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"
49 #include "xfs_buf_item.h"
50 #include "xfs_utils.h"
51 #include "xfs_vnodeops.h"
52 #include "xfs_vfsops.h"
53 #include "xfs_version.h"
54 #include "xfs_log_priv.h"
55 #include "xfs_trans_priv.h"
56 #include "xfs_filestream.h"
57 #include "xfs_da_btree.h"
58 #include "xfs_dir2_trace.h"
59 #include "xfs_extfree_item.h"
60 #include "xfs_mru_cache.h"
61 #include "xfs_inode_item.h"
64 #include <linux/namei.h>
65 #include <linux/init.h>
66 #include <linux/mount.h>
67 #include <linux/mempool.h>
68 #include <linux/writeback.h>
69 #include <linux/kthread.h>
70 #include <linux/freezer.h>
71 #include <linux/parser.h>
73 static struct quotactl_ops xfs_quotactl_operations;
74 static struct super_operations xfs_super_operations;
75 static kmem_zone_t *xfs_vnode_zone;
76 static kmem_zone_t *xfs_ioend_zone;
77 mempool_t *xfs_ioend_pool;
79 STATIC struct xfs_mount_args *
81 struct super_block *sb,
84 struct xfs_mount_args *args;
86 args = kzalloc(sizeof(struct xfs_mount_args), GFP_KERNEL);
90 args->logbufs = args->logbufsize = -1;
91 strncpy(args->fsname, sb->s_id, MAXNAMELEN);
93 /* Copy the already-parsed mount(2) flags we're interested in */
94 if (sb->s_flags & MS_DIRSYNC)
95 args->flags |= XFSMNT_DIRSYNC;
96 if (sb->s_flags & MS_SYNCHRONOUS)
97 args->flags |= XFSMNT_WSYNC;
99 args->flags |= XFSMNT_QUIET;
100 args->flags |= XFSMNT_32BITINODES;
105 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
106 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
107 #define MNTOPT_LOGDEV "logdev" /* log device */
108 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
109 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
110 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
111 #define MNTOPT_INO64 "ino64" /* force inodes into 64-bit range */
112 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
113 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
114 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
115 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
116 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
117 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
118 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
119 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
120 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
121 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
122 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
123 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
124 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
125 * unwritten extent conversion */
126 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
127 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
128 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
129 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
130 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
131 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
132 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
134 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
135 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
136 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
137 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
138 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
139 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
140 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
141 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
142 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
143 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
144 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
145 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
146 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
147 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
148 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
149 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
150 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
151 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
154 * Table driven mount option parser.
156 * Currently only used for remount, but it will be used for mount
157 * in the future, too.
160 Opt_barrier, Opt_nobarrier, Opt_err
163 static const match_table_t tokens = {
164 {Opt_barrier, "barrier"},
165 {Opt_nobarrier, "nobarrier"},
171 suffix_strtoul(char *s, char **endp, unsigned int base)
173 int last, shift_left_factor = 0;
176 last = strlen(value) - 1;
177 if (value[last] == 'K' || value[last] == 'k') {
178 shift_left_factor = 10;
181 if (value[last] == 'M' || value[last] == 'm') {
182 shift_left_factor = 20;
185 if (value[last] == 'G' || value[last] == 'g') {
186 shift_left_factor = 30;
190 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
195 struct xfs_mount *mp,
197 struct xfs_mount_args *args,
200 char *this_char, *value, *eov;
201 int dsunit, dswidth, vol_dsunit, vol_dswidth;
203 int dmapi_implies_ikeep = 1;
205 args->flags |= XFSMNT_BARRIER;
206 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
211 iosize = dsunit = dswidth = vol_dsunit = vol_dswidth = 0;
213 while ((this_char = strsep(&options, ",")) != NULL) {
216 if ((value = strchr(this_char, '=')) != NULL)
219 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
220 if (!value || !*value) {
222 "XFS: %s option requires an argument",
226 args->logbufs = simple_strtoul(value, &eov, 10);
227 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
228 if (!value || !*value) {
230 "XFS: %s option requires an argument",
234 args->logbufsize = suffix_strtoul(value, &eov, 10);
235 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
236 if (!value || !*value) {
238 "XFS: %s option requires an argument",
242 strncpy(args->logname, value, MAXNAMELEN);
243 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
244 if (!value || !*value) {
246 "XFS: %s option requires an argument",
250 strncpy(args->mtpt, value, MAXNAMELEN);
251 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
252 if (!value || !*value) {
254 "XFS: %s option requires an argument",
258 strncpy(args->rtname, value, MAXNAMELEN);
259 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
260 if (!value || !*value) {
262 "XFS: %s option requires an argument",
266 iosize = simple_strtoul(value, &eov, 10);
267 args->flags |= XFSMNT_IOSIZE;
268 args->iosizelog = (uint8_t) iosize;
269 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
270 if (!value || !*value) {
272 "XFS: %s option requires an argument",
276 iosize = suffix_strtoul(value, &eov, 10);
277 args->flags |= XFSMNT_IOSIZE;
278 args->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 args->flags |= XFSMNT_WSYNC;
287 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
288 args->flags |= XFSMNT_OSYNCISOSYNC;
289 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
290 args->flags |= XFSMNT_NORECOVERY;
291 } else if (!strcmp(this_char, MNTOPT_INO64)) {
292 args->flags |= XFSMNT_INO64;
295 "XFS: %s option not allowed on this system",
299 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
300 args->flags |= XFSMNT_NOALIGN;
301 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
302 args->flags |= XFSMNT_SWALLOC;
303 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
304 if (!value || !*value) {
306 "XFS: %s option requires an argument",
310 dsunit = simple_strtoul(value, &eov, 10);
311 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
312 if (!value || !*value) {
314 "XFS: %s option requires an argument",
318 dswidth = simple_strtoul(value, &eov, 10);
319 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
320 args->flags &= ~XFSMNT_32BITINODES;
323 "XFS: %s option not allowed on this system",
327 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
328 args->flags |= XFSMNT_NOUUID;
329 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
330 args->flags |= XFSMNT_BARRIER;
331 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
332 args->flags &= ~XFSMNT_BARRIER;
333 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
334 args->flags |= XFSMNT_IKEEP;
335 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
336 dmapi_implies_ikeep = 0;
337 args->flags &= ~XFSMNT_IKEEP;
338 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
339 args->flags2 &= ~XFSMNT2_COMPAT_IOSIZE;
340 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
341 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
342 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
343 args->flags |= XFSMNT_ATTR2;
344 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
345 args->flags &= ~XFSMNT_ATTR2;
346 args->flags |= XFSMNT_NOATTR2;
347 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
348 args->flags2 |= XFSMNT2_FILESTREAMS;
349 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
350 args->flags &= ~(XFSMNT_UQUOTAENF|XFSMNT_UQUOTA);
351 args->flags &= ~(XFSMNT_GQUOTAENF|XFSMNT_GQUOTA);
352 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
353 !strcmp(this_char, MNTOPT_UQUOTA) ||
354 !strcmp(this_char, MNTOPT_USRQUOTA)) {
355 args->flags |= XFSMNT_UQUOTA | XFSMNT_UQUOTAENF;
356 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
357 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
358 args->flags |= XFSMNT_UQUOTA;
359 args->flags &= ~XFSMNT_UQUOTAENF;
360 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
361 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
362 args->flags |= XFSMNT_PQUOTA | XFSMNT_PQUOTAENF;
363 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
364 args->flags |= XFSMNT_PQUOTA;
365 args->flags &= ~XFSMNT_PQUOTAENF;
366 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
367 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
368 args->flags |= XFSMNT_GQUOTA | XFSMNT_GQUOTAENF;
369 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
370 args->flags |= XFSMNT_GQUOTA;
371 args->flags &= ~XFSMNT_GQUOTAENF;
372 } else if (!strcmp(this_char, MNTOPT_DMAPI)) {
373 args->flags |= XFSMNT_DMAPI;
374 } else if (!strcmp(this_char, MNTOPT_XDSM)) {
375 args->flags |= XFSMNT_DMAPI;
376 } else if (!strcmp(this_char, MNTOPT_DMI)) {
377 args->flags |= XFSMNT_DMAPI;
378 } else if (!strcmp(this_char, "ihashsize")) {
380 "XFS: ihashsize no longer used, option is deprecated.");
381 } else if (!strcmp(this_char, "osyncisdsync")) {
382 /* no-op, this is now the default */
384 "XFS: osyncisdsync is now the default, option is deprecated.");
385 } else if (!strcmp(this_char, "irixsgid")) {
387 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
390 "XFS: unknown mount option [%s].", this_char);
395 if (args->flags & XFSMNT_NORECOVERY) {
396 if ((mp->m_flags & XFS_MOUNT_RDONLY) == 0) {
398 "XFS: no-recovery mounts must be read-only.");
403 if ((args->flags & XFSMNT_NOALIGN) && (dsunit || dswidth)) {
405 "XFS: sunit and swidth options incompatible with the noalign option");
409 if ((args->flags & XFSMNT_GQUOTA) && (args->flags & XFSMNT_PQUOTA)) {
411 "XFS: cannot mount with both project and group quota");
415 if ((args->flags & XFSMNT_DMAPI) && *args->mtpt == '\0') {
416 printk("XFS: %s option needs the mount point option as well\n",
421 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
423 "XFS: sunit and swidth must be specified together");
427 if (dsunit && (dswidth % dsunit != 0)) {
429 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
435 * Applications using DMI filesystems often expect the
436 * inode generation number to be monotonically increasing.
437 * If we delete inode chunks we break this assumption, so
438 * keep unused inode chunks on disk for DMI filesystems
439 * until we come up with a better solution.
440 * Note that if "ikeep" or "noikeep" mount options are
441 * supplied, then they are honored.
443 if ((args->flags & XFSMNT_DMAPI) && dmapi_implies_ikeep)
444 args->flags |= XFSMNT_IKEEP;
446 if ((args->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
448 args->sunit = dsunit;
449 args->flags |= XFSMNT_RETERR;
451 args->sunit = vol_dsunit;
453 dswidth ? (args->swidth = dswidth) :
454 (args->swidth = vol_dswidth);
456 args->sunit = args->swidth = 0;
460 if (args->flags & XFSMNT_32BITINODES)
461 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
463 args->flags |= XFSMNT_FLAGS2;
467 struct proc_xfs_info {
474 struct xfs_mount *mp,
477 static struct proc_xfs_info xfs_info_set[] = {
478 /* the few simple ones we can get from the mount struct */
479 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
480 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
481 { XFS_MOUNT_INO64, "," MNTOPT_INO64 },
482 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
483 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
484 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
485 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
486 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
487 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
488 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
489 { XFS_MOUNT_DMAPI, "," MNTOPT_DMAPI },
490 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
493 static struct proc_xfs_info xfs_info_unset[] = {
494 /* the few simple ones we can get from the mount struct */
495 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
496 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
497 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
500 struct proc_xfs_info *xfs_infop;
502 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
503 if (mp->m_flags & xfs_infop->flag)
504 seq_puts(m, xfs_infop->str);
506 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
507 if (!(mp->m_flags & xfs_infop->flag))
508 seq_puts(m, xfs_infop->str);
511 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
512 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
513 (int)(1 << mp->m_writeio_log) >> 10);
515 if (mp->m_logbufs > 0)
516 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
517 if (mp->m_logbsize > 0)
518 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
521 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
523 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
525 if (mp->m_dalign > 0)
526 seq_printf(m, "," MNTOPT_SUNIT "=%d",
527 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
528 if (mp->m_swidth > 0)
529 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
530 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
532 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
533 seq_puts(m, "," MNTOPT_USRQUOTA);
534 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
535 seq_puts(m, "," MNTOPT_UQUOTANOENF);
537 if (mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
538 seq_puts(m, "," MNTOPT_PRJQUOTA);
539 else if (mp->m_qflags & XFS_PQUOTA_ACCT)
540 seq_puts(m, "," MNTOPT_PQUOTANOENF);
542 if (mp->m_qflags & (XFS_GQUOTA_ACCT|XFS_OQUOTA_ENFD))
543 seq_puts(m, "," MNTOPT_GRPQUOTA);
544 else if (mp->m_qflags & XFS_GQUOTA_ACCT)
545 seq_puts(m, "," MNTOPT_GQUOTANOENF);
547 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
548 seq_puts(m, "," MNTOPT_NOQUOTA);
554 unsigned int blockshift)
556 unsigned int pagefactor = 1;
557 unsigned int bitshift = BITS_PER_LONG - 1;
559 /* Figure out maximum filesize, on Linux this can depend on
560 * the filesystem blocksize (on 32 bit platforms).
561 * __block_prepare_write does this in an [unsigned] long...
562 * page->index << (PAGE_CACHE_SHIFT - bbits)
563 * So, for page sized blocks (4K on 32 bit platforms),
564 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
565 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
566 * but for smaller blocksizes it is less (bbits = log2 bsize).
567 * Note1: get_block_t takes a long (implicit cast from above)
568 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
569 * can optionally convert the [unsigned] long from above into
570 * an [unsigned] long long.
573 #if BITS_PER_LONG == 32
574 # if defined(CONFIG_LBD)
575 ASSERT(sizeof(sector_t) == 8);
576 pagefactor = PAGE_CACHE_SIZE;
577 bitshift = BITS_PER_LONG;
579 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
583 return (((__uint64_t)pagefactor) << bitshift) - 1;
590 struct block_device **bdevp)
594 *bdevp = open_bdev_exclusive(name, FMODE_READ|FMODE_WRITE, mp);
595 if (IS_ERR(*bdevp)) {
596 error = PTR_ERR(*bdevp);
597 printk("XFS: Invalid device [%s], error=%d\n", name, error);
605 struct block_device *bdev)
608 close_bdev_exclusive(bdev, FMODE_READ|FMODE_WRITE);
612 * Try to write out the superblock using barriers.
618 xfs_buf_t *sbp = xfs_getsb(mp, 0);
623 XFS_BUF_UNDELAYWRITE(sbp);
625 XFS_BUF_UNASYNC(sbp);
626 XFS_BUF_ORDERED(sbp);
629 error = xfs_iowait(sbp);
632 * Clear all the flags we set and possible error state in the
633 * buffer. We only did the write to try out whether barriers
634 * worked and shouldn't leave any traces in the superblock
638 XFS_BUF_ERROR(sbp, 0);
639 XFS_BUF_UNORDERED(sbp);
646 xfs_mountfs_check_barriers(xfs_mount_t *mp)
650 if (mp->m_logdev_targp != mp->m_ddev_targp) {
651 xfs_fs_cmn_err(CE_NOTE, mp,
652 "Disabling barriers, not supported with external log device");
653 mp->m_flags &= ~XFS_MOUNT_BARRIER;
657 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
658 xfs_fs_cmn_err(CE_NOTE, mp,
659 "Disabling barriers, underlying device is readonly");
660 mp->m_flags &= ~XFS_MOUNT_BARRIER;
664 error = xfs_barrier_test(mp);
666 xfs_fs_cmn_err(CE_NOTE, mp,
667 "Disabling barriers, trial barrier write failed");
668 mp->m_flags &= ~XFS_MOUNT_BARRIER;
674 xfs_blkdev_issue_flush(
675 xfs_buftarg_t *buftarg)
677 blkdev_issue_flush(buftarg->bt_bdev, NULL);
682 struct xfs_mount *mp)
684 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
685 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
686 xfs_free_buftarg(mp->m_logdev_targp);
687 xfs_blkdev_put(logdev);
689 if (mp->m_rtdev_targp) {
690 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
691 xfs_free_buftarg(mp->m_rtdev_targp);
692 xfs_blkdev_put(rtdev);
694 xfs_free_buftarg(mp->m_ddev_targp);
698 * The file system configurations are:
699 * (1) device (partition) with data and internal log
700 * (2) logical volume with data and log subvolumes.
701 * (3) logical volume with data, log, and realtime subvolumes.
703 * We only have to handle opening the log and realtime volumes here if
704 * they are present. The data subvolume has already been opened by
705 * get_sb_bdev() and is stored in sb->s_bdev.
709 struct xfs_mount *mp,
710 struct xfs_mount_args *args)
712 struct block_device *ddev = mp->m_super->s_bdev;
713 struct block_device *logdev = NULL, *rtdev = NULL;
717 * Open real time and log devices - order is important.
719 if (args->logname[0]) {
720 error = xfs_blkdev_get(mp, args->logname, &logdev);
725 if (args->rtname[0]) {
726 error = xfs_blkdev_get(mp, args->rtname, &rtdev);
728 goto out_close_logdev;
730 if (rtdev == ddev || rtdev == logdev) {
732 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
734 goto out_close_rtdev;
739 * Setup xfs_mount buffer target pointers
742 mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0);
743 if (!mp->m_ddev_targp)
744 goto out_close_rtdev;
747 mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1);
748 if (!mp->m_rtdev_targp)
749 goto out_free_ddev_targ;
752 if (logdev && logdev != ddev) {
753 mp->m_logdev_targp = xfs_alloc_buftarg(logdev, 1);
754 if (!mp->m_logdev_targp)
755 goto out_free_rtdev_targ;
757 mp->m_logdev_targp = mp->m_ddev_targp;
763 if (mp->m_rtdev_targp)
764 xfs_free_buftarg(mp->m_rtdev_targp);
766 xfs_free_buftarg(mp->m_ddev_targp);
769 xfs_blkdev_put(rtdev);
771 if (logdev && logdev != ddev)
772 xfs_blkdev_put(logdev);
778 * Setup xfs_mount buffer target pointers based on superblock
782 struct xfs_mount *mp)
786 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
787 mp->m_sb.sb_sectsize);
791 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
792 unsigned int log_sector_size = BBSIZE;
794 if (xfs_sb_version_hassector(&mp->m_sb))
795 log_sector_size = mp->m_sb.sb_logsectsize;
796 error = xfs_setsize_buftarg(mp->m_logdev_targp,
797 mp->m_sb.sb_blocksize,
802 if (mp->m_rtdev_targp) {
803 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
804 mp->m_sb.sb_blocksize,
805 mp->m_sb.sb_sectsize);
814 * XFS AIL push thread support
819 xfs_lsn_t threshold_lsn)
821 mp->m_ail.xa_target = threshold_lsn;
822 wake_up_process(mp->m_ail.xa_task);
829 xfs_mount_t *mp = (xfs_mount_t *)data;
830 xfs_lsn_t last_pushed_lsn = 0;
833 while (!kthread_should_stop()) {
835 schedule_timeout_interruptible(msecs_to_jiffies(tout));
842 if (XFS_FORCED_SHUTDOWN(mp))
845 tout = xfsaild_push(mp, &last_pushed_lsn);
855 mp->m_ail.xa_target = 0;
856 mp->m_ail.xa_task = kthread_run(xfsaild, mp, "xfsaild");
857 if (IS_ERR(mp->m_ail.xa_task))
858 return -PTR_ERR(mp->m_ail.xa_task);
866 kthread_stop(mp->m_ail.xa_task);
871 STATIC struct inode *
873 struct super_block *sb)
875 return kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP);
879 xfs_fs_destroy_inode(
882 kmem_zone_free(xfs_vnode_zone, inode);
886 xfs_fs_inode_init_once(
889 inode_init_once((struct inode *)vnode);
894 * Slab object creation initialisation for the XFS inode.
895 * This covers only the idempotent fields in the XFS inode;
896 * all other fields need to be initialised on allocation
897 * from the slab. This avoids the need to repeatedly intialise
898 * fields in the xfs inode that left in the initialise state
899 * when freeing the inode.
905 struct xfs_inode *ip = inode;
907 memset(ip, 0, sizeof(struct xfs_inode));
908 atomic_set(&ip->i_iocount, 0);
909 atomic_set(&ip->i_pincount, 0);
910 spin_lock_init(&ip->i_flags_lock);
911 INIT_LIST_HEAD(&ip->i_reclaim);
912 init_waitqueue_head(&ip->i_ipin_wait);
914 * Because we want to use a counting completion, complete
915 * the flush completion once to allow a single access to
916 * the flush completion without blocking.
918 init_completion(&ip->i_flush);
919 complete(&ip->i_flush);
921 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
922 "xfsino", ip->i_ino);
923 mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
927 * Attempt to flush the inode, this will actually fail
928 * if the inode is pinned, but we dirty the inode again
929 * at the point when it is unpinned after a log write,
930 * since this is when the inode itself becomes flushable.
940 xfs_itrace_entry(XFS_I(inode));
942 filemap_fdatawait(inode->i_mapping);
945 error = xfs_inode_flush(XFS_I(inode), flags);
947 * if we failed to write out the inode then mark
948 * it dirty again so we'll try again later.
951 mark_inode_dirty_sync(inode);
960 xfs_inode_t *ip = XFS_I(inode);
963 * ip can be null when xfs_iget_core calls xfs_idestroy if we
964 * find an inode with di_mode == 0 but without IGET_CREATE set.
967 xfs_itrace_entry(ip);
968 XFS_STATS_INC(vn_rele);
969 XFS_STATS_INC(vn_remove);
970 XFS_STATS_INC(vn_reclaim);
971 XFS_STATS_DEC(vn_active);
974 xfs_iflags_clear(ip, XFS_IMODIFIED);
976 panic("%s: cannot reclaim 0x%p\n", __func__, inode);
979 ASSERT(XFS_I(inode) == NULL);
984 struct xfs_mount *mp)
988 kfree(mp->m_logname);
993 struct super_block *sb)
995 struct xfs_mount *mp = XFS_M(sb);
996 struct xfs_inode *rip = mp->m_rootip;
997 int unmount_event_flags = 0;
1001 xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI);
1004 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1005 unmount_event_flags =
1006 (mp->m_dmevmask & (1 << DM_EVENT_UNMOUNT)) ?
1007 0 : DM_FLAGS_UNWANTED;
1009 * Ignore error from dmapi here, first unmount is not allowed
1010 * to fail anyway, and second we wouldn't want to fail a
1011 * unmount because of dmapi.
1013 XFS_SEND_PREUNMOUNT(mp, rip, DM_RIGHT_NULL, rip, DM_RIGHT_NULL,
1014 NULL, NULL, 0, 0, unmount_event_flags);
1019 * Blow away any referenced inode in the filestreams cache.
1020 * This can and will cause log traffic as inodes go inactive
1023 xfs_filestream_unmount(mp);
1025 XFS_bflush(mp->m_ddev_targp);
1026 error = xfs_unmount_flush(mp, 0);
1030 * If we're forcing a shutdown, typically because of a media error,
1031 * we want to make sure we invalidate dirty pages that belong to
1032 * referenced vnodes as well.
1034 if (XFS_FORCED_SHUTDOWN(mp)) {
1035 error = xfs_sync(mp, SYNC_WAIT | SYNC_CLOSE);
1036 ASSERT(error != EFSCORRUPTED);
1039 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1040 XFS_SEND_UNMOUNT(mp, rip, DM_RIGHT_NULL, 0, 0,
1041 unmount_event_flags);
1046 xfs_icsb_destroy_counters(mp);
1047 xfs_close_devices(mp);
1050 xfs_free_fsname(mp);
1056 struct super_block *sb)
1058 if (!(sb->s_flags & MS_RDONLY))
1059 xfs_sync(XFS_M(sb), SYNC_FSDATA);
1065 struct super_block *sb,
1068 struct xfs_mount *mp = XFS_M(sb);
1073 * Treat a sync operation like a freeze. This is to work
1074 * around a race in sync_inodes() which works in two phases
1075 * - an asynchronous flush, which can write out an inode
1076 * without waiting for file size updates to complete, and a
1077 * synchronous flush, which wont do anything because the
1078 * async flush removed the inode's dirty flag. Also
1079 * sync_inodes() will not see any files that just have
1080 * outstanding transactions to be flushed because we don't
1081 * dirty the Linux inode until after the transaction I/O
1084 if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE)) {
1086 * First stage of freeze - no more writers will make progress
1087 * now we are here, so we flush delwri and delalloc buffers
1088 * here, then wait for all I/O to complete. Data is frozen at
1089 * that point. Metadata is not frozen, transactions can still
1090 * occur here so don't bother flushing the buftarg (i.e
1091 * SYNC_QUIESCE) because it'll just get dirty again.
1093 flags = SYNC_DATA_QUIESCE;
1095 flags = SYNC_FSDATA;
1097 error = xfs_sync(mp, flags);
1100 if (unlikely(laptop_mode)) {
1101 int prev_sync_seq = mp->m_sync_seq;
1104 * The disk must be active because we're syncing.
1105 * We schedule xfssyncd now (now that the disk is
1106 * active) instead of later (when it might not be).
1108 wake_up_process(mp->m_sync_task);
1110 * We have to wait for the sync iteration to complete.
1111 * If we don't, the disk activity caused by the sync
1112 * will come after the sync is completed, and that
1113 * triggers another sync from laptop mode.
1115 wait_event(mp->m_wait_single_sync_task,
1116 mp->m_sync_seq != prev_sync_seq);
1124 struct dentry *dentry,
1125 struct kstatfs *statp)
1127 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1128 xfs_sb_t *sbp = &mp->m_sb;
1129 __uint64_t fakeinos, id;
1132 statp->f_type = XFS_SB_MAGIC;
1133 statp->f_namelen = MAXNAMELEN - 1;
1135 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1136 statp->f_fsid.val[0] = (u32)id;
1137 statp->f_fsid.val[1] = (u32)(id >> 32);
1139 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1141 spin_lock(&mp->m_sb_lock);
1142 statp->f_bsize = sbp->sb_blocksize;
1143 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1144 statp->f_blocks = sbp->sb_dblocks - lsize;
1145 statp->f_bfree = statp->f_bavail =
1146 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1147 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1149 fakeinos += mp->m_inoadd;
1152 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1153 if (mp->m_maxicount)
1157 statp->f_files = min_t(typeof(statp->f_files),
1160 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1161 spin_unlock(&mp->m_sb_lock);
1163 XFS_QM_DQSTATVFS(XFS_I(dentry->d_inode), statp);
1169 struct super_block *sb,
1173 struct xfs_mount *mp = XFS_M(sb);
1174 substring_t args[MAX_OPT_ARGS];
1177 while ((p = strsep(&options, ",")) != NULL) {
1183 token = match_token(p, tokens, args);
1186 mp->m_flags |= XFS_MOUNT_BARRIER;
1189 * Test if barriers are actually working if we can,
1190 * else delay this check until the filesystem is
1193 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1194 xfs_mountfs_check_barriers(mp);
1197 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1201 * Logically we would return an error here to prevent
1202 * users from believing they might have changed
1203 * mount options using remount which can't be changed.
1205 * But unfortunately mount(8) adds all options from
1206 * mtab and fstab to the mount arguments in some cases
1207 * so we can't blindly reject options, but have to
1208 * check for each specified option if it actually
1209 * differs from the currently set option and only
1210 * reject it if that's the case.
1212 * Until that is implemented we return success for
1213 * every remount request, and silently ignore all
1214 * options that we can't actually change.
1218 "XFS: mount option \"%s\" not supported for remount\n", p);
1227 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1228 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1229 if (mp->m_flags & XFS_MOUNT_BARRIER)
1230 xfs_mountfs_check_barriers(mp);
1234 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1235 xfs_filestream_flush(mp);
1236 xfs_sync(mp, SYNC_DATA_QUIESCE);
1237 xfs_attr_quiesce(mp);
1238 mp->m_flags |= XFS_MOUNT_RDONLY;
1245 * Second stage of a freeze. The data is already frozen so we only
1246 * need to take care of themetadata. Once that's done write a dummy
1247 * record to dirty the log in case of a crash while frozen.
1251 struct super_block *sb)
1253 struct xfs_mount *mp = XFS_M(sb);
1255 xfs_attr_quiesce(mp);
1256 xfs_fs_log_dummy(mp);
1260 xfs_fs_show_options(
1262 struct vfsmount *mnt)
1264 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1269 struct super_block *sb,
1272 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XQUOTASYNC, 0, NULL);
1277 struct super_block *sb,
1278 struct fs_quota_stat *fqs)
1280 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XGETQSTAT, 0, (caddr_t)fqs);
1285 struct super_block *sb,
1289 return -XFS_QM_QUOTACTL(XFS_M(sb), op, 0, (caddr_t)&flags);
1294 struct super_block *sb,
1297 struct fs_disk_quota *fdq)
1299 return -XFS_QM_QUOTACTL(XFS_M(sb),
1300 (type == USRQUOTA) ? Q_XGETQUOTA :
1301 ((type == GRPQUOTA) ? Q_XGETGQUOTA :
1302 Q_XGETPQUOTA), id, (caddr_t)fdq);
1307 struct super_block *sb,
1310 struct fs_disk_quota *fdq)
1312 return -XFS_QM_QUOTACTL(XFS_M(sb),
1313 (type == USRQUOTA) ? Q_XSETQLIM :
1314 ((type == GRPQUOTA) ? Q_XSETGQLIM :
1315 Q_XSETPQLIM), id, (caddr_t)fdq);
1319 * This function fills in xfs_mount_t fields based on mount args.
1320 * Note: the superblock has _not_ yet been read in.
1324 struct xfs_mount_args *ap,
1325 struct xfs_mount *mp)
1329 /* Values are in BBs */
1330 if ((ap->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
1332 * At this point the superblock has not been read
1333 * in, therefore we do not know the block size.
1334 * Before the mount call ends we will convert
1337 mp->m_dalign = ap->sunit;
1338 mp->m_swidth = ap->swidth;
1341 if (ap->logbufs != -1 &&
1343 (ap->logbufs < XLOG_MIN_ICLOGS ||
1344 ap->logbufs > XLOG_MAX_ICLOGS)) {
1346 "XFS: invalid logbufs value: %d [not %d-%d]",
1347 ap->logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
1348 return XFS_ERROR(EINVAL);
1350 mp->m_logbufs = ap->logbufs;
1351 if (ap->logbufsize != -1 &&
1352 ap->logbufsize != 0 &&
1353 (ap->logbufsize < XLOG_MIN_RECORD_BSIZE ||
1354 ap->logbufsize > XLOG_MAX_RECORD_BSIZE ||
1355 !is_power_of_2(ap->logbufsize))) {
1357 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
1359 return XFS_ERROR(EINVAL);
1364 mp->m_logbsize = ap->logbufsize;
1365 mp->m_fsname_len = strlen(ap->fsname) + 1;
1367 mp->m_fsname = kstrdup(ap->fsname, GFP_KERNEL);
1371 if (ap->rtname[0]) {
1372 mp->m_rtname = kstrdup(ap->rtname, GFP_KERNEL);
1374 goto out_free_fsname;
1378 if (ap->logname[0]) {
1379 mp->m_logname = kstrdup(ap->logname, GFP_KERNEL);
1381 goto out_free_rtname;
1384 if (ap->flags & XFSMNT_WSYNC)
1385 mp->m_flags |= XFS_MOUNT_WSYNC;
1387 if (ap->flags & XFSMNT_INO64) {
1388 mp->m_flags |= XFS_MOUNT_INO64;
1389 mp->m_inoadd = XFS_INO64_OFFSET;
1392 if (ap->flags & XFSMNT_RETERR)
1393 mp->m_flags |= XFS_MOUNT_RETERR;
1394 if (ap->flags & XFSMNT_NOALIGN)
1395 mp->m_flags |= XFS_MOUNT_NOALIGN;
1396 if (ap->flags & XFSMNT_SWALLOC)
1397 mp->m_flags |= XFS_MOUNT_SWALLOC;
1398 if (ap->flags & XFSMNT_OSYNCISOSYNC)
1399 mp->m_flags |= XFS_MOUNT_OSYNCISOSYNC;
1400 if (ap->flags & XFSMNT_32BITINODES)
1401 mp->m_flags |= XFS_MOUNT_32BITINODES;
1403 if (ap->flags & XFSMNT_IOSIZE) {
1404 if (ap->iosizelog > XFS_MAX_IO_LOG ||
1405 ap->iosizelog < XFS_MIN_IO_LOG) {
1407 "XFS: invalid log iosize: %d [not %d-%d]",
1408 ap->iosizelog, XFS_MIN_IO_LOG,
1410 return XFS_ERROR(EINVAL);
1413 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
1414 mp->m_readio_log = mp->m_writeio_log = ap->iosizelog;
1417 if (ap->flags & XFSMNT_IKEEP)
1418 mp->m_flags |= XFS_MOUNT_IKEEP;
1419 if (ap->flags & XFSMNT_DIRSYNC)
1420 mp->m_flags |= XFS_MOUNT_DIRSYNC;
1421 if (ap->flags & XFSMNT_ATTR2)
1422 mp->m_flags |= XFS_MOUNT_ATTR2;
1423 if (ap->flags & XFSMNT_NOATTR2)
1424 mp->m_flags |= XFS_MOUNT_NOATTR2;
1426 if (ap->flags2 & XFSMNT2_COMPAT_IOSIZE)
1427 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
1430 * no recovery flag requires a read-only mount
1432 if (ap->flags & XFSMNT_NORECOVERY) {
1433 if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
1435 "XFS: tried to mount a FS read-write without recovery!");
1436 return XFS_ERROR(EINVAL);
1438 mp->m_flags |= XFS_MOUNT_NORECOVERY;
1441 if (ap->flags & XFSMNT_NOUUID)
1442 mp->m_flags |= XFS_MOUNT_NOUUID;
1443 if (ap->flags & XFSMNT_BARRIER)
1444 mp->m_flags |= XFS_MOUNT_BARRIER;
1446 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1448 if (ap->flags2 & XFSMNT2_FILESTREAMS)
1449 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
1451 if (ap->flags & XFSMNT_DMAPI)
1452 mp->m_flags |= XFS_MOUNT_DMAPI;
1457 kfree(mp->m_rtname);
1459 kfree(mp->m_fsname);
1465 * This function fills in xfs_mount_t fields based on mount args.
1466 * Note: the superblock _has_ now been read in.
1470 struct xfs_mount_args *ap,
1471 struct xfs_mount *mp)
1473 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1475 /* Fail a mount where the logbuf is smaller then the log stripe */
1476 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1477 if ((ap->logbufsize <= 0) &&
1478 (mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE)) {
1479 mp->m_logbsize = mp->m_sb.sb_logsunit;
1480 } else if (ap->logbufsize > 0 &&
1481 ap->logbufsize < mp->m_sb.sb_logsunit) {
1483 "XFS: logbuf size must be greater than or equal to log stripe size");
1484 return XFS_ERROR(EINVAL);
1487 /* Fail a mount if the logbuf is larger than 32K */
1488 if (ap->logbufsize > XLOG_BIG_RECORD_BSIZE) {
1490 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1491 return XFS_ERROR(EINVAL);
1496 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1497 * told by noattr2 to turn it off
1499 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1500 !(ap->flags & XFSMNT_NOATTR2))
1501 mp->m_flags |= XFS_MOUNT_ATTR2;
1504 * prohibit r/w mounts of read-only filesystems
1506 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1508 "XFS: cannot mount a read-only filesystem as read-write");
1509 return XFS_ERROR(EROFS);
1513 * check for shared mount.
1515 if (ap->flags & XFSMNT_SHARED) {
1516 if (!xfs_sb_version_hasshared(&mp->m_sb))
1517 return XFS_ERROR(EINVAL);
1520 * For IRIX 6.5, shared mounts must have the shared
1521 * version bit set, have the persistent readonly
1522 * field set, must be version 0 and can only be mounted
1525 if (!ronly || !(mp->m_sb.sb_flags & XFS_SBF_READONLY) ||
1526 (mp->m_sb.sb_shared_vn != 0))
1527 return XFS_ERROR(EINVAL);
1529 mp->m_flags |= XFS_MOUNT_SHARED;
1532 * Shared XFS V0 can't deal with DMI. Return EINVAL.
1534 if (mp->m_sb.sb_shared_vn == 0 && (ap->flags & XFSMNT_DMAPI))
1535 return XFS_ERROR(EINVAL);
1538 if (ap->flags & XFSMNT_UQUOTA) {
1539 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
1540 if (ap->flags & XFSMNT_UQUOTAENF)
1541 mp->m_qflags |= XFS_UQUOTA_ENFD;
1544 if (ap->flags & XFSMNT_GQUOTA) {
1545 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
1546 if (ap->flags & XFSMNT_GQUOTAENF)
1547 mp->m_qflags |= XFS_OQUOTA_ENFD;
1548 } else if (ap->flags & XFSMNT_PQUOTA) {
1549 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
1550 if (ap->flags & XFSMNT_PQUOTAENF)
1551 mp->m_qflags |= XFS_OQUOTA_ENFD;
1559 struct super_block *sb,
1564 struct xfs_mount *mp = NULL;
1565 struct xfs_mount_args *args;
1566 int flags = 0, error = ENOMEM;
1568 args = xfs_args_allocate(sb, silent);
1572 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1576 spin_lock_init(&mp->m_sb_lock);
1577 mutex_init(&mp->m_ilock);
1578 mutex_init(&mp->m_growlock);
1579 atomic_set(&mp->m_active_trans, 0);
1580 INIT_LIST_HEAD(&mp->m_sync_list);
1581 spin_lock_init(&mp->m_sync_lock);
1582 init_waitqueue_head(&mp->m_wait_single_sync_task);
1587 if (sb->s_flags & MS_RDONLY)
1588 mp->m_flags |= XFS_MOUNT_RDONLY;
1590 error = xfs_parseargs(mp, (char *)data, args, 0);
1594 sb_min_blocksize(sb, BBSIZE);
1595 sb->s_xattr = xfs_xattr_handlers;
1596 sb->s_export_op = &xfs_export_operations;
1597 sb->s_qcop = &xfs_quotactl_operations;
1598 sb->s_op = &xfs_super_operations;
1600 error = xfs_dmops_get(mp, args);
1603 error = xfs_qmops_get(mp, args);
1607 if (args->flags & XFSMNT_QUIET)
1608 flags |= XFS_MFSI_QUIET;
1610 error = xfs_open_devices(mp, args);
1614 if (xfs_icsb_init_counters(mp))
1615 mp->m_flags |= XFS_MOUNT_NO_PERCPU_SB;
1618 * Setup flags based on mount(2) options and then the superblock
1620 error = xfs_start_flags(args, mp);
1622 goto out_free_fsname;
1623 error = xfs_readsb(mp, flags);
1625 goto out_free_fsname;
1626 error = xfs_finish_flags(args, mp);
1630 error = xfs_setup_devices(mp);
1634 if (mp->m_flags & XFS_MOUNT_BARRIER)
1635 xfs_mountfs_check_barriers(mp);
1637 error = xfs_filestream_mount(mp);
1641 error = xfs_mountfs(mp);
1643 goto out_filestream_unmount;
1645 XFS_SEND_MOUNT(mp, DM_RIGHT_NULL, args->mtpt, args->fsname);
1648 sb->s_magic = XFS_SB_MAGIC;
1649 sb->s_blocksize = mp->m_sb.sb_blocksize;
1650 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1651 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1652 sb->s_time_gran = 1;
1653 set_posix_acl_flag(sb);
1655 root = igrab(VFS_I(mp->m_rootip));
1660 if (is_bad_inode(root)) {
1664 sb->s_root = d_alloc_root(root);
1670 error = xfs_syncd_init(mp);
1674 xfs_itrace_exit(XFS_I(sb->s_root->d_inode));
1679 out_filestream_unmount:
1680 xfs_filestream_unmount(mp);
1684 xfs_free_fsname(mp);
1685 xfs_icsb_destroy_counters(mp);
1686 xfs_close_devices(mp);
1707 * Blow away any referenced inode in the filestreams cache.
1708 * This can and will cause log traffic as inodes go inactive
1711 xfs_filestream_unmount(mp);
1713 XFS_bflush(mp->m_ddev_targp);
1714 error = xfs_unmount_flush(mp, 0);
1723 struct file_system_type *fs_type,
1725 const char *dev_name,
1727 struct vfsmount *mnt)
1729 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1733 static struct super_operations xfs_super_operations = {
1734 .alloc_inode = xfs_fs_alloc_inode,
1735 .destroy_inode = xfs_fs_destroy_inode,
1736 .write_inode = xfs_fs_write_inode,
1737 .clear_inode = xfs_fs_clear_inode,
1738 .put_super = xfs_fs_put_super,
1739 .write_super = xfs_fs_write_super,
1740 .sync_fs = xfs_fs_sync_super,
1741 .write_super_lockfs = xfs_fs_lockfs,
1742 .statfs = xfs_fs_statfs,
1743 .remount_fs = xfs_fs_remount,
1744 .show_options = xfs_fs_show_options,
1747 static struct quotactl_ops xfs_quotactl_operations = {
1748 .quota_sync = xfs_fs_quotasync,
1749 .get_xstate = xfs_fs_getxstate,
1750 .set_xstate = xfs_fs_setxstate,
1751 .get_xquota = xfs_fs_getxquota,
1752 .set_xquota = xfs_fs_setxquota,
1755 static struct file_system_type xfs_fs_type = {
1756 .owner = THIS_MODULE,
1758 .get_sb = xfs_fs_get_sb,
1759 .kill_sb = kill_block_super,
1760 .fs_flags = FS_REQUIRES_DEV,
1764 xfs_alloc_trace_bufs(void)
1766 #ifdef XFS_ALLOC_TRACE
1767 xfs_alloc_trace_buf = ktrace_alloc(XFS_ALLOC_TRACE_SIZE, KM_MAYFAIL);
1768 if (!xfs_alloc_trace_buf)
1771 #ifdef XFS_BMAP_TRACE
1772 xfs_bmap_trace_buf = ktrace_alloc(XFS_BMAP_TRACE_SIZE, KM_MAYFAIL);
1773 if (!xfs_bmap_trace_buf)
1774 goto out_free_alloc_trace;
1776 #ifdef XFS_BTREE_TRACE
1777 xfs_allocbt_trace_buf = ktrace_alloc(XFS_ALLOCBT_TRACE_SIZE,
1779 if (!xfs_allocbt_trace_buf)
1780 goto out_free_bmap_trace;
1782 xfs_inobt_trace_buf = ktrace_alloc(XFS_INOBT_TRACE_SIZE, KM_MAYFAIL);
1783 if (!xfs_inobt_trace_buf)
1784 goto out_free_allocbt_trace;
1786 xfs_bmbt_trace_buf = ktrace_alloc(XFS_BMBT_TRACE_SIZE, KM_MAYFAIL);
1787 if (!xfs_bmbt_trace_buf)
1788 goto out_free_inobt_trace;
1790 #ifdef XFS_ATTR_TRACE
1791 xfs_attr_trace_buf = ktrace_alloc(XFS_ATTR_TRACE_SIZE, KM_MAYFAIL);
1792 if (!xfs_attr_trace_buf)
1793 goto out_free_bmbt_trace;
1795 #ifdef XFS_DIR2_TRACE
1796 xfs_dir2_trace_buf = ktrace_alloc(XFS_DIR2_GTRACE_SIZE, KM_MAYFAIL);
1797 if (!xfs_dir2_trace_buf)
1798 goto out_free_attr_trace;
1803 #ifdef XFS_DIR2_TRACE
1804 out_free_attr_trace:
1806 #ifdef XFS_ATTR_TRACE
1807 ktrace_free(xfs_attr_trace_buf);
1808 out_free_bmbt_trace:
1810 #ifdef XFS_BTREE_TRACE
1811 ktrace_free(xfs_bmbt_trace_buf);
1812 out_free_inobt_trace:
1813 ktrace_free(xfs_inobt_trace_buf);
1814 out_free_allocbt_trace:
1815 ktrace_free(xfs_allocbt_trace_buf);
1816 out_free_bmap_trace:
1818 #ifdef XFS_BMAP_TRACE
1819 ktrace_free(xfs_bmap_trace_buf);
1820 out_free_alloc_trace:
1822 #ifdef XFS_ALLOC_TRACE
1823 ktrace_free(xfs_alloc_trace_buf);
1830 xfs_free_trace_bufs(void)
1832 #ifdef XFS_DIR2_TRACE
1833 ktrace_free(xfs_dir2_trace_buf);
1835 #ifdef XFS_ATTR_TRACE
1836 ktrace_free(xfs_attr_trace_buf);
1838 #ifdef XFS_BTREE_TRACE
1839 ktrace_free(xfs_bmbt_trace_buf);
1840 ktrace_free(xfs_inobt_trace_buf);
1841 ktrace_free(xfs_allocbt_trace_buf);
1843 #ifdef XFS_BMAP_TRACE
1844 ktrace_free(xfs_bmap_trace_buf);
1846 #ifdef XFS_ALLOC_TRACE
1847 ktrace_free(xfs_alloc_trace_buf);
1852 xfs_init_zones(void)
1854 xfs_vnode_zone = kmem_zone_init_flags(sizeof(struct inode), "xfs_vnode",
1855 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM |
1857 xfs_fs_inode_init_once);
1858 if (!xfs_vnode_zone)
1861 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1862 if (!xfs_ioend_zone)
1863 goto out_destroy_vnode_zone;
1865 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1867 if (!xfs_ioend_pool)
1868 goto out_destroy_ioend_zone;
1870 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1872 if (!xfs_log_ticket_zone)
1873 goto out_destroy_ioend_pool;
1875 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1876 "xfs_bmap_free_item");
1877 if (!xfs_bmap_free_item_zone)
1878 goto out_destroy_log_ticket_zone;
1879 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1881 if (!xfs_btree_cur_zone)
1882 goto out_destroy_bmap_free_item_zone;
1884 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1886 if (!xfs_da_state_zone)
1887 goto out_destroy_btree_cur_zone;
1889 xfs_dabuf_zone = kmem_zone_init(sizeof(xfs_dabuf_t), "xfs_dabuf");
1890 if (!xfs_dabuf_zone)
1891 goto out_destroy_da_state_zone;
1893 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1894 if (!xfs_ifork_zone)
1895 goto out_destroy_dabuf_zone;
1897 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1898 if (!xfs_trans_zone)
1899 goto out_destroy_ifork_zone;
1902 * The size of the zone allocated buf log item is the maximum
1903 * size possible under XFS. This wastes a little bit of memory,
1904 * but it is much faster.
1906 xfs_buf_item_zone = kmem_zone_init((sizeof(xfs_buf_log_item_t) +
1907 (((XFS_MAX_BLOCKSIZE / XFS_BLI_CHUNK) /
1908 NBWORD) * sizeof(int))), "xfs_buf_item");
1909 if (!xfs_buf_item_zone)
1910 goto out_destroy_trans_zone;
1912 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1913 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1914 sizeof(xfs_extent_t))), "xfs_efd_item");
1916 goto out_destroy_buf_item_zone;
1918 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1919 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1920 sizeof(xfs_extent_t))), "xfs_efi_item");
1922 goto out_destroy_efd_zone;
1925 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1926 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM |
1927 KM_ZONE_SPREAD, xfs_inode_init_once);
1928 if (!xfs_inode_zone)
1929 goto out_destroy_efi_zone;
1932 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1933 KM_ZONE_SPREAD, NULL);
1935 goto out_destroy_inode_zone;
1937 #ifdef CONFIG_XFS_POSIX_ACL
1938 xfs_acl_zone = kmem_zone_init(sizeof(xfs_acl_t), "xfs_acl");
1940 goto out_destroy_ili_zone;
1945 #ifdef CONFIG_XFS_POSIX_ACL
1946 out_destroy_ili_zone:
1948 kmem_zone_destroy(xfs_ili_zone);
1949 out_destroy_inode_zone:
1950 kmem_zone_destroy(xfs_inode_zone);
1951 out_destroy_efi_zone:
1952 kmem_zone_destroy(xfs_efi_zone);
1953 out_destroy_efd_zone:
1954 kmem_zone_destroy(xfs_efd_zone);
1955 out_destroy_buf_item_zone:
1956 kmem_zone_destroy(xfs_buf_item_zone);
1957 out_destroy_trans_zone:
1958 kmem_zone_destroy(xfs_trans_zone);
1959 out_destroy_ifork_zone:
1960 kmem_zone_destroy(xfs_ifork_zone);
1961 out_destroy_dabuf_zone:
1962 kmem_zone_destroy(xfs_dabuf_zone);
1963 out_destroy_da_state_zone:
1964 kmem_zone_destroy(xfs_da_state_zone);
1965 out_destroy_btree_cur_zone:
1966 kmem_zone_destroy(xfs_btree_cur_zone);
1967 out_destroy_bmap_free_item_zone:
1968 kmem_zone_destroy(xfs_bmap_free_item_zone);
1969 out_destroy_log_ticket_zone:
1970 kmem_zone_destroy(xfs_log_ticket_zone);
1971 out_destroy_ioend_pool:
1972 mempool_destroy(xfs_ioend_pool);
1973 out_destroy_ioend_zone:
1974 kmem_zone_destroy(xfs_ioend_zone);
1975 out_destroy_vnode_zone:
1976 kmem_zone_destroy(xfs_vnode_zone);
1982 xfs_destroy_zones(void)
1984 #ifdef CONFIG_XFS_POSIX_ACL
1985 kmem_zone_destroy(xfs_acl_zone);
1987 kmem_zone_destroy(xfs_ili_zone);
1988 kmem_zone_destroy(xfs_inode_zone);
1989 kmem_zone_destroy(xfs_efi_zone);
1990 kmem_zone_destroy(xfs_efd_zone);
1991 kmem_zone_destroy(xfs_buf_item_zone);
1992 kmem_zone_destroy(xfs_trans_zone);
1993 kmem_zone_destroy(xfs_ifork_zone);
1994 kmem_zone_destroy(xfs_dabuf_zone);
1995 kmem_zone_destroy(xfs_da_state_zone);
1996 kmem_zone_destroy(xfs_btree_cur_zone);
1997 kmem_zone_destroy(xfs_bmap_free_item_zone);
1998 kmem_zone_destroy(xfs_log_ticket_zone);
1999 mempool_destroy(xfs_ioend_pool);
2000 kmem_zone_destroy(xfs_ioend_zone);
2001 kmem_zone_destroy(xfs_vnode_zone);
2009 static char message[] __initdata = KERN_INFO \
2010 XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n";
2018 error = xfs_init_zones();
2022 error = xfs_alloc_trace_bufs();
2024 goto out_destroy_zones;
2026 error = xfs_mru_cache_init();
2028 goto out_free_trace_buffers;
2030 error = xfs_filestream_init();
2032 goto out_mru_cache_uninit;
2034 error = xfs_buf_init();
2036 goto out_filestream_uninit;
2038 error = xfs_init_procfs();
2040 goto out_buf_terminate;
2042 error = xfs_sysctl_register();
2044 goto out_cleanup_procfs;
2048 error = register_filesystem(&xfs_fs_type);
2050 goto out_sysctl_unregister;
2053 out_sysctl_unregister:
2054 xfs_sysctl_unregister();
2056 xfs_cleanup_procfs();
2058 xfs_buf_terminate();
2059 out_filestream_uninit:
2060 xfs_filestream_uninit();
2061 out_mru_cache_uninit:
2062 xfs_mru_cache_uninit();
2063 out_free_trace_buffers:
2064 xfs_free_trace_bufs();
2066 xfs_destroy_zones();
2075 unregister_filesystem(&xfs_fs_type);
2076 xfs_sysctl_unregister();
2077 xfs_cleanup_procfs();
2078 xfs_buf_terminate();
2079 xfs_filestream_uninit();
2080 xfs_mru_cache_uninit();
2081 xfs_free_trace_bufs();
2082 xfs_destroy_zones();
2086 module_init(init_xfs_fs);
2087 module_exit(exit_xfs_fs);
2089 MODULE_AUTHOR("Silicon Graphics, Inc.");
2090 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
2091 MODULE_LICENSE("GPL");