HAMMER VFS: record the name of the mount directory
[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.
427e5fc6
MD
33 */
34
35#include <sys/param.h>
36#include <sys/systm.h>
37#include <sys/kernel.h>
38#include <sys/vnode.h>
39#include <sys/mount.h>
40#include <sys/malloc.h>
41#include <sys/nlookup.h>
42#include <sys/fcntl.h>
b3deaf57 43#include <sys/sysctl.h>
427e5fc6
MD
44#include <sys/buf.h>
45#include <sys/buf2.h>
46#include "hammer.h"
47
aac0aabd
MD
48/*
49 * NOTE! Global statistics may not be MPSAFE so HAMMER never uses them
50 * in conditionals.
51 */
02428fb6 52int hammer_supported_version = HAMMER_VOL_VERSION_DEFAULT;
2f85fa4d 53int hammer_debug_io;
d5ef456e 54int hammer_debug_general;
1b0ab2c3 55int hammer_debug_debug = 1; /* medium-error panics */
e8599db1 56int hammer_debug_inode;
7d683b0f 57int hammer_debug_locks;
b3deaf57 58int hammer_debug_btree;
d113fda1 59int hammer_debug_tid;
cebe9493 60int hammer_debug_recover; /* -1 will disable, +1 will force */
46fe7ae1 61int hammer_debug_recover_faults;
fc73edd8 62int hammer_debug_critical; /* non-zero enter debugger on error */
1b0ab2c3 63int hammer_cluster_enable = 1; /* enable read clustering by default */
507df98a 64int hammer_live_dedup = 0;
e2a02b72 65int hammer_tdmux_ticks;
7a61b85d 66int hammer_count_fsyncs;
b3deaf57 67int hammer_count_inodes;
af209b0f 68int hammer_count_iqueued;
e2a02b72 69int hammer_count_reclaims;
b3deaf57
MD
70int hammer_count_records;
71int hammer_count_record_datas;
72int hammer_count_volumes;
b3deaf57
MD
73int hammer_count_buffers;
74int hammer_count_nodes;
a7e9bef1 75int64_t hammer_count_extra_space_used;
cb51be26
MD
76int64_t hammer_stats_btree_lookups;
77int64_t hammer_stats_btree_searches;
78int64_t hammer_stats_btree_inserts;
79int64_t hammer_stats_btree_deletes;
80int64_t hammer_stats_btree_elements;
81int64_t hammer_stats_btree_splits;
82int64_t hammer_stats_btree_iterations;
39d8fd63 83int64_t hammer_stats_btree_root_iterations;
cb51be26 84int64_t hammer_stats_record_iterations;
ce0138a6
MD
85
86int64_t hammer_stats_file_read;
87int64_t hammer_stats_file_write;
88int64_t hammer_stats_file_iopsr;
89int64_t hammer_stats_file_iopsw;
90int64_t hammer_stats_disk_read;
91int64_t hammer_stats_disk_write;
92int64_t hammer_stats_inode_flushes;
93int64_t hammer_stats_commits;
89e744ce 94int64_t hammer_stats_undo;
6048b411 95int64_t hammer_stats_redo;
ce0138a6 96
3583bbb4 97long hammer_count_dirtybufspace; /* global */
a99b9ea2 98int hammer_count_refedbufs; /* global */
0832c9bb 99int hammer_count_reservations;
3583bbb4
MD
100long hammer_count_io_running_read;
101long hammer_count_io_running_write;
a99b9ea2 102int hammer_count_io_locked;
3583bbb4 103long hammer_limit_dirtybufspace; /* per-mount */
0832c9bb 104int hammer_limit_recs; /* as a whole XXX */
a117fbeb 105int hammer_limit_inode_recs = 2048; /* per inode */
e2a02b72 106int hammer_limit_reclaims;
e2ef7a95 107int hammer_live_dedup_cache_size = DEDUP_CACHE_SIZE;
6048b411 108int hammer_limit_redo = 4096 * 1024; /* per inode */
37646115 109int hammer_autoflush = 500; /* auto flush (typ on reclaim) */
1f07f686 110int hammer_bio_count;
cb51be26 111int hammer_verify_zone;
1b0ab2c3 112int hammer_verify_data = 1;
cb51be26 113int hammer_write_mode;
b4f86ea3 114int hammer_double_buffer;
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,
f06b3d6d 162 &hammer_live_dedup, 0, "Enable live dedup");
e2a02b72
MD
163SYSCTL_INT(_vfs_hammer, OID_AUTO, tdmux_ticks, CTLFLAG_RW,
164 &hammer_tdmux_ticks, 0, "Hammer tdmux ticks");
9480ff55 165
3583bbb4 166SYSCTL_LONG(_vfs_hammer, OID_AUTO, limit_dirtybufspace, CTLFLAG_RW,
f5a07a7a 167 &hammer_limit_dirtybufspace, 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, "");
e2a02b72
MD
172SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_reclaims, CTLFLAG_RW,
173 &hammer_limit_reclaims, 0, "");
e2ef7a95 174SYSCTL_INT(_vfs_hammer, OID_AUTO, live_dedup_cache_size, CTLFLAG_RW,
2b34db9b
SG
175 &hammer_live_dedup_cache_size, 0,
176 "Number of cache entries");
6048b411
MD
177SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_redo, CTLFLAG_RW,
178 &hammer_limit_redo, 0, "");
9480ff55 179
7a61b85d
MD
180SYSCTL_INT(_vfs_hammer, OID_AUTO, count_fsyncs, CTLFLAG_RD,
181 &hammer_count_fsyncs, 0, "");
b3deaf57
MD
182SYSCTL_INT(_vfs_hammer, OID_AUTO, count_inodes, CTLFLAG_RD,
183 &hammer_count_inodes, 0, "");
af209b0f
MD
184SYSCTL_INT(_vfs_hammer, OID_AUTO, count_iqueued, CTLFLAG_RD,
185 &hammer_count_iqueued, 0, "");
e2a02b72
MD
186SYSCTL_INT(_vfs_hammer, OID_AUTO, count_reclaims, CTLFLAG_RD,
187 &hammer_count_reclaims, 0, "");
b3deaf57
MD
188SYSCTL_INT(_vfs_hammer, OID_AUTO, count_records, CTLFLAG_RD,
189 &hammer_count_records, 0, "");
190SYSCTL_INT(_vfs_hammer, OID_AUTO, count_record_datas, CTLFLAG_RD,
191 &hammer_count_record_datas, 0, "");
192SYSCTL_INT(_vfs_hammer, OID_AUTO, count_volumes, CTLFLAG_RD,
193 &hammer_count_volumes, 0, "");
b3deaf57
MD
194SYSCTL_INT(_vfs_hammer, OID_AUTO, count_buffers, CTLFLAG_RD,
195 &hammer_count_buffers, 0, "");
196SYSCTL_INT(_vfs_hammer, OID_AUTO, count_nodes, CTLFLAG_RD,
197 &hammer_count_nodes, 0, "");
a7e9bef1
MD
198SYSCTL_QUAD(_vfs_hammer, OID_AUTO, count_extra_space_used, CTLFLAG_RD,
199 &hammer_count_extra_space_used, 0, "");
ce0138a6 200
cb51be26
MD
201SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_searches, CTLFLAG_RD,
202 &hammer_stats_btree_searches, 0, "");
203SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_lookups, CTLFLAG_RD,
204 &hammer_stats_btree_lookups, 0, "");
205SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_inserts, CTLFLAG_RD,
206 &hammer_stats_btree_inserts, 0, "");
207SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_deletes, CTLFLAG_RD,
208 &hammer_stats_btree_deletes, 0, "");
209SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_elements, CTLFLAG_RD,
210 &hammer_stats_btree_elements, 0, "");
211SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_splits, CTLFLAG_RD,
212 &hammer_stats_btree_splits, 0, "");
213SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_iterations, CTLFLAG_RD,
214 &hammer_stats_btree_iterations, 0, "");
39d8fd63
MD
215SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_root_iterations, CTLFLAG_RD,
216 &hammer_stats_btree_root_iterations, 0, "");
cb51be26
MD
217SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_record_iterations, CTLFLAG_RD,
218 &hammer_stats_record_iterations, 0, "");
ce0138a6
MD
219
220SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_read, CTLFLAG_RD,
221 &hammer_stats_file_read, 0, "");
222SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_write, CTLFLAG_RD,
223 &hammer_stats_file_write, 0, "");
224SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_iopsr, CTLFLAG_RD,
225 &hammer_stats_file_iopsr, 0, "");
226SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_iopsw, CTLFLAG_RD,
227 &hammer_stats_file_iopsw, 0, "");
228SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_disk_read, CTLFLAG_RD,
229 &hammer_stats_disk_read, 0, "");
230SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_disk_write, CTLFLAG_RD,
231 &hammer_stats_disk_write, 0, "");
232SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_inode_flushes, CTLFLAG_RD,
233 &hammer_stats_inode_flushes, 0, "");
234SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_commits, CTLFLAG_RD,
235 &hammer_stats_commits, 0, "");
89e744ce
MD
236SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_undo, CTLFLAG_RD,
237 &hammer_stats_undo, 0, "");
6048b411
MD
238SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_redo, CTLFLAG_RD,
239 &hammer_stats_redo, 0, "");
ce0138a6 240
507df98a 241SYSCTL_QUAD(_vfs_hammer, OID_AUTO, live_dedup_vnode_bcmps, CTLFLAG_RW,
f06b3d6d
SG
242 &hammer_live_dedup_vnode_bcmps, 0,
243 "successful vnode buffer comparisons");
507df98a 244SYSCTL_QUAD(_vfs_hammer, OID_AUTO, live_dedup_device_bcmps, CTLFLAG_RW,
f06b3d6d
SG
245 &hammer_live_dedup_device_bcmps, 0,
246 "successful device buffer comparisons");
507df98a 247SYSCTL_QUAD(_vfs_hammer, OID_AUTO, live_dedup_findblk_failures, CTLFLAG_RW,
f06b3d6d
SG
248 &hammer_live_dedup_findblk_failures, 0,
249 "block lookup failures for comparison");
507df98a 250SYSCTL_QUAD(_vfs_hammer, OID_AUTO, live_dedup_bmap_saves, CTLFLAG_RW,
f06b3d6d
SG
251 &hammer_live_dedup_bmap_saves, 0,
252 "useful physical block lookups");
507df98a 253
3583bbb4 254SYSCTL_LONG(_vfs_hammer, OID_AUTO, count_dirtybufspace, CTLFLAG_RD,
f5a07a7a 255 &hammer_count_dirtybufspace, 0, "");
a99b9ea2
MD
256SYSCTL_INT(_vfs_hammer, OID_AUTO, count_refedbufs, CTLFLAG_RD,
257 &hammer_count_refedbufs, 0, "");
0832c9bb
MD
258SYSCTL_INT(_vfs_hammer, OID_AUTO, count_reservations, CTLFLAG_RD,
259 &hammer_count_reservations, 0, "");
3583bbb4 260SYSCTL_LONG(_vfs_hammer, OID_AUTO, count_io_running_read, CTLFLAG_RD,
a99b9ea2
MD
261 &hammer_count_io_running_read, 0, "");
262SYSCTL_INT(_vfs_hammer, OID_AUTO, count_io_locked, CTLFLAG_RD,
263 &hammer_count_io_locked, 0, "");
3583bbb4 264SYSCTL_LONG(_vfs_hammer, OID_AUTO, count_io_running_write, CTLFLAG_RD,
a99b9ea2 265 &hammer_count_io_running_write, 0, "");
f03c9cf4
MD
266SYSCTL_QUAD(_vfs_hammer, OID_AUTO, zone_limit, CTLFLAG_RW,
267 &hammer_zone_limit, 0, "");
7d683b0f
MD
268SYSCTL_QUAD(_vfs_hammer, OID_AUTO, contention_count, CTLFLAG_RW,
269 &hammer_contention_count, 0, "");
21fde338
MD
270SYSCTL_INT(_vfs_hammer, OID_AUTO, autoflush, CTLFLAG_RW,
271 &hammer_autoflush, 0, "");
cb51be26
MD
272SYSCTL_INT(_vfs_hammer, OID_AUTO, verify_zone, CTLFLAG_RW,
273 &hammer_verify_zone, 0, "");
1b0ab2c3
MD
274SYSCTL_INT(_vfs_hammer, OID_AUTO, verify_data, CTLFLAG_RW,
275 &hammer_verify_data, 0, "");
cb51be26
MD
276SYSCTL_INT(_vfs_hammer, OID_AUTO, write_mode, CTLFLAG_RW,
277 &hammer_write_mode, 0, "");
b4f86ea3
MD
278SYSCTL_INT(_vfs_hammer, OID_AUTO, double_buffer, CTLFLAG_RW,
279 &hammer_double_buffer, 0, "");
3e583440
MD
280SYSCTL_INT(_vfs_hammer, OID_AUTO, yield_check, CTLFLAG_RW,
281 &hammer_yield_check, 0, "");
6f3d87c0
MD
282SYSCTL_INT(_vfs_hammer, OID_AUTO, fsync_mode, CTLFLAG_RW,
283 &hammer_fsync_mode, 0, "");
b3deaf57 284
879a1b60 285/* KTR_INFO_MASTER(hammer); */
bcac4bbb 286
427e5fc6
MD
287/*
288 * VFS ABI
289 */
290static void hammer_free_hmp(struct mount *mp);
291
292static int hammer_vfs_mount(struct mount *mp, char *path, caddr_t data,
293 struct ucred *cred);
294static int hammer_vfs_unmount(struct mount *mp, int mntflags);
295static int hammer_vfs_root(struct mount *mp, struct vnode **vpp);
296static int hammer_vfs_statfs(struct mount *mp, struct statfs *sbp,
297 struct ucred *cred);
6f97fce3
MD
298static int hammer_vfs_statvfs(struct mount *mp, struct statvfs *sbp,
299 struct ucred *cred);
427e5fc6 300static int hammer_vfs_sync(struct mount *mp, int waitfor);
b9b0a6d0
MD
301static int hammer_vfs_vget(struct mount *mp, struct vnode *dvp,
302 ino_t ino, struct vnode **vpp);
427e5fc6 303static int hammer_vfs_init(struct vfsconf *conf);
67863d04
MD
304static int hammer_vfs_fhtovp(struct mount *mp, struct vnode *rootvp,
305 struct fid *fhp, struct vnode **vpp);
513ca7d7
MD
306static int hammer_vfs_vptofh(struct vnode *vp, struct fid *fhp);
307static int hammer_vfs_checkexp(struct mount *mp, struct sockaddr *nam,
308 int *exflagsp, struct ucred **credanonp);
309
427e5fc6
MD
310
311static struct vfsops hammer_vfsops = {
312 .vfs_mount = hammer_vfs_mount,
313 .vfs_unmount = hammer_vfs_unmount,
314 .vfs_root = hammer_vfs_root,
315 .vfs_statfs = hammer_vfs_statfs,
6f97fce3 316 .vfs_statvfs = hammer_vfs_statvfs,
427e5fc6
MD
317 .vfs_sync = hammer_vfs_sync,
318 .vfs_vget = hammer_vfs_vget,
513ca7d7
MD
319 .vfs_init = hammer_vfs_init,
320 .vfs_vptofh = hammer_vfs_vptofh,
321 .vfs_fhtovp = hammer_vfs_fhtovp,
322 .vfs_checkexp = hammer_vfs_checkexp
427e5fc6
MD
323};
324
bac808fe 325MALLOC_DEFINE(M_HAMMER, "HAMMER-mount", "");
427e5fc6
MD
326
327VFS_SET(hammer_vfsops, hammer, 0);
328MODULE_VERSION(hammer, 1);
329
330static int
331hammer_vfs_init(struct vfsconf *conf)
332{
3098dc2f
MD
333 int n;
334
e2a02b72
MD
335 /*
336 * Wait up to this long for an exclusive deadlock to clear
337 * before acquiring a new shared lock on the ip. The deadlock
338 * may have occured on a b-tree node related to the ip.
339 */
340 if (hammer_tdmux_ticks == 0)
341 hammer_tdmux_ticks = hz / 5;
342
343 /*
344 * Autosize
345 */
3098dc2f 346 if (hammer_limit_recs == 0) {
a99b9ea2 347 hammer_limit_recs = nbuf * 25;
3098dc2f
MD
348 n = kmalloc_limit(M_HAMMER) / 512;
349 if (hammer_limit_recs > n)
350 hammer_limit_recs = n;
351 }
f5a07a7a
MD
352 if (hammer_limit_dirtybufspace == 0) {
353 hammer_limit_dirtybufspace = hidirtybufspace / 2;
354 if (hammer_limit_dirtybufspace < 100)
355 hammer_limit_dirtybufspace = 100;
9f5097dc 356 }
ba298df1
MD
357
358 /*
37646115
MD
359 * The hammer_inode structure detaches from the vnode on reclaim.
360 * This limits the number of inodes in this state to prevent a
361 * memory pool blowout.
362 */
e2a02b72
MD
363 if (hammer_limit_reclaims == 0)
364 hammer_limit_reclaims = desiredvnodes / 10;
37646115 365
427e5fc6
MD
366 return(0);
367}
368
369static int
370hammer_vfs_mount(struct mount *mp, char *mntpt, caddr_t data,
371 struct ucred *cred)
372{
373 struct hammer_mount_info info;
a89aec1b
MD
374 hammer_mount_t hmp;
375 hammer_volume_t rootvol;
27ea2398 376 struct vnode *rootvp;
7c19b529 377 struct vnode *devvp = NULL;
427e5fc6
MD
378 const char *upath; /* volume name in userspace */
379 char *path; /* volume name in system space */
380 int error;
381 int i;
732a1697 382 int master_id;
104cb849 383 char *next_volume_ptr = NULL;
ab0bf4a3
MD
384
385 /*
386 * Accept hammer_mount_info. mntpt is NULL for root mounts at boot.
387 */
7c19b529 388 if (mntpt == NULL) {
7c19b529
MN
389 bzero(&info, sizeof(info));
390 info.asof = 0;
391 info.hflags = 0;
392 info.nvolumes = 1;
104cb849
MN
393
394 next_volume_ptr = mp->mnt_stat.f_mntfromname;
395
396 /* Count number of volumes separated by ':' */
397 for (char *p = next_volume_ptr; *p != '\0'; ++p) {
398 if (*p == ':') {
399 ++info.nvolumes;
400 }
401 }
402
403 mp->mnt_flag &= ~MNT_RDONLY; /* mount R/W */
7c19b529
MN
404 } else {
405 if ((error = copyin(data, &info, sizeof(info))) != 0)
406 return (error);
407 }
732a1697
MD
408
409 /*
410 * updating or new mount
411 */
412 if (mp->mnt_flag & MNT_UPDATE) {
413 hmp = (void *)mp->mnt_data;
414 KKASSERT(hmp != NULL);
415 } else {
51c35492
MD
416 if (info.nvolumes <= 0 || info.nvolumes >= 32768)
417 return (EINVAL);
732a1697 418 hmp = NULL;
51c35492 419 }
732a1697
MD
420
421 /*
422 * master-id validation. The master id may not be changed by a
423 * mount update.
424 */
425 if (info.hflags & HMNT_MASTERID) {
426 if (hmp && hmp->master_id != info.master_id) {
427 kprintf("hammer: cannot change master id "
428 "with mount update\n");
429 return(EINVAL);
430 }
431 master_id = info.master_id;
432 if (master_id < -1 || master_id >= HAMMER_MAX_MASTERS)
5de0c0e5 433 return (EINVAL);
732a1697
MD
434 } else {
435 if (hmp)
436 master_id = hmp->master_id;
437 else
438 master_id = 0;
5de0c0e5 439 }
427e5fc6
MD
440
441 /*
b0aab9b9 442 * Internal mount data structure
427e5fc6 443 */
732a1697 444 if (hmp == NULL) {
195c19a1
MD
445 hmp = kmalloc(sizeof(*hmp), M_HAMMER, M_WAITOK | M_ZERO);
446 mp->mnt_data = (qaddr_t)hmp;
447 hmp->mp = mp;
46fe7ae1 448 /*TAILQ_INIT(&hmp->recycle_list);*/
47197d71 449
bac808fe 450 /*
7c457ac8
MD
451 * Make sure kmalloc type limits are set appropriately.
452 *
453 * Our inode kmalloc group is sized based on maxvnodes
454 * (controlled by the system, not us).
bac808fe
MD
455 */
456 kmalloc_create(&hmp->m_misc, "HAMMER-others");
457 kmalloc_create(&hmp->m_inodes, "HAMMER-inodes");
458
7c457ac8 459 kmalloc_raise_limit(hmp->m_inodes, 0); /* unlimited */
bac808fe 460
dd94f1b1 461 hmp->root_btree_beg.localization = 0x00000000U;
47197d71
MD
462 hmp->root_btree_beg.obj_id = -0x8000000000000000LL;
463 hmp->root_btree_beg.key = -0x8000000000000000LL;
464 hmp->root_btree_beg.create_tid = 1;
465 hmp->root_btree_beg.delete_tid = 1;
466 hmp->root_btree_beg.rec_type = 0;
467 hmp->root_btree_beg.obj_type = 0;
468
dd94f1b1 469 hmp->root_btree_end.localization = 0xFFFFFFFFU;
47197d71
MD
470 hmp->root_btree_end.obj_id = 0x7FFFFFFFFFFFFFFFLL;
471 hmp->root_btree_end.key = 0x7FFFFFFFFFFFFFFFLL;
472 hmp->root_btree_end.create_tid = 0xFFFFFFFFFFFFFFFFULL;
473 hmp->root_btree_end.delete_tid = 0; /* special case */
474 hmp->root_btree_end.rec_type = 0xFFFFU;
475 hmp->root_btree_end.obj_type = 0;
f03c9cf4 476
cdb6e4e6
MD
477 hmp->krate.freq = 1; /* maximum reporting rate (hz) */
478 hmp->krate.count = -16; /* initial burst */
479
9480ff55 480 hmp->sync_lock.refs = 1;
c9b9e29d 481 hmp->free_lock.refs = 1;
d99d6bf5
MD
482 hmp->undo_lock.refs = 1;
483 hmp->blkmap_lock.refs = 1;
83f2a3aa 484 hmp->snapshot_lock.refs = 1;
52e547e3 485 hmp->volume_lock.refs = 1;
9480ff55 486
cebe9493 487 TAILQ_INIT(&hmp->delay_list);
7a61b85d 488 TAILQ_INIT(&hmp->flush_group_list);
0729c8c8 489 TAILQ_INIT(&hmp->objid_cache_list);
e8599db1 490 TAILQ_INIT(&hmp->undo_lru_list);
7bc5b8c2 491 TAILQ_INIT(&hmp->reclaim_list);
507df98a
ID
492
493 RB_INIT(&hmp->rb_dedup_crc_root);
494 RB_INIT(&hmp->rb_dedup_off_root);
495 TAILQ_INIT(&hmp->dedup_lru_list);
195c19a1 496 }
51c35492
MD
497 hmp->hflags &= ~HMNT_USERFLAGS;
498 hmp->hflags |= info.hflags & HMNT_USERFLAGS;
732a1697
MD
499
500 hmp->master_id = master_id;
501
7f7c1f84
MD
502 if (info.asof) {
503 mp->mnt_flag |= MNT_RDONLY;
504 hmp->asof = info.asof;
505 } else {
506 hmp->asof = HAMMER_MAX_TID;
507 }
195c19a1 508
865c9609
MN
509 hmp->volume_to_remove = -1;
510
195c19a1 511 /*
51c35492 512 * Re-open read-write if originally read-only, or vise-versa.
02428fb6
MD
513 *
514 * When going from read-only to read-write execute the stage2
515 * recovery if it has not already been run.
195c19a1
MD
516 */
517 if (mp->mnt_flag & MNT_UPDATE) {
b0aab9b9 518 lwkt_gettoken(&hmp->fs_token);
51c35492
MD
519 error = 0;
520 if (hmp->ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
521 kprintf("HAMMER read-only -> read-write\n");
522 hmp->ronly = 0;
523 RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
524 hammer_adjust_volume_mode, NULL);
525 rootvol = hammer_get_root_volume(hmp, &error);
526 if (rootvol) {
06ad81ff 527 hammer_recover_flush_buffers(hmp, rootvol, 1);
02428fb6 528 error = hammer_recover_stage2(hmp, rootvol);
9f5097dc
MD
529 bcopy(rootvol->ondisk->vol0_blockmap,
530 hmp->blockmap,
531 sizeof(hmp->blockmap));
51c35492
MD
532 hammer_rel_volume(rootvol, 0);
533 }
534 RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
535 hammer_reload_inode, NULL);
536 /* kernel clears MNT_RDONLY */
537 } else if (hmp->ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
538 kprintf("HAMMER read-write -> read-only\n");
539 hmp->ronly = 1; /* messy */
540 RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
541 hammer_reload_inode, NULL);
195c19a1 542 hmp->ronly = 0;
51c35492
MD
543 hammer_flusher_sync(hmp);
544 hammer_flusher_sync(hmp);
545 hammer_flusher_sync(hmp);
546 hmp->ronly = 1;
547 RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
548 hammer_adjust_volume_mode, NULL);
195c19a1 549 }
b0aab9b9 550 lwkt_reltoken(&hmp->fs_token);
51c35492 551 return(error);
195c19a1
MD
552 }
553
427e5fc6
MD
554 RB_INIT(&hmp->rb_vols_root);
555 RB_INIT(&hmp->rb_inos_root);
73896937 556 RB_INIT(&hmp->rb_redo_root);
40043e7f 557 RB_INIT(&hmp->rb_nods_root);
e8599db1 558 RB_INIT(&hmp->rb_undo_root);
0832c9bb
MD
559 RB_INIT(&hmp->rb_resv_root);
560 RB_INIT(&hmp->rb_bufs_root);
5fa5c92f 561 RB_INIT(&hmp->rb_pfsm_root);
0832c9bb 562
195c19a1 563 hmp->ronly = ((mp->mnt_flag & MNT_RDONLY) != 0);
427e5fc6 564
1afb73cf
MD
565 RB_INIT(&hmp->volu_root);
566 RB_INIT(&hmp->undo_root);
567 RB_INIT(&hmp->data_root);
568 RB_INIT(&hmp->meta_root);
569 RB_INIT(&hmp->lose_root);
eddadaee 570 TAILQ_INIT(&hmp->iorun_list);
10a5d1ba 571
a3c18566
MD
572 lwkt_token_init(&hmp->fs_token, "hammerfs");
573 lwkt_token_init(&hmp->io_token, "hammerio");
b0aab9b9
MD
574
575 lwkt_gettoken(&hmp->fs_token);
576
427e5fc6
MD
577 /*
578 * Load volumes
579 */
580 path = objcache_get(namei_oc, M_WAITOK);
1b0ab2c3 581 hmp->nvolumes = -1;
427e5fc6 582 for (i = 0; i < info.nvolumes; ++i) {
7c19b529
MN
583 if (mntpt == NULL) {
584 /*
585 * Root mount.
7c19b529 586 */
104cb849 587 KKASSERT(next_volume_ptr != NULL);
a407819f
MN
588 strcpy(path, "");
589 if (*next_volume_ptr != '/') {
590 /* relative path */
591 strcpy(path, "/dev/");
592 }
104cb849
MN
593 int k;
594 for (k = strlen(path); k < MAXPATHLEN-1; ++k) {
595 if (*next_volume_ptr == '\0') {
596 break;
597 } else if (*next_volume_ptr == ':') {
598 ++next_volume_ptr;
599 break;
600 } else {
601 path[k] = *next_volume_ptr;
602 ++next_volume_ptr;
603 }
604 }
605 path[k] = '\0';
606
7c19b529 607 error = 0;
104cb849
MN
608 cdev_t dev = kgetdiskbyname(path);
609 error = bdevvp(dev, &devvp);
610 if (error) {
611 kprintf("hammer_mountroot: can't find devvp\n");
612 }
7c19b529
MN
613 } else {
614 error = copyin(&info.volumes[i], &upath,
615 sizeof(char *));
616 if (error == 0)
617 error = copyinstr(upath, path,
618 MAXPATHLEN, NULL);
619 }
427e5fc6 620 if (error == 0)
7c19b529 621 error = hammer_install_volume(hmp, path, devvp);
427e5fc6
MD
622 if (error)
623 break;
624 }
625 objcache_put(namei_oc, path);
626
627 /*
628 * Make sure we found a root volume
629 */
630 if (error == 0 && hmp->rootvol == NULL) {
631 kprintf("hammer_mount: No root volume found!\n");
632 error = EINVAL;
633 }
1b0ab2c3
MD
634
635 /*
636 * Check that all required volumes are available
637 */
638 if (error == 0 && hammer_mountcheck_volumes(hmp)) {
639 kprintf("hammer_mount: Missing volumes, cannot mount!\n");
640 error = EINVAL;
641 }
642
427e5fc6 643 if (error) {
b0aab9b9 644 /* called with fs_token held */
427e5fc6
MD
645 hammer_free_hmp(mp);
646 return (error);
647 }
648
649 /*
27ea2398
MD
650 * No errors, setup enough of the mount point so we can lookup the
651 * root vnode.
427e5fc6 652 */
27ea2398
MD
653 mp->mnt_iosize_max = MAXPHYS;
654 mp->mnt_kern_flag |= MNTK_FSMID;
c0ade690 655
aac0aabd
MD
656 /*
657 * MPSAFE code. Note that VOPs and VFSops which are not MPSAFE
658 * will acquire a per-mount token prior to entry and release it
659 * on return, so even if we do not specify it we no longer get
660 * the BGL regardlless of how we are flagged.
661 */
b0aab9b9
MD
662 mp->mnt_kern_flag |= MNTK_ALL_MPSAFE;
663 /*MNTK_RD_MPSAFE | MNTK_GA_MPSAFE | MNTK_IN_MPSAFE;*/
aac0aabd 664
c0ade690
MD
665 /*
666 * note: f_iosize is used by vnode_pager_haspage() when constructing
667 * its VOP_BMAP call.
668 */
669 mp->mnt_stat.f_iosize = HAMMER_BUFSIZE;
fbc6e32a 670 mp->mnt_stat.f_bsize = HAMMER_BUFSIZE;
6f97fce3
MD
671
672 mp->mnt_vstat.f_frsize = HAMMER_BUFSIZE;
673 mp->mnt_vstat.f_bsize = HAMMER_BUFSIZE;
674
27ea2398
MD
675 mp->mnt_maxsymlinklen = 255;
676 mp->mnt_flag |= MNT_LOCAL;
677
678 vfs_add_vnodeops(mp, &hammer_vnode_vops, &mp->mnt_vn_norm_ops);
7a04d74f
MD
679 vfs_add_vnodeops(mp, &hammer_spec_vops, &mp->mnt_vn_spec_ops);
680 vfs_add_vnodeops(mp, &hammer_fifo_vops, &mp->mnt_vn_fifo_ops);
27ea2398
MD
681
682 /*
a89aec1b
MD
683 * The root volume's ondisk pointer is only valid if we hold a
684 * reference to it.
685 */
686 rootvol = hammer_get_root_volume(hmp, &error);
687 if (error)
f90dde4c
MD
688 goto failed;
689
690 /*
9f5097dc 691 * Perform any necessary UNDO operations. The recovery code does
0729c8c8
MD
692 * call hammer_undo_lookup() so we have to pre-cache the blockmap,
693 * and then re-copy it again after recovery is complete.
c9b9e29d 694 *
51c35492
MD
695 * If this is a read-only mount the UNDO information is retained
696 * in memory in the form of dirty buffer cache buffers, and not
697 * written back to the media.
f90dde4c 698 */
0729c8c8
MD
699 bcopy(rootvol->ondisk->vol0_blockmap, hmp->blockmap,
700 sizeof(hmp->blockmap));
c9b9e29d 701
7a61b85d 702 /*
44a83111
MD
703 * Check filesystem version
704 */
705 hmp->version = rootvol->ondisk->vol_version;
706 if (hmp->version < HAMMER_VOL_VERSION_MIN ||
707 hmp->version > HAMMER_VOL_VERSION_MAX) {
708 kprintf("HAMMER: mount unsupported fs version %d\n",
709 hmp->version);
710 error = ERANGE;
711 goto done;
712 }
713
714 /*
7a61b85d
MD
715 * The undo_rec_limit limits the size of flush groups to avoid
716 * blowing out the UNDO FIFO. This calculation is typically in
717 * the tens of thousands and is designed primarily when small
718 * HAMMER filesystems are created.
719 */
720 hmp->undo_rec_limit = hammer_undo_max(hmp) / 8192 + 100;
721 if (hammer_debug_general & 0x0001)
722 kprintf("HAMMER: undo_rec_limit %d\n", hmp->undo_rec_limit);
723
02428fb6
MD
724 /*
725 * NOTE: Recover stage1 not only handles meta-data recovery, it
726 * also sets hmp->undo_seqno for HAMMER VERSION 4+ filesystems.
727 */
728 error = hammer_recover_stage1(hmp, rootvol);
f90dde4c
MD
729 if (error) {
730 kprintf("Failed to recover HAMMER filesystem on mount\n");
a89aec1b 731 goto done;
f90dde4c
MD
732 }
733
734 /*
adf01747
MD
735 * Finish setup now that we have a good root volume.
736 *
737 * The top 16 bits of fsid.val[1] is a pfs id.
f90dde4c 738 */
a89aec1b
MD
739 ksnprintf(mp->mnt_stat.f_mntfromname,
740 sizeof(mp->mnt_stat.f_mntfromname), "%s",
741 rootvol->ondisk->vol_name);
513ca7d7
MD
742 mp->mnt_stat.f_fsid.val[0] =
743 crc32((char *)&rootvol->ondisk->vol_fsid + 0, 8);
744 mp->mnt_stat.f_fsid.val[1] =
745 crc32((char *)&rootvol->ondisk->vol_fsid + 8, 8);
adf01747 746 mp->mnt_stat.f_fsid.val[1] &= 0x0000FFFF;
b84de5af 747
6f97fce3
MD
748 mp->mnt_vstat.f_fsid_uuid = rootvol->ondisk->vol_fsid;
749 mp->mnt_vstat.f_fsid = crc32(&mp->mnt_vstat.f_fsid_uuid,
750 sizeof(mp->mnt_vstat.f_fsid_uuid));
751
0729c8c8
MD
752 /*
753 * Certain often-modified fields in the root volume are cached in
754 * the hammer_mount structure so we do not have to generate lots
755 * of little UNDO structures for them.
c9b9e29d 756 *
9f5097dc
MD
757 * Recopy after recovery. This also has the side effect of
758 * setting our cached undo FIFO's first_offset, which serves to
759 * placemark the FIFO start for the NEXT flush cycle while the
760 * on-disk first_offset represents the LAST flush cycle.
0729c8c8 761 */
b84de5af 762 hmp->next_tid = rootvol->ondisk->vol0_next_tid;
4889cbd4
MD
763 hmp->flush_tid1 = hmp->next_tid;
764 hmp->flush_tid2 = hmp->next_tid;
0729c8c8
MD
765 bcopy(rootvol->ondisk->vol0_blockmap, hmp->blockmap,
766 sizeof(hmp->blockmap));
e63644f0 767 hmp->copy_stat_freebigblocks = rootvol->ondisk->vol0_stat_freebigblocks;
0729c8c8 768
059819e3
MD
769 hammer_flusher_create(hmp);
770
a89aec1b 771 /*
27ea2398
MD
772 * Locate the root directory using the root cluster's B-Tree as a
773 * starting point. The root directory uses an obj_id of 1.
774 *
775 * FUTURE: Leave the root directory cached referenced but unlocked
776 * in hmp->rootvp (need to flush it on unmount).
777 */
b9b0a6d0 778 error = hammer_vfs_vget(mp, NULL, 1, &rootvp);
a89aec1b
MD
779 if (error)
780 goto done;
781 vput(rootvp);
27ea2398 782 /*vn_unlock(hmp->rootvp);*/
88c39f64
TN
783 if (hmp->ronly == 0)
784 error = hammer_recover_stage2(hmp, rootvol);
427e5fc6 785
9fa352a2
MD
786 /*
787 * If the stage2 recovery fails be sure to clean out all cached
788 * vnodes before throwing away the mount structure or bad things
789 * will happen.
790 */
791 if (error)
792 vflush(mp, 0, 0);
793
a89aec1b 794done:
98ad7ade
FT
795 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
796 /* New mount */
797
798 /* Populate info for mount point (NULL pad)*/
799 bzero(mp->mnt_stat.f_mntonname, MNAMELEN);
800 size_t size;
801 if (mntpt) {
802 copyinstr(mntpt, mp->mnt_stat.f_mntonname,
803 MNAMELEN -1, &size);
804 } else { /* Root mount */
805 mp->mnt_stat.f_mntonname[0] = '/';
806 }
807 }
808 (void)VFS_STATFS(mp, &mp->mnt_stat, cred);
f90dde4c
MD
809 hammer_rel_volume(rootvol, 0);
810failed:
427e5fc6 811 /*
27ea2398 812 * Cleanup and return.
427e5fc6 813 */
b0aab9b9
MD
814 if (error) {
815 /* called with fs_token held */
427e5fc6 816 hammer_free_hmp(mp);
b0aab9b9
MD
817 } else {
818 lwkt_reltoken(&hmp->fs_token);
819 }
427e5fc6
MD
820 return (error);
821}
822
823static int
824hammer_vfs_unmount(struct mount *mp, int mntflags)
825{
b0aab9b9 826 hammer_mount_t hmp = (void *)mp->mnt_data;
427e5fc6 827 int flags;
66325755 828 int error;
27ea2398
MD
829
830 /*
427e5fc6
MD
831 * Clean out the vnodes
832 */
b0aab9b9 833 lwkt_gettoken(&hmp->fs_token);
66325755
MD
834 flags = 0;
835 if (mntflags & MNT_FORCE)
836 flags |= FORCECLOSE;
b0aab9b9 837 error = vflush(mp, 0, flags);
427e5fc6
MD
838
839 /*
840 * Clean up the internal mount structure and related entities. This
841 * may issue I/O.
842 */
b0aab9b9
MD
843 if (error == 0) {
844 /* called with fs_token held */
845 hammer_free_hmp(mp);
846 } else {
847 lwkt_reltoken(&hmp->fs_token);
848 }
849 return(error);
427e5fc6
MD
850}
851
852/*
853 * Clean up the internal mount structure and disassociate it from the mount.
854 * This may issue I/O.
b0aab9b9
MD
855 *
856 * Called with fs_token held.
427e5fc6
MD
857 */
858static void
859hammer_free_hmp(struct mount *mp)
860{
b0aab9b9 861 hammer_mount_t hmp = (void *)mp->mnt_data;
cdb6e4e6 862 hammer_flush_group_t flg;
1b0ab2c3 863 int count;
d40bfeca 864 int dummy;
427e5fc6
MD
865
866 /*
cdb6e4e6
MD
867 * Flush anything dirty. This won't even run if the
868 * filesystem errored-out.
427e5fc6 869 */
1b0ab2c3
MD
870 count = 0;
871 while (hammer_flusher_haswork(hmp)) {
872 hammer_flusher_sync(hmp);
873 ++count;
874 if (count >= 5) {
875 if (count == 5)
876 kprintf("HAMMER: umount flushing.");
877 else
878 kprintf(".");
d40bfeca 879 tsleep(&dummy, 0, "hmrufl", hz);
1b0ab2c3
MD
880 }
881 if (count == 30) {
882 kprintf("giving up\n");
883 break;
884 }
885 }
886 if (count >= 5 && count < 30)
887 kprintf("\n");
427e5fc6 888
cdb6e4e6
MD
889 /*
890 * If the mount had a critical error we have to destroy any
891 * remaining inodes before we can finish cleaning up the flusher.
892 */
893 if (hmp->flags & HAMMER_MOUNT_CRITICAL_ERROR) {
894 RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
895 hammer_destroy_inode_callback, NULL);
896 }
b84de5af 897
427e5fc6 898 /*
cdb6e4e6
MD
899 * There shouldn't be any inodes left now and any left over
900 * flush groups should now be empty.
427e5fc6 901 */
cdb6e4e6
MD
902 KKASSERT(RB_EMPTY(&hmp->rb_inos_root));
903 while ((flg = TAILQ_FIRST(&hmp->flush_group_list)) != NULL) {
904 TAILQ_REMOVE(&hmp->flush_group_list, flg, flush_entry);
ff003b11 905 KKASSERT(RB_EMPTY(&flg->flush_tree));
cdb6e4e6
MD
906 if (flg->refs) {
907 kprintf("HAMMER: Warning, flush_group %p was "
908 "not empty on umount!\n", flg);
909 }
bac808fe 910 kfree(flg, hmp->m_misc);
cdb6e4e6 911 }
427e5fc6
MD
912
913 /*
cdb6e4e6 914 * We can finally destroy the flusher
427e5fc6 915 */
cdb6e4e6
MD
916 hammer_flusher_destroy(hmp);
917
b84de5af 918 /*
00f16fad
MD
919 * We may have held recovered buffers due to a read-only mount.
920 * These must be discarded.
921 */
922 if (hmp->ronly)
923 hammer_recover_flush_buffers(hmp, NULL, -1);
924
925 /*
0832c9bb 926 * Unload buffers and then volumes
b84de5af 927 */
0832c9bb
MD
928 RB_SCAN(hammer_buf_rb_tree, &hmp->rb_bufs_root, NULL,
929 hammer_unload_buffer, NULL);
427e5fc6
MD
930 RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
931 hammer_unload_volume, NULL);
932
933 mp->mnt_data = NULL;
66325755 934 mp->mnt_flag &= ~MNT_LOCAL;
427e5fc6 935 hmp->mp = NULL;
0729c8c8 936 hammer_destroy_objid_cache(hmp);
507df98a
ID
937 hammer_destroy_dedup_cache(hmp);
938 if (hmp->dedup_free_cache != NULL) {
939 kfree(hmp->dedup_free_cache, hmp->m_misc);
940 hmp->dedup_free_cache = NULL;
941 }
bac808fe
MD
942 kmalloc_destroy(&hmp->m_misc);
943 kmalloc_destroy(&hmp->m_inodes);
b0aab9b9 944 lwkt_reltoken(&hmp->fs_token);
427e5fc6
MD
945 kfree(hmp, M_HAMMER);
946}
947
948/*
cdb6e4e6
MD
949 * Report critical errors. ip may be NULL.
950 */
951void
952hammer_critical_error(hammer_mount_t hmp, hammer_inode_t ip,
953 int error, const char *msg)
954{
955 hmp->flags |= HAMMER_MOUNT_CRITICAL_ERROR;
c9ce54d6 956
cdb6e4e6 957 krateprintf(&hmp->krate,
c9ce54d6
MD
958 "HAMMER(%s): Critical error inode=%jd error=%d %s\n",
959 hmp->mp->mnt_stat.f_mntfromname,
960 (intmax_t)(ip ? ip->obj_id : -1),
961 error, msg);
962
cdb6e4e6
MD
963 if (hmp->ronly == 0) {
964 hmp->ronly = 2; /* special errored read-only mode */
965 hmp->mp->mnt_flag |= MNT_RDONLY;
dfec479f
MD
966 RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
967 hammer_adjust_volume_mode, NULL);
cdb6e4e6
MD
968 kprintf("HAMMER(%s): Forcing read-only mode\n",
969 hmp->mp->mnt_stat.f_mntfromname);
970 }
971 hmp->error = error;
fc73edd8 972 if (hammer_debug_critical)
c9ce54d6 973 Debugger("Entering debugger");
cdb6e4e6
MD
974}
975
976
977/*
513ca7d7
MD
978 * Obtain a vnode for the specified inode number. An exclusively locked
979 * vnode is returned.
980 */
981int
b9b0a6d0
MD
982hammer_vfs_vget(struct mount *mp, struct vnode *dvp,
983 ino_t ino, struct vnode **vpp)
513ca7d7 984{
36f82b23 985 struct hammer_transaction trans;
513ca7d7
MD
986 struct hammer_mount *hmp = (void *)mp->mnt_data;
987 struct hammer_inode *ip;
988 int error;
b9b0a6d0 989 u_int32_t localization;
513ca7d7 990
b0aab9b9 991 lwkt_gettoken(&hmp->fs_token);
36f82b23
MD
992 hammer_simple_transaction(&trans, hmp);
993
513ca7d7 994 /*
b9b0a6d0
MD
995 * If a directory vnode is supplied (mainly NFS) then we can acquire
996 * the PFS domain from it. Otherwise we would only be able to vget
997 * inodes in the root PFS.
998 */
999 if (dvp) {
1000 localization = HAMMER_DEF_LOCALIZATION +
1001 VTOI(dvp)->obj_localization;
1002 } else {
1003 localization = HAMMER_DEF_LOCALIZATION;
1004 }
1005
1006 /*
36f82b23
MD
1007 * Lookup the requested HAMMER inode. The structure must be
1008 * left unlocked while we manipulate the related vnode to avoid
1009 * a deadlock.
513ca7d7 1010 */
ddfdf542 1011 ip = hammer_get_inode(&trans, NULL, ino,
b9b0a6d0 1012 hmp->asof, localization,
ddfdf542 1013 0, &error);
513ca7d7
MD
1014 if (ip == NULL) {
1015 *vpp = NULL;
b0aab9b9
MD
1016 } else {
1017 error = hammer_get_vnode(ip, vpp);
1018 hammer_rel_inode(ip, 0);
513ca7d7 1019 }
b84de5af 1020 hammer_done_transaction(&trans);
b0aab9b9 1021 lwkt_reltoken(&hmp->fs_token);
513ca7d7
MD
1022 return (error);
1023}
1024
1025/*
427e5fc6
MD
1026 * Return the root vnode for the filesystem.
1027 *
1028 * HAMMER stores the root vnode in the hammer_mount structure so
1029 * getting it is easy.
1030 */
1031static int
1032hammer_vfs_root(struct mount *mp, struct vnode **vpp)
1033{
27ea2398 1034 int error;
427e5fc6 1035
b9b0a6d0 1036 error = hammer_vfs_vget(mp, NULL, 1, vpp);
27ea2398 1037 return (error);
427e5fc6
MD
1038}
1039
1040static int
1041hammer_vfs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
1042{
fbc6e32a
MD
1043 struct hammer_mount *hmp = (void *)mp->mnt_data;
1044 hammer_volume_t volume;
1045 hammer_volume_ondisk_t ondisk;
1046 int error;
47197d71 1047 int64_t bfree;
31a56ce2 1048 int64_t breserved;
fbc6e32a 1049
b0aab9b9 1050 lwkt_gettoken(&hmp->fs_token);
fbc6e32a 1051 volume = hammer_get_root_volume(hmp, &error);
b0aab9b9
MD
1052 if (error) {
1053 lwkt_reltoken(&hmp->fs_token);
fbc6e32a 1054 return(error);
b0aab9b9 1055 }
fbc6e32a
MD
1056 ondisk = volume->ondisk;
1057
47197d71
MD
1058 /*
1059 * Basic stats
1060 */
31a56ce2 1061 _hammer_checkspace(hmp, HAMMER_CHKSPC_WRITE, &breserved);
fbc6e32a 1062 mp->mnt_stat.f_files = ondisk->vol0_stat_inodes;
c3be93f2 1063 bfree = ondisk->vol0_stat_freebigblocks * HAMMER_LARGEBLOCK_SIZE;
40043e7f 1064 hammer_rel_volume(volume, 0);
47197d71 1065
31a56ce2 1066 mp->mnt_stat.f_bfree = (bfree - breserved) / HAMMER_BUFSIZE;
47197d71 1067 mp->mnt_stat.f_bavail = mp->mnt_stat.f_bfree;
fbc6e32a
MD
1068 if (mp->mnt_stat.f_files < 0)
1069 mp->mnt_stat.f_files = 0;
1070
27ea2398 1071 *sbp = mp->mnt_stat;
b0aab9b9 1072 lwkt_reltoken(&hmp->fs_token);
27ea2398 1073 return(0);
427e5fc6
MD
1074}
1075
6f97fce3
MD
1076static int
1077hammer_vfs_statvfs(struct mount *mp, struct statvfs *sbp, struct ucred *cred)
1078{
1079 struct hammer_mount *hmp = (void *)mp->mnt_data;
1080 hammer_volume_t volume;
1081 hammer_volume_ondisk_t ondisk;
1082 int error;
1083 int64_t bfree;
0f65be10 1084 int64_t breserved;
6f97fce3 1085
b0aab9b9 1086 lwkt_gettoken(&hmp->fs_token);
6f97fce3 1087 volume = hammer_get_root_volume(hmp, &error);
b0aab9b9
MD
1088 if (error) {
1089 lwkt_reltoken(&hmp->fs_token);
6f97fce3 1090 return(error);
b0aab9b9 1091 }
6f97fce3
MD
1092 ondisk = volume->ondisk;
1093
1094 /*
1095 * Basic stats
1096 */
0f65be10 1097 _hammer_checkspace(hmp, HAMMER_CHKSPC_WRITE, &breserved);
6f97fce3
MD
1098 mp->mnt_vstat.f_files = ondisk->vol0_stat_inodes;
1099 bfree = ondisk->vol0_stat_freebigblocks * HAMMER_LARGEBLOCK_SIZE;
1100 hammer_rel_volume(volume, 0);
1101
0f65be10 1102 mp->mnt_vstat.f_bfree = (bfree - breserved) / HAMMER_BUFSIZE;
c0763659 1103 mp->mnt_vstat.f_bavail = mp->mnt_vstat.f_bfree;
6f97fce3
MD
1104 if (mp->mnt_vstat.f_files < 0)
1105 mp->mnt_vstat.f_files = 0;
1106 *sbp = mp->mnt_vstat;
b0aab9b9 1107 lwkt_reltoken(&hmp->fs_token);
6f97fce3
MD
1108 return(0);
1109}
1110
0729c8c8
MD
1111/*
1112 * Sync the filesystem. Currently we have to run it twice, the second
1113 * one will advance the undo start index to the end index, so if a crash
1114 * occurs no undos will be run on mount.
77062c8a
MD
1115 *
1116 * We do not sync the filesystem if we are called from a panic. If we did
1117 * we might end up blowing up a sync that was already in progress.
0729c8c8 1118 */
427e5fc6
MD
1119static int
1120hammer_vfs_sync(struct mount *mp, int waitfor)
1121{
fbc6e32a 1122 struct hammer_mount *hmp = (void *)mp->mnt_data;
0729c8c8
MD
1123 int error;
1124
b0aab9b9 1125 lwkt_gettoken(&hmp->fs_token);
77062c8a 1126 if (panicstr == NULL) {
0729c8c8 1127 error = hammer_sync_hmp(hmp, waitfor);
77062c8a
MD
1128 } else {
1129 error = EIO;
77062c8a 1130 }
b0aab9b9 1131 lwkt_reltoken(&hmp->fs_token);
0729c8c8 1132 return (error);
427e5fc6
MD
1133}
1134
513ca7d7
MD
1135/*
1136 * Convert a vnode to a file handle.
b0aab9b9
MD
1137 *
1138 * Accesses read-only fields on already-referenced structures so
1139 * no token is needed.
513ca7d7
MD
1140 */
1141static int
1142hammer_vfs_vptofh(struct vnode *vp, struct fid *fhp)
1143{
1144 hammer_inode_t ip;
1145
1146 KKASSERT(MAXFIDSZ >= 16);
1147 ip = VTOI(vp);
1148 fhp->fid_len = offsetof(struct fid, fid_data[16]);
adf01747 1149 fhp->fid_ext = ip->obj_localization >> 16;
513ca7d7
MD
1150 bcopy(&ip->obj_id, fhp->fid_data + 0, sizeof(ip->obj_id));
1151 bcopy(&ip->obj_asof, fhp->fid_data + 8, sizeof(ip->obj_asof));
1152 return(0);
1153}
1154
1155
1156/*
1157 * Convert a file handle back to a vnode.
67863d04
MD
1158 *
1159 * Use rootvp to enforce PFS isolation when a PFS is exported via a
1160 * null mount.
513ca7d7
MD
1161 */
1162static int
67863d04
MD
1163hammer_vfs_fhtovp(struct mount *mp, struct vnode *rootvp,
1164 struct fid *fhp, struct vnode **vpp)
513ca7d7 1165{
b0aab9b9 1166 hammer_mount_t hmp = (void *)mp->mnt_data;
36f82b23 1167 struct hammer_transaction trans;
513ca7d7
MD
1168 struct hammer_inode *ip;
1169 struct hammer_inode_info info;
1170 int error;
adf01747 1171 u_int32_t localization;
513ca7d7
MD
1172
1173 bcopy(fhp->fid_data + 0, &info.obj_id, sizeof(info.obj_id));
1174 bcopy(fhp->fid_data + 8, &info.obj_asof, sizeof(info.obj_asof));
67863d04
MD
1175 if (rootvp)
1176 localization = VTOI(rootvp)->obj_localization;
1177 else
1178 localization = (u_int32_t)fhp->fid_ext << 16;
513ca7d7 1179
b0aab9b9
MD
1180 lwkt_gettoken(&hmp->fs_token);
1181 hammer_simple_transaction(&trans, hmp);
36f82b23 1182
513ca7d7
MD
1183 /*
1184 * Get/allocate the hammer_inode structure. The structure must be
1185 * unlocked while we manipulate the related vnode to avoid a
1186 * deadlock.
1187 */
ddfdf542 1188 ip = hammer_get_inode(&trans, NULL, info.obj_id,
adf01747 1189 info.obj_asof, localization, 0, &error);
d978e7cf
MD
1190 if (ip) {
1191 error = hammer_get_vnode(ip, vpp);
1192 hammer_rel_inode(ip, 0);
1193 } else {
513ca7d7 1194 *vpp = NULL;
513ca7d7 1195 }
b84de5af 1196 hammer_done_transaction(&trans);
b0aab9b9 1197 lwkt_reltoken(&hmp->fs_token);
513ca7d7
MD
1198 return (error);
1199}
1200
1201static int
1202hammer_vfs_checkexp(struct mount *mp, struct sockaddr *nam,
1203 int *exflagsp, struct ucred **credanonp)
1204{
1205 hammer_mount_t hmp = (void *)mp->mnt_data;
1206 struct netcred *np;
1207 int error;
1208
b0aab9b9 1209 lwkt_gettoken(&hmp->fs_token);
513ca7d7
MD
1210 np = vfs_export_lookup(mp, &hmp->export, nam);
1211 if (np) {
1212 *exflagsp = np->netc_exflags;
1213 *credanonp = &np->netc_anon;
1214 error = 0;
1215 } else {
1216 error = EACCES;
1217 }
b0aab9b9 1218 lwkt_reltoken(&hmp->fs_token);
513ca7d7
MD
1219 return (error);
1220
1221}
1222
1223int
1224hammer_vfs_export(struct mount *mp, int op, const struct export_args *export)
1225{
1226 hammer_mount_t hmp = (void *)mp->mnt_data;
1227 int error;
1228
b0aab9b9
MD
1229 lwkt_gettoken(&hmp->fs_token);
1230
513ca7d7
MD
1231 switch(op) {
1232 case MOUNTCTL_SET_EXPORT:
1233 error = vfs_export(mp, &hmp->export, export);
1234 break;
1235 default:
1236 error = EOPNOTSUPP;
1237 break;
1238 }
b0aab9b9
MD
1239 lwkt_reltoken(&hmp->fs_token);
1240
513ca7d7
MD
1241 return(error);
1242}
1243