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 spin_lock_init(&xfs_dabuf_global_lock);
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);
842 * xfs_sync flushes any pending I/O to file system vfsp.
844 * This routine is called by vfs_sync() to make sure that things make it
845 * out to disk eventually, on sync() system calls to flush out everything,
846 * and when the file system is unmounted. For the vfs_sync() case, all
847 * we really need to do is sync out the log to make all of our meta-data
848 * updates permanent (except for timestamps). For calls from pflushd(),
849 * dirty pages are kept moving by calling pdflush() on the inodes
850 * containing them. We also flush the inodes that we can lock without
851 * sleeping and the superblock if we can lock it without sleeping from
852 * vfs_sync() so that items at the tail of the log are always moving out.
855 * SYNC_BDFLUSH - We're being called from vfs_sync() so we don't want
856 * to sleep if we can help it. All we really need
857 * to do is ensure that the log is synced at least
858 * periodically. We also push the inodes and
859 * superblock if we can lock them without sleeping
860 * and they are not pinned.
861 * SYNC_ATTR - We need to flush the inodes. If SYNC_BDFLUSH is not
862 * set, then we really want to lock each inode and flush
864 * SYNC_WAIT - All the flushes that take place in this call should
866 * SYNC_DELWRI - This tells us to push dirty pages associated with
867 * inodes. SYNC_WAIT and SYNC_BDFLUSH are used to
868 * determine if they should be flushed sync, async, or
870 * SYNC_CLOSE - This flag is passed when the system is being
871 * unmounted. We should sync and invalidate everything.
872 * SYNC_FSDATA - This indicates that the caller would like to make
873 * sure the superblock is safe on disk. We can ensure
874 * this by simply making sure the log gets flushed
875 * if SYNC_BDFLUSH is set, and by actually writing it
877 * SYNC_IOWAIT - The caller wants us to wait for all data I/O to complete
878 * before we return (including direct I/O). Forms the drain
879 * side of the write barrier needed to safely quiesce the
891 * Get the Quota Manager to flush the dquots.
893 * If XFS quota support is not enabled or this filesystem
894 * instance does not use quotas XFS_QM_DQSYNC will always
897 error = XFS_QM_DQSYNC(mp, flags);
900 * If we got an IO error, we will be shutting down.
901 * So, there's nothing more for us to do here.
903 ASSERT(error != EIO || XFS_FORCED_SHUTDOWN(mp));
904 if (XFS_FORCED_SHUTDOWN(mp))
905 return XFS_ERROR(error);
908 if (flags & SYNC_IOWAIT)
909 xfs_filestream_flush(mp);
911 return xfs_syncsub(mp, flags, NULL);
915 * xfs sync routine for internal use
917 * This routine supports all of the flags defined for the generic vfs_sync
918 * interface as explained above under xfs_sync.
927 xfs_inode_t *ip = NULL;
928 bhv_vnode_t *vp = NULL;
933 uint base_lock_flags;
934 boolean_t mount_locked;
935 boolean_t vnode_refed;
937 xfs_iptr_t *ipointer;
939 boolean_t ipointer_in = B_FALSE;
941 #define IPOINTER_SET ipointer_in = B_TRUE
942 #define IPOINTER_CLR ipointer_in = B_FALSE
949 /* Insert a marker record into the inode list after inode ip. The list
950 * must be locked when this is called. After the call the list will no
953 #define IPOINTER_INSERT(ip, mp) { \
954 ASSERT(ipointer_in == B_FALSE); \
955 ipointer->ip_mnext = ip->i_mnext; \
956 ipointer->ip_mprev = ip; \
957 ip->i_mnext = (xfs_inode_t *)ipointer; \
958 ipointer->ip_mnext->i_mprev = (xfs_inode_t *)ipointer; \
960 XFS_MOUNT_IUNLOCK(mp); \
961 mount_locked = B_FALSE; \
965 /* Remove the marker from the inode list. If the marker was the only item
966 * in the list then there are no remaining inodes and we should zero out
967 * the whole list. If we are the current head of the list then move the head
970 #define IPOINTER_REMOVE(ip, mp) { \
971 ASSERT(ipointer_in == B_TRUE); \
972 if (ipointer->ip_mnext != (xfs_inode_t *)ipointer) { \
973 ip = ipointer->ip_mnext; \
974 ip->i_mprev = ipointer->ip_mprev; \
975 ipointer->ip_mprev->i_mnext = ip; \
976 if (mp->m_inodes == (xfs_inode_t *)ipointer) { \
980 ASSERT(mp->m_inodes == (xfs_inode_t *)ipointer); \
981 mp->m_inodes = NULL; \
987 #define XFS_PREEMPT_MASK 0x7f
989 ASSERT(!(flags & SYNC_BDFLUSH));
993 if (mp->m_flags & XFS_MOUNT_RDONLY)
999 /* Allocate a reference marker */
1000 ipointer = (xfs_iptr_t *)kmem_zalloc(sizeof(xfs_iptr_t), KM_SLEEP);
1002 fflag = XFS_B_ASYNC; /* default is don't wait */
1003 if (flags & SYNC_DELWRI)
1004 fflag = XFS_B_DELWRI;
1005 if (flags & SYNC_WAIT)
1006 fflag = 0; /* synchronous overrides all */
1008 base_lock_flags = XFS_ILOCK_SHARED;
1009 if (flags & (SYNC_DELWRI | SYNC_CLOSE)) {
1011 * We need the I/O lock if we're going to call any of
1012 * the flush/inval routines.
1014 base_lock_flags |= XFS_IOLOCK_SHARED;
1017 XFS_MOUNT_ILOCK(mp);
1021 mount_locked = B_TRUE;
1022 vnode_refed = B_FALSE;
1027 ASSERT(ipointer_in == B_FALSE);
1028 ASSERT(vnode_refed == B_FALSE);
1030 lock_flags = base_lock_flags;
1033 * There were no inodes in the list, just break out
1041 * We found another sync thread marker - skip it
1043 if (ip->i_mount == NULL) {
1048 vp = XFS_ITOV_NULL(ip);
1051 * If the vnode is gone then this is being torn down,
1052 * call reclaim if it is flushed, else let regular flush
1053 * code deal with it later in the loop.
1057 /* Skip ones already in reclaim */
1058 if (ip->i_flags & XFS_IRECLAIM) {
1062 if (xfs_ilock_nowait(ip, XFS_ILOCK_EXCL) == 0) {
1064 } else if ((xfs_ipincount(ip) == 0) &&
1065 xfs_iflock_nowait(ip)) {
1066 IPOINTER_INSERT(ip, mp);
1068 xfs_finish_reclaim(ip, 1,
1069 XFS_IFLUSH_DELWRI_ELSE_ASYNC);
1071 XFS_MOUNT_ILOCK(mp);
1072 mount_locked = B_TRUE;
1073 IPOINTER_REMOVE(ip, mp);
1075 xfs_iunlock(ip, XFS_ILOCK_EXCL);
1086 if (XFS_FORCED_SHUTDOWN(mp) && !(flags & SYNC_CLOSE)) {
1087 XFS_MOUNT_IUNLOCK(mp);
1088 kmem_free(ipointer, sizeof(xfs_iptr_t));
1093 * Try to lock without sleeping. We're out of order with
1094 * the inode list lock here, so if we fail we need to drop
1095 * the mount lock and try again. If we're called from
1096 * bdflush() here, then don't bother.
1098 * The inode lock here actually coordinates with the
1099 * almost spurious inode lock in xfs_ireclaim() to prevent
1100 * the vnode we handle here without a reference from
1101 * being freed while we reference it. If we lock the inode
1102 * while it's on the mount list here, then the spurious inode
1103 * lock in xfs_ireclaim() after the inode is pulled from
1104 * the mount list will sleep until we release it here.
1105 * This keeps the vnode from being freed while we reference
1108 if (xfs_ilock_nowait(ip, lock_flags) == 0) {
1120 IPOINTER_INSERT(ip, mp);
1121 xfs_ilock(ip, lock_flags);
1123 ASSERT(vp == XFS_ITOV(ip));
1124 ASSERT(ip->i_mount == mp);
1126 vnode_refed = B_TRUE;
1129 /* From here on in the loop we may have a marker record
1130 * in the inode list.
1134 * If we have to flush data or wait for I/O completion
1135 * we need to drop the ilock that we currently hold.
1136 * If we need to drop the lock, insert a marker if we
1137 * have not already done so.
1139 if ((flags & (SYNC_CLOSE|SYNC_IOWAIT)) ||
1140 ((flags & SYNC_DELWRI) && VN_DIRTY(vp))) {
1142 IPOINTER_INSERT(ip, mp);
1144 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1146 if (flags & SYNC_CLOSE) {
1147 /* Shutdown case. Flush and invalidate. */
1148 if (XFS_FORCED_SHUTDOWN(mp))
1149 xfs_tosspages(ip, 0, -1,
1152 error = xfs_flushinval_pages(ip,
1154 } else if ((flags & SYNC_DELWRI) && VN_DIRTY(vp)) {
1155 error = xfs_flush_pages(ip, 0,
1156 -1, fflag, FI_NONE);
1160 * When freezing, we need to wait ensure all I/O (including direct
1161 * I/O) is complete to ensure no further data modification can take
1162 * place after this point
1164 if (flags & SYNC_IOWAIT)
1167 xfs_ilock(ip, XFS_ILOCK_SHARED);
1170 if ((flags & SYNC_ATTR) &&
1171 (ip->i_update_core ||
1172 (ip->i_itemp && ip->i_itemp->ili_format.ilf_fields))) {
1174 IPOINTER_INSERT(ip, mp);
1176 if (flags & SYNC_WAIT) {
1178 error = xfs_iflush(ip, XFS_IFLUSH_SYNC);
1181 * If we can't acquire the flush lock, then the inode
1182 * is already being flushed so don't bother waiting.
1184 * If we can lock it then do a delwri flush so we can
1185 * combine multiple inode flushes in each disk write.
1187 } else if (xfs_iflock_nowait(ip)) {
1188 error = xfs_iflush(ip, XFS_IFLUSH_DELWRI);
1189 } else if (bypassed) {
1194 if (lock_flags != 0) {
1195 xfs_iunlock(ip, lock_flags);
1200 * If we had to take a reference on the vnode
1201 * above, then wait until after we've unlocked
1202 * the inode to release the reference. This is
1203 * because we can be already holding the inode
1204 * lock when VN_RELE() calls xfs_inactive().
1206 * Make sure to drop the mount lock before calling
1207 * VN_RELE() so that we don't trip over ourselves if
1208 * we have to go for the mount lock again in the
1212 IPOINTER_INSERT(ip, mp);
1217 vnode_refed = B_FALSE;
1225 * bail out if the filesystem is corrupted.
1227 if (error == EFSCORRUPTED) {
1228 if (!mount_locked) {
1229 XFS_MOUNT_ILOCK(mp);
1230 IPOINTER_REMOVE(ip, mp);
1232 XFS_MOUNT_IUNLOCK(mp);
1233 ASSERT(ipointer_in == B_FALSE);
1234 kmem_free(ipointer, sizeof(xfs_iptr_t));
1235 return XFS_ERROR(error);
1238 /* Let other threads have a chance at the mount lock
1239 * if we have looped many times without dropping the
1242 if ((++preempt & XFS_PREEMPT_MASK) == 0) {
1244 IPOINTER_INSERT(ip, mp);
1248 if (mount_locked == B_FALSE) {
1249 XFS_MOUNT_ILOCK(mp);
1250 mount_locked = B_TRUE;
1251 IPOINTER_REMOVE(ip, mp);
1255 ASSERT(ipointer_in == B_FALSE);
1258 } while (ip != mp->m_inodes);
1260 XFS_MOUNT_IUNLOCK(mp);
1262 ASSERT(ipointer_in == B_FALSE);
1264 kmem_free(ipointer, sizeof(xfs_iptr_t));
1265 return XFS_ERROR(last_error);
1269 * xfs sync routine for internal use
1271 * This routine supports all of the flags defined for the generic vfs_sync
1272 * interface as explained above under xfs_sync.
1283 uint log_flags = XFS_LOG_FORCE;
1285 xfs_buf_log_item_t *bip;
1288 * Sync out the log. This ensures that the log is periodically
1289 * flushed even if there is not enough activity to fill it up.
1291 if (flags & SYNC_WAIT)
1292 log_flags |= XFS_LOG_SYNC;
1294 xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
1296 if (flags & (SYNC_ATTR|SYNC_DELWRI)) {
1297 if (flags & SYNC_BDFLUSH)
1298 xfs_finish_reclaim_all(mp, 1);
1300 error = xfs_sync_inodes(mp, flags, bypassed);
1304 * Flushing out dirty data above probably generated more
1305 * log activity, so if this isn't vfs_sync() then flush
1308 if (flags & SYNC_DELWRI) {
1309 xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
1312 if (flags & SYNC_FSDATA) {
1314 * If this is vfs_sync() then only sync the superblock
1315 * if we can lock it without sleeping and it is not pinned.
1317 if (flags & SYNC_BDFLUSH) {
1318 bp = xfs_getsb(mp, XFS_BUF_TRYLOCK);
1320 bip = XFS_BUF_FSPRIVATE(bp,xfs_buf_log_item_t*);
1321 if ((bip != NULL) &&
1322 xfs_buf_item_dirty(bip)) {
1323 if (!(XFS_BUF_ISPINNED(bp))) {
1325 error = xfs_bwrite(mp, bp);
1334 bp = xfs_getsb(mp, 0);
1336 * If the buffer is pinned then push on the log so
1337 * we won't get stuck waiting in the write for
1338 * someone, maybe ourselves, to flush the log.
1339 * Even though we just pushed the log above, we
1340 * did not have the superblock buffer locked at
1341 * that point so it can become pinned in between
1344 if (XFS_BUF_ISPINNED(bp))
1345 xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE);
1346 if (flags & SYNC_WAIT)
1347 XFS_BUF_UNASYNC(bp);
1350 error = xfs_bwrite(mp, bp);
1358 * If this is the periodic sync, then kick some entries out of
1359 * the reference cache. This ensures that idle entries are
1360 * eventually kicked out of the cache.
1362 if (flags & SYNC_REFCACHE) {
1363 if (flags & SYNC_WAIT)
1364 xfs_refcache_purge_mp(mp);
1366 xfs_refcache_purge_some(mp);
1370 * If asked, update the disk superblock with incore counter values if we
1371 * are using non-persistent counters so that they don't get too far out
1372 * of sync if we crash or get a forced shutdown. We don't want to force
1373 * this to disk, just get a transaction into the iclogs....
1375 if (flags & SYNC_SUPER)
1376 xfs_log_sbcount(mp, 0);
1379 * Now check to see if the log needs a "dummy" transaction.
1382 if (!(flags & SYNC_REMOUNT) && xfs_log_need_covered(mp)) {
1387 * Put a dummy transaction in the log to tell
1388 * recovery that all others are OK.
1390 tp = xfs_trans_alloc(mp, XFS_TRANS_DUMMY1);
1391 if ((error = xfs_trans_reserve(tp, 0,
1392 XFS_ICHANGE_LOG_RES(mp),
1394 xfs_trans_cancel(tp, 0);
1399 xfs_ilock(ip, XFS_ILOCK_EXCL);
1401 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
1402 xfs_trans_ihold(tp, ip);
1403 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1404 error = xfs_trans_commit(tp, 0);
1405 xfs_iunlock(ip, XFS_ILOCK_EXCL);
1406 xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
1410 * When shutting down, we need to insure that the AIL is pushed
1411 * to disk or the filesystem can appear corrupt from the PROM.
1413 if ((flags & (SYNC_CLOSE|SYNC_WAIT)) == (SYNC_CLOSE|SYNC_WAIT)) {
1414 XFS_bflush(mp->m_ddev_targp);
1415 if (mp->m_rtdev_targp) {
1416 XFS_bflush(mp->m_rtdev_targp);
1420 return XFS_ERROR(last_error);
1424 * xfs_vget - called by DMAPI and NFSD to get vnode from file handle
1438 * Invalid. Since handles can be created in user space and passed in
1439 * via gethandle(), this is not cause for a panic.
1441 if (xfid->fid_len != sizeof(*xfid) - sizeof(xfid->fid_len))
1442 return XFS_ERROR(EINVAL);
1444 ino = xfid->fid_ino;
1445 igen = xfid->fid_gen;
1448 * NFS can sometimes send requests for ino 0. Fail them gracefully.
1451 return XFS_ERROR(ESTALE);
1453 error = xfs_iget(mp, NULL, ino, 0, XFS_ILOCK_SHARED, &ip, 0);
1461 return XFS_ERROR(EIO);
1464 if (ip->i_d.di_mode == 0 || ip->i_d.di_gen != igen) {
1465 xfs_iput_new(ip, XFS_ILOCK_SHARED);
1467 return XFS_ERROR(ENOENT);
1470 *vpp = XFS_ITOV(ip);
1471 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1476 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
1477 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
1478 #define MNTOPT_LOGDEV "logdev" /* log device */
1479 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
1480 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
1481 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
1482 #define MNTOPT_INO64 "ino64" /* force inodes into 64-bit range */
1483 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
1484 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
1485 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
1486 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
1487 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
1488 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
1489 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
1490 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
1491 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
1492 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
1493 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
1494 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
1495 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
1496 * unwritten extent conversion */
1497 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
1498 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
1499 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
1500 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
1501 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
1502 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
1503 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
1505 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
1506 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
1507 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
1508 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
1509 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
1510 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
1511 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
1512 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
1513 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
1514 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
1515 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
1516 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
1517 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
1518 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
1519 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
1520 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
1521 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
1522 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
1524 STATIC unsigned long
1525 suffix_strtoul(char *s, char **endp, unsigned int base)
1527 int last, shift_left_factor = 0;
1530 last = strlen(value) - 1;
1531 if (value[last] == 'K' || value[last] == 'k') {
1532 shift_left_factor = 10;
1535 if (value[last] == 'M' || value[last] == 'm') {
1536 shift_left_factor = 20;
1539 if (value[last] == 'G' || value[last] == 'g') {
1540 shift_left_factor = 30;
1544 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
1549 struct xfs_mount *mp,
1551 struct xfs_mount_args *args,
1554 char *this_char, *value, *eov;
1555 int dsunit, dswidth, vol_dsunit, vol_dswidth;
1559 args->flags |= XFSMNT_BARRIER;
1560 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
1565 iosize = dsunit = dswidth = vol_dsunit = vol_dswidth = 0;
1567 while ((this_char = strsep(&options, ",")) != NULL) {
1570 if ((value = strchr(this_char, '=')) != NULL)
1573 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
1574 if (!value || !*value) {
1576 "XFS: %s option requires an argument",
1580 args->logbufs = simple_strtoul(value, &eov, 10);
1581 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
1582 if (!value || !*value) {
1584 "XFS: %s option requires an argument",
1588 args->logbufsize = suffix_strtoul(value, &eov, 10);
1589 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
1590 if (!value || !*value) {
1592 "XFS: %s option requires an argument",
1596 strncpy(args->logname, value, MAXNAMELEN);
1597 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
1598 if (!value || !*value) {
1600 "XFS: %s option requires an argument",
1604 strncpy(args->mtpt, value, MAXNAMELEN);
1605 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
1606 if (!value || !*value) {
1608 "XFS: %s option requires an argument",
1612 strncpy(args->rtname, value, MAXNAMELEN);
1613 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
1614 if (!value || !*value) {
1616 "XFS: %s option requires an argument",
1620 iosize = simple_strtoul(value, &eov, 10);
1621 args->flags |= XFSMNT_IOSIZE;
1622 args->iosizelog = (uint8_t) iosize;
1623 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
1624 if (!value || !*value) {
1626 "XFS: %s option requires an argument",
1630 iosize = suffix_strtoul(value, &eov, 10);
1631 args->flags |= XFSMNT_IOSIZE;
1632 args->iosizelog = ffs(iosize) - 1;
1633 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
1634 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
1635 mp->m_flags |= XFS_MOUNT_GRPID;
1636 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
1637 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
1638 mp->m_flags &= ~XFS_MOUNT_GRPID;
1639 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
1640 args->flags |= XFSMNT_WSYNC;
1641 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
1642 args->flags |= XFSMNT_OSYNCISOSYNC;
1643 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
1644 args->flags |= XFSMNT_NORECOVERY;
1645 } else if (!strcmp(this_char, MNTOPT_INO64)) {
1646 args->flags |= XFSMNT_INO64;
1649 "XFS: %s option not allowed on this system",
1653 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
1654 args->flags |= XFSMNT_NOALIGN;
1655 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
1656 args->flags |= XFSMNT_SWALLOC;
1657 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
1658 if (!value || !*value) {
1660 "XFS: %s option requires an argument",
1664 dsunit = simple_strtoul(value, &eov, 10);
1665 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
1666 if (!value || !*value) {
1668 "XFS: %s option requires an argument",
1672 dswidth = simple_strtoul(value, &eov, 10);
1673 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
1674 args->flags &= ~XFSMNT_32BITINODES;
1677 "XFS: %s option not allowed on this system",
1681 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
1682 args->flags |= XFSMNT_NOUUID;
1683 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
1684 args->flags |= XFSMNT_BARRIER;
1685 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
1686 args->flags &= ~XFSMNT_BARRIER;
1687 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
1689 args->flags &= ~XFSMNT_IDELETE;
1690 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
1691 args->flags |= XFSMNT_IDELETE;
1692 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
1693 args->flags2 &= ~XFSMNT2_COMPAT_IOSIZE;
1694 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
1695 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
1696 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
1697 args->flags |= XFSMNT_ATTR2;
1698 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
1699 args->flags &= ~XFSMNT_ATTR2;
1700 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
1701 args->flags2 |= XFSMNT2_FILESTREAMS;
1702 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
1703 args->flags &= ~(XFSMNT_UQUOTAENF|XFSMNT_UQUOTA);
1704 args->flags &= ~(XFSMNT_GQUOTAENF|XFSMNT_GQUOTA);
1705 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
1706 !strcmp(this_char, MNTOPT_UQUOTA) ||
1707 !strcmp(this_char, MNTOPT_USRQUOTA)) {
1708 args->flags |= XFSMNT_UQUOTA | XFSMNT_UQUOTAENF;
1709 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
1710 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
1711 args->flags |= XFSMNT_UQUOTA;
1712 args->flags &= ~XFSMNT_UQUOTAENF;
1713 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
1714 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
1715 args->flags |= XFSMNT_PQUOTA | XFSMNT_PQUOTAENF;
1716 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
1717 args->flags |= XFSMNT_PQUOTA;
1718 args->flags &= ~XFSMNT_PQUOTAENF;
1719 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
1720 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
1721 args->flags |= XFSMNT_GQUOTA | XFSMNT_GQUOTAENF;
1722 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
1723 args->flags |= XFSMNT_GQUOTA;
1724 args->flags &= ~XFSMNT_GQUOTAENF;
1725 } else if (!strcmp(this_char, MNTOPT_DMAPI)) {
1726 args->flags |= XFSMNT_DMAPI;
1727 } else if (!strcmp(this_char, MNTOPT_XDSM)) {
1728 args->flags |= XFSMNT_DMAPI;
1729 } else if (!strcmp(this_char, MNTOPT_DMI)) {
1730 args->flags |= XFSMNT_DMAPI;
1731 } else if (!strcmp(this_char, "ihashsize")) {
1733 "XFS: ihashsize no longer used, option is deprecated.");
1734 } else if (!strcmp(this_char, "osyncisdsync")) {
1735 /* no-op, this is now the default */
1737 "XFS: osyncisdsync is now the default, option is deprecated.");
1738 } else if (!strcmp(this_char, "irixsgid")) {
1740 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
1743 "XFS: unknown mount option [%s].", this_char);
1748 if (args->flags & XFSMNT_NORECOVERY) {
1749 if ((mp->m_flags & XFS_MOUNT_RDONLY) == 0) {
1751 "XFS: no-recovery mounts must be read-only.");
1756 if ((args->flags & XFSMNT_NOALIGN) && (dsunit || dswidth)) {
1758 "XFS: sunit and swidth options incompatible with the noalign option");
1762 if ((args->flags & XFSMNT_GQUOTA) && (args->flags & XFSMNT_PQUOTA)) {
1764 "XFS: cannot mount with both project and group quota");
1768 if ((args->flags & XFSMNT_DMAPI) && *args->mtpt == '\0') {
1769 printk("XFS: %s option needs the mount point option as well\n",
1774 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
1776 "XFS: sunit and swidth must be specified together");
1780 if (dsunit && (dswidth % dsunit != 0)) {
1782 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
1788 * Applications using DMI filesystems often expect the
1789 * inode generation number to be monotonically increasing.
1790 * If we delete inode chunks we break this assumption, so
1791 * keep unused inode chunks on disk for DMI filesystems
1792 * until we come up with a better solution.
1793 * Note that if "ikeep" or "noikeep" mount options are
1794 * supplied, then they are honored.
1796 if (!(args->flags & XFSMNT_DMAPI) && !ikeep)
1797 args->flags |= XFSMNT_IDELETE;
1799 if ((args->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
1801 args->sunit = dsunit;
1802 args->flags |= XFSMNT_RETERR;
1804 args->sunit = vol_dsunit;
1806 dswidth ? (args->swidth = dswidth) :
1807 (args->swidth = vol_dswidth);
1809 args->sunit = args->swidth = 0;
1813 if (args->flags & XFSMNT_32BITINODES)
1814 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
1816 args->flags |= XFSMNT_FLAGS2;
1822 struct xfs_mount *mp,
1825 static struct proc_xfs_info {
1829 /* the few simple ones we can get from the mount struct */
1830 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
1831 { XFS_MOUNT_INO64, "," MNTOPT_INO64 },
1832 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
1833 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
1834 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
1835 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
1836 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
1839 struct proc_xfs_info *xfs_infop;
1841 for (xfs_infop = xfs_info; xfs_infop->flag; xfs_infop++) {
1842 if (mp->m_flags & xfs_infop->flag)
1843 seq_puts(m, xfs_infop->str);
1846 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
1847 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
1848 (int)(1 << mp->m_writeio_log) >> 10);
1850 if (mp->m_logbufs > 0)
1851 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
1852 if (mp->m_logbsize > 0)
1853 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
1856 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
1858 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
1860 if (mp->m_dalign > 0)
1861 seq_printf(m, "," MNTOPT_SUNIT "=%d",
1862 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
1863 if (mp->m_swidth > 0)
1864 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
1865 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
1867 if (!(mp->m_flags & XFS_MOUNT_IDELETE))
1868 seq_printf(m, "," MNTOPT_IKEEP);
1869 if (!(mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE))
1870 seq_printf(m, "," MNTOPT_LARGEIO);
1872 if (!(mp->m_flags & XFS_MOUNT_SMALL_INUMS))
1873 seq_printf(m, "," MNTOPT_64BITINODE);
1874 if (mp->m_flags & XFS_MOUNT_GRPID)
1875 seq_printf(m, "," MNTOPT_GRPID);
1877 if (mp->m_qflags & XFS_UQUOTA_ACCT) {
1878 if (mp->m_qflags & XFS_UQUOTA_ENFD)
1879 seq_puts(m, "," MNTOPT_USRQUOTA);
1881 seq_puts(m, "," MNTOPT_UQUOTANOENF);
1884 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
1885 if (mp->m_qflags & XFS_OQUOTA_ENFD)
1886 seq_puts(m, "," MNTOPT_PRJQUOTA);
1888 seq_puts(m, "," MNTOPT_PQUOTANOENF);
1891 if (mp->m_qflags & XFS_GQUOTA_ACCT) {
1892 if (mp->m_qflags & XFS_OQUOTA_ENFD)
1893 seq_puts(m, "," MNTOPT_GRPQUOTA);
1895 seq_puts(m, "," MNTOPT_GQUOTANOENF);
1898 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
1899 seq_puts(m, "," MNTOPT_NOQUOTA);
1901 if (mp->m_flags & XFS_MOUNT_DMAPI)
1902 seq_puts(m, "," MNTOPT_DMAPI);