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_dmapi.h"
29 #include "xfs_quota.h"
30 #include "xfs_mount.h"
31 #include "xfs_bmap_btree.h"
32 #include "xfs_alloc_btree.h"
33 #include "xfs_ialloc_btree.h"
34 #include "xfs_dir2_sf.h"
35 #include "xfs_attr_sf.h"
36 #include "xfs_dinode.h"
37 #include "xfs_inode.h"
38 #include "xfs_btree.h"
39 #include "xfs_btree_trace.h"
40 #include "xfs_ialloc.h"
42 #include "xfs_rtalloc.h"
43 #include "xfs_error.h"
44 #include "xfs_itable.h"
45 #include "xfs_fsops.h"
48 #include "xfs_buf_item.h"
49 #include "xfs_utils.h"
50 #include "xfs_vnodeops.h"
51 #include "xfs_version.h"
52 #include "xfs_log_priv.h"
53 #include "xfs_trans_priv.h"
54 #include "xfs_filestream.h"
55 #include "xfs_da_btree.h"
56 #include "xfs_extfree_item.h"
57 #include "xfs_mru_cache.h"
58 #include "xfs_inode_item.h"
60 #include "xfs_trace.h"
62 #include <linux/namei.h>
63 #include <linux/init.h>
64 #include <linux/mount.h>
65 #include <linux/mempool.h>
66 #include <linux/writeback.h>
67 #include <linux/kthread.h>
68 #include <linux/freezer.h>
69 #include <linux/parser.h>
71 static const struct super_operations xfs_super_operations;
72 static kmem_zone_t *xfs_ioend_zone;
73 mempool_t *xfs_ioend_pool;
75 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
76 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
77 #define MNTOPT_LOGDEV "logdev" /* log device */
78 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
79 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
80 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
81 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
82 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
83 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
84 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
85 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
86 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
87 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
88 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
89 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
90 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
91 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
92 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
93 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
94 * unwritten extent conversion */
95 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
96 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
97 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
98 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
99 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
100 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
101 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
103 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
104 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
105 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
106 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
107 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
108 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
109 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
110 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
111 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
112 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
113 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
114 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
115 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
116 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
117 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
118 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
119 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
120 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
123 * Table driven mount option parser.
125 * Currently only used for remount, but it will be used for mount
126 * in the future, too.
129 Opt_barrier, Opt_nobarrier, Opt_err
132 static const match_table_t tokens = {
133 {Opt_barrier, "barrier"},
134 {Opt_nobarrier, "nobarrier"},
140 suffix_strtoul(char *s, char **endp, unsigned int base)
142 int last, shift_left_factor = 0;
145 last = strlen(value) - 1;
146 if (value[last] == 'K' || value[last] == 'k') {
147 shift_left_factor = 10;
150 if (value[last] == 'M' || value[last] == 'm') {
151 shift_left_factor = 20;
154 if (value[last] == 'G' || value[last] == 'g') {
155 shift_left_factor = 30;
159 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
163 * This function fills in xfs_mount_t fields based on mount args.
164 * Note: the superblock has _not_ yet been read in.
166 * Note that this function leaks the various device name allocations on
167 * failure. The caller takes care of them.
171 struct xfs_mount *mp,
175 struct super_block *sb = mp->m_super;
176 char *this_char, *value, *eov;
180 int dmapi_implies_ikeep = 1;
181 __uint8_t iosizelog = 0;
184 * Copy binary VFS mount flags we are interested in.
186 if (sb->s_flags & MS_RDONLY)
187 mp->m_flags |= XFS_MOUNT_RDONLY;
188 if (sb->s_flags & MS_DIRSYNC)
189 mp->m_flags |= XFS_MOUNT_DIRSYNC;
190 if (sb->s_flags & MS_SYNCHRONOUS)
191 mp->m_flags |= XFS_MOUNT_WSYNC;
194 * Set some default flags that could be cleared by the mount option
197 mp->m_flags |= XFS_MOUNT_BARRIER;
198 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
199 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
202 * These can be overridden by the mount option parsing.
210 while ((this_char = strsep(&options, ",")) != NULL) {
213 if ((value = strchr(this_char, '=')) != NULL)
216 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
217 if (!value || !*value) {
219 "XFS: %s option requires an argument",
223 mp->m_logbufs = simple_strtoul(value, &eov, 10);
224 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
225 if (!value || !*value) {
227 "XFS: %s option requires an argument",
231 mp->m_logbsize = suffix_strtoul(value, &eov, 10);
232 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
233 if (!value || !*value) {
235 "XFS: %s option requires an argument",
239 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
242 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
243 if (!value || !*value) {
245 "XFS: %s option requires an argument",
249 *mtpt = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
252 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
253 if (!value || !*value) {
255 "XFS: %s option requires an argument",
259 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
262 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
263 if (!value || !*value) {
265 "XFS: %s option requires an argument",
269 iosize = simple_strtoul(value, &eov, 10);
270 iosizelog = ffs(iosize) - 1;
271 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
272 if (!value || !*value) {
274 "XFS: %s option requires an argument",
278 iosize = suffix_strtoul(value, &eov, 10);
279 iosizelog = ffs(iosize) - 1;
280 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
281 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
282 mp->m_flags |= XFS_MOUNT_GRPID;
283 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
284 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
285 mp->m_flags &= ~XFS_MOUNT_GRPID;
286 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
287 mp->m_flags |= XFS_MOUNT_WSYNC;
288 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
289 mp->m_flags |= XFS_MOUNT_OSYNCISOSYNC;
290 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
291 mp->m_flags |= XFS_MOUNT_NORECOVERY;
292 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
293 mp->m_flags |= XFS_MOUNT_NOALIGN;
294 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
295 mp->m_flags |= XFS_MOUNT_SWALLOC;
296 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
297 if (!value || !*value) {
299 "XFS: %s option requires an argument",
303 dsunit = simple_strtoul(value, &eov, 10);
304 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
305 if (!value || !*value) {
307 "XFS: %s option requires an argument",
311 dswidth = simple_strtoul(value, &eov, 10);
312 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
313 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
316 "XFS: %s option not allowed on this system",
320 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
321 mp->m_flags |= XFS_MOUNT_NOUUID;
322 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
323 mp->m_flags |= XFS_MOUNT_BARRIER;
324 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
325 mp->m_flags &= ~XFS_MOUNT_BARRIER;
326 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
327 mp->m_flags |= XFS_MOUNT_IKEEP;
328 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
329 dmapi_implies_ikeep = 0;
330 mp->m_flags &= ~XFS_MOUNT_IKEEP;
331 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
332 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
333 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
334 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
335 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
336 mp->m_flags |= XFS_MOUNT_ATTR2;
337 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
338 mp->m_flags &= ~XFS_MOUNT_ATTR2;
339 mp->m_flags |= XFS_MOUNT_NOATTR2;
340 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
341 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
342 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
343 mp->m_qflags &= ~(XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
344 XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
345 XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
346 XFS_UQUOTA_ENFD | XFS_OQUOTA_ENFD);
347 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
348 !strcmp(this_char, MNTOPT_UQUOTA) ||
349 !strcmp(this_char, MNTOPT_USRQUOTA)) {
350 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
352 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
353 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
354 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
355 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
356 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
357 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
358 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
360 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
361 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
362 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
363 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
364 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
365 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
367 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
368 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
369 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
370 } else if (!strcmp(this_char, MNTOPT_DMAPI)) {
371 mp->m_flags |= XFS_MOUNT_DMAPI;
372 } else if (!strcmp(this_char, MNTOPT_XDSM)) {
373 mp->m_flags |= XFS_MOUNT_DMAPI;
374 } else if (!strcmp(this_char, MNTOPT_DMI)) {
375 mp->m_flags |= XFS_MOUNT_DMAPI;
376 } else if (!strcmp(this_char, "ihashsize")) {
378 "XFS: ihashsize no longer used, option is deprecated.");
379 } else if (!strcmp(this_char, "osyncisdsync")) {
380 /* no-op, this is now the default */
382 "XFS: osyncisdsync is now the default, option is deprecated.");
383 } else if (!strcmp(this_char, "irixsgid")) {
385 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
388 "XFS: unknown mount option [%s].", this_char);
394 * no recovery flag requires a read-only mount
396 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
397 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
398 cmn_err(CE_WARN, "XFS: no-recovery mounts must be read-only.");
402 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
404 "XFS: sunit and swidth options incompatible with the noalign option");
408 #ifndef CONFIG_XFS_QUOTA
409 if (XFS_IS_QUOTA_RUNNING(mp)) {
411 "XFS: quota support not available in this kernel.");
416 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
417 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE))) {
419 "XFS: cannot mount with both project and group quota");
423 if ((mp->m_flags & XFS_MOUNT_DMAPI) && (!*mtpt || *mtpt[0] == '\0')) {
424 printk("XFS: %s option needs the mount point option as well\n",
429 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
431 "XFS: sunit and swidth must be specified together");
435 if (dsunit && (dswidth % dsunit != 0)) {
437 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
443 * Applications using DMI filesystems often expect the
444 * inode generation number to be monotonically increasing.
445 * If we delete inode chunks we break this assumption, so
446 * keep unused inode chunks on disk for DMI filesystems
447 * until we come up with a better solution.
448 * Note that if "ikeep" or "noikeep" mount options are
449 * supplied, then they are honored.
451 if ((mp->m_flags & XFS_MOUNT_DMAPI) && dmapi_implies_ikeep)
452 mp->m_flags |= XFS_MOUNT_IKEEP;
455 if (!(mp->m_flags & XFS_MOUNT_NOALIGN)) {
457 * At this point the superblock has not been read
458 * in, therefore we do not know the block size.
459 * Before the mount call ends we will convert
463 mp->m_dalign = dsunit;
464 mp->m_flags |= XFS_MOUNT_RETERR;
468 mp->m_swidth = dswidth;
471 if (mp->m_logbufs != -1 &&
472 mp->m_logbufs != 0 &&
473 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
474 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
476 "XFS: invalid logbufs value: %d [not %d-%d]",
477 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
478 return XFS_ERROR(EINVAL);
480 if (mp->m_logbsize != -1 &&
481 mp->m_logbsize != 0 &&
482 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
483 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
484 !is_power_of_2(mp->m_logbsize))) {
486 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
488 return XFS_ERROR(EINVAL);
491 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
494 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
497 if (iosizelog > XFS_MAX_IO_LOG ||
498 iosizelog < XFS_MIN_IO_LOG) {
500 "XFS: invalid log iosize: %d [not %d-%d]",
501 iosizelog, XFS_MIN_IO_LOG,
503 return XFS_ERROR(EINVAL);
506 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
507 mp->m_readio_log = iosizelog;
508 mp->m_writeio_log = iosizelog;
514 struct proc_xfs_info {
521 struct xfs_mount *mp,
524 static struct proc_xfs_info xfs_info_set[] = {
525 /* the few simple ones we can get from the mount struct */
526 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
527 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
528 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
529 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
530 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
531 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
532 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
533 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
534 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
535 { XFS_MOUNT_DMAPI, "," MNTOPT_DMAPI },
536 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
539 static struct proc_xfs_info xfs_info_unset[] = {
540 /* the few simple ones we can get from the mount struct */
541 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
542 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
543 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
546 struct proc_xfs_info *xfs_infop;
548 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
549 if (mp->m_flags & xfs_infop->flag)
550 seq_puts(m, xfs_infop->str);
552 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
553 if (!(mp->m_flags & xfs_infop->flag))
554 seq_puts(m, xfs_infop->str);
557 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
558 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
559 (int)(1 << mp->m_writeio_log) >> 10);
561 if (mp->m_logbufs > 0)
562 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
563 if (mp->m_logbsize > 0)
564 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
567 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
569 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
571 if (mp->m_dalign > 0)
572 seq_printf(m, "," MNTOPT_SUNIT "=%d",
573 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
574 if (mp->m_swidth > 0)
575 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
576 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
578 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
579 seq_puts(m, "," MNTOPT_USRQUOTA);
580 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
581 seq_puts(m, "," MNTOPT_UQUOTANOENF);
583 /* Either project or group quotas can be active, not both */
585 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
586 if (mp->m_qflags & XFS_OQUOTA_ENFD)
587 seq_puts(m, "," MNTOPT_PRJQUOTA);
589 seq_puts(m, "," MNTOPT_PQUOTANOENF);
590 } else if (mp->m_qflags & XFS_GQUOTA_ACCT) {
591 if (mp->m_qflags & XFS_OQUOTA_ENFD)
592 seq_puts(m, "," MNTOPT_GRPQUOTA);
594 seq_puts(m, "," MNTOPT_GQUOTANOENF);
597 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
598 seq_puts(m, "," MNTOPT_NOQUOTA);
604 unsigned int blockshift)
606 unsigned int pagefactor = 1;
607 unsigned int bitshift = BITS_PER_LONG - 1;
609 /* Figure out maximum filesize, on Linux this can depend on
610 * the filesystem blocksize (on 32 bit platforms).
611 * __block_prepare_write does this in an [unsigned] long...
612 * page->index << (PAGE_CACHE_SHIFT - bbits)
613 * So, for page sized blocks (4K on 32 bit platforms),
614 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
615 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
616 * but for smaller blocksizes it is less (bbits = log2 bsize).
617 * Note1: get_block_t takes a long (implicit cast from above)
618 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
619 * can optionally convert the [unsigned] long from above into
620 * an [unsigned] long long.
623 #if BITS_PER_LONG == 32
624 # if defined(CONFIG_LBDAF)
625 ASSERT(sizeof(sector_t) == 8);
626 pagefactor = PAGE_CACHE_SIZE;
627 bitshift = BITS_PER_LONG;
629 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
633 return (((__uint64_t)pagefactor) << bitshift) - 1;
640 struct block_device **bdevp)
644 *bdevp = open_bdev_exclusive(name, FMODE_READ|FMODE_WRITE, mp);
645 if (IS_ERR(*bdevp)) {
646 error = PTR_ERR(*bdevp);
647 printk("XFS: Invalid device [%s], error=%d\n", name, error);
655 struct block_device *bdev)
658 close_bdev_exclusive(bdev, FMODE_READ|FMODE_WRITE);
662 * Try to write out the superblock using barriers.
668 xfs_buf_t *sbp = xfs_getsb(mp, 0);
673 XFS_BUF_UNDELAYWRITE(sbp);
675 XFS_BUF_UNASYNC(sbp);
676 XFS_BUF_ORDERED(sbp);
679 error = xfs_iowait(sbp);
682 * Clear all the flags we set and possible error state in the
683 * buffer. We only did the write to try out whether barriers
684 * worked and shouldn't leave any traces in the superblock
688 XFS_BUF_ERROR(sbp, 0);
689 XFS_BUF_UNORDERED(sbp);
696 xfs_mountfs_check_barriers(xfs_mount_t *mp)
700 if (mp->m_logdev_targp != mp->m_ddev_targp) {
701 xfs_fs_cmn_err(CE_NOTE, mp,
702 "Disabling barriers, not supported with external log device");
703 mp->m_flags &= ~XFS_MOUNT_BARRIER;
707 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
708 xfs_fs_cmn_err(CE_NOTE, mp,
709 "Disabling barriers, underlying device is readonly");
710 mp->m_flags &= ~XFS_MOUNT_BARRIER;
714 error = xfs_barrier_test(mp);
716 xfs_fs_cmn_err(CE_NOTE, mp,
717 "Disabling barriers, trial barrier write failed");
718 mp->m_flags &= ~XFS_MOUNT_BARRIER;
724 xfs_blkdev_issue_flush(
725 xfs_buftarg_t *buftarg)
727 blkdev_issue_flush(buftarg->bt_bdev, NULL);
732 struct xfs_mount *mp)
734 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
735 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
736 xfs_free_buftarg(mp, mp->m_logdev_targp);
737 xfs_blkdev_put(logdev);
739 if (mp->m_rtdev_targp) {
740 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
741 xfs_free_buftarg(mp, mp->m_rtdev_targp);
742 xfs_blkdev_put(rtdev);
744 xfs_free_buftarg(mp, mp->m_ddev_targp);
748 * The file system configurations are:
749 * (1) device (partition) with data and internal log
750 * (2) logical volume with data and log subvolumes.
751 * (3) logical volume with data, log, and realtime subvolumes.
753 * We only have to handle opening the log and realtime volumes here if
754 * they are present. The data subvolume has already been opened by
755 * get_sb_bdev() and is stored in sb->s_bdev.
759 struct xfs_mount *mp)
761 struct block_device *ddev = mp->m_super->s_bdev;
762 struct block_device *logdev = NULL, *rtdev = NULL;
766 * Open real time and log devices - order is important.
769 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
775 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
777 goto out_close_logdev;
779 if (rtdev == ddev || rtdev == logdev) {
781 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
783 goto out_close_rtdev;
788 * Setup xfs_mount buffer target pointers
791 mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0);
792 if (!mp->m_ddev_targp)
793 goto out_close_rtdev;
796 mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1);
797 if (!mp->m_rtdev_targp)
798 goto out_free_ddev_targ;
801 if (logdev && logdev != ddev) {
802 mp->m_logdev_targp = xfs_alloc_buftarg(logdev, 1);
803 if (!mp->m_logdev_targp)
804 goto out_free_rtdev_targ;
806 mp->m_logdev_targp = mp->m_ddev_targp;
812 if (mp->m_rtdev_targp)
813 xfs_free_buftarg(mp, mp->m_rtdev_targp);
815 xfs_free_buftarg(mp, mp->m_ddev_targp);
818 xfs_blkdev_put(rtdev);
820 if (logdev && logdev != ddev)
821 xfs_blkdev_put(logdev);
827 * Setup xfs_mount buffer target pointers based on superblock
831 struct xfs_mount *mp)
835 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
836 mp->m_sb.sb_sectsize);
840 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
841 unsigned int log_sector_size = BBSIZE;
843 if (xfs_sb_version_hassector(&mp->m_sb))
844 log_sector_size = mp->m_sb.sb_logsectsize;
845 error = xfs_setsize_buftarg(mp->m_logdev_targp,
846 mp->m_sb.sb_blocksize,
851 if (mp->m_rtdev_targp) {
852 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
853 mp->m_sb.sb_blocksize,
854 mp->m_sb.sb_sectsize);
863 * XFS AIL push thread support
867 struct xfs_ail *ailp,
868 xfs_lsn_t threshold_lsn)
870 ailp->xa_target = threshold_lsn;
871 wake_up_process(ailp->xa_task);
878 struct xfs_ail *ailp = data;
879 xfs_lsn_t last_pushed_lsn = 0;
880 long tout = 0; /* milliseconds */
882 while (!kthread_should_stop()) {
883 schedule_timeout_interruptible(tout ?
884 msecs_to_jiffies(tout) : MAX_SCHEDULE_TIMEOUT);
889 ASSERT(ailp->xa_mount->m_log);
890 if (XFS_FORCED_SHUTDOWN(ailp->xa_mount))
893 tout = xfsaild_push(ailp, &last_pushed_lsn);
901 struct xfs_ail *ailp)
904 ailp->xa_task = kthread_run(xfsaild, ailp, "xfsaild");
905 if (IS_ERR(ailp->xa_task))
906 return -PTR_ERR(ailp->xa_task);
912 struct xfs_ail *ailp)
914 kthread_stop(ailp->xa_task);
918 /* Catch misguided souls that try to use this interface on XFS */
919 STATIC struct inode *
921 struct super_block *sb)
928 * Now that the generic code is guaranteed not to be accessing
929 * the linux inode, we can reclaim the inode.
932 xfs_fs_destroy_inode(
935 struct xfs_inode *ip = XFS_I(inode);
937 xfs_itrace_entry(ip);
939 XFS_STATS_INC(vn_reclaim);
941 /* bad inode, get out here ASAP */
942 if (is_bad_inode(inode))
947 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
950 * We should never get here with one of the reclaim flags already set.
952 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
953 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
956 * We always use background reclaim here because even if the
957 * inode is clean, it still may be under IO and hence we have
958 * to take the flush lock. The background reclaim path handles
959 * this more efficiently than we can here, so simply let background
960 * reclaim tear down all inodes.
963 xfs_inode_set_reclaim_tag(ip);
967 * Slab object creation initialisation for the XFS inode.
968 * This covers only the idempotent fields in the XFS inode;
969 * all other fields need to be initialised on allocation
970 * from the slab. This avoids the need to repeatedly intialise
971 * fields in the xfs inode that left in the initialise state
972 * when freeing the inode.
975 xfs_fs_inode_init_once(
978 struct xfs_inode *ip = inode;
980 memset(ip, 0, sizeof(struct xfs_inode));
983 inode_init_once(VFS_I(ip));
986 atomic_set(&ip->i_iocount, 0);
987 atomic_set(&ip->i_pincount, 0);
988 spin_lock_init(&ip->i_flags_lock);
989 init_waitqueue_head(&ip->i_ipin_wait);
991 * Because we want to use a counting completion, complete
992 * the flush completion once to allow a single access to
993 * the flush completion without blocking.
995 init_completion(&ip->i_flush);
996 complete(&ip->i_flush);
998 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
999 "xfsino", ip->i_ino);
1003 * Dirty the XFS inode when mark_inode_dirty_sync() is called so that
1004 * we catch unlogged VFS level updates to the inode. Care must be taken
1005 * here - the transaction code calls mark_inode_dirty_sync() to mark the
1006 * VFS inode dirty in a transaction and clears the i_update_core field;
1007 * it must clear the field after calling mark_inode_dirty_sync() to
1008 * correctly indicate that the dirty state has been propagated into the
1011 * We need the barrier() to maintain correct ordering between unlogged
1012 * updates and the transaction commit code that clears the i_update_core
1013 * field. This requires all updates to be completed before marking the
1018 struct inode *inode)
1021 XFS_I(inode)->i_update_core = 1;
1026 struct xfs_inode *ip)
1028 struct xfs_mount *mp = ip->i_mount;
1029 struct xfs_trans *tp;
1032 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1033 tp = xfs_trans_alloc(mp, XFS_TRANS_FSYNC_TS);
1034 error = xfs_trans_reserve(tp, 0, XFS_FSYNC_TS_LOG_RES(mp), 0, 0, 0);
1037 xfs_trans_cancel(tp, 0);
1038 /* we need to return with the lock hold shared */
1039 xfs_ilock(ip, XFS_ILOCK_SHARED);
1043 xfs_ilock(ip, XFS_ILOCK_EXCL);
1046 * Note - it's possible that we might have pushed ourselves out of the
1047 * way during trans_reserve which would flush the inode. But there's
1048 * no guarantee that the inode buffer has actually gone out yet (it's
1049 * delwri). Plus the buffer could be pinned anyway if it's part of
1050 * an inode in another recent transaction. So we play it safe and
1051 * fire off the transaction anyway.
1053 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
1054 xfs_trans_ihold(tp, ip);
1055 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1056 xfs_trans_set_sync(tp);
1057 error = xfs_trans_commit(tp, 0);
1058 xfs_ilock_demote(ip, XFS_ILOCK_EXCL);
1065 struct inode *inode,
1066 struct writeback_control *wbc)
1068 struct xfs_inode *ip = XFS_I(inode);
1069 struct xfs_mount *mp = ip->i_mount;
1072 xfs_itrace_entry(ip);
1074 if (XFS_FORCED_SHUTDOWN(mp))
1075 return XFS_ERROR(EIO);
1077 if (wbc->sync_mode == WB_SYNC_ALL) {
1079 * Make sure the inode has hit stable storage. By using the
1080 * log and the fsync transactions we reduce the IOs we have
1081 * to do here from two (log and inode) to just the log.
1083 * Note: We still need to do a delwri write of the inode after
1084 * this to flush it to the backing buffer so that bulkstat
1085 * works properly if this is the first time the inode has been
1086 * written. Because we hold the ilock atomically over the
1087 * transaction commit and the inode flush we are guaranteed
1088 * that the inode is not pinned when it returns. If the flush
1089 * lock is already held, then the inode has already been
1090 * flushed once and we don't need to flush it again. Hence
1091 * the code will only flush the inode if it isn't already
1094 xfs_ilock(ip, XFS_ILOCK_SHARED);
1095 if (ip->i_update_core) {
1096 error = xfs_log_inode(ip);
1102 * We make this non-blocking if the inode is contended, return
1103 * EAGAIN to indicate to the caller that they did not succeed.
1104 * This prevents the flush path from blocking on inodes inside
1105 * another operation right now, they get caught later by xfs_sync.
1107 if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED))
1111 if (xfs_ipincount(ip) || !xfs_iflock_nowait(ip))
1115 * Now we have the flush lock and the inode is not pinned, we can check
1116 * if the inode is really clean as we know that there are no pending
1117 * transaction completions, it is not waiting on the delayed write
1118 * queue and there is no IO in progress.
1120 if (xfs_inode_clean(ip)) {
1125 error = xfs_iflush(ip, 0);
1128 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1131 * if we failed to write out the inode then mark
1132 * it dirty again so we'll try again later.
1135 xfs_mark_inode_dirty_sync(ip);
1141 struct inode *inode)
1143 xfs_inode_t *ip = XFS_I(inode);
1145 xfs_itrace_entry(ip);
1146 XFS_STATS_INC(vn_rele);
1147 XFS_STATS_INC(vn_remove);
1148 XFS_STATS_DEC(vn_active);
1151 * The iolock is used by the file system to coordinate reads,
1152 * writes, and block truncates. Up to this point the lock
1153 * protected concurrent accesses by users of the inode. But
1154 * from here forward we're doing some final processing of the
1155 * inode because we're done with it, and although we reuse the
1156 * iolock for protection it is really a distinct lock class
1157 * (in the lockdep sense) from before. To keep lockdep happy
1158 * (and basically indicate what we are doing), we explicitly
1159 * re-init the iolock here.
1161 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
1162 mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
1169 struct xfs_mount *mp)
1171 kfree(mp->m_fsname);
1172 kfree(mp->m_rtname);
1173 kfree(mp->m_logname);
1178 struct super_block *sb)
1180 struct xfs_mount *mp = XFS_M(sb);
1184 if (!(sb->s_flags & MS_RDONLY)) {
1186 * XXX(hch): this should be SYNC_WAIT.
1188 * Or more likely not needed at all because the VFS is already
1189 * calling ->sync_fs after shutting down all filestem
1190 * operations and just before calling ->put_super.
1192 xfs_sync_data(mp, 0);
1193 xfs_sync_attr(mp, 0);
1196 XFS_SEND_PREUNMOUNT(mp);
1199 * Blow away any referenced inode in the filestreams cache.
1200 * This can and will cause log traffic as inodes go inactive
1203 xfs_filestream_unmount(mp);
1205 XFS_bflush(mp->m_ddev_targp);
1207 XFS_SEND_UNMOUNT(mp);
1211 xfs_icsb_destroy_counters(mp);
1212 xfs_close_devices(mp);
1214 xfs_free_fsname(mp);
1220 struct super_block *sb,
1223 struct xfs_mount *mp = XFS_M(sb);
1227 * Not much we can do for the first async pass. Writing out the
1228 * superblock would be counter-productive as we are going to redirty
1229 * when writing out other data and metadata (and writing out a single
1230 * block is quite fast anyway).
1232 * Try to asynchronously kick off quota syncing at least.
1235 xfs_qm_sync(mp, SYNC_TRYLOCK);
1239 error = xfs_quiesce_data(mp);
1244 int prev_sync_seq = mp->m_sync_seq;
1247 * The disk must be active because we're syncing.
1248 * We schedule xfssyncd now (now that the disk is
1249 * active) instead of later (when it might not be).
1251 wake_up_process(mp->m_sync_task);
1253 * We have to wait for the sync iteration to complete.
1254 * If we don't, the disk activity caused by the sync
1255 * will come after the sync is completed, and that
1256 * triggers another sync from laptop mode.
1258 wait_event(mp->m_wait_single_sync_task,
1259 mp->m_sync_seq != prev_sync_seq);
1267 struct dentry *dentry,
1268 struct kstatfs *statp)
1270 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1271 xfs_sb_t *sbp = &mp->m_sb;
1272 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1273 __uint64_t fakeinos, id;
1276 statp->f_type = XFS_SB_MAGIC;
1277 statp->f_namelen = MAXNAMELEN - 1;
1279 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1280 statp->f_fsid.val[0] = (u32)id;
1281 statp->f_fsid.val[1] = (u32)(id >> 32);
1283 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1285 spin_lock(&mp->m_sb_lock);
1286 statp->f_bsize = sbp->sb_blocksize;
1287 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1288 statp->f_blocks = sbp->sb_dblocks - lsize;
1289 statp->f_bfree = statp->f_bavail =
1290 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1291 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1293 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1294 if (mp->m_maxicount)
1295 statp->f_files = min_t(typeof(statp->f_files),
1298 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1299 spin_unlock(&mp->m_sb_lock);
1301 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) ||
1302 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))) ==
1303 (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
1304 xfs_qm_statvfs(ip, statp);
1309 xfs_save_resvblks(struct xfs_mount *mp)
1311 __uint64_t resblks = 0;
1313 mp->m_resblks_save = mp->m_resblks;
1314 xfs_reserve_blocks(mp, &resblks, NULL);
1318 xfs_restore_resvblks(struct xfs_mount *mp)
1322 if (mp->m_resblks_save) {
1323 resblks = mp->m_resblks_save;
1324 mp->m_resblks_save = 0;
1326 resblks = xfs_default_resblks(mp);
1328 xfs_reserve_blocks(mp, &resblks, NULL);
1333 struct super_block *sb,
1337 struct xfs_mount *mp = XFS_M(sb);
1338 substring_t args[MAX_OPT_ARGS];
1342 while ((p = strsep(&options, ",")) != NULL) {
1348 token = match_token(p, tokens, args);
1351 mp->m_flags |= XFS_MOUNT_BARRIER;
1354 * Test if barriers are actually working if we can,
1355 * else delay this check until the filesystem is
1358 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1359 xfs_mountfs_check_barriers(mp);
1362 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1366 * Logically we would return an error here to prevent
1367 * users from believing they might have changed
1368 * mount options using remount which can't be changed.
1370 * But unfortunately mount(8) adds all options from
1371 * mtab and fstab to the mount arguments in some cases
1372 * so we can't blindly reject options, but have to
1373 * check for each specified option if it actually
1374 * differs from the currently set option and only
1375 * reject it if that's the case.
1377 * Until that is implemented we return success for
1378 * every remount request, and silently ignore all
1379 * options that we can't actually change.
1383 "XFS: mount option \"%s\" not supported for remount\n", p);
1392 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1393 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1394 if (mp->m_flags & XFS_MOUNT_BARRIER)
1395 xfs_mountfs_check_barriers(mp);
1398 * If this is the first remount to writeable state we
1399 * might have some superblock changes to update.
1401 if (mp->m_update_flags) {
1402 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1405 "XFS: failed to write sb changes");
1408 mp->m_update_flags = 0;
1412 * Fill out the reserve pool if it is empty. Use the stashed
1413 * value if it is non-zero, otherwise go with the default.
1415 xfs_restore_resvblks(mp);
1419 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1421 * After we have synced the data but before we sync the
1422 * metadata, we need to free up the reserve block pool so that
1423 * the used block count in the superblock on disk is correct at
1424 * the end of the remount. Stash the current reserve pool size
1425 * so that if we get remounted rw, we can return it to the same
1429 xfs_quiesce_data(mp);
1430 xfs_save_resvblks(mp);
1431 xfs_quiesce_attr(mp);
1432 mp->m_flags |= XFS_MOUNT_RDONLY;
1439 * Second stage of a freeze. The data is already frozen so we only
1440 * need to take care of the metadata. Once that's done write a dummy
1441 * record to dirty the log in case of a crash while frozen.
1445 struct super_block *sb)
1447 struct xfs_mount *mp = XFS_M(sb);
1449 xfs_save_resvblks(mp);
1450 xfs_quiesce_attr(mp);
1451 return -xfs_fs_log_dummy(mp);
1456 struct super_block *sb)
1458 struct xfs_mount *mp = XFS_M(sb);
1460 xfs_restore_resvblks(mp);
1465 xfs_fs_show_options(
1467 struct vfsmount *mnt)
1469 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1473 * This function fills in xfs_mount_t fields based on mount args.
1474 * Note: the superblock _has_ now been read in.
1478 struct xfs_mount *mp)
1480 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1482 /* Fail a mount where the logbuf is smaller than the log stripe */
1483 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1484 if (mp->m_logbsize <= 0 &&
1485 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1486 mp->m_logbsize = mp->m_sb.sb_logsunit;
1487 } else if (mp->m_logbsize > 0 &&
1488 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1490 "XFS: logbuf size must be greater than or equal to log stripe size");
1491 return XFS_ERROR(EINVAL);
1494 /* Fail a mount if the logbuf is larger than 32K */
1495 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1497 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1498 return XFS_ERROR(EINVAL);
1503 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1504 * told by noattr2 to turn it off
1506 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1507 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1508 mp->m_flags |= XFS_MOUNT_ATTR2;
1511 * prohibit r/w mounts of read-only filesystems
1513 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1515 "XFS: cannot mount a read-only filesystem as read-write");
1516 return XFS_ERROR(EROFS);
1524 struct super_block *sb,
1529 struct xfs_mount *mp = NULL;
1530 int flags = 0, error = ENOMEM;
1533 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1537 spin_lock_init(&mp->m_sb_lock);
1538 mutex_init(&mp->m_growlock);
1539 atomic_set(&mp->m_active_trans, 0);
1540 INIT_LIST_HEAD(&mp->m_sync_list);
1541 spin_lock_init(&mp->m_sync_lock);
1542 init_waitqueue_head(&mp->m_wait_single_sync_task);
1547 error = xfs_parseargs(mp, (char *)data, &mtpt);
1549 goto out_free_fsname;
1551 sb_min_blocksize(sb, BBSIZE);
1552 sb->s_xattr = xfs_xattr_handlers;
1553 sb->s_export_op = &xfs_export_operations;
1554 #ifdef CONFIG_XFS_QUOTA
1555 sb->s_qcop = &xfs_quotactl_operations;
1557 sb->s_op = &xfs_super_operations;
1559 error = xfs_dmops_get(mp);
1561 goto out_free_fsname;
1564 flags |= XFS_MFSI_QUIET;
1566 error = xfs_open_devices(mp);
1570 if (xfs_icsb_init_counters(mp))
1571 mp->m_flags |= XFS_MOUNT_NO_PERCPU_SB;
1573 error = xfs_readsb(mp, flags);
1575 goto out_destroy_counters;
1577 error = xfs_finish_flags(mp);
1581 error = xfs_setup_devices(mp);
1585 if (mp->m_flags & XFS_MOUNT_BARRIER)
1586 xfs_mountfs_check_barriers(mp);
1588 error = xfs_filestream_mount(mp);
1592 error = xfs_mountfs(mp);
1594 goto out_filestream_unmount;
1596 XFS_SEND_MOUNT(mp, DM_RIGHT_NULL, mtpt, mp->m_fsname);
1598 sb->s_magic = XFS_SB_MAGIC;
1599 sb->s_blocksize = mp->m_sb.sb_blocksize;
1600 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1601 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1602 sb->s_time_gran = 1;
1603 set_posix_acl_flag(sb);
1605 root = igrab(VFS_I(mp->m_rootip));
1610 if (is_bad_inode(root)) {
1614 sb->s_root = d_alloc_root(root);
1620 error = xfs_syncd_init(mp);
1627 out_filestream_unmount:
1628 xfs_filestream_unmount(mp);
1631 out_destroy_counters:
1632 xfs_icsb_destroy_counters(mp);
1633 xfs_close_devices(mp);
1637 xfs_free_fsname(mp);
1653 * Blow away any referenced inode in the filestreams cache.
1654 * This can and will cause log traffic as inodes go inactive
1657 xfs_filestream_unmount(mp);
1659 XFS_bflush(mp->m_ddev_targp);
1667 struct file_system_type *fs_type,
1669 const char *dev_name,
1671 struct vfsmount *mnt)
1673 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1677 static const struct super_operations xfs_super_operations = {
1678 .alloc_inode = xfs_fs_alloc_inode,
1679 .destroy_inode = xfs_fs_destroy_inode,
1680 .dirty_inode = xfs_fs_dirty_inode,
1681 .write_inode = xfs_fs_write_inode,
1682 .clear_inode = xfs_fs_clear_inode,
1683 .put_super = xfs_fs_put_super,
1684 .sync_fs = xfs_fs_sync_fs,
1685 .freeze_fs = xfs_fs_freeze,
1686 .unfreeze_fs = xfs_fs_unfreeze,
1687 .statfs = xfs_fs_statfs,
1688 .remount_fs = xfs_fs_remount,
1689 .show_options = xfs_fs_show_options,
1692 static struct file_system_type xfs_fs_type = {
1693 .owner = THIS_MODULE,
1695 .get_sb = xfs_fs_get_sb,
1696 .kill_sb = kill_block_super,
1697 .fs_flags = FS_REQUIRES_DEV,
1701 xfs_init_zones(void)
1704 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1705 if (!xfs_ioend_zone)
1708 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1710 if (!xfs_ioend_pool)
1711 goto out_destroy_ioend_zone;
1713 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1715 if (!xfs_log_ticket_zone)
1716 goto out_destroy_ioend_pool;
1718 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1719 "xfs_bmap_free_item");
1720 if (!xfs_bmap_free_item_zone)
1721 goto out_destroy_log_ticket_zone;
1723 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1725 if (!xfs_btree_cur_zone)
1726 goto out_destroy_bmap_free_item_zone;
1728 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1730 if (!xfs_da_state_zone)
1731 goto out_destroy_btree_cur_zone;
1733 xfs_dabuf_zone = kmem_zone_init(sizeof(xfs_dabuf_t), "xfs_dabuf");
1734 if (!xfs_dabuf_zone)
1735 goto out_destroy_da_state_zone;
1737 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1738 if (!xfs_ifork_zone)
1739 goto out_destroy_dabuf_zone;
1741 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1742 if (!xfs_trans_zone)
1743 goto out_destroy_ifork_zone;
1746 * The size of the zone allocated buf log item is the maximum
1747 * size possible under XFS. This wastes a little bit of memory,
1748 * but it is much faster.
1750 xfs_buf_item_zone = kmem_zone_init((sizeof(xfs_buf_log_item_t) +
1751 (((XFS_MAX_BLOCKSIZE / XFS_BLI_CHUNK) /
1752 NBWORD) * sizeof(int))), "xfs_buf_item");
1753 if (!xfs_buf_item_zone)
1754 goto out_destroy_trans_zone;
1756 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1757 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1758 sizeof(xfs_extent_t))), "xfs_efd_item");
1760 goto out_destroy_buf_item_zone;
1762 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1763 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1764 sizeof(xfs_extent_t))), "xfs_efi_item");
1766 goto out_destroy_efd_zone;
1769 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1770 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1771 xfs_fs_inode_init_once);
1772 if (!xfs_inode_zone)
1773 goto out_destroy_efi_zone;
1776 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1777 KM_ZONE_SPREAD, NULL);
1779 goto out_destroy_inode_zone;
1783 out_destroy_inode_zone:
1784 kmem_zone_destroy(xfs_inode_zone);
1785 out_destroy_efi_zone:
1786 kmem_zone_destroy(xfs_efi_zone);
1787 out_destroy_efd_zone:
1788 kmem_zone_destroy(xfs_efd_zone);
1789 out_destroy_buf_item_zone:
1790 kmem_zone_destroy(xfs_buf_item_zone);
1791 out_destroy_trans_zone:
1792 kmem_zone_destroy(xfs_trans_zone);
1793 out_destroy_ifork_zone:
1794 kmem_zone_destroy(xfs_ifork_zone);
1795 out_destroy_dabuf_zone:
1796 kmem_zone_destroy(xfs_dabuf_zone);
1797 out_destroy_da_state_zone:
1798 kmem_zone_destroy(xfs_da_state_zone);
1799 out_destroy_btree_cur_zone:
1800 kmem_zone_destroy(xfs_btree_cur_zone);
1801 out_destroy_bmap_free_item_zone:
1802 kmem_zone_destroy(xfs_bmap_free_item_zone);
1803 out_destroy_log_ticket_zone:
1804 kmem_zone_destroy(xfs_log_ticket_zone);
1805 out_destroy_ioend_pool:
1806 mempool_destroy(xfs_ioend_pool);
1807 out_destroy_ioend_zone:
1808 kmem_zone_destroy(xfs_ioend_zone);
1814 xfs_destroy_zones(void)
1816 kmem_zone_destroy(xfs_ili_zone);
1817 kmem_zone_destroy(xfs_inode_zone);
1818 kmem_zone_destroy(xfs_efi_zone);
1819 kmem_zone_destroy(xfs_efd_zone);
1820 kmem_zone_destroy(xfs_buf_item_zone);
1821 kmem_zone_destroy(xfs_trans_zone);
1822 kmem_zone_destroy(xfs_ifork_zone);
1823 kmem_zone_destroy(xfs_dabuf_zone);
1824 kmem_zone_destroy(xfs_da_state_zone);
1825 kmem_zone_destroy(xfs_btree_cur_zone);
1826 kmem_zone_destroy(xfs_bmap_free_item_zone);
1827 kmem_zone_destroy(xfs_log_ticket_zone);
1828 mempool_destroy(xfs_ioend_pool);
1829 kmem_zone_destroy(xfs_ioend_zone);
1838 printk(KERN_INFO XFS_VERSION_STRING " with "
1839 XFS_BUILD_OPTIONS " enabled\n");
1844 error = xfs_init_zones();
1848 error = xfs_mru_cache_init();
1850 goto out_destroy_zones;
1852 error = xfs_filestream_init();
1854 goto out_mru_cache_uninit;
1856 error = xfs_buf_init();
1858 goto out_filestream_uninit;
1860 error = xfs_init_procfs();
1862 goto out_buf_terminate;
1864 error = xfs_sysctl_register();
1866 goto out_cleanup_procfs;
1870 error = register_filesystem(&xfs_fs_type);
1872 goto out_sysctl_unregister;
1875 out_sysctl_unregister:
1876 xfs_sysctl_unregister();
1878 xfs_cleanup_procfs();
1880 xfs_buf_terminate();
1881 out_filestream_uninit:
1882 xfs_filestream_uninit();
1883 out_mru_cache_uninit:
1884 xfs_mru_cache_uninit();
1886 xfs_destroy_zones();
1895 unregister_filesystem(&xfs_fs_type);
1896 xfs_sysctl_unregister();
1897 xfs_cleanup_procfs();
1898 xfs_buf_terminate();
1899 xfs_filestream_uninit();
1900 xfs_mru_cache_uninit();
1901 xfs_destroy_zones();
1904 module_init(init_xfs_fs);
1905 module_exit(exit_xfs_fs);
1907 MODULE_AUTHOR("Silicon Graphics, Inc.");
1908 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1909 MODULE_LICENSE("GPL");