2 * Copyright (c) 2007-2008 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * $DragonFly: src/sys/vfs/hammer/hammer_vfsops.c,v 1.63.2.2 2008/07/18 00:21:09 dillon Exp $
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/kernel.h>
40 #include <sys/vnode.h>
41 #include <sys/mount.h>
42 #include <sys/malloc.h>
43 #include <sys/nlookup.h>
44 #include <sys/fcntl.h>
45 #include <sys/sysctl.h>
51 int hammer_debug_general
;
52 int hammer_debug_debug
= 1; /* medium-error panics */
53 int hammer_debug_inode
;
54 int hammer_debug_locks
;
55 int hammer_debug_btree
;
57 int hammer_debug_recover
; /* -1 will disable, +1 will force */
58 int hammer_debug_recover_faults
;
59 int hammer_cluster_enable
= 1; /* enable read clustering by default */
60 int hammer_count_fsyncs
;
61 int hammer_count_inodes
;
62 int hammer_count_iqueued
;
63 int hammer_count_reclaiming
;
64 int hammer_count_records
;
65 int hammer_count_record_datas
;
66 int hammer_count_volumes
;
67 int hammer_count_buffers
;
68 int hammer_count_nodes
;
69 int64_t hammer_count_extra_space_used
;
70 int64_t hammer_stats_btree_lookups
;
71 int64_t hammer_stats_btree_searches
;
72 int64_t hammer_stats_btree_inserts
;
73 int64_t hammer_stats_btree_deletes
;
74 int64_t hammer_stats_btree_elements
;
75 int64_t hammer_stats_btree_splits
;
76 int64_t hammer_stats_btree_iterations
;
77 int64_t hammer_stats_record_iterations
;
79 int64_t hammer_stats_file_read
;
80 int64_t hammer_stats_file_write
;
81 int64_t hammer_stats_file_iopsr
;
82 int64_t hammer_stats_file_iopsw
;
83 int64_t hammer_stats_disk_read
;
84 int64_t hammer_stats_disk_write
;
85 int64_t hammer_stats_inode_flushes
;
86 int64_t hammer_stats_commits
;
88 int hammer_count_dirtybufspace
; /* global */
89 int hammer_count_refedbufs
; /* global */
90 int hammer_count_reservations
;
91 int hammer_count_io_running_read
;
92 int hammer_count_io_running_write
;
93 int hammer_count_io_locked
;
94 int hammer_limit_dirtybufspace
; /* per-mount */
95 int hammer_limit_recs
; /* as a whole XXX */
96 int hammer_limit_iqueued
; /* per-mount */
98 int hammer_verify_zone
;
99 int hammer_verify_data
= 1;
100 int hammer_write_mode
;
101 int64_t hammer_contention_count
;
102 int64_t hammer_zone_limit
;
104 SYSCTL_NODE(_vfs
, OID_AUTO
, hammer
, CTLFLAG_RW
, 0, "HAMMER filesystem");
105 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_general
, CTLFLAG_RW
,
106 &hammer_debug_general
, 0, "");
107 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_io
, CTLFLAG_RW
,
108 &hammer_debug_io
, 0, "");
109 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_debug
, CTLFLAG_RW
,
110 &hammer_debug_debug
, 0, "");
111 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_inode
, CTLFLAG_RW
,
112 &hammer_debug_inode
, 0, "");
113 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_locks
, CTLFLAG_RW
,
114 &hammer_debug_locks
, 0, "");
115 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_btree
, CTLFLAG_RW
,
116 &hammer_debug_btree
, 0, "");
117 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_tid
, CTLFLAG_RW
,
118 &hammer_debug_tid
, 0, "");
119 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_recover
, CTLFLAG_RW
,
120 &hammer_debug_recover
, 0, "");
121 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_recover_faults
, CTLFLAG_RW
,
122 &hammer_debug_recover_faults
, 0, "");
123 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, cluster_enable
, CTLFLAG_RW
,
124 &hammer_cluster_enable
, 0, "");
126 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, limit_dirtybufspace
, CTLFLAG_RW
,
127 &hammer_limit_dirtybufspace
, 0, "");
128 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, limit_recs
, CTLFLAG_RW
,
129 &hammer_limit_recs
, 0, "");
130 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, limit_iqueued
, CTLFLAG_RW
,
131 &hammer_limit_iqueued
, 0, "");
133 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_fsyncs
, CTLFLAG_RD
,
134 &hammer_count_fsyncs
, 0, "");
135 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_inodes
, CTLFLAG_RD
,
136 &hammer_count_inodes
, 0, "");
137 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_iqueued
, CTLFLAG_RD
,
138 &hammer_count_iqueued
, 0, "");
139 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_reclaiming
, CTLFLAG_RD
,
140 &hammer_count_reclaiming
, 0, "");
141 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_records
, CTLFLAG_RD
,
142 &hammer_count_records
, 0, "");
143 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_record_datas
, CTLFLAG_RD
,
144 &hammer_count_record_datas
, 0, "");
145 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_volumes
, CTLFLAG_RD
,
146 &hammer_count_volumes
, 0, "");
147 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_buffers
, CTLFLAG_RD
,
148 &hammer_count_buffers
, 0, "");
149 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_nodes
, CTLFLAG_RD
,
150 &hammer_count_nodes
, 0, "");
151 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, count_extra_space_used
, CTLFLAG_RD
,
152 &hammer_count_extra_space_used
, 0, "");
154 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_searches
, CTLFLAG_RD
,
155 &hammer_stats_btree_searches
, 0, "");
156 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_lookups
, CTLFLAG_RD
,
157 &hammer_stats_btree_lookups
, 0, "");
158 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_inserts
, CTLFLAG_RD
,
159 &hammer_stats_btree_inserts
, 0, "");
160 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_deletes
, CTLFLAG_RD
,
161 &hammer_stats_btree_deletes
, 0, "");
162 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_elements
, CTLFLAG_RD
,
163 &hammer_stats_btree_elements
, 0, "");
164 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_splits
, CTLFLAG_RD
,
165 &hammer_stats_btree_splits
, 0, "");
166 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_iterations
, CTLFLAG_RD
,
167 &hammer_stats_btree_iterations
, 0, "");
168 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_record_iterations
, CTLFLAG_RD
,
169 &hammer_stats_record_iterations
, 0, "");
171 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_file_read
, CTLFLAG_RD
,
172 &hammer_stats_file_read
, 0, "");
173 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_file_write
, CTLFLAG_RD
,
174 &hammer_stats_file_write
, 0, "");
175 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_file_iopsr
, CTLFLAG_RD
,
176 &hammer_stats_file_iopsr
, 0, "");
177 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_file_iopsw
, CTLFLAG_RD
,
178 &hammer_stats_file_iopsw
, 0, "");
179 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_disk_read
, CTLFLAG_RD
,
180 &hammer_stats_disk_read
, 0, "");
181 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_disk_write
, CTLFLAG_RD
,
182 &hammer_stats_disk_write
, 0, "");
183 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_inode_flushes
, CTLFLAG_RD
,
184 &hammer_stats_inode_flushes
, 0, "");
185 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_commits
, CTLFLAG_RD
,
186 &hammer_stats_commits
, 0, "");
188 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_dirtybufspace
, CTLFLAG_RD
,
189 &hammer_count_dirtybufspace
, 0, "");
190 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_refedbufs
, CTLFLAG_RD
,
191 &hammer_count_refedbufs
, 0, "");
192 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_reservations
, CTLFLAG_RD
,
193 &hammer_count_reservations
, 0, "");
194 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_io_running_read
, CTLFLAG_RD
,
195 &hammer_count_io_running_read
, 0, "");
196 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_io_locked
, CTLFLAG_RD
,
197 &hammer_count_io_locked
, 0, "");
198 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_io_running_write
, CTLFLAG_RD
,
199 &hammer_count_io_running_write
, 0, "");
200 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, zone_limit
, CTLFLAG_RW
,
201 &hammer_zone_limit
, 0, "");
202 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, contention_count
, CTLFLAG_RW
,
203 &hammer_contention_count
, 0, "");
204 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, verify_zone
, CTLFLAG_RW
,
205 &hammer_verify_zone
, 0, "");
206 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, verify_data
, CTLFLAG_RW
,
207 &hammer_verify_data
, 0, "");
208 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, write_mode
, CTLFLAG_RW
,
209 &hammer_write_mode
, 0, "");
211 KTR_INFO_MASTER(hammer
);
216 static void hammer_free_hmp(struct mount
*mp
);
218 static int hammer_vfs_mount(struct mount
*mp
, char *path
, caddr_t data
,
220 static int hammer_vfs_unmount(struct mount
*mp
, int mntflags
);
221 static int hammer_vfs_root(struct mount
*mp
, struct vnode
**vpp
);
222 static int hammer_vfs_statfs(struct mount
*mp
, struct statfs
*sbp
,
224 static int hammer_vfs_statvfs(struct mount
*mp
, struct statvfs
*sbp
,
226 static int hammer_vfs_sync(struct mount
*mp
, int waitfor
);
227 static int hammer_vfs_vget(struct mount
*mp
, ino_t ino
,
229 static int hammer_vfs_init(struct vfsconf
*conf
);
230 static int hammer_vfs_fhtovp(struct mount
*mp
, struct fid
*fhp
,
232 static int hammer_vfs_vptofh(struct vnode
*vp
, struct fid
*fhp
);
233 static int hammer_vfs_checkexp(struct mount
*mp
, struct sockaddr
*nam
,
234 int *exflagsp
, struct ucred
**credanonp
);
237 static struct vfsops hammer_vfsops
= {
238 .vfs_mount
= hammer_vfs_mount
,
239 .vfs_unmount
= hammer_vfs_unmount
,
240 .vfs_root
= hammer_vfs_root
,
241 .vfs_statfs
= hammer_vfs_statfs
,
242 .vfs_statvfs
= hammer_vfs_statvfs
,
243 .vfs_sync
= hammer_vfs_sync
,
244 .vfs_vget
= hammer_vfs_vget
,
245 .vfs_init
= hammer_vfs_init
,
246 .vfs_vptofh
= hammer_vfs_vptofh
,
247 .vfs_fhtovp
= hammer_vfs_fhtovp
,
248 .vfs_checkexp
= hammer_vfs_checkexp
251 MALLOC_DEFINE(M_HAMMER
, "hammer-mount", "hammer mount");
253 VFS_SET(hammer_vfsops
, hammer
, 0);
254 MODULE_VERSION(hammer
, 1);
257 hammer_vfs_init(struct vfsconf
*conf
)
261 if (hammer_limit_recs
== 0) {
262 hammer_limit_recs
= nbuf
* 25;
263 n
= kmalloc_limit(M_HAMMER
) / 512;
264 if (hammer_limit_recs
> n
)
265 hammer_limit_recs
= n
;
267 if (hammer_limit_dirtybufspace
== 0) {
268 hammer_limit_dirtybufspace
= hidirtybufspace
/ 2;
269 if (hammer_limit_dirtybufspace
< 100)
270 hammer_limit_dirtybufspace
= 100;
272 if (hammer_limit_iqueued
== 0)
273 hammer_limit_iqueued
= desiredvnodes
/ 5;
278 hammer_vfs_mount(struct mount
*mp
, char *mntpt
, caddr_t data
,
281 struct hammer_mount_info info
;
283 hammer_volume_t rootvol
;
284 struct vnode
*rootvp
;
285 const char *upath
; /* volume name in userspace */
286 char *path
; /* volume name in system space */
290 if ((error
= copyin(data
, &info
, sizeof(info
))) != 0)
292 if ((mp
->mnt_flag
& MNT_UPDATE
) == 0) {
293 if (info
.nvolumes
<= 0 || info
.nvolumes
>= 32768)
296 if ((info
.hflags
& HMNT_MASTERID
) &&
297 (info
.masterid
< -1 || info
.masterid
>= HAMMER_MAX_MASTERS
)) {
302 * Interal mount data structure
304 if (mp
->mnt_flag
& MNT_UPDATE
) {
305 hmp
= (void *)mp
->mnt_data
;
306 KKASSERT(hmp
!= NULL
);
308 hmp
= kmalloc(sizeof(*hmp
), M_HAMMER
, M_WAITOK
| M_ZERO
);
309 mp
->mnt_data
= (qaddr_t
)hmp
;
311 hmp
->namekey_iterator
= mycpu
->gd_time_seconds
;
312 /*TAILQ_INIT(&hmp->recycle_list);*/
314 hmp
->root_btree_beg
.localization
= 0x00000000U
;
315 hmp
->root_btree_beg
.obj_id
= -0x8000000000000000LL
;
316 hmp
->root_btree_beg
.key
= -0x8000000000000000LL
;
317 hmp
->root_btree_beg
.create_tid
= 1;
318 hmp
->root_btree_beg
.delete_tid
= 1;
319 hmp
->root_btree_beg
.rec_type
= 0;
320 hmp
->root_btree_beg
.obj_type
= 0;
322 hmp
->root_btree_end
.localization
= 0xFFFFFFFFU
;
323 hmp
->root_btree_end
.obj_id
= 0x7FFFFFFFFFFFFFFFLL
;
324 hmp
->root_btree_end
.key
= 0x7FFFFFFFFFFFFFFFLL
;
325 hmp
->root_btree_end
.create_tid
= 0xFFFFFFFFFFFFFFFFULL
;
326 hmp
->root_btree_end
.delete_tid
= 0; /* special case */
327 hmp
->root_btree_end
.rec_type
= 0xFFFFU
;
328 hmp
->root_btree_end
.obj_type
= 0;
330 hmp
->krate
.freq
= 1; /* maximum reporting rate (hz) */
331 hmp
->krate
.count
= -16; /* initial burst */
333 hmp
->sync_lock
.refs
= 1;
334 hmp
->free_lock
.refs
= 1;
335 hmp
->undo_lock
.refs
= 1;
336 hmp
->blkmap_lock
.refs
= 1;
338 TAILQ_INIT(&hmp
->delay_list
);
339 TAILQ_INIT(&hmp
->flush_group_list
);
340 TAILQ_INIT(&hmp
->objid_cache_list
);
341 TAILQ_INIT(&hmp
->undo_lru_list
);
342 TAILQ_INIT(&hmp
->reclaim_list
);
344 hmp
->hflags
&= ~HMNT_USERFLAGS
;
345 hmp
->hflags
|= info
.hflags
& HMNT_USERFLAGS
;
346 if (info
.hflags
& HMNT_MASTERID
)
347 hmp
->masterid
= info
.masterid
;
352 mp
->mnt_flag
|= MNT_RDONLY
;
353 hmp
->asof
= info
.asof
;
355 hmp
->asof
= HAMMER_MAX_TID
;
359 * Re-open read-write if originally read-only, or vise-versa.
361 if (mp
->mnt_flag
& MNT_UPDATE
) {
363 if (hmp
->ronly
&& (mp
->mnt_kern_flag
& MNTK_WANTRDWR
)) {
364 kprintf("HAMMER read-only -> read-write\n");
366 RB_SCAN(hammer_vol_rb_tree
, &hmp
->rb_vols_root
, NULL
,
367 hammer_adjust_volume_mode
, NULL
);
368 rootvol
= hammer_get_root_volume(hmp
, &error
);
370 hammer_recover_flush_buffers(hmp
, rootvol
, 1);
371 bcopy(rootvol
->ondisk
->vol0_blockmap
,
373 sizeof(hmp
->blockmap
));
374 hammer_rel_volume(rootvol
, 0);
376 RB_SCAN(hammer_ino_rb_tree
, &hmp
->rb_inos_root
, NULL
,
377 hammer_reload_inode
, NULL
);
378 /* kernel clears MNT_RDONLY */
379 } else if (hmp
->ronly
== 0 && (mp
->mnt_flag
& MNT_RDONLY
)) {
380 kprintf("HAMMER read-write -> read-only\n");
381 hmp
->ronly
= 1; /* messy */
382 RB_SCAN(hammer_ino_rb_tree
, &hmp
->rb_inos_root
, NULL
,
383 hammer_reload_inode
, NULL
);
385 hammer_flusher_sync(hmp
);
386 hammer_flusher_sync(hmp
);
387 hammer_flusher_sync(hmp
);
389 RB_SCAN(hammer_vol_rb_tree
, &hmp
->rb_vols_root
, NULL
,
390 hammer_adjust_volume_mode
, NULL
);
395 RB_INIT(&hmp
->rb_vols_root
);
396 RB_INIT(&hmp
->rb_inos_root
);
397 RB_INIT(&hmp
->rb_nods_root
);
398 RB_INIT(&hmp
->rb_undo_root
);
399 RB_INIT(&hmp
->rb_resv_root
);
400 RB_INIT(&hmp
->rb_bufs_root
);
401 RB_INIT(&hmp
->rb_pfsm_root
);
403 hmp
->ronly
= ((mp
->mnt_flag
& MNT_RDONLY
) != 0);
405 TAILQ_INIT(&hmp
->volu_list
);
406 TAILQ_INIT(&hmp
->undo_list
);
407 TAILQ_INIT(&hmp
->data_list
);
408 TAILQ_INIT(&hmp
->meta_list
);
409 TAILQ_INIT(&hmp
->lose_list
);
414 path
= objcache_get(namei_oc
, M_WAITOK
);
416 for (i
= 0; i
< info
.nvolumes
; ++i
) {
417 error
= copyin(&info
.volumes
[i
], &upath
, sizeof(char *));
419 error
= copyinstr(upath
, path
, MAXPATHLEN
, NULL
);
421 error
= hammer_install_volume(hmp
, path
);
425 objcache_put(namei_oc
, path
);
428 * Make sure we found a root volume
430 if (error
== 0 && hmp
->rootvol
== NULL
) {
431 kprintf("hammer_mount: No root volume found!\n");
436 * Check that all required volumes are available
438 if (error
== 0 && hammer_mountcheck_volumes(hmp
)) {
439 kprintf("hammer_mount: Missing volumes, cannot mount!\n");
449 * No errors, setup enough of the mount point so we can lookup the
452 mp
->mnt_iosize_max
= MAXPHYS
;
453 mp
->mnt_kern_flag
|= MNTK_FSMID
;
456 * note: f_iosize is used by vnode_pager_haspage() when constructing
459 mp
->mnt_stat
.f_iosize
= HAMMER_BUFSIZE
;
460 mp
->mnt_stat
.f_bsize
= HAMMER_BUFSIZE
;
462 mp
->mnt_vstat
.f_frsize
= HAMMER_BUFSIZE
;
463 mp
->mnt_vstat
.f_bsize
= HAMMER_BUFSIZE
;
465 mp
->mnt_maxsymlinklen
= 255;
466 mp
->mnt_flag
|= MNT_LOCAL
;
468 vfs_add_vnodeops(mp
, &hammer_vnode_vops
, &mp
->mnt_vn_norm_ops
);
469 vfs_add_vnodeops(mp
, &hammer_spec_vops
, &mp
->mnt_vn_spec_ops
);
470 vfs_add_vnodeops(mp
, &hammer_fifo_vops
, &mp
->mnt_vn_fifo_ops
);
473 * The root volume's ondisk pointer is only valid if we hold a
476 rootvol
= hammer_get_root_volume(hmp
, &error
);
481 * Perform any necessary UNDO operations. The recovery code does
482 * call hammer_undo_lookup() so we have to pre-cache the blockmap,
483 * and then re-copy it again after recovery is complete.
485 * If this is a read-only mount the UNDO information is retained
486 * in memory in the form of dirty buffer cache buffers, and not
487 * written back to the media.
489 bcopy(rootvol
->ondisk
->vol0_blockmap
, hmp
->blockmap
,
490 sizeof(hmp
->blockmap
));
493 * The undo_rec_limit limits the size of flush groups to avoid
494 * blowing out the UNDO FIFO. This calculation is typically in
495 * the tens of thousands and is designed primarily when small
496 * HAMMER filesystems are created.
498 hmp
->undo_rec_limit
= hammer_undo_max(hmp
) / 8192 + 100;
499 if (hammer_debug_general
& 0x0001)
500 kprintf("HAMMER: undo_rec_limit %d\n", hmp
->undo_rec_limit
);
502 error
= hammer_recover(hmp
, rootvol
);
504 kprintf("Failed to recover HAMMER filesystem on mount\n");
509 * Finish setup now that we have a good root volume.
511 * The top 16 bits of fsid.val[1] is a pfs id.
513 ksnprintf(mp
->mnt_stat
.f_mntfromname
,
514 sizeof(mp
->mnt_stat
.f_mntfromname
), "%s",
515 rootvol
->ondisk
->vol_name
);
516 mp
->mnt_stat
.f_fsid
.val
[0] =
517 crc32((char *)&rootvol
->ondisk
->vol_fsid
+ 0, 8);
518 mp
->mnt_stat
.f_fsid
.val
[1] =
519 crc32((char *)&rootvol
->ondisk
->vol_fsid
+ 8, 8);
520 mp
->mnt_stat
.f_fsid
.val
[1] &= 0x0000FFFF;
522 mp
->mnt_vstat
.f_fsid_uuid
= rootvol
->ondisk
->vol_fsid
;
523 mp
->mnt_vstat
.f_fsid
= crc32(&mp
->mnt_vstat
.f_fsid_uuid
,
524 sizeof(mp
->mnt_vstat
.f_fsid_uuid
));
527 * Certain often-modified fields in the root volume are cached in
528 * the hammer_mount structure so we do not have to generate lots
529 * of little UNDO structures for them.
531 * Recopy after recovery. This also has the side effect of
532 * setting our cached undo FIFO's first_offset, which serves to
533 * placemark the FIFO start for the NEXT flush cycle while the
534 * on-disk first_offset represents the LAST flush cycle.
536 hmp
->next_tid
= rootvol
->ondisk
->vol0_next_tid
;
537 bcopy(rootvol
->ondisk
->vol0_blockmap
, hmp
->blockmap
,
538 sizeof(hmp
->blockmap
));
539 hmp
->copy_stat_freebigblocks
= rootvol
->ondisk
->vol0_stat_freebigblocks
;
541 hammer_flusher_create(hmp
);
544 * Locate the root directory using the root cluster's B-Tree as a
545 * starting point. The root directory uses an obj_id of 1.
547 * FUTURE: Leave the root directory cached referenced but unlocked
548 * in hmp->rootvp (need to flush it on unmount).
550 error
= hammer_vfs_vget(mp
, 1, &rootvp
);
554 /*vn_unlock(hmp->rootvp);*/
557 hammer_rel_volume(rootvol
, 0);
560 * Cleanup and return.
568 hammer_vfs_unmount(struct mount
*mp
, int mntflags
)
571 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
577 * Clean out the vnodes
580 if (mntflags
& MNT_FORCE
)
582 if ((error
= vflush(mp
, 0, flags
)) != 0)
586 * Clean up the internal mount structure and related entities. This
594 * Clean up the internal mount structure and disassociate it from the mount.
595 * This may issue I/O.
598 hammer_free_hmp(struct mount
*mp
)
600 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
601 hammer_flush_group_t flg
;
605 * Flush anything dirty. This won't even run if the
606 * filesystem errored-out.
609 while (hammer_flusher_haswork(hmp
)) {
610 hammer_flusher_sync(hmp
);
614 kprintf("HAMMER: umount flushing.");
617 tsleep(hmp
, 0, "hmrufl", hz
);
620 kprintf("giving up\n");
624 if (count
>= 5 && count
< 30)
628 * If the mount had a critical error we have to destroy any
629 * remaining inodes before we can finish cleaning up the flusher.
631 if (hmp
->flags
& HAMMER_MOUNT_CRITICAL_ERROR
) {
632 RB_SCAN(hammer_ino_rb_tree
, &hmp
->rb_inos_root
, NULL
,
633 hammer_destroy_inode_callback
, NULL
);
637 * There shouldn't be any inodes left now and any left over
638 * flush groups should now be empty.
640 KKASSERT(RB_EMPTY(&hmp
->rb_inos_root
));
641 while ((flg
= TAILQ_FIRST(&hmp
->flush_group_list
)) != NULL
) {
642 TAILQ_REMOVE(&hmp
->flush_group_list
, flg
, flush_entry
);
643 KKASSERT(TAILQ_EMPTY(&flg
->flush_list
));
645 kprintf("HAMMER: Warning, flush_group %p was "
646 "not empty on umount!\n", flg
);
648 kfree(flg
, M_HAMMER
);
652 * We can finally destroy the flusher
654 hammer_flusher_destroy(hmp
);
657 * Unload buffers and then volumes
659 RB_SCAN(hammer_buf_rb_tree
, &hmp
->rb_bufs_root
, NULL
,
660 hammer_unload_buffer
, NULL
);
661 RB_SCAN(hammer_vol_rb_tree
, &hmp
->rb_vols_root
, NULL
,
662 hammer_unload_volume
, NULL
);
665 mp
->mnt_flag
&= ~MNT_LOCAL
;
667 hammer_destroy_objid_cache(hmp
);
668 kfree(hmp
, M_HAMMER
);
672 * Report critical errors. ip may be NULL.
675 hammer_critical_error(hammer_mount_t hmp
, hammer_inode_t ip
,
676 int error
, const char *msg
)
678 hmp
->flags
|= HAMMER_MOUNT_CRITICAL_ERROR
;
679 krateprintf(&hmp
->krate
,
680 "HAMMER(%s): Critical error inode=%lld %s\n",
681 hmp
->mp
->mnt_stat
.f_mntfromname
,
682 (ip
? ip
->obj_id
: -1), msg
);
683 if (hmp
->ronly
== 0) {
684 hmp
->ronly
= 2; /* special errored read-only mode */
685 hmp
->mp
->mnt_flag
|= MNT_RDONLY
;
686 kprintf("HAMMER(%s): Forcing read-only mode\n",
687 hmp
->mp
->mnt_stat
.f_mntfromname
);
694 * Obtain a vnode for the specified inode number. An exclusively locked
698 hammer_vfs_vget(struct mount
*mp
, ino_t ino
, struct vnode
**vpp
)
700 struct hammer_transaction trans
;
701 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
702 struct hammer_inode
*ip
;
705 hammer_simple_transaction(&trans
, hmp
);
708 * Lookup the requested HAMMER inode. The structure must be
709 * left unlocked while we manipulate the related vnode to avoid
712 ip
= hammer_get_inode(&trans
, NULL
, ino
,
713 hmp
->asof
, HAMMER_DEF_LOCALIZATION
,
719 error
= hammer_get_vnode(ip
, vpp
);
720 hammer_rel_inode(ip
, 0);
721 hammer_done_transaction(&trans
);
726 * Return the root vnode for the filesystem.
728 * HAMMER stores the root vnode in the hammer_mount structure so
729 * getting it is easy.
732 hammer_vfs_root(struct mount
*mp
, struct vnode
**vpp
)
735 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
739 error
= hammer_vfs_vget(mp
, 1, vpp
);
744 hammer_vfs_statfs(struct mount
*mp
, struct statfs
*sbp
, struct ucred
*cred
)
746 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
747 hammer_volume_t volume
;
748 hammer_volume_ondisk_t ondisk
;
752 volume
= hammer_get_root_volume(hmp
, &error
);
755 ondisk
= volume
->ondisk
;
760 mp
->mnt_stat
.f_files
= ondisk
->vol0_stat_inodes
;
761 bfree
= ondisk
->vol0_stat_freebigblocks
* HAMMER_LARGEBLOCK_SIZE
;
762 hammer_rel_volume(volume
, 0);
764 mp
->mnt_stat
.f_bfree
= bfree
/ HAMMER_BUFSIZE
;
765 mp
->mnt_stat
.f_bavail
= mp
->mnt_stat
.f_bfree
;
766 if (mp
->mnt_stat
.f_files
< 0)
767 mp
->mnt_stat
.f_files
= 0;
774 hammer_vfs_statvfs(struct mount
*mp
, struct statvfs
*sbp
, struct ucred
*cred
)
776 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
777 hammer_volume_t volume
;
778 hammer_volume_ondisk_t ondisk
;
782 volume
= hammer_get_root_volume(hmp
, &error
);
785 ondisk
= volume
->ondisk
;
790 mp
->mnt_vstat
.f_files
= ondisk
->vol0_stat_inodes
;
791 bfree
= ondisk
->vol0_stat_freebigblocks
* HAMMER_LARGEBLOCK_SIZE
;
792 hammer_rel_volume(volume
, 0);
794 mp
->mnt_vstat
.f_bfree
= bfree
/ HAMMER_BUFSIZE
;
795 mp
->mnt_vstat
.f_bavail
= mp
->mnt_stat
.f_bfree
;
796 if (mp
->mnt_vstat
.f_files
< 0)
797 mp
->mnt_vstat
.f_files
= 0;
798 *sbp
= mp
->mnt_vstat
;
803 * Sync the filesystem. Currently we have to run it twice, the second
804 * one will advance the undo start index to the end index, so if a crash
805 * occurs no undos will be run on mount.
807 * We do not sync the filesystem if we are called from a panic. If we did
808 * we might end up blowing up a sync that was already in progress.
811 hammer_vfs_sync(struct mount
*mp
, int waitfor
)
813 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
816 if (panicstr
== NULL
) {
817 error
= hammer_sync_hmp(hmp
, waitfor
);
819 error
= hammer_sync_hmp(hmp
, waitfor
);
827 * Convert a vnode to a file handle.
830 hammer_vfs_vptofh(struct vnode
*vp
, struct fid
*fhp
)
834 KKASSERT(MAXFIDSZ
>= 16);
836 fhp
->fid_len
= offsetof(struct fid
, fid_data
[16]);
837 fhp
->fid_ext
= ip
->obj_localization
>> 16;
838 bcopy(&ip
->obj_id
, fhp
->fid_data
+ 0, sizeof(ip
->obj_id
));
839 bcopy(&ip
->obj_asof
, fhp
->fid_data
+ 8, sizeof(ip
->obj_asof
));
845 * Convert a file handle back to a vnode.
848 hammer_vfs_fhtovp(struct mount
*mp
, struct fid
*fhp
, struct vnode
**vpp
)
850 struct hammer_transaction trans
;
851 struct hammer_inode
*ip
;
852 struct hammer_inode_info info
;
854 u_int32_t localization
;
856 bcopy(fhp
->fid_data
+ 0, &info
.obj_id
, sizeof(info
.obj_id
));
857 bcopy(fhp
->fid_data
+ 8, &info
.obj_asof
, sizeof(info
.obj_asof
));
858 localization
= (u_int32_t
)fhp
->fid_ext
<< 16;
860 hammer_simple_transaction(&trans
, (void *)mp
->mnt_data
);
863 * Get/allocate the hammer_inode structure. The structure must be
864 * unlocked while we manipulate the related vnode to avoid a
867 ip
= hammer_get_inode(&trans
, NULL
, info
.obj_id
,
868 info
.obj_asof
, localization
, 0, &error
);
873 error
= hammer_get_vnode(ip
, vpp
);
874 hammer_rel_inode(ip
, 0);
875 hammer_done_transaction(&trans
);
880 hammer_vfs_checkexp(struct mount
*mp
, struct sockaddr
*nam
,
881 int *exflagsp
, struct ucred
**credanonp
)
883 hammer_mount_t hmp
= (void *)mp
->mnt_data
;
887 np
= vfs_export_lookup(mp
, &hmp
->export
, nam
);
889 *exflagsp
= np
->netc_exflags
;
890 *credanonp
= &np
->netc_anon
;
900 hammer_vfs_export(struct mount
*mp
, int op
, const struct export_args
*export
)
902 hammer_mount_t hmp
= (void *)mp
->mnt_data
;
906 case MOUNTCTL_SET_EXPORT
:
907 error
= vfs_export(mp
, &hmp
->export
, export
);