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_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
94 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
95 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
96 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
97 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
98 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
100 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
101 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
102 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
103 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
104 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
105 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
106 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
107 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
108 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
109 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
110 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
111 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
112 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
113 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
114 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
115 #define MNTOPT_DELAYLOG "delaylog" /* Delayed loging enabled */
116 #define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed loging disabled */
119 * Table driven mount option parser.
121 * Currently only used for remount, but it will be used for mount
122 * in the future, too.
125 Opt_barrier, Opt_nobarrier, Opt_err
128 static const match_table_t tokens = {
129 {Opt_barrier, "barrier"},
130 {Opt_nobarrier, "nobarrier"},
136 suffix_strtoul(char *s, char **endp, unsigned int base)
138 int last, shift_left_factor = 0;
141 last = strlen(value) - 1;
142 if (value[last] == 'K' || value[last] == 'k') {
143 shift_left_factor = 10;
146 if (value[last] == 'M' || value[last] == 'm') {
147 shift_left_factor = 20;
150 if (value[last] == 'G' || value[last] == 'g') {
151 shift_left_factor = 30;
155 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
159 * This function fills in xfs_mount_t fields based on mount args.
160 * Note: the superblock has _not_ yet been read in.
162 * Note that this function leaks the various device name allocations on
163 * failure. The caller takes care of them.
167 struct xfs_mount *mp,
170 struct super_block *sb = mp->m_super;
171 char *this_char, *value, *eov;
175 __uint8_t iosizelog = 0;
178 * Copy binary VFS mount flags we are interested in.
180 if (sb->s_flags & MS_RDONLY)
181 mp->m_flags |= XFS_MOUNT_RDONLY;
182 if (sb->s_flags & MS_DIRSYNC)
183 mp->m_flags |= XFS_MOUNT_DIRSYNC;
184 if (sb->s_flags & MS_SYNCHRONOUS)
185 mp->m_flags |= XFS_MOUNT_WSYNC;
188 * Set some default flags that could be cleared by the mount option
191 mp->m_flags |= XFS_MOUNT_BARRIER;
192 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
193 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
196 * These can be overridden by the mount option parsing.
204 while ((this_char = strsep(&options, ",")) != NULL) {
207 if ((value = strchr(this_char, '=')) != NULL)
210 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
211 if (!value || !*value) {
213 "XFS: %s option requires an argument",
217 mp->m_logbufs = simple_strtoul(value, &eov, 10);
218 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
219 if (!value || !*value) {
221 "XFS: %s option requires an argument",
225 mp->m_logbsize = suffix_strtoul(value, &eov, 10);
226 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
227 if (!value || !*value) {
229 "XFS: %s option requires an argument",
233 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
236 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
238 "XFS: %s option not allowed on this system",
241 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
242 if (!value || !*value) {
244 "XFS: %s option requires an argument",
248 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
251 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
252 if (!value || !*value) {
254 "XFS: %s option requires an argument",
258 iosize = simple_strtoul(value, &eov, 10);
259 iosizelog = ffs(iosize) - 1;
260 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
261 if (!value || !*value) {
263 "XFS: %s option requires an argument",
267 iosize = suffix_strtoul(value, &eov, 10);
268 iosizelog = ffs(iosize) - 1;
269 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
270 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
271 mp->m_flags |= XFS_MOUNT_GRPID;
272 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
273 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
274 mp->m_flags &= ~XFS_MOUNT_GRPID;
275 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
276 mp->m_flags |= XFS_MOUNT_WSYNC;
277 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
278 mp->m_flags |= XFS_MOUNT_OSYNCISOSYNC;
279 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
280 mp->m_flags |= XFS_MOUNT_NORECOVERY;
281 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
282 mp->m_flags |= XFS_MOUNT_NOALIGN;
283 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
284 mp->m_flags |= XFS_MOUNT_SWALLOC;
285 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
286 if (!value || !*value) {
288 "XFS: %s option requires an argument",
292 dsunit = simple_strtoul(value, &eov, 10);
293 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
294 if (!value || !*value) {
296 "XFS: %s option requires an argument",
300 dswidth = simple_strtoul(value, &eov, 10);
301 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
302 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
305 "XFS: %s option not allowed on this system",
309 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
310 mp->m_flags |= XFS_MOUNT_NOUUID;
311 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
312 mp->m_flags |= XFS_MOUNT_BARRIER;
313 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
314 mp->m_flags &= ~XFS_MOUNT_BARRIER;
315 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
316 mp->m_flags |= XFS_MOUNT_IKEEP;
317 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
318 mp->m_flags &= ~XFS_MOUNT_IKEEP;
319 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
320 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
321 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
322 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
323 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
324 mp->m_flags |= XFS_MOUNT_ATTR2;
325 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
326 mp->m_flags &= ~XFS_MOUNT_ATTR2;
327 mp->m_flags |= XFS_MOUNT_NOATTR2;
328 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
329 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
330 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
331 mp->m_qflags &= ~(XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
332 XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
333 XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
334 XFS_UQUOTA_ENFD | XFS_OQUOTA_ENFD);
335 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
336 !strcmp(this_char, MNTOPT_UQUOTA) ||
337 !strcmp(this_char, MNTOPT_USRQUOTA)) {
338 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
340 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
341 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
342 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
343 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
344 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
345 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
346 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
348 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
349 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
350 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
351 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
352 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
353 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
355 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
356 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
357 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
358 } else if (!strcmp(this_char, MNTOPT_DELAYLOG)) {
359 mp->m_flags |= XFS_MOUNT_DELAYLOG;
361 "Enabling EXPERIMENTAL delayed logging feature "
362 "- use at your own risk.\n");
363 } else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) {
364 mp->m_flags &= ~XFS_MOUNT_DELAYLOG;
365 } else if (!strcmp(this_char, "ihashsize")) {
367 "XFS: ihashsize no longer used, option is deprecated.");
368 } else if (!strcmp(this_char, "osyncisdsync")) {
369 /* no-op, this is now the default */
371 "XFS: osyncisdsync is now the default, option is deprecated.");
372 } else if (!strcmp(this_char, "irixsgid")) {
374 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
377 "XFS: unknown mount option [%s].", this_char);
383 * no recovery flag requires a read-only mount
385 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
386 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
387 cmn_err(CE_WARN, "XFS: no-recovery mounts must be read-only.");
391 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
393 "XFS: sunit and swidth options incompatible with the noalign option");
397 #ifndef CONFIG_XFS_QUOTA
398 if (XFS_IS_QUOTA_RUNNING(mp)) {
400 "XFS: quota support not available in this kernel.");
405 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
406 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE))) {
408 "XFS: cannot mount with both project and group quota");
412 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
414 "XFS: sunit and swidth must be specified together");
418 if (dsunit && (dswidth % dsunit != 0)) {
420 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
426 if (!(mp->m_flags & XFS_MOUNT_NOALIGN)) {
428 * At this point the superblock has not been read
429 * in, therefore we do not know the block size.
430 * Before the mount call ends we will convert
434 mp->m_dalign = dsunit;
435 mp->m_flags |= XFS_MOUNT_RETERR;
439 mp->m_swidth = dswidth;
442 if (mp->m_logbufs != -1 &&
443 mp->m_logbufs != 0 &&
444 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
445 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
447 "XFS: invalid logbufs value: %d [not %d-%d]",
448 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
449 return XFS_ERROR(EINVAL);
451 if (mp->m_logbsize != -1 &&
452 mp->m_logbsize != 0 &&
453 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
454 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
455 !is_power_of_2(mp->m_logbsize))) {
457 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
459 return XFS_ERROR(EINVAL);
462 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
465 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
468 if (iosizelog > XFS_MAX_IO_LOG ||
469 iosizelog < XFS_MIN_IO_LOG) {
471 "XFS: invalid log iosize: %d [not %d-%d]",
472 iosizelog, XFS_MIN_IO_LOG,
474 return XFS_ERROR(EINVAL);
477 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
478 mp->m_readio_log = iosizelog;
479 mp->m_writeio_log = iosizelog;
485 struct proc_xfs_info {
492 struct xfs_mount *mp,
495 static struct proc_xfs_info xfs_info_set[] = {
496 /* the few simple ones we can get from the mount struct */
497 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
498 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
499 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
500 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
501 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
502 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
503 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
504 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
505 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
506 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
507 { XFS_MOUNT_DELAYLOG, "," MNTOPT_DELAYLOG },
510 static struct proc_xfs_info xfs_info_unset[] = {
511 /* the few simple ones we can get from the mount struct */
512 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
513 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
514 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
517 struct proc_xfs_info *xfs_infop;
519 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
520 if (mp->m_flags & xfs_infop->flag)
521 seq_puts(m, xfs_infop->str);
523 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
524 if (!(mp->m_flags & xfs_infop->flag))
525 seq_puts(m, xfs_infop->str);
528 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
529 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
530 (int)(1 << mp->m_writeio_log) >> 10);
532 if (mp->m_logbufs > 0)
533 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
534 if (mp->m_logbsize > 0)
535 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
538 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
540 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
542 if (mp->m_dalign > 0)
543 seq_printf(m, "," MNTOPT_SUNIT "=%d",
544 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
545 if (mp->m_swidth > 0)
546 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
547 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
549 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
550 seq_puts(m, "," MNTOPT_USRQUOTA);
551 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
552 seq_puts(m, "," MNTOPT_UQUOTANOENF);
554 /* Either project or group quotas can be active, not both */
556 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
557 if (mp->m_qflags & XFS_OQUOTA_ENFD)
558 seq_puts(m, "," MNTOPT_PRJQUOTA);
560 seq_puts(m, "," MNTOPT_PQUOTANOENF);
561 } else if (mp->m_qflags & XFS_GQUOTA_ACCT) {
562 if (mp->m_qflags & XFS_OQUOTA_ENFD)
563 seq_puts(m, "," MNTOPT_GRPQUOTA);
565 seq_puts(m, "," MNTOPT_GQUOTANOENF);
568 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
569 seq_puts(m, "," MNTOPT_NOQUOTA);
575 unsigned int blockshift)
577 unsigned int pagefactor = 1;
578 unsigned int bitshift = BITS_PER_LONG - 1;
580 /* Figure out maximum filesize, on Linux this can depend on
581 * the filesystem blocksize (on 32 bit platforms).
582 * __block_prepare_write does this in an [unsigned] long...
583 * page->index << (PAGE_CACHE_SHIFT - bbits)
584 * So, for page sized blocks (4K on 32 bit platforms),
585 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
586 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
587 * but for smaller blocksizes it is less (bbits = log2 bsize).
588 * Note1: get_block_t takes a long (implicit cast from above)
589 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
590 * can optionally convert the [unsigned] long from above into
591 * an [unsigned] long long.
594 #if BITS_PER_LONG == 32
595 # if defined(CONFIG_LBDAF)
596 ASSERT(sizeof(sector_t) == 8);
597 pagefactor = PAGE_CACHE_SIZE;
598 bitshift = BITS_PER_LONG;
600 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
604 return (((__uint64_t)pagefactor) << bitshift) - 1;
611 struct block_device **bdevp)
615 *bdevp = open_bdev_exclusive(name, FMODE_READ|FMODE_WRITE, mp);
616 if (IS_ERR(*bdevp)) {
617 error = PTR_ERR(*bdevp);
618 printk("XFS: Invalid device [%s], error=%d\n", name, error);
626 struct block_device *bdev)
629 close_bdev_exclusive(bdev, FMODE_READ|FMODE_WRITE);
633 * Try to write out the superblock using barriers.
639 xfs_buf_t *sbp = xfs_getsb(mp, 0);
644 XFS_BUF_UNDELAYWRITE(sbp);
646 XFS_BUF_UNASYNC(sbp);
647 XFS_BUF_ORDERED(sbp);
650 error = xfs_iowait(sbp);
653 * Clear all the flags we set and possible error state in the
654 * buffer. We only did the write to try out whether barriers
655 * worked and shouldn't leave any traces in the superblock
659 XFS_BUF_ERROR(sbp, 0);
660 XFS_BUF_UNORDERED(sbp);
667 xfs_mountfs_check_barriers(xfs_mount_t *mp)
671 if (mp->m_logdev_targp != mp->m_ddev_targp) {
672 xfs_fs_cmn_err(CE_NOTE, mp,
673 "Disabling barriers, not supported with external log device");
674 mp->m_flags &= ~XFS_MOUNT_BARRIER;
678 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
679 xfs_fs_cmn_err(CE_NOTE, mp,
680 "Disabling barriers, underlying device is readonly");
681 mp->m_flags &= ~XFS_MOUNT_BARRIER;
685 error = xfs_barrier_test(mp);
687 xfs_fs_cmn_err(CE_NOTE, mp,
688 "Disabling barriers, trial barrier write failed");
689 mp->m_flags &= ~XFS_MOUNT_BARRIER;
695 xfs_blkdev_issue_flush(
696 xfs_buftarg_t *buftarg)
698 blkdev_issue_flush(buftarg->bt_bdev, GFP_KERNEL, NULL,
704 struct xfs_mount *mp)
706 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
707 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
708 xfs_free_buftarg(mp, mp->m_logdev_targp);
709 xfs_blkdev_put(logdev);
711 if (mp->m_rtdev_targp) {
712 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
713 xfs_free_buftarg(mp, mp->m_rtdev_targp);
714 xfs_blkdev_put(rtdev);
716 xfs_free_buftarg(mp, mp->m_ddev_targp);
720 * The file system configurations are:
721 * (1) device (partition) with data and internal log
722 * (2) logical volume with data and log subvolumes.
723 * (3) logical volume with data, log, and realtime subvolumes.
725 * We only have to handle opening the log and realtime volumes here if
726 * they are present. The data subvolume has already been opened by
727 * get_sb_bdev() and is stored in sb->s_bdev.
731 struct xfs_mount *mp)
733 struct block_device *ddev = mp->m_super->s_bdev;
734 struct block_device *logdev = NULL, *rtdev = NULL;
738 * Open real time and log devices - order is important.
741 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
747 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
749 goto out_close_logdev;
751 if (rtdev == ddev || rtdev == logdev) {
753 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
755 goto out_close_rtdev;
760 * Setup xfs_mount buffer target pointers
763 mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0, mp->m_fsname);
764 if (!mp->m_ddev_targp)
765 goto out_close_rtdev;
768 mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1, mp->m_fsname);
769 if (!mp->m_rtdev_targp)
770 goto out_free_ddev_targ;
773 if (logdev && logdev != ddev) {
774 mp->m_logdev_targp = xfs_alloc_buftarg(logdev, 1, mp->m_fsname);
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 xfs_itrace_entry(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. Care must be taken
978 * here - the transaction code calls mark_inode_dirty_sync() to mark the
979 * VFS inode dirty in a transaction and clears the i_update_core field;
980 * it must clear the field after calling mark_inode_dirty_sync() to
981 * correctly indicate that the dirty state has been propagated into the
984 * We need the barrier() to maintain correct ordering between unlogged
985 * updates and the transaction commit code that clears the i_update_core
986 * field. This requires all updates to be completed before marking the
994 XFS_I(inode)->i_update_core = 1;
999 struct xfs_inode *ip)
1001 struct xfs_mount *mp = ip->i_mount;
1002 struct xfs_trans *tp;
1005 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1006 tp = xfs_trans_alloc(mp, XFS_TRANS_FSYNC_TS);
1007 error = xfs_trans_reserve(tp, 0, XFS_FSYNC_TS_LOG_RES(mp), 0, 0, 0);
1010 xfs_trans_cancel(tp, 0);
1011 /* we need to return with the lock hold shared */
1012 xfs_ilock(ip, XFS_ILOCK_SHARED);
1016 xfs_ilock(ip, XFS_ILOCK_EXCL);
1019 * Note - it's possible that we might have pushed ourselves out of the
1020 * way during trans_reserve which would flush the inode. But there's
1021 * no guarantee that the inode buffer has actually gone out yet (it's
1022 * delwri). Plus the buffer could be pinned anyway if it's part of
1023 * an inode in another recent transaction. So we play it safe and
1024 * fire off the transaction anyway.
1026 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
1027 xfs_trans_ihold(tp, ip);
1028 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1029 xfs_trans_set_sync(tp);
1030 error = xfs_trans_commit(tp, 0);
1031 xfs_ilock_demote(ip, XFS_ILOCK_EXCL);
1038 struct inode *inode,
1039 struct writeback_control *wbc)
1041 struct xfs_inode *ip = XFS_I(inode);
1042 struct xfs_mount *mp = ip->i_mount;
1045 xfs_itrace_entry(ip);
1047 if (XFS_FORCED_SHUTDOWN(mp))
1048 return XFS_ERROR(EIO);
1050 if (wbc->sync_mode == WB_SYNC_ALL) {
1052 * Make sure the inode has hit stable storage. By using the
1053 * log and the fsync transactions we reduce the IOs we have
1054 * to do here from two (log and inode) to just the log.
1056 * Note: We still need to do a delwri write of the inode after
1057 * this to flush it to the backing buffer so that bulkstat
1058 * works properly if this is the first time the inode has been
1059 * written. Because we hold the ilock atomically over the
1060 * transaction commit and the inode flush we are guaranteed
1061 * that the inode is not pinned when it returns. If the flush
1062 * lock is already held, then the inode has already been
1063 * flushed once and we don't need to flush it again. Hence
1064 * the code will only flush the inode if it isn't already
1068 xfs_ilock(ip, XFS_ILOCK_SHARED);
1069 if (ip->i_update_core) {
1070 error = xfs_log_inode(ip);
1076 * We make this non-blocking if the inode is contended, return
1077 * EAGAIN to indicate to the caller that they did not succeed.
1078 * This prevents the flush path from blocking on inodes inside
1079 * another operation right now, they get caught later by xfs_sync.
1081 if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED))
1085 if (xfs_ipincount(ip) || !xfs_iflock_nowait(ip))
1089 * Now we have the flush lock and the inode is not pinned, we can check
1090 * if the inode is really clean as we know that there are no pending
1091 * transaction completions, it is not waiting on the delayed write
1092 * queue and there is no IO in progress.
1094 if (xfs_inode_clean(ip)) {
1099 error = xfs_iflush(ip, 0);
1102 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1105 * if we failed to write out the inode then mark
1106 * it dirty again so we'll try again later.
1109 xfs_mark_inode_dirty_sync(ip);
1115 struct inode *inode)
1117 xfs_inode_t *ip = XFS_I(inode);
1119 xfs_itrace_entry(ip);
1120 XFS_STATS_INC(vn_rele);
1121 XFS_STATS_INC(vn_remove);
1122 XFS_STATS_DEC(vn_active);
1125 * The iolock is used by the file system to coordinate reads,
1126 * writes, and block truncates. Up to this point the lock
1127 * protected concurrent accesses by users of the inode. But
1128 * from here forward we're doing some final processing of the
1129 * inode because we're done with it, and although we reuse the
1130 * iolock for protection it is really a distinct lock class
1131 * (in the lockdep sense) from before. To keep lockdep happy
1132 * (and basically indicate what we are doing), we explicitly
1133 * re-init the iolock here.
1135 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
1136 mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
1143 struct xfs_mount *mp)
1145 kfree(mp->m_fsname);
1146 kfree(mp->m_rtname);
1147 kfree(mp->m_logname);
1152 struct super_block *sb)
1154 struct xfs_mount *mp = XFS_M(sb);
1158 if (!(sb->s_flags & MS_RDONLY)) {
1160 * XXX(hch): this should be SYNC_WAIT.
1162 * Or more likely not needed at all because the VFS is already
1163 * calling ->sync_fs after shutting down all filestem
1164 * operations and just before calling ->put_super.
1166 xfs_sync_data(mp, 0);
1167 xfs_sync_attr(mp, 0);
1171 * Blow away any referenced inode in the filestreams cache.
1172 * This can and will cause log traffic as inodes go inactive
1175 xfs_filestream_unmount(mp);
1177 XFS_bflush(mp->m_ddev_targp);
1181 xfs_inode_shrinker_unregister(mp);
1182 xfs_icsb_destroy_counters(mp);
1183 xfs_close_devices(mp);
1184 xfs_free_fsname(mp);
1190 struct super_block *sb,
1193 struct xfs_mount *mp = XFS_M(sb);
1197 * Not much we can do for the first async pass. Writing out the
1198 * superblock would be counter-productive as we are going to redirty
1199 * when writing out other data and metadata (and writing out a single
1200 * block is quite fast anyway).
1202 * Try to asynchronously kick off quota syncing at least.
1205 xfs_qm_sync(mp, SYNC_TRYLOCK);
1209 error = xfs_quiesce_data(mp);
1214 int prev_sync_seq = mp->m_sync_seq;
1217 * The disk must be active because we're syncing.
1218 * We schedule xfssyncd now (now that the disk is
1219 * active) instead of later (when it might not be).
1221 wake_up_process(mp->m_sync_task);
1223 * We have to wait for the sync iteration to complete.
1224 * If we don't, the disk activity caused by the sync
1225 * will come after the sync is completed, and that
1226 * triggers another sync from laptop mode.
1228 wait_event(mp->m_wait_single_sync_task,
1229 mp->m_sync_seq != prev_sync_seq);
1237 struct dentry *dentry,
1238 struct kstatfs *statp)
1240 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1241 xfs_sb_t *sbp = &mp->m_sb;
1242 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1243 __uint64_t fakeinos, id;
1246 statp->f_type = XFS_SB_MAGIC;
1247 statp->f_namelen = MAXNAMELEN - 1;
1249 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1250 statp->f_fsid.val[0] = (u32)id;
1251 statp->f_fsid.val[1] = (u32)(id >> 32);
1253 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1255 spin_lock(&mp->m_sb_lock);
1256 statp->f_bsize = sbp->sb_blocksize;
1257 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1258 statp->f_blocks = sbp->sb_dblocks - lsize;
1259 statp->f_bfree = statp->f_bavail =
1260 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1261 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1263 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1264 if (mp->m_maxicount)
1265 statp->f_files = min_t(typeof(statp->f_files),
1268 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1269 spin_unlock(&mp->m_sb_lock);
1271 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) ||
1272 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))) ==
1273 (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
1274 xfs_qm_statvfs(ip, statp);
1279 xfs_save_resvblks(struct xfs_mount *mp)
1281 __uint64_t resblks = 0;
1283 mp->m_resblks_save = mp->m_resblks;
1284 xfs_reserve_blocks(mp, &resblks, NULL);
1288 xfs_restore_resvblks(struct xfs_mount *mp)
1292 if (mp->m_resblks_save) {
1293 resblks = mp->m_resblks_save;
1294 mp->m_resblks_save = 0;
1296 resblks = xfs_default_resblks(mp);
1298 xfs_reserve_blocks(mp, &resblks, NULL);
1303 struct super_block *sb,
1307 struct xfs_mount *mp = XFS_M(sb);
1308 substring_t args[MAX_OPT_ARGS];
1312 while ((p = strsep(&options, ",")) != NULL) {
1318 token = match_token(p, tokens, args);
1321 mp->m_flags |= XFS_MOUNT_BARRIER;
1324 * Test if barriers are actually working if we can,
1325 * else delay this check until the filesystem is
1328 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1329 xfs_mountfs_check_barriers(mp);
1332 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1336 * Logically we would return an error here to prevent
1337 * users from believing they might have changed
1338 * mount options using remount which can't be changed.
1340 * But unfortunately mount(8) adds all options from
1341 * mtab and fstab to the mount arguments in some cases
1342 * so we can't blindly reject options, but have to
1343 * check for each specified option if it actually
1344 * differs from the currently set option and only
1345 * reject it if that's the case.
1347 * Until that is implemented we return success for
1348 * every remount request, and silently ignore all
1349 * options that we can't actually change.
1353 "XFS: mount option \"%s\" not supported for remount\n", p);
1362 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1363 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1364 if (mp->m_flags & XFS_MOUNT_BARRIER)
1365 xfs_mountfs_check_barriers(mp);
1368 * If this is the first remount to writeable state we
1369 * might have some superblock changes to update.
1371 if (mp->m_update_flags) {
1372 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1375 "XFS: failed to write sb changes");
1378 mp->m_update_flags = 0;
1382 * Fill out the reserve pool if it is empty. Use the stashed
1383 * value if it is non-zero, otherwise go with the default.
1385 xfs_restore_resvblks(mp);
1389 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1391 * After we have synced the data but before we sync the
1392 * metadata, we need to free up the reserve block pool so that
1393 * the used block count in the superblock on disk is correct at
1394 * the end of the remount. Stash the current reserve pool size
1395 * so that if we get remounted rw, we can return it to the same
1399 xfs_quiesce_data(mp);
1400 xfs_save_resvblks(mp);
1401 xfs_quiesce_attr(mp);
1402 mp->m_flags |= XFS_MOUNT_RDONLY;
1409 * Second stage of a freeze. The data is already frozen so we only
1410 * need to take care of the metadata. Once that's done write a dummy
1411 * record to dirty the log in case of a crash while frozen.
1415 struct super_block *sb)
1417 struct xfs_mount *mp = XFS_M(sb);
1419 xfs_save_resvblks(mp);
1420 xfs_quiesce_attr(mp);
1421 return -xfs_fs_log_dummy(mp);
1426 struct super_block *sb)
1428 struct xfs_mount *mp = XFS_M(sb);
1430 xfs_restore_resvblks(mp);
1435 xfs_fs_show_options(
1437 struct vfsmount *mnt)
1439 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1443 * This function fills in xfs_mount_t fields based on mount args.
1444 * Note: the superblock _has_ now been read in.
1448 struct xfs_mount *mp)
1450 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1452 /* Fail a mount where the logbuf is smaller than the log stripe */
1453 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1454 if (mp->m_logbsize <= 0 &&
1455 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1456 mp->m_logbsize = mp->m_sb.sb_logsunit;
1457 } else if (mp->m_logbsize > 0 &&
1458 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1460 "XFS: logbuf size must be greater than or equal to log stripe size");
1461 return XFS_ERROR(EINVAL);
1464 /* Fail a mount if the logbuf is larger than 32K */
1465 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1467 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1468 return XFS_ERROR(EINVAL);
1473 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1474 * told by noattr2 to turn it off
1476 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1477 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1478 mp->m_flags |= XFS_MOUNT_ATTR2;
1481 * prohibit r/w mounts of read-only filesystems
1483 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1485 "XFS: cannot mount a read-only filesystem as read-write");
1486 return XFS_ERROR(EROFS);
1494 struct super_block *sb,
1499 struct xfs_mount *mp = NULL;
1500 int flags = 0, error = ENOMEM;
1502 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1506 spin_lock_init(&mp->m_sb_lock);
1507 mutex_init(&mp->m_growlock);
1508 atomic_set(&mp->m_active_trans, 0);
1509 INIT_LIST_HEAD(&mp->m_sync_list);
1510 spin_lock_init(&mp->m_sync_lock);
1511 init_waitqueue_head(&mp->m_wait_single_sync_task);
1516 error = xfs_parseargs(mp, (char *)data);
1518 goto out_free_fsname;
1520 sb_min_blocksize(sb, BBSIZE);
1521 sb->s_xattr = xfs_xattr_handlers;
1522 sb->s_export_op = &xfs_export_operations;
1523 #ifdef CONFIG_XFS_QUOTA
1524 sb->s_qcop = &xfs_quotactl_operations;
1526 sb->s_op = &xfs_super_operations;
1529 flags |= XFS_MFSI_QUIET;
1531 error = xfs_open_devices(mp);
1533 goto out_free_fsname;
1535 if (xfs_icsb_init_counters(mp))
1536 mp->m_flags |= XFS_MOUNT_NO_PERCPU_SB;
1538 error = xfs_readsb(mp, flags);
1540 goto out_destroy_counters;
1542 error = xfs_finish_flags(mp);
1546 error = xfs_setup_devices(mp);
1550 if (mp->m_flags & XFS_MOUNT_BARRIER)
1551 xfs_mountfs_check_barriers(mp);
1553 error = xfs_filestream_mount(mp);
1557 error = xfs_mountfs(mp);
1559 goto out_filestream_unmount;
1561 sb->s_magic = XFS_SB_MAGIC;
1562 sb->s_blocksize = mp->m_sb.sb_blocksize;
1563 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1564 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1565 sb->s_time_gran = 1;
1566 set_posix_acl_flag(sb);
1568 root = igrab(VFS_I(mp->m_rootip));
1573 if (is_bad_inode(root)) {
1577 sb->s_root = d_alloc_root(root);
1583 error = xfs_syncd_init(mp);
1587 xfs_inode_shrinker_register(mp);
1591 out_filestream_unmount:
1592 xfs_filestream_unmount(mp);
1595 out_destroy_counters:
1596 xfs_icsb_destroy_counters(mp);
1597 xfs_close_devices(mp);
1599 xfs_free_fsname(mp);
1614 * Blow away any referenced inode in the filestreams cache.
1615 * This can and will cause log traffic as inodes go inactive
1618 xfs_filestream_unmount(mp);
1620 XFS_bflush(mp->m_ddev_targp);
1628 struct file_system_type *fs_type,
1630 const char *dev_name,
1632 struct vfsmount *mnt)
1634 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1638 static const struct super_operations xfs_super_operations = {
1639 .alloc_inode = xfs_fs_alloc_inode,
1640 .destroy_inode = xfs_fs_destroy_inode,
1641 .dirty_inode = xfs_fs_dirty_inode,
1642 .write_inode = xfs_fs_write_inode,
1643 .clear_inode = xfs_fs_clear_inode,
1644 .put_super = xfs_fs_put_super,
1645 .sync_fs = xfs_fs_sync_fs,
1646 .freeze_fs = xfs_fs_freeze,
1647 .unfreeze_fs = xfs_fs_unfreeze,
1648 .statfs = xfs_fs_statfs,
1649 .remount_fs = xfs_fs_remount,
1650 .show_options = xfs_fs_show_options,
1653 static struct file_system_type xfs_fs_type = {
1654 .owner = THIS_MODULE,
1656 .get_sb = xfs_fs_get_sb,
1657 .kill_sb = kill_block_super,
1658 .fs_flags = FS_REQUIRES_DEV,
1662 xfs_init_zones(void)
1665 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1666 if (!xfs_ioend_zone)
1669 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1671 if (!xfs_ioend_pool)
1672 goto out_destroy_ioend_zone;
1674 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1676 if (!xfs_log_ticket_zone)
1677 goto out_destroy_ioend_pool;
1679 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1680 "xfs_bmap_free_item");
1681 if (!xfs_bmap_free_item_zone)
1682 goto out_destroy_log_ticket_zone;
1684 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1686 if (!xfs_btree_cur_zone)
1687 goto out_destroy_bmap_free_item_zone;
1689 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1691 if (!xfs_da_state_zone)
1692 goto out_destroy_btree_cur_zone;
1694 xfs_dabuf_zone = kmem_zone_init(sizeof(xfs_dabuf_t), "xfs_dabuf");
1695 if (!xfs_dabuf_zone)
1696 goto out_destroy_da_state_zone;
1698 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1699 if (!xfs_ifork_zone)
1700 goto out_destroy_dabuf_zone;
1702 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1703 if (!xfs_trans_zone)
1704 goto out_destroy_ifork_zone;
1706 xfs_log_item_desc_zone =
1707 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1708 "xfs_log_item_desc");
1709 if (!xfs_log_item_desc_zone)
1710 goto out_destroy_trans_zone;
1713 * The size of the zone allocated buf log item is the maximum
1714 * size possible under XFS. This wastes a little bit of memory,
1715 * but it is much faster.
1717 xfs_buf_item_zone = kmem_zone_init((sizeof(xfs_buf_log_item_t) +
1718 (((XFS_MAX_BLOCKSIZE / XFS_BLF_CHUNK) /
1719 NBWORD) * sizeof(int))), "xfs_buf_item");
1720 if (!xfs_buf_item_zone)
1721 goto out_destroy_log_item_desc_zone;
1723 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1724 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1725 sizeof(xfs_extent_t))), "xfs_efd_item");
1727 goto out_destroy_buf_item_zone;
1729 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1730 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1731 sizeof(xfs_extent_t))), "xfs_efi_item");
1733 goto out_destroy_efd_zone;
1736 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1737 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1738 xfs_fs_inode_init_once);
1739 if (!xfs_inode_zone)
1740 goto out_destroy_efi_zone;
1743 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1744 KM_ZONE_SPREAD, NULL);
1746 goto out_destroy_inode_zone;
1750 out_destroy_inode_zone:
1751 kmem_zone_destroy(xfs_inode_zone);
1752 out_destroy_efi_zone:
1753 kmem_zone_destroy(xfs_efi_zone);
1754 out_destroy_efd_zone:
1755 kmem_zone_destroy(xfs_efd_zone);
1756 out_destroy_buf_item_zone:
1757 kmem_zone_destroy(xfs_buf_item_zone);
1758 out_destroy_log_item_desc_zone:
1759 kmem_zone_destroy(xfs_log_item_desc_zone);
1760 out_destroy_trans_zone:
1761 kmem_zone_destroy(xfs_trans_zone);
1762 out_destroy_ifork_zone:
1763 kmem_zone_destroy(xfs_ifork_zone);
1764 out_destroy_dabuf_zone:
1765 kmem_zone_destroy(xfs_dabuf_zone);
1766 out_destroy_da_state_zone:
1767 kmem_zone_destroy(xfs_da_state_zone);
1768 out_destroy_btree_cur_zone:
1769 kmem_zone_destroy(xfs_btree_cur_zone);
1770 out_destroy_bmap_free_item_zone:
1771 kmem_zone_destroy(xfs_bmap_free_item_zone);
1772 out_destroy_log_ticket_zone:
1773 kmem_zone_destroy(xfs_log_ticket_zone);
1774 out_destroy_ioend_pool:
1775 mempool_destroy(xfs_ioend_pool);
1776 out_destroy_ioend_zone:
1777 kmem_zone_destroy(xfs_ioend_zone);
1783 xfs_destroy_zones(void)
1785 kmem_zone_destroy(xfs_ili_zone);
1786 kmem_zone_destroy(xfs_inode_zone);
1787 kmem_zone_destroy(xfs_efi_zone);
1788 kmem_zone_destroy(xfs_efd_zone);
1789 kmem_zone_destroy(xfs_buf_item_zone);
1790 kmem_zone_destroy(xfs_log_item_desc_zone);
1791 kmem_zone_destroy(xfs_trans_zone);
1792 kmem_zone_destroy(xfs_ifork_zone);
1793 kmem_zone_destroy(xfs_dabuf_zone);
1794 kmem_zone_destroy(xfs_da_state_zone);
1795 kmem_zone_destroy(xfs_btree_cur_zone);
1796 kmem_zone_destroy(xfs_bmap_free_item_zone);
1797 kmem_zone_destroy(xfs_log_ticket_zone);
1798 mempool_destroy(xfs_ioend_pool);
1799 kmem_zone_destroy(xfs_ioend_zone);
1808 printk(KERN_INFO XFS_VERSION_STRING " with "
1809 XFS_BUILD_OPTIONS " enabled\n");
1814 error = xfs_init_zones();
1818 error = xfs_mru_cache_init();
1820 goto out_destroy_zones;
1822 error = xfs_filestream_init();
1824 goto out_mru_cache_uninit;
1826 error = xfs_buf_init();
1828 goto out_filestream_uninit;
1830 error = xfs_init_procfs();
1832 goto out_buf_terminate;
1834 error = xfs_sysctl_register();
1836 goto out_cleanup_procfs;
1840 error = register_filesystem(&xfs_fs_type);
1842 goto out_sysctl_unregister;
1845 out_sysctl_unregister:
1846 xfs_sysctl_unregister();
1848 xfs_cleanup_procfs();
1850 xfs_buf_terminate();
1851 out_filestream_uninit:
1852 xfs_filestream_uninit();
1853 out_mru_cache_uninit:
1854 xfs_mru_cache_uninit();
1856 xfs_destroy_zones();
1865 unregister_filesystem(&xfs_fs_type);
1866 xfs_sysctl_unregister();
1867 xfs_cleanup_procfs();
1868 xfs_buf_terminate();
1869 xfs_filestream_uninit();
1870 xfs_mru_cache_uninit();
1871 xfs_destroy_zones();
1874 module_init(init_xfs_fs);
1875 module_exit(exit_xfs_fs);
1877 MODULE_AUTHOR("Silicon Graphics, Inc.");
1878 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1879 MODULE_LICENSE("GPL");