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
22 #include "xfs_trans.h"
26 #include "xfs_alloc.h"
27 #include "xfs_dmapi.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_dir2_sf.h"
34 #include "xfs_attr_sf.h"
35 #include "xfs_dinode.h"
36 #include "xfs_inode.h"
37 #include "xfs_btree.h"
38 #include "xfs_btree_trace.h"
39 #include "xfs_ialloc.h"
41 #include "xfs_rtalloc.h"
42 #include "xfs_error.h"
43 #include "xfs_itable.h"
44 #include "xfs_fsops.h"
47 #include "xfs_buf_item.h"
48 #include "xfs_utils.h"
49 #include "xfs_vnodeops.h"
50 #include "xfs_version.h"
51 #include "xfs_log_priv.h"
52 #include "xfs_trans_priv.h"
53 #include "xfs_filestream.h"
54 #include "xfs_da_btree.h"
55 #include "xfs_dir2_trace.h"
56 #include "xfs_extfree_item.h"
57 #include "xfs_mru_cache.h"
58 #include "xfs_inode_item.h"
61 #include <linux/namei.h>
62 #include <linux/init.h>
63 #include <linux/mount.h>
64 #include <linux/mempool.h>
65 #include <linux/writeback.h>
66 #include <linux/kthread.h>
67 #include <linux/freezer.h>
68 #include <linux/parser.h>
70 static const struct super_operations xfs_super_operations;
71 static kmem_zone_t *xfs_ioend_zone;
72 mempool_t *xfs_ioend_pool;
74 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
75 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
76 #define MNTOPT_LOGDEV "logdev" /* log device */
77 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
78 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
79 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
80 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
81 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
82 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
83 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
84 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
85 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
86 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
87 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
88 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
89 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
90 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
91 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
92 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
93 * unwritten extent conversion */
94 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
95 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
96 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
97 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
98 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
99 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
100 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
102 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
103 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
104 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
105 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
106 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
107 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
108 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
109 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
110 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
111 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
112 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
113 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
114 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
115 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
116 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
117 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
118 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
119 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
122 * Table driven mount option parser.
124 * Currently only used for remount, but it will be used for mount
125 * in the future, too.
128 Opt_barrier, Opt_nobarrier, Opt_err
131 static const match_table_t tokens = {
132 {Opt_barrier, "barrier"},
133 {Opt_nobarrier, "nobarrier"},
139 suffix_strtoul(char *s, char **endp, unsigned int base)
141 int last, shift_left_factor = 0;
144 last = strlen(value) - 1;
145 if (value[last] == 'K' || value[last] == 'k') {
146 shift_left_factor = 10;
149 if (value[last] == 'M' || value[last] == 'm') {
150 shift_left_factor = 20;
153 if (value[last] == 'G' || value[last] == 'g') {
154 shift_left_factor = 30;
158 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
162 * This function fills in xfs_mount_t fields based on mount args.
163 * Note: the superblock has _not_ yet been read in.
165 * Note that this function leaks the various device name allocations on
166 * failure. The caller takes care of them.
170 struct xfs_mount *mp,
174 struct super_block *sb = mp->m_super;
175 char *this_char, *value, *eov;
179 int dmapi_implies_ikeep = 1;
180 __uint8_t iosizelog = 0;
183 * Copy binary VFS mount flags we are interested in.
185 if (sb->s_flags & MS_RDONLY)
186 mp->m_flags |= XFS_MOUNT_RDONLY;
187 if (sb->s_flags & MS_DIRSYNC)
188 mp->m_flags |= XFS_MOUNT_DIRSYNC;
189 if (sb->s_flags & MS_SYNCHRONOUS)
190 mp->m_flags |= XFS_MOUNT_WSYNC;
193 * Set some default flags that could be cleared by the mount option
196 mp->m_flags |= XFS_MOUNT_BARRIER;
197 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
198 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
201 * These can be overridden by the mount option parsing.
209 while ((this_char = strsep(&options, ",")) != NULL) {
212 if ((value = strchr(this_char, '=')) != NULL)
215 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
216 if (!value || !*value) {
218 "XFS: %s option requires an argument",
222 mp->m_logbufs = simple_strtoul(value, &eov, 10);
223 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
224 if (!value || !*value) {
226 "XFS: %s option requires an argument",
230 mp->m_logbsize = suffix_strtoul(value, &eov, 10);
231 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
232 if (!value || !*value) {
234 "XFS: %s option requires an argument",
238 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
241 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
242 if (!value || !*value) {
244 "XFS: %s option requires an argument",
248 *mtpt = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
251 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
252 if (!value || !*value) {
254 "XFS: %s option requires an argument",
258 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
261 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
262 if (!value || !*value) {
264 "XFS: %s option requires an argument",
268 iosize = simple_strtoul(value, &eov, 10);
269 iosizelog = ffs(iosize) - 1;
270 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
271 if (!value || !*value) {
273 "XFS: %s option requires an argument",
277 iosize = suffix_strtoul(value, &eov, 10);
278 iosizelog = ffs(iosize) - 1;
279 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
280 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
281 mp->m_flags |= XFS_MOUNT_GRPID;
282 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
283 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
284 mp->m_flags &= ~XFS_MOUNT_GRPID;
285 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
286 mp->m_flags |= XFS_MOUNT_WSYNC;
287 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
288 mp->m_flags |= XFS_MOUNT_OSYNCISOSYNC;
289 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
290 mp->m_flags |= XFS_MOUNT_NORECOVERY;
291 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
292 mp->m_flags |= XFS_MOUNT_NOALIGN;
293 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
294 mp->m_flags |= XFS_MOUNT_SWALLOC;
295 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
296 if (!value || !*value) {
298 "XFS: %s option requires an argument",
302 dsunit = simple_strtoul(value, &eov, 10);
303 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
304 if (!value || !*value) {
306 "XFS: %s option requires an argument",
310 dswidth = simple_strtoul(value, &eov, 10);
311 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
312 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
315 "XFS: %s option not allowed on this system",
319 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
320 mp->m_flags |= XFS_MOUNT_NOUUID;
321 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
322 mp->m_flags |= XFS_MOUNT_BARRIER;
323 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
324 mp->m_flags &= ~XFS_MOUNT_BARRIER;
325 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
326 mp->m_flags |= XFS_MOUNT_IKEEP;
327 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
328 dmapi_implies_ikeep = 0;
329 mp->m_flags &= ~XFS_MOUNT_IKEEP;
330 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
331 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
332 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
333 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
334 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
335 mp->m_flags |= XFS_MOUNT_ATTR2;
336 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
337 mp->m_flags &= ~XFS_MOUNT_ATTR2;
338 mp->m_flags |= XFS_MOUNT_NOATTR2;
339 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
340 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
341 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
342 mp->m_qflags &= ~(XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
343 XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
344 XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
345 XFS_UQUOTA_ENFD | XFS_OQUOTA_ENFD);
346 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
347 !strcmp(this_char, MNTOPT_UQUOTA) ||
348 !strcmp(this_char, MNTOPT_USRQUOTA)) {
349 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
351 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
352 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
353 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
354 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
355 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
356 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
357 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
359 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
360 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
361 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
362 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
363 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
364 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
366 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
367 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
368 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
369 } else if (!strcmp(this_char, MNTOPT_DMAPI)) {
370 mp->m_flags |= XFS_MOUNT_DMAPI;
371 } else if (!strcmp(this_char, MNTOPT_XDSM)) {
372 mp->m_flags |= XFS_MOUNT_DMAPI;
373 } else if (!strcmp(this_char, MNTOPT_DMI)) {
374 mp->m_flags |= XFS_MOUNT_DMAPI;
375 } else if (!strcmp(this_char, "ihashsize")) {
377 "XFS: ihashsize no longer used, option is deprecated.");
378 } else if (!strcmp(this_char, "osyncisdsync")) {
379 /* no-op, this is now the default */
381 "XFS: osyncisdsync is now the default, option is deprecated.");
382 } else if (!strcmp(this_char, "irixsgid")) {
384 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
387 "XFS: unknown mount option [%s].", this_char);
393 * no recovery flag requires a read-only mount
395 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
396 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
397 cmn_err(CE_WARN, "XFS: no-recovery mounts must be read-only.");
401 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
403 "XFS: sunit and swidth options incompatible with the noalign option");
407 #ifndef CONFIG_XFS_QUOTA
408 if (XFS_IS_QUOTA_RUNNING(mp)) {
410 "XFS: quota support not available in this kernel.");
415 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
416 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE))) {
418 "XFS: cannot mount with both project and group quota");
422 if ((mp->m_flags & XFS_MOUNT_DMAPI) && (!*mtpt || *mtpt[0] == '\0')) {
423 printk("XFS: %s option needs the mount point option as well\n",
428 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
430 "XFS: sunit and swidth must be specified together");
434 if (dsunit && (dswidth % dsunit != 0)) {
436 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
442 * Applications using DMI filesystems often expect the
443 * inode generation number to be monotonically increasing.
444 * If we delete inode chunks we break this assumption, so
445 * keep unused inode chunks on disk for DMI filesystems
446 * until we come up with a better solution.
447 * Note that if "ikeep" or "noikeep" mount options are
448 * supplied, then they are honored.
450 if ((mp->m_flags & XFS_MOUNT_DMAPI) && dmapi_implies_ikeep)
451 mp->m_flags |= XFS_MOUNT_IKEEP;
454 if (!(mp->m_flags & XFS_MOUNT_NOALIGN)) {
456 * At this point the superblock has not been read
457 * in, therefore we do not know the block size.
458 * Before the mount call ends we will convert
462 mp->m_dalign = dsunit;
463 mp->m_flags |= XFS_MOUNT_RETERR;
467 mp->m_swidth = dswidth;
470 if (mp->m_logbufs != -1 &&
471 mp->m_logbufs != 0 &&
472 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
473 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
475 "XFS: invalid logbufs value: %d [not %d-%d]",
476 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
477 return XFS_ERROR(EINVAL);
479 if (mp->m_logbsize != -1 &&
480 mp->m_logbsize != 0 &&
481 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
482 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
483 !is_power_of_2(mp->m_logbsize))) {
485 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
487 return XFS_ERROR(EINVAL);
490 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
493 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
496 if (iosizelog > XFS_MAX_IO_LOG ||
497 iosizelog < XFS_MIN_IO_LOG) {
499 "XFS: invalid log iosize: %d [not %d-%d]",
500 iosizelog, XFS_MIN_IO_LOG,
502 return XFS_ERROR(EINVAL);
505 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
506 mp->m_readio_log = iosizelog;
507 mp->m_writeio_log = iosizelog;
513 struct proc_xfs_info {
520 struct xfs_mount *mp,
523 static struct proc_xfs_info xfs_info_set[] = {
524 /* the few simple ones we can get from the mount struct */
525 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
526 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
527 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
528 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
529 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
530 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
531 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
532 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
533 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
534 { XFS_MOUNT_DMAPI, "," MNTOPT_DMAPI },
535 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
538 static struct proc_xfs_info xfs_info_unset[] = {
539 /* the few simple ones we can get from the mount struct */
540 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
541 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
542 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
545 struct proc_xfs_info *xfs_infop;
547 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
548 if (mp->m_flags & xfs_infop->flag)
549 seq_puts(m, xfs_infop->str);
551 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
552 if (!(mp->m_flags & xfs_infop->flag))
553 seq_puts(m, xfs_infop->str);
556 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
557 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
558 (int)(1 << mp->m_writeio_log) >> 10);
560 if (mp->m_logbufs > 0)
561 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
562 if (mp->m_logbsize > 0)
563 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
566 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
568 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
570 if (mp->m_dalign > 0)
571 seq_printf(m, "," MNTOPT_SUNIT "=%d",
572 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
573 if (mp->m_swidth > 0)
574 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
575 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
577 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
578 seq_puts(m, "," MNTOPT_USRQUOTA);
579 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
580 seq_puts(m, "," MNTOPT_UQUOTANOENF);
582 /* Either project or group quotas can be active, not both */
584 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
585 if (mp->m_qflags & XFS_OQUOTA_ENFD)
586 seq_puts(m, "," MNTOPT_PRJQUOTA);
588 seq_puts(m, "," MNTOPT_PQUOTANOENF);
589 } else if (mp->m_qflags & XFS_GQUOTA_ACCT) {
590 if (mp->m_qflags & XFS_OQUOTA_ENFD)
591 seq_puts(m, "," MNTOPT_GRPQUOTA);
593 seq_puts(m, "," MNTOPT_GQUOTANOENF);
596 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
597 seq_puts(m, "," MNTOPT_NOQUOTA);
603 unsigned int blockshift)
605 unsigned int pagefactor = 1;
606 unsigned int bitshift = BITS_PER_LONG - 1;
608 /* Figure out maximum filesize, on Linux this can depend on
609 * the filesystem blocksize (on 32 bit platforms).
610 * __block_prepare_write does this in an [unsigned] long...
611 * page->index << (PAGE_CACHE_SHIFT - bbits)
612 * So, for page sized blocks (4K on 32 bit platforms),
613 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
614 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
615 * but for smaller blocksizes it is less (bbits = log2 bsize).
616 * Note1: get_block_t takes a long (implicit cast from above)
617 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
618 * can optionally convert the [unsigned] long from above into
619 * an [unsigned] long long.
622 #if BITS_PER_LONG == 32
623 # if defined(CONFIG_LBDAF)
624 ASSERT(sizeof(sector_t) == 8);
625 pagefactor = PAGE_CACHE_SIZE;
626 bitshift = BITS_PER_LONG;
628 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
632 return (((__uint64_t)pagefactor) << bitshift) - 1;
639 struct block_device **bdevp)
643 *bdevp = open_bdev_exclusive(name, FMODE_READ|FMODE_WRITE, mp);
644 if (IS_ERR(*bdevp)) {
645 error = PTR_ERR(*bdevp);
646 printk("XFS: Invalid device [%s], error=%d\n", name, error);
654 struct block_device *bdev)
657 close_bdev_exclusive(bdev, FMODE_READ|FMODE_WRITE);
661 * Try to write out the superblock using barriers.
667 xfs_buf_t *sbp = xfs_getsb(mp, 0);
672 XFS_BUF_UNDELAYWRITE(sbp);
674 XFS_BUF_UNASYNC(sbp);
675 XFS_BUF_ORDERED(sbp);
678 error = xfs_iowait(sbp);
681 * Clear all the flags we set and possible error state in the
682 * buffer. We only did the write to try out whether barriers
683 * worked and shouldn't leave any traces in the superblock
687 XFS_BUF_ERROR(sbp, 0);
688 XFS_BUF_UNORDERED(sbp);
695 xfs_mountfs_check_barriers(xfs_mount_t *mp)
699 if (mp->m_logdev_targp != mp->m_ddev_targp) {
700 xfs_fs_cmn_err(CE_NOTE, mp,
701 "Disabling barriers, not supported with external log device");
702 mp->m_flags &= ~XFS_MOUNT_BARRIER;
706 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
707 xfs_fs_cmn_err(CE_NOTE, mp,
708 "Disabling barriers, underlying device is readonly");
709 mp->m_flags &= ~XFS_MOUNT_BARRIER;
713 error = xfs_barrier_test(mp);
715 xfs_fs_cmn_err(CE_NOTE, mp,
716 "Disabling barriers, trial barrier write failed");
717 mp->m_flags &= ~XFS_MOUNT_BARRIER;
723 xfs_blkdev_issue_flush(
724 xfs_buftarg_t *buftarg)
726 blkdev_issue_flush(buftarg->bt_bdev, NULL);
731 struct xfs_mount *mp)
733 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
734 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
735 xfs_free_buftarg(mp, mp->m_logdev_targp);
736 xfs_blkdev_put(logdev);
738 if (mp->m_rtdev_targp) {
739 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
740 xfs_free_buftarg(mp, mp->m_rtdev_targp);
741 xfs_blkdev_put(rtdev);
743 xfs_free_buftarg(mp, mp->m_ddev_targp);
747 * The file system configurations are:
748 * (1) device (partition) with data and internal log
749 * (2) logical volume with data and log subvolumes.
750 * (3) logical volume with data, log, and realtime subvolumes.
752 * We only have to handle opening the log and realtime volumes here if
753 * they are present. The data subvolume has already been opened by
754 * get_sb_bdev() and is stored in sb->s_bdev.
758 struct xfs_mount *mp)
760 struct block_device *ddev = mp->m_super->s_bdev;
761 struct block_device *logdev = NULL, *rtdev = NULL;
765 * Open real time and log devices - order is important.
768 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
774 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
776 goto out_close_logdev;
778 if (rtdev == ddev || rtdev == logdev) {
780 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
782 goto out_close_rtdev;
787 * Setup xfs_mount buffer target pointers
790 mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0);
791 if (!mp->m_ddev_targp)
792 goto out_close_rtdev;
795 mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1);
796 if (!mp->m_rtdev_targp)
797 goto out_free_ddev_targ;
800 if (logdev && logdev != ddev) {
801 mp->m_logdev_targp = xfs_alloc_buftarg(logdev, 1);
802 if (!mp->m_logdev_targp)
803 goto out_free_rtdev_targ;
805 mp->m_logdev_targp = mp->m_ddev_targp;
811 if (mp->m_rtdev_targp)
812 xfs_free_buftarg(mp, mp->m_rtdev_targp);
814 xfs_free_buftarg(mp, mp->m_ddev_targp);
817 xfs_blkdev_put(rtdev);
819 if (logdev && logdev != ddev)
820 xfs_blkdev_put(logdev);
826 * Setup xfs_mount buffer target pointers based on superblock
830 struct xfs_mount *mp)
834 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
835 mp->m_sb.sb_sectsize);
839 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
840 unsigned int log_sector_size = BBSIZE;
842 if (xfs_sb_version_hassector(&mp->m_sb))
843 log_sector_size = mp->m_sb.sb_logsectsize;
844 error = xfs_setsize_buftarg(mp->m_logdev_targp,
845 mp->m_sb.sb_blocksize,
850 if (mp->m_rtdev_targp) {
851 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
852 mp->m_sb.sb_blocksize,
853 mp->m_sb.sb_sectsize);
862 * XFS AIL push thread support
866 struct xfs_ail *ailp,
867 xfs_lsn_t threshold_lsn)
869 ailp->xa_target = threshold_lsn;
870 wake_up_process(ailp->xa_task);
877 struct xfs_ail *ailp = data;
878 xfs_lsn_t last_pushed_lsn = 0;
881 while (!kthread_should_stop()) {
883 schedule_timeout_interruptible(msecs_to_jiffies(tout));
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 * If we have nothing to flush with this inode then complete the
957 * teardown now, otherwise delay the flush operation.
959 if (!xfs_inode_clean(ip)) {
960 xfs_inode_set_reclaim_tag(ip);
969 * Slab object creation initialisation for the XFS inode.
970 * This covers only the idempotent fields in the XFS inode;
971 * all other fields need to be initialised on allocation
972 * from the slab. This avoids the need to repeatedly intialise
973 * fields in the xfs inode that left in the initialise state
974 * when freeing the inode.
977 xfs_fs_inode_init_once(
980 struct xfs_inode *ip = inode;
982 memset(ip, 0, sizeof(struct xfs_inode));
985 inode_init_once(VFS_I(ip));
988 atomic_set(&ip->i_iocount, 0);
989 atomic_set(&ip->i_pincount, 0);
990 spin_lock_init(&ip->i_flags_lock);
991 init_waitqueue_head(&ip->i_ipin_wait);
993 * Because we want to use a counting completion, complete
994 * the flush completion once to allow a single access to
995 * the flush completion without blocking.
997 init_completion(&ip->i_flush);
998 complete(&ip->i_flush);
1000 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
1001 "xfsino", ip->i_ino);
1002 mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
1006 * Dirty the XFS inode when mark_inode_dirty_sync() is called so that
1007 * we catch unlogged VFS level updates to the inode. Care must be taken
1008 * here - the transaction code calls mark_inode_dirty_sync() to mark the
1009 * VFS inode dirty in a transaction and clears the i_update_core field;
1010 * it must clear the field after calling mark_inode_dirty_sync() to
1011 * correctly indicate that the dirty state has been propagated into the
1014 * We need the barrier() to maintain correct ordering between unlogged
1015 * updates and the transaction commit code that clears the i_update_core
1016 * field. This requires all updates to be completed before marking the
1021 struct inode *inode)
1024 XFS_I(inode)->i_update_core = 1;
1028 * Attempt to flush the inode, this will actually fail
1029 * if the inode is pinned, but we dirty the inode again
1030 * at the point when it is unpinned after a log write,
1031 * since this is when the inode itself becomes flushable.
1035 struct inode *inode,
1038 struct xfs_inode *ip = XFS_I(inode);
1039 struct xfs_mount *mp = ip->i_mount;
1042 xfs_itrace_entry(ip);
1044 if (XFS_FORCED_SHUTDOWN(mp))
1045 return XFS_ERROR(EIO);
1048 error = xfs_wait_on_pages(ip, 0, -1);
1054 * Bypass inodes which have already been cleaned by
1055 * the inode flush clustering code inside xfs_iflush
1057 if (xfs_inode_clean(ip))
1061 * We make this non-blocking if the inode is contended, return
1062 * EAGAIN to indicate to the caller that they did not succeed.
1063 * This prevents the flush path from blocking on inodes inside
1064 * another operation right now, they get caught later by xfs_sync.
1067 xfs_ilock(ip, XFS_ILOCK_SHARED);
1070 error = xfs_iflush(ip, XFS_IFLUSH_SYNC);
1073 if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED))
1075 if (xfs_ipincount(ip) || !xfs_iflock_nowait(ip))
1078 error = xfs_iflush(ip, XFS_IFLUSH_ASYNC_NOBLOCK);
1082 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1085 * if we failed to write out the inode then mark
1086 * it dirty again so we'll try again later.
1089 xfs_mark_inode_dirty_sync(ip);
1095 struct inode *inode)
1097 xfs_inode_t *ip = XFS_I(inode);
1099 xfs_itrace_entry(ip);
1100 XFS_STATS_INC(vn_rele);
1101 XFS_STATS_INC(vn_remove);
1102 XFS_STATS_DEC(vn_active);
1109 struct xfs_mount *mp)
1111 kfree(mp->m_fsname);
1112 kfree(mp->m_rtname);
1113 kfree(mp->m_logname);
1118 struct super_block *sb)
1120 struct xfs_mount *mp = XFS_M(sb);
1121 struct xfs_inode *rip = mp->m_rootip;
1122 int unmount_event_flags = 0;
1126 if (!(sb->s_flags & MS_RDONLY)) {
1128 * XXX(hch): this should be SYNC_WAIT.
1130 * Or more likely not needed at all because the VFS is already
1131 * calling ->sync_fs after shutting down all filestem
1132 * operations and just before calling ->put_super.
1134 xfs_sync_data(mp, 0);
1135 xfs_sync_attr(mp, 0);
1139 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1140 unmount_event_flags =
1141 (mp->m_dmevmask & (1 << DM_EVENT_UNMOUNT)) ?
1142 0 : DM_FLAGS_UNWANTED;
1144 * Ignore error from dmapi here, first unmount is not allowed
1145 * to fail anyway, and second we wouldn't want to fail a
1146 * unmount because of dmapi.
1148 XFS_SEND_PREUNMOUNT(mp, rip, DM_RIGHT_NULL, rip, DM_RIGHT_NULL,
1149 NULL, NULL, 0, 0, unmount_event_flags);
1154 * Blow away any referenced inode in the filestreams cache.
1155 * This can and will cause log traffic as inodes go inactive
1158 xfs_filestream_unmount(mp);
1160 XFS_bflush(mp->m_ddev_targp);
1162 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1163 XFS_SEND_UNMOUNT(mp, rip, DM_RIGHT_NULL, 0, 0,
1164 unmount_event_flags);
1169 xfs_icsb_destroy_counters(mp);
1170 xfs_close_devices(mp);
1172 xfs_free_fsname(mp);
1178 struct super_block *sb,
1181 struct xfs_mount *mp = XFS_M(sb);
1185 * Not much we can do for the first async pass. Writing out the
1186 * superblock would be counter-productive as we are going to redirty
1187 * when writing out other data and metadata (and writing out a single
1188 * block is quite fast anyway).
1190 * Try to asynchronously kick off quota syncing at least.
1193 xfs_qm_sync(mp, SYNC_TRYLOCK);
1197 error = xfs_quiesce_data(mp);
1202 int prev_sync_seq = mp->m_sync_seq;
1205 * The disk must be active because we're syncing.
1206 * We schedule xfssyncd now (now that the disk is
1207 * active) instead of later (when it might not be).
1209 wake_up_process(mp->m_sync_task);
1211 * We have to wait for the sync iteration to complete.
1212 * If we don't, the disk activity caused by the sync
1213 * will come after the sync is completed, and that
1214 * triggers another sync from laptop mode.
1216 wait_event(mp->m_wait_single_sync_task,
1217 mp->m_sync_seq != prev_sync_seq);
1225 struct dentry *dentry,
1226 struct kstatfs *statp)
1228 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1229 xfs_sb_t *sbp = &mp->m_sb;
1230 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1231 __uint64_t fakeinos, id;
1234 statp->f_type = XFS_SB_MAGIC;
1235 statp->f_namelen = MAXNAMELEN - 1;
1237 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1238 statp->f_fsid.val[0] = (u32)id;
1239 statp->f_fsid.val[1] = (u32)(id >> 32);
1241 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1243 spin_lock(&mp->m_sb_lock);
1244 statp->f_bsize = sbp->sb_blocksize;
1245 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1246 statp->f_blocks = sbp->sb_dblocks - lsize;
1247 statp->f_bfree = statp->f_bavail =
1248 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1249 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1251 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1252 if (mp->m_maxicount)
1253 statp->f_files = min_t(typeof(statp->f_files),
1256 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1257 spin_unlock(&mp->m_sb_lock);
1259 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) ||
1260 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))) ==
1261 (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
1262 xfs_qm_statvfs(ip, statp);
1268 struct super_block *sb,
1272 struct xfs_mount *mp = XFS_M(sb);
1273 substring_t args[MAX_OPT_ARGS];
1277 while ((p = strsep(&options, ",")) != NULL) {
1283 token = match_token(p, tokens, args);
1286 mp->m_flags |= XFS_MOUNT_BARRIER;
1289 * Test if barriers are actually working if we can,
1290 * else delay this check until the filesystem is
1293 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1294 xfs_mountfs_check_barriers(mp);
1297 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1301 * Logically we would return an error here to prevent
1302 * users from believing they might have changed
1303 * mount options using remount which can't be changed.
1305 * But unfortunately mount(8) adds all options from
1306 * mtab and fstab to the mount arguments in some cases
1307 * so we can't blindly reject options, but have to
1308 * check for each specified option if it actually
1309 * differs from the currently set option and only
1310 * reject it if that's the case.
1312 * Until that is implemented we return success for
1313 * every remount request, and silently ignore all
1314 * options that we can't actually change.
1318 "XFS: mount option \"%s\" not supported for remount\n", p);
1327 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1328 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1329 if (mp->m_flags & XFS_MOUNT_BARRIER)
1330 xfs_mountfs_check_barriers(mp);
1333 * If this is the first remount to writeable state we
1334 * might have some superblock changes to update.
1336 if (mp->m_update_flags) {
1337 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1340 "XFS: failed to write sb changes");
1343 mp->m_update_flags = 0;
1348 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1349 xfs_quiesce_data(mp);
1350 xfs_quiesce_attr(mp);
1351 mp->m_flags |= XFS_MOUNT_RDONLY;
1358 * Second stage of a freeze. The data is already frozen so we only
1359 * need to take care of the metadata. Once that's done write a dummy
1360 * record to dirty the log in case of a crash while frozen.
1364 struct super_block *sb)
1366 struct xfs_mount *mp = XFS_M(sb);
1368 xfs_quiesce_attr(mp);
1369 return -xfs_fs_log_dummy(mp);
1373 xfs_fs_show_options(
1375 struct vfsmount *mnt)
1377 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1381 * This function fills in xfs_mount_t fields based on mount args.
1382 * Note: the superblock _has_ now been read in.
1386 struct xfs_mount *mp)
1388 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1390 /* Fail a mount where the logbuf is smaller than the log stripe */
1391 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1392 if (mp->m_logbsize <= 0 &&
1393 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1394 mp->m_logbsize = mp->m_sb.sb_logsunit;
1395 } else if (mp->m_logbsize > 0 &&
1396 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1398 "XFS: logbuf size must be greater than or equal to log stripe size");
1399 return XFS_ERROR(EINVAL);
1402 /* Fail a mount if the logbuf is larger than 32K */
1403 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1405 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1406 return XFS_ERROR(EINVAL);
1411 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1412 * told by noattr2 to turn it off
1414 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1415 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1416 mp->m_flags |= XFS_MOUNT_ATTR2;
1419 * prohibit r/w mounts of read-only filesystems
1421 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1423 "XFS: cannot mount a read-only filesystem as read-write");
1424 return XFS_ERROR(EROFS);
1432 struct super_block *sb,
1437 struct xfs_mount *mp = NULL;
1438 int flags = 0, error = ENOMEM;
1441 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1445 spin_lock_init(&mp->m_sb_lock);
1446 mutex_init(&mp->m_growlock);
1447 atomic_set(&mp->m_active_trans, 0);
1448 INIT_LIST_HEAD(&mp->m_sync_list);
1449 spin_lock_init(&mp->m_sync_lock);
1450 init_waitqueue_head(&mp->m_wait_single_sync_task);
1455 error = xfs_parseargs(mp, (char *)data, &mtpt);
1457 goto out_free_fsname;
1459 sb_min_blocksize(sb, BBSIZE);
1460 sb->s_xattr = xfs_xattr_handlers;
1461 sb->s_export_op = &xfs_export_operations;
1462 #ifdef CONFIG_XFS_QUOTA
1463 sb->s_qcop = &xfs_quotactl_operations;
1465 sb->s_op = &xfs_super_operations;
1467 error = xfs_dmops_get(mp);
1469 goto out_free_fsname;
1472 flags |= XFS_MFSI_QUIET;
1474 error = xfs_open_devices(mp);
1478 if (xfs_icsb_init_counters(mp))
1479 mp->m_flags |= XFS_MOUNT_NO_PERCPU_SB;
1481 error = xfs_readsb(mp, flags);
1483 goto out_destroy_counters;
1485 error = xfs_finish_flags(mp);
1489 error = xfs_setup_devices(mp);
1493 if (mp->m_flags & XFS_MOUNT_BARRIER)
1494 xfs_mountfs_check_barriers(mp);
1496 error = xfs_filestream_mount(mp);
1500 error = xfs_mountfs(mp);
1502 goto out_filestream_unmount;
1504 XFS_SEND_MOUNT(mp, DM_RIGHT_NULL, mtpt, mp->m_fsname);
1506 sb->s_magic = XFS_SB_MAGIC;
1507 sb->s_blocksize = mp->m_sb.sb_blocksize;
1508 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1509 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1510 sb->s_time_gran = 1;
1511 set_posix_acl_flag(sb);
1513 root = igrab(VFS_I(mp->m_rootip));
1518 if (is_bad_inode(root)) {
1522 sb->s_root = d_alloc_root(root);
1528 error = xfs_syncd_init(mp);
1534 xfs_itrace_exit(XFS_I(sb->s_root->d_inode));
1537 out_filestream_unmount:
1538 xfs_filestream_unmount(mp);
1541 out_destroy_counters:
1542 xfs_icsb_destroy_counters(mp);
1543 xfs_close_devices(mp);
1547 xfs_free_fsname(mp);
1563 * Blow away any referenced inode in the filestreams cache.
1564 * This can and will cause log traffic as inodes go inactive
1567 xfs_filestream_unmount(mp);
1569 XFS_bflush(mp->m_ddev_targp);
1577 struct file_system_type *fs_type,
1579 const char *dev_name,
1581 struct vfsmount *mnt)
1583 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1587 static const struct super_operations xfs_super_operations = {
1588 .alloc_inode = xfs_fs_alloc_inode,
1589 .destroy_inode = xfs_fs_destroy_inode,
1590 .dirty_inode = xfs_fs_dirty_inode,
1591 .write_inode = xfs_fs_write_inode,
1592 .clear_inode = xfs_fs_clear_inode,
1593 .put_super = xfs_fs_put_super,
1594 .sync_fs = xfs_fs_sync_fs,
1595 .freeze_fs = xfs_fs_freeze,
1596 .statfs = xfs_fs_statfs,
1597 .remount_fs = xfs_fs_remount,
1598 .show_options = xfs_fs_show_options,
1601 static struct file_system_type xfs_fs_type = {
1602 .owner = THIS_MODULE,
1604 .get_sb = xfs_fs_get_sb,
1605 .kill_sb = kill_block_super,
1606 .fs_flags = FS_REQUIRES_DEV,
1610 xfs_alloc_trace_bufs(void)
1612 #ifdef XFS_ALLOC_TRACE
1613 xfs_alloc_trace_buf = ktrace_alloc(XFS_ALLOC_TRACE_SIZE, KM_MAYFAIL);
1614 if (!xfs_alloc_trace_buf)
1617 #ifdef XFS_BMAP_TRACE
1618 xfs_bmap_trace_buf = ktrace_alloc(XFS_BMAP_TRACE_SIZE, KM_MAYFAIL);
1619 if (!xfs_bmap_trace_buf)
1620 goto out_free_alloc_trace;
1622 #ifdef XFS_BTREE_TRACE
1623 xfs_allocbt_trace_buf = ktrace_alloc(XFS_ALLOCBT_TRACE_SIZE,
1625 if (!xfs_allocbt_trace_buf)
1626 goto out_free_bmap_trace;
1628 xfs_inobt_trace_buf = ktrace_alloc(XFS_INOBT_TRACE_SIZE, KM_MAYFAIL);
1629 if (!xfs_inobt_trace_buf)
1630 goto out_free_allocbt_trace;
1632 xfs_bmbt_trace_buf = ktrace_alloc(XFS_BMBT_TRACE_SIZE, KM_MAYFAIL);
1633 if (!xfs_bmbt_trace_buf)
1634 goto out_free_inobt_trace;
1636 #ifdef XFS_ATTR_TRACE
1637 xfs_attr_trace_buf = ktrace_alloc(XFS_ATTR_TRACE_SIZE, KM_MAYFAIL);
1638 if (!xfs_attr_trace_buf)
1639 goto out_free_bmbt_trace;
1641 #ifdef XFS_DIR2_TRACE
1642 xfs_dir2_trace_buf = ktrace_alloc(XFS_DIR2_GTRACE_SIZE, KM_MAYFAIL);
1643 if (!xfs_dir2_trace_buf)
1644 goto out_free_attr_trace;
1649 #ifdef XFS_DIR2_TRACE
1650 out_free_attr_trace:
1652 #ifdef XFS_ATTR_TRACE
1653 ktrace_free(xfs_attr_trace_buf);
1654 out_free_bmbt_trace:
1656 #ifdef XFS_BTREE_TRACE
1657 ktrace_free(xfs_bmbt_trace_buf);
1658 out_free_inobt_trace:
1659 ktrace_free(xfs_inobt_trace_buf);
1660 out_free_allocbt_trace:
1661 ktrace_free(xfs_allocbt_trace_buf);
1662 out_free_bmap_trace:
1664 #ifdef XFS_BMAP_TRACE
1665 ktrace_free(xfs_bmap_trace_buf);
1666 out_free_alloc_trace:
1668 #ifdef XFS_ALLOC_TRACE
1669 ktrace_free(xfs_alloc_trace_buf);
1676 xfs_free_trace_bufs(void)
1678 #ifdef XFS_DIR2_TRACE
1679 ktrace_free(xfs_dir2_trace_buf);
1681 #ifdef XFS_ATTR_TRACE
1682 ktrace_free(xfs_attr_trace_buf);
1684 #ifdef XFS_BTREE_TRACE
1685 ktrace_free(xfs_bmbt_trace_buf);
1686 ktrace_free(xfs_inobt_trace_buf);
1687 ktrace_free(xfs_allocbt_trace_buf);
1689 #ifdef XFS_BMAP_TRACE
1690 ktrace_free(xfs_bmap_trace_buf);
1692 #ifdef XFS_ALLOC_TRACE
1693 ktrace_free(xfs_alloc_trace_buf);
1698 xfs_init_zones(void)
1701 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1702 if (!xfs_ioend_zone)
1705 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1707 if (!xfs_ioend_pool)
1708 goto out_destroy_ioend_zone;
1710 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1712 if (!xfs_log_ticket_zone)
1713 goto out_destroy_ioend_pool;
1715 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1716 "xfs_bmap_free_item");
1717 if (!xfs_bmap_free_item_zone)
1718 goto out_destroy_log_ticket_zone;
1720 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1722 if (!xfs_btree_cur_zone)
1723 goto out_destroy_bmap_free_item_zone;
1725 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1727 if (!xfs_da_state_zone)
1728 goto out_destroy_btree_cur_zone;
1730 xfs_dabuf_zone = kmem_zone_init(sizeof(xfs_dabuf_t), "xfs_dabuf");
1731 if (!xfs_dabuf_zone)
1732 goto out_destroy_da_state_zone;
1734 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1735 if (!xfs_ifork_zone)
1736 goto out_destroy_dabuf_zone;
1738 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1739 if (!xfs_trans_zone)
1740 goto out_destroy_ifork_zone;
1743 * The size of the zone allocated buf log item is the maximum
1744 * size possible under XFS. This wastes a little bit of memory,
1745 * but it is much faster.
1747 xfs_buf_item_zone = kmem_zone_init((sizeof(xfs_buf_log_item_t) +
1748 (((XFS_MAX_BLOCKSIZE / XFS_BLI_CHUNK) /
1749 NBWORD) * sizeof(int))), "xfs_buf_item");
1750 if (!xfs_buf_item_zone)
1751 goto out_destroy_trans_zone;
1753 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1754 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1755 sizeof(xfs_extent_t))), "xfs_efd_item");
1757 goto out_destroy_buf_item_zone;
1759 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1760 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1761 sizeof(xfs_extent_t))), "xfs_efi_item");
1763 goto out_destroy_efd_zone;
1766 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1767 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1768 xfs_fs_inode_init_once);
1769 if (!xfs_inode_zone)
1770 goto out_destroy_efi_zone;
1773 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1774 KM_ZONE_SPREAD, NULL);
1776 goto out_destroy_inode_zone;
1780 out_destroy_inode_zone:
1781 kmem_zone_destroy(xfs_inode_zone);
1782 out_destroy_efi_zone:
1783 kmem_zone_destroy(xfs_efi_zone);
1784 out_destroy_efd_zone:
1785 kmem_zone_destroy(xfs_efd_zone);
1786 out_destroy_buf_item_zone:
1787 kmem_zone_destroy(xfs_buf_item_zone);
1788 out_destroy_trans_zone:
1789 kmem_zone_destroy(xfs_trans_zone);
1790 out_destroy_ifork_zone:
1791 kmem_zone_destroy(xfs_ifork_zone);
1792 out_destroy_dabuf_zone:
1793 kmem_zone_destroy(xfs_dabuf_zone);
1794 out_destroy_da_state_zone:
1795 kmem_zone_destroy(xfs_da_state_zone);
1796 out_destroy_btree_cur_zone:
1797 kmem_zone_destroy(xfs_btree_cur_zone);
1798 out_destroy_bmap_free_item_zone:
1799 kmem_zone_destroy(xfs_bmap_free_item_zone);
1800 out_destroy_log_ticket_zone:
1801 kmem_zone_destroy(xfs_log_ticket_zone);
1802 out_destroy_ioend_pool:
1803 mempool_destroy(xfs_ioend_pool);
1804 out_destroy_ioend_zone:
1805 kmem_zone_destroy(xfs_ioend_zone);
1811 xfs_destroy_zones(void)
1813 kmem_zone_destroy(xfs_ili_zone);
1814 kmem_zone_destroy(xfs_inode_zone);
1815 kmem_zone_destroy(xfs_efi_zone);
1816 kmem_zone_destroy(xfs_efd_zone);
1817 kmem_zone_destroy(xfs_buf_item_zone);
1818 kmem_zone_destroy(xfs_trans_zone);
1819 kmem_zone_destroy(xfs_ifork_zone);
1820 kmem_zone_destroy(xfs_dabuf_zone);
1821 kmem_zone_destroy(xfs_da_state_zone);
1822 kmem_zone_destroy(xfs_btree_cur_zone);
1823 kmem_zone_destroy(xfs_bmap_free_item_zone);
1824 kmem_zone_destroy(xfs_log_ticket_zone);
1825 mempool_destroy(xfs_ioend_pool);
1826 kmem_zone_destroy(xfs_ioend_zone);
1835 printk(KERN_INFO XFS_VERSION_STRING " with "
1836 XFS_BUILD_OPTIONS " enabled\n");
1842 error = xfs_init_zones();
1846 error = xfs_alloc_trace_bufs();
1848 goto out_destroy_zones;
1850 error = xfs_mru_cache_init();
1852 goto out_free_trace_buffers;
1854 error = xfs_filestream_init();
1856 goto out_mru_cache_uninit;
1858 error = xfs_buf_init();
1860 goto out_filestream_uninit;
1862 error = xfs_init_procfs();
1864 goto out_buf_terminate;
1866 error = xfs_sysctl_register();
1868 goto out_cleanup_procfs;
1872 error = register_filesystem(&xfs_fs_type);
1874 goto out_sysctl_unregister;
1877 out_sysctl_unregister:
1878 xfs_sysctl_unregister();
1880 xfs_cleanup_procfs();
1882 xfs_buf_terminate();
1883 out_filestream_uninit:
1884 xfs_filestream_uninit();
1885 out_mru_cache_uninit:
1886 xfs_mru_cache_uninit();
1887 out_free_trace_buffers:
1888 xfs_free_trace_bufs();
1890 xfs_destroy_zones();
1899 unregister_filesystem(&xfs_fs_type);
1900 xfs_sysctl_unregister();
1901 xfs_cleanup_procfs();
1902 xfs_buf_terminate();
1903 xfs_filestream_uninit();
1904 xfs_mru_cache_uninit();
1905 xfs_free_trace_bufs();
1906 xfs_destroy_zones();
1910 module_init(init_xfs_fs);
1911 module_exit(exit_xfs_fs);
1913 MODULE_AUTHOR("Silicon Graphics, Inc.");
1914 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1915 MODULE_LICENSE("GPL");