2 * Copyright (c) 2000-2005 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
20 #include "xfs_types.h"
24 #include "xfs_trans.h"
28 #include "xfs_dmapi.h"
29 #include "xfs_mount.h"
30 #include "xfs_da_btree.h"
31 #include "xfs_bmap_btree.h"
32 #include "xfs_ialloc_btree.h"
33 #include "xfs_alloc_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_inode_item.h"
39 #include "xfs_btree.h"
40 #include "xfs_alloc.h"
41 #include "xfs_ialloc.h"
42 #include "xfs_quota.h"
43 #include "xfs_error.h"
46 #include "xfs_refcache.h"
47 #include "xfs_buf_item.h"
48 #include "xfs_log_priv.h"
49 #include "xfs_dir2_trace.h"
50 #include "xfs_extfree_item.h"
54 #include "xfs_mru_cache.h"
55 #include "xfs_filestream.h"
56 #include "xfs_fsops.h"
57 #include "xfs_vnodeops.h"
58 #include "xfs_vfsops.h"
64 extern kmem_zone_t
*xfs_bmap_free_item_zone
;
65 extern kmem_zone_t
*xfs_btree_cur_zone
;
66 extern kmem_zone_t
*xfs_trans_zone
;
67 extern kmem_zone_t
*xfs_buf_item_zone
;
68 extern kmem_zone_t
*xfs_dabuf_zone
;
69 #ifdef XFS_DABUF_DEBUG
70 extern spinlock_t xfs_dabuf_global_lock
;
71 spinlock_init(&xfs_dabuf_global_lock
, "xfsda");
75 * Initialize all of the zone allocators we use.
77 xfs_bmap_free_item_zone
= kmem_zone_init(sizeof(xfs_bmap_free_item_t
),
78 "xfs_bmap_free_item");
79 xfs_btree_cur_zone
= kmem_zone_init(sizeof(xfs_btree_cur_t
),
81 xfs_trans_zone
= kmem_zone_init(sizeof(xfs_trans_t
), "xfs_trans");
83 kmem_zone_init(sizeof(xfs_da_state_t
), "xfs_da_state");
84 xfs_dabuf_zone
= kmem_zone_init(sizeof(xfs_dabuf_t
), "xfs_dabuf");
85 xfs_ifork_zone
= kmem_zone_init(sizeof(xfs_ifork_t
), "xfs_ifork");
86 xfs_acl_zone_init(xfs_acl_zone
, "xfs_acl");
88 xfs_filestream_init();
91 * The size of the zone allocated buf log item is the maximum
92 * size possible under XFS. This wastes a little bit of memory,
93 * but it is much faster.
96 kmem_zone_init((sizeof(xfs_buf_log_item_t
) +
97 (((XFS_MAX_BLOCKSIZE
/ XFS_BLI_CHUNK
) /
98 NBWORD
) * sizeof(int))),
101 kmem_zone_init((sizeof(xfs_efd_log_item_t
) +
102 ((XFS_EFD_MAX_FAST_EXTENTS
- 1) *
103 sizeof(xfs_extent_t
))),
106 kmem_zone_init((sizeof(xfs_efi_log_item_t
) +
107 ((XFS_EFI_MAX_FAST_EXTENTS
- 1) *
108 sizeof(xfs_extent_t
))),
112 * These zones warrant special memory allocator hints
115 kmem_zone_init_flags(sizeof(xfs_inode_t
), "xfs_inode",
116 KM_ZONE_HWALIGN
| KM_ZONE_RECLAIM
|
117 KM_ZONE_SPREAD
, NULL
);
119 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t
), "xfs_ili",
120 KM_ZONE_SPREAD
, NULL
);
122 kmem_zone_init_flags(sizeof(xfs_icluster_t
), "xfs_icluster",
123 KM_ZONE_SPREAD
, NULL
);
126 * Allocate global trace buffers.
128 #ifdef XFS_ALLOC_TRACE
129 xfs_alloc_trace_buf
= ktrace_alloc(XFS_ALLOC_TRACE_SIZE
, KM_SLEEP
);
131 #ifdef XFS_BMAP_TRACE
132 xfs_bmap_trace_buf
= ktrace_alloc(XFS_BMAP_TRACE_SIZE
, KM_SLEEP
);
134 #ifdef XFS_BMBT_TRACE
135 xfs_bmbt_trace_buf
= ktrace_alloc(XFS_BMBT_TRACE_SIZE
, KM_SLEEP
);
137 #ifdef XFS_ATTR_TRACE
138 xfs_attr_trace_buf
= ktrace_alloc(XFS_ATTR_TRACE_SIZE
, KM_SLEEP
);
140 #ifdef XFS_DIR2_TRACE
141 xfs_dir2_trace_buf
= ktrace_alloc(XFS_DIR2_GTRACE_SIZE
, KM_SLEEP
);
146 #if (defined(DEBUG) || defined(INDUCE_IO_ERROR))
147 xfs_error_test_init();
148 #endif /* DEBUG || INDUCE_IO_ERROR */
151 xfs_sysctl_register();
158 extern kmem_zone_t
*xfs_bmap_free_item_zone
;
159 extern kmem_zone_t
*xfs_btree_cur_zone
;
160 extern kmem_zone_t
*xfs_inode_zone
;
161 extern kmem_zone_t
*xfs_trans_zone
;
162 extern kmem_zone_t
*xfs_da_state_zone
;
163 extern kmem_zone_t
*xfs_dabuf_zone
;
164 extern kmem_zone_t
*xfs_efd_zone
;
165 extern kmem_zone_t
*xfs_efi_zone
;
166 extern kmem_zone_t
*xfs_buf_item_zone
;
167 extern kmem_zone_t
*xfs_icluster_zone
;
169 xfs_cleanup_procfs();
170 xfs_sysctl_unregister();
171 xfs_refcache_destroy();
172 xfs_filestream_uninit();
173 xfs_mru_cache_uninit();
174 xfs_acl_zone_destroy(xfs_acl_zone
);
176 #ifdef XFS_DIR2_TRACE
177 ktrace_free(xfs_dir2_trace_buf
);
179 #ifdef XFS_ATTR_TRACE
180 ktrace_free(xfs_attr_trace_buf
);
182 #ifdef XFS_BMBT_TRACE
183 ktrace_free(xfs_bmbt_trace_buf
);
185 #ifdef XFS_BMAP_TRACE
186 ktrace_free(xfs_bmap_trace_buf
);
188 #ifdef XFS_ALLOC_TRACE
189 ktrace_free(xfs_alloc_trace_buf
);
192 kmem_zone_destroy(xfs_bmap_free_item_zone
);
193 kmem_zone_destroy(xfs_btree_cur_zone
);
194 kmem_zone_destroy(xfs_inode_zone
);
195 kmem_zone_destroy(xfs_trans_zone
);
196 kmem_zone_destroy(xfs_da_state_zone
);
197 kmem_zone_destroy(xfs_dabuf_zone
);
198 kmem_zone_destroy(xfs_buf_item_zone
);
199 kmem_zone_destroy(xfs_efd_zone
);
200 kmem_zone_destroy(xfs_efi_zone
);
201 kmem_zone_destroy(xfs_ifork_zone
);
202 kmem_zone_destroy(xfs_ili_zone
);
203 kmem_zone_destroy(xfs_icluster_zone
);
209 * This function fills in xfs_mount_t fields based on mount args.
210 * Note: the superblock has _not_ yet been read in.
214 struct xfs_mount_args
*ap
,
215 struct xfs_mount
*mp
)
217 /* Values are in BBs */
218 if ((ap
->flags
& XFSMNT_NOALIGN
) != XFSMNT_NOALIGN
) {
220 * At this point the superblock has not been read
221 * in, therefore we do not know the block size.
222 * Before the mount call ends we will convert
225 mp
->m_dalign
= ap
->sunit
;
226 mp
->m_swidth
= ap
->swidth
;
229 if (ap
->logbufs
!= -1 &&
231 (ap
->logbufs
< XLOG_MIN_ICLOGS
||
232 ap
->logbufs
> XLOG_MAX_ICLOGS
)) {
234 "XFS: invalid logbufs value: %d [not %d-%d]",
235 ap
->logbufs
, XLOG_MIN_ICLOGS
, XLOG_MAX_ICLOGS
);
236 return XFS_ERROR(EINVAL
);
238 mp
->m_logbufs
= ap
->logbufs
;
239 if (ap
->logbufsize
!= -1 &&
240 ap
->logbufsize
!= 0 &&
241 (ap
->logbufsize
< XLOG_MIN_RECORD_BSIZE
||
242 ap
->logbufsize
> XLOG_MAX_RECORD_BSIZE
||
243 !is_power_of_2(ap
->logbufsize
))) {
245 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
247 return XFS_ERROR(EINVAL
);
249 mp
->m_logbsize
= ap
->logbufsize
;
250 mp
->m_fsname_len
= strlen(ap
->fsname
) + 1;
251 mp
->m_fsname
= kmem_alloc(mp
->m_fsname_len
, KM_SLEEP
);
252 strcpy(mp
->m_fsname
, ap
->fsname
);
254 mp
->m_rtname
= kmem_alloc(strlen(ap
->rtname
) + 1, KM_SLEEP
);
255 strcpy(mp
->m_rtname
, ap
->rtname
);
257 if (ap
->logname
[0]) {
258 mp
->m_logname
= kmem_alloc(strlen(ap
->logname
) + 1, KM_SLEEP
);
259 strcpy(mp
->m_logname
, ap
->logname
);
262 if (ap
->flags
& XFSMNT_WSYNC
)
263 mp
->m_flags
|= XFS_MOUNT_WSYNC
;
265 if (ap
->flags
& XFSMNT_INO64
) {
266 mp
->m_flags
|= XFS_MOUNT_INO64
;
267 mp
->m_inoadd
= XFS_INO64_OFFSET
;
270 if (ap
->flags
& XFSMNT_RETERR
)
271 mp
->m_flags
|= XFS_MOUNT_RETERR
;
272 if (ap
->flags
& XFSMNT_NOALIGN
)
273 mp
->m_flags
|= XFS_MOUNT_NOALIGN
;
274 if (ap
->flags
& XFSMNT_SWALLOC
)
275 mp
->m_flags
|= XFS_MOUNT_SWALLOC
;
276 if (ap
->flags
& XFSMNT_OSYNCISOSYNC
)
277 mp
->m_flags
|= XFS_MOUNT_OSYNCISOSYNC
;
278 if (ap
->flags
& XFSMNT_32BITINODES
)
279 mp
->m_flags
|= XFS_MOUNT_32BITINODES
;
281 if (ap
->flags
& XFSMNT_IOSIZE
) {
282 if (ap
->iosizelog
> XFS_MAX_IO_LOG
||
283 ap
->iosizelog
< XFS_MIN_IO_LOG
) {
285 "XFS: invalid log iosize: %d [not %d-%d]",
286 ap
->iosizelog
, XFS_MIN_IO_LOG
,
288 return XFS_ERROR(EINVAL
);
291 mp
->m_flags
|= XFS_MOUNT_DFLT_IOSIZE
;
292 mp
->m_readio_log
= mp
->m_writeio_log
= ap
->iosizelog
;
295 if (ap
->flags
& XFSMNT_IDELETE
)
296 mp
->m_flags
|= XFS_MOUNT_IDELETE
;
297 if (ap
->flags
& XFSMNT_DIRSYNC
)
298 mp
->m_flags
|= XFS_MOUNT_DIRSYNC
;
299 if (ap
->flags
& XFSMNT_ATTR2
)
300 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
302 if (ap
->flags2
& XFSMNT2_COMPAT_IOSIZE
)
303 mp
->m_flags
|= XFS_MOUNT_COMPAT_IOSIZE
;
306 * no recovery flag requires a read-only mount
308 if (ap
->flags
& XFSMNT_NORECOVERY
) {
309 if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
)) {
311 "XFS: tried to mount a FS read-write without recovery!");
312 return XFS_ERROR(EINVAL
);
314 mp
->m_flags
|= XFS_MOUNT_NORECOVERY
;
317 if (ap
->flags
& XFSMNT_NOUUID
)
318 mp
->m_flags
|= XFS_MOUNT_NOUUID
;
319 if (ap
->flags
& XFSMNT_BARRIER
)
320 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
322 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
324 if (ap
->flags2
& XFSMNT2_FILESTREAMS
)
325 mp
->m_flags
|= XFS_MOUNT_FILESTREAMS
;
327 if (ap
->flags
& XFSMNT_DMAPI
)
328 mp
->m_flags
|= XFS_MOUNT_DMAPI
;
333 * This function fills in xfs_mount_t fields based on mount args.
334 * Note: the superblock _has_ now been read in.
338 struct xfs_mount_args
*ap
,
339 struct xfs_mount
*mp
)
341 int ronly
= (mp
->m_flags
& XFS_MOUNT_RDONLY
);
343 /* Fail a mount where the logbuf is smaller then the log stripe */
344 if (XFS_SB_VERSION_HASLOGV2(&mp
->m_sb
)) {
345 if ((ap
->logbufsize
<= 0) &&
346 (mp
->m_sb
.sb_logsunit
> XLOG_BIG_RECORD_BSIZE
)) {
347 mp
->m_logbsize
= mp
->m_sb
.sb_logsunit
;
348 } else if (ap
->logbufsize
> 0 &&
349 ap
->logbufsize
< mp
->m_sb
.sb_logsunit
) {
351 "XFS: logbuf size must be greater than or equal to log stripe size");
352 return XFS_ERROR(EINVAL
);
355 /* Fail a mount if the logbuf is larger than 32K */
356 if (ap
->logbufsize
> XLOG_BIG_RECORD_BSIZE
) {
358 "XFS: logbuf size for version 1 logs must be 16K or 32K");
359 return XFS_ERROR(EINVAL
);
363 if (XFS_SB_VERSION_HASATTR2(&mp
->m_sb
)) {
364 mp
->m_flags
|= XFS_MOUNT_ATTR2
;
368 * prohibit r/w mounts of read-only filesystems
370 if ((mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) && !ronly
) {
372 "XFS: cannot mount a read-only filesystem as read-write");
373 return XFS_ERROR(EROFS
);
377 * check for shared mount.
379 if (ap
->flags
& XFSMNT_SHARED
) {
380 if (!XFS_SB_VERSION_HASSHARED(&mp
->m_sb
))
381 return XFS_ERROR(EINVAL
);
384 * For IRIX 6.5, shared mounts must have the shared
385 * version bit set, have the persistent readonly
386 * field set, must be version 0 and can only be mounted
389 if (!ronly
|| !(mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) ||
390 (mp
->m_sb
.sb_shared_vn
!= 0))
391 return XFS_ERROR(EINVAL
);
393 mp
->m_flags
|= XFS_MOUNT_SHARED
;
396 * Shared XFS V0 can't deal with DMI. Return EINVAL.
398 if (mp
->m_sb
.sb_shared_vn
== 0 && (ap
->flags
& XFSMNT_DMAPI
))
399 return XFS_ERROR(EINVAL
);
402 if (ap
->flags
& XFSMNT_UQUOTA
) {
403 mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ACTIVE
);
404 if (ap
->flags
& XFSMNT_UQUOTAENF
)
405 mp
->m_qflags
|= XFS_UQUOTA_ENFD
;
408 if (ap
->flags
& XFSMNT_GQUOTA
) {
409 mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ACTIVE
);
410 if (ap
->flags
& XFSMNT_GQUOTAENF
)
411 mp
->m_qflags
|= XFS_OQUOTA_ENFD
;
412 } else if (ap
->flags
& XFSMNT_PQUOTA
) {
413 mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ACTIVE
);
414 if (ap
->flags
& XFSMNT_PQUOTAENF
)
415 mp
->m_qflags
|= XFS_OQUOTA_ENFD
;
424 * The file system configurations are:
425 * (1) device (partition) with data and internal log
426 * (2) logical volume with data and log subvolumes.
427 * (3) logical volume with data, log, and realtime subvolumes.
429 * We only have to handle opening the log and realtime volumes here if
430 * they are present. The data subvolume has already been opened by
431 * get_sb_bdev() and is stored in vfsp->vfs_super->s_bdev.
435 struct xfs_mount
*mp
,
436 struct xfs_mount_args
*args
,
439 struct block_device
*ddev
, *logdev
, *rtdev
;
440 int flags
= 0, error
;
442 ddev
= mp
->m_super
->s_bdev
;
443 logdev
= rtdev
= NULL
;
445 error
= xfs_dmops_get(mp
, args
);
448 error
= xfs_qmops_get(mp
, args
);
452 if (args
->flags
& XFSMNT_QUIET
)
453 flags
|= XFS_MFSI_QUIET
;
456 * Open real time and log devices - order is important.
458 if (args
->logname
[0]) {
459 error
= xfs_blkdev_get(mp
, args
->logname
, &logdev
);
463 if (args
->rtname
[0]) {
464 error
= xfs_blkdev_get(mp
, args
->rtname
, &rtdev
);
466 xfs_blkdev_put(logdev
);
470 if (rtdev
== ddev
|| rtdev
== logdev
) {
472 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
473 xfs_blkdev_put(logdev
);
474 xfs_blkdev_put(rtdev
);
480 * Setup xfs_mount buffer target pointers
483 mp
->m_ddev_targp
= xfs_alloc_buftarg(ddev
, 0);
484 if (!mp
->m_ddev_targp
) {
485 xfs_blkdev_put(logdev
);
486 xfs_blkdev_put(rtdev
);
490 mp
->m_rtdev_targp
= xfs_alloc_buftarg(rtdev
, 1);
491 if (!mp
->m_rtdev_targp
) {
492 xfs_blkdev_put(logdev
);
493 xfs_blkdev_put(rtdev
);
497 mp
->m_logdev_targp
= (logdev
&& logdev
!= ddev
) ?
498 xfs_alloc_buftarg(logdev
, 1) : mp
->m_ddev_targp
;
499 if (!mp
->m_logdev_targp
) {
500 xfs_blkdev_put(logdev
);
501 xfs_blkdev_put(rtdev
);
506 * Setup flags based on mount(2) options and then the superblock
508 error
= xfs_start_flags(args
, mp
);
511 error
= xfs_readsb(mp
, flags
);
514 error
= xfs_finish_flags(args
, mp
);
519 * Setup xfs_mount buffer target pointers based on superblock
521 error
= xfs_setsize_buftarg(mp
->m_ddev_targp
, mp
->m_sb
.sb_blocksize
,
522 mp
->m_sb
.sb_sectsize
);
523 if (!error
&& logdev
&& logdev
!= ddev
) {
524 unsigned int log_sector_size
= BBSIZE
;
526 if (XFS_SB_VERSION_HASSECTOR(&mp
->m_sb
))
527 log_sector_size
= mp
->m_sb
.sb_logsectsize
;
528 error
= xfs_setsize_buftarg(mp
->m_logdev_targp
,
529 mp
->m_sb
.sb_blocksize
,
533 error
= xfs_setsize_buftarg(mp
->m_rtdev_targp
,
534 mp
->m_sb
.sb_blocksize
,
535 mp
->m_sb
.sb_sectsize
);
539 if (mp
->m_flags
& XFS_MOUNT_BARRIER
)
540 xfs_mountfs_check_barriers(mp
);
542 if ((error
= xfs_filestream_mount(mp
)))
545 error
= xfs_mountfs(mp
, flags
);
549 XFS_SEND_MOUNT(mp
, DM_RIGHT_NULL
, args
->mtpt
, args
->fsname
);
557 xfs_binval(mp
->m_ddev_targp
);
558 if (logdev
&& logdev
!= ddev
)
559 xfs_binval(mp
->m_logdev_targp
);
561 xfs_binval(mp
->m_rtdev_targp
);
563 xfs_unmountfs_close(mp
, credp
);
577 int unmount_event_wanted
= 0;
578 int unmount_event_flags
= 0;
579 int xfs_unmountfs_needed
= 0;
586 if (mp
->m_flags
& XFS_MOUNT_DMAPI
) {
587 error
= XFS_SEND_PREUNMOUNT(mp
,
588 rvp
, DM_RIGHT_NULL
, rvp
, DM_RIGHT_NULL
,
590 (mp
->m_dmevmask
& (1<<DM_EVENT_PREUNMOUNT
))?
591 0:DM_FLAGS_UNWANTED
);
593 return XFS_ERROR(error
);
594 unmount_event_wanted
= 1;
595 unmount_event_flags
= (mp
->m_dmevmask
& (1<<DM_EVENT_UNMOUNT
))?
596 0 : DM_FLAGS_UNWANTED
;
600 * First blow any referenced inode from this file system
601 * out of the reference cache, and delete the timer.
603 xfs_refcache_purge_mp(mp
);
606 * Blow away any referenced inode in the filestreams cache.
607 * This can and will cause log traffic as inodes go inactive
610 xfs_filestream_unmount(mp
);
612 XFS_bflush(mp
->m_ddev_targp
);
613 error
= xfs_unmount_flush(mp
, 0);
617 ASSERT(vn_count(rvp
) == 1);
620 * Drop the reference count
625 * If we're forcing a shutdown, typically because of a media error,
626 * we want to make sure we invalidate dirty pages that belong to
627 * referenced vnodes as well.
629 if (XFS_FORCED_SHUTDOWN(mp
)) {
630 error
= xfs_sync(mp
, SYNC_WAIT
| SYNC_CLOSE
);
631 ASSERT(error
!= EFSCORRUPTED
);
633 xfs_unmountfs_needed
= 1;
636 /* Send DMAPI event, if required.
637 * Then do xfs_unmountfs() if needed.
638 * Then return error (or zero).
640 if (unmount_event_wanted
) {
641 /* Note: mp structure must still exist for
642 * XFS_SEND_UNMOUNT() call.
644 XFS_SEND_UNMOUNT(mp
, error
== 0 ? rvp
: NULL
,
645 DM_RIGHT_NULL
, 0, error
, unmount_event_flags
);
647 if (xfs_unmountfs_needed
) {
649 * Call common unmount function to flush to disk
650 * and free the super block buffer & mount structures.
652 xfs_unmountfs(mp
, credp
);
655 kmem_free(mp
, sizeof(xfs_mount_t
));
658 return XFS_ERROR(error
);
665 int count
= 0, pincount
;
667 xfs_refcache_purge_mp(mp
);
668 xfs_flush_buftarg(mp
->m_ddev_targp
, 0);
669 xfs_finish_reclaim_all(mp
, 0);
671 /* This loop must run at least twice.
672 * The first instance of the loop will flush
673 * most meta data but that will generate more
674 * meta data (typically directory updates).
675 * Which then must be flushed and logged before
676 * we can write the unmount record.
679 xfs_syncsub(mp
, SYNC_INODE_QUIESCE
, NULL
);
680 pincount
= xfs_flush_buftarg(mp
->m_ddev_targp
, 1);
691 * Second stage of a quiesce. The data is already synced, now we have to take
692 * care of the metadata. New transactions are already blocked, so we need to
693 * wait for any remaining transactions to drain out before proceding.
699 /* wait for all modifications to complete */
700 while (atomic_read(&mp
->m_active_trans
) > 0)
703 /* flush inodes and push all remaining buffers out to disk */
706 ASSERT_ALWAYS(atomic_read(&mp
->m_active_trans
) == 0);
708 /* Push the superblock and write an unmount record */
709 xfs_log_sbcount(mp
, 1);
710 xfs_log_unmount_write(mp
);
711 xfs_unmountfs_writesb(mp
);
716 struct xfs_mount
*mp
,
718 struct xfs_mount_args
*args
)
720 if (!(*flags
& MS_RDONLY
)) { /* rw/ro -> rw */
721 if (mp
->m_flags
& XFS_MOUNT_RDONLY
)
722 mp
->m_flags
&= ~XFS_MOUNT_RDONLY
;
723 if (args
->flags
& XFSMNT_BARRIER
) {
724 mp
->m_flags
|= XFS_MOUNT_BARRIER
;
725 xfs_mountfs_check_barriers(mp
);
727 mp
->m_flags
&= ~XFS_MOUNT_BARRIER
;
729 } else if (!(mp
->m_flags
& XFS_MOUNT_RDONLY
)) { /* rw -> ro */
730 xfs_filestream_flush(mp
);
731 xfs_sync(mp
, SYNC_DATA_QUIESCE
);
732 xfs_attr_quiesce(mp
);
733 mp
->m_flags
|= XFS_MOUNT_RDONLY
;
739 * xfs_unmount_flush implements a set of flush operation on special
740 * inodes, which are needed as a separate set of operations so that
741 * they can be called as part of relocation process.
745 xfs_mount_t
*mp
, /* Mount structure we are getting
747 int relocation
) /* Called from vfs relocation. */
749 xfs_inode_t
*rip
= mp
->m_rootip
;
751 xfs_inode_t
*rsumip
= NULL
;
752 bhv_vnode_t
*rvp
= XFS_ITOV(rip
);
755 xfs_ilock(rip
, XFS_ILOCK_EXCL
| XFS_ILOCK_PARENT
);
759 * Flush out the real time inodes.
761 if ((rbmip
= mp
->m_rbmip
) != NULL
) {
762 xfs_ilock(rbmip
, XFS_ILOCK_EXCL
);
764 error
= xfs_iflush(rbmip
, XFS_IFLUSH_SYNC
);
765 xfs_iunlock(rbmip
, XFS_ILOCK_EXCL
);
767 if (error
== EFSCORRUPTED
)
770 ASSERT(vn_count(XFS_ITOV(rbmip
)) == 1);
772 rsumip
= mp
->m_rsumip
;
773 xfs_ilock(rsumip
, XFS_ILOCK_EXCL
);
775 error
= xfs_iflush(rsumip
, XFS_IFLUSH_SYNC
);
776 xfs_iunlock(rsumip
, XFS_ILOCK_EXCL
);
778 if (error
== EFSCORRUPTED
)
781 ASSERT(vn_count(XFS_ITOV(rsumip
)) == 1);
785 * Synchronously flush root inode to disk
787 error
= xfs_iflush(rip
, XFS_IFLUSH_SYNC
);
788 if (error
== EFSCORRUPTED
)
791 if (vn_count(rvp
) != 1 && !relocation
) {
792 xfs_iunlock(rip
, XFS_ILOCK_EXCL
);
793 return XFS_ERROR(EBUSY
);
797 * Release dquot that rootinode, rbmino and rsumino might be holding,
798 * flush and purge the quota inodes.
800 error
= XFS_QM_UNMOUNT(mp
);
801 if (error
== EFSCORRUPTED
)
805 VN_RELE(XFS_ITOV(rbmip
));
806 VN_RELE(XFS_ITOV(rsumip
));
809 xfs_iunlock(rip
, XFS_ILOCK_EXCL
);
816 xfs_iunlock(rip
, XFS_ILOCK_EXCL
);
818 return XFS_ERROR(EFSCORRUPTED
);
822 * xfs_root extracts the root vnode from a vfs.
824 * vfsp -- the vfs struct for the desired file system
825 * vpp -- address of the caller's vnode pointer which should be
826 * set to the desired fs root vnode
835 vp
= XFS_ITOV(mp
->m_rootip
);
844 * Fill in the statvfs structure for the given file system. We use
845 * the superblock lock in the mount structure to ensure a consistent
846 * snapshot of the counters returned.
851 bhv_statvfs_t
*statp
,
860 statp
->f_type
= XFS_SB_MAGIC
;
862 xfs_icsb_sync_counters_flags(mp
, XFS_ICSB_LAZY_COUNT
);
863 spin_lock(&mp
->m_sb_lock
);
864 statp
->f_bsize
= sbp
->sb_blocksize
;
865 lsize
= sbp
->sb_logstart
? sbp
->sb_logblocks
: 0;
866 statp
->f_blocks
= sbp
->sb_dblocks
- lsize
;
867 statp
->f_bfree
= statp
->f_bavail
=
868 sbp
->sb_fdblocks
- XFS_ALLOC_SET_ASIDE(mp
);
869 fakeinos
= statp
->f_bfree
<< sbp
->sb_inopblog
;
871 fakeinos
+= mp
->m_inoadd
;
874 MIN(sbp
->sb_icount
+ fakeinos
, (__uint64_t
)XFS_MAXINUMBER
);
879 statp
->f_files
= min_t(typeof(statp
->f_files
),
882 statp
->f_ffree
= statp
->f_files
- (sbp
->sb_icount
- sbp
->sb_ifree
);
883 spin_unlock(&mp
->m_sb_lock
);
885 xfs_statvfs_fsid(statp
, mp
);
886 statp
->f_namelen
= MAXNAMELEN
- 1;
889 XFS_QM_DQSTATVFS(xfs_vtoi(vp
), statp
);
895 * xfs_sync flushes any pending I/O to file system vfsp.
897 * This routine is called by vfs_sync() to make sure that things make it
898 * out to disk eventually, on sync() system calls to flush out everything,
899 * and when the file system is unmounted. For the vfs_sync() case, all
900 * we really need to do is sync out the log to make all of our meta-data
901 * updates permanent (except for timestamps). For calls from pflushd(),
902 * dirty pages are kept moving by calling pdflush() on the inodes
903 * containing them. We also flush the inodes that we can lock without
904 * sleeping and the superblock if we can lock it without sleeping from
905 * vfs_sync() so that items at the tail of the log are always moving out.
908 * SYNC_BDFLUSH - We're being called from vfs_sync() so we don't want
909 * to sleep if we can help it. All we really need
910 * to do is ensure that the log is synced at least
911 * periodically. We also push the inodes and
912 * superblock if we can lock them without sleeping
913 * and they are not pinned.
914 * SYNC_ATTR - We need to flush the inodes. If SYNC_BDFLUSH is not
915 * set, then we really want to lock each inode and flush
917 * SYNC_WAIT - All the flushes that take place in this call should
919 * SYNC_DELWRI - This tells us to push dirty pages associated with
920 * inodes. SYNC_WAIT and SYNC_BDFLUSH are used to
921 * determine if they should be flushed sync, async, or
923 * SYNC_CLOSE - This flag is passed when the system is being
924 * unmounted. We should sync and invalidate everything.
925 * SYNC_FSDATA - This indicates that the caller would like to make
926 * sure the superblock is safe on disk. We can ensure
927 * this by simply making sure the log gets flushed
928 * if SYNC_BDFLUSH is set, and by actually writing it
930 * SYNC_IOWAIT - The caller wants us to wait for all data I/O to complete
931 * before we return (including direct I/O). Forms the drain
932 * side of the write barrier needed to safely quiesce the
944 * Get the Quota Manager to flush the dquots.
946 * If XFS quota support is not enabled or this filesystem
947 * instance does not use quotas XFS_QM_DQSYNC will always
950 error
= XFS_QM_DQSYNC(mp
, flags
);
953 * If we got an IO error, we will be shutting down.
954 * So, there's nothing more for us to do here.
956 ASSERT(error
!= EIO
|| XFS_FORCED_SHUTDOWN(mp
));
957 if (XFS_FORCED_SHUTDOWN(mp
))
958 return XFS_ERROR(error
);
961 if (flags
& SYNC_IOWAIT
)
962 xfs_filestream_flush(mp
);
964 return xfs_syncsub(mp
, flags
, NULL
);
968 * xfs sync routine for internal use
970 * This routine supports all of the flags defined for the generic vfs_sync
971 * interface as explained above under xfs_sync.
980 xfs_inode_t
*ip
= NULL
;
981 bhv_vnode_t
*vp
= NULL
;
986 uint base_lock_flags
;
987 boolean_t mount_locked
;
988 boolean_t vnode_refed
;
990 xfs_iptr_t
*ipointer
;
992 boolean_t ipointer_in
= B_FALSE
;
994 #define IPOINTER_SET ipointer_in = B_TRUE
995 #define IPOINTER_CLR ipointer_in = B_FALSE
1002 /* Insert a marker record into the inode list after inode ip. The list
1003 * must be locked when this is called. After the call the list will no
1006 #define IPOINTER_INSERT(ip, mp) { \
1007 ASSERT(ipointer_in == B_FALSE); \
1008 ipointer->ip_mnext = ip->i_mnext; \
1009 ipointer->ip_mprev = ip; \
1010 ip->i_mnext = (xfs_inode_t *)ipointer; \
1011 ipointer->ip_mnext->i_mprev = (xfs_inode_t *)ipointer; \
1013 XFS_MOUNT_IUNLOCK(mp); \
1014 mount_locked = B_FALSE; \
1018 /* Remove the marker from the inode list. If the marker was the only item
1019 * in the list then there are no remaining inodes and we should zero out
1020 * the whole list. If we are the current head of the list then move the head
1023 #define IPOINTER_REMOVE(ip, mp) { \
1024 ASSERT(ipointer_in == B_TRUE); \
1025 if (ipointer->ip_mnext != (xfs_inode_t *)ipointer) { \
1026 ip = ipointer->ip_mnext; \
1027 ip->i_mprev = ipointer->ip_mprev; \
1028 ipointer->ip_mprev->i_mnext = ip; \
1029 if (mp->m_inodes == (xfs_inode_t *)ipointer) { \
1030 mp->m_inodes = ip; \
1033 ASSERT(mp->m_inodes == (xfs_inode_t *)ipointer); \
1034 mp->m_inodes = NULL; \
1040 #define XFS_PREEMPT_MASK 0x7f
1042 ASSERT(!(flags
& SYNC_BDFLUSH
));
1046 if (mp
->m_flags
& XFS_MOUNT_RDONLY
)
1052 /* Allocate a reference marker */
1053 ipointer
= (xfs_iptr_t
*)kmem_zalloc(sizeof(xfs_iptr_t
), KM_SLEEP
);
1055 fflag
= XFS_B_ASYNC
; /* default is don't wait */
1056 if (flags
& SYNC_DELWRI
)
1057 fflag
= XFS_B_DELWRI
;
1058 if (flags
& SYNC_WAIT
)
1059 fflag
= 0; /* synchronous overrides all */
1061 base_lock_flags
= XFS_ILOCK_SHARED
;
1062 if (flags
& (SYNC_DELWRI
| SYNC_CLOSE
)) {
1064 * We need the I/O lock if we're going to call any of
1065 * the flush/inval routines.
1067 base_lock_flags
|= XFS_IOLOCK_SHARED
;
1070 XFS_MOUNT_ILOCK(mp
);
1074 mount_locked
= B_TRUE
;
1075 vnode_refed
= B_FALSE
;
1080 ASSERT(ipointer_in
== B_FALSE
);
1081 ASSERT(vnode_refed
== B_FALSE
);
1083 lock_flags
= base_lock_flags
;
1086 * There were no inodes in the list, just break out
1094 * We found another sync thread marker - skip it
1096 if (ip
->i_mount
== NULL
) {
1101 vp
= XFS_ITOV_NULL(ip
);
1104 * If the vnode is gone then this is being torn down,
1105 * call reclaim if it is flushed, else let regular flush
1106 * code deal with it later in the loop.
1110 /* Skip ones already in reclaim */
1111 if (ip
->i_flags
& XFS_IRECLAIM
) {
1115 if (xfs_ilock_nowait(ip
, XFS_ILOCK_EXCL
) == 0) {
1117 } else if ((xfs_ipincount(ip
) == 0) &&
1118 xfs_iflock_nowait(ip
)) {
1119 IPOINTER_INSERT(ip
, mp
);
1121 xfs_finish_reclaim(ip
, 1,
1122 XFS_IFLUSH_DELWRI_ELSE_ASYNC
);
1124 XFS_MOUNT_ILOCK(mp
);
1125 mount_locked
= B_TRUE
;
1126 IPOINTER_REMOVE(ip
, mp
);
1128 xfs_iunlock(ip
, XFS_ILOCK_EXCL
);
1139 if (XFS_FORCED_SHUTDOWN(mp
) && !(flags
& SYNC_CLOSE
)) {
1140 XFS_MOUNT_IUNLOCK(mp
);
1141 kmem_free(ipointer
, sizeof(xfs_iptr_t
));
1146 * Try to lock without sleeping. We're out of order with
1147 * the inode list lock here, so if we fail we need to drop
1148 * the mount lock and try again. If we're called from
1149 * bdflush() here, then don't bother.
1151 * The inode lock here actually coordinates with the
1152 * almost spurious inode lock in xfs_ireclaim() to prevent
1153 * the vnode we handle here without a reference from
1154 * being freed while we reference it. If we lock the inode
1155 * while it's on the mount list here, then the spurious inode
1156 * lock in xfs_ireclaim() after the inode is pulled from
1157 * the mount list will sleep until we release it here.
1158 * This keeps the vnode from being freed while we reference
1161 if (xfs_ilock_nowait(ip
, lock_flags
) == 0) {
1173 IPOINTER_INSERT(ip
, mp
);
1174 xfs_ilock(ip
, lock_flags
);
1176 ASSERT(vp
== XFS_ITOV(ip
));
1177 ASSERT(ip
->i_mount
== mp
);
1179 vnode_refed
= B_TRUE
;
1182 /* From here on in the loop we may have a marker record
1183 * in the inode list.
1187 * If we have to flush data or wait for I/O completion
1188 * we need to drop the ilock that we currently hold.
1189 * If we need to drop the lock, insert a marker if we
1190 * have not already done so.
1192 if ((flags
& (SYNC_CLOSE
|SYNC_IOWAIT
)) ||
1193 ((flags
& SYNC_DELWRI
) && VN_DIRTY(vp
))) {
1195 IPOINTER_INSERT(ip
, mp
);
1197 xfs_iunlock(ip
, XFS_ILOCK_SHARED
);
1199 if (flags
& SYNC_CLOSE
) {
1200 /* Shutdown case. Flush and invalidate. */
1201 if (XFS_FORCED_SHUTDOWN(mp
))
1202 xfs_tosspages(ip
, 0, -1,
1205 error
= xfs_flushinval_pages(ip
,
1207 } else if ((flags
& SYNC_DELWRI
) && VN_DIRTY(vp
)) {
1208 error
= xfs_flush_pages(ip
, 0,
1209 -1, fflag
, FI_NONE
);
1213 * When freezing, we need to wait ensure all I/O (including direct
1214 * I/O) is complete to ensure no further data modification can take
1215 * place after this point
1217 if (flags
& SYNC_IOWAIT
)
1220 xfs_ilock(ip
, XFS_ILOCK_SHARED
);
1223 if ((flags
& SYNC_ATTR
) &&
1224 (ip
->i_update_core
||
1225 (ip
->i_itemp
&& ip
->i_itemp
->ili_format
.ilf_fields
))) {
1227 IPOINTER_INSERT(ip
, mp
);
1229 if (flags
& SYNC_WAIT
) {
1231 error
= xfs_iflush(ip
, XFS_IFLUSH_SYNC
);
1234 * If we can't acquire the flush lock, then the inode
1235 * is already being flushed so don't bother waiting.
1237 * If we can lock it then do a delwri flush so we can
1238 * combine multiple inode flushes in each disk write.
1240 } else if (xfs_iflock_nowait(ip
)) {
1241 error
= xfs_iflush(ip
, XFS_IFLUSH_DELWRI
);
1242 } else if (bypassed
) {
1247 if (lock_flags
!= 0) {
1248 xfs_iunlock(ip
, lock_flags
);
1253 * If we had to take a reference on the vnode
1254 * above, then wait until after we've unlocked
1255 * the inode to release the reference. This is
1256 * because we can be already holding the inode
1257 * lock when VN_RELE() calls xfs_inactive().
1259 * Make sure to drop the mount lock before calling
1260 * VN_RELE() so that we don't trip over ourselves if
1261 * we have to go for the mount lock again in the
1265 IPOINTER_INSERT(ip
, mp
);
1270 vnode_refed
= B_FALSE
;
1278 * bail out if the filesystem is corrupted.
1280 if (error
== EFSCORRUPTED
) {
1281 if (!mount_locked
) {
1282 XFS_MOUNT_ILOCK(mp
);
1283 IPOINTER_REMOVE(ip
, mp
);
1285 XFS_MOUNT_IUNLOCK(mp
);
1286 ASSERT(ipointer_in
== B_FALSE
);
1287 kmem_free(ipointer
, sizeof(xfs_iptr_t
));
1288 return XFS_ERROR(error
);
1291 /* Let other threads have a chance at the mount lock
1292 * if we have looped many times without dropping the
1295 if ((++preempt
& XFS_PREEMPT_MASK
) == 0) {
1297 IPOINTER_INSERT(ip
, mp
);
1301 if (mount_locked
== B_FALSE
) {
1302 XFS_MOUNT_ILOCK(mp
);
1303 mount_locked
= B_TRUE
;
1304 IPOINTER_REMOVE(ip
, mp
);
1308 ASSERT(ipointer_in
== B_FALSE
);
1311 } while (ip
!= mp
->m_inodes
);
1313 XFS_MOUNT_IUNLOCK(mp
);
1315 ASSERT(ipointer_in
== B_FALSE
);
1317 kmem_free(ipointer
, sizeof(xfs_iptr_t
));
1318 return XFS_ERROR(last_error
);
1322 * xfs sync routine for internal use
1324 * This routine supports all of the flags defined for the generic vfs_sync
1325 * interface as explained above under xfs_sync.
1336 uint log_flags
= XFS_LOG_FORCE
;
1338 xfs_buf_log_item_t
*bip
;
1341 * Sync out the log. This ensures that the log is periodically
1342 * flushed even if there is not enough activity to fill it up.
1344 if (flags
& SYNC_WAIT
)
1345 log_flags
|= XFS_LOG_SYNC
;
1347 xfs_log_force(mp
, (xfs_lsn_t
)0, log_flags
);
1349 if (flags
& (SYNC_ATTR
|SYNC_DELWRI
)) {
1350 if (flags
& SYNC_BDFLUSH
)
1351 xfs_finish_reclaim_all(mp
, 1);
1353 error
= xfs_sync_inodes(mp
, flags
, bypassed
);
1357 * Flushing out dirty data above probably generated more
1358 * log activity, so if this isn't vfs_sync() then flush
1361 if (flags
& SYNC_DELWRI
) {
1362 xfs_log_force(mp
, (xfs_lsn_t
)0, log_flags
);
1365 if (flags
& SYNC_FSDATA
) {
1367 * If this is vfs_sync() then only sync the superblock
1368 * if we can lock it without sleeping and it is not pinned.
1370 if (flags
& SYNC_BDFLUSH
) {
1371 bp
= xfs_getsb(mp
, XFS_BUF_TRYLOCK
);
1373 bip
= XFS_BUF_FSPRIVATE(bp
,xfs_buf_log_item_t
*);
1374 if ((bip
!= NULL
) &&
1375 xfs_buf_item_dirty(bip
)) {
1376 if (!(XFS_BUF_ISPINNED(bp
))) {
1378 error
= xfs_bwrite(mp
, bp
);
1387 bp
= xfs_getsb(mp
, 0);
1389 * If the buffer is pinned then push on the log so
1390 * we won't get stuck waiting in the write for
1391 * someone, maybe ourselves, to flush the log.
1392 * Even though we just pushed the log above, we
1393 * did not have the superblock buffer locked at
1394 * that point so it can become pinned in between
1397 if (XFS_BUF_ISPINNED(bp
))
1398 xfs_log_force(mp
, (xfs_lsn_t
)0, XFS_LOG_FORCE
);
1399 if (flags
& SYNC_WAIT
)
1400 XFS_BUF_UNASYNC(bp
);
1403 error
= xfs_bwrite(mp
, bp
);
1411 * If this is the periodic sync, then kick some entries out of
1412 * the reference cache. This ensures that idle entries are
1413 * eventually kicked out of the cache.
1415 if (flags
& SYNC_REFCACHE
) {
1416 if (flags
& SYNC_WAIT
)
1417 xfs_refcache_purge_mp(mp
);
1419 xfs_refcache_purge_some(mp
);
1423 * If asked, update the disk superblock with incore counter values if we
1424 * are using non-persistent counters so that they don't get too far out
1425 * of sync if we crash or get a forced shutdown. We don't want to force
1426 * this to disk, just get a transaction into the iclogs....
1428 if (flags
& SYNC_SUPER
)
1429 xfs_log_sbcount(mp
, 0);
1432 * Now check to see if the log needs a "dummy" transaction.
1435 if (!(flags
& SYNC_REMOUNT
) && xfs_log_need_covered(mp
)) {
1440 * Put a dummy transaction in the log to tell
1441 * recovery that all others are OK.
1443 tp
= xfs_trans_alloc(mp
, XFS_TRANS_DUMMY1
);
1444 if ((error
= xfs_trans_reserve(tp
, 0,
1445 XFS_ICHANGE_LOG_RES(mp
),
1447 xfs_trans_cancel(tp
, 0);
1452 xfs_ilock(ip
, XFS_ILOCK_EXCL
);
1454 xfs_trans_ijoin(tp
, ip
, XFS_ILOCK_EXCL
);
1455 xfs_trans_ihold(tp
, ip
);
1456 xfs_trans_log_inode(tp
, ip
, XFS_ILOG_CORE
);
1457 error
= xfs_trans_commit(tp
, 0);
1458 xfs_iunlock(ip
, XFS_ILOCK_EXCL
);
1459 xfs_log_force(mp
, (xfs_lsn_t
)0, log_flags
);
1463 * When shutting down, we need to insure that the AIL is pushed
1464 * to disk or the filesystem can appear corrupt from the PROM.
1466 if ((flags
& (SYNC_CLOSE
|SYNC_WAIT
)) == (SYNC_CLOSE
|SYNC_WAIT
)) {
1467 XFS_bflush(mp
->m_ddev_targp
);
1468 if (mp
->m_rtdev_targp
) {
1469 XFS_bflush(mp
->m_rtdev_targp
);
1473 return XFS_ERROR(last_error
);
1477 * xfs_vget - called by DMAPI and NFSD to get vnode from file handle
1491 * Invalid. Since handles can be created in user space and passed in
1492 * via gethandle(), this is not cause for a panic.
1494 if (xfid
->fid_len
!= sizeof(*xfid
) - sizeof(xfid
->fid_len
))
1495 return XFS_ERROR(EINVAL
);
1497 ino
= xfid
->fid_ino
;
1498 igen
= xfid
->fid_gen
;
1501 * NFS can sometimes send requests for ino 0. Fail them gracefully.
1504 return XFS_ERROR(ESTALE
);
1506 error
= xfs_iget(mp
, NULL
, ino
, 0, XFS_ILOCK_SHARED
, &ip
, 0);
1514 return XFS_ERROR(EIO
);
1517 if (ip
->i_d
.di_mode
== 0 || ip
->i_d
.di_gen
!= igen
) {
1518 xfs_iput_new(ip
, XFS_ILOCK_SHARED
);
1520 return XFS_ERROR(ENOENT
);
1523 *vpp
= XFS_ITOV(ip
);
1524 xfs_iunlock(ip
, XFS_ILOCK_SHARED
);
1529 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
1530 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
1531 #define MNTOPT_LOGDEV "logdev" /* log device */
1532 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
1533 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
1534 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
1535 #define MNTOPT_INO64 "ino64" /* force inodes into 64-bit range */
1536 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
1537 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
1538 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
1539 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
1540 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
1541 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
1542 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
1543 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
1544 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
1545 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
1546 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
1547 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
1548 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
1549 * unwritten extent conversion */
1550 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
1551 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
1552 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
1553 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
1554 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
1555 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
1556 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
1558 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
1559 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
1560 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
1561 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
1562 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
1563 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
1564 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
1565 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
1566 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
1567 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
1568 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
1569 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
1570 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
1571 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
1572 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
1573 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
1574 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
1575 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
1577 STATIC
unsigned long
1578 suffix_strtoul(char *s
, char **endp
, unsigned int base
)
1580 int last
, shift_left_factor
= 0;
1583 last
= strlen(value
) - 1;
1584 if (value
[last
] == 'K' || value
[last
] == 'k') {
1585 shift_left_factor
= 10;
1588 if (value
[last
] == 'M' || value
[last
] == 'm') {
1589 shift_left_factor
= 20;
1592 if (value
[last
] == 'G' || value
[last
] == 'g') {
1593 shift_left_factor
= 30;
1597 return simple_strtoul((const char *)s
, endp
, base
) << shift_left_factor
;
1602 struct xfs_mount
*mp
,
1604 struct xfs_mount_args
*args
,
1607 char *this_char
, *value
, *eov
;
1608 int dsunit
, dswidth
, vol_dsunit
, vol_dswidth
;
1612 args
->flags
|= XFSMNT_BARRIER
;
1613 args
->flags2
|= XFSMNT2_COMPAT_IOSIZE
;
1618 iosize
= dsunit
= dswidth
= vol_dsunit
= vol_dswidth
= 0;
1620 while ((this_char
= strsep(&options
, ",")) != NULL
) {
1623 if ((value
= strchr(this_char
, '=')) != NULL
)
1626 if (!strcmp(this_char
, MNTOPT_LOGBUFS
)) {
1627 if (!value
|| !*value
) {
1629 "XFS: %s option requires an argument",
1633 args
->logbufs
= simple_strtoul(value
, &eov
, 10);
1634 } else if (!strcmp(this_char
, MNTOPT_LOGBSIZE
)) {
1635 if (!value
|| !*value
) {
1637 "XFS: %s option requires an argument",
1641 args
->logbufsize
= suffix_strtoul(value
, &eov
, 10);
1642 } else if (!strcmp(this_char
, MNTOPT_LOGDEV
)) {
1643 if (!value
|| !*value
) {
1645 "XFS: %s option requires an argument",
1649 strncpy(args
->logname
, value
, MAXNAMELEN
);
1650 } else if (!strcmp(this_char
, MNTOPT_MTPT
)) {
1651 if (!value
|| !*value
) {
1653 "XFS: %s option requires an argument",
1657 strncpy(args
->mtpt
, value
, MAXNAMELEN
);
1658 } else if (!strcmp(this_char
, MNTOPT_RTDEV
)) {
1659 if (!value
|| !*value
) {
1661 "XFS: %s option requires an argument",
1665 strncpy(args
->rtname
, value
, MAXNAMELEN
);
1666 } else if (!strcmp(this_char
, MNTOPT_BIOSIZE
)) {
1667 if (!value
|| !*value
) {
1669 "XFS: %s option requires an argument",
1673 iosize
= simple_strtoul(value
, &eov
, 10);
1674 args
->flags
|= XFSMNT_IOSIZE
;
1675 args
->iosizelog
= (uint8_t) iosize
;
1676 } else if (!strcmp(this_char
, MNTOPT_ALLOCSIZE
)) {
1677 if (!value
|| !*value
) {
1679 "XFS: %s option requires an argument",
1683 iosize
= suffix_strtoul(value
, &eov
, 10);
1684 args
->flags
|= XFSMNT_IOSIZE
;
1685 args
->iosizelog
= ffs(iosize
) - 1;
1686 } else if (!strcmp(this_char
, MNTOPT_GRPID
) ||
1687 !strcmp(this_char
, MNTOPT_BSDGROUPS
)) {
1688 mp
->m_flags
|= XFS_MOUNT_GRPID
;
1689 } else if (!strcmp(this_char
, MNTOPT_NOGRPID
) ||
1690 !strcmp(this_char
, MNTOPT_SYSVGROUPS
)) {
1691 mp
->m_flags
&= ~XFS_MOUNT_GRPID
;
1692 } else if (!strcmp(this_char
, MNTOPT_WSYNC
)) {
1693 args
->flags
|= XFSMNT_WSYNC
;
1694 } else if (!strcmp(this_char
, MNTOPT_OSYNCISOSYNC
)) {
1695 args
->flags
|= XFSMNT_OSYNCISOSYNC
;
1696 } else if (!strcmp(this_char
, MNTOPT_NORECOVERY
)) {
1697 args
->flags
|= XFSMNT_NORECOVERY
;
1698 } else if (!strcmp(this_char
, MNTOPT_INO64
)) {
1699 args
->flags
|= XFSMNT_INO64
;
1702 "XFS: %s option not allowed on this system",
1706 } else if (!strcmp(this_char
, MNTOPT_NOALIGN
)) {
1707 args
->flags
|= XFSMNT_NOALIGN
;
1708 } else if (!strcmp(this_char
, MNTOPT_SWALLOC
)) {
1709 args
->flags
|= XFSMNT_SWALLOC
;
1710 } else if (!strcmp(this_char
, MNTOPT_SUNIT
)) {
1711 if (!value
|| !*value
) {
1713 "XFS: %s option requires an argument",
1717 dsunit
= simple_strtoul(value
, &eov
, 10);
1718 } else if (!strcmp(this_char
, MNTOPT_SWIDTH
)) {
1719 if (!value
|| !*value
) {
1721 "XFS: %s option requires an argument",
1725 dswidth
= simple_strtoul(value
, &eov
, 10);
1726 } else if (!strcmp(this_char
, MNTOPT_64BITINODE
)) {
1727 args
->flags
&= ~XFSMNT_32BITINODES
;
1730 "XFS: %s option not allowed on this system",
1734 } else if (!strcmp(this_char
, MNTOPT_NOUUID
)) {
1735 args
->flags
|= XFSMNT_NOUUID
;
1736 } else if (!strcmp(this_char
, MNTOPT_BARRIER
)) {
1737 args
->flags
|= XFSMNT_BARRIER
;
1738 } else if (!strcmp(this_char
, MNTOPT_NOBARRIER
)) {
1739 args
->flags
&= ~XFSMNT_BARRIER
;
1740 } else if (!strcmp(this_char
, MNTOPT_IKEEP
)) {
1742 args
->flags
&= ~XFSMNT_IDELETE
;
1743 } else if (!strcmp(this_char
, MNTOPT_NOIKEEP
)) {
1744 args
->flags
|= XFSMNT_IDELETE
;
1745 } else if (!strcmp(this_char
, MNTOPT_LARGEIO
)) {
1746 args
->flags2
&= ~XFSMNT2_COMPAT_IOSIZE
;
1747 } else if (!strcmp(this_char
, MNTOPT_NOLARGEIO
)) {
1748 args
->flags2
|= XFSMNT2_COMPAT_IOSIZE
;
1749 } else if (!strcmp(this_char
, MNTOPT_ATTR2
)) {
1750 args
->flags
|= XFSMNT_ATTR2
;
1751 } else if (!strcmp(this_char
, MNTOPT_NOATTR2
)) {
1752 args
->flags
&= ~XFSMNT_ATTR2
;
1753 } else if (!strcmp(this_char
, MNTOPT_FILESTREAM
)) {
1754 args
->flags2
|= XFSMNT2_FILESTREAMS
;
1755 } else if (!strcmp(this_char
, MNTOPT_NOQUOTA
)) {
1756 args
->flags
&= ~(XFSMNT_UQUOTAENF
|XFSMNT_UQUOTA
);
1757 args
->flags
&= ~(XFSMNT_GQUOTAENF
|XFSMNT_GQUOTA
);
1758 } else if (!strcmp(this_char
, MNTOPT_QUOTA
) ||
1759 !strcmp(this_char
, MNTOPT_UQUOTA
) ||
1760 !strcmp(this_char
, MNTOPT_USRQUOTA
)) {
1761 args
->flags
|= XFSMNT_UQUOTA
| XFSMNT_UQUOTAENF
;
1762 } else if (!strcmp(this_char
, MNTOPT_QUOTANOENF
) ||
1763 !strcmp(this_char
, MNTOPT_UQUOTANOENF
)) {
1764 args
->flags
|= XFSMNT_UQUOTA
;
1765 args
->flags
&= ~XFSMNT_UQUOTAENF
;
1766 } else if (!strcmp(this_char
, MNTOPT_PQUOTA
) ||
1767 !strcmp(this_char
, MNTOPT_PRJQUOTA
)) {
1768 args
->flags
|= XFSMNT_PQUOTA
| XFSMNT_PQUOTAENF
;
1769 } else if (!strcmp(this_char
, MNTOPT_PQUOTANOENF
)) {
1770 args
->flags
|= XFSMNT_PQUOTA
;
1771 args
->flags
&= ~XFSMNT_PQUOTAENF
;
1772 } else if (!strcmp(this_char
, MNTOPT_GQUOTA
) ||
1773 !strcmp(this_char
, MNTOPT_GRPQUOTA
)) {
1774 args
->flags
|= XFSMNT_GQUOTA
| XFSMNT_GQUOTAENF
;
1775 } else if (!strcmp(this_char
, MNTOPT_GQUOTANOENF
)) {
1776 args
->flags
|= XFSMNT_GQUOTA
;
1777 args
->flags
&= ~XFSMNT_GQUOTAENF
;
1778 } else if (!strcmp(this_char
, MNTOPT_DMAPI
)) {
1779 args
->flags
|= XFSMNT_DMAPI
;
1780 } else if (!strcmp(this_char
, MNTOPT_XDSM
)) {
1781 args
->flags
|= XFSMNT_DMAPI
;
1782 } else if (!strcmp(this_char
, MNTOPT_DMI
)) {
1783 args
->flags
|= XFSMNT_DMAPI
;
1784 } else if (!strcmp(this_char
, "ihashsize")) {
1786 "XFS: ihashsize no longer used, option is deprecated.");
1787 } else if (!strcmp(this_char
, "osyncisdsync")) {
1788 /* no-op, this is now the default */
1790 "XFS: osyncisdsync is now the default, option is deprecated.");
1791 } else if (!strcmp(this_char
, "irixsgid")) {
1793 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
1796 "XFS: unknown mount option [%s].", this_char
);
1801 if (args
->flags
& XFSMNT_NORECOVERY
) {
1802 if ((mp
->m_flags
& XFS_MOUNT_RDONLY
) == 0) {
1804 "XFS: no-recovery mounts must be read-only.");
1809 if ((args
->flags
& XFSMNT_NOALIGN
) && (dsunit
|| dswidth
)) {
1811 "XFS: sunit and swidth options incompatible with the noalign option");
1815 if ((args
->flags
& XFSMNT_GQUOTA
) && (args
->flags
& XFSMNT_PQUOTA
)) {
1817 "XFS: cannot mount with both project and group quota");
1821 if ((args
->flags
& XFSMNT_DMAPI
) && *args
->mtpt
== '\0') {
1822 printk("XFS: %s option needs the mount point option as well\n",
1827 if ((dsunit
&& !dswidth
) || (!dsunit
&& dswidth
)) {
1829 "XFS: sunit and swidth must be specified together");
1833 if (dsunit
&& (dswidth
% dsunit
!= 0)) {
1835 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
1841 * Applications using DMI filesystems often expect the
1842 * inode generation number to be monotonically increasing.
1843 * If we delete inode chunks we break this assumption, so
1844 * keep unused inode chunks on disk for DMI filesystems
1845 * until we come up with a better solution.
1846 * Note that if "ikeep" or "noikeep" mount options are
1847 * supplied, then they are honored.
1849 if (!(args
->flags
& XFSMNT_DMAPI
) && !ikeep
)
1850 args
->flags
|= XFSMNT_IDELETE
;
1852 if ((args
->flags
& XFSMNT_NOALIGN
) != XFSMNT_NOALIGN
) {
1854 args
->sunit
= dsunit
;
1855 args
->flags
|= XFSMNT_RETERR
;
1857 args
->sunit
= vol_dsunit
;
1859 dswidth
? (args
->swidth
= dswidth
) :
1860 (args
->swidth
= vol_dswidth
);
1862 args
->sunit
= args
->swidth
= 0;
1866 if (args
->flags
& XFSMNT_32BITINODES
)
1867 mp
->m_flags
|= XFS_MOUNT_SMALL_INUMS
;
1869 args
->flags
|= XFSMNT_FLAGS2
;
1875 struct xfs_mount
*mp
,
1878 static struct proc_xfs_info
{
1882 /* the few simple ones we can get from the mount struct */
1883 { XFS_MOUNT_WSYNC
, "," MNTOPT_WSYNC
},
1884 { XFS_MOUNT_INO64
, "," MNTOPT_INO64
},
1885 { XFS_MOUNT_NOALIGN
, "," MNTOPT_NOALIGN
},
1886 { XFS_MOUNT_SWALLOC
, "," MNTOPT_SWALLOC
},
1887 { XFS_MOUNT_NOUUID
, "," MNTOPT_NOUUID
},
1888 { XFS_MOUNT_NORECOVERY
, "," MNTOPT_NORECOVERY
},
1889 { XFS_MOUNT_OSYNCISOSYNC
, "," MNTOPT_OSYNCISOSYNC
},
1892 struct proc_xfs_info
*xfs_infop
;
1894 for (xfs_infop
= xfs_info
; xfs_infop
->flag
; xfs_infop
++) {
1895 if (mp
->m_flags
& xfs_infop
->flag
)
1896 seq_puts(m
, xfs_infop
->str
);
1899 if (mp
->m_flags
& XFS_MOUNT_DFLT_IOSIZE
)
1900 seq_printf(m
, "," MNTOPT_ALLOCSIZE
"=%dk",
1901 (int)(1 << mp
->m_writeio_log
) >> 10);
1903 if (mp
->m_logbufs
> 0)
1904 seq_printf(m
, "," MNTOPT_LOGBUFS
"=%d", mp
->m_logbufs
);
1905 if (mp
->m_logbsize
> 0)
1906 seq_printf(m
, "," MNTOPT_LOGBSIZE
"=%dk", mp
->m_logbsize
>> 10);
1909 seq_printf(m
, "," MNTOPT_LOGDEV
"=%s", mp
->m_logname
);
1911 seq_printf(m
, "," MNTOPT_RTDEV
"=%s", mp
->m_rtname
);
1913 if (mp
->m_dalign
> 0)
1914 seq_printf(m
, "," MNTOPT_SUNIT
"=%d",
1915 (int)XFS_FSB_TO_BB(mp
, mp
->m_dalign
));
1916 if (mp
->m_swidth
> 0)
1917 seq_printf(m
, "," MNTOPT_SWIDTH
"=%d",
1918 (int)XFS_FSB_TO_BB(mp
, mp
->m_swidth
));
1920 if (!(mp
->m_flags
& XFS_MOUNT_IDELETE
))
1921 seq_printf(m
, "," MNTOPT_IKEEP
);
1922 if (!(mp
->m_flags
& XFS_MOUNT_COMPAT_IOSIZE
))
1923 seq_printf(m
, "," MNTOPT_LARGEIO
);
1925 if (!(mp
->m_flags
& XFS_MOUNT_SMALL_INUMS
))
1926 seq_printf(m
, "," MNTOPT_64BITINODE
);
1927 if (mp
->m_flags
& XFS_MOUNT_GRPID
)
1928 seq_printf(m
, "," MNTOPT_GRPID
);
1930 if (mp
->m_qflags
& XFS_UQUOTA_ACCT
) {
1931 if (mp
->m_qflags
& XFS_UQUOTA_ENFD
)
1932 seq_puts(m
, "," MNTOPT_USRQUOTA
);
1934 seq_puts(m
, "," MNTOPT_UQUOTANOENF
);
1937 if (mp
->m_qflags
& XFS_PQUOTA_ACCT
) {
1938 if (mp
->m_qflags
& XFS_OQUOTA_ENFD
)
1939 seq_puts(m
, "," MNTOPT_PRJQUOTA
);
1941 seq_puts(m
, "," MNTOPT_PQUOTANOENF
);
1944 if (mp
->m_qflags
& XFS_GQUOTA_ACCT
) {
1945 if (mp
->m_qflags
& XFS_OQUOTA_ENFD
)
1946 seq_puts(m
, "," MNTOPT_GRPQUOTA
);
1948 seq_puts(m
, "," MNTOPT_GQUOTANOENF
);
1951 if (!(mp
->m_qflags
& XFS_ALL_QUOTA_ACCT
))
1952 seq_puts(m
, "," MNTOPT_NOQUOTA
);
1954 if (mp
->m_flags
& XFS_MOUNT_DMAPI
)
1955 seq_puts(m
, "," MNTOPT_DMAPI
);
1960 * Second stage of a freeze. The data is already frozen so we only
1961 * need to take care of themetadata. Once that's done write a dummy
1962 * record to dirty the log in case of a crash while frozen.
1968 xfs_attr_quiesce(mp
);
1969 xfs_fs_log_dummy(mp
);