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_ialloc.h"
41 #include "xfs_rtalloc.h"
42 #include "xfs_error.h"
43 #include "xfs_itable.h"
44 #include "xfs_fsops.h"
48 #include "xfs_buf_item.h"
49 #include "xfs_utils.h"
50 #include "xfs_vnodeops.h"
51 #include "xfs_vfsops.h"
52 #include "xfs_version.h"
53 #include "xfs_log_priv.h"
54 #include "xfs_trans_priv.h"
55 #include "xfs_filestream.h"
57 #include <linux/namei.h>
58 #include <linux/init.h>
59 #include <linux/mount.h>
60 #include <linux/mempool.h>
61 #include <linux/writeback.h>
62 #include <linux/kthread.h>
63 #include <linux/freezer.h>
65 static struct quotactl_ops xfs_quotactl_operations;
66 static struct super_operations xfs_super_operations;
67 static kmem_zone_t *xfs_vnode_zone;
68 static kmem_zone_t *xfs_ioend_zone;
69 mempool_t *xfs_ioend_pool;
71 STATIC struct xfs_mount_args *
73 struct super_block *sb,
76 struct xfs_mount_args *args;
78 args = kmem_zalloc(sizeof(struct xfs_mount_args), KM_SLEEP);
79 args->logbufs = args->logbufsize = -1;
80 strncpy(args->fsname, sb->s_id, MAXNAMELEN);
82 /* Copy the already-parsed mount(2) flags we're interested in */
83 if (sb->s_flags & MS_DIRSYNC)
84 args->flags |= XFSMNT_DIRSYNC;
85 if (sb->s_flags & MS_SYNCHRONOUS)
86 args->flags |= XFSMNT_WSYNC;
88 args->flags |= XFSMNT_QUIET;
89 args->flags |= XFSMNT_32BITINODES;
94 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
95 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
96 #define MNTOPT_LOGDEV "logdev" /* log device */
97 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
98 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
99 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
100 #define MNTOPT_INO64 "ino64" /* force inodes into 64-bit range */
101 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
102 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
103 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
104 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
105 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
106 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
107 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
108 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
109 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
110 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
111 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
112 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
113 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
114 * unwritten extent conversion */
115 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
116 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
117 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
118 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
119 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
120 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
121 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
123 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
124 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
125 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
126 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
127 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
128 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
129 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
130 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
131 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
132 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
133 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
134 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
135 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
136 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
137 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
138 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
139 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
140 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
143 suffix_strtoul(char *s, char **endp, unsigned int base)
145 int last, shift_left_factor = 0;
148 last = strlen(value) - 1;
149 if (value[last] == 'K' || value[last] == 'k') {
150 shift_left_factor = 10;
153 if (value[last] == 'M' || value[last] == 'm') {
154 shift_left_factor = 20;
157 if (value[last] == 'G' || value[last] == 'g') {
158 shift_left_factor = 30;
162 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
167 struct xfs_mount *mp,
169 struct xfs_mount_args *args,
172 char *this_char, *value, *eov;
173 int dsunit, dswidth, vol_dsunit, vol_dswidth;
175 int dmapi_implies_ikeep = 1;
177 args->flags |= XFSMNT_BARRIER;
178 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
183 iosize = dsunit = dswidth = vol_dsunit = vol_dswidth = 0;
185 while ((this_char = strsep(&options, ",")) != NULL) {
188 if ((value = strchr(this_char, '=')) != NULL)
191 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
192 if (!value || !*value) {
194 "XFS: %s option requires an argument",
198 args->logbufs = simple_strtoul(value, &eov, 10);
199 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
200 if (!value || !*value) {
202 "XFS: %s option requires an argument",
206 args->logbufsize = suffix_strtoul(value, &eov, 10);
207 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
208 if (!value || !*value) {
210 "XFS: %s option requires an argument",
214 strncpy(args->logname, value, MAXNAMELEN);
215 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
216 if (!value || !*value) {
218 "XFS: %s option requires an argument",
222 strncpy(args->mtpt, value, MAXNAMELEN);
223 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
224 if (!value || !*value) {
226 "XFS: %s option requires an argument",
230 strncpy(args->rtname, value, MAXNAMELEN);
231 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
232 if (!value || !*value) {
234 "XFS: %s option requires an argument",
238 iosize = simple_strtoul(value, &eov, 10);
239 args->flags |= XFSMNT_IOSIZE;
240 args->iosizelog = (uint8_t) iosize;
241 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
242 if (!value || !*value) {
244 "XFS: %s option requires an argument",
248 iosize = suffix_strtoul(value, &eov, 10);
249 args->flags |= XFSMNT_IOSIZE;
250 args->iosizelog = ffs(iosize) - 1;
251 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
252 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
253 mp->m_flags |= XFS_MOUNT_GRPID;
254 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
255 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
256 mp->m_flags &= ~XFS_MOUNT_GRPID;
257 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
258 args->flags |= XFSMNT_WSYNC;
259 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
260 args->flags |= XFSMNT_OSYNCISOSYNC;
261 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
262 args->flags |= XFSMNT_NORECOVERY;
263 } else if (!strcmp(this_char, MNTOPT_INO64)) {
264 args->flags |= XFSMNT_INO64;
267 "XFS: %s option not allowed on this system",
271 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
272 args->flags |= XFSMNT_NOALIGN;
273 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
274 args->flags |= XFSMNT_SWALLOC;
275 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
276 if (!value || !*value) {
278 "XFS: %s option requires an argument",
282 dsunit = simple_strtoul(value, &eov, 10);
283 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
284 if (!value || !*value) {
286 "XFS: %s option requires an argument",
290 dswidth = simple_strtoul(value, &eov, 10);
291 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
292 args->flags &= ~XFSMNT_32BITINODES;
295 "XFS: %s option not allowed on this system",
299 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
300 args->flags |= XFSMNT_NOUUID;
301 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
302 args->flags |= XFSMNT_BARRIER;
303 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
304 args->flags &= ~XFSMNT_BARRIER;
305 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
306 args->flags |= XFSMNT_IKEEP;
307 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
308 dmapi_implies_ikeep = 0;
309 args->flags &= ~XFSMNT_IKEEP;
310 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
311 args->flags2 &= ~XFSMNT2_COMPAT_IOSIZE;
312 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
313 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
314 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
315 args->flags |= XFSMNT_ATTR2;
316 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
317 args->flags &= ~XFSMNT_ATTR2;
318 args->flags |= XFSMNT_NOATTR2;
319 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
320 args->flags2 |= XFSMNT2_FILESTREAMS;
321 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
322 args->flags &= ~(XFSMNT_UQUOTAENF|XFSMNT_UQUOTA);
323 args->flags &= ~(XFSMNT_GQUOTAENF|XFSMNT_GQUOTA);
324 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
325 !strcmp(this_char, MNTOPT_UQUOTA) ||
326 !strcmp(this_char, MNTOPT_USRQUOTA)) {
327 args->flags |= XFSMNT_UQUOTA | XFSMNT_UQUOTAENF;
328 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
329 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
330 args->flags |= XFSMNT_UQUOTA;
331 args->flags &= ~XFSMNT_UQUOTAENF;
332 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
333 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
334 args->flags |= XFSMNT_PQUOTA | XFSMNT_PQUOTAENF;
335 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
336 args->flags |= XFSMNT_PQUOTA;
337 args->flags &= ~XFSMNT_PQUOTAENF;
338 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
339 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
340 args->flags |= XFSMNT_GQUOTA | XFSMNT_GQUOTAENF;
341 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
342 args->flags |= XFSMNT_GQUOTA;
343 args->flags &= ~XFSMNT_GQUOTAENF;
344 } else if (!strcmp(this_char, MNTOPT_DMAPI)) {
345 args->flags |= XFSMNT_DMAPI;
346 } else if (!strcmp(this_char, MNTOPT_XDSM)) {
347 args->flags |= XFSMNT_DMAPI;
348 } else if (!strcmp(this_char, MNTOPT_DMI)) {
349 args->flags |= XFSMNT_DMAPI;
350 } else if (!strcmp(this_char, "ihashsize")) {
352 "XFS: ihashsize no longer used, option is deprecated.");
353 } else if (!strcmp(this_char, "osyncisdsync")) {
354 /* no-op, this is now the default */
356 "XFS: osyncisdsync is now the default, option is deprecated.");
357 } else if (!strcmp(this_char, "irixsgid")) {
359 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
362 "XFS: unknown mount option [%s].", this_char);
367 if (args->flags & XFSMNT_NORECOVERY) {
368 if ((mp->m_flags & XFS_MOUNT_RDONLY) == 0) {
370 "XFS: no-recovery mounts must be read-only.");
375 if ((args->flags & XFSMNT_NOALIGN) && (dsunit || dswidth)) {
377 "XFS: sunit and swidth options incompatible with the noalign option");
381 if ((args->flags & XFSMNT_GQUOTA) && (args->flags & XFSMNT_PQUOTA)) {
383 "XFS: cannot mount with both project and group quota");
387 if ((args->flags & XFSMNT_DMAPI) && *args->mtpt == '\0') {
388 printk("XFS: %s option needs the mount point option as well\n",
393 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
395 "XFS: sunit and swidth must be specified together");
399 if (dsunit && (dswidth % dsunit != 0)) {
401 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
407 * Applications using DMI filesystems often expect the
408 * inode generation number to be monotonically increasing.
409 * If we delete inode chunks we break this assumption, so
410 * keep unused inode chunks on disk for DMI filesystems
411 * until we come up with a better solution.
412 * Note that if "ikeep" or "noikeep" mount options are
413 * supplied, then they are honored.
415 if ((args->flags & XFSMNT_DMAPI) && dmapi_implies_ikeep)
416 args->flags |= XFSMNT_IKEEP;
418 if ((args->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
420 args->sunit = dsunit;
421 args->flags |= XFSMNT_RETERR;
423 args->sunit = vol_dsunit;
425 dswidth ? (args->swidth = dswidth) :
426 (args->swidth = vol_dswidth);
428 args->sunit = args->swidth = 0;
432 if (args->flags & XFSMNT_32BITINODES)
433 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
435 args->flags |= XFSMNT_FLAGS2;
439 struct proc_xfs_info {
446 struct xfs_mount *mp,
449 static struct proc_xfs_info xfs_info_set[] = {
450 /* the few simple ones we can get from the mount struct */
451 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
452 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
453 { XFS_MOUNT_INO64, "," MNTOPT_INO64 },
454 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
455 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
456 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
457 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
458 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
459 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
460 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
461 { XFS_MOUNT_DMAPI, "," MNTOPT_DMAPI },
462 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
465 static struct proc_xfs_info xfs_info_unset[] = {
466 /* the few simple ones we can get from the mount struct */
467 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
468 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
469 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
472 struct proc_xfs_info *xfs_infop;
474 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
475 if (mp->m_flags & xfs_infop->flag)
476 seq_puts(m, xfs_infop->str);
478 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
479 if (!(mp->m_flags & xfs_infop->flag))
480 seq_puts(m, xfs_infop->str);
483 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
484 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
485 (int)(1 << mp->m_writeio_log) >> 10);
487 if (mp->m_logbufs > 0)
488 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
489 if (mp->m_logbsize > 0)
490 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
493 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
495 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
497 if (mp->m_dalign > 0)
498 seq_printf(m, "," MNTOPT_SUNIT "=%d",
499 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
500 if (mp->m_swidth > 0)
501 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
502 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
504 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
505 seq_puts(m, "," MNTOPT_USRQUOTA);
506 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
507 seq_puts(m, "," MNTOPT_UQUOTANOENF);
509 if (mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
510 seq_puts(m, "," MNTOPT_PRJQUOTA);
511 else if (mp->m_qflags & XFS_PQUOTA_ACCT)
512 seq_puts(m, "," MNTOPT_PQUOTANOENF);
514 if (mp->m_qflags & (XFS_GQUOTA_ACCT|XFS_OQUOTA_ENFD))
515 seq_puts(m, "," MNTOPT_GRPQUOTA);
516 else if (mp->m_qflags & XFS_GQUOTA_ACCT)
517 seq_puts(m, "," MNTOPT_GQUOTANOENF);
519 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
520 seq_puts(m, "," MNTOPT_NOQUOTA);
526 unsigned int blockshift)
528 unsigned int pagefactor = 1;
529 unsigned int bitshift = BITS_PER_LONG - 1;
531 /* Figure out maximum filesize, on Linux this can depend on
532 * the filesystem blocksize (on 32 bit platforms).
533 * __block_prepare_write does this in an [unsigned] long...
534 * page->index << (PAGE_CACHE_SHIFT - bbits)
535 * So, for page sized blocks (4K on 32 bit platforms),
536 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
537 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
538 * but for smaller blocksizes it is less (bbits = log2 bsize).
539 * Note1: get_block_t takes a long (implicit cast from above)
540 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
541 * can optionally convert the [unsigned] long from above into
542 * an [unsigned] long long.
545 #if BITS_PER_LONG == 32
546 # if defined(CONFIG_LBD)
547 ASSERT(sizeof(sector_t) == 8);
548 pagefactor = PAGE_CACHE_SIZE;
549 bitshift = BITS_PER_LONG;
551 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
555 return (((__uint64_t)pagefactor) << bitshift) - 1;
562 switch (inode->i_mode & S_IFMT) {
564 inode->i_op = &xfs_inode_operations;
565 inode->i_fop = &xfs_file_operations;
566 inode->i_mapping->a_ops = &xfs_address_space_operations;
569 inode->i_op = &xfs_dir_inode_operations;
570 inode->i_fop = &xfs_dir_file_operations;
573 inode->i_op = &xfs_symlink_inode_operations;
574 if (!(XFS_I(inode)->i_df.if_flags & XFS_IFINLINE))
575 inode->i_mapping->a_ops = &xfs_address_space_operations;
578 inode->i_op = &xfs_inode_operations;
579 init_special_inode(inode, inode->i_mode, inode->i_rdev);
585 xfs_revalidate_inode(
590 struct inode *inode = vn_to_inode(vp);
592 inode->i_mode = ip->i_d.di_mode;
593 inode->i_nlink = ip->i_d.di_nlink;
594 inode->i_uid = ip->i_d.di_uid;
595 inode->i_gid = ip->i_d.di_gid;
597 switch (inode->i_mode & S_IFMT) {
601 MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
602 sysv_minor(ip->i_df.if_u2.if_rdev));
609 inode->i_generation = ip->i_d.di_gen;
610 i_size_write(inode, ip->i_d.di_size);
611 inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec;
612 inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec;
613 inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
614 inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
615 inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
616 inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
617 if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
618 inode->i_flags |= S_IMMUTABLE;
620 inode->i_flags &= ~S_IMMUTABLE;
621 if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
622 inode->i_flags |= S_APPEND;
624 inode->i_flags &= ~S_APPEND;
625 if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
626 inode->i_flags |= S_SYNC;
628 inode->i_flags &= ~S_SYNC;
629 if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
630 inode->i_flags |= S_NOATIME;
632 inode->i_flags &= ~S_NOATIME;
633 xfs_iflags_clear(ip, XFS_IMODIFIED);
637 xfs_initialize_vnode(
638 struct xfs_mount *mp,
640 struct xfs_inode *ip)
642 struct inode *inode = vn_to_inode(vp);
646 inode->i_private = ip;
650 * We need to set the ops vectors, and unlock the inode, but if
651 * we have been called during the new inode create process, it is
652 * too early to fill in the Linux inode. We will get called a
653 * second time once the inode is properly set up, and then we can
656 if (ip->i_d.di_mode != 0 && (inode->i_state & I_NEW)) {
657 xfs_revalidate_inode(mp, vp, ip);
658 xfs_set_inodeops(inode);
660 xfs_iflags_clear(ip, XFS_INEW);
663 unlock_new_inode(inode);
671 struct block_device **bdevp)
675 *bdevp = open_bdev_excl(name, 0, mp);
676 if (IS_ERR(*bdevp)) {
677 error = PTR_ERR(*bdevp);
678 printk("XFS: Invalid device [%s], error=%d\n", name, error);
686 struct block_device *bdev)
689 close_bdev_excl(bdev);
693 * Try to write out the superblock using barriers.
699 xfs_buf_t *sbp = xfs_getsb(mp, 0);
704 XFS_BUF_UNDELAYWRITE(sbp);
706 XFS_BUF_UNASYNC(sbp);
707 XFS_BUF_ORDERED(sbp);
710 error = xfs_iowait(sbp);
713 * Clear all the flags we set and possible error state in the
714 * buffer. We only did the write to try out whether barriers
715 * worked and shouldn't leave any traces in the superblock
719 XFS_BUF_ERROR(sbp, 0);
720 XFS_BUF_UNORDERED(sbp);
727 xfs_mountfs_check_barriers(xfs_mount_t *mp)
731 if (mp->m_logdev_targp != mp->m_ddev_targp) {
732 xfs_fs_cmn_err(CE_NOTE, mp,
733 "Disabling barriers, not supported with external log device");
734 mp->m_flags &= ~XFS_MOUNT_BARRIER;
738 if (mp->m_ddev_targp->bt_bdev->bd_disk->queue->ordered ==
739 QUEUE_ORDERED_NONE) {
740 xfs_fs_cmn_err(CE_NOTE, mp,
741 "Disabling barriers, not supported by the underlying device");
742 mp->m_flags &= ~XFS_MOUNT_BARRIER;
746 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
747 xfs_fs_cmn_err(CE_NOTE, mp,
748 "Disabling barriers, underlying device is readonly");
749 mp->m_flags &= ~XFS_MOUNT_BARRIER;
753 error = xfs_barrier_test(mp);
755 xfs_fs_cmn_err(CE_NOTE, mp,
756 "Disabling barriers, trial barrier write failed");
757 mp->m_flags &= ~XFS_MOUNT_BARRIER;
763 xfs_blkdev_issue_flush(
764 xfs_buftarg_t *buftarg)
766 blkdev_issue_flush(buftarg->bt_bdev, NULL);
771 struct xfs_mount *mp)
773 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
774 xfs_free_buftarg(mp->m_logdev_targp);
775 xfs_blkdev_put(mp->m_logdev_targp->bt_bdev);
777 if (mp->m_rtdev_targp) {
778 xfs_free_buftarg(mp->m_rtdev_targp);
779 xfs_blkdev_put(mp->m_rtdev_targp->bt_bdev);
781 xfs_free_buftarg(mp->m_ddev_targp);
785 * The file system configurations are:
786 * (1) device (partition) with data and internal log
787 * (2) logical volume with data and log subvolumes.
788 * (3) logical volume with data, log, and realtime subvolumes.
790 * We only have to handle opening the log and realtime volumes here if
791 * they are present. The data subvolume has already been opened by
792 * get_sb_bdev() and is stored in sb->s_bdev.
796 struct xfs_mount *mp,
797 struct xfs_mount_args *args)
799 struct block_device *ddev = mp->m_super->s_bdev;
800 struct block_device *logdev = NULL, *rtdev = NULL;
804 * Open real time and log devices - order is important.
806 if (args->logname[0]) {
807 error = xfs_blkdev_get(mp, args->logname, &logdev);
812 if (args->rtname[0]) {
813 error = xfs_blkdev_get(mp, args->rtname, &rtdev);
815 goto out_close_logdev;
817 if (rtdev == ddev || rtdev == logdev) {
819 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
821 goto out_close_rtdev;
826 * Setup xfs_mount buffer target pointers
829 mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0);
830 if (!mp->m_ddev_targp)
831 goto out_close_rtdev;
834 mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1);
835 if (!mp->m_rtdev_targp)
836 goto out_free_ddev_targ;
839 if (logdev && logdev != ddev) {
840 mp->m_logdev_targp = xfs_alloc_buftarg(logdev, 1);
841 if (!mp->m_logdev_targp)
842 goto out_free_rtdev_targ;
844 mp->m_logdev_targp = mp->m_ddev_targp;
850 if (mp->m_rtdev_targp)
851 xfs_free_buftarg(mp->m_rtdev_targp);
853 xfs_free_buftarg(mp->m_ddev_targp);
856 xfs_blkdev_put(rtdev);
858 if (logdev && logdev != ddev)
859 xfs_blkdev_put(logdev);
865 * Setup xfs_mount buffer target pointers based on superblock
869 struct xfs_mount *mp)
873 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
874 mp->m_sb.sb_sectsize);
878 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
879 unsigned int log_sector_size = BBSIZE;
881 if (xfs_sb_version_hassector(&mp->m_sb))
882 log_sector_size = mp->m_sb.sb_logsectsize;
883 error = xfs_setsize_buftarg(mp->m_logdev_targp,
884 mp->m_sb.sb_blocksize,
889 if (mp->m_rtdev_targp) {
890 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
891 mp->m_sb.sb_blocksize,
892 mp->m_sb.sb_sectsize);
901 * XFS AIL push thread support
906 xfs_lsn_t threshold_lsn)
908 mp->m_ail.xa_target = threshold_lsn;
909 wake_up_process(mp->m_ail.xa_task);
916 xfs_mount_t *mp = (xfs_mount_t *)data;
917 xfs_lsn_t last_pushed_lsn = 0;
920 while (!kthread_should_stop()) {
922 schedule_timeout_interruptible(msecs_to_jiffies(tout));
929 if (XFS_FORCED_SHUTDOWN(mp))
932 tout = xfsaild_push(mp, &last_pushed_lsn);
942 mp->m_ail.xa_target = 0;
943 mp->m_ail.xa_task = kthread_run(xfsaild, mp, "xfsaild");
944 if (IS_ERR(mp->m_ail.xa_task))
945 return -PTR_ERR(mp->m_ail.xa_task);
953 kthread_stop(mp->m_ail.xa_task);
958 STATIC struct inode *
960 struct super_block *sb)
964 vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP);
967 return vn_to_inode(vp);
971 xfs_fs_destroy_inode(
974 kmem_zone_free(xfs_vnode_zone, vn_from_inode(inode));
978 xfs_fs_inode_init_once(
981 inode_init_once(vn_to_inode((bhv_vnode_t *)vnode));
987 xfs_vnode_zone = kmem_zone_init_flags(sizeof(bhv_vnode_t), "xfs_vnode",
988 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM |
990 xfs_fs_inode_init_once);
994 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
996 goto out_destroy_vnode_zone;
998 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1000 if (!xfs_ioend_pool)
1001 goto out_free_ioend_zone;
1004 out_free_ioend_zone:
1005 kmem_zone_destroy(xfs_ioend_zone);
1006 out_destroy_vnode_zone:
1007 kmem_zone_destroy(xfs_vnode_zone);
1013 xfs_destroy_zones(void)
1015 mempool_destroy(xfs_ioend_pool);
1016 kmem_zone_destroy(xfs_vnode_zone);
1017 kmem_zone_destroy(xfs_ioend_zone);
1021 * Attempt to flush the inode, this will actually fail
1022 * if the inode is pinned, but we dirty the inode again
1023 * at the point when it is unpinned after a log write,
1024 * since this is when the inode itself becomes flushable.
1028 struct inode *inode,
1034 xfs_itrace_entry(XFS_I(inode));
1036 filemap_fdatawait(inode->i_mapping);
1037 flags |= FLUSH_SYNC;
1039 error = xfs_inode_flush(XFS_I(inode), flags);
1041 * if we failed to write out the inode then mark
1042 * it dirty again so we'll try again later.
1045 mark_inode_dirty_sync(inode);
1052 struct inode *inode)
1054 xfs_inode_t *ip = XFS_I(inode);
1057 * ip can be null when xfs_iget_core calls xfs_idestroy if we
1058 * find an inode with di_mode == 0 but without IGET_CREATE set.
1061 xfs_itrace_entry(ip);
1062 XFS_STATS_INC(vn_rele);
1063 XFS_STATS_INC(vn_remove);
1064 XFS_STATS_INC(vn_reclaim);
1065 XFS_STATS_DEC(vn_active);
1068 xfs_iflags_clear(ip, XFS_IMODIFIED);
1069 if (xfs_reclaim(ip))
1070 panic("%s: cannot reclaim 0x%p\n", __func__, inode);
1073 ASSERT(XFS_I(inode) == NULL);
1077 * Enqueue a work item to be picked up by the vfs xfssyncd thread.
1078 * Doing this has two advantages:
1079 * - It saves on stack space, which is tight in certain situations
1080 * - It can be used (with care) as a mechanism to avoid deadlocks.
1081 * Flushing while allocating in a full filesystem requires both.
1084 xfs_syncd_queue_work(
1085 struct xfs_mount *mp,
1087 void (*syncer)(struct xfs_mount *, void *))
1089 struct bhv_vfs_sync_work *work;
1091 work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP);
1092 INIT_LIST_HEAD(&work->w_list);
1093 work->w_syncer = syncer;
1094 work->w_data = data;
1096 spin_lock(&mp->m_sync_lock);
1097 list_add_tail(&work->w_list, &mp->m_sync_list);
1098 spin_unlock(&mp->m_sync_lock);
1099 wake_up_process(mp->m_sync_task);
1103 * Flush delayed allocate data, attempting to free up reserved space
1104 * from existing allocations. At this point a new allocation attempt
1105 * has failed with ENOSPC and we are in the process of scratching our
1106 * heads, looking about for more room...
1109 xfs_flush_inode_work(
1110 struct xfs_mount *mp,
1113 struct inode *inode = arg;
1114 filemap_flush(inode->i_mapping);
1122 struct inode *inode = ip->i_vnode;
1125 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_inode_work);
1126 delay(msecs_to_jiffies(500));
1130 * This is the "bigger hammer" version of xfs_flush_inode_work...
1131 * (IOW, "If at first you don't succeed, use a Bigger Hammer").
1134 xfs_flush_device_work(
1135 struct xfs_mount *mp,
1138 struct inode *inode = arg;
1139 sync_blockdev(mp->m_super->s_bdev);
1147 struct inode *inode = vn_to_inode(XFS_ITOV(ip));
1150 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_device_work);
1151 delay(msecs_to_jiffies(500));
1152 xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC);
1157 struct xfs_mount *mp,
1162 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1163 error = xfs_sync(mp, SYNC_FSDATA | SYNC_BDFLUSH | SYNC_ATTR);
1165 wake_up(&mp->m_wait_single_sync_task);
1172 struct xfs_mount *mp = arg;
1174 bhv_vfs_sync_work_t *work, *n;
1178 timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10);
1180 timeleft = schedule_timeout_interruptible(timeleft);
1183 if (kthread_should_stop() && list_empty(&mp->m_sync_list))
1186 spin_lock(&mp->m_sync_lock);
1188 * We can get woken by laptop mode, to do a sync -
1189 * that's the (only!) case where the list would be
1190 * empty with time remaining.
1192 if (!timeleft || list_empty(&mp->m_sync_list)) {
1194 timeleft = xfs_syncd_centisecs *
1195 msecs_to_jiffies(10);
1196 INIT_LIST_HEAD(&mp->m_sync_work.w_list);
1197 list_add_tail(&mp->m_sync_work.w_list,
1200 list_for_each_entry_safe(work, n, &mp->m_sync_list, w_list)
1201 list_move(&work->w_list, &tmp);
1202 spin_unlock(&mp->m_sync_lock);
1204 list_for_each_entry_safe(work, n, &tmp, w_list) {
1205 (*work->w_syncer)(mp, work->w_data);
1206 list_del(&work->w_list);
1207 if (work == &mp->m_sync_work)
1218 struct super_block *sb)
1220 struct xfs_mount *mp = XFS_M(sb);
1221 struct xfs_inode *rip = mp->m_rootip;
1222 int unmount_event_flags = 0;
1225 kthread_stop(mp->m_sync_task);
1227 xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI);
1230 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1231 unmount_event_flags =
1232 (mp->m_dmevmask & (1 << DM_EVENT_UNMOUNT)) ?
1233 0 : DM_FLAGS_UNWANTED;
1235 * Ignore error from dmapi here, first unmount is not allowed
1236 * to fail anyway, and second we wouldn't want to fail a
1237 * unmount because of dmapi.
1239 XFS_SEND_PREUNMOUNT(mp, rip, DM_RIGHT_NULL, rip, DM_RIGHT_NULL,
1240 NULL, NULL, 0, 0, unmount_event_flags);
1245 * Blow away any referenced inode in the filestreams cache.
1246 * This can and will cause log traffic as inodes go inactive
1249 xfs_filestream_unmount(mp);
1251 XFS_bflush(mp->m_ddev_targp);
1252 error = xfs_unmount_flush(mp, 0);
1258 * If we're forcing a shutdown, typically because of a media error,
1259 * we want to make sure we invalidate dirty pages that belong to
1260 * referenced vnodes as well.
1262 if (XFS_FORCED_SHUTDOWN(mp)) {
1263 error = xfs_sync(mp, SYNC_WAIT | SYNC_CLOSE);
1264 ASSERT(error != EFSCORRUPTED);
1267 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1268 XFS_SEND_UNMOUNT(mp, rip, DM_RIGHT_NULL, 0, 0,
1269 unmount_event_flags);
1273 xfs_close_devices(mp);
1281 struct super_block *sb)
1283 if (!(sb->s_flags & MS_RDONLY))
1284 xfs_sync(XFS_M(sb), SYNC_FSDATA);
1290 struct super_block *sb,
1293 struct xfs_mount *mp = XFS_M(sb);
1298 * Treat a sync operation like a freeze. This is to work
1299 * around a race in sync_inodes() which works in two phases
1300 * - an asynchronous flush, which can write out an inode
1301 * without waiting for file size updates to complete, and a
1302 * synchronous flush, which wont do anything because the
1303 * async flush removed the inode's dirty flag. Also
1304 * sync_inodes() will not see any files that just have
1305 * outstanding transactions to be flushed because we don't
1306 * dirty the Linux inode until after the transaction I/O
1309 if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE)) {
1311 * First stage of freeze - no more writers will make progress
1312 * now we are here, so we flush delwri and delalloc buffers
1313 * here, then wait for all I/O to complete. Data is frozen at
1314 * that point. Metadata is not frozen, transactions can still
1315 * occur here so don't bother flushing the buftarg (i.e
1316 * SYNC_QUIESCE) because it'll just get dirty again.
1318 flags = SYNC_DATA_QUIESCE;
1320 flags = SYNC_FSDATA;
1322 error = xfs_sync(mp, flags);
1325 if (unlikely(laptop_mode)) {
1326 int prev_sync_seq = mp->m_sync_seq;
1329 * The disk must be active because we're syncing.
1330 * We schedule xfssyncd now (now that the disk is
1331 * active) instead of later (when it might not be).
1333 wake_up_process(mp->m_sync_task);
1335 * We have to wait for the sync iteration to complete.
1336 * If we don't, the disk activity caused by the sync
1337 * will come after the sync is completed, and that
1338 * triggers another sync from laptop mode.
1340 wait_event(mp->m_wait_single_sync_task,
1341 mp->m_sync_seq != prev_sync_seq);
1349 struct dentry *dentry,
1350 struct kstatfs *statp)
1352 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1353 xfs_sb_t *sbp = &mp->m_sb;
1354 __uint64_t fakeinos, id;
1357 statp->f_type = XFS_SB_MAGIC;
1358 statp->f_namelen = MAXNAMELEN - 1;
1360 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1361 statp->f_fsid.val[0] = (u32)id;
1362 statp->f_fsid.val[1] = (u32)(id >> 32);
1364 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1366 spin_lock(&mp->m_sb_lock);
1367 statp->f_bsize = sbp->sb_blocksize;
1368 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1369 statp->f_blocks = sbp->sb_dblocks - lsize;
1370 statp->f_bfree = statp->f_bavail =
1371 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1372 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1374 fakeinos += mp->m_inoadd;
1377 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1378 if (mp->m_maxicount)
1382 statp->f_files = min_t(typeof(statp->f_files),
1385 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1386 spin_unlock(&mp->m_sb_lock);
1388 XFS_QM_DQSTATVFS(XFS_I(dentry->d_inode), statp);
1394 struct super_block *sb,
1398 struct xfs_mount *mp = XFS_M(sb);
1399 struct xfs_mount_args *args = xfs_args_allocate(sb, 0);
1402 error = xfs_parseargs(mp, options, args, 1);
1406 if (!(*flags & MS_RDONLY)) { /* rw/ro -> rw */
1407 if (mp->m_flags & XFS_MOUNT_RDONLY)
1408 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1409 if (args->flags & XFSMNT_BARRIER) {
1410 mp->m_flags |= XFS_MOUNT_BARRIER;
1411 xfs_mountfs_check_barriers(mp);
1413 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1415 } else if (!(mp->m_flags & XFS_MOUNT_RDONLY)) { /* rw -> ro */
1416 xfs_filestream_flush(mp);
1417 xfs_sync(mp, SYNC_DATA_QUIESCE);
1418 xfs_attr_quiesce(mp);
1419 mp->m_flags |= XFS_MOUNT_RDONLY;
1428 * Second stage of a freeze. The data is already frozen so we only
1429 * need to take care of themetadata. Once that's done write a dummy
1430 * record to dirty the log in case of a crash while frozen.
1434 struct super_block *sb)
1436 struct xfs_mount *mp = XFS_M(sb);
1438 xfs_attr_quiesce(mp);
1439 xfs_fs_log_dummy(mp);
1443 xfs_fs_show_options(
1445 struct vfsmount *mnt)
1447 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1452 struct super_block *sb,
1455 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XQUOTASYNC, 0, NULL);
1460 struct super_block *sb,
1461 struct fs_quota_stat *fqs)
1463 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XGETQSTAT, 0, (caddr_t)fqs);
1468 struct super_block *sb,
1472 return -XFS_QM_QUOTACTL(XFS_M(sb), op, 0, (caddr_t)&flags);
1477 struct super_block *sb,
1480 struct fs_disk_quota *fdq)
1482 return -XFS_QM_QUOTACTL(XFS_M(sb),
1483 (type == USRQUOTA) ? Q_XGETQUOTA :
1484 ((type == GRPQUOTA) ? Q_XGETGQUOTA :
1485 Q_XGETPQUOTA), id, (caddr_t)fdq);
1490 struct super_block *sb,
1493 struct fs_disk_quota *fdq)
1495 return -XFS_QM_QUOTACTL(XFS_M(sb),
1496 (type == USRQUOTA) ? Q_XSETQLIM :
1497 ((type == GRPQUOTA) ? Q_XSETGQLIM :
1498 Q_XSETPQLIM), id, (caddr_t)fdq);
1502 * This function fills in xfs_mount_t fields based on mount args.
1503 * Note: the superblock has _not_ yet been read in.
1507 struct xfs_mount_args *ap,
1508 struct xfs_mount *mp)
1510 /* Values are in BBs */
1511 if ((ap->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
1513 * At this point the superblock has not been read
1514 * in, therefore we do not know the block size.
1515 * Before the mount call ends we will convert
1518 mp->m_dalign = ap->sunit;
1519 mp->m_swidth = ap->swidth;
1522 if (ap->logbufs != -1 &&
1524 (ap->logbufs < XLOG_MIN_ICLOGS ||
1525 ap->logbufs > XLOG_MAX_ICLOGS)) {
1527 "XFS: invalid logbufs value: %d [not %d-%d]",
1528 ap->logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
1529 return XFS_ERROR(EINVAL);
1531 mp->m_logbufs = ap->logbufs;
1532 if (ap->logbufsize != -1 &&
1533 ap->logbufsize != 0 &&
1534 (ap->logbufsize < XLOG_MIN_RECORD_BSIZE ||
1535 ap->logbufsize > XLOG_MAX_RECORD_BSIZE ||
1536 !is_power_of_2(ap->logbufsize))) {
1538 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
1540 return XFS_ERROR(EINVAL);
1542 mp->m_logbsize = ap->logbufsize;
1543 mp->m_fsname_len = strlen(ap->fsname) + 1;
1544 mp->m_fsname = kmem_alloc(mp->m_fsname_len, KM_SLEEP);
1545 strcpy(mp->m_fsname, ap->fsname);
1546 if (ap->rtname[0]) {
1547 mp->m_rtname = kmem_alloc(strlen(ap->rtname) + 1, KM_SLEEP);
1548 strcpy(mp->m_rtname, ap->rtname);
1550 if (ap->logname[0]) {
1551 mp->m_logname = kmem_alloc(strlen(ap->logname) + 1, KM_SLEEP);
1552 strcpy(mp->m_logname, ap->logname);
1555 if (ap->flags & XFSMNT_WSYNC)
1556 mp->m_flags |= XFS_MOUNT_WSYNC;
1558 if (ap->flags & XFSMNT_INO64) {
1559 mp->m_flags |= XFS_MOUNT_INO64;
1560 mp->m_inoadd = XFS_INO64_OFFSET;
1563 if (ap->flags & XFSMNT_RETERR)
1564 mp->m_flags |= XFS_MOUNT_RETERR;
1565 if (ap->flags & XFSMNT_NOALIGN)
1566 mp->m_flags |= XFS_MOUNT_NOALIGN;
1567 if (ap->flags & XFSMNT_SWALLOC)
1568 mp->m_flags |= XFS_MOUNT_SWALLOC;
1569 if (ap->flags & XFSMNT_OSYNCISOSYNC)
1570 mp->m_flags |= XFS_MOUNT_OSYNCISOSYNC;
1571 if (ap->flags & XFSMNT_32BITINODES)
1572 mp->m_flags |= XFS_MOUNT_32BITINODES;
1574 if (ap->flags & XFSMNT_IOSIZE) {
1575 if (ap->iosizelog > XFS_MAX_IO_LOG ||
1576 ap->iosizelog < XFS_MIN_IO_LOG) {
1578 "XFS: invalid log iosize: %d [not %d-%d]",
1579 ap->iosizelog, XFS_MIN_IO_LOG,
1581 return XFS_ERROR(EINVAL);
1584 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
1585 mp->m_readio_log = mp->m_writeio_log = ap->iosizelog;
1588 if (ap->flags & XFSMNT_IKEEP)
1589 mp->m_flags |= XFS_MOUNT_IKEEP;
1590 if (ap->flags & XFSMNT_DIRSYNC)
1591 mp->m_flags |= XFS_MOUNT_DIRSYNC;
1592 if (ap->flags & XFSMNT_ATTR2)
1593 mp->m_flags |= XFS_MOUNT_ATTR2;
1594 if (ap->flags & XFSMNT_NOATTR2)
1595 mp->m_flags |= XFS_MOUNT_NOATTR2;
1597 if (ap->flags2 & XFSMNT2_COMPAT_IOSIZE)
1598 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
1601 * no recovery flag requires a read-only mount
1603 if (ap->flags & XFSMNT_NORECOVERY) {
1604 if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
1606 "XFS: tried to mount a FS read-write without recovery!");
1607 return XFS_ERROR(EINVAL);
1609 mp->m_flags |= XFS_MOUNT_NORECOVERY;
1612 if (ap->flags & XFSMNT_NOUUID)
1613 mp->m_flags |= XFS_MOUNT_NOUUID;
1614 if (ap->flags & XFSMNT_BARRIER)
1615 mp->m_flags |= XFS_MOUNT_BARRIER;
1617 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1619 if (ap->flags2 & XFSMNT2_FILESTREAMS)
1620 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
1622 if (ap->flags & XFSMNT_DMAPI)
1623 mp->m_flags |= XFS_MOUNT_DMAPI;
1628 * This function fills in xfs_mount_t fields based on mount args.
1629 * Note: the superblock _has_ now been read in.
1633 struct xfs_mount_args *ap,
1634 struct xfs_mount *mp)
1636 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1638 /* Fail a mount where the logbuf is smaller then the log stripe */
1639 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1640 if ((ap->logbufsize <= 0) &&
1641 (mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE)) {
1642 mp->m_logbsize = mp->m_sb.sb_logsunit;
1643 } else if (ap->logbufsize > 0 &&
1644 ap->logbufsize < mp->m_sb.sb_logsunit) {
1646 "XFS: logbuf size must be greater than or equal to log stripe size");
1647 return XFS_ERROR(EINVAL);
1650 /* Fail a mount if the logbuf is larger than 32K */
1651 if (ap->logbufsize > XLOG_BIG_RECORD_BSIZE) {
1653 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1654 return XFS_ERROR(EINVAL);
1659 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1660 * told by noattr2 to turn it off
1662 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1663 !(ap->flags & XFSMNT_NOATTR2))
1664 mp->m_flags |= XFS_MOUNT_ATTR2;
1667 * prohibit r/w mounts of read-only filesystems
1669 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1671 "XFS: cannot mount a read-only filesystem as read-write");
1672 return XFS_ERROR(EROFS);
1676 * check for shared mount.
1678 if (ap->flags & XFSMNT_SHARED) {
1679 if (!xfs_sb_version_hasshared(&mp->m_sb))
1680 return XFS_ERROR(EINVAL);
1683 * For IRIX 6.5, shared mounts must have the shared
1684 * version bit set, have the persistent readonly
1685 * field set, must be version 0 and can only be mounted
1688 if (!ronly || !(mp->m_sb.sb_flags & XFS_SBF_READONLY) ||
1689 (mp->m_sb.sb_shared_vn != 0))
1690 return XFS_ERROR(EINVAL);
1692 mp->m_flags |= XFS_MOUNT_SHARED;
1695 * Shared XFS V0 can't deal with DMI. Return EINVAL.
1697 if (mp->m_sb.sb_shared_vn == 0 && (ap->flags & XFSMNT_DMAPI))
1698 return XFS_ERROR(EINVAL);
1701 if (ap->flags & XFSMNT_UQUOTA) {
1702 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
1703 if (ap->flags & XFSMNT_UQUOTAENF)
1704 mp->m_qflags |= XFS_UQUOTA_ENFD;
1707 if (ap->flags & XFSMNT_GQUOTA) {
1708 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
1709 if (ap->flags & XFSMNT_GQUOTAENF)
1710 mp->m_qflags |= XFS_OQUOTA_ENFD;
1711 } else if (ap->flags & XFSMNT_PQUOTA) {
1712 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
1713 if (ap->flags & XFSMNT_PQUOTAENF)
1714 mp->m_qflags |= XFS_OQUOTA_ENFD;
1722 struct super_block *sb,
1727 struct xfs_mount *mp = NULL;
1728 struct xfs_mount_args *args = xfs_args_allocate(sb, silent);
1729 int flags = 0, error;
1731 mp = xfs_mount_init();
1733 INIT_LIST_HEAD(&mp->m_sync_list);
1734 spin_lock_init(&mp->m_sync_lock);
1735 init_waitqueue_head(&mp->m_wait_single_sync_task);
1740 if (sb->s_flags & MS_RDONLY)
1741 mp->m_flags |= XFS_MOUNT_RDONLY;
1743 error = xfs_parseargs(mp, (char *)data, args, 0);
1747 sb_min_blocksize(sb, BBSIZE);
1748 sb->s_export_op = &xfs_export_operations;
1749 sb->s_qcop = &xfs_quotactl_operations;
1750 sb->s_op = &xfs_super_operations;
1752 error = xfs_dmops_get(mp, args);
1755 error = xfs_qmops_get(mp, args);
1759 if (args->flags & XFSMNT_QUIET)
1760 flags |= XFS_MFSI_QUIET;
1762 error = xfs_open_devices(mp, args);
1767 * Setup flags based on mount(2) options and then the superblock
1769 error = xfs_start_flags(args, mp);
1772 error = xfs_readsb(mp, flags);
1775 error = xfs_finish_flags(args, mp);
1779 error = xfs_setup_devices(mp);
1783 if (mp->m_flags & XFS_MOUNT_BARRIER)
1784 xfs_mountfs_check_barriers(mp);
1786 error = xfs_filestream_mount(mp);
1790 error = xfs_mountfs(mp, flags);
1794 XFS_SEND_MOUNT(mp, DM_RIGHT_NULL, args->mtpt, args->fsname);
1797 sb->s_magic = XFS_SB_MAGIC;
1798 sb->s_blocksize = mp->m_sb.sb_blocksize;
1799 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1800 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1801 sb->s_time_gran = 1;
1802 set_posix_acl_flag(sb);
1804 root = igrab(mp->m_rootip->i_vnode);
1809 if (is_bad_inode(root)) {
1813 sb->s_root = d_alloc_root(root);
1819 mp->m_sync_work.w_syncer = xfs_sync_worker;
1820 mp->m_sync_work.w_mount = mp;
1821 mp->m_sync_task = kthread_run(xfssyncd, mp, "xfssyncd");
1822 if (IS_ERR(mp->m_sync_task)) {
1823 error = -PTR_ERR(mp->m_sync_task);
1827 xfs_itrace_exit(XFS_I(sb->s_root->d_inode));
1836 xfs_binval(mp->m_ddev_targp);
1837 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp)
1838 xfs_binval(mp->m_logdev_targp);
1839 if (mp->m_rtdev_targp)
1840 xfs_binval(mp->m_rtdev_targp);
1841 xfs_close_devices(mp);
1858 * Blow away any referenced inode in the filestreams cache.
1859 * This can and will cause log traffic as inodes go inactive
1862 xfs_filestream_unmount(mp);
1864 XFS_bflush(mp->m_ddev_targp);
1865 error = xfs_unmount_flush(mp, 0);
1868 IRELE(mp->m_rootip);
1871 xfs_close_devices(mp);
1883 struct file_system_type *fs_type,
1885 const char *dev_name,
1887 struct vfsmount *mnt)
1889 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1893 static struct super_operations xfs_super_operations = {
1894 .alloc_inode = xfs_fs_alloc_inode,
1895 .destroy_inode = xfs_fs_destroy_inode,
1896 .write_inode = xfs_fs_write_inode,
1897 .clear_inode = xfs_fs_clear_inode,
1898 .put_super = xfs_fs_put_super,
1899 .write_super = xfs_fs_write_super,
1900 .sync_fs = xfs_fs_sync_super,
1901 .write_super_lockfs = xfs_fs_lockfs,
1902 .statfs = xfs_fs_statfs,
1903 .remount_fs = xfs_fs_remount,
1904 .show_options = xfs_fs_show_options,
1907 static struct quotactl_ops xfs_quotactl_operations = {
1908 .quota_sync = xfs_fs_quotasync,
1909 .get_xstate = xfs_fs_getxstate,
1910 .set_xstate = xfs_fs_setxstate,
1911 .get_xquota = xfs_fs_getxquota,
1912 .set_xquota = xfs_fs_setxquota,
1915 static struct file_system_type xfs_fs_type = {
1916 .owner = THIS_MODULE,
1918 .get_sb = xfs_fs_get_sb,
1919 .kill_sb = kill_block_super,
1920 .fs_flags = FS_REQUIRES_DEV,
1928 static char message[] __initdata = KERN_INFO \
1929 XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n";
1935 error = xfs_init_zones();
1939 error = xfs_buf_init();
1948 error = register_filesystem(&xfs_fs_type);
1954 xfs_buf_terminate();
1957 xfs_destroy_zones();
1967 unregister_filesystem(&xfs_fs_type);
1969 xfs_buf_terminate();
1970 xfs_destroy_zones();
1974 module_init(init_xfs_fs);
1975 module_exit(exit_xfs_fs);
1977 MODULE_AUTHOR("Silicon Graphics, Inc.");
1978 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1979 MODULE_LICENSE("GPL");