HAMMER - Add live dedup sysctl and support
[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
aac0aabd
MD
50/*
51 * NOTE! Global statistics may not be MPSAFE so HAMMER never uses them
52 * in conditionals.
53 */
02428fb6 54int hammer_supported_version = HAMMER_VOL_VERSION_DEFAULT;
2f85fa4d 55int hammer_debug_io;
d5ef456e 56int hammer_debug_general;
1b0ab2c3 57int hammer_debug_debug = 1; /* medium-error panics */
e8599db1 58int hammer_debug_inode;
7d683b0f 59int hammer_debug_locks;
b3deaf57 60int hammer_debug_btree;
d113fda1 61int hammer_debug_tid;
cebe9493 62int hammer_debug_recover; /* -1 will disable, +1 will force */
46fe7ae1 63int hammer_debug_recover_faults;
fc73edd8 64int hammer_debug_critical; /* non-zero enter debugger on error */
1b0ab2c3 65int hammer_cluster_enable = 1; /* enable read clustering by default */
507df98a 66int hammer_live_dedup = 0;
7a61b85d 67int hammer_count_fsyncs;
b3deaf57 68int hammer_count_inodes;
af209b0f 69int hammer_count_iqueued;
9f5097dc 70int hammer_count_reclaiming;
b3deaf57
MD
71int hammer_count_records;
72int hammer_count_record_datas;
73int hammer_count_volumes;
b3deaf57
MD
74int hammer_count_buffers;
75int hammer_count_nodes;
a7e9bef1 76int64_t hammer_count_extra_space_used;
cb51be26
MD
77int64_t hammer_stats_btree_lookups;
78int64_t hammer_stats_btree_searches;
79int64_t hammer_stats_btree_inserts;
80int64_t hammer_stats_btree_deletes;
81int64_t hammer_stats_btree_elements;
82int64_t hammer_stats_btree_splits;
83int64_t hammer_stats_btree_iterations;
39d8fd63 84int64_t hammer_stats_btree_root_iterations;
cb51be26 85int64_t hammer_stats_record_iterations;
ce0138a6
MD
86
87int64_t hammer_stats_file_read;
88int64_t hammer_stats_file_write;
89int64_t hammer_stats_file_iopsr;
90int64_t hammer_stats_file_iopsw;
91int64_t hammer_stats_disk_read;
92int64_t hammer_stats_disk_write;
93int64_t hammer_stats_inode_flushes;
94int64_t hammer_stats_commits;
89e744ce 95int64_t hammer_stats_undo;
6048b411 96int64_t hammer_stats_redo;
ce0138a6 97
f5a07a7a 98int hammer_count_dirtybufspace; /* global */
a99b9ea2 99int hammer_count_refedbufs; /* global */
0832c9bb 100int hammer_count_reservations;
a99b9ea2
MD
101int hammer_count_io_running_read;
102int hammer_count_io_running_write;
103int hammer_count_io_locked;
f5a07a7a 104int hammer_limit_dirtybufspace; /* per-mount */
ba298df1 105int hammer_limit_running_io; /* per-mount */
0832c9bb 106int hammer_limit_recs; /* as a whole XXX */
de996e86 107int hammer_limit_inode_recs = 1024; /* per inode */
ff003b11 108int hammer_limit_reclaim = HAMMER_RECLAIM_WAIT;
6048b411 109int hammer_limit_redo = 4096 * 1024; /* per inode */
21fde338 110int hammer_autoflush = 2000; /* auto flush */
1f07f686 111int hammer_bio_count;
cb51be26 112int hammer_verify_zone;
1b0ab2c3 113int hammer_verify_data = 1;
cb51be26 114int hammer_write_mode;
3e583440 115int hammer_yield_check = 16;
9d0a6205 116int hammer_fsync_mode = 3;
7d683b0f 117int64_t hammer_contention_count;
f03c9cf4 118int64_t hammer_zone_limit;
b3deaf57 119
507df98a
ID
120/*
121 * Live dedup debug counters (sysctls are writable so that counters
122 * can be reset from userspace).
123 */
124int64_t hammer_live_dedup_vnode_bcmps = 0;
125int64_t hammer_live_dedup_device_bcmps = 0;
126int64_t hammer_live_dedup_findblk_failures = 0;
127int64_t hammer_live_dedup_bmap_saves = 0;
128
129
b3deaf57 130SYSCTL_NODE(_vfs, OID_AUTO, hammer, CTLFLAG_RW, 0, "HAMMER filesystem");
507df98a 131
5987cc42
MD
132SYSCTL_INT(_vfs_hammer, OID_AUTO, supported_version, CTLFLAG_RD,
133 &hammer_supported_version, 0, "");
d5ef456e
MD
134SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_general, CTLFLAG_RW,
135 &hammer_debug_general, 0, "");
2f85fa4d
MD
136SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_io, CTLFLAG_RW,
137 &hammer_debug_io, 0, "");
77062c8a
MD
138SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_debug, CTLFLAG_RW,
139 &hammer_debug_debug, 0, "");
e8599db1
MD
140SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_inode, CTLFLAG_RW,
141 &hammer_debug_inode, 0, "");
7d683b0f
MD
142SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_locks, CTLFLAG_RW,
143 &hammer_debug_locks, 0, "");
b3deaf57
MD
144SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_btree, CTLFLAG_RW,
145 &hammer_debug_btree, 0, "");
d113fda1
MD
146SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_tid, CTLFLAG_RW,
147 &hammer_debug_tid, 0, "");
b33e2cc0
MD
148SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_recover, CTLFLAG_RW,
149 &hammer_debug_recover, 0, "");
46fe7ae1
MD
150SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_recover_faults, CTLFLAG_RW,
151 &hammer_debug_recover_faults, 0, "");
fc73edd8
MD
152SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_critical, CTLFLAG_RW,
153 &hammer_debug_critical, 0, "");
1b0ab2c3
MD
154SYSCTL_INT(_vfs_hammer, OID_AUTO, cluster_enable, CTLFLAG_RW,
155 &hammer_cluster_enable, 0, "");
507df98a
ID
156/*
157 * 0 - live dedup is disabled
158 * 1 - dedup cache is populated on reads only
159 * 2 - dedup cache is populated on both reads and writes
160 */
161SYSCTL_INT(_vfs_hammer, OID_AUTO, live_dedup, CTLFLAG_RW,
162 &hammer_live_dedup, 0, "");
9480ff55 163
f5a07a7a
MD
164SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_dirtybufspace, CTLFLAG_RW,
165 &hammer_limit_dirtybufspace, 0, "");
ba298df1
MD
166SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_running_io, CTLFLAG_RW,
167 &hammer_limit_running_io, 0, "");
47637bff
MD
168SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_recs, CTLFLAG_RW,
169 &hammer_limit_recs, 0, "");
de996e86
MD
170SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_inode_recs, CTLFLAG_RW,
171 &hammer_limit_inode_recs, 0, "");
ff003b11
MD
172SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_reclaim, CTLFLAG_RW,
173 &hammer_limit_reclaim, 0, "");
6048b411
MD
174SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_redo, CTLFLAG_RW,
175 &hammer_limit_redo, 0, "");
9480ff55 176
7a61b85d
MD
177SYSCTL_INT(_vfs_hammer, OID_AUTO, count_fsyncs, CTLFLAG_RD,
178 &hammer_count_fsyncs, 0, "");
b3deaf57
MD
179SYSCTL_INT(_vfs_hammer, OID_AUTO, count_inodes, CTLFLAG_RD,
180 &hammer_count_inodes, 0, "");
af209b0f
MD
181SYSCTL_INT(_vfs_hammer, OID_AUTO, count_iqueued, CTLFLAG_RD,
182 &hammer_count_iqueued, 0, "");
9f5097dc
MD
183SYSCTL_INT(_vfs_hammer, OID_AUTO, count_reclaiming, CTLFLAG_RD,
184 &hammer_count_reclaiming, 0, "");
b3deaf57
MD
185SYSCTL_INT(_vfs_hammer, OID_AUTO, count_records, CTLFLAG_RD,
186 &hammer_count_records, 0, "");
187SYSCTL_INT(_vfs_hammer, OID_AUTO, count_record_datas, CTLFLAG_RD,
188 &hammer_count_record_datas, 0, "");
189SYSCTL_INT(_vfs_hammer, OID_AUTO, count_volumes, CTLFLAG_RD,
190 &hammer_count_volumes, 0, "");
b3deaf57
MD
191SYSCTL_INT(_vfs_hammer, OID_AUTO, count_buffers, CTLFLAG_RD,
192 &hammer_count_buffers, 0, "");
193SYSCTL_INT(_vfs_hammer, OID_AUTO, count_nodes, CTLFLAG_RD,
194 &hammer_count_nodes, 0, "");
a7e9bef1
MD
195SYSCTL_QUAD(_vfs_hammer, OID_AUTO, count_extra_space_used, CTLFLAG_RD,
196 &hammer_count_extra_space_used, 0, "");
ce0138a6 197
cb51be26
MD
198SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_searches, CTLFLAG_RD,
199 &hammer_stats_btree_searches, 0, "");
200SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_lookups, CTLFLAG_RD,
201 &hammer_stats_btree_lookups, 0, "");
202SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_inserts, CTLFLAG_RD,
203 &hammer_stats_btree_inserts, 0, "");
204SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_deletes, CTLFLAG_RD,
205 &hammer_stats_btree_deletes, 0, "");
206SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_elements, CTLFLAG_RD,
207 &hammer_stats_btree_elements, 0, "");
208SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_splits, CTLFLAG_RD,
209 &hammer_stats_btree_splits, 0, "");
210SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_iterations, CTLFLAG_RD,
211 &hammer_stats_btree_iterations, 0, "");
39d8fd63
MD
212SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_root_iterations, CTLFLAG_RD,
213 &hammer_stats_btree_root_iterations, 0, "");
cb51be26
MD
214SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_record_iterations, CTLFLAG_RD,
215 &hammer_stats_record_iterations, 0, "");
ce0138a6
MD
216
217SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_read, CTLFLAG_RD,
218 &hammer_stats_file_read, 0, "");
219SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_write, CTLFLAG_RD,
220 &hammer_stats_file_write, 0, "");
221SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_iopsr, CTLFLAG_RD,
222 &hammer_stats_file_iopsr, 0, "");
223SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_iopsw, CTLFLAG_RD,
224 &hammer_stats_file_iopsw, 0, "");
225SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_disk_read, CTLFLAG_RD,
226 &hammer_stats_disk_read, 0, "");
227SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_disk_write, CTLFLAG_RD,
228 &hammer_stats_disk_write, 0, "");
229SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_inode_flushes, CTLFLAG_RD,
230 &hammer_stats_inode_flushes, 0, "");
231SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_commits, CTLFLAG_RD,
232 &hammer_stats_commits, 0, "");
89e744ce
MD
233SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_undo, CTLFLAG_RD,
234 &hammer_stats_undo, 0, "");
6048b411
MD
235SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_redo, CTLFLAG_RD,
236 &hammer_stats_redo, 0, "");
ce0138a6 237
507df98a
ID
238SYSCTL_QUAD(_vfs_hammer, OID_AUTO, live_dedup_vnode_bcmps, CTLFLAG_RW,
239 &hammer_live_dedup_vnode_bcmps, 0, "");
240SYSCTL_QUAD(_vfs_hammer, OID_AUTO, live_dedup_device_bcmps, CTLFLAG_RW,
241 &hammer_live_dedup_device_bcmps, 0, "");
242SYSCTL_QUAD(_vfs_hammer, OID_AUTO, live_dedup_findblk_failures, CTLFLAG_RW,
243 &hammer_live_dedup_findblk_failures, 0, "");
244SYSCTL_QUAD(_vfs_hammer, OID_AUTO, live_dedup_bmap_saves, CTLFLAG_RW,
245 &hammer_live_dedup_bmap_saves, 0, "");
246
f5a07a7a
MD
247SYSCTL_INT(_vfs_hammer, OID_AUTO, count_dirtybufspace, CTLFLAG_RD,
248 &hammer_count_dirtybufspace, 0, "");
a99b9ea2
MD
249SYSCTL_INT(_vfs_hammer, OID_AUTO, count_refedbufs, CTLFLAG_RD,
250 &hammer_count_refedbufs, 0, "");
0832c9bb
MD
251SYSCTL_INT(_vfs_hammer, OID_AUTO, count_reservations, CTLFLAG_RD,
252 &hammer_count_reservations, 0, "");
a99b9ea2
MD
253SYSCTL_INT(_vfs_hammer, OID_AUTO, count_io_running_read, CTLFLAG_RD,
254 &hammer_count_io_running_read, 0, "");
255SYSCTL_INT(_vfs_hammer, OID_AUTO, count_io_locked, CTLFLAG_RD,
256 &hammer_count_io_locked, 0, "");
257SYSCTL_INT(_vfs_hammer, OID_AUTO, count_io_running_write, CTLFLAG_RD,
258 &hammer_count_io_running_write, 0, "");
f03c9cf4
MD
259SYSCTL_QUAD(_vfs_hammer, OID_AUTO, zone_limit, CTLFLAG_RW,
260 &hammer_zone_limit, 0, "");
7d683b0f
MD
261SYSCTL_QUAD(_vfs_hammer, OID_AUTO, contention_count, CTLFLAG_RW,
262 &hammer_contention_count, 0, "");
21fde338
MD
263SYSCTL_INT(_vfs_hammer, OID_AUTO, autoflush, CTLFLAG_RW,
264 &hammer_autoflush, 0, "");
cb51be26
MD
265SYSCTL_INT(_vfs_hammer, OID_AUTO, verify_zone, CTLFLAG_RW,
266 &hammer_verify_zone, 0, "");
1b0ab2c3
MD
267SYSCTL_INT(_vfs_hammer, OID_AUTO, verify_data, CTLFLAG_RW,
268 &hammer_verify_data, 0, "");
cb51be26
MD
269SYSCTL_INT(_vfs_hammer, OID_AUTO, write_mode, CTLFLAG_RW,
270 &hammer_write_mode, 0, "");
3e583440
MD
271SYSCTL_INT(_vfs_hammer, OID_AUTO, yield_check, CTLFLAG_RW,
272 &hammer_yield_check, 0, "");
6f3d87c0
MD
273SYSCTL_INT(_vfs_hammer, OID_AUTO, fsync_mode, CTLFLAG_RW,
274 &hammer_fsync_mode, 0, "");
b3deaf57 275
bcac4bbb
MD
276KTR_INFO_MASTER(hammer);
277
427e5fc6
MD
278/*
279 * VFS ABI
280 */
281static void hammer_free_hmp(struct mount *mp);
282
283static int hammer_vfs_mount(struct mount *mp, char *path, caddr_t data,
284 struct ucred *cred);
285static int hammer_vfs_unmount(struct mount *mp, int mntflags);
286static int hammer_vfs_root(struct mount *mp, struct vnode **vpp);
287static int hammer_vfs_statfs(struct mount *mp, struct statfs *sbp,
288 struct ucred *cred);
6f97fce3
MD
289static int hammer_vfs_statvfs(struct mount *mp, struct statvfs *sbp,
290 struct ucred *cred);
427e5fc6 291static int hammer_vfs_sync(struct mount *mp, int waitfor);
b9b0a6d0
MD
292static int hammer_vfs_vget(struct mount *mp, struct vnode *dvp,
293 ino_t ino, struct vnode **vpp);
427e5fc6 294static int hammer_vfs_init(struct vfsconf *conf);
67863d04
MD
295static int hammer_vfs_fhtovp(struct mount *mp, struct vnode *rootvp,
296 struct fid *fhp, struct vnode **vpp);
513ca7d7
MD
297static int hammer_vfs_vptofh(struct vnode *vp, struct fid *fhp);
298static int hammer_vfs_checkexp(struct mount *mp, struct sockaddr *nam,
299 int *exflagsp, struct ucred **credanonp);
300
427e5fc6
MD
301
302static struct vfsops hammer_vfsops = {
303 .vfs_mount = hammer_vfs_mount,
304 .vfs_unmount = hammer_vfs_unmount,
305 .vfs_root = hammer_vfs_root,
306 .vfs_statfs = hammer_vfs_statfs,
6f97fce3 307 .vfs_statvfs = hammer_vfs_statvfs,
427e5fc6
MD
308 .vfs_sync = hammer_vfs_sync,
309 .vfs_vget = hammer_vfs_vget,
513ca7d7
MD
310 .vfs_init = hammer_vfs_init,
311 .vfs_vptofh = hammer_vfs_vptofh,
312 .vfs_fhtovp = hammer_vfs_fhtovp,
313 .vfs_checkexp = hammer_vfs_checkexp
427e5fc6
MD
314};
315
bac808fe 316MALLOC_DEFINE(M_HAMMER, "HAMMER-mount", "");
427e5fc6
MD
317
318VFS_SET(hammer_vfsops, hammer, 0);
319MODULE_VERSION(hammer, 1);
320
321static int
322hammer_vfs_init(struct vfsconf *conf)
323{
3098dc2f
MD
324 int n;
325
326 if (hammer_limit_recs == 0) {
a99b9ea2 327 hammer_limit_recs = nbuf * 25;
3098dc2f
MD
328 n = kmalloc_limit(M_HAMMER) / 512;
329 if (hammer_limit_recs > n)
330 hammer_limit_recs = n;
331 }
f5a07a7a
MD
332 if (hammer_limit_dirtybufspace == 0) {
333 hammer_limit_dirtybufspace = hidirtybufspace / 2;
334 if (hammer_limit_dirtybufspace < 100)
335 hammer_limit_dirtybufspace = 100;
9f5097dc 336 }
ba298df1
MD
337
338 /*
339 * Set reasonable limits to maintain an I/O pipeline. This is
340 * used by the flush code which explicitly initiates I/O, and
341 * is per-mount.
342 *
343 * The system-driven buffer cache uses vfs.lorunningspace and
344 * vfs.hirunningspace globally.
345 */
346 if (hammer_limit_running_io == 0)
347 hammer_limit_running_io = hammer_limit_dirtybufspace;
348 if (hammer_limit_running_io > 10 * 1024 * 1024)
349 hammer_limit_running_io = 10 * 1024 * 1024;
427e5fc6
MD
350 return(0);
351}
352
353static int
354hammer_vfs_mount(struct mount *mp, char *mntpt, caddr_t data,
355 struct ucred *cred)
356{
357 struct hammer_mount_info info;
a89aec1b
MD
358 hammer_mount_t hmp;
359 hammer_volume_t rootvol;
27ea2398 360 struct vnode *rootvp;
7c19b529 361 struct vnode *devvp = NULL;
427e5fc6
MD
362 const char *upath; /* volume name in userspace */
363 char *path; /* volume name in system space */
364 int error;
365 int i;
732a1697 366 int master_id;
104cb849 367 char *next_volume_ptr = NULL;
ab0bf4a3
MD
368
369 /*
370 * Accept hammer_mount_info. mntpt is NULL for root mounts at boot.
371 */
7c19b529 372 if (mntpt == NULL) {
7c19b529
MN
373 bzero(&info, sizeof(info));
374 info.asof = 0;
375 info.hflags = 0;
376 info.nvolumes = 1;
104cb849
MN
377
378 next_volume_ptr = mp->mnt_stat.f_mntfromname;
379
380 /* Count number of volumes separated by ':' */
381 for (char *p = next_volume_ptr; *p != '\0'; ++p) {
382 if (*p == ':') {
383 ++info.nvolumes;
384 }
385 }
386
387 mp->mnt_flag &= ~MNT_RDONLY; /* mount R/W */
7c19b529
MN
388 } else {
389 if ((error = copyin(data, &info, sizeof(info))) != 0)
390 return (error);
391 }
732a1697
MD
392
393 /*
394 * updating or new mount
395 */
396 if (mp->mnt_flag & MNT_UPDATE) {
397 hmp = (void *)mp->mnt_data;
398 KKASSERT(hmp != NULL);
399 } else {
51c35492
MD
400 if (info.nvolumes <= 0 || info.nvolumes >= 32768)
401 return (EINVAL);
732a1697 402 hmp = NULL;
51c35492 403 }
732a1697
MD
404
405 /*
406 * master-id validation. The master id may not be changed by a
407 * mount update.
408 */
409 if (info.hflags & HMNT_MASTERID) {
410 if (hmp && hmp->master_id != info.master_id) {
411 kprintf("hammer: cannot change master id "
412 "with mount update\n");
413 return(EINVAL);
414 }
415 master_id = info.master_id;
416 if (master_id < -1 || master_id >= HAMMER_MAX_MASTERS)
5de0c0e5 417 return (EINVAL);
732a1697
MD
418 } else {
419 if (hmp)
420 master_id = hmp->master_id;
421 else
422 master_id = 0;
5de0c0e5 423 }
427e5fc6
MD
424
425 /*
b0aab9b9 426 * Internal mount data structure
427e5fc6 427 */
732a1697 428 if (hmp == NULL) {
195c19a1
MD
429 hmp = kmalloc(sizeof(*hmp), M_HAMMER, M_WAITOK | M_ZERO);
430 mp->mnt_data = (qaddr_t)hmp;
431 hmp->mp = mp;
46fe7ae1 432 /*TAILQ_INIT(&hmp->recycle_list);*/
47197d71 433
bac808fe 434 /*
7c457ac8
MD
435 * Make sure kmalloc type limits are set appropriately.
436 *
437 * Our inode kmalloc group is sized based on maxvnodes
438 * (controlled by the system, not us).
bac808fe
MD
439 */
440 kmalloc_create(&hmp->m_misc, "HAMMER-others");
441 kmalloc_create(&hmp->m_inodes, "HAMMER-inodes");
442
7c457ac8 443 kmalloc_raise_limit(hmp->m_inodes, 0); /* unlimited */
bac808fe 444
dd94f1b1 445 hmp->root_btree_beg.localization = 0x00000000U;
47197d71
MD
446 hmp->root_btree_beg.obj_id = -0x8000000000000000LL;
447 hmp->root_btree_beg.key = -0x8000000000000000LL;
448 hmp->root_btree_beg.create_tid = 1;
449 hmp->root_btree_beg.delete_tid = 1;
450 hmp->root_btree_beg.rec_type = 0;
451 hmp->root_btree_beg.obj_type = 0;
452
dd94f1b1 453 hmp->root_btree_end.localization = 0xFFFFFFFFU;
47197d71
MD
454 hmp->root_btree_end.obj_id = 0x7FFFFFFFFFFFFFFFLL;
455 hmp->root_btree_end.key = 0x7FFFFFFFFFFFFFFFLL;
456 hmp->root_btree_end.create_tid = 0xFFFFFFFFFFFFFFFFULL;
457 hmp->root_btree_end.delete_tid = 0; /* special case */
458 hmp->root_btree_end.rec_type = 0xFFFFU;
459 hmp->root_btree_end.obj_type = 0;
f03c9cf4 460
cdb6e4e6
MD
461 hmp->krate.freq = 1; /* maximum reporting rate (hz) */
462 hmp->krate.count = -16; /* initial burst */
463
9480ff55 464 hmp->sync_lock.refs = 1;
c9b9e29d 465 hmp->free_lock.refs = 1;
d99d6bf5
MD
466 hmp->undo_lock.refs = 1;
467 hmp->blkmap_lock.refs = 1;
83f2a3aa 468 hmp->snapshot_lock.refs = 1;
52e547e3 469 hmp->volume_lock.refs = 1;
9480ff55 470
cebe9493 471 TAILQ_INIT(&hmp->delay_list);
7a61b85d 472 TAILQ_INIT(&hmp->flush_group_list);
0729c8c8 473 TAILQ_INIT(&hmp->objid_cache_list);
e8599db1 474 TAILQ_INIT(&hmp->undo_lru_list);
7bc5b8c2 475 TAILQ_INIT(&hmp->reclaim_list);
507df98a
ID
476
477 RB_INIT(&hmp->rb_dedup_crc_root);
478 RB_INIT(&hmp->rb_dedup_off_root);
479 TAILQ_INIT(&hmp->dedup_lru_list);
195c19a1 480 }
51c35492
MD
481 hmp->hflags &= ~HMNT_USERFLAGS;
482 hmp->hflags |= info.hflags & HMNT_USERFLAGS;
732a1697
MD
483
484 hmp->master_id = master_id;
485
7f7c1f84
MD
486 if (info.asof) {
487 mp->mnt_flag |= MNT_RDONLY;
488 hmp->asof = info.asof;
489 } else {
490 hmp->asof = HAMMER_MAX_TID;
491 }
195c19a1 492
865c9609
MN
493 hmp->volume_to_remove = -1;
494
195c19a1 495 /*
51c35492 496 * Re-open read-write if originally read-only, or vise-versa.
02428fb6
MD
497 *
498 * When going from read-only to read-write execute the stage2
499 * recovery if it has not already been run.
195c19a1
MD
500 */
501 if (mp->mnt_flag & MNT_UPDATE) {
b0aab9b9 502 lwkt_gettoken(&hmp->fs_token);
51c35492
MD
503 error = 0;
504 if (hmp->ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
505 kprintf("HAMMER read-only -> read-write\n");
506 hmp->ronly = 0;
507 RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
508 hammer_adjust_volume_mode, NULL);
509 rootvol = hammer_get_root_volume(hmp, &error);
510 if (rootvol) {
06ad81ff 511 hammer_recover_flush_buffers(hmp, rootvol, 1);
02428fb6 512 error = hammer_recover_stage2(hmp, rootvol);
9f5097dc
MD
513 bcopy(rootvol->ondisk->vol0_blockmap,
514 hmp->blockmap,
515 sizeof(hmp->blockmap));
51c35492
MD
516 hammer_rel_volume(rootvol, 0);
517 }
518 RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
519 hammer_reload_inode, NULL);
520 /* kernel clears MNT_RDONLY */
521 } else if (hmp->ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
522 kprintf("HAMMER read-write -> read-only\n");
523 hmp->ronly = 1; /* messy */
524 RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
525 hammer_reload_inode, NULL);
195c19a1 526 hmp->ronly = 0;
51c35492
MD
527 hammer_flusher_sync(hmp);
528 hammer_flusher_sync(hmp);
529 hammer_flusher_sync(hmp);
530 hmp->ronly = 1;
531 RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
532 hammer_adjust_volume_mode, NULL);
195c19a1 533 }
b0aab9b9 534 lwkt_reltoken(&hmp->fs_token);
51c35492 535 return(error);
195c19a1
MD
536 }
537
427e5fc6
MD
538 RB_INIT(&hmp->rb_vols_root);
539 RB_INIT(&hmp->rb_inos_root);
73896937 540 RB_INIT(&hmp->rb_redo_root);
40043e7f 541 RB_INIT(&hmp->rb_nods_root);
e8599db1 542 RB_INIT(&hmp->rb_undo_root);
0832c9bb
MD
543 RB_INIT(&hmp->rb_resv_root);
544 RB_INIT(&hmp->rb_bufs_root);
5fa5c92f 545 RB_INIT(&hmp->rb_pfsm_root);
0832c9bb 546
195c19a1 547 hmp->ronly = ((mp->mnt_flag & MNT_RDONLY) != 0);
427e5fc6 548
10a5d1ba
MD
549 TAILQ_INIT(&hmp->volu_list);
550 TAILQ_INIT(&hmp->undo_list);
551 TAILQ_INIT(&hmp->data_list);
552 TAILQ_INIT(&hmp->meta_list);
553 TAILQ_INIT(&hmp->lose_list);
eddadaee 554 TAILQ_INIT(&hmp->iorun_list);
10a5d1ba 555
b0aab9b9
MD
556 lwkt_token_init(&hmp->fs_token, 1, "hammerfs");
557 lwkt_token_init(&hmp->io_token, 1, "hammerio");
558
559 lwkt_gettoken(&hmp->fs_token);
560
427e5fc6
MD
561 /*
562 * Load volumes
563 */
564 path = objcache_get(namei_oc, M_WAITOK);
1b0ab2c3 565 hmp->nvolumes = -1;
427e5fc6 566 for (i = 0; i < info.nvolumes; ++i) {
7c19b529
MN
567 if (mntpt == NULL) {
568 /*
569 * Root mount.
7c19b529 570 */
104cb849 571 KKASSERT(next_volume_ptr != NULL);
a407819f
MN
572 strcpy(path, "");
573 if (*next_volume_ptr != '/') {
574 /* relative path */
575 strcpy(path, "/dev/");
576 }
104cb849
MN
577 int k;
578 for (k = strlen(path); k < MAXPATHLEN-1; ++k) {
579 if (*next_volume_ptr == '\0') {
580 break;
581 } else if (*next_volume_ptr == ':') {
582 ++next_volume_ptr;
583 break;
584 } else {
585 path[k] = *next_volume_ptr;
586 ++next_volume_ptr;
587 }
588 }
589 path[k] = '\0';
590
7c19b529 591 error = 0;
104cb849
MN
592 cdev_t dev = kgetdiskbyname(path);
593 error = bdevvp(dev, &devvp);
594 if (error) {
595 kprintf("hammer_mountroot: can't find devvp\n");
596 }
7c19b529
MN
597 } else {
598 error = copyin(&info.volumes[i], &upath,
599 sizeof(char *));
600 if (error == 0)
601 error = copyinstr(upath, path,
602 MAXPATHLEN, NULL);
603 }
427e5fc6 604 if (error == 0)
7c19b529 605 error = hammer_install_volume(hmp, path, devvp);
427e5fc6
MD
606 if (error)
607 break;
608 }
609 objcache_put(namei_oc, path);
610
611 /*
612 * Make sure we found a root volume
613 */
614 if (error == 0 && hmp->rootvol == NULL) {
615 kprintf("hammer_mount: No root volume found!\n");
616 error = EINVAL;
617 }
1b0ab2c3
MD
618
619 /*
620 * Check that all required volumes are available
621 */
622 if (error == 0 && hammer_mountcheck_volumes(hmp)) {
623 kprintf("hammer_mount: Missing volumes, cannot mount!\n");
624 error = EINVAL;
625 }
626
427e5fc6 627 if (error) {
b0aab9b9 628 /* called with fs_token held */
427e5fc6
MD
629 hammer_free_hmp(mp);
630 return (error);
631 }
632
633 /*
27ea2398
MD
634 * No errors, setup enough of the mount point so we can lookup the
635 * root vnode.
427e5fc6 636 */
27ea2398
MD
637 mp->mnt_iosize_max = MAXPHYS;
638 mp->mnt_kern_flag |= MNTK_FSMID;
c0ade690 639
aac0aabd
MD
640 /*
641 * MPSAFE code. Note that VOPs and VFSops which are not MPSAFE
642 * will acquire a per-mount token prior to entry and release it
643 * on return, so even if we do not specify it we no longer get
644 * the BGL regardlless of how we are flagged.
645 */
b0aab9b9
MD
646 mp->mnt_kern_flag |= MNTK_ALL_MPSAFE;
647 /*MNTK_RD_MPSAFE | MNTK_GA_MPSAFE | MNTK_IN_MPSAFE;*/
aac0aabd 648
c0ade690
MD
649 /*
650 * note: f_iosize is used by vnode_pager_haspage() when constructing
651 * its VOP_BMAP call.
652 */
653 mp->mnt_stat.f_iosize = HAMMER_BUFSIZE;
fbc6e32a 654 mp->mnt_stat.f_bsize = HAMMER_BUFSIZE;
6f97fce3
MD
655
656 mp->mnt_vstat.f_frsize = HAMMER_BUFSIZE;
657 mp->mnt_vstat.f_bsize = HAMMER_BUFSIZE;
658
27ea2398
MD
659 mp->mnt_maxsymlinklen = 255;
660 mp->mnt_flag |= MNT_LOCAL;
661
662 vfs_add_vnodeops(mp, &hammer_vnode_vops, &mp->mnt_vn_norm_ops);
7a04d74f
MD
663 vfs_add_vnodeops(mp, &hammer_spec_vops, &mp->mnt_vn_spec_ops);
664 vfs_add_vnodeops(mp, &hammer_fifo_vops, &mp->mnt_vn_fifo_ops);
27ea2398
MD
665
666 /*
a89aec1b
MD
667 * The root volume's ondisk pointer is only valid if we hold a
668 * reference to it.
669 */
670 rootvol = hammer_get_root_volume(hmp, &error);
671 if (error)
f90dde4c
MD
672 goto failed;
673
674 /*
9f5097dc 675 * Perform any necessary UNDO operations. The recovery code does
0729c8c8
MD
676 * call hammer_undo_lookup() so we have to pre-cache the blockmap,
677 * and then re-copy it again after recovery is complete.
c9b9e29d 678 *
51c35492
MD
679 * If this is a read-only mount the UNDO information is retained
680 * in memory in the form of dirty buffer cache buffers, and not
681 * written back to the media.
f90dde4c 682 */
0729c8c8
MD
683 bcopy(rootvol->ondisk->vol0_blockmap, hmp->blockmap,
684 sizeof(hmp->blockmap));
c9b9e29d 685
7a61b85d 686 /*
44a83111
MD
687 * Check filesystem version
688 */
689 hmp->version = rootvol->ondisk->vol_version;
690 if (hmp->version < HAMMER_VOL_VERSION_MIN ||
691 hmp->version > HAMMER_VOL_VERSION_MAX) {
692 kprintf("HAMMER: mount unsupported fs version %d\n",
693 hmp->version);
694 error = ERANGE;
695 goto done;
696 }
697
698 /*
7a61b85d
MD
699 * The undo_rec_limit limits the size of flush groups to avoid
700 * blowing out the UNDO FIFO. This calculation is typically in
701 * the tens of thousands and is designed primarily when small
702 * HAMMER filesystems are created.
703 */
704 hmp->undo_rec_limit = hammer_undo_max(hmp) / 8192 + 100;
705 if (hammer_debug_general & 0x0001)
706 kprintf("HAMMER: undo_rec_limit %d\n", hmp->undo_rec_limit);
707
02428fb6
MD
708 /*
709 * NOTE: Recover stage1 not only handles meta-data recovery, it
710 * also sets hmp->undo_seqno for HAMMER VERSION 4+ filesystems.
711 */
712 error = hammer_recover_stage1(hmp, rootvol);
f90dde4c
MD
713 if (error) {
714 kprintf("Failed to recover HAMMER filesystem on mount\n");
a89aec1b 715 goto done;
f90dde4c
MD
716 }
717
718 /*
adf01747
MD
719 * Finish setup now that we have a good root volume.
720 *
721 * The top 16 bits of fsid.val[1] is a pfs id.
f90dde4c 722 */
a89aec1b
MD
723 ksnprintf(mp->mnt_stat.f_mntfromname,
724 sizeof(mp->mnt_stat.f_mntfromname), "%s",
725 rootvol->ondisk->vol_name);
513ca7d7
MD
726 mp->mnt_stat.f_fsid.val[0] =
727 crc32((char *)&rootvol->ondisk->vol_fsid + 0, 8);
728 mp->mnt_stat.f_fsid.val[1] =
729 crc32((char *)&rootvol->ondisk->vol_fsid + 8, 8);
adf01747 730 mp->mnt_stat.f_fsid.val[1] &= 0x0000FFFF;
b84de5af 731
6f97fce3
MD
732 mp->mnt_vstat.f_fsid_uuid = rootvol->ondisk->vol_fsid;
733 mp->mnt_vstat.f_fsid = crc32(&mp->mnt_vstat.f_fsid_uuid,
734 sizeof(mp->mnt_vstat.f_fsid_uuid));
735
0729c8c8
MD
736 /*
737 * Certain often-modified fields in the root volume are cached in
738 * the hammer_mount structure so we do not have to generate lots
739 * of little UNDO structures for them.
c9b9e29d 740 *
9f5097dc
MD
741 * Recopy after recovery. This also has the side effect of
742 * setting our cached undo FIFO's first_offset, which serves to
743 * placemark the FIFO start for the NEXT flush cycle while the
744 * on-disk first_offset represents the LAST flush cycle.
0729c8c8 745 */
b84de5af 746 hmp->next_tid = rootvol->ondisk->vol0_next_tid;
4889cbd4
MD
747 hmp->flush_tid1 = hmp->next_tid;
748 hmp->flush_tid2 = hmp->next_tid;
0729c8c8
MD
749 bcopy(rootvol->ondisk->vol0_blockmap, hmp->blockmap,
750 sizeof(hmp->blockmap));
e63644f0 751 hmp->copy_stat_freebigblocks = rootvol->ondisk->vol0_stat_freebigblocks;
0729c8c8 752
059819e3
MD
753 hammer_flusher_create(hmp);
754
a89aec1b 755 /*
27ea2398
MD
756 * Locate the root directory using the root cluster's B-Tree as a
757 * starting point. The root directory uses an obj_id of 1.
758 *
759 * FUTURE: Leave the root directory cached referenced but unlocked
760 * in hmp->rootvp (need to flush it on unmount).
761 */
b9b0a6d0 762 error = hammer_vfs_vget(mp, NULL, 1, &rootvp);
a89aec1b
MD
763 if (error)
764 goto done;
765 vput(rootvp);
27ea2398 766 /*vn_unlock(hmp->rootvp);*/
88c39f64
TN
767 if (hmp->ronly == 0)
768 error = hammer_recover_stage2(hmp, rootvol);
427e5fc6 769
9fa352a2
MD
770 /*
771 * If the stage2 recovery fails be sure to clean out all cached
772 * vnodes before throwing away the mount structure or bad things
773 * will happen.
774 */
775 if (error)
776 vflush(mp, 0, 0);
777
a89aec1b 778done:
f90dde4c
MD
779 hammer_rel_volume(rootvol, 0);
780failed:
427e5fc6 781 /*
27ea2398 782 * Cleanup and return.
427e5fc6 783 */
b0aab9b9
MD
784 if (error) {
785 /* called with fs_token held */
427e5fc6 786 hammer_free_hmp(mp);
b0aab9b9
MD
787 } else {
788 lwkt_reltoken(&hmp->fs_token);
789 }
427e5fc6
MD
790 return (error);
791}
792
793static int
794hammer_vfs_unmount(struct mount *mp, int mntflags)
795{
b0aab9b9 796 hammer_mount_t hmp = (void *)mp->mnt_data;
427e5fc6 797 int flags;
66325755 798 int error;
27ea2398
MD
799
800 /*
427e5fc6
MD
801 * Clean out the vnodes
802 */
b0aab9b9 803 lwkt_gettoken(&hmp->fs_token);
66325755
MD
804 flags = 0;
805 if (mntflags & MNT_FORCE)
806 flags |= FORCECLOSE;
b0aab9b9 807 error = vflush(mp, 0, flags);
427e5fc6
MD
808
809 /*
810 * Clean up the internal mount structure and related entities. This
811 * may issue I/O.
812 */
b0aab9b9
MD
813 if (error == 0) {
814 /* called with fs_token held */
815 hammer_free_hmp(mp);
816 } else {
817 lwkt_reltoken(&hmp->fs_token);
818 }
819 return(error);
427e5fc6
MD
820}
821
822/*
823 * Clean up the internal mount structure and disassociate it from the mount.
824 * This may issue I/O.
b0aab9b9
MD
825 *
826 * Called with fs_token held.
427e5fc6
MD
827 */
828static void
829hammer_free_hmp(struct mount *mp)
830{
b0aab9b9 831 hammer_mount_t hmp = (void *)mp->mnt_data;
cdb6e4e6 832 hammer_flush_group_t flg;
1b0ab2c3 833 int count;
d40bfeca 834 int dummy;
427e5fc6
MD
835
836 /*
cdb6e4e6
MD
837 * Flush anything dirty. This won't even run if the
838 * filesystem errored-out.
427e5fc6 839 */
1b0ab2c3
MD
840 count = 0;
841 while (hammer_flusher_haswork(hmp)) {
842 hammer_flusher_sync(hmp);
843 ++count;
844 if (count >= 5) {
845 if (count == 5)
846 kprintf("HAMMER: umount flushing.");
847 else
848 kprintf(".");
d40bfeca 849 tsleep(&dummy, 0, "hmrufl", hz);
1b0ab2c3
MD
850 }
851 if (count == 30) {
852 kprintf("giving up\n");
853 break;
854 }
855 }
856 if (count >= 5 && count < 30)
857 kprintf("\n");
427e5fc6 858
cdb6e4e6
MD
859 /*
860 * If the mount had a critical error we have to destroy any
861 * remaining inodes before we can finish cleaning up the flusher.
862 */
863 if (hmp->flags & HAMMER_MOUNT_CRITICAL_ERROR) {
864 RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
865 hammer_destroy_inode_callback, NULL);
866 }
b84de5af 867
427e5fc6 868 /*
cdb6e4e6
MD
869 * There shouldn't be any inodes left now and any left over
870 * flush groups should now be empty.
427e5fc6 871 */
cdb6e4e6
MD
872 KKASSERT(RB_EMPTY(&hmp->rb_inos_root));
873 while ((flg = TAILQ_FIRST(&hmp->flush_group_list)) != NULL) {
874 TAILQ_REMOVE(&hmp->flush_group_list, flg, flush_entry);
ff003b11 875 KKASSERT(RB_EMPTY(&flg->flush_tree));
cdb6e4e6
MD
876 if (flg->refs) {
877 kprintf("HAMMER: Warning, flush_group %p was "
878 "not empty on umount!\n", flg);
879 }
bac808fe 880 kfree(flg, hmp->m_misc);
cdb6e4e6 881 }
427e5fc6
MD
882
883 /*
cdb6e4e6 884 * We can finally destroy the flusher
427e5fc6 885 */
cdb6e4e6
MD
886 hammer_flusher_destroy(hmp);
887
b84de5af 888 /*
00f16fad
MD
889 * We may have held recovered buffers due to a read-only mount.
890 * These must be discarded.
891 */
892 if (hmp->ronly)
893 hammer_recover_flush_buffers(hmp, NULL, -1);
894
895 /*
0832c9bb 896 * Unload buffers and then volumes
b84de5af 897 */
0832c9bb
MD
898 RB_SCAN(hammer_buf_rb_tree, &hmp->rb_bufs_root, NULL,
899 hammer_unload_buffer, NULL);
427e5fc6
MD
900 RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
901 hammer_unload_volume, NULL);
902
903 mp->mnt_data = NULL;
66325755 904 mp->mnt_flag &= ~MNT_LOCAL;
427e5fc6 905 hmp->mp = NULL;
0729c8c8 906 hammer_destroy_objid_cache(hmp);
507df98a
ID
907 hammer_destroy_dedup_cache(hmp);
908 if (hmp->dedup_free_cache != NULL) {
909 kfree(hmp->dedup_free_cache, hmp->m_misc);
910 hmp->dedup_free_cache = NULL;
911 }
bac808fe
MD
912 kmalloc_destroy(&hmp->m_misc);
913 kmalloc_destroy(&hmp->m_inodes);
b0aab9b9 914 lwkt_reltoken(&hmp->fs_token);
427e5fc6
MD
915 kfree(hmp, M_HAMMER);
916}
917
918/*
cdb6e4e6
MD
919 * Report critical errors. ip may be NULL.
920 */
921void
922hammer_critical_error(hammer_mount_t hmp, hammer_inode_t ip,
923 int error, const char *msg)
924{
925 hmp->flags |= HAMMER_MOUNT_CRITICAL_ERROR;
c9ce54d6 926
cdb6e4e6 927 krateprintf(&hmp->krate,
c9ce54d6
MD
928 "HAMMER(%s): Critical error inode=%jd error=%d %s\n",
929 hmp->mp->mnt_stat.f_mntfromname,
930 (intmax_t)(ip ? ip->obj_id : -1),
931 error, msg);
932
cdb6e4e6
MD
933 if (hmp->ronly == 0) {
934 hmp->ronly = 2; /* special errored read-only mode */
935 hmp->mp->mnt_flag |= MNT_RDONLY;
936 kprintf("HAMMER(%s): Forcing read-only mode\n",
937 hmp->mp->mnt_stat.f_mntfromname);
938 }
939 hmp->error = error;
fc73edd8 940 if (hammer_debug_critical)
c9ce54d6 941 Debugger("Entering debugger");
cdb6e4e6
MD
942}
943
944
945/*
513ca7d7
MD
946 * Obtain a vnode for the specified inode number. An exclusively locked
947 * vnode is returned.
948 */
949int
b9b0a6d0
MD
950hammer_vfs_vget(struct mount *mp, struct vnode *dvp,
951 ino_t ino, struct vnode **vpp)
513ca7d7 952{
36f82b23 953 struct hammer_transaction trans;
513ca7d7
MD
954 struct hammer_mount *hmp = (void *)mp->mnt_data;
955 struct hammer_inode *ip;
956 int error;
b9b0a6d0 957 u_int32_t localization;
513ca7d7 958
b0aab9b9 959 lwkt_gettoken(&hmp->fs_token);
36f82b23
MD
960 hammer_simple_transaction(&trans, hmp);
961
513ca7d7 962 /*
b9b0a6d0
MD
963 * If a directory vnode is supplied (mainly NFS) then we can acquire
964 * the PFS domain from it. Otherwise we would only be able to vget
965 * inodes in the root PFS.
966 */
967 if (dvp) {
968 localization = HAMMER_DEF_LOCALIZATION +
969 VTOI(dvp)->obj_localization;
970 } else {
971 localization = HAMMER_DEF_LOCALIZATION;
972 }
973
974 /*
36f82b23
MD
975 * Lookup the requested HAMMER inode. The structure must be
976 * left unlocked while we manipulate the related vnode to avoid
977 * a deadlock.
513ca7d7 978 */
ddfdf542 979 ip = hammer_get_inode(&trans, NULL, ino,
b9b0a6d0 980 hmp->asof, localization,
ddfdf542 981 0, &error);
513ca7d7
MD
982 if (ip == NULL) {
983 *vpp = NULL;
b0aab9b9
MD
984 } else {
985 error = hammer_get_vnode(ip, vpp);
986 hammer_rel_inode(ip, 0);
513ca7d7 987 }
b84de5af 988 hammer_done_transaction(&trans);
b0aab9b9 989 lwkt_reltoken(&hmp->fs_token);
513ca7d7
MD
990 return (error);
991}
992
993/*
427e5fc6
MD
994 * Return the root vnode for the filesystem.
995 *
996 * HAMMER stores the root vnode in the hammer_mount structure so
997 * getting it is easy.
998 */
999static int
1000hammer_vfs_root(struct mount *mp, struct vnode **vpp)
1001{
27ea2398 1002 int error;
427e5fc6 1003
b9b0a6d0 1004 error = hammer_vfs_vget(mp, NULL, 1, vpp);
27ea2398 1005 return (error);
427e5fc6
MD
1006}
1007
1008static int
1009hammer_vfs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
1010{
fbc6e32a
MD
1011 struct hammer_mount *hmp = (void *)mp->mnt_data;
1012 hammer_volume_t volume;
1013 hammer_volume_ondisk_t ondisk;
1014 int error;
47197d71 1015 int64_t bfree;
31a56ce2 1016 int64_t breserved;
fbc6e32a 1017
b0aab9b9 1018 lwkt_gettoken(&hmp->fs_token);
fbc6e32a 1019 volume = hammer_get_root_volume(hmp, &error);
b0aab9b9
MD
1020 if (error) {
1021 lwkt_reltoken(&hmp->fs_token);
fbc6e32a 1022 return(error);
b0aab9b9 1023 }
fbc6e32a
MD
1024 ondisk = volume->ondisk;
1025
47197d71
MD
1026 /*
1027 * Basic stats
1028 */
31a56ce2 1029 _hammer_checkspace(hmp, HAMMER_CHKSPC_WRITE, &breserved);
fbc6e32a 1030 mp->mnt_stat.f_files = ondisk->vol0_stat_inodes;
c3be93f2 1031 bfree = ondisk->vol0_stat_freebigblocks * HAMMER_LARGEBLOCK_SIZE;
40043e7f 1032 hammer_rel_volume(volume, 0);
47197d71 1033
31a56ce2 1034 mp->mnt_stat.f_bfree = (bfree - breserved) / HAMMER_BUFSIZE;
47197d71 1035 mp->mnt_stat.f_bavail = mp->mnt_stat.f_bfree;
fbc6e32a
MD
1036 if (mp->mnt_stat.f_files < 0)
1037 mp->mnt_stat.f_files = 0;
1038
27ea2398 1039 *sbp = mp->mnt_stat;
b0aab9b9 1040 lwkt_reltoken(&hmp->fs_token);
27ea2398 1041 return(0);
427e5fc6
MD
1042}
1043
6f97fce3
MD
1044static int
1045hammer_vfs_statvfs(struct mount *mp, struct statvfs *sbp, struct ucred *cred)
1046{
1047 struct hammer_mount *hmp = (void *)mp->mnt_data;
1048 hammer_volume_t volume;
1049 hammer_volume_ondisk_t ondisk;
1050 int error;
1051 int64_t bfree;
0f65be10 1052 int64_t breserved;
6f97fce3 1053
b0aab9b9 1054 lwkt_gettoken(&hmp->fs_token);
6f97fce3 1055 volume = hammer_get_root_volume(hmp, &error);
b0aab9b9
MD
1056 if (error) {
1057 lwkt_reltoken(&hmp->fs_token);
6f97fce3 1058 return(error);
b0aab9b9 1059 }
6f97fce3
MD
1060 ondisk = volume->ondisk;
1061
1062 /*
1063 * Basic stats
1064 */
0f65be10 1065 _hammer_checkspace(hmp, HAMMER_CHKSPC_WRITE, &breserved);
6f97fce3
MD
1066 mp->mnt_vstat.f_files = ondisk->vol0_stat_inodes;
1067 bfree = ondisk->vol0_stat_freebigblocks * HAMMER_LARGEBLOCK_SIZE;
1068 hammer_rel_volume(volume, 0);
1069
0f65be10 1070 mp->mnt_vstat.f_bfree = (bfree - breserved) / HAMMER_BUFSIZE;
c0763659 1071 mp->mnt_vstat.f_bavail = mp->mnt_vstat.f_bfree;
6f97fce3
MD
1072 if (mp->mnt_vstat.f_files < 0)
1073 mp->mnt_vstat.f_files = 0;
1074 *sbp = mp->mnt_vstat;
b0aab9b9 1075 lwkt_reltoken(&hmp->fs_token);
6f97fce3
MD
1076 return(0);
1077}
1078
0729c8c8
MD
1079/*
1080 * Sync the filesystem. Currently we have to run it twice, the second
1081 * one will advance the undo start index to the end index, so if a crash
1082 * occurs no undos will be run on mount.
77062c8a
MD
1083 *
1084 * We do not sync the filesystem if we are called from a panic. If we did
1085 * we might end up blowing up a sync that was already in progress.
0729c8c8 1086 */
427e5fc6
MD
1087static int
1088hammer_vfs_sync(struct mount *mp, int waitfor)
1089{
fbc6e32a 1090 struct hammer_mount *hmp = (void *)mp->mnt_data;
0729c8c8
MD
1091 int error;
1092
b0aab9b9 1093 lwkt_gettoken(&hmp->fs_token);
77062c8a 1094 if (panicstr == NULL) {
0729c8c8 1095 error = hammer_sync_hmp(hmp, waitfor);
77062c8a
MD
1096 } else {
1097 error = EIO;
77062c8a 1098 }
b0aab9b9 1099 lwkt_reltoken(&hmp->fs_token);
0729c8c8 1100 return (error);
427e5fc6
MD
1101}
1102
513ca7d7
MD
1103/*
1104 * Convert a vnode to a file handle.
b0aab9b9
MD
1105 *
1106 * Accesses read-only fields on already-referenced structures so
1107 * no token is needed.
513ca7d7
MD
1108 */
1109static int
1110hammer_vfs_vptofh(struct vnode *vp, struct fid *fhp)
1111{
1112 hammer_inode_t ip;
1113
1114 KKASSERT(MAXFIDSZ >= 16);
1115 ip = VTOI(vp);
1116 fhp->fid_len = offsetof(struct fid, fid_data[16]);
adf01747 1117 fhp->fid_ext = ip->obj_localization >> 16;
513ca7d7
MD
1118 bcopy(&ip->obj_id, fhp->fid_data + 0, sizeof(ip->obj_id));
1119 bcopy(&ip->obj_asof, fhp->fid_data + 8, sizeof(ip->obj_asof));
1120 return(0);
1121}
1122
1123
1124/*
1125 * Convert a file handle back to a vnode.
67863d04
MD
1126 *
1127 * Use rootvp to enforce PFS isolation when a PFS is exported via a
1128 * null mount.
513ca7d7
MD
1129 */
1130static int
67863d04
MD
1131hammer_vfs_fhtovp(struct mount *mp, struct vnode *rootvp,
1132 struct fid *fhp, struct vnode **vpp)
513ca7d7 1133{
b0aab9b9 1134 hammer_mount_t hmp = (void *)mp->mnt_data;
36f82b23 1135 struct hammer_transaction trans;
513ca7d7
MD
1136 struct hammer_inode *ip;
1137 struct hammer_inode_info info;
1138 int error;
adf01747 1139 u_int32_t localization;
513ca7d7
MD
1140
1141 bcopy(fhp->fid_data + 0, &info.obj_id, sizeof(info.obj_id));
1142 bcopy(fhp->fid_data + 8, &info.obj_asof, sizeof(info.obj_asof));
67863d04
MD
1143 if (rootvp)
1144 localization = VTOI(rootvp)->obj_localization;
1145 else
1146 localization = (u_int32_t)fhp->fid_ext << 16;
513ca7d7 1147
b0aab9b9
MD
1148 lwkt_gettoken(&hmp->fs_token);
1149 hammer_simple_transaction(&trans, hmp);
36f82b23 1150
513ca7d7
MD
1151 /*
1152 * Get/allocate the hammer_inode structure. The structure must be
1153 * unlocked while we manipulate the related vnode to avoid a
1154 * deadlock.
1155 */
ddfdf542 1156 ip = hammer_get_inode(&trans, NULL, info.obj_id,
adf01747 1157 info.obj_asof, localization, 0, &error);
d978e7cf
MD
1158 if (ip) {
1159 error = hammer_get_vnode(ip, vpp);
1160 hammer_rel_inode(ip, 0);
1161 } else {
513ca7d7 1162 *vpp = NULL;
513ca7d7 1163 }
b84de5af 1164 hammer_done_transaction(&trans);
b0aab9b9 1165 lwkt_reltoken(&hmp->fs_token);
513ca7d7
MD
1166 return (error);
1167}
1168
1169static int
1170hammer_vfs_checkexp(struct mount *mp, struct sockaddr *nam,
1171 int *exflagsp, struct ucred **credanonp)
1172{
1173 hammer_mount_t hmp = (void *)mp->mnt_data;
1174 struct netcred *np;
1175 int error;
1176
b0aab9b9 1177 lwkt_gettoken(&hmp->fs_token);
513ca7d7
MD
1178 np = vfs_export_lookup(mp, &hmp->export, nam);
1179 if (np) {
1180 *exflagsp = np->netc_exflags;
1181 *credanonp = &np->netc_anon;
1182 error = 0;
1183 } else {
1184 error = EACCES;
1185 }
b0aab9b9 1186 lwkt_reltoken(&hmp->fs_token);
513ca7d7
MD
1187 return (error);
1188
1189}
1190
1191int
1192hammer_vfs_export(struct mount *mp, int op, const struct export_args *export)
1193{
1194 hammer_mount_t hmp = (void *)mp->mnt_data;
1195 int error;
1196
b0aab9b9
MD
1197 lwkt_gettoken(&hmp->fs_token);
1198
513ca7d7
MD
1199 switch(op) {
1200 case MOUNTCTL_SET_EXPORT:
1201 error = vfs_export(mp, &hmp->export, export);
1202 break;
1203 default:
1204 error = EOPNOTSUPP;
1205 break;
1206 }
b0aab9b9
MD
1207 lwkt_reltoken(&hmp->fs_token);
1208
513ca7d7
MD
1209 return(error);
1210}
1211