HAMMER VFS - The backend flusher now sorts inodes
[dragonfly.git] / sys / vfs / hammer / hammer_vfsops.c
CommitLineData
427e5fc6 1/*
b84de5af 2 * Copyright (c) 2007-2008 The DragonFly Project. All rights reserved.
427e5fc6
MD
3 *
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 *
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
16 * distribution.
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.
20 *
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
32 * SUCH DAMAGE.
33 *
44a83111 34 * $DragonFly: src/sys/vfs/hammer/hammer_vfsops.c,v 1.74 2008/11/13 02:18:43 dillon Exp $
427e5fc6
MD
35 */
36
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>
b3deaf57 45#include <sys/sysctl.h>
427e5fc6
MD
46#include <sys/buf.h>
47#include <sys/buf2.h>
48#include "hammer.h"
49
02428fb6 50int hammer_supported_version = HAMMER_VOL_VERSION_DEFAULT;
2f85fa4d 51int hammer_debug_io;
d5ef456e 52int hammer_debug_general;
1b0ab2c3 53int hammer_debug_debug = 1; /* medium-error panics */
e8599db1 54int hammer_debug_inode;
7d683b0f 55int hammer_debug_locks;
b3deaf57 56int hammer_debug_btree;
d113fda1 57int hammer_debug_tid;
cebe9493 58int hammer_debug_recover; /* -1 will disable, +1 will force */
46fe7ae1 59int hammer_debug_recover_faults;
c9ce54d6 60int hammer_error_panic; /* panic on error levels */
1b0ab2c3 61int hammer_cluster_enable = 1; /* enable read clustering by default */
7a61b85d 62int hammer_count_fsyncs;
b3deaf57 63int hammer_count_inodes;
af209b0f 64int hammer_count_iqueued;
9f5097dc 65int hammer_count_reclaiming;
b3deaf57
MD
66int hammer_count_records;
67int hammer_count_record_datas;
68int hammer_count_volumes;
b3deaf57
MD
69int hammer_count_buffers;
70int hammer_count_nodes;
a7e9bef1 71int64_t hammer_count_extra_space_used;
cb51be26
MD
72int64_t hammer_stats_btree_lookups;
73int64_t hammer_stats_btree_searches;
74int64_t hammer_stats_btree_inserts;
75int64_t hammer_stats_btree_deletes;
76int64_t hammer_stats_btree_elements;
77int64_t hammer_stats_btree_splits;
78int64_t hammer_stats_btree_iterations;
39d8fd63 79int64_t hammer_stats_btree_root_iterations;
cb51be26 80int64_t hammer_stats_record_iterations;
ce0138a6
MD
81
82int64_t hammer_stats_file_read;
83int64_t hammer_stats_file_write;
84int64_t hammer_stats_file_iopsr;
85int64_t hammer_stats_file_iopsw;
86int64_t hammer_stats_disk_read;
87int64_t hammer_stats_disk_write;
88int64_t hammer_stats_inode_flushes;
89int64_t hammer_stats_commits;
89e744ce 90int64_t hammer_stats_undo;
ce0138a6 91
f5a07a7a 92int hammer_count_dirtybufspace; /* global */
a99b9ea2 93int hammer_count_refedbufs; /* global */
0832c9bb 94int hammer_count_reservations;
a99b9ea2
MD
95int hammer_count_io_running_read;
96int hammer_count_io_running_write;
97int hammer_count_io_locked;
f5a07a7a 98int hammer_limit_dirtybufspace; /* per-mount */
0832c9bb 99int hammer_limit_recs; /* as a whole XXX */
de996e86 100int hammer_limit_inode_recs = 1024; /* per inode */
ff003b11 101int hammer_limit_reclaim = HAMMER_RECLAIM_WAIT;
21fde338 102int hammer_autoflush = 2000; /* auto flush */
1f07f686 103int hammer_bio_count;
cb51be26 104int hammer_verify_zone;
1b0ab2c3 105int hammer_verify_data = 1;
cb51be26 106int hammer_write_mode;
3e583440 107int hammer_yield_check = 16;
6f3d87c0 108int hammer_fsync_mode;
7d683b0f 109int64_t hammer_contention_count;
f03c9cf4 110int64_t hammer_zone_limit;
b3deaf57
MD
111
112SYSCTL_NODE(_vfs, OID_AUTO, hammer, CTLFLAG_RW, 0, "HAMMER filesystem");
5987cc42
MD
113SYSCTL_INT(_vfs_hammer, OID_AUTO, supported_version, CTLFLAG_RD,
114 &hammer_supported_version, 0, "");
d5ef456e
MD
115SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_general, CTLFLAG_RW,
116 &hammer_debug_general, 0, "");
2f85fa4d
MD
117SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_io, CTLFLAG_RW,
118 &hammer_debug_io, 0, "");
77062c8a
MD
119SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_debug, CTLFLAG_RW,
120 &hammer_debug_debug, 0, "");
e8599db1
MD
121SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_inode, CTLFLAG_RW,
122 &hammer_debug_inode, 0, "");
7d683b0f
MD
123SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_locks, CTLFLAG_RW,
124 &hammer_debug_locks, 0, "");
b3deaf57
MD
125SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_btree, CTLFLAG_RW,
126 &hammer_debug_btree, 0, "");
d113fda1
MD
127SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_tid, CTLFLAG_RW,
128 &hammer_debug_tid, 0, "");
b33e2cc0
MD
129SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_recover, CTLFLAG_RW,
130 &hammer_debug_recover, 0, "");
46fe7ae1
MD
131SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_recover_faults, CTLFLAG_RW,
132 &hammer_debug_recover_faults, 0, "");
c9ce54d6
MD
133SYSCTL_INT(_vfs_hammer, OID_AUTO, error_panic, CTLFLAG_RW,
134 &hammer_error_panic, 0, "");
1b0ab2c3
MD
135SYSCTL_INT(_vfs_hammer, OID_AUTO, cluster_enable, CTLFLAG_RW,
136 &hammer_cluster_enable, 0, "");
9480ff55 137
f5a07a7a
MD
138SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_dirtybufspace, CTLFLAG_RW,
139 &hammer_limit_dirtybufspace, 0, "");
47637bff
MD
140SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_recs, CTLFLAG_RW,
141 &hammer_limit_recs, 0, "");
de996e86
MD
142SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_inode_recs, CTLFLAG_RW,
143 &hammer_limit_inode_recs, 0, "");
ff003b11
MD
144SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_reclaim, CTLFLAG_RW,
145 &hammer_limit_reclaim, 0, "");
9480ff55 146
7a61b85d
MD
147SYSCTL_INT(_vfs_hammer, OID_AUTO, count_fsyncs, CTLFLAG_RD,
148 &hammer_count_fsyncs, 0, "");
b3deaf57
MD
149SYSCTL_INT(_vfs_hammer, OID_AUTO, count_inodes, CTLFLAG_RD,
150 &hammer_count_inodes, 0, "");
af209b0f
MD
151SYSCTL_INT(_vfs_hammer, OID_AUTO, count_iqueued, CTLFLAG_RD,
152 &hammer_count_iqueued, 0, "");
9f5097dc
MD
153SYSCTL_INT(_vfs_hammer, OID_AUTO, count_reclaiming, CTLFLAG_RD,
154 &hammer_count_reclaiming, 0, "");
b3deaf57
MD
155SYSCTL_INT(_vfs_hammer, OID_AUTO, count_records, CTLFLAG_RD,
156 &hammer_count_records, 0, "");
157SYSCTL_INT(_vfs_hammer, OID_AUTO, count_record_datas, CTLFLAG_RD,
158 &hammer_count_record_datas, 0, "");
159SYSCTL_INT(_vfs_hammer, OID_AUTO, count_volumes, CTLFLAG_RD,
160 &hammer_count_volumes, 0, "");
b3deaf57
MD
161SYSCTL_INT(_vfs_hammer, OID_AUTO, count_buffers, CTLFLAG_RD,
162 &hammer_count_buffers, 0, "");
163SYSCTL_INT(_vfs_hammer, OID_AUTO, count_nodes, CTLFLAG_RD,
164 &hammer_count_nodes, 0, "");
a7e9bef1
MD
165SYSCTL_QUAD(_vfs_hammer, OID_AUTO, count_extra_space_used, CTLFLAG_RD,
166 &hammer_count_extra_space_used, 0, "");
ce0138a6 167
cb51be26
MD
168SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_searches, CTLFLAG_RD,
169 &hammer_stats_btree_searches, 0, "");
170SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_lookups, CTLFLAG_RD,
171 &hammer_stats_btree_lookups, 0, "");
172SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_inserts, CTLFLAG_RD,
173 &hammer_stats_btree_inserts, 0, "");
174SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_deletes, CTLFLAG_RD,
175 &hammer_stats_btree_deletes, 0, "");
176SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_elements, CTLFLAG_RD,
177 &hammer_stats_btree_elements, 0, "");
178SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_splits, CTLFLAG_RD,
179 &hammer_stats_btree_splits, 0, "");
180SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_iterations, CTLFLAG_RD,
181 &hammer_stats_btree_iterations, 0, "");
39d8fd63
MD
182SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_root_iterations, CTLFLAG_RD,
183 &hammer_stats_btree_root_iterations, 0, "");
cb51be26
MD
184SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_record_iterations, CTLFLAG_RD,
185 &hammer_stats_record_iterations, 0, "");
ce0138a6
MD
186
187SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_read, CTLFLAG_RD,
188 &hammer_stats_file_read, 0, "");
189SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_write, CTLFLAG_RD,
190 &hammer_stats_file_write, 0, "");
191SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_iopsr, CTLFLAG_RD,
192 &hammer_stats_file_iopsr, 0, "");
193SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_iopsw, CTLFLAG_RD,
194 &hammer_stats_file_iopsw, 0, "");
195SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_disk_read, CTLFLAG_RD,
196 &hammer_stats_disk_read, 0, "");
197SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_disk_write, CTLFLAG_RD,
198 &hammer_stats_disk_write, 0, "");
199SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_inode_flushes, CTLFLAG_RD,
200 &hammer_stats_inode_flushes, 0, "");
201SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_commits, CTLFLAG_RD,
202 &hammer_stats_commits, 0, "");
89e744ce
MD
203SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_undo, CTLFLAG_RD,
204 &hammer_stats_undo, 0, "");
ce0138a6 205
f5a07a7a
MD
206SYSCTL_INT(_vfs_hammer, OID_AUTO, count_dirtybufspace, CTLFLAG_RD,
207 &hammer_count_dirtybufspace, 0, "");
a99b9ea2
MD
208SYSCTL_INT(_vfs_hammer, OID_AUTO, count_refedbufs, CTLFLAG_RD,
209 &hammer_count_refedbufs, 0, "");
0832c9bb
MD
210SYSCTL_INT(_vfs_hammer, OID_AUTO, count_reservations, CTLFLAG_RD,
211 &hammer_count_reservations, 0, "");
a99b9ea2
MD
212SYSCTL_INT(_vfs_hammer, OID_AUTO, count_io_running_read, CTLFLAG_RD,
213 &hammer_count_io_running_read, 0, "");
214SYSCTL_INT(_vfs_hammer, OID_AUTO, count_io_locked, CTLFLAG_RD,
215 &hammer_count_io_locked, 0, "");
216SYSCTL_INT(_vfs_hammer, OID_AUTO, count_io_running_write, CTLFLAG_RD,
217 &hammer_count_io_running_write, 0, "");
f03c9cf4
MD
218SYSCTL_QUAD(_vfs_hammer, OID_AUTO, zone_limit, CTLFLAG_RW,
219 &hammer_zone_limit, 0, "");
7d683b0f
MD
220SYSCTL_QUAD(_vfs_hammer, OID_AUTO, contention_count, CTLFLAG_RW,
221 &hammer_contention_count, 0, "");
21fde338
MD
222SYSCTL_INT(_vfs_hammer, OID_AUTO, autoflush, CTLFLAG_RW,
223 &hammer_autoflush, 0, "");
cb51be26
MD
224SYSCTL_INT(_vfs_hammer, OID_AUTO, verify_zone, CTLFLAG_RW,
225 &hammer_verify_zone, 0, "");
1b0ab2c3
MD
226SYSCTL_INT(_vfs_hammer, OID_AUTO, verify_data, CTLFLAG_RW,
227 &hammer_verify_data, 0, "");
cb51be26
MD
228SYSCTL_INT(_vfs_hammer, OID_AUTO, write_mode, CTLFLAG_RW,
229 &hammer_write_mode, 0, "");
3e583440
MD
230SYSCTL_INT(_vfs_hammer, OID_AUTO, yield_check, CTLFLAG_RW,
231 &hammer_yield_check, 0, "");
6f3d87c0
MD
232SYSCTL_INT(_vfs_hammer, OID_AUTO, fsync_mode, CTLFLAG_RW,
233 &hammer_fsync_mode, 0, "");
b3deaf57 234
bcac4bbb
MD
235KTR_INFO_MASTER(hammer);
236
427e5fc6
MD
237/*
238 * VFS ABI
239 */
240static void hammer_free_hmp(struct mount *mp);
241
242static int hammer_vfs_mount(struct mount *mp, char *path, caddr_t data,
243 struct ucred *cred);
244static int hammer_vfs_unmount(struct mount *mp, int mntflags);
245static int hammer_vfs_root(struct mount *mp, struct vnode **vpp);
246static int hammer_vfs_statfs(struct mount *mp, struct statfs *sbp,
247 struct ucred *cred);
6f97fce3
MD
248static int hammer_vfs_statvfs(struct mount *mp, struct statvfs *sbp,
249 struct ucred *cred);
427e5fc6 250static int hammer_vfs_sync(struct mount *mp, int waitfor);
b9b0a6d0
MD
251static int hammer_vfs_vget(struct mount *mp, struct vnode *dvp,
252 ino_t ino, struct vnode **vpp);
427e5fc6 253static int hammer_vfs_init(struct vfsconf *conf);
67863d04
MD
254static int hammer_vfs_fhtovp(struct mount *mp, struct vnode *rootvp,
255 struct fid *fhp, struct vnode **vpp);
513ca7d7
MD
256static int hammer_vfs_vptofh(struct vnode *vp, struct fid *fhp);
257static int hammer_vfs_checkexp(struct mount *mp, struct sockaddr *nam,
258 int *exflagsp, struct ucred **credanonp);
259
427e5fc6
MD
260
261static struct vfsops hammer_vfsops = {
262 .vfs_mount = hammer_vfs_mount,
263 .vfs_unmount = hammer_vfs_unmount,
264 .vfs_root = hammer_vfs_root,
265 .vfs_statfs = hammer_vfs_statfs,
6f97fce3 266 .vfs_statvfs = hammer_vfs_statvfs,
427e5fc6
MD
267 .vfs_sync = hammer_vfs_sync,
268 .vfs_vget = hammer_vfs_vget,
513ca7d7
MD
269 .vfs_init = hammer_vfs_init,
270 .vfs_vptofh = hammer_vfs_vptofh,
271 .vfs_fhtovp = hammer_vfs_fhtovp,
272 .vfs_checkexp = hammer_vfs_checkexp
427e5fc6
MD
273};
274
bac808fe 275MALLOC_DEFINE(M_HAMMER, "HAMMER-mount", "");
427e5fc6
MD
276
277VFS_SET(hammer_vfsops, hammer, 0);
278MODULE_VERSION(hammer, 1);
279
280static int
281hammer_vfs_init(struct vfsconf *conf)
282{
3098dc2f
MD
283 int n;
284
285 if (hammer_limit_recs == 0) {
a99b9ea2 286 hammer_limit_recs = nbuf * 25;
3098dc2f
MD
287 n = kmalloc_limit(M_HAMMER) / 512;
288 if (hammer_limit_recs > n)
289 hammer_limit_recs = n;
290 }
f5a07a7a
MD
291 if (hammer_limit_dirtybufspace == 0) {
292 hammer_limit_dirtybufspace = hidirtybufspace / 2;
293 if (hammer_limit_dirtybufspace < 100)
294 hammer_limit_dirtybufspace = 100;
9f5097dc 295 }
427e5fc6
MD
296 return(0);
297}
298
299static int
300hammer_vfs_mount(struct mount *mp, char *mntpt, caddr_t data,
301 struct ucred *cred)
302{
303 struct hammer_mount_info info;
a89aec1b
MD
304 hammer_mount_t hmp;
305 hammer_volume_t rootvol;
27ea2398 306 struct vnode *rootvp;
7c19b529 307 struct vnode *devvp = NULL;
427e5fc6
MD
308 const char *upath; /* volume name in userspace */
309 char *path; /* volume name in system space */
310 int error;
311 int i;
732a1697 312 int master_id;
bac808fe 313 int maxinodes;
ab0bf4a3
MD
314
315 /*
316 * Accept hammer_mount_info. mntpt is NULL for root mounts at boot.
317 */
7c19b529 318 if (mntpt == NULL) {
7c19b529
MN
319 if ((error = bdevvp(rootdev, &devvp))) {
320 kprintf("hammer_mountroot: can't find devvp\n");
321 return (error);
322 }
323 mp->mnt_flag &= ~MNT_RDONLY; /* mount R/W */
324 bzero(&info, sizeof(info));
325 info.asof = 0;
326 info.hflags = 0;
327 info.nvolumes = 1;
328 } else {
329 if ((error = copyin(data, &info, sizeof(info))) != 0)
330 return (error);
331 }
732a1697
MD
332
333 /*
334 * updating or new mount
335 */
336 if (mp->mnt_flag & MNT_UPDATE) {
337 hmp = (void *)mp->mnt_data;
338 KKASSERT(hmp != NULL);
339 } else {
51c35492
MD
340 if (info.nvolumes <= 0 || info.nvolumes >= 32768)
341 return (EINVAL);
732a1697 342 hmp = NULL;
51c35492 343 }
732a1697
MD
344
345 /*
346 * master-id validation. The master id may not be changed by a
347 * mount update.
348 */
349 if (info.hflags & HMNT_MASTERID) {
350 if (hmp && hmp->master_id != info.master_id) {
351 kprintf("hammer: cannot change master id "
352 "with mount update\n");
353 return(EINVAL);
354 }
355 master_id = info.master_id;
356 if (master_id < -1 || master_id >= HAMMER_MAX_MASTERS)
5de0c0e5 357 return (EINVAL);
732a1697
MD
358 } else {
359 if (hmp)
360 master_id = hmp->master_id;
361 else
362 master_id = 0;
5de0c0e5 363 }
427e5fc6
MD
364
365 /*
366 * Interal mount data structure
367 */
732a1697 368 if (hmp == NULL) {
195c19a1
MD
369 hmp = kmalloc(sizeof(*hmp), M_HAMMER, M_WAITOK | M_ZERO);
370 mp->mnt_data = (qaddr_t)hmp;
371 hmp->mp = mp;
46fe7ae1 372 /*TAILQ_INIT(&hmp->recycle_list);*/
47197d71 373
bac808fe
MD
374 /*
375 * Make sure kmalloc type limits are set appropriately. If root
376 * increases the vnode limit you may have to do a dummy remount
377 * to adjust the HAMMER inode limit.
378 */
379 kmalloc_create(&hmp->m_misc, "HAMMER-others");
380 kmalloc_create(&hmp->m_inodes, "HAMMER-inodes");
381
382 maxinodes = desiredvnodes + desiredvnodes / 5 +
ff003b11 383 hammer_limit_reclaim * 2;
bac808fe
MD
384 kmalloc_raise_limit(hmp->m_inodes,
385 maxinodes * sizeof(struct hammer_inode));
386
dd94f1b1 387 hmp->root_btree_beg.localization = 0x00000000U;
47197d71
MD
388 hmp->root_btree_beg.obj_id = -0x8000000000000000LL;
389 hmp->root_btree_beg.key = -0x8000000000000000LL;
390 hmp->root_btree_beg.create_tid = 1;
391 hmp->root_btree_beg.delete_tid = 1;
392 hmp->root_btree_beg.rec_type = 0;
393 hmp->root_btree_beg.obj_type = 0;
394
dd94f1b1 395 hmp->root_btree_end.localization = 0xFFFFFFFFU;
47197d71
MD
396 hmp->root_btree_end.obj_id = 0x7FFFFFFFFFFFFFFFLL;
397 hmp->root_btree_end.key = 0x7FFFFFFFFFFFFFFFLL;
398 hmp->root_btree_end.create_tid = 0xFFFFFFFFFFFFFFFFULL;
399 hmp->root_btree_end.delete_tid = 0; /* special case */
400 hmp->root_btree_end.rec_type = 0xFFFFU;
401 hmp->root_btree_end.obj_type = 0;
f03c9cf4 402
cdb6e4e6
MD
403 hmp->krate.freq = 1; /* maximum reporting rate (hz) */
404 hmp->krate.count = -16; /* initial burst */
405
9480ff55 406 hmp->sync_lock.refs = 1;
c9b9e29d 407 hmp->free_lock.refs = 1;
d99d6bf5
MD
408 hmp->undo_lock.refs = 1;
409 hmp->blkmap_lock.refs = 1;
83f2a3aa 410 hmp->snapshot_lock.refs = 1;
9480ff55 411
cebe9493 412 TAILQ_INIT(&hmp->delay_list);
7a61b85d 413 TAILQ_INIT(&hmp->flush_group_list);
0729c8c8 414 TAILQ_INIT(&hmp->objid_cache_list);
e8599db1 415 TAILQ_INIT(&hmp->undo_lru_list);
7bc5b8c2 416 TAILQ_INIT(&hmp->reclaim_list);
195c19a1 417 }
51c35492
MD
418 hmp->hflags &= ~HMNT_USERFLAGS;
419 hmp->hflags |= info.hflags & HMNT_USERFLAGS;
732a1697
MD
420
421 hmp->master_id = master_id;
422
7f7c1f84
MD
423 if (info.asof) {
424 mp->mnt_flag |= MNT_RDONLY;
425 hmp->asof = info.asof;
426 } else {
427 hmp->asof = HAMMER_MAX_TID;
428 }
195c19a1
MD
429
430 /*
51c35492 431 * Re-open read-write if originally read-only, or vise-versa.
02428fb6
MD
432 *
433 * When going from read-only to read-write execute the stage2
434 * recovery if it has not already been run.
195c19a1
MD
435 */
436 if (mp->mnt_flag & MNT_UPDATE) {
51c35492
MD
437 error = 0;
438 if (hmp->ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
439 kprintf("HAMMER read-only -> read-write\n");
440 hmp->ronly = 0;
441 RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
442 hammer_adjust_volume_mode, NULL);
443 rootvol = hammer_get_root_volume(hmp, &error);
444 if (rootvol) {
06ad81ff 445 hammer_recover_flush_buffers(hmp, rootvol, 1);
02428fb6 446 error = hammer_recover_stage2(hmp, rootvol);
9f5097dc
MD
447 bcopy(rootvol->ondisk->vol0_blockmap,
448 hmp->blockmap,
449 sizeof(hmp->blockmap));
51c35492
MD
450 hammer_rel_volume(rootvol, 0);
451 }
452 RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
453 hammer_reload_inode, NULL);
454 /* kernel clears MNT_RDONLY */
455 } else if (hmp->ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
456 kprintf("HAMMER read-write -> read-only\n");
457 hmp->ronly = 1; /* messy */
458 RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
459 hammer_reload_inode, NULL);
195c19a1 460 hmp->ronly = 0;
51c35492
MD
461 hammer_flusher_sync(hmp);
462 hammer_flusher_sync(hmp);
463 hammer_flusher_sync(hmp);
464 hmp->ronly = 1;
465 RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
466 hammer_adjust_volume_mode, NULL);
195c19a1 467 }
51c35492 468 return(error);
195c19a1
MD
469 }
470
427e5fc6
MD
471 RB_INIT(&hmp->rb_vols_root);
472 RB_INIT(&hmp->rb_inos_root);
40043e7f 473 RB_INIT(&hmp->rb_nods_root);
e8599db1 474 RB_INIT(&hmp->rb_undo_root);
0832c9bb
MD
475 RB_INIT(&hmp->rb_resv_root);
476 RB_INIT(&hmp->rb_bufs_root);
5fa5c92f 477 RB_INIT(&hmp->rb_pfsm_root);
0832c9bb 478
195c19a1 479 hmp->ronly = ((mp->mnt_flag & MNT_RDONLY) != 0);
427e5fc6 480
10a5d1ba
MD
481 TAILQ_INIT(&hmp->volu_list);
482 TAILQ_INIT(&hmp->undo_list);
483 TAILQ_INIT(&hmp->data_list);
484 TAILQ_INIT(&hmp->meta_list);
485 TAILQ_INIT(&hmp->lose_list);
486
427e5fc6
MD
487 /*
488 * Load volumes
489 */
490 path = objcache_get(namei_oc, M_WAITOK);
1b0ab2c3 491 hmp->nvolumes = -1;
427e5fc6 492 for (i = 0; i < info.nvolumes; ++i) {
7c19b529
MN
493 if (mntpt == NULL) {
494 /*
495 * Root mount.
496 * Only one volume; and no need for copyin.
497 */
498 KKASSERT(info.nvolumes == 1);
499 ksnprintf(path, MAXPATHLEN, "/dev/%s",
500 mp->mnt_stat.f_mntfromname);
501 error = 0;
502 } else {
503 error = copyin(&info.volumes[i], &upath,
504 sizeof(char *));
505 if (error == 0)
506 error = copyinstr(upath, path,
507 MAXPATHLEN, NULL);
508 }
427e5fc6 509 if (error == 0)
7c19b529 510 error = hammer_install_volume(hmp, path, devvp);
427e5fc6
MD
511 if (error)
512 break;
513 }
514 objcache_put(namei_oc, path);
515
516 /*
517 * Make sure we found a root volume
518 */
519 if (error == 0 && hmp->rootvol == NULL) {
520 kprintf("hammer_mount: No root volume found!\n");
521 error = EINVAL;
522 }
1b0ab2c3
MD
523
524 /*
525 * Check that all required volumes are available
526 */
527 if (error == 0 && hammer_mountcheck_volumes(hmp)) {
528 kprintf("hammer_mount: Missing volumes, cannot mount!\n");
529 error = EINVAL;
530 }
531
427e5fc6
MD
532 if (error) {
533 hammer_free_hmp(mp);
534 return (error);
535 }
536
537 /*
27ea2398
MD
538 * No errors, setup enough of the mount point so we can lookup the
539 * root vnode.
427e5fc6 540 */
27ea2398
MD
541 mp->mnt_iosize_max = MAXPHYS;
542 mp->mnt_kern_flag |= MNTK_FSMID;
c0ade690
MD
543
544 /*
545 * note: f_iosize is used by vnode_pager_haspage() when constructing
546 * its VOP_BMAP call.
547 */
548 mp->mnt_stat.f_iosize = HAMMER_BUFSIZE;
fbc6e32a 549 mp->mnt_stat.f_bsize = HAMMER_BUFSIZE;
6f97fce3
MD
550
551 mp->mnt_vstat.f_frsize = HAMMER_BUFSIZE;
552 mp->mnt_vstat.f_bsize = HAMMER_BUFSIZE;
553
27ea2398
MD
554 mp->mnt_maxsymlinklen = 255;
555 mp->mnt_flag |= MNT_LOCAL;
556
557 vfs_add_vnodeops(mp, &hammer_vnode_vops, &mp->mnt_vn_norm_ops);
7a04d74f
MD
558 vfs_add_vnodeops(mp, &hammer_spec_vops, &mp->mnt_vn_spec_ops);
559 vfs_add_vnodeops(mp, &hammer_fifo_vops, &mp->mnt_vn_fifo_ops);
27ea2398
MD
560
561 /*
a89aec1b
MD
562 * The root volume's ondisk pointer is only valid if we hold a
563 * reference to it.
564 */
565 rootvol = hammer_get_root_volume(hmp, &error);
566 if (error)
f90dde4c
MD
567 goto failed;
568
569 /*
9f5097dc 570 * Perform any necessary UNDO operations. The recovery code does
0729c8c8
MD
571 * call hammer_undo_lookup() so we have to pre-cache the blockmap,
572 * and then re-copy it again after recovery is complete.
c9b9e29d 573 *
51c35492
MD
574 * If this is a read-only mount the UNDO information is retained
575 * in memory in the form of dirty buffer cache buffers, and not
576 * written back to the media.
f90dde4c 577 */
0729c8c8
MD
578 bcopy(rootvol->ondisk->vol0_blockmap, hmp->blockmap,
579 sizeof(hmp->blockmap));
c9b9e29d 580
7a61b85d 581 /*
44a83111
MD
582 * Check filesystem version
583 */
584 hmp->version = rootvol->ondisk->vol_version;
585 if (hmp->version < HAMMER_VOL_VERSION_MIN ||
586 hmp->version > HAMMER_VOL_VERSION_MAX) {
587 kprintf("HAMMER: mount unsupported fs version %d\n",
588 hmp->version);
589 error = ERANGE;
590 goto done;
591 }
592
593 /*
7a61b85d
MD
594 * The undo_rec_limit limits the size of flush groups to avoid
595 * blowing out the UNDO FIFO. This calculation is typically in
596 * the tens of thousands and is designed primarily when small
597 * HAMMER filesystems are created.
598 */
599 hmp->undo_rec_limit = hammer_undo_max(hmp) / 8192 + 100;
600 if (hammer_debug_general & 0x0001)
601 kprintf("HAMMER: undo_rec_limit %d\n", hmp->undo_rec_limit);
602
02428fb6
MD
603 /*
604 * NOTE: Recover stage1 not only handles meta-data recovery, it
605 * also sets hmp->undo_seqno for HAMMER VERSION 4+ filesystems.
606 */
607 error = hammer_recover_stage1(hmp, rootvol);
f90dde4c
MD
608 if (error) {
609 kprintf("Failed to recover HAMMER filesystem on mount\n");
a89aec1b 610 goto done;
f90dde4c
MD
611 }
612
613 /*
adf01747
MD
614 * Finish setup now that we have a good root volume.
615 *
616 * The top 16 bits of fsid.val[1] is a pfs id.
f90dde4c 617 */
a89aec1b
MD
618 ksnprintf(mp->mnt_stat.f_mntfromname,
619 sizeof(mp->mnt_stat.f_mntfromname), "%s",
620 rootvol->ondisk->vol_name);
513ca7d7
MD
621 mp->mnt_stat.f_fsid.val[0] =
622 crc32((char *)&rootvol->ondisk->vol_fsid + 0, 8);
623 mp->mnt_stat.f_fsid.val[1] =
624 crc32((char *)&rootvol->ondisk->vol_fsid + 8, 8);
adf01747 625 mp->mnt_stat.f_fsid.val[1] &= 0x0000FFFF;
b84de5af 626
6f97fce3
MD
627 mp->mnt_vstat.f_fsid_uuid = rootvol->ondisk->vol_fsid;
628 mp->mnt_vstat.f_fsid = crc32(&mp->mnt_vstat.f_fsid_uuid,
629 sizeof(mp->mnt_vstat.f_fsid_uuid));
630
0729c8c8
MD
631 /*
632 * Certain often-modified fields in the root volume are cached in
633 * the hammer_mount structure so we do not have to generate lots
634 * of little UNDO structures for them.
c9b9e29d 635 *
9f5097dc
MD
636 * Recopy after recovery. This also has the side effect of
637 * setting our cached undo FIFO's first_offset, which serves to
638 * placemark the FIFO start for the NEXT flush cycle while the
639 * on-disk first_offset represents the LAST flush cycle.
0729c8c8 640 */
b84de5af 641 hmp->next_tid = rootvol->ondisk->vol0_next_tid;
4889cbd4
MD
642 hmp->flush_tid1 = hmp->next_tid;
643 hmp->flush_tid2 = hmp->next_tid;
0729c8c8
MD
644 bcopy(rootvol->ondisk->vol0_blockmap, hmp->blockmap,
645 sizeof(hmp->blockmap));
e63644f0 646 hmp->copy_stat_freebigblocks = rootvol->ondisk->vol0_stat_freebigblocks;
0729c8c8 647
059819e3
MD
648 hammer_flusher_create(hmp);
649
a89aec1b 650 /*
27ea2398
MD
651 * Locate the root directory using the root cluster's B-Tree as a
652 * starting point. The root directory uses an obj_id of 1.
653 *
654 * FUTURE: Leave the root directory cached referenced but unlocked
655 * in hmp->rootvp (need to flush it on unmount).
656 */
b9b0a6d0 657 error = hammer_vfs_vget(mp, NULL, 1, &rootvp);
a89aec1b
MD
658 if (error)
659 goto done;
660 vput(rootvp);
27ea2398 661 /*vn_unlock(hmp->rootvp);*/
02428fb6 662 error = hammer_recover_stage2(hmp, rootvol);
427e5fc6 663
a89aec1b 664done:
f90dde4c
MD
665 hammer_rel_volume(rootvol, 0);
666failed:
427e5fc6 667 /*
27ea2398 668 * Cleanup and return.
427e5fc6 669 */
27ea2398 670 if (error)
427e5fc6 671 hammer_free_hmp(mp);
427e5fc6
MD
672 return (error);
673}
674
675static int
676hammer_vfs_unmount(struct mount *mp, int mntflags)
677{
678#if 0
679 struct hammer_mount *hmp = (void *)mp->mnt_data;
680#endif
681 int flags;
66325755 682 int error;
27ea2398
MD
683
684 /*
427e5fc6
MD
685 * Clean out the vnodes
686 */
66325755
MD
687 flags = 0;
688 if (mntflags & MNT_FORCE)
689 flags |= FORCECLOSE;
690 if ((error = vflush(mp, 0, flags)) != 0)
691 return (error);
427e5fc6
MD
692
693 /*
694 * Clean up the internal mount structure and related entities. This
695 * may issue I/O.
696 */
697 hammer_free_hmp(mp);
698 return(0);
699}
700
701/*
702 * Clean up the internal mount structure and disassociate it from the mount.
703 * This may issue I/O.
704 */
705static void
706hammer_free_hmp(struct mount *mp)
707{
708 struct hammer_mount *hmp = (void *)mp->mnt_data;
cdb6e4e6 709 hammer_flush_group_t flg;
1b0ab2c3 710 int count;
d40bfeca 711 int dummy;
427e5fc6
MD
712
713 /*
cdb6e4e6
MD
714 * Flush anything dirty. This won't even run if the
715 * filesystem errored-out.
427e5fc6 716 */
1b0ab2c3
MD
717 count = 0;
718 while (hammer_flusher_haswork(hmp)) {
719 hammer_flusher_sync(hmp);
720 ++count;
721 if (count >= 5) {
722 if (count == 5)
723 kprintf("HAMMER: umount flushing.");
724 else
725 kprintf(".");
d40bfeca 726 tsleep(&dummy, 0, "hmrufl", hz);
1b0ab2c3
MD
727 }
728 if (count == 30) {
729 kprintf("giving up\n");
730 break;
731 }
732 }
733 if (count >= 5 && count < 30)
734 kprintf("\n");
427e5fc6 735
cdb6e4e6
MD
736 /*
737 * If the mount had a critical error we have to destroy any
738 * remaining inodes before we can finish cleaning up the flusher.
739 */
740 if (hmp->flags & HAMMER_MOUNT_CRITICAL_ERROR) {
741 RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
742 hammer_destroy_inode_callback, NULL);
743 }
b84de5af 744
427e5fc6 745 /*
cdb6e4e6
MD
746 * There shouldn't be any inodes left now and any left over
747 * flush groups should now be empty.
427e5fc6 748 */
cdb6e4e6
MD
749 KKASSERT(RB_EMPTY(&hmp->rb_inos_root));
750 while ((flg = TAILQ_FIRST(&hmp->flush_group_list)) != NULL) {
751 TAILQ_REMOVE(&hmp->flush_group_list, flg, flush_entry);
ff003b11 752 KKASSERT(RB_EMPTY(&flg->flush_tree));
cdb6e4e6
MD
753 if (flg->refs) {
754 kprintf("HAMMER: Warning, flush_group %p was "
755 "not empty on umount!\n", flg);
756 }
bac808fe 757 kfree(flg, hmp->m_misc);
cdb6e4e6 758 }
427e5fc6
MD
759
760 /*
cdb6e4e6 761 * We can finally destroy the flusher
427e5fc6 762 */
cdb6e4e6
MD
763 hammer_flusher_destroy(hmp);
764
b84de5af 765 /*
00f16fad
MD
766 * We may have held recovered buffers due to a read-only mount.
767 * These must be discarded.
768 */
769 if (hmp->ronly)
770 hammer_recover_flush_buffers(hmp, NULL, -1);
771
772 /*
0832c9bb 773 * Unload buffers and then volumes
b84de5af 774 */
0832c9bb
MD
775 RB_SCAN(hammer_buf_rb_tree, &hmp->rb_bufs_root, NULL,
776 hammer_unload_buffer, NULL);
427e5fc6
MD
777 RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
778 hammer_unload_volume, NULL);
779
780 mp->mnt_data = NULL;
66325755 781 mp->mnt_flag &= ~MNT_LOCAL;
427e5fc6 782 hmp->mp = NULL;
0729c8c8 783 hammer_destroy_objid_cache(hmp);
bac808fe
MD
784 kmalloc_destroy(&hmp->m_misc);
785 kmalloc_destroy(&hmp->m_inodes);
427e5fc6
MD
786 kfree(hmp, M_HAMMER);
787}
788
789/*
cdb6e4e6
MD
790 * Report critical errors. ip may be NULL.
791 */
792void
793hammer_critical_error(hammer_mount_t hmp, hammer_inode_t ip,
794 int error, const char *msg)
795{
796 hmp->flags |= HAMMER_MOUNT_CRITICAL_ERROR;
c9ce54d6 797
cdb6e4e6 798 krateprintf(&hmp->krate,
c9ce54d6
MD
799 "HAMMER(%s): Critical error inode=%jd error=%d %s\n",
800 hmp->mp->mnt_stat.f_mntfromname,
801 (intmax_t)(ip ? ip->obj_id : -1),
802 error, msg);
803
cdb6e4e6
MD
804 if (hmp->ronly == 0) {
805 hmp->ronly = 2; /* special errored read-only mode */
806 hmp->mp->mnt_flag |= MNT_RDONLY;
807 kprintf("HAMMER(%s): Forcing read-only mode\n",
808 hmp->mp->mnt_stat.f_mntfromname);
809 }
810 hmp->error = error;
c9ce54d6
MD
811 if (hammer_error_panic > 2)
812 Debugger("Entering debugger");
cdb6e4e6
MD
813}
814
815
816/*
513ca7d7
MD
817 * Obtain a vnode for the specified inode number. An exclusively locked
818 * vnode is returned.
819 */
820int
b9b0a6d0
MD
821hammer_vfs_vget(struct mount *mp, struct vnode *dvp,
822 ino_t ino, struct vnode **vpp)
513ca7d7 823{
36f82b23 824 struct hammer_transaction trans;
513ca7d7
MD
825 struct hammer_mount *hmp = (void *)mp->mnt_data;
826 struct hammer_inode *ip;
827 int error;
b9b0a6d0 828 u_int32_t localization;
513ca7d7 829
36f82b23
MD
830 hammer_simple_transaction(&trans, hmp);
831
513ca7d7 832 /*
b9b0a6d0
MD
833 * If a directory vnode is supplied (mainly NFS) then we can acquire
834 * the PFS domain from it. Otherwise we would only be able to vget
835 * inodes in the root PFS.
836 */
837 if (dvp) {
838 localization = HAMMER_DEF_LOCALIZATION +
839 VTOI(dvp)->obj_localization;
840 } else {
841 localization = HAMMER_DEF_LOCALIZATION;
842 }
843
844 /*
36f82b23
MD
845 * Lookup the requested HAMMER inode. The structure must be
846 * left unlocked while we manipulate the related vnode to avoid
847 * a deadlock.
513ca7d7 848 */
ddfdf542 849 ip = hammer_get_inode(&trans, NULL, ino,
b9b0a6d0 850 hmp->asof, localization,
ddfdf542 851 0, &error);
513ca7d7
MD
852 if (ip == NULL) {
853 *vpp = NULL;
00f16fad 854 hammer_done_transaction(&trans);
513ca7d7
MD
855 return(error);
856 }
e8599db1 857 error = hammer_get_vnode(ip, vpp);
513ca7d7 858 hammer_rel_inode(ip, 0);
b84de5af 859 hammer_done_transaction(&trans);
513ca7d7
MD
860 return (error);
861}
862
863/*
427e5fc6
MD
864 * Return the root vnode for the filesystem.
865 *
866 * HAMMER stores the root vnode in the hammer_mount structure so
867 * getting it is easy.
868 */
869static int
870hammer_vfs_root(struct mount *mp, struct vnode **vpp)
871{
47197d71 872#if 0
427e5fc6 873 struct hammer_mount *hmp = (void *)mp->mnt_data;
47197d71 874#endif
27ea2398 875 int error;
427e5fc6 876
b9b0a6d0 877 error = hammer_vfs_vget(mp, NULL, 1, vpp);
27ea2398 878 return (error);
427e5fc6
MD
879}
880
881static int
882hammer_vfs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
883{
fbc6e32a
MD
884 struct hammer_mount *hmp = (void *)mp->mnt_data;
885 hammer_volume_t volume;
886 hammer_volume_ondisk_t ondisk;
887 int error;
47197d71 888 int64_t bfree;
31a56ce2 889 int64_t breserved;
fbc6e32a
MD
890
891 volume = hammer_get_root_volume(hmp, &error);
892 if (error)
893 return(error);
fbc6e32a
MD
894 ondisk = volume->ondisk;
895
47197d71
MD
896 /*
897 * Basic stats
898 */
31a56ce2 899 _hammer_checkspace(hmp, HAMMER_CHKSPC_WRITE, &breserved);
fbc6e32a 900 mp->mnt_stat.f_files = ondisk->vol0_stat_inodes;
c3be93f2 901 bfree = ondisk->vol0_stat_freebigblocks * HAMMER_LARGEBLOCK_SIZE;
40043e7f 902 hammer_rel_volume(volume, 0);
47197d71 903
31a56ce2 904 mp->mnt_stat.f_bfree = (bfree - breserved) / HAMMER_BUFSIZE;
47197d71 905 mp->mnt_stat.f_bavail = mp->mnt_stat.f_bfree;
fbc6e32a
MD
906 if (mp->mnt_stat.f_files < 0)
907 mp->mnt_stat.f_files = 0;
908
27ea2398
MD
909 *sbp = mp->mnt_stat;
910 return(0);
427e5fc6
MD
911}
912
6f97fce3
MD
913static int
914hammer_vfs_statvfs(struct mount *mp, struct statvfs *sbp, struct ucred *cred)
915{
916 struct hammer_mount *hmp = (void *)mp->mnt_data;
917 hammer_volume_t volume;
918 hammer_volume_ondisk_t ondisk;
919 int error;
920 int64_t bfree;
0f65be10 921 int64_t breserved;
6f97fce3
MD
922
923 volume = hammer_get_root_volume(hmp, &error);
924 if (error)
925 return(error);
926 ondisk = volume->ondisk;
927
928 /*
929 * Basic stats
930 */
0f65be10 931 _hammer_checkspace(hmp, HAMMER_CHKSPC_WRITE, &breserved);
6f97fce3
MD
932 mp->mnt_vstat.f_files = ondisk->vol0_stat_inodes;
933 bfree = ondisk->vol0_stat_freebigblocks * HAMMER_LARGEBLOCK_SIZE;
934 hammer_rel_volume(volume, 0);
935
0f65be10 936 mp->mnt_vstat.f_bfree = (bfree - breserved) / HAMMER_BUFSIZE;
6f97fce3
MD
937 mp->mnt_vstat.f_bavail = mp->mnt_stat.f_bfree;
938 if (mp->mnt_vstat.f_files < 0)
939 mp->mnt_vstat.f_files = 0;
940 *sbp = mp->mnt_vstat;
941 return(0);
942}
943
0729c8c8
MD
944/*
945 * Sync the filesystem. Currently we have to run it twice, the second
946 * one will advance the undo start index to the end index, so if a crash
947 * occurs no undos will be run on mount.
77062c8a
MD
948 *
949 * We do not sync the filesystem if we are called from a panic. If we did
950 * we might end up blowing up a sync that was already in progress.
0729c8c8 951 */
427e5fc6
MD
952static int
953hammer_vfs_sync(struct mount *mp, int waitfor)
954{
fbc6e32a 955 struct hammer_mount *hmp = (void *)mp->mnt_data;
0729c8c8
MD
956 int error;
957
77062c8a 958 if (panicstr == NULL) {
0729c8c8 959 error = hammer_sync_hmp(hmp, waitfor);
77062c8a
MD
960 } else {
961 error = EIO;
77062c8a 962 }
0729c8c8 963 return (error);
427e5fc6
MD
964}
965
513ca7d7
MD
966/*
967 * Convert a vnode to a file handle.
968 */
969static int
970hammer_vfs_vptofh(struct vnode *vp, struct fid *fhp)
971{
972 hammer_inode_t ip;
973
974 KKASSERT(MAXFIDSZ >= 16);
975 ip = VTOI(vp);
976 fhp->fid_len = offsetof(struct fid, fid_data[16]);
adf01747 977 fhp->fid_ext = ip->obj_localization >> 16;
513ca7d7
MD
978 bcopy(&ip->obj_id, fhp->fid_data + 0, sizeof(ip->obj_id));
979 bcopy(&ip->obj_asof, fhp->fid_data + 8, sizeof(ip->obj_asof));
980 return(0);
981}
982
983
984/*
985 * Convert a file handle back to a vnode.
67863d04
MD
986 *
987 * Use rootvp to enforce PFS isolation when a PFS is exported via a
988 * null mount.
513ca7d7
MD
989 */
990static int
67863d04
MD
991hammer_vfs_fhtovp(struct mount *mp, struct vnode *rootvp,
992 struct fid *fhp, struct vnode **vpp)
513ca7d7 993{
36f82b23 994 struct hammer_transaction trans;
513ca7d7
MD
995 struct hammer_inode *ip;
996 struct hammer_inode_info info;
997 int error;
adf01747 998 u_int32_t localization;
513ca7d7
MD
999
1000 bcopy(fhp->fid_data + 0, &info.obj_id, sizeof(info.obj_id));
1001 bcopy(fhp->fid_data + 8, &info.obj_asof, sizeof(info.obj_asof));
67863d04
MD
1002 if (rootvp)
1003 localization = VTOI(rootvp)->obj_localization;
1004 else
1005 localization = (u_int32_t)fhp->fid_ext << 16;
513ca7d7 1006
36f82b23
MD
1007 hammer_simple_transaction(&trans, (void *)mp->mnt_data);
1008
513ca7d7
MD
1009 /*
1010 * Get/allocate the hammer_inode structure. The structure must be
1011 * unlocked while we manipulate the related vnode to avoid a
1012 * deadlock.
1013 */
ddfdf542 1014 ip = hammer_get_inode(&trans, NULL, info.obj_id,
adf01747 1015 info.obj_asof, localization, 0, &error);
513ca7d7
MD
1016 if (ip == NULL) {
1017 *vpp = NULL;
1018 return(error);
1019 }
e8599db1 1020 error = hammer_get_vnode(ip, vpp);
513ca7d7 1021 hammer_rel_inode(ip, 0);
b84de5af 1022 hammer_done_transaction(&trans);
513ca7d7
MD
1023 return (error);
1024}
1025
1026static int
1027hammer_vfs_checkexp(struct mount *mp, struct sockaddr *nam,
1028 int *exflagsp, struct ucred **credanonp)
1029{
1030 hammer_mount_t hmp = (void *)mp->mnt_data;
1031 struct netcred *np;
1032 int error;
1033
1034 np = vfs_export_lookup(mp, &hmp->export, nam);
1035 if (np) {
1036 *exflagsp = np->netc_exflags;
1037 *credanonp = &np->netc_anon;
1038 error = 0;
1039 } else {
1040 error = EACCES;
1041 }
1042 return (error);
1043
1044}
1045
1046int
1047hammer_vfs_export(struct mount *mp, int op, const struct export_args *export)
1048{
1049 hammer_mount_t hmp = (void *)mp->mnt_data;
1050 int error;
1051
1052 switch(op) {
1053 case MOUNTCTL_SET_EXPORT:
1054 error = vfs_export(mp, &hmp->export, export);
1055 break;
1056 default:
1057 error = EOPNOTSUPP;
1058 break;
1059 }
1060 return(error);
1061}
1062