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 = kzalloc(sizeof(struct xfs_mount_args), GFP_KERNEL);
82 args->logbufs = args->logbufsize = -1;
83 strncpy(args->fsname, sb->s_id, MAXNAMELEN);
85 /* Copy the already-parsed mount(2) flags we're interested in */
86 if (sb->s_flags & MS_DIRSYNC)
87 args->flags |= XFSMNT_DIRSYNC;
88 if (sb->s_flags & MS_SYNCHRONOUS)
89 args->flags |= XFSMNT_WSYNC;
91 args->flags |= XFSMNT_QUIET;
92 args->flags |= XFSMNT_32BITINODES;
97 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
98 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
99 #define MNTOPT_LOGDEV "logdev" /* log device */
100 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
101 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
102 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
103 #define MNTOPT_INO64 "ino64" /* force inodes into 64-bit range */
104 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
105 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
106 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
107 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
108 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
109 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
110 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
111 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
112 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
113 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
114 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
115 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
116 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
117 * unwritten extent conversion */
118 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
119 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
120 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
121 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
122 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
123 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
124 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
126 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
127 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
128 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
129 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
130 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
131 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
132 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
133 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
134 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
135 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
136 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
137 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
138 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
139 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
140 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
141 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
142 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
143 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
146 suffix_strtoul(char *s, char **endp, unsigned int base)
148 int last, shift_left_factor = 0;
151 last = strlen(value) - 1;
152 if (value[last] == 'K' || value[last] == 'k') {
153 shift_left_factor = 10;
156 if (value[last] == 'M' || value[last] == 'm') {
157 shift_left_factor = 20;
160 if (value[last] == 'G' || value[last] == 'g') {
161 shift_left_factor = 30;
165 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
170 struct xfs_mount *mp,
172 struct xfs_mount_args *args,
175 char *this_char, *value, *eov;
176 int dsunit, dswidth, vol_dsunit, vol_dswidth;
178 int dmapi_implies_ikeep = 1;
180 args->flags |= XFSMNT_BARRIER;
181 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
186 iosize = dsunit = dswidth = vol_dsunit = vol_dswidth = 0;
188 while ((this_char = strsep(&options, ",")) != NULL) {
191 if ((value = strchr(this_char, '=')) != NULL)
194 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
195 if (!value || !*value) {
197 "XFS: %s option requires an argument",
201 args->logbufs = simple_strtoul(value, &eov, 10);
202 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
203 if (!value || !*value) {
205 "XFS: %s option requires an argument",
209 args->logbufsize = suffix_strtoul(value, &eov, 10);
210 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
211 if (!value || !*value) {
213 "XFS: %s option requires an argument",
217 strncpy(args->logname, value, MAXNAMELEN);
218 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
219 if (!value || !*value) {
221 "XFS: %s option requires an argument",
225 strncpy(args->mtpt, value, MAXNAMELEN);
226 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
227 if (!value || !*value) {
229 "XFS: %s option requires an argument",
233 strncpy(args->rtname, value, MAXNAMELEN);
234 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
235 if (!value || !*value) {
237 "XFS: %s option requires an argument",
241 iosize = simple_strtoul(value, &eov, 10);
242 args->flags |= XFSMNT_IOSIZE;
243 args->iosizelog = (uint8_t) iosize;
244 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
245 if (!value || !*value) {
247 "XFS: %s option requires an argument",
251 iosize = suffix_strtoul(value, &eov, 10);
252 args->flags |= XFSMNT_IOSIZE;
253 args->iosizelog = ffs(iosize) - 1;
254 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
255 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
256 mp->m_flags |= XFS_MOUNT_GRPID;
257 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
258 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
259 mp->m_flags &= ~XFS_MOUNT_GRPID;
260 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
261 args->flags |= XFSMNT_WSYNC;
262 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
263 args->flags |= XFSMNT_OSYNCISOSYNC;
264 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
265 args->flags |= XFSMNT_NORECOVERY;
266 } else if (!strcmp(this_char, MNTOPT_INO64)) {
267 args->flags |= XFSMNT_INO64;
270 "XFS: %s option not allowed on this system",
274 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
275 args->flags |= XFSMNT_NOALIGN;
276 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
277 args->flags |= XFSMNT_SWALLOC;
278 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
279 if (!value || !*value) {
281 "XFS: %s option requires an argument",
285 dsunit = simple_strtoul(value, &eov, 10);
286 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
287 if (!value || !*value) {
289 "XFS: %s option requires an argument",
293 dswidth = simple_strtoul(value, &eov, 10);
294 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
295 args->flags &= ~XFSMNT_32BITINODES;
298 "XFS: %s option not allowed on this system",
302 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
303 args->flags |= XFSMNT_NOUUID;
304 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
305 args->flags |= XFSMNT_BARRIER;
306 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
307 args->flags &= ~XFSMNT_BARRIER;
308 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
309 args->flags |= XFSMNT_IKEEP;
310 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
311 dmapi_implies_ikeep = 0;
312 args->flags &= ~XFSMNT_IKEEP;
313 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
314 args->flags2 &= ~XFSMNT2_COMPAT_IOSIZE;
315 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
316 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
317 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
318 args->flags |= XFSMNT_ATTR2;
319 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
320 args->flags &= ~XFSMNT_ATTR2;
321 args->flags |= XFSMNT_NOATTR2;
322 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
323 args->flags2 |= XFSMNT2_FILESTREAMS;
324 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
325 args->flags &= ~(XFSMNT_UQUOTAENF|XFSMNT_UQUOTA);
326 args->flags &= ~(XFSMNT_GQUOTAENF|XFSMNT_GQUOTA);
327 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
328 !strcmp(this_char, MNTOPT_UQUOTA) ||
329 !strcmp(this_char, MNTOPT_USRQUOTA)) {
330 args->flags |= XFSMNT_UQUOTA | XFSMNT_UQUOTAENF;
331 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
332 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
333 args->flags |= XFSMNT_UQUOTA;
334 args->flags &= ~XFSMNT_UQUOTAENF;
335 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
336 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
337 args->flags |= XFSMNT_PQUOTA | XFSMNT_PQUOTAENF;
338 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
339 args->flags |= XFSMNT_PQUOTA;
340 args->flags &= ~XFSMNT_PQUOTAENF;
341 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
342 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
343 args->flags |= XFSMNT_GQUOTA | XFSMNT_GQUOTAENF;
344 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
345 args->flags |= XFSMNT_GQUOTA;
346 args->flags &= ~XFSMNT_GQUOTAENF;
347 } else if (!strcmp(this_char, MNTOPT_DMAPI)) {
348 args->flags |= XFSMNT_DMAPI;
349 } else if (!strcmp(this_char, MNTOPT_XDSM)) {
350 args->flags |= XFSMNT_DMAPI;
351 } else if (!strcmp(this_char, MNTOPT_DMI)) {
352 args->flags |= XFSMNT_DMAPI;
353 } else if (!strcmp(this_char, "ihashsize")) {
355 "XFS: ihashsize no longer used, option is deprecated.");
356 } else if (!strcmp(this_char, "osyncisdsync")) {
357 /* no-op, this is now the default */
359 "XFS: osyncisdsync is now the default, option is deprecated.");
360 } else if (!strcmp(this_char, "irixsgid")) {
362 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
365 "XFS: unknown mount option [%s].", this_char);
370 if (args->flags & XFSMNT_NORECOVERY) {
371 if ((mp->m_flags & XFS_MOUNT_RDONLY) == 0) {
373 "XFS: no-recovery mounts must be read-only.");
378 if ((args->flags & XFSMNT_NOALIGN) && (dsunit || dswidth)) {
380 "XFS: sunit and swidth options incompatible with the noalign option");
384 if ((args->flags & XFSMNT_GQUOTA) && (args->flags & XFSMNT_PQUOTA)) {
386 "XFS: cannot mount with both project and group quota");
390 if ((args->flags & XFSMNT_DMAPI) && *args->mtpt == '\0') {
391 printk("XFS: %s option needs the mount point option as well\n",
396 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
398 "XFS: sunit and swidth must be specified together");
402 if (dsunit && (dswidth % dsunit != 0)) {
404 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
410 * Applications using DMI filesystems often expect the
411 * inode generation number to be monotonically increasing.
412 * If we delete inode chunks we break this assumption, so
413 * keep unused inode chunks on disk for DMI filesystems
414 * until we come up with a better solution.
415 * Note that if "ikeep" or "noikeep" mount options are
416 * supplied, then they are honored.
418 if ((args->flags & XFSMNT_DMAPI) && dmapi_implies_ikeep)
419 args->flags |= XFSMNT_IKEEP;
421 if ((args->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
423 args->sunit = dsunit;
424 args->flags |= XFSMNT_RETERR;
426 args->sunit = vol_dsunit;
428 dswidth ? (args->swidth = dswidth) :
429 (args->swidth = vol_dswidth);
431 args->sunit = args->swidth = 0;
435 if (args->flags & XFSMNT_32BITINODES)
436 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
438 args->flags |= XFSMNT_FLAGS2;
442 struct proc_xfs_info {
449 struct xfs_mount *mp,
452 static struct proc_xfs_info xfs_info_set[] = {
453 /* the few simple ones we can get from the mount struct */
454 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
455 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
456 { XFS_MOUNT_INO64, "," MNTOPT_INO64 },
457 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
458 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
459 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
460 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
461 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
462 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
463 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
464 { XFS_MOUNT_DMAPI, "," MNTOPT_DMAPI },
465 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
468 static struct proc_xfs_info xfs_info_unset[] = {
469 /* the few simple ones we can get from the mount struct */
470 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
471 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
472 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
475 struct proc_xfs_info *xfs_infop;
477 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
478 if (mp->m_flags & xfs_infop->flag)
479 seq_puts(m, xfs_infop->str);
481 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
482 if (!(mp->m_flags & xfs_infop->flag))
483 seq_puts(m, xfs_infop->str);
486 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
487 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
488 (int)(1 << mp->m_writeio_log) >> 10);
490 if (mp->m_logbufs > 0)
491 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
492 if (mp->m_logbsize > 0)
493 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
496 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
498 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
500 if (mp->m_dalign > 0)
501 seq_printf(m, "," MNTOPT_SUNIT "=%d",
502 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
503 if (mp->m_swidth > 0)
504 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
505 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
507 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
508 seq_puts(m, "," MNTOPT_USRQUOTA);
509 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
510 seq_puts(m, "," MNTOPT_UQUOTANOENF);
512 if (mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
513 seq_puts(m, "," MNTOPT_PRJQUOTA);
514 else if (mp->m_qflags & XFS_PQUOTA_ACCT)
515 seq_puts(m, "," MNTOPT_PQUOTANOENF);
517 if (mp->m_qflags & (XFS_GQUOTA_ACCT|XFS_OQUOTA_ENFD))
518 seq_puts(m, "," MNTOPT_GRPQUOTA);
519 else if (mp->m_qflags & XFS_GQUOTA_ACCT)
520 seq_puts(m, "," MNTOPT_GQUOTANOENF);
522 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
523 seq_puts(m, "," MNTOPT_NOQUOTA);
529 unsigned int blockshift)
531 unsigned int pagefactor = 1;
532 unsigned int bitshift = BITS_PER_LONG - 1;
534 /* Figure out maximum filesize, on Linux this can depend on
535 * the filesystem blocksize (on 32 bit platforms).
536 * __block_prepare_write does this in an [unsigned] long...
537 * page->index << (PAGE_CACHE_SHIFT - bbits)
538 * So, for page sized blocks (4K on 32 bit platforms),
539 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
540 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
541 * but for smaller blocksizes it is less (bbits = log2 bsize).
542 * Note1: get_block_t takes a long (implicit cast from above)
543 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
544 * can optionally convert the [unsigned] long from above into
545 * an [unsigned] long long.
548 #if BITS_PER_LONG == 32
549 # if defined(CONFIG_LBD)
550 ASSERT(sizeof(sector_t) == 8);
551 pagefactor = PAGE_CACHE_SIZE;
552 bitshift = BITS_PER_LONG;
554 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
558 return (((__uint64_t)pagefactor) << bitshift) - 1;
565 switch (inode->i_mode & S_IFMT) {
567 inode->i_op = &xfs_inode_operations;
568 inode->i_fop = &xfs_file_operations;
569 inode->i_mapping->a_ops = &xfs_address_space_operations;
572 inode->i_op = &xfs_dir_inode_operations;
573 inode->i_fop = &xfs_dir_file_operations;
576 inode->i_op = &xfs_symlink_inode_operations;
577 if (!(XFS_I(inode)->i_df.if_flags & XFS_IFINLINE))
578 inode->i_mapping->a_ops = &xfs_address_space_operations;
581 inode->i_op = &xfs_inode_operations;
582 init_special_inode(inode, inode->i_mode, inode->i_rdev);
588 xfs_revalidate_inode(
593 struct inode *inode = vn_to_inode(vp);
595 inode->i_mode = ip->i_d.di_mode;
596 inode->i_nlink = ip->i_d.di_nlink;
597 inode->i_uid = ip->i_d.di_uid;
598 inode->i_gid = ip->i_d.di_gid;
600 switch (inode->i_mode & S_IFMT) {
604 MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
605 sysv_minor(ip->i_df.if_u2.if_rdev));
612 inode->i_generation = ip->i_d.di_gen;
613 i_size_write(inode, ip->i_d.di_size);
614 inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec;
615 inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec;
616 inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
617 inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
618 inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
619 inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
620 if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
621 inode->i_flags |= S_IMMUTABLE;
623 inode->i_flags &= ~S_IMMUTABLE;
624 if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
625 inode->i_flags |= S_APPEND;
627 inode->i_flags &= ~S_APPEND;
628 if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
629 inode->i_flags |= S_SYNC;
631 inode->i_flags &= ~S_SYNC;
632 if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
633 inode->i_flags |= S_NOATIME;
635 inode->i_flags &= ~S_NOATIME;
636 xfs_iflags_clear(ip, XFS_IMODIFIED);
640 xfs_initialize_vnode(
641 struct xfs_mount *mp,
643 struct xfs_inode *ip)
645 struct inode *inode = vn_to_inode(vp);
649 inode->i_private = ip;
653 * We need to set the ops vectors, and unlock the inode, but if
654 * we have been called during the new inode create process, it is
655 * too early to fill in the Linux inode. We will get called a
656 * second time once the inode is properly set up, and then we can
659 if (ip->i_d.di_mode != 0 && (inode->i_state & I_NEW)) {
660 xfs_revalidate_inode(mp, vp, ip);
661 xfs_set_inodeops(inode);
663 xfs_iflags_clear(ip, XFS_INEW);
666 unlock_new_inode(inode);
674 struct block_device **bdevp)
678 *bdevp = open_bdev_excl(name, 0, mp);
679 if (IS_ERR(*bdevp)) {
680 error = PTR_ERR(*bdevp);
681 printk("XFS: Invalid device [%s], error=%d\n", name, error);
689 struct block_device *bdev)
692 close_bdev_excl(bdev);
696 * Try to write out the superblock using barriers.
702 xfs_buf_t *sbp = xfs_getsb(mp, 0);
707 XFS_BUF_UNDELAYWRITE(sbp);
709 XFS_BUF_UNASYNC(sbp);
710 XFS_BUF_ORDERED(sbp);
713 error = xfs_iowait(sbp);
716 * Clear all the flags we set and possible error state in the
717 * buffer. We only did the write to try out whether barriers
718 * worked and shouldn't leave any traces in the superblock
722 XFS_BUF_ERROR(sbp, 0);
723 XFS_BUF_UNORDERED(sbp);
730 xfs_mountfs_check_barriers(xfs_mount_t *mp)
734 if (mp->m_logdev_targp != mp->m_ddev_targp) {
735 xfs_fs_cmn_err(CE_NOTE, mp,
736 "Disabling barriers, not supported with external log device");
737 mp->m_flags &= ~XFS_MOUNT_BARRIER;
741 if (mp->m_ddev_targp->bt_bdev->bd_disk->queue->ordered ==
742 QUEUE_ORDERED_NONE) {
743 xfs_fs_cmn_err(CE_NOTE, mp,
744 "Disabling barriers, not supported by the underlying device");
745 mp->m_flags &= ~XFS_MOUNT_BARRIER;
749 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
750 xfs_fs_cmn_err(CE_NOTE, mp,
751 "Disabling barriers, underlying device is readonly");
752 mp->m_flags &= ~XFS_MOUNT_BARRIER;
756 error = xfs_barrier_test(mp);
758 xfs_fs_cmn_err(CE_NOTE, mp,
759 "Disabling barriers, trial barrier write failed");
760 mp->m_flags &= ~XFS_MOUNT_BARRIER;
766 xfs_blkdev_issue_flush(
767 xfs_buftarg_t *buftarg)
769 blkdev_issue_flush(buftarg->bt_bdev, NULL);
774 struct xfs_mount *mp)
776 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
777 xfs_free_buftarg(mp->m_logdev_targp);
778 xfs_blkdev_put(mp->m_logdev_targp->bt_bdev);
780 if (mp->m_rtdev_targp) {
781 xfs_free_buftarg(mp->m_rtdev_targp);
782 xfs_blkdev_put(mp->m_rtdev_targp->bt_bdev);
784 xfs_free_buftarg(mp->m_ddev_targp);
788 * The file system configurations are:
789 * (1) device (partition) with data and internal log
790 * (2) logical volume with data and log subvolumes.
791 * (3) logical volume with data, log, and realtime subvolumes.
793 * We only have to handle opening the log and realtime volumes here if
794 * they are present. The data subvolume has already been opened by
795 * get_sb_bdev() and is stored in sb->s_bdev.
799 struct xfs_mount *mp,
800 struct xfs_mount_args *args)
802 struct block_device *ddev = mp->m_super->s_bdev;
803 struct block_device *logdev = NULL, *rtdev = NULL;
807 * Open real time and log devices - order is important.
809 if (args->logname[0]) {
810 error = xfs_blkdev_get(mp, args->logname, &logdev);
815 if (args->rtname[0]) {
816 error = xfs_blkdev_get(mp, args->rtname, &rtdev);
818 goto out_close_logdev;
820 if (rtdev == ddev || rtdev == logdev) {
822 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
824 goto out_close_rtdev;
829 * Setup xfs_mount buffer target pointers
832 mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0);
833 if (!mp->m_ddev_targp)
834 goto out_close_rtdev;
837 mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1);
838 if (!mp->m_rtdev_targp)
839 goto out_free_ddev_targ;
842 if (logdev && logdev != ddev) {
843 mp->m_logdev_targp = xfs_alloc_buftarg(logdev, 1);
844 if (!mp->m_logdev_targp)
845 goto out_free_rtdev_targ;
847 mp->m_logdev_targp = mp->m_ddev_targp;
853 if (mp->m_rtdev_targp)
854 xfs_free_buftarg(mp->m_rtdev_targp);
856 xfs_free_buftarg(mp->m_ddev_targp);
859 xfs_blkdev_put(rtdev);
861 if (logdev && logdev != ddev)
862 xfs_blkdev_put(logdev);
868 * Setup xfs_mount buffer target pointers based on superblock
872 struct xfs_mount *mp)
876 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
877 mp->m_sb.sb_sectsize);
881 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
882 unsigned int log_sector_size = BBSIZE;
884 if (xfs_sb_version_hassector(&mp->m_sb))
885 log_sector_size = mp->m_sb.sb_logsectsize;
886 error = xfs_setsize_buftarg(mp->m_logdev_targp,
887 mp->m_sb.sb_blocksize,
892 if (mp->m_rtdev_targp) {
893 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
894 mp->m_sb.sb_blocksize,
895 mp->m_sb.sb_sectsize);
904 * XFS AIL push thread support
909 xfs_lsn_t threshold_lsn)
911 mp->m_ail.xa_target = threshold_lsn;
912 wake_up_process(mp->m_ail.xa_task);
919 xfs_mount_t *mp = (xfs_mount_t *)data;
920 xfs_lsn_t last_pushed_lsn = 0;
923 while (!kthread_should_stop()) {
925 schedule_timeout_interruptible(msecs_to_jiffies(tout));
932 if (XFS_FORCED_SHUTDOWN(mp))
935 tout = xfsaild_push(mp, &last_pushed_lsn);
945 mp->m_ail.xa_target = 0;
946 mp->m_ail.xa_task = kthread_run(xfsaild, mp, "xfsaild");
947 if (IS_ERR(mp->m_ail.xa_task))
948 return -PTR_ERR(mp->m_ail.xa_task);
956 kthread_stop(mp->m_ail.xa_task);
961 STATIC struct inode *
963 struct super_block *sb)
967 vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP);
970 return vn_to_inode(vp);
974 xfs_fs_destroy_inode(
977 kmem_zone_free(xfs_vnode_zone, vn_from_inode(inode));
981 xfs_fs_inode_init_once(
984 inode_init_once(vn_to_inode((bhv_vnode_t *)vnode));
990 xfs_vnode_zone = kmem_zone_init_flags(sizeof(bhv_vnode_t), "xfs_vnode",
991 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM |
993 xfs_fs_inode_init_once);
997 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
999 goto out_destroy_vnode_zone;
1001 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1003 if (!xfs_ioend_pool)
1004 goto out_free_ioend_zone;
1007 out_free_ioend_zone:
1008 kmem_zone_destroy(xfs_ioend_zone);
1009 out_destroy_vnode_zone:
1010 kmem_zone_destroy(xfs_vnode_zone);
1016 xfs_destroy_zones(void)
1018 mempool_destroy(xfs_ioend_pool);
1019 kmem_zone_destroy(xfs_vnode_zone);
1020 kmem_zone_destroy(xfs_ioend_zone);
1024 * Attempt to flush the inode, this will actually fail
1025 * if the inode is pinned, but we dirty the inode again
1026 * at the point when it is unpinned after a log write,
1027 * since this is when the inode itself becomes flushable.
1031 struct inode *inode,
1037 xfs_itrace_entry(XFS_I(inode));
1039 filemap_fdatawait(inode->i_mapping);
1040 flags |= FLUSH_SYNC;
1042 error = xfs_inode_flush(XFS_I(inode), flags);
1044 * if we failed to write out the inode then mark
1045 * it dirty again so we'll try again later.
1048 mark_inode_dirty_sync(inode);
1055 struct inode *inode)
1057 xfs_inode_t *ip = XFS_I(inode);
1060 * ip can be null when xfs_iget_core calls xfs_idestroy if we
1061 * find an inode with di_mode == 0 but without IGET_CREATE set.
1064 xfs_itrace_entry(ip);
1065 XFS_STATS_INC(vn_rele);
1066 XFS_STATS_INC(vn_remove);
1067 XFS_STATS_INC(vn_reclaim);
1068 XFS_STATS_DEC(vn_active);
1071 xfs_iflags_clear(ip, XFS_IMODIFIED);
1072 if (xfs_reclaim(ip))
1073 panic("%s: cannot reclaim 0x%p\n", __func__, inode);
1076 ASSERT(XFS_I(inode) == NULL);
1080 * Enqueue a work item to be picked up by the vfs xfssyncd thread.
1081 * Doing this has two advantages:
1082 * - It saves on stack space, which is tight in certain situations
1083 * - It can be used (with care) as a mechanism to avoid deadlocks.
1084 * Flushing while allocating in a full filesystem requires both.
1087 xfs_syncd_queue_work(
1088 struct xfs_mount *mp,
1090 void (*syncer)(struct xfs_mount *, void *))
1092 struct bhv_vfs_sync_work *work;
1094 work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP);
1095 INIT_LIST_HEAD(&work->w_list);
1096 work->w_syncer = syncer;
1097 work->w_data = data;
1099 spin_lock(&mp->m_sync_lock);
1100 list_add_tail(&work->w_list, &mp->m_sync_list);
1101 spin_unlock(&mp->m_sync_lock);
1102 wake_up_process(mp->m_sync_task);
1106 * Flush delayed allocate data, attempting to free up reserved space
1107 * from existing allocations. At this point a new allocation attempt
1108 * has failed with ENOSPC and we are in the process of scratching our
1109 * heads, looking about for more room...
1112 xfs_flush_inode_work(
1113 struct xfs_mount *mp,
1116 struct inode *inode = arg;
1117 filemap_flush(inode->i_mapping);
1125 struct inode *inode = ip->i_vnode;
1128 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_inode_work);
1129 delay(msecs_to_jiffies(500));
1133 * This is the "bigger hammer" version of xfs_flush_inode_work...
1134 * (IOW, "If at first you don't succeed, use a Bigger Hammer").
1137 xfs_flush_device_work(
1138 struct xfs_mount *mp,
1141 struct inode *inode = arg;
1142 sync_blockdev(mp->m_super->s_bdev);
1150 struct inode *inode = vn_to_inode(XFS_ITOV(ip));
1153 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_device_work);
1154 delay(msecs_to_jiffies(500));
1155 xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC);
1160 struct xfs_mount *mp,
1165 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1166 error = xfs_sync(mp, SYNC_FSDATA | SYNC_BDFLUSH | SYNC_ATTR);
1168 wake_up(&mp->m_wait_single_sync_task);
1175 struct xfs_mount *mp = arg;
1177 bhv_vfs_sync_work_t *work, *n;
1181 timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10);
1183 timeleft = schedule_timeout_interruptible(timeleft);
1186 if (kthread_should_stop() && list_empty(&mp->m_sync_list))
1189 spin_lock(&mp->m_sync_lock);
1191 * We can get woken by laptop mode, to do a sync -
1192 * that's the (only!) case where the list would be
1193 * empty with time remaining.
1195 if (!timeleft || list_empty(&mp->m_sync_list)) {
1197 timeleft = xfs_syncd_centisecs *
1198 msecs_to_jiffies(10);
1199 INIT_LIST_HEAD(&mp->m_sync_work.w_list);
1200 list_add_tail(&mp->m_sync_work.w_list,
1203 list_for_each_entry_safe(work, n, &mp->m_sync_list, w_list)
1204 list_move(&work->w_list, &tmp);
1205 spin_unlock(&mp->m_sync_lock);
1207 list_for_each_entry_safe(work, n, &tmp, w_list) {
1208 (*work->w_syncer)(mp, work->w_data);
1209 list_del(&work->w_list);
1210 if (work == &mp->m_sync_work)
1221 struct super_block *sb)
1223 struct xfs_mount *mp = XFS_M(sb);
1224 struct xfs_inode *rip = mp->m_rootip;
1225 int unmount_event_flags = 0;
1228 kthread_stop(mp->m_sync_task);
1230 xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI);
1233 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1234 unmount_event_flags =
1235 (mp->m_dmevmask & (1 << DM_EVENT_UNMOUNT)) ?
1236 0 : DM_FLAGS_UNWANTED;
1238 * Ignore error from dmapi here, first unmount is not allowed
1239 * to fail anyway, and second we wouldn't want to fail a
1240 * unmount because of dmapi.
1242 XFS_SEND_PREUNMOUNT(mp, rip, DM_RIGHT_NULL, rip, DM_RIGHT_NULL,
1243 NULL, NULL, 0, 0, unmount_event_flags);
1248 * Blow away any referenced inode in the filestreams cache.
1249 * This can and will cause log traffic as inodes go inactive
1252 xfs_filestream_unmount(mp);
1254 XFS_bflush(mp->m_ddev_targp);
1255 error = xfs_unmount_flush(mp, 0);
1261 * If we're forcing a shutdown, typically because of a media error,
1262 * we want to make sure we invalidate dirty pages that belong to
1263 * referenced vnodes as well.
1265 if (XFS_FORCED_SHUTDOWN(mp)) {
1266 error = xfs_sync(mp, SYNC_WAIT | SYNC_CLOSE);
1267 ASSERT(error != EFSCORRUPTED);
1270 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1271 XFS_SEND_UNMOUNT(mp, rip, DM_RIGHT_NULL, 0, 0,
1272 unmount_event_flags);
1276 xfs_icsb_destroy_counters(mp);
1277 xfs_close_devices(mp);
1285 struct super_block *sb)
1287 if (!(sb->s_flags & MS_RDONLY))
1288 xfs_sync(XFS_M(sb), SYNC_FSDATA);
1294 struct super_block *sb,
1297 struct xfs_mount *mp = XFS_M(sb);
1302 * Treat a sync operation like a freeze. This is to work
1303 * around a race in sync_inodes() which works in two phases
1304 * - an asynchronous flush, which can write out an inode
1305 * without waiting for file size updates to complete, and a
1306 * synchronous flush, which wont do anything because the
1307 * async flush removed the inode's dirty flag. Also
1308 * sync_inodes() will not see any files that just have
1309 * outstanding transactions to be flushed because we don't
1310 * dirty the Linux inode until after the transaction I/O
1313 if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE)) {
1315 * First stage of freeze - no more writers will make progress
1316 * now we are here, so we flush delwri and delalloc buffers
1317 * here, then wait for all I/O to complete. Data is frozen at
1318 * that point. Metadata is not frozen, transactions can still
1319 * occur here so don't bother flushing the buftarg (i.e
1320 * SYNC_QUIESCE) because it'll just get dirty again.
1322 flags = SYNC_DATA_QUIESCE;
1324 flags = SYNC_FSDATA;
1326 error = xfs_sync(mp, flags);
1329 if (unlikely(laptop_mode)) {
1330 int prev_sync_seq = mp->m_sync_seq;
1333 * The disk must be active because we're syncing.
1334 * We schedule xfssyncd now (now that the disk is
1335 * active) instead of later (when it might not be).
1337 wake_up_process(mp->m_sync_task);
1339 * We have to wait for the sync iteration to complete.
1340 * If we don't, the disk activity caused by the sync
1341 * will come after the sync is completed, and that
1342 * triggers another sync from laptop mode.
1344 wait_event(mp->m_wait_single_sync_task,
1345 mp->m_sync_seq != prev_sync_seq);
1353 struct dentry *dentry,
1354 struct kstatfs *statp)
1356 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1357 xfs_sb_t *sbp = &mp->m_sb;
1358 __uint64_t fakeinos, id;
1361 statp->f_type = XFS_SB_MAGIC;
1362 statp->f_namelen = MAXNAMELEN - 1;
1364 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1365 statp->f_fsid.val[0] = (u32)id;
1366 statp->f_fsid.val[1] = (u32)(id >> 32);
1368 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1370 spin_lock(&mp->m_sb_lock);
1371 statp->f_bsize = sbp->sb_blocksize;
1372 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1373 statp->f_blocks = sbp->sb_dblocks - lsize;
1374 statp->f_bfree = statp->f_bavail =
1375 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1376 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1378 fakeinos += mp->m_inoadd;
1381 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1382 if (mp->m_maxicount)
1386 statp->f_files = min_t(typeof(statp->f_files),
1389 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1390 spin_unlock(&mp->m_sb_lock);
1392 XFS_QM_DQSTATVFS(XFS_I(dentry->d_inode), statp);
1398 struct super_block *sb,
1402 struct xfs_mount *mp = XFS_M(sb);
1403 struct xfs_mount_args *args;
1406 args = xfs_args_allocate(sb, 0);
1410 error = xfs_parseargs(mp, options, args, 1);
1414 if (!(*flags & MS_RDONLY)) { /* rw/ro -> rw */
1415 if (mp->m_flags & XFS_MOUNT_RDONLY)
1416 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1417 if (args->flags & XFSMNT_BARRIER) {
1418 mp->m_flags |= XFS_MOUNT_BARRIER;
1419 xfs_mountfs_check_barriers(mp);
1421 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1423 } else if (!(mp->m_flags & XFS_MOUNT_RDONLY)) { /* rw -> ro */
1424 xfs_filestream_flush(mp);
1425 xfs_sync(mp, SYNC_DATA_QUIESCE);
1426 xfs_attr_quiesce(mp);
1427 mp->m_flags |= XFS_MOUNT_RDONLY;
1436 * Second stage of a freeze. The data is already frozen so we only
1437 * need to take care of themetadata. Once that's done write a dummy
1438 * record to dirty the log in case of a crash while frozen.
1442 struct super_block *sb)
1444 struct xfs_mount *mp = XFS_M(sb);
1446 xfs_attr_quiesce(mp);
1447 xfs_fs_log_dummy(mp);
1451 xfs_fs_show_options(
1453 struct vfsmount *mnt)
1455 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1460 struct super_block *sb,
1463 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XQUOTASYNC, 0, NULL);
1468 struct super_block *sb,
1469 struct fs_quota_stat *fqs)
1471 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XGETQSTAT, 0, (caddr_t)fqs);
1476 struct super_block *sb,
1480 return -XFS_QM_QUOTACTL(XFS_M(sb), op, 0, (caddr_t)&flags);
1485 struct super_block *sb,
1488 struct fs_disk_quota *fdq)
1490 return -XFS_QM_QUOTACTL(XFS_M(sb),
1491 (type == USRQUOTA) ? Q_XGETQUOTA :
1492 ((type == GRPQUOTA) ? Q_XGETGQUOTA :
1493 Q_XGETPQUOTA), id, (caddr_t)fdq);
1498 struct super_block *sb,
1501 struct fs_disk_quota *fdq)
1503 return -XFS_QM_QUOTACTL(XFS_M(sb),
1504 (type == USRQUOTA) ? Q_XSETQLIM :
1505 ((type == GRPQUOTA) ? Q_XSETGQLIM :
1506 Q_XSETPQLIM), id, (caddr_t)fdq);
1510 * This function fills in xfs_mount_t fields based on mount args.
1511 * Note: the superblock has _not_ yet been read in.
1515 struct xfs_mount_args *ap,
1516 struct xfs_mount *mp)
1518 /* Values are in BBs */
1519 if ((ap->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
1521 * At this point the superblock has not been read
1522 * in, therefore we do not know the block size.
1523 * Before the mount call ends we will convert
1526 mp->m_dalign = ap->sunit;
1527 mp->m_swidth = ap->swidth;
1530 if (ap->logbufs != -1 &&
1532 (ap->logbufs < XLOG_MIN_ICLOGS ||
1533 ap->logbufs > XLOG_MAX_ICLOGS)) {
1535 "XFS: invalid logbufs value: %d [not %d-%d]",
1536 ap->logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
1537 return XFS_ERROR(EINVAL);
1539 mp->m_logbufs = ap->logbufs;
1540 if (ap->logbufsize != -1 &&
1541 ap->logbufsize != 0 &&
1542 (ap->logbufsize < XLOG_MIN_RECORD_BSIZE ||
1543 ap->logbufsize > XLOG_MAX_RECORD_BSIZE ||
1544 !is_power_of_2(ap->logbufsize))) {
1546 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
1548 return XFS_ERROR(EINVAL);
1550 mp->m_logbsize = ap->logbufsize;
1551 mp->m_fsname_len = strlen(ap->fsname) + 1;
1552 mp->m_fsname = kmem_alloc(mp->m_fsname_len, KM_SLEEP);
1553 strcpy(mp->m_fsname, ap->fsname);
1554 if (ap->rtname[0]) {
1555 mp->m_rtname = kmem_alloc(strlen(ap->rtname) + 1, KM_SLEEP);
1556 strcpy(mp->m_rtname, ap->rtname);
1558 if (ap->logname[0]) {
1559 mp->m_logname = kmem_alloc(strlen(ap->logname) + 1, KM_SLEEP);
1560 strcpy(mp->m_logname, ap->logname);
1563 if (ap->flags & XFSMNT_WSYNC)
1564 mp->m_flags |= XFS_MOUNT_WSYNC;
1566 if (ap->flags & XFSMNT_INO64) {
1567 mp->m_flags |= XFS_MOUNT_INO64;
1568 mp->m_inoadd = XFS_INO64_OFFSET;
1571 if (ap->flags & XFSMNT_RETERR)
1572 mp->m_flags |= XFS_MOUNT_RETERR;
1573 if (ap->flags & XFSMNT_NOALIGN)
1574 mp->m_flags |= XFS_MOUNT_NOALIGN;
1575 if (ap->flags & XFSMNT_SWALLOC)
1576 mp->m_flags |= XFS_MOUNT_SWALLOC;
1577 if (ap->flags & XFSMNT_OSYNCISOSYNC)
1578 mp->m_flags |= XFS_MOUNT_OSYNCISOSYNC;
1579 if (ap->flags & XFSMNT_32BITINODES)
1580 mp->m_flags |= XFS_MOUNT_32BITINODES;
1582 if (ap->flags & XFSMNT_IOSIZE) {
1583 if (ap->iosizelog > XFS_MAX_IO_LOG ||
1584 ap->iosizelog < XFS_MIN_IO_LOG) {
1586 "XFS: invalid log iosize: %d [not %d-%d]",
1587 ap->iosizelog, XFS_MIN_IO_LOG,
1589 return XFS_ERROR(EINVAL);
1592 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
1593 mp->m_readio_log = mp->m_writeio_log = ap->iosizelog;
1596 if (ap->flags & XFSMNT_IKEEP)
1597 mp->m_flags |= XFS_MOUNT_IKEEP;
1598 if (ap->flags & XFSMNT_DIRSYNC)
1599 mp->m_flags |= XFS_MOUNT_DIRSYNC;
1600 if (ap->flags & XFSMNT_ATTR2)
1601 mp->m_flags |= XFS_MOUNT_ATTR2;
1602 if (ap->flags & XFSMNT_NOATTR2)
1603 mp->m_flags |= XFS_MOUNT_NOATTR2;
1605 if (ap->flags2 & XFSMNT2_COMPAT_IOSIZE)
1606 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
1609 * no recovery flag requires a read-only mount
1611 if (ap->flags & XFSMNT_NORECOVERY) {
1612 if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
1614 "XFS: tried to mount a FS read-write without recovery!");
1615 return XFS_ERROR(EINVAL);
1617 mp->m_flags |= XFS_MOUNT_NORECOVERY;
1620 if (ap->flags & XFSMNT_NOUUID)
1621 mp->m_flags |= XFS_MOUNT_NOUUID;
1622 if (ap->flags & XFSMNT_BARRIER)
1623 mp->m_flags |= XFS_MOUNT_BARRIER;
1625 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1627 if (ap->flags2 & XFSMNT2_FILESTREAMS)
1628 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
1630 if (ap->flags & XFSMNT_DMAPI)
1631 mp->m_flags |= XFS_MOUNT_DMAPI;
1636 * This function fills in xfs_mount_t fields based on mount args.
1637 * Note: the superblock _has_ now been read in.
1641 struct xfs_mount_args *ap,
1642 struct xfs_mount *mp)
1644 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1646 /* Fail a mount where the logbuf is smaller then the log stripe */
1647 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1648 if ((ap->logbufsize <= 0) &&
1649 (mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE)) {
1650 mp->m_logbsize = mp->m_sb.sb_logsunit;
1651 } else if (ap->logbufsize > 0 &&
1652 ap->logbufsize < mp->m_sb.sb_logsunit) {
1654 "XFS: logbuf size must be greater than or equal to log stripe size");
1655 return XFS_ERROR(EINVAL);
1658 /* Fail a mount if the logbuf is larger than 32K */
1659 if (ap->logbufsize > XLOG_BIG_RECORD_BSIZE) {
1661 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1662 return XFS_ERROR(EINVAL);
1667 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1668 * told by noattr2 to turn it off
1670 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1671 !(ap->flags & XFSMNT_NOATTR2))
1672 mp->m_flags |= XFS_MOUNT_ATTR2;
1675 * prohibit r/w mounts of read-only filesystems
1677 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1679 "XFS: cannot mount a read-only filesystem as read-write");
1680 return XFS_ERROR(EROFS);
1684 * check for shared mount.
1686 if (ap->flags & XFSMNT_SHARED) {
1687 if (!xfs_sb_version_hasshared(&mp->m_sb))
1688 return XFS_ERROR(EINVAL);
1691 * For IRIX 6.5, shared mounts must have the shared
1692 * version bit set, have the persistent readonly
1693 * field set, must be version 0 and can only be mounted
1696 if (!ronly || !(mp->m_sb.sb_flags & XFS_SBF_READONLY) ||
1697 (mp->m_sb.sb_shared_vn != 0))
1698 return XFS_ERROR(EINVAL);
1700 mp->m_flags |= XFS_MOUNT_SHARED;
1703 * Shared XFS V0 can't deal with DMI. Return EINVAL.
1705 if (mp->m_sb.sb_shared_vn == 0 && (ap->flags & XFSMNT_DMAPI))
1706 return XFS_ERROR(EINVAL);
1709 if (ap->flags & XFSMNT_UQUOTA) {
1710 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
1711 if (ap->flags & XFSMNT_UQUOTAENF)
1712 mp->m_qflags |= XFS_UQUOTA_ENFD;
1715 if (ap->flags & XFSMNT_GQUOTA) {
1716 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
1717 if (ap->flags & XFSMNT_GQUOTAENF)
1718 mp->m_qflags |= XFS_OQUOTA_ENFD;
1719 } else if (ap->flags & XFSMNT_PQUOTA) {
1720 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
1721 if (ap->flags & XFSMNT_PQUOTAENF)
1722 mp->m_qflags |= XFS_OQUOTA_ENFD;
1730 struct super_block *sb,
1735 struct xfs_mount *mp = NULL;
1736 struct xfs_mount_args *args;
1737 int flags = 0, error = ENOMEM;
1739 args = xfs_args_allocate(sb, silent);
1743 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1747 spin_lock_init(&mp->m_sb_lock);
1748 mutex_init(&mp->m_ilock);
1749 mutex_init(&mp->m_growlock);
1750 atomic_set(&mp->m_active_trans, 0);
1751 INIT_LIST_HEAD(&mp->m_sync_list);
1752 spin_lock_init(&mp->m_sync_lock);
1753 init_waitqueue_head(&mp->m_wait_single_sync_task);
1758 if (sb->s_flags & MS_RDONLY)
1759 mp->m_flags |= XFS_MOUNT_RDONLY;
1761 error = xfs_parseargs(mp, (char *)data, args, 0);
1765 sb_min_blocksize(sb, BBSIZE);
1766 sb->s_export_op = &xfs_export_operations;
1767 sb->s_qcop = &xfs_quotactl_operations;
1768 sb->s_op = &xfs_super_operations;
1770 error = xfs_dmops_get(mp, args);
1773 error = xfs_qmops_get(mp, args);
1777 if (args->flags & XFSMNT_QUIET)
1778 flags |= XFS_MFSI_QUIET;
1780 error = xfs_open_devices(mp, args);
1784 if (xfs_icsb_init_counters(mp))
1785 mp->m_flags |= XFS_MOUNT_NO_PERCPU_SB;
1788 * Setup flags based on mount(2) options and then the superblock
1790 error = xfs_start_flags(args, mp);
1792 goto out_destroy_counters;
1793 error = xfs_readsb(mp, flags);
1795 goto out_destroy_counters;
1796 error = xfs_finish_flags(args, mp);
1800 error = xfs_setup_devices(mp);
1804 if (mp->m_flags & XFS_MOUNT_BARRIER)
1805 xfs_mountfs_check_barriers(mp);
1807 error = xfs_filestream_mount(mp);
1811 error = xfs_mountfs(mp, flags);
1813 goto out_filestream_unmount;
1815 XFS_SEND_MOUNT(mp, DM_RIGHT_NULL, args->mtpt, args->fsname);
1818 sb->s_magic = XFS_SB_MAGIC;
1819 sb->s_blocksize = mp->m_sb.sb_blocksize;
1820 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1821 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1822 sb->s_time_gran = 1;
1823 set_posix_acl_flag(sb);
1825 root = igrab(mp->m_rootip->i_vnode);
1830 if (is_bad_inode(root)) {
1834 sb->s_root = d_alloc_root(root);
1840 mp->m_sync_work.w_syncer = xfs_sync_worker;
1841 mp->m_sync_work.w_mount = mp;
1842 mp->m_sync_task = kthread_run(xfssyncd, mp, "xfssyncd");
1843 if (IS_ERR(mp->m_sync_task)) {
1844 error = -PTR_ERR(mp->m_sync_task);
1848 xfs_itrace_exit(XFS_I(sb->s_root->d_inode));
1853 out_filestream_unmount:
1854 xfs_filestream_unmount(mp);
1857 out_destroy_counters:
1858 xfs_icsb_destroy_counters(mp);
1859 xfs_close_devices(mp);
1880 * Blow away any referenced inode in the filestreams cache.
1881 * This can and will cause log traffic as inodes go inactive
1884 xfs_filestream_unmount(mp);
1886 XFS_bflush(mp->m_ddev_targp);
1887 error = xfs_unmount_flush(mp, 0);
1890 IRELE(mp->m_rootip);
1893 goto out_destroy_counters;
1898 struct file_system_type *fs_type,
1900 const char *dev_name,
1902 struct vfsmount *mnt)
1904 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1908 static struct super_operations xfs_super_operations = {
1909 .alloc_inode = xfs_fs_alloc_inode,
1910 .destroy_inode = xfs_fs_destroy_inode,
1911 .write_inode = xfs_fs_write_inode,
1912 .clear_inode = xfs_fs_clear_inode,
1913 .put_super = xfs_fs_put_super,
1914 .write_super = xfs_fs_write_super,
1915 .sync_fs = xfs_fs_sync_super,
1916 .write_super_lockfs = xfs_fs_lockfs,
1917 .statfs = xfs_fs_statfs,
1918 .remount_fs = xfs_fs_remount,
1919 .show_options = xfs_fs_show_options,
1922 static struct quotactl_ops xfs_quotactl_operations = {
1923 .quota_sync = xfs_fs_quotasync,
1924 .get_xstate = xfs_fs_getxstate,
1925 .set_xstate = xfs_fs_setxstate,
1926 .get_xquota = xfs_fs_getxquota,
1927 .set_xquota = xfs_fs_setxquota,
1930 static struct file_system_type xfs_fs_type = {
1931 .owner = THIS_MODULE,
1933 .get_sb = xfs_fs_get_sb,
1934 .kill_sb = kill_block_super,
1935 .fs_flags = FS_REQUIRES_DEV,
1943 static char message[] __initdata = KERN_INFO \
1944 XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n";
1950 error = xfs_init_zones();
1954 error = xfs_buf_init();
1963 error = register_filesystem(&xfs_fs_type);
1969 xfs_buf_terminate();
1972 xfs_destroy_zones();
1982 unregister_filesystem(&xfs_fs_type);
1984 xfs_buf_terminate();
1985 xfs_destroy_zones();
1989 module_init(init_xfs_fs);
1990 module_exit(exit_xfs_fs);
1992 MODULE_AUTHOR("Silicon Graphics, Inc.");
1993 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1994 MODULE_LICENSE("GPL");