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_quota.h"
28 #include "xfs_mount.h"
29 #include "xfs_bmap_btree.h"
30 #include "xfs_alloc_btree.h"
31 #include "xfs_ialloc_btree.h"
32 #include "xfs_dinode.h"
33 #include "xfs_inode.h"
34 #include "xfs_btree.h"
35 #include "xfs_ialloc.h"
37 #include "xfs_rtalloc.h"
38 #include "xfs_error.h"
39 #include "xfs_itable.h"
40 #include "xfs_fsops.h"
42 #include "xfs_buf_item.h"
43 #include "xfs_utils.h"
44 #include "xfs_vnodeops.h"
45 #include "xfs_log_priv.h"
46 #include "xfs_trans_priv.h"
47 #include "xfs_filestream.h"
48 #include "xfs_da_btree.h"
49 #include "xfs_extfree_item.h"
50 #include "xfs_mru_cache.h"
51 #include "xfs_inode_item.h"
52 #include "xfs_icache.h"
53 #include "xfs_trace.h"
55 #include <linux/namei.h>
56 #include <linux/init.h>
57 #include <linux/slab.h>
58 #include <linux/mount.h>
59 #include <linux/mempool.h>
60 #include <linux/writeback.h>
61 #include <linux/kthread.h>
62 #include <linux/freezer.h>
63 #include <linux/parser.h>
65 static const struct super_operations xfs_super_operations;
66 static kmem_zone_t *xfs_ioend_zone;
67 mempool_t *xfs_ioend_pool;
69 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
70 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
71 #define MNTOPT_LOGDEV "logdev" /* log device */
72 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
73 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
74 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
75 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
76 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
77 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
78 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
79 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
80 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
81 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
82 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
83 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
84 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
85 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
86 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
87 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
88 * unwritten extent conversion */
89 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
90 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
91 #define MNTOPT_32BITINODE "inode32" /* inode allocation limited to
92 * XFS_MAXINUMBER_32 */
93 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
94 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
95 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
96 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
98 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
99 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
100 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
101 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
102 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
103 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
104 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
105 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
106 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
107 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
108 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
109 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
110 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
111 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
112 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
113 #define MNTOPT_DELAYLOG "delaylog" /* Delayed logging enabled */
114 #define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed logging disabled */
115 #define MNTOPT_DISCARD "discard" /* Discard unused blocks */
116 #define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */
119 * Table driven mount option parser.
121 * Currently only used for remount, but it will be used for mount
122 * in the future, too.
132 static const match_table_t tokens = {
133 {Opt_barrier, "barrier"},
134 {Opt_nobarrier, "nobarrier"},
135 {Opt_inode64, "inode64"},
136 {Opt_inode32, "inode32"},
142 suffix_strtoul(char *s, char **endp, unsigned int base)
144 int last, shift_left_factor = 0;
147 last = strlen(value) - 1;
148 if (value[last] == 'K' || value[last] == 'k') {
149 shift_left_factor = 10;
152 if (value[last] == 'M' || value[last] == 'm') {
153 shift_left_factor = 20;
156 if (value[last] == 'G' || value[last] == 'g') {
157 shift_left_factor = 30;
161 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
165 * This function fills in xfs_mount_t fields based on mount args.
166 * Note: the superblock has _not_ yet been read in.
168 * Note that this function leaks the various device name allocations on
169 * failure. The caller takes care of them.
173 struct xfs_mount *mp,
176 struct super_block *sb = mp->m_super;
177 char *this_char, *value, *eov;
181 __uint8_t iosizelog = 0;
184 * set up the mount name first so all the errors will refer to the
187 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
190 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
193 * Copy binary VFS mount flags we are interested in.
195 if (sb->s_flags & MS_RDONLY)
196 mp->m_flags |= XFS_MOUNT_RDONLY;
197 if (sb->s_flags & MS_DIRSYNC)
198 mp->m_flags |= XFS_MOUNT_DIRSYNC;
199 if (sb->s_flags & MS_SYNCHRONOUS)
200 mp->m_flags |= XFS_MOUNT_WSYNC;
203 * Set some default flags that could be cleared by the mount option
206 mp->m_flags |= XFS_MOUNT_BARRIER;
207 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
209 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
213 * These can be overridden by the mount option parsing.
221 while ((this_char = strsep(&options, ",")) != NULL) {
224 if ((value = strchr(this_char, '=')) != NULL)
227 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
228 if (!value || !*value) {
229 xfs_warn(mp, "%s option requires an argument",
233 mp->m_logbufs = simple_strtoul(value, &eov, 10);
234 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
235 if (!value || !*value) {
236 xfs_warn(mp, "%s option requires an argument",
240 mp->m_logbsize = suffix_strtoul(value, &eov, 10);
241 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
242 if (!value || !*value) {
243 xfs_warn(mp, "%s option requires an argument",
247 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
250 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
251 xfs_warn(mp, "%s option not allowed on this system",
254 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
255 if (!value || !*value) {
256 xfs_warn(mp, "%s option requires an argument",
260 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
263 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
264 if (!value || !*value) {
265 xfs_warn(mp, "%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) {
273 xfs_warn(mp, "%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_NORECOVERY)) {
288 mp->m_flags |= XFS_MOUNT_NORECOVERY;
289 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
290 mp->m_flags |= XFS_MOUNT_NOALIGN;
291 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
292 mp->m_flags |= XFS_MOUNT_SWALLOC;
293 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
294 if (!value || !*value) {
295 xfs_warn(mp, "%s option requires an argument",
299 dsunit = simple_strtoul(value, &eov, 10);
300 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
301 if (!value || !*value) {
302 xfs_warn(mp, "%s option requires an argument",
306 dswidth = simple_strtoul(value, &eov, 10);
307 } else if (!strcmp(this_char, MNTOPT_32BITINODE)) {
308 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
309 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
310 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
312 xfs_warn(mp, "%s option not allowed on this system",
316 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
317 mp->m_flags |= XFS_MOUNT_NOUUID;
318 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
319 mp->m_flags |= XFS_MOUNT_BARRIER;
320 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
321 mp->m_flags &= ~XFS_MOUNT_BARRIER;
322 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
323 mp->m_flags |= XFS_MOUNT_IKEEP;
324 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
325 mp->m_flags &= ~XFS_MOUNT_IKEEP;
326 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
327 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
328 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
329 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
330 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
331 mp->m_flags |= XFS_MOUNT_ATTR2;
332 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
333 mp->m_flags &= ~XFS_MOUNT_ATTR2;
334 mp->m_flags |= XFS_MOUNT_NOATTR2;
335 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
336 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
337 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
338 mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT;
339 mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD;
340 mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE;
341 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
342 !strcmp(this_char, MNTOPT_UQUOTA) ||
343 !strcmp(this_char, MNTOPT_USRQUOTA)) {
344 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
346 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
347 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
348 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
349 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
350 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
351 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
352 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
354 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
355 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
356 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
357 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
358 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
359 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
361 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
362 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
363 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
364 } else if (!strcmp(this_char, MNTOPT_DELAYLOG)) {
366 "delaylog is the default now, option is deprecated.");
367 } else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) {
369 "nodelaylog support has been removed, option is deprecated.");
370 } else if (!strcmp(this_char, MNTOPT_DISCARD)) {
371 mp->m_flags |= XFS_MOUNT_DISCARD;
372 } else if (!strcmp(this_char, MNTOPT_NODISCARD)) {
373 mp->m_flags &= ~XFS_MOUNT_DISCARD;
374 } else if (!strcmp(this_char, "ihashsize")) {
376 "ihashsize no longer used, option is deprecated.");
377 } else if (!strcmp(this_char, "osyncisdsync")) {
379 "osyncisdsync has no effect, option is deprecated.");
380 } else if (!strcmp(this_char, "osyncisosync")) {
382 "osyncisosync has no effect, option is deprecated.");
383 } else if (!strcmp(this_char, "irixsgid")) {
385 "irixsgid is now a sysctl(2) variable, option is deprecated.");
387 xfs_warn(mp, "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 xfs_warn(mp, "no-recovery mounts must be read-only.");
401 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
403 "sunit and swidth options incompatible with the noalign option");
407 #ifndef CONFIG_XFS_QUOTA
408 if (XFS_IS_QUOTA_RUNNING(mp)) {
409 xfs_warn(mp, "quota support not available in this kernel.");
414 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
415 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE))) {
416 xfs_warn(mp, "cannot mount with both project and group quota");
420 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
421 xfs_warn(mp, "sunit and swidth must be specified together");
425 if (dsunit && (dswidth % dsunit != 0)) {
427 "stripe width (%d) must be a multiple of the stripe unit (%d)",
433 if (!(mp->m_flags & XFS_MOUNT_NOALIGN)) {
435 * At this point the superblock has not been read
436 * in, therefore we do not know the block size.
437 * Before the mount call ends we will convert
441 mp->m_dalign = dsunit;
442 mp->m_flags |= XFS_MOUNT_RETERR;
446 mp->m_swidth = dswidth;
449 if (mp->m_logbufs != -1 &&
450 mp->m_logbufs != 0 &&
451 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
452 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
453 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
454 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
455 return XFS_ERROR(EINVAL);
457 if (mp->m_logbsize != -1 &&
458 mp->m_logbsize != 0 &&
459 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
460 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
461 !is_power_of_2(mp->m_logbsize))) {
463 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
465 return XFS_ERROR(EINVAL);
469 if (iosizelog > XFS_MAX_IO_LOG ||
470 iosizelog < XFS_MIN_IO_LOG) {
471 xfs_warn(mp, "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_ATTR2, "," MNTOPT_ATTR2 },
504 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
505 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
506 { XFS_MOUNT_DISCARD, "," MNTOPT_DISCARD },
507 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_32BITINODE },
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_write_begin 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;
608 xfs_set_inode32(struct xfs_mount *mp)
610 xfs_agnumber_t index = 0;
611 xfs_agnumber_t maxagi = 0;
612 xfs_sb_t *sbp = &mp->m_sb;
613 xfs_agnumber_t max_metadata;
614 xfs_agino_t agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks -1, 0);
615 xfs_ino_t ino = XFS_AGINO_TO_INO(mp, sbp->sb_agcount -1, agino);
618 /* Calculate how much should be reserved for inodes to meet
619 * the max inode percentage.
621 if (mp->m_maxicount) {
624 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
626 icount += sbp->sb_agblocks - 1;
627 do_div(icount, sbp->sb_agblocks);
628 max_metadata = icount;
630 max_metadata = sbp->sb_agcount;
633 for (index = 0; index < sbp->sb_agcount; index++) {
634 ino = XFS_AGINO_TO_INO(mp, index, agino);
636 if (ino > XFS_MAXINUMBER_32) {
637 pag = xfs_perag_get(mp, index);
638 pag->pagi_inodeok = 0;
639 pag->pagf_metadata = 0;
644 pag = xfs_perag_get(mp, index);
645 pag->pagi_inodeok = 1;
647 if (index < max_metadata)
648 pag->pagf_metadata = 1;
651 mp->m_flags |= (XFS_MOUNT_32BITINODES |
652 XFS_MOUNT_SMALL_INUMS);
658 xfs_set_inode64(struct xfs_mount *mp)
660 xfs_agnumber_t index = 0;
662 for (index = 0; index < mp->m_sb.sb_agcount; index++) {
663 struct xfs_perag *pag;
665 pag = xfs_perag_get(mp, index);
666 pag->pagi_inodeok = 1;
667 pag->pagf_metadata = 0;
671 /* There is no need for lock protection on m_flags,
672 * the rw_semaphore of the VFS superblock is locked
673 * during mount/umount/remount operations, so this is
674 * enough to avoid concurency on the m_flags field
676 mp->m_flags &= ~(XFS_MOUNT_32BITINODES |
677 XFS_MOUNT_SMALL_INUMS);
685 struct block_device **bdevp)
689 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
691 if (IS_ERR(*bdevp)) {
692 error = PTR_ERR(*bdevp);
693 xfs_warn(mp, "Invalid device [%s], error=%d\n", name, error);
701 struct block_device *bdev)
704 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
708 xfs_blkdev_issue_flush(
709 xfs_buftarg_t *buftarg)
711 blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL);
716 struct xfs_mount *mp)
718 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
719 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
720 xfs_free_buftarg(mp, mp->m_logdev_targp);
721 xfs_blkdev_put(logdev);
723 if (mp->m_rtdev_targp) {
724 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
725 xfs_free_buftarg(mp, mp->m_rtdev_targp);
726 xfs_blkdev_put(rtdev);
728 xfs_free_buftarg(mp, mp->m_ddev_targp);
732 * The file system configurations are:
733 * (1) device (partition) with data and internal log
734 * (2) logical volume with data and log subvolumes.
735 * (3) logical volume with data, log, and realtime subvolumes.
737 * We only have to handle opening the log and realtime volumes here if
738 * they are present. The data subvolume has already been opened by
739 * get_sb_bdev() and is stored in sb->s_bdev.
743 struct xfs_mount *mp)
745 struct block_device *ddev = mp->m_super->s_bdev;
746 struct block_device *logdev = NULL, *rtdev = NULL;
750 * Open real time and log devices - order is important.
753 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
759 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
761 goto out_close_logdev;
763 if (rtdev == ddev || rtdev == logdev) {
765 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
767 goto out_close_rtdev;
772 * Setup xfs_mount buffer target pointers
775 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, 0, mp->m_fsname);
776 if (!mp->m_ddev_targp)
777 goto out_close_rtdev;
780 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, 1,
782 if (!mp->m_rtdev_targp)
783 goto out_free_ddev_targ;
786 if (logdev && logdev != ddev) {
787 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, 1,
789 if (!mp->m_logdev_targp)
790 goto out_free_rtdev_targ;
792 mp->m_logdev_targp = mp->m_ddev_targp;
798 if (mp->m_rtdev_targp)
799 xfs_free_buftarg(mp, mp->m_rtdev_targp);
801 xfs_free_buftarg(mp, mp->m_ddev_targp);
804 xfs_blkdev_put(rtdev);
806 if (logdev && logdev != ddev)
807 xfs_blkdev_put(logdev);
813 * Setup xfs_mount buffer target pointers based on superblock
817 struct xfs_mount *mp)
821 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
822 mp->m_sb.sb_sectsize);
826 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
827 unsigned int log_sector_size = BBSIZE;
829 if (xfs_sb_version_hassector(&mp->m_sb))
830 log_sector_size = mp->m_sb.sb_logsectsize;
831 error = xfs_setsize_buftarg(mp->m_logdev_targp,
832 mp->m_sb.sb_blocksize,
837 if (mp->m_rtdev_targp) {
838 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
839 mp->m_sb.sb_blocksize,
840 mp->m_sb.sb_sectsize);
849 xfs_init_mount_workqueues(
850 struct xfs_mount *mp)
852 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s",
853 WQ_MEM_RECLAIM, 0, mp->m_fsname);
854 if (!mp->m_data_workqueue)
857 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
858 WQ_MEM_RECLAIM, 0, mp->m_fsname);
859 if (!mp->m_unwritten_workqueue)
860 goto out_destroy_data_iodone_queue;
862 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s",
863 WQ_MEM_RECLAIM, 0, mp->m_fsname);
864 if (!mp->m_cil_workqueue)
865 goto out_destroy_unwritten;
867 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
868 WQ_NON_REENTRANT, 0, mp->m_fsname);
869 if (!mp->m_reclaim_workqueue)
870 goto out_destroy_cil;
872 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
873 WQ_NON_REENTRANT, 0, mp->m_fsname);
874 if (!mp->m_log_workqueue)
875 goto out_destroy_reclaim;
877 mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s",
878 WQ_NON_REENTRANT, 0, mp->m_fsname);
879 if (!mp->m_eofblocks_workqueue)
880 goto out_destroy_log;
885 destroy_workqueue(mp->m_log_workqueue);
887 destroy_workqueue(mp->m_reclaim_workqueue);
889 destroy_workqueue(mp->m_cil_workqueue);
890 out_destroy_unwritten:
891 destroy_workqueue(mp->m_unwritten_workqueue);
892 out_destroy_data_iodone_queue:
893 destroy_workqueue(mp->m_data_workqueue);
899 xfs_destroy_mount_workqueues(
900 struct xfs_mount *mp)
902 destroy_workqueue(mp->m_eofblocks_workqueue);
903 destroy_workqueue(mp->m_log_workqueue);
904 destroy_workqueue(mp->m_reclaim_workqueue);
905 destroy_workqueue(mp->m_cil_workqueue);
906 destroy_workqueue(mp->m_data_workqueue);
907 destroy_workqueue(mp->m_unwritten_workqueue);
911 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
912 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
913 * for IO to complete so that we effectively throttle multiple callers to the
914 * rate at which IO is completing.
918 struct xfs_mount *mp)
920 struct super_block *sb = mp->m_super;
922 if (down_read_trylock(&sb->s_umount)) {
924 up_read(&sb->s_umount);
928 /* Catch misguided souls that try to use this interface on XFS */
929 STATIC struct inode *
931 struct super_block *sb)
938 * Now that the generic code is guaranteed not to be accessing
939 * the linux inode, we can reclaim the inode.
942 xfs_fs_destroy_inode(
945 struct xfs_inode *ip = XFS_I(inode);
947 trace_xfs_destroy_inode(ip);
949 XFS_STATS_INC(vn_reclaim);
951 /* bad inode, get out here ASAP */
952 if (is_bad_inode(inode))
955 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
958 * We should never get here with one of the reclaim flags already set.
960 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
961 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
964 * We always use background reclaim here because even if the
965 * inode is clean, it still may be under IO and hence we have
966 * to take the flush lock. The background reclaim path handles
967 * this more efficiently than we can here, so simply let background
968 * reclaim tear down all inodes.
971 xfs_inode_set_reclaim_tag(ip);
975 * Slab object creation initialisation for the XFS inode.
976 * This covers only the idempotent fields in the XFS inode;
977 * all other fields need to be initialised on allocation
978 * from the slab. This avoids the need to repeatedly initialise
979 * fields in the xfs inode that left in the initialise state
980 * when freeing the inode.
983 xfs_fs_inode_init_once(
986 struct xfs_inode *ip = inode;
988 memset(ip, 0, sizeof(struct xfs_inode));
991 inode_init_once(VFS_I(ip));
994 atomic_set(&ip->i_pincount, 0);
995 spin_lock_init(&ip->i_flags_lock);
997 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
998 "xfsino", ip->i_ino);
1003 struct inode *inode)
1005 xfs_inode_t *ip = XFS_I(inode);
1007 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
1009 trace_xfs_evict_inode(ip);
1011 truncate_inode_pages(&inode->i_data, 0);
1013 XFS_STATS_INC(vn_rele);
1014 XFS_STATS_INC(vn_remove);
1015 XFS_STATS_DEC(vn_active);
1021 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1022 * serialised against cache hits here via the inode->i_lock and igrab() in
1023 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1024 * racing with us, and it avoids needing to grab a spinlock here for every inode
1025 * we drop the final reference on.
1029 struct inode *inode)
1031 struct xfs_inode *ip = XFS_I(inode);
1033 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);
1038 struct xfs_mount *mp)
1040 kfree(mp->m_fsname);
1041 kfree(mp->m_rtname);
1042 kfree(mp->m_logname);
1047 struct super_block *sb)
1049 struct xfs_mount *mp = XFS_M(sb);
1051 xfs_filestream_unmount(mp);
1055 xfs_icsb_destroy_counters(mp);
1056 xfs_destroy_mount_workqueues(mp);
1057 xfs_close_devices(mp);
1058 xfs_free_fsname(mp);
1064 struct super_block *sb,
1067 struct xfs_mount *mp = XFS_M(sb);
1070 * Doing anything during the async pass would be counterproductive.
1075 xfs_log_force(mp, XFS_LOG_SYNC);
1078 * The disk must be active because we're syncing.
1079 * We schedule log work now (now that the disk is
1080 * active) instead of later (when it might not be).
1082 flush_delayed_work(&mp->m_log->l_work);
1090 struct dentry *dentry,
1091 struct kstatfs *statp)
1093 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1094 xfs_sb_t *sbp = &mp->m_sb;
1095 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1096 __uint64_t fakeinos, id;
1100 statp->f_type = XFS_SB_MAGIC;
1101 statp->f_namelen = MAXNAMELEN - 1;
1103 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1104 statp->f_fsid.val[0] = (u32)id;
1105 statp->f_fsid.val[1] = (u32)(id >> 32);
1107 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1109 spin_lock(&mp->m_sb_lock);
1110 statp->f_bsize = sbp->sb_blocksize;
1111 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1112 statp->f_blocks = sbp->sb_dblocks - lsize;
1113 statp->f_bfree = statp->f_bavail =
1114 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1115 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1117 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1118 if (mp->m_maxicount)
1119 statp->f_files = min_t(typeof(statp->f_files),
1123 /* make sure statp->f_ffree does not underflow */
1124 ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1125 statp->f_ffree = max_t(__int64_t, ffree, 0);
1127 spin_unlock(&mp->m_sb_lock);
1129 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
1130 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))) ==
1131 (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
1132 xfs_qm_statvfs(ip, statp);
1137 xfs_save_resvblks(struct xfs_mount *mp)
1139 __uint64_t resblks = 0;
1141 mp->m_resblks_save = mp->m_resblks;
1142 xfs_reserve_blocks(mp, &resblks, NULL);
1146 xfs_restore_resvblks(struct xfs_mount *mp)
1150 if (mp->m_resblks_save) {
1151 resblks = mp->m_resblks_save;
1152 mp->m_resblks_save = 0;
1154 resblks = xfs_default_resblks(mp);
1156 xfs_reserve_blocks(mp, &resblks, NULL);
1160 * Trigger writeback of all the dirty metadata in the file system.
1162 * This ensures that the metadata is written to their location on disk rather
1163 * than just existing in transactions in the log. This means after a quiesce
1164 * there is no log replay required to write the inodes to disk - this is the
1165 * primary difference between a sync and a quiesce.
1167 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1168 * it is started again when appropriate.
1172 struct xfs_mount *mp)
1176 /* wait for all modifications to complete */
1177 while (atomic_read(&mp->m_active_trans) > 0)
1180 /* force the log to unpin objects from the now complete transactions */
1181 xfs_log_force(mp, XFS_LOG_SYNC);
1183 /* reclaim inodes to do any IO before the freeze completes */
1184 xfs_reclaim_inodes(mp, 0);
1185 xfs_reclaim_inodes(mp, SYNC_WAIT);
1187 /* Push the superblock and write an unmount record */
1188 error = xfs_log_sbcount(mp);
1190 xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
1191 "Frozen image may not be consistent.");
1193 * Just warn here till VFS can correctly support
1194 * read-only remount without racing.
1196 WARN_ON(atomic_read(&mp->m_active_trans) != 0);
1198 xfs_log_quiesce(mp);
1203 struct super_block *sb,
1207 struct xfs_mount *mp = XFS_M(sb);
1208 substring_t args[MAX_OPT_ARGS];
1212 while ((p = strsep(&options, ",")) != NULL) {
1218 token = match_token(p, tokens, args);
1221 mp->m_flags |= XFS_MOUNT_BARRIER;
1224 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1227 mp->m_maxagi = xfs_set_inode64(mp);
1230 mp->m_maxagi = xfs_set_inode32(mp);
1234 * Logically we would return an error here to prevent
1235 * users from believing they might have changed
1236 * mount options using remount which can't be changed.
1238 * But unfortunately mount(8) adds all options from
1239 * mtab and fstab to the mount arguments in some cases
1240 * so we can't blindly reject options, but have to
1241 * check for each specified option if it actually
1242 * differs from the currently set option and only
1243 * reject it if that's the case.
1245 * Until that is implemented we return success for
1246 * every remount request, and silently ignore all
1247 * options that we can't actually change.
1251 "mount option \"%s\" not supported for remount\n", p);
1260 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1261 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1264 * If this is the first remount to writeable state we
1265 * might have some superblock changes to update.
1267 if (mp->m_update_flags) {
1268 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1270 xfs_warn(mp, "failed to write sb changes");
1273 mp->m_update_flags = 0;
1277 * Fill out the reserve pool if it is empty. Use the stashed
1278 * value if it is non-zero, otherwise go with the default.
1280 xfs_restore_resvblks(mp);
1281 xfs_log_work_queue(mp);
1285 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1287 * Before we sync the metadata, we need to free up the reserve
1288 * block pool so that the used block count in the superblock on
1289 * disk is correct at the end of the remount. Stash the current
1290 * reserve pool size so that if we get remounted rw, we can
1291 * return it to the same size.
1293 xfs_save_resvblks(mp);
1294 xfs_quiesce_attr(mp);
1295 mp->m_flags |= XFS_MOUNT_RDONLY;
1302 * Second stage of a freeze. The data is already frozen so we only
1303 * need to take care of the metadata. Once that's done write a dummy
1304 * record to dirty the log in case of a crash while frozen.
1308 struct super_block *sb)
1310 struct xfs_mount *mp = XFS_M(sb);
1312 xfs_save_resvblks(mp);
1313 xfs_quiesce_attr(mp);
1314 return -xfs_fs_log_dummy(mp);
1319 struct super_block *sb)
1321 struct xfs_mount *mp = XFS_M(sb);
1323 xfs_restore_resvblks(mp);
1324 xfs_log_work_queue(mp);
1329 xfs_fs_show_options(
1331 struct dentry *root)
1333 return -xfs_showargs(XFS_M(root->d_sb), m);
1337 * This function fills in xfs_mount_t fields based on mount args.
1338 * Note: the superblock _has_ now been read in.
1342 struct xfs_mount *mp)
1344 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1346 /* Fail a mount where the logbuf is smaller than the log stripe */
1347 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1348 if (mp->m_logbsize <= 0 &&
1349 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1350 mp->m_logbsize = mp->m_sb.sb_logsunit;
1351 } else if (mp->m_logbsize > 0 &&
1352 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1354 "logbuf size must be greater than or equal to log stripe size");
1355 return XFS_ERROR(EINVAL);
1358 /* Fail a mount if the logbuf is larger than 32K */
1359 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1361 "logbuf size for version 1 logs must be 16K or 32K");
1362 return XFS_ERROR(EINVAL);
1367 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1368 * told by noattr2 to turn it off
1370 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1371 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1372 mp->m_flags |= XFS_MOUNT_ATTR2;
1375 * prohibit r/w mounts of read-only filesystems
1377 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1379 "cannot mount a read-only filesystem as read-write");
1380 return XFS_ERROR(EROFS);
1388 struct super_block *sb,
1393 struct xfs_mount *mp = NULL;
1394 int flags = 0, error = ENOMEM;
1396 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1400 spin_lock_init(&mp->m_sb_lock);
1401 mutex_init(&mp->m_growlock);
1402 atomic_set(&mp->m_active_trans, 0);
1403 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1404 INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
1409 error = xfs_parseargs(mp, (char *)data);
1411 goto out_free_fsname;
1413 sb_min_blocksize(sb, BBSIZE);
1414 sb->s_xattr = xfs_xattr_handlers;
1415 sb->s_export_op = &xfs_export_operations;
1416 #ifdef CONFIG_XFS_QUOTA
1417 sb->s_qcop = &xfs_quotactl_operations;
1419 sb->s_op = &xfs_super_operations;
1422 flags |= XFS_MFSI_QUIET;
1424 error = xfs_open_devices(mp);
1426 goto out_free_fsname;
1428 error = xfs_init_mount_workqueues(mp);
1430 goto out_close_devices;
1432 error = xfs_icsb_init_counters(mp);
1434 goto out_destroy_workqueues;
1436 error = xfs_readsb(mp, flags);
1438 goto out_destroy_counters;
1440 error = xfs_finish_flags(mp);
1444 error = xfs_setup_devices(mp);
1448 error = xfs_filestream_mount(mp);
1453 * we must configure the block size in the superblock before we run the
1454 * full mount process as the mount process can lookup and cache inodes.
1456 sb->s_magic = XFS_SB_MAGIC;
1457 sb->s_blocksize = mp->m_sb.sb_blocksize;
1458 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1459 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1460 sb->s_max_links = XFS_MAXLINK;
1461 sb->s_time_gran = 1;
1462 set_posix_acl_flag(sb);
1464 error = xfs_mountfs(mp);
1466 goto out_filestream_unmount;
1468 root = igrab(VFS_I(mp->m_rootip));
1473 if (is_bad_inode(root)) {
1477 sb->s_root = d_make_root(root);
1485 out_filestream_unmount:
1486 xfs_filestream_unmount(mp);
1489 out_destroy_counters:
1490 xfs_icsb_destroy_counters(mp);
1491 out_destroy_workqueues:
1492 xfs_destroy_mount_workqueues(mp);
1494 xfs_close_devices(mp);
1496 xfs_free_fsname(mp);
1502 xfs_filestream_unmount(mp);
1507 STATIC struct dentry *
1509 struct file_system_type *fs_type,
1511 const char *dev_name,
1514 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1518 xfs_fs_nr_cached_objects(
1519 struct super_block *sb)
1521 return xfs_reclaim_inodes_count(XFS_M(sb));
1525 xfs_fs_free_cached_objects(
1526 struct super_block *sb,
1529 xfs_reclaim_inodes_nr(XFS_M(sb), nr_to_scan);
1532 static const struct super_operations xfs_super_operations = {
1533 .alloc_inode = xfs_fs_alloc_inode,
1534 .destroy_inode = xfs_fs_destroy_inode,
1535 .evict_inode = xfs_fs_evict_inode,
1536 .drop_inode = xfs_fs_drop_inode,
1537 .put_super = xfs_fs_put_super,
1538 .sync_fs = xfs_fs_sync_fs,
1539 .freeze_fs = xfs_fs_freeze,
1540 .unfreeze_fs = xfs_fs_unfreeze,
1541 .statfs = xfs_fs_statfs,
1542 .remount_fs = xfs_fs_remount,
1543 .show_options = xfs_fs_show_options,
1544 .nr_cached_objects = xfs_fs_nr_cached_objects,
1545 .free_cached_objects = xfs_fs_free_cached_objects,
1548 static struct file_system_type xfs_fs_type = {
1549 .owner = THIS_MODULE,
1551 .mount = xfs_fs_mount,
1552 .kill_sb = kill_block_super,
1553 .fs_flags = FS_REQUIRES_DEV,
1557 xfs_init_zones(void)
1560 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1561 if (!xfs_ioend_zone)
1564 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1566 if (!xfs_ioend_pool)
1567 goto out_destroy_ioend_zone;
1569 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1571 if (!xfs_log_ticket_zone)
1572 goto out_destroy_ioend_pool;
1574 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1575 "xfs_bmap_free_item");
1576 if (!xfs_bmap_free_item_zone)
1577 goto out_destroy_log_ticket_zone;
1579 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1581 if (!xfs_btree_cur_zone)
1582 goto out_destroy_bmap_free_item_zone;
1584 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1586 if (!xfs_da_state_zone)
1587 goto out_destroy_btree_cur_zone;
1589 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1590 if (!xfs_ifork_zone)
1591 goto out_destroy_da_state_zone;
1593 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1594 if (!xfs_trans_zone)
1595 goto out_destroy_ifork_zone;
1597 xfs_log_item_desc_zone =
1598 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1599 "xfs_log_item_desc");
1600 if (!xfs_log_item_desc_zone)
1601 goto out_destroy_trans_zone;
1604 * The size of the zone allocated buf log item is the maximum
1605 * size possible under XFS. This wastes a little bit of memory,
1606 * but it is much faster.
1608 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item),
1610 if (!xfs_buf_item_zone)
1611 goto out_destroy_log_item_desc_zone;
1613 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1614 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1615 sizeof(xfs_extent_t))), "xfs_efd_item");
1617 goto out_destroy_buf_item_zone;
1619 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1620 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1621 sizeof(xfs_extent_t))), "xfs_efi_item");
1623 goto out_destroy_efd_zone;
1626 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1627 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1628 xfs_fs_inode_init_once);
1629 if (!xfs_inode_zone)
1630 goto out_destroy_efi_zone;
1633 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1634 KM_ZONE_SPREAD, NULL);
1636 goto out_destroy_inode_zone;
1640 out_destroy_inode_zone:
1641 kmem_zone_destroy(xfs_inode_zone);
1642 out_destroy_efi_zone:
1643 kmem_zone_destroy(xfs_efi_zone);
1644 out_destroy_efd_zone:
1645 kmem_zone_destroy(xfs_efd_zone);
1646 out_destroy_buf_item_zone:
1647 kmem_zone_destroy(xfs_buf_item_zone);
1648 out_destroy_log_item_desc_zone:
1649 kmem_zone_destroy(xfs_log_item_desc_zone);
1650 out_destroy_trans_zone:
1651 kmem_zone_destroy(xfs_trans_zone);
1652 out_destroy_ifork_zone:
1653 kmem_zone_destroy(xfs_ifork_zone);
1654 out_destroy_da_state_zone:
1655 kmem_zone_destroy(xfs_da_state_zone);
1656 out_destroy_btree_cur_zone:
1657 kmem_zone_destroy(xfs_btree_cur_zone);
1658 out_destroy_bmap_free_item_zone:
1659 kmem_zone_destroy(xfs_bmap_free_item_zone);
1660 out_destroy_log_ticket_zone:
1661 kmem_zone_destroy(xfs_log_ticket_zone);
1662 out_destroy_ioend_pool:
1663 mempool_destroy(xfs_ioend_pool);
1664 out_destroy_ioend_zone:
1665 kmem_zone_destroy(xfs_ioend_zone);
1671 xfs_destroy_zones(void)
1674 * Make sure all delayed rcu free are flushed before we
1678 kmem_zone_destroy(xfs_ili_zone);
1679 kmem_zone_destroy(xfs_inode_zone);
1680 kmem_zone_destroy(xfs_efi_zone);
1681 kmem_zone_destroy(xfs_efd_zone);
1682 kmem_zone_destroy(xfs_buf_item_zone);
1683 kmem_zone_destroy(xfs_log_item_desc_zone);
1684 kmem_zone_destroy(xfs_trans_zone);
1685 kmem_zone_destroy(xfs_ifork_zone);
1686 kmem_zone_destroy(xfs_da_state_zone);
1687 kmem_zone_destroy(xfs_btree_cur_zone);
1688 kmem_zone_destroy(xfs_bmap_free_item_zone);
1689 kmem_zone_destroy(xfs_log_ticket_zone);
1690 mempool_destroy(xfs_ioend_pool);
1691 kmem_zone_destroy(xfs_ioend_zone);
1696 xfs_init_workqueues(void)
1699 * The allocation workqueue can be used in memory reclaim situations
1700 * (writepage path), and parallelism is only limited by the number of
1701 * AGs in all the filesystems mounted. Hence use the default large
1702 * max_active value for this workqueue.
1704 xfs_alloc_wq = alloc_workqueue("xfsalloc", WQ_MEM_RECLAIM, 0);
1712 xfs_destroy_workqueues(void)
1714 destroy_workqueue(xfs_alloc_wq);
1722 printk(KERN_INFO XFS_VERSION_STRING " with "
1723 XFS_BUILD_OPTIONS " enabled\n");
1727 error = xfs_init_zones();
1731 error = xfs_init_workqueues();
1733 goto out_destroy_zones;
1735 error = xfs_mru_cache_init();
1737 goto out_destroy_wq;
1739 error = xfs_filestream_init();
1741 goto out_mru_cache_uninit;
1743 error = xfs_buf_init();
1745 goto out_filestream_uninit;
1747 error = xfs_init_procfs();
1749 goto out_buf_terminate;
1751 error = xfs_sysctl_register();
1753 goto out_cleanup_procfs;
1755 error = xfs_qm_init();
1757 goto out_sysctl_unregister;
1759 error = register_filesystem(&xfs_fs_type);
1766 out_sysctl_unregister:
1767 xfs_sysctl_unregister();
1769 xfs_cleanup_procfs();
1771 xfs_buf_terminate();
1772 out_filestream_uninit:
1773 xfs_filestream_uninit();
1774 out_mru_cache_uninit:
1775 xfs_mru_cache_uninit();
1777 xfs_destroy_workqueues();
1779 xfs_destroy_zones();
1788 unregister_filesystem(&xfs_fs_type);
1789 xfs_sysctl_unregister();
1790 xfs_cleanup_procfs();
1791 xfs_buf_terminate();
1792 xfs_filestream_uninit();
1793 xfs_mru_cache_uninit();
1794 xfs_destroy_workqueues();
1795 xfs_destroy_zones();
1798 module_init(init_xfs_fs);
1799 module_exit(exit_xfs_fs);
1801 MODULE_AUTHOR("Silicon Graphics, Inc.");
1802 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1803 MODULE_LICENSE("GPL");