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 1.63.2.7 2008/09/25 01:42:52 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 */
97 int hammer_autoflush
= 2000; /* auto flush */
99 int hammer_verify_zone
;
100 int hammer_verify_data
= 1;
101 int hammer_write_mode
;
102 int64_t hammer_contention_count
;
103 int64_t hammer_zone_limit
;
105 SYSCTL_NODE(_vfs
, OID_AUTO
, hammer
, CTLFLAG_RW
, 0, "HAMMER filesystem");
106 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_general
, CTLFLAG_RW
,
107 &hammer_debug_general
, 0, "");
108 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_io
, CTLFLAG_RW
,
109 &hammer_debug_io
, 0, "");
110 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_debug
, CTLFLAG_RW
,
111 &hammer_debug_debug
, 0, "");
112 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_inode
, CTLFLAG_RW
,
113 &hammer_debug_inode
, 0, "");
114 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_locks
, CTLFLAG_RW
,
115 &hammer_debug_locks
, 0, "");
116 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_btree
, CTLFLAG_RW
,
117 &hammer_debug_btree
, 0, "");
118 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_tid
, CTLFLAG_RW
,
119 &hammer_debug_tid
, 0, "");
120 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_recover
, CTLFLAG_RW
,
121 &hammer_debug_recover
, 0, "");
122 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, debug_recover_faults
, CTLFLAG_RW
,
123 &hammer_debug_recover_faults
, 0, "");
124 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, cluster_enable
, CTLFLAG_RW
,
125 &hammer_cluster_enable
, 0, "");
127 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, limit_dirtybufspace
, CTLFLAG_RW
,
128 &hammer_limit_dirtybufspace
, 0, "");
129 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, limit_recs
, CTLFLAG_RW
,
130 &hammer_limit_recs
, 0, "");
131 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, limit_iqueued
, CTLFLAG_RW
,
132 &hammer_limit_iqueued
, 0, "");
134 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_fsyncs
, CTLFLAG_RD
,
135 &hammer_count_fsyncs
, 0, "");
136 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_inodes
, CTLFLAG_RD
,
137 &hammer_count_inodes
, 0, "");
138 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_iqueued
, CTLFLAG_RD
,
139 &hammer_count_iqueued
, 0, "");
140 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_reclaiming
, CTLFLAG_RD
,
141 &hammer_count_reclaiming
, 0, "");
142 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_records
, CTLFLAG_RD
,
143 &hammer_count_records
, 0, "");
144 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_record_datas
, CTLFLAG_RD
,
145 &hammer_count_record_datas
, 0, "");
146 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_volumes
, CTLFLAG_RD
,
147 &hammer_count_volumes
, 0, "");
148 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_buffers
, CTLFLAG_RD
,
149 &hammer_count_buffers
, 0, "");
150 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_nodes
, CTLFLAG_RD
,
151 &hammer_count_nodes
, 0, "");
152 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, count_extra_space_used
, CTLFLAG_RD
,
153 &hammer_count_extra_space_used
, 0, "");
155 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_searches
, CTLFLAG_RD
,
156 &hammer_stats_btree_searches
, 0, "");
157 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_lookups
, CTLFLAG_RD
,
158 &hammer_stats_btree_lookups
, 0, "");
159 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_inserts
, CTLFLAG_RD
,
160 &hammer_stats_btree_inserts
, 0, "");
161 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_deletes
, CTLFLAG_RD
,
162 &hammer_stats_btree_deletes
, 0, "");
163 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_elements
, CTLFLAG_RD
,
164 &hammer_stats_btree_elements
, 0, "");
165 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_splits
, CTLFLAG_RD
,
166 &hammer_stats_btree_splits
, 0, "");
167 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_btree_iterations
, CTLFLAG_RD
,
168 &hammer_stats_btree_iterations
, 0, "");
169 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_record_iterations
, CTLFLAG_RD
,
170 &hammer_stats_record_iterations
, 0, "");
172 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_file_read
, CTLFLAG_RD
,
173 &hammer_stats_file_read
, 0, "");
174 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_file_write
, CTLFLAG_RD
,
175 &hammer_stats_file_write
, 0, "");
176 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_file_iopsr
, CTLFLAG_RD
,
177 &hammer_stats_file_iopsr
, 0, "");
178 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_file_iopsw
, CTLFLAG_RD
,
179 &hammer_stats_file_iopsw
, 0, "");
180 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_disk_read
, CTLFLAG_RD
,
181 &hammer_stats_disk_read
, 0, "");
182 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_disk_write
, CTLFLAG_RD
,
183 &hammer_stats_disk_write
, 0, "");
184 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_inode_flushes
, CTLFLAG_RD
,
185 &hammer_stats_inode_flushes
, 0, "");
186 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, stats_commits
, CTLFLAG_RD
,
187 &hammer_stats_commits
, 0, "");
189 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_dirtybufspace
, CTLFLAG_RD
,
190 &hammer_count_dirtybufspace
, 0, "");
191 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_refedbufs
, CTLFLAG_RD
,
192 &hammer_count_refedbufs
, 0, "");
193 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_reservations
, CTLFLAG_RD
,
194 &hammer_count_reservations
, 0, "");
195 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_io_running_read
, CTLFLAG_RD
,
196 &hammer_count_io_running_read
, 0, "");
197 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_io_locked
, CTLFLAG_RD
,
198 &hammer_count_io_locked
, 0, "");
199 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, count_io_running_write
, CTLFLAG_RD
,
200 &hammer_count_io_running_write
, 0, "");
201 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, zone_limit
, CTLFLAG_RW
,
202 &hammer_zone_limit
, 0, "");
203 SYSCTL_QUAD(_vfs_hammer
, OID_AUTO
, contention_count
, CTLFLAG_RW
,
204 &hammer_contention_count
, 0, "");
205 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, autoflush
, CTLFLAG_RW
,
206 &hammer_autoflush
, 0, "");
207 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, verify_zone
, CTLFLAG_RW
,
208 &hammer_verify_zone
, 0, "");
209 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, verify_data
, CTLFLAG_RW
,
210 &hammer_verify_data
, 0, "");
211 SYSCTL_INT(_vfs_hammer
, OID_AUTO
, write_mode
, CTLFLAG_RW
,
212 &hammer_write_mode
, 0, "");
214 KTR_INFO_MASTER(hammer
);
219 static void hammer_free_hmp(struct mount
*mp
);
221 static int hammer_vfs_mount(struct mount
*mp
, char *path
, caddr_t data
,
223 static int hammer_vfs_unmount(struct mount
*mp
, int mntflags
);
224 static int hammer_vfs_root(struct mount
*mp
, struct vnode
**vpp
);
225 static int hammer_vfs_statfs(struct mount
*mp
, struct statfs
*sbp
,
227 static int hammer_vfs_statvfs(struct mount
*mp
, struct statvfs
*sbp
,
229 static int hammer_vfs_sync(struct mount
*mp
, int waitfor
);
230 static int hammer_vfs_vget(struct mount
*mp
, ino_t ino
,
232 static int hammer_vfs_init(struct vfsconf
*conf
);
233 static int hammer_vfs_fhtovp(struct mount
*mp
, struct vnode
*rootvp
,
234 struct fid
*fhp
, struct vnode
**vpp
);
235 static int hammer_vfs_vptofh(struct vnode
*vp
, struct fid
*fhp
);
236 static int hammer_vfs_checkexp(struct mount
*mp
, struct sockaddr
*nam
,
237 int *exflagsp
, struct ucred
**credanonp
);
240 static struct vfsops hammer_vfsops
= {
241 .vfs_mount
= hammer_vfs_mount
,
242 .vfs_unmount
= hammer_vfs_unmount
,
243 .vfs_root
= hammer_vfs_root
,
244 .vfs_statfs
= hammer_vfs_statfs
,
245 .vfs_statvfs
= hammer_vfs_statvfs
,
246 .vfs_sync
= hammer_vfs_sync
,
247 .vfs_vget
= hammer_vfs_vget
,
248 .vfs_init
= hammer_vfs_init
,
249 .vfs_vptofh
= hammer_vfs_vptofh
,
250 .vfs_fhtovp
= hammer_vfs_fhtovp
,
251 .vfs_checkexp
= hammer_vfs_checkexp
254 MALLOC_DEFINE(M_HAMMER
, "hammer-general", "hammer general");
255 MALLOC_DEFINE(M_HAMMER_INO
, "hammer-inodes", "hammer inodes");
257 VFS_SET(hammer_vfsops
, hammer
, 0);
258 MODULE_VERSION(hammer
, 1);
261 hammer_vfs_init(struct vfsconf
*conf
)
265 if (hammer_limit_recs
== 0) {
266 hammer_limit_recs
= nbuf
* 25;
267 n
= kmalloc_limit(M_HAMMER
) / 512;
268 if (hammer_limit_recs
> n
)
269 hammer_limit_recs
= n
;
271 if (hammer_limit_dirtybufspace
== 0) {
272 hammer_limit_dirtybufspace
= hidirtybufspace
/ 2;
273 if (hammer_limit_dirtybufspace
< 100)
274 hammer_limit_dirtybufspace
= 100;
276 if (hammer_limit_iqueued
== 0)
277 hammer_limit_iqueued
= desiredvnodes
/ 5;
282 hammer_vfs_mount(struct mount
*mp
, char *mntpt
, caddr_t data
,
285 struct hammer_mount_info info
;
287 hammer_volume_t rootvol
;
288 struct vnode
*rootvp
;
289 struct vnode
*devvp
= NULL
;
290 const char *upath
; /* volume name in userspace */
291 char *path
; /* volume name in system space */
299 if ((error
= bdevvp(rootdev
, &devvp
))) {
300 kprintf("hammer_mountroot: can't find devvp\n");
303 mp
->mnt_flag
&= ~MNT_RDONLY
; /* mount R/W */
304 bzero(&info
, sizeof(info
));
309 if ((error
= copyin(data
, &info
, sizeof(info
))) != 0)
314 * updating or new mount
316 if (mp
->mnt_flag
& MNT_UPDATE
) {
317 hmp
= (void *)mp
->mnt_data
;
318 KKASSERT(hmp
!= NULL
);
320 if (info
.nvolumes
<= 0 || info
.nvolumes
>= 32768)
326 * master-id validation. The master id may not be changed by a
329 if (info
.hflags
& HMNT_MASTERID
) {
330 if (hmp
&& hmp
->master_id
!= info
.master_id
) {
331 kprintf("hammer: cannot change master id "
332 "with mount update\n");
335 master_id
= info
.master_id
;
336 if (master_id
< -1 || master_id
>= HAMMER_MAX_MASTERS
)
340 master_id
= hmp
->master_id
;
346 * Interal mount data structure
349 hmp
= kmalloc(sizeof(*hmp
), M_HAMMER
, M_WAITOK
| M_ZERO
);
350 mp
->mnt_data
= (qaddr_t
)hmp
;
352 hmp
->namekey_iterator
= mycpu
->gd_time_seconds
;
353 /*TAILQ_INIT(&hmp->recycle_list);*/
355 hmp
->root_btree_beg
.localization
= 0x00000000U
;
356 hmp
->root_btree_beg
.obj_id
= -0x8000000000000000LL
;
357 hmp
->root_btree_beg
.key
= -0x8000000000000000LL
;
358 hmp
->root_btree_beg
.create_tid
= 1;
359 hmp
->root_btree_beg
.delete_tid
= 1;
360 hmp
->root_btree_beg
.rec_type
= 0;
361 hmp
->root_btree_beg
.obj_type
= 0;
363 hmp
->root_btree_end
.localization
= 0xFFFFFFFFU
;
364 hmp
->root_btree_end
.obj_id
= 0x7FFFFFFFFFFFFFFFLL
;
365 hmp
->root_btree_end
.key
= 0x7FFFFFFFFFFFFFFFLL
;
366 hmp
->root_btree_end
.create_tid
= 0xFFFFFFFFFFFFFFFFULL
;
367 hmp
->root_btree_end
.delete_tid
= 0; /* special case */
368 hmp
->root_btree_end
.rec_type
= 0xFFFFU
;
369 hmp
->root_btree_end
.obj_type
= 0;
371 hmp
->krate
.freq
= 1; /* maximum reporting rate (hz) */
372 hmp
->krate
.count
= -16; /* initial burst */
374 hmp
->sync_lock
.refs
= 1;
375 hmp
->free_lock
.refs
= 1;
376 hmp
->undo_lock
.refs
= 1;
377 hmp
->blkmap_lock
.refs
= 1;
379 TAILQ_INIT(&hmp
->delay_list
);
380 TAILQ_INIT(&hmp
->flush_group_list
);
381 TAILQ_INIT(&hmp
->objid_cache_list
);
382 TAILQ_INIT(&hmp
->undo_lru_list
);
383 TAILQ_INIT(&hmp
->reclaim_list
);
385 hmp
->hflags
&= ~HMNT_USERFLAGS
;
386 hmp
->hflags
|= info
.hflags
& HMNT_USERFLAGS
;
388 hmp
->master_id
= master_id
;
391 mp
->mnt_flag
|= MNT_RDONLY
;
392 hmp
->asof
= info
.asof
;
394 hmp
->asof
= HAMMER_MAX_TID
;
398 * Re-open read-write if originally read-only, or vise-versa.
400 if (mp
->mnt_flag
& MNT_UPDATE
) {
402 if (hmp
->ronly
&& (mp
->mnt_kern_flag
& MNTK_WANTRDWR
)) {
403 kprintf("HAMMER read-only -> read-write\n");
405 RB_SCAN(hammer_vol_rb_tree
, &hmp
->rb_vols_root
, NULL
,
406 hammer_adjust_volume_mode
, NULL
);
407 rootvol
= hammer_get_root_volume(hmp
, &error
);
409 hammer_recover_flush_buffers(hmp
, rootvol
, 1);
410 bcopy(rootvol
->ondisk
->vol0_blockmap
,
412 sizeof(hmp
->blockmap
));
413 hammer_rel_volume(rootvol
, 0);
415 RB_SCAN(hammer_ino_rb_tree
, &hmp
->rb_inos_root
, NULL
,
416 hammer_reload_inode
, NULL
);
417 /* kernel clears MNT_RDONLY */
418 } else if (hmp
->ronly
== 0 && (mp
->mnt_flag
& MNT_RDONLY
)) {
419 kprintf("HAMMER read-write -> read-only\n");
420 hmp
->ronly
= 1; /* messy */
421 RB_SCAN(hammer_ino_rb_tree
, &hmp
->rb_inos_root
, NULL
,
422 hammer_reload_inode
, NULL
);
424 hammer_flusher_sync(hmp
);
425 hammer_flusher_sync(hmp
);
426 hammer_flusher_sync(hmp
);
428 RB_SCAN(hammer_vol_rb_tree
, &hmp
->rb_vols_root
, NULL
,
429 hammer_adjust_volume_mode
, NULL
);
434 RB_INIT(&hmp
->rb_vols_root
);
435 RB_INIT(&hmp
->rb_inos_root
);
436 RB_INIT(&hmp
->rb_nods_root
);
437 RB_INIT(&hmp
->rb_undo_root
);
438 RB_INIT(&hmp
->rb_resv_root
);
439 RB_INIT(&hmp
->rb_bufs_root
);
440 RB_INIT(&hmp
->rb_pfsm_root
);
442 hmp
->ronly
= ((mp
->mnt_flag
& MNT_RDONLY
) != 0);
444 TAILQ_INIT(&hmp
->volu_list
);
445 TAILQ_INIT(&hmp
->undo_list
);
446 TAILQ_INIT(&hmp
->data_list
);
447 TAILQ_INIT(&hmp
->meta_list
);
448 TAILQ_INIT(&hmp
->lose_list
);
453 path
= objcache_get(namei_oc
, M_WAITOK
);
455 for (i
= 0; i
< info
.nvolumes
; ++i
) {
459 * Only one volume; and no need for copyin.
461 KKASSERT(info
.nvolumes
== 1);
462 ksnprintf(path
, MAXPATHLEN
, "/dev/%s",
463 mp
->mnt_stat
.f_mntfromname
);
466 error
= copyin(&info
.volumes
[i
], &upath
,
469 error
= copyinstr(upath
, path
,
473 error
= hammer_install_volume(hmp
, path
, devvp
);
477 objcache_put(namei_oc
, path
);
480 * Make sure we found a root volume
482 if (error
== 0 && hmp
->rootvol
== NULL
) {
483 kprintf("hammer_mount: No root volume found!\n");
488 * Check that all required volumes are available
490 if (error
== 0 && hammer_mountcheck_volumes(hmp
)) {
491 kprintf("hammer_mount: Missing volumes, cannot mount!\n");
501 * No errors, setup enough of the mount point so we can lookup the
504 mp
->mnt_iosize_max
= MAXPHYS
;
505 mp
->mnt_kern_flag
|= MNTK_FSMID
;
508 * note: f_iosize is used by vnode_pager_haspage() when constructing
511 mp
->mnt_stat
.f_iosize
= HAMMER_BUFSIZE
;
512 mp
->mnt_stat
.f_bsize
= HAMMER_BUFSIZE
;
514 mp
->mnt_vstat
.f_frsize
= HAMMER_BUFSIZE
;
515 mp
->mnt_vstat
.f_bsize
= HAMMER_BUFSIZE
;
517 mp
->mnt_maxsymlinklen
= 255;
518 mp
->mnt_flag
|= MNT_LOCAL
;
520 vfs_add_vnodeops(mp
, &hammer_vnode_vops
, &mp
->mnt_vn_norm_ops
);
521 vfs_add_vnodeops(mp
, &hammer_spec_vops
, &mp
->mnt_vn_spec_ops
);
522 vfs_add_vnodeops(mp
, &hammer_fifo_vops
, &mp
->mnt_vn_fifo_ops
);
525 * The root volume's ondisk pointer is only valid if we hold a
528 rootvol
= hammer_get_root_volume(hmp
, &error
);
533 * Perform any necessary UNDO operations. The recovery code does
534 * call hammer_undo_lookup() so we have to pre-cache the blockmap,
535 * and then re-copy it again after recovery is complete.
537 * If this is a read-only mount the UNDO information is retained
538 * in memory in the form of dirty buffer cache buffers, and not
539 * written back to the media.
541 bcopy(rootvol
->ondisk
->vol0_blockmap
, hmp
->blockmap
,
542 sizeof(hmp
->blockmap
));
545 * The undo_rec_limit limits the size of flush groups to avoid
546 * blowing out the UNDO FIFO. This calculation is typically in
547 * the tens of thousands and is designed primarily when small
548 * HAMMER filesystems are created.
550 hmp
->undo_rec_limit
= hammer_undo_max(hmp
) / 8192 + 100;
551 if (hammer_debug_general
& 0x0001)
552 kprintf("HAMMER: undo_rec_limit %d\n", hmp
->undo_rec_limit
);
554 error
= hammer_recover(hmp
, rootvol
);
556 kprintf("Failed to recover HAMMER filesystem on mount\n");
561 * Finish setup now that we have a good root volume.
563 * The top 16 bits of fsid.val[1] is a pfs id.
565 ksnprintf(mp
->mnt_stat
.f_mntfromname
,
566 sizeof(mp
->mnt_stat
.f_mntfromname
), "%s",
567 rootvol
->ondisk
->vol_name
);
568 mp
->mnt_stat
.f_fsid
.val
[0] =
569 crc32((char *)&rootvol
->ondisk
->vol_fsid
+ 0, 8);
570 mp
->mnt_stat
.f_fsid
.val
[1] =
571 crc32((char *)&rootvol
->ondisk
->vol_fsid
+ 8, 8);
572 mp
->mnt_stat
.f_fsid
.val
[1] &= 0x0000FFFF;
574 mp
->mnt_vstat
.f_fsid_uuid
= rootvol
->ondisk
->vol_fsid
;
575 mp
->mnt_vstat
.f_fsid
= crc32(&mp
->mnt_vstat
.f_fsid_uuid
,
576 sizeof(mp
->mnt_vstat
.f_fsid_uuid
));
579 * Certain often-modified fields in the root volume are cached in
580 * the hammer_mount structure so we do not have to generate lots
581 * of little UNDO structures for them.
583 * Recopy after recovery. This also has the side effect of
584 * setting our cached undo FIFO's first_offset, which serves to
585 * placemark the FIFO start for the NEXT flush cycle while the
586 * on-disk first_offset represents the LAST flush cycle.
588 hmp
->next_tid
= rootvol
->ondisk
->vol0_next_tid
;
589 hmp
->flush_tid1
= hmp
->next_tid
;
590 hmp
->flush_tid2
= hmp
->next_tid
;
591 bcopy(rootvol
->ondisk
->vol0_blockmap
, hmp
->blockmap
,
592 sizeof(hmp
->blockmap
));
593 hmp
->copy_stat_freebigblocks
= rootvol
->ondisk
->vol0_stat_freebigblocks
;
595 hammer_flusher_create(hmp
);
598 * Locate the root directory using the root cluster's B-Tree as a
599 * starting point. The root directory uses an obj_id of 1.
601 * FUTURE: Leave the root directory cached referenced but unlocked
602 * in hmp->rootvp (need to flush it on unmount).
604 error
= hammer_vfs_vget(mp
, 1, &rootvp
);
608 /*vn_unlock(hmp->rootvp);*/
611 hammer_rel_volume(rootvol
, 0);
614 * Cleanup and return.
622 hammer_vfs_unmount(struct mount
*mp
, int mntflags
)
625 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
631 * Clean out the vnodes
634 if (mntflags
& MNT_FORCE
)
636 if ((error
= vflush(mp
, 0, flags
)) != 0)
640 * Clean up the internal mount structure and related entities. This
648 * Clean up the internal mount structure and disassociate it from the mount.
649 * This may issue I/O.
652 hammer_free_hmp(struct mount
*mp
)
654 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
655 hammer_flush_group_t flg
;
659 * Flush anything dirty. This won't even run if the
660 * filesystem errored-out.
663 while (hammer_flusher_haswork(hmp
)) {
664 hammer_flusher_sync(hmp
);
668 kprintf("HAMMER: umount flushing.");
671 tsleep(hmp
, 0, "hmrufl", hz
);
674 kprintf("giving up\n");
678 if (count
>= 5 && count
< 30)
682 * If the mount had a critical error we have to destroy any
683 * remaining inodes before we can finish cleaning up the flusher.
685 if (hmp
->flags
& HAMMER_MOUNT_CRITICAL_ERROR
) {
686 RB_SCAN(hammer_ino_rb_tree
, &hmp
->rb_inos_root
, NULL
,
687 hammer_destroy_inode_callback
, NULL
);
691 * There shouldn't be any inodes left now and any left over
692 * flush groups should now be empty.
694 KKASSERT(RB_EMPTY(&hmp
->rb_inos_root
));
695 while ((flg
= TAILQ_FIRST(&hmp
->flush_group_list
)) != NULL
) {
696 TAILQ_REMOVE(&hmp
->flush_group_list
, flg
, flush_entry
);
697 KKASSERT(TAILQ_EMPTY(&flg
->flush_list
));
699 kprintf("HAMMER: Warning, flush_group %p was "
700 "not empty on umount!\n", flg
);
702 kfree(flg
, M_HAMMER
);
706 * We can finally destroy the flusher
708 hammer_flusher_destroy(hmp
);
711 * We may have held recovered buffers due to a read-only mount.
712 * These must be discarded.
715 hammer_recover_flush_buffers(hmp
, NULL
, -1);
718 * Unload buffers and then volumes
720 RB_SCAN(hammer_buf_rb_tree
, &hmp
->rb_bufs_root
, NULL
,
721 hammer_unload_buffer
, NULL
);
722 RB_SCAN(hammer_vol_rb_tree
, &hmp
->rb_vols_root
, NULL
,
723 hammer_unload_volume
, NULL
);
726 mp
->mnt_flag
&= ~MNT_LOCAL
;
728 hammer_destroy_objid_cache(hmp
);
729 kfree(hmp
, M_HAMMER
);
733 * Report critical errors. ip may be NULL.
736 hammer_critical_error(hammer_mount_t hmp
, hammer_inode_t ip
,
737 int error
, const char *msg
)
739 hmp
->flags
|= HAMMER_MOUNT_CRITICAL_ERROR
;
740 krateprintf(&hmp
->krate
,
741 "HAMMER(%s): Critical error inode=%lld %s\n",
742 hmp
->mp
->mnt_stat
.f_mntfromname
,
743 (ip
? ip
->obj_id
: -1), msg
);
744 if (hmp
->ronly
== 0) {
745 hmp
->ronly
= 2; /* special errored read-only mode */
746 hmp
->mp
->mnt_flag
|= MNT_RDONLY
;
747 kprintf("HAMMER(%s): Forcing read-only mode\n",
748 hmp
->mp
->mnt_stat
.f_mntfromname
);
755 * Obtain a vnode for the specified inode number. An exclusively locked
759 hammer_vfs_vget(struct mount
*mp
, ino_t ino
, struct vnode
**vpp
)
761 struct hammer_transaction trans
;
762 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
763 struct hammer_inode
*ip
;
766 hammer_simple_transaction(&trans
, hmp
);
769 * Lookup the requested HAMMER inode. The structure must be
770 * left unlocked while we manipulate the related vnode to avoid
773 ip
= hammer_get_inode(&trans
, NULL
, ino
,
774 hmp
->asof
, HAMMER_DEF_LOCALIZATION
,
778 hammer_done_transaction(&trans
);
781 error
= hammer_get_vnode(ip
, vpp
);
782 hammer_rel_inode(ip
, 0);
783 hammer_done_transaction(&trans
);
788 * Return the root vnode for the filesystem.
790 * HAMMER stores the root vnode in the hammer_mount structure so
791 * getting it is easy.
794 hammer_vfs_root(struct mount
*mp
, struct vnode
**vpp
)
797 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
801 error
= hammer_vfs_vget(mp
, 1, vpp
);
806 hammer_vfs_statfs(struct mount
*mp
, struct statfs
*sbp
, struct ucred
*cred
)
808 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
809 hammer_volume_t volume
;
810 hammer_volume_ondisk_t ondisk
;
814 volume
= hammer_get_root_volume(hmp
, &error
);
817 ondisk
= volume
->ondisk
;
822 mp
->mnt_stat
.f_files
= ondisk
->vol0_stat_inodes
;
823 bfree
= ondisk
->vol0_stat_freebigblocks
* HAMMER_LARGEBLOCK_SIZE
;
824 hammer_rel_volume(volume
, 0);
826 mp
->mnt_stat
.f_bfree
= bfree
/ HAMMER_BUFSIZE
;
827 mp
->mnt_stat
.f_bavail
= mp
->mnt_stat
.f_bfree
;
828 if (mp
->mnt_stat
.f_files
< 0)
829 mp
->mnt_stat
.f_files
= 0;
836 hammer_vfs_statvfs(struct mount
*mp
, struct statvfs
*sbp
, struct ucred
*cred
)
838 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
839 hammer_volume_t volume
;
840 hammer_volume_ondisk_t ondisk
;
844 volume
= hammer_get_root_volume(hmp
, &error
);
847 ondisk
= volume
->ondisk
;
852 mp
->mnt_vstat
.f_files
= ondisk
->vol0_stat_inodes
;
853 bfree
= ondisk
->vol0_stat_freebigblocks
* HAMMER_LARGEBLOCK_SIZE
;
854 hammer_rel_volume(volume
, 0);
856 mp
->mnt_vstat
.f_bfree
= bfree
/ HAMMER_BUFSIZE
;
857 mp
->mnt_vstat
.f_bavail
= mp
->mnt_stat
.f_bfree
;
858 if (mp
->mnt_vstat
.f_files
< 0)
859 mp
->mnt_vstat
.f_files
= 0;
860 *sbp
= mp
->mnt_vstat
;
865 * Sync the filesystem. Currently we have to run it twice, the second
866 * one will advance the undo start index to the end index, so if a crash
867 * occurs no undos will be run on mount.
869 * We do not sync the filesystem if we are called from a panic. If we did
870 * we might end up blowing up a sync that was already in progress.
873 hammer_vfs_sync(struct mount
*mp
, int waitfor
)
875 struct hammer_mount
*hmp
= (void *)mp
->mnt_data
;
878 if (panicstr
== NULL
) {
879 error
= hammer_sync_hmp(hmp
, waitfor
);
887 * Convert a vnode to a file handle.
890 hammer_vfs_vptofh(struct vnode
*vp
, struct fid
*fhp
)
894 KKASSERT(MAXFIDSZ
>= 16);
896 fhp
->fid_len
= offsetof(struct fid
, fid_data
[16]);
897 fhp
->fid_ext
= ip
->obj_localization
>> 16;
898 bcopy(&ip
->obj_id
, fhp
->fid_data
+ 0, sizeof(ip
->obj_id
));
899 bcopy(&ip
->obj_asof
, fhp
->fid_data
+ 8, sizeof(ip
->obj_asof
));
905 * Convert a file handle back to a vnode.
907 * Use rootvp to enforce PFS isolation when a PFS is exported via a
911 hammer_vfs_fhtovp(struct mount
*mp
, struct vnode
*rootvp
,
912 struct fid
*fhp
, struct vnode
**vpp
)
914 struct hammer_transaction trans
;
915 struct hammer_inode
*ip
;
916 struct hammer_inode_info info
;
918 u_int32_t localization
;
920 bcopy(fhp
->fid_data
+ 0, &info
.obj_id
, sizeof(info
.obj_id
));
921 bcopy(fhp
->fid_data
+ 8, &info
.obj_asof
, sizeof(info
.obj_asof
));
923 localization
= VTOI(rootvp
)->obj_localization
;
925 localization
= (u_int32_t
)fhp
->fid_ext
<< 16;
927 hammer_simple_transaction(&trans
, (void *)mp
->mnt_data
);
930 * Get/allocate the hammer_inode structure. The structure must be
931 * unlocked while we manipulate the related vnode to avoid a
934 ip
= hammer_get_inode(&trans
, NULL
, info
.obj_id
,
935 info
.obj_asof
, localization
, 0, &error
);
940 error
= hammer_get_vnode(ip
, vpp
);
941 hammer_rel_inode(ip
, 0);
942 hammer_done_transaction(&trans
);
947 hammer_vfs_checkexp(struct mount
*mp
, struct sockaddr
*nam
,
948 int *exflagsp
, struct ucred
**credanonp
)
950 hammer_mount_t hmp
= (void *)mp
->mnt_data
;
954 np
= vfs_export_lookup(mp
, &hmp
->export
, nam
);
956 *exflagsp
= np
->netc_exflags
;
957 *credanonp
= &np
->netc_anon
;
967 hammer_vfs_export(struct mount
*mp
, int op
, const struct export_args
*export
)
969 hammer_mount_t hmp
= (void *)mp
->mnt_data
;
973 case MOUNTCTL_SET_EXPORT
:
974 error
= vfs_export(mp
, &hmp
->export
, export
);