfs_btree_cur_zone);
1792 kmem_zone_destroy(xfs_bmap_free_item_zone);
1793 kmem_zone_destroy(xfs_log_ticket_zone);
1794 mempool_destroy(xfs_ioend_pool);
1795 kmem_zone_destroy(xfs_ioend_zone);
1804 printk(KERN_INFO XFS_VERSION_STRING " with "
1805 XFS_BUILD_OPTIONS " enabled\n");
1810 error = xfs_init_zones();
1814 error = xfs_mru_cache_init();
1816 goto out_destroy_zones;
1818 error = xfs_filestream_init();
1820 goto out_mru_cache_uninit;
1822 error = xfs_buf_init();
1824 goto out_filestream_uninit;
1826 error = xfs_init_procfs();
1828 goto out_buf_terminate;
1830 error = xfs_sysctl_register();
1832 goto out_cleanup_procfs;
1836 error = register_filesystem(&xfs_fs_type);
1838 goto out_sysctl_unregister;
1841 out_sysctl_unregister:
1842 xfs_sysctl_unregister();
1844 xfs_cleanup_procfs();
1846 xfs_buf_terminate();
1847 out_filestream_uninit:
1848 xfs_filestream_uninit();
1849 out_mru_cache_uninit:
1850 xfs_mru_cache_uninit();
1852 xfs_destroy_zones();
1861 unregister_filesystem(&xfs_fs_type);
1862 xfs_sysctl_unregister();
1863 xfs_cleanup_procfs();
1864 xfs_buf_terminate();
1865 xfs_filestream_uninit();
1866 xfs_mru_cache_uninit();
1867 xfs_destroy_zones();
1870 module_init(init_xfs_fs);
1871 module_exit(exit_xfs_fs);
1873 MODULE_AUTHOR("Silicon Graphics, Inc.");
1874 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1875 MODULE_LICENSE("GPL");