HAMMER VFS - Throw away cached vnodes on stage2 mount failure
[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 */
7a61b85d 66int hammer_count_fsyncs;
b3deaf57 67int hammer_count_inodes;
af209b0f 68int hammer_count_iqueued;
9f5097dc 69int hammer_count_reclaiming;
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
f5a07a7a 97int hammer_count_dirtybufspace; /* global */
a99b9ea2 98int hammer_count_refedbufs; /* global */
0832c9bb 99int hammer_count_reservations;
a99b9ea2
MD
100int hammer_count_io_running_read;
101int hammer_count_io_running_write;
102int hammer_count_io_locked;
f5a07a7a 103int hammer_limit_dirtybufspace; /* per-mount */
ba298df1 104int hammer_limit_running_io; /* per-mount */
0832c9bb 105int hammer_limit_recs; /* as a whole XXX */
de996e86 106int hammer_limit_inode_recs = 1024; /* per inode */
ff003b11 107int hammer_limit_reclaim = HAMMER_RECLAIM_WAIT;
6048b411 108int hammer_limit_redo = 4096 * 1024; /* per inode */
21fde338 109int hammer_autoflush = 2000; /* auto flush */
1f07f686 110int hammer_bio_count;
cb51be26 111int hammer_verify_zone;
1b0ab2c3 112int hammer_verify_data = 1;
cb51be26 113int hammer_write_mode;
3e583440 114int hammer_yield_check = 16;
9d0a6205 115int hammer_fsync_mode = 3;
7d683b0f 116int64_t hammer_contention_count;
f03c9cf4 117int64_t hammer_zone_limit;
b3deaf57
MD
118
119SYSCTL_NODE(_vfs, OID_AUTO, hammer, CTLFLAG_RW, 0, "HAMMER filesystem");
5987cc42
MD
120SYSCTL_INT(_vfs_hammer, OID_AUTO, supported_version, CTLFLAG_RD,
121 &hammer_supported_version, 0, "");
d5ef456e
MD
122SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_general, CTLFLAG_RW,
123 &hammer_debug_general, 0, "");
2f85fa4d
MD
124SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_io, CTLFLAG_RW,
125 &hammer_debug_io, 0, "");
77062c8a
MD
126SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_debug, CTLFLAG_RW,
127 &hammer_debug_debug, 0, "");
e8599db1
MD
128SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_inode, CTLFLAG_RW,
129 &hammer_debug_inode, 0, "");
7d683b0f
MD
130SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_locks, CTLFLAG_RW,
131 &hammer_debug_locks, 0, "");
b3deaf57
MD
132SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_btree, CTLFLAG_RW,
133 &hammer_debug_btree, 0, "");
d113fda1
MD
134SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_tid, CTLFLAG_RW,
135 &hammer_debug_tid, 0, "");
b33e2cc0
MD
136SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_recover, CTLFLAG_RW,
137 &hammer_debug_recover, 0, "");
46fe7ae1
MD
138SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_recover_faults, CTLFLAG_RW,
139 &hammer_debug_recover_faults, 0, "");
fc73edd8
MD
140SYSCTL_INT(_vfs_hammer, OID_AUTO, debug_critical, CTLFLAG_RW,
141 &hammer_debug_critical, 0, "");
1b0ab2c3
MD
142SYSCTL_INT(_vfs_hammer, OID_AUTO, cluster_enable, CTLFLAG_RW,
143 &hammer_cluster_enable, 0, "");
9480ff55 144
f5a07a7a
MD
145SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_dirtybufspace, CTLFLAG_RW,
146 &hammer_limit_dirtybufspace, 0, "");
ba298df1
MD
147SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_running_io, CTLFLAG_RW,
148 &hammer_limit_running_io, 0, "");
47637bff
MD
149SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_recs, CTLFLAG_RW,
150 &hammer_limit_recs, 0, "");
de996e86
MD
151SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_inode_recs, CTLFLAG_RW,
152 &hammer_limit_inode_recs, 0, "");
ff003b11
MD
153SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_reclaim, CTLFLAG_RW,
154 &hammer_limit_reclaim, 0, "");
6048b411
MD
155SYSCTL_INT(_vfs_hammer, OID_AUTO, limit_redo, CTLFLAG_RW,
156 &hammer_limit_redo, 0, "");
9480ff55 157
7a61b85d
MD
158SYSCTL_INT(_vfs_hammer, OID_AUTO, count_fsyncs, CTLFLAG_RD,
159 &hammer_count_fsyncs, 0, "");
b3deaf57
MD
160SYSCTL_INT(_vfs_hammer, OID_AUTO, count_inodes, CTLFLAG_RD,
161 &hammer_count_inodes, 0, "");
af209b0f
MD
162SYSCTL_INT(_vfs_hammer, OID_AUTO, count_iqueued, CTLFLAG_RD,
163 &hammer_count_iqueued, 0, "");
9f5097dc
MD
164SYSCTL_INT(_vfs_hammer, OID_AUTO, count_reclaiming, CTLFLAG_RD,
165 &hammer_count_reclaiming, 0, "");
b3deaf57
MD
166SYSCTL_INT(_vfs_hammer, OID_AUTO, count_records, CTLFLAG_RD,
167 &hammer_count_records, 0, "");
168SYSCTL_INT(_vfs_hammer, OID_AUTO, count_record_datas, CTLFLAG_RD,
169 &hammer_count_record_datas, 0, "");
170SYSCTL_INT(_vfs_hammer, OID_AUTO, count_volumes, CTLFLAG_RD,
171 &hammer_count_volumes, 0, "");
b3deaf57
MD
172SYSCTL_INT(_vfs_hammer, OID_AUTO, count_buffers, CTLFLAG_RD,
173 &hammer_count_buffers, 0, "");
174SYSCTL_INT(_vfs_hammer, OID_AUTO, count_nodes, CTLFLAG_RD,
175 &hammer_count_nodes, 0, "");
a7e9bef1
MD
176SYSCTL_QUAD(_vfs_hammer, OID_AUTO, count_extra_space_used, CTLFLAG_RD,
177 &hammer_count_extra_space_used, 0, "");
ce0138a6 178
cb51be26
MD
179SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_searches, CTLFLAG_RD,
180 &hammer_stats_btree_searches, 0, "");
181SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_lookups, CTLFLAG_RD,
182 &hammer_stats_btree_lookups, 0, "");
183SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_inserts, CTLFLAG_RD,
184 &hammer_stats_btree_inserts, 0, "");
185SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_deletes, CTLFLAG_RD,
186 &hammer_stats_btree_deletes, 0, "");
187SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_elements, CTLFLAG_RD,
188 &hammer_stats_btree_elements, 0, "");
189SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_splits, CTLFLAG_RD,
190 &hammer_stats_btree_splits, 0, "");
191SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_iterations, CTLFLAG_RD,
192 &hammer_stats_btree_iterations, 0, "");
39d8fd63
MD
193SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_btree_root_iterations, CTLFLAG_RD,
194 &hammer_stats_btree_root_iterations, 0, "");
cb51be26
MD
195SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_record_iterations, CTLFLAG_RD,
196 &hammer_stats_record_iterations, 0, "");
ce0138a6
MD
197
198SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_read, CTLFLAG_RD,
199 &hammer_stats_file_read, 0, "");
200SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_write, CTLFLAG_RD,
201 &hammer_stats_file_write, 0, "");
202SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_iopsr, CTLFLAG_RD,
203 &hammer_stats_file_iopsr, 0, "");
204SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_file_iopsw, CTLFLAG_RD,
205 &hammer_stats_file_iopsw, 0, "");
206SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_disk_read, CTLFLAG_RD,
207 &hammer_stats_disk_read, 0, "");
208SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_disk_write, CTLFLAG_RD,
209 &hammer_stats_disk_write, 0, "");
210SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_inode_flushes, CTLFLAG_RD,
211 &hammer_stats_inode_flushes, 0, "");
212SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_commits, CTLFLAG_RD,
213 &hammer_stats_commits, 0, "");
89e744ce
MD
214SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_undo, CTLFLAG_RD,
215 &hammer_stats_undo, 0, "");
6048b411
MD
216SYSCTL_QUAD(_vfs_hammer, OID_AUTO, stats_redo, CTLFLAG_RD,
217 &hammer_stats_redo, 0, "");
ce0138a6 218
f5a07a7a
MD
219SYSCTL_INT(_vfs_hammer, OID_AUTO, count_dirtybufspace, CTLFLAG_RD,
220 &hammer_count_dirtybufspace, 0, "");
a99b9ea2
MD
221SYSCTL_INT(_vfs_hammer, OID_AUTO, count_refedbufs, CTLFLAG_RD,
222 &hammer_count_refedbufs, 0, "");
0832c9bb
MD
223SYSCTL_INT(_vfs_hammer, OID_AUTO, count_reservations, CTLFLAG_RD,
224 &hammer_count_reservations, 0, "");
a99b9ea2
MD
225SYSCTL_INT(_vfs_hammer, OID_AUTO, count_io_running_read, CTLFLAG_RD,
226 &hammer_count_io_running_read, 0, "");
227SYSCTL_INT(_vfs_hammer, OID_AUTO, count_io_locked, CTLFLAG_RD,
228 &hammer_count_io_locked, 0, "");
229SYSCTL_INT(_vfs_hammer, OID_AUTO, count_io_running_write, CTLFLAG_RD,
230 &hammer_count_io_running_write, 0, "");
f03c9cf4
MD
231SYSCTL_QUAD(_vfs_hammer, OID_AUTO, zone_limit, CTLFLAG_RW,
232 &hammer_zone_limit, 0, "");
7d683b0f
MD
233SYSCTL_QUAD(_vfs_hammer, OID_AUTO, contention_count, CTLFLAG_RW,
234 &hammer_contention_count, 0, "");
21fde338
MD
235SYSCTL_INT(_vfs_hammer, OID_AUTO, autoflush, CTLFLAG_RW,
236 &hammer_autoflush, 0, "");
cb51be26
MD
237SYSCTL_INT(_vfs_hammer, OID_AUTO, verify_zone, CTLFLAG_RW,
238 &hammer_verify_zone, 0, "");
1b0ab2c3
MD
239SYSCTL_INT(_vfs_hammer, OID_AUTO, verify_data, CTLFLAG_RW,
240 &hammer_verify_data, 0, "");
cb51be26
MD
241SYSCTL_INT(_vfs_hammer, OID_AUTO, write_mode, CTLFLAG_RW,
242 &hammer_write_mode, 0, "");
3e583440
MD
243SYSCTL_INT(_vfs_hammer, OID_AUTO, yield_check, CTLFLAG_RW,
244 &hammer_yield_check, 0, "");
6f3d87c0
MD
245SYSCTL_INT(_vfs_hammer, OID_AUTO, fsync_mode, CTLFLAG_RW,
246 &hammer_fsync_mode, 0, "");
b3deaf57 247
bcac4bbb
MD
248KTR_INFO_MASTER(hammer);
249
427e5fc6
MD
250/*
251 * VFS ABI
252 */
253static void hammer_free_hmp(struct mount *mp);
254
255static int hammer_vfs_mount(struct mount *mp, char *path, caddr_t data,
256 struct ucred *cred);
257static int hammer_vfs_unmount(struct mount *mp, int mntflags);
258static int hammer_vfs_root(struct mount *mp, struct vnode **vpp);
259static int hammer_vfs_statfs(struct mount *mp, struct statfs *sbp,
260 struct ucred *cred);
6f97fce3
MD
261static int hammer_vfs_statvfs(struct mount *mp, struct statvfs *sbp,
262 struct ucred *cred);
427e5fc6 263static int hammer_vfs_sync(struct mount *mp, int waitfor);
b9b0a6d0
MD
264static int hammer_vfs_vget(struct mount *mp, struct vnode *dvp,
265 ino_t ino, struct vnode **vpp);
427e5fc6 266static int hammer_vfs_init(struct vfsconf *conf);
67863d04
MD
267static int hammer_vfs_fhtovp(struct mount *mp, struct vnode *rootvp,
268 struct fid *fhp, struct vnode **vpp);
513ca7d7
MD
269static int hammer_vfs_vptofh(struct vnode *vp, struct fid *fhp);
270static int hammer_vfs_checkexp(struct mount *mp, struct sockaddr *nam,
271 int *exflagsp, struct ucred **credanonp);
272
427e5fc6
MD
273
274static struct vfsops hammer_vfsops = {
275 .vfs_mount = hammer_vfs_mount,
276 .vfs_unmount = hammer_vfs_unmount,
277 .vfs_root = hammer_vfs_root,
278 .vfs_statfs = hammer_vfs_statfs,
6f97fce3 279 .vfs_statvfs = hammer_vfs_statvfs,
427e5fc6
MD
280 .vfs_sync = hammer_vfs_sync,
281 .vfs_vget = hammer_vfs_vget,
513ca7d7
MD
282 .vfs_init = hammer_vfs_init,
283 .vfs_vptofh = hammer_vfs_vptofh,
284 .vfs_fhtovp = hammer_vfs_fhtovp,
285 .vfs_checkexp = hammer_vfs_checkexp
427e5fc6
MD
286};
287
bac808fe 288MALLOC_DEFINE(M_HAMMER, "HAMMER-mount", "");
427e5fc6
MD
289
290VFS_SET(hammer_vfsops, hammer, 0);
291MODULE_VERSION(hammer, 1);
292
293static int
294hammer_vfs_init(struct vfsconf *conf)
295{
3098dc2f
MD
296 int n;
297
298 if (hammer_limit_recs == 0) {
a99b9ea2 299 hammer_limit_recs = nbuf * 25;
3098dc2f
MD
300 n = kmalloc_limit(M_HAMMER) / 512;
301 if (hammer_limit_recs > n)
302 hammer_limit_recs = n;
303 }
f5a07a7a
MD
304 if (hammer_limit_dirtybufspace == 0) {
305 hammer_limit_dirtybufspace = hidirtybufspace / 2;
306 if (hammer_limit_dirtybufspace < 100)
307 hammer_limit_dirtybufspace = 100;
9f5097dc 308 }
ba298df1
MD
309
310 /*
311 * Set reasonable limits to maintain an I/O pipeline. This is
312 * used by the flush code which explicitly initiates I/O, and
313 * is per-mount.
314 *
315 * The system-driven buffer cache uses vfs.lorunningspace and
316 * vfs.hirunningspace globally.
317 */
318 if (hammer_limit_running_io == 0)
319 hammer_limit_running_io = hammer_limit_dirtybufspace;
320 if (hammer_limit_running_io > 10 * 1024 * 1024)
321 hammer_limit_running_io = 10 * 1024 * 1024;
427e5fc6
MD
322 return(0);
323}
324
325static int
326hammer_vfs_mount(struct mount *mp, char *mntpt, caddr_t data,
327 struct ucred *cred)
328{
329 struct hammer_mount_info info;
a89aec1b
MD
330 hammer_mount_t hmp;
331 hammer_volume_t rootvol;
27ea2398 332 struct vnode *rootvp;
7c19b529 333 struct vnode *devvp = NULL;
427e5fc6
MD
334 const char *upath; /* volume name in userspace */
335 char *path; /* volume name in system space */
336 int error;
337 int i;
732a1697 338 int master_id;
104cb849 339 char *next_volume_ptr = NULL;
ab0bf4a3
MD
340
341 /*
342 * Accept hammer_mount_info. mntpt is NULL for root mounts at boot.
343 */
7c19b529 344 if (mntpt == NULL) {
7c19b529
MN
345 bzero(&info, sizeof(info));
346 info.asof = 0;
347 info.hflags = 0;
348 info.nvolumes = 1;
104cb849
MN
349
350 next_volume_ptr = mp->mnt_stat.f_mntfromname;
351
352 /* Count number of volumes separated by ':' */
353 for (char *p = next_volume_ptr; *p != '\0'; ++p) {
354 if (*p == ':') {
355 ++info.nvolumes;
356 }
357 }
358
359 mp->mnt_flag &= ~MNT_RDONLY; /* mount R/W */
7c19b529
MN
360 } else {
361 if ((error = copyin(data, &info, sizeof(info))) != 0)
362 return (error);
363 }
732a1697
MD
364
365 /*
366 * updating or new mount
367 */
368 if (mp->mnt_flag & MNT_UPDATE) {
369 hmp = (void *)mp->mnt_data;
370 KKASSERT(hmp != NULL);
371 } else {
51c35492
MD
372 if (info.nvolumes <= 0 || info.nvolumes >= 32768)
373 return (EINVAL);
732a1697 374 hmp = NULL;
51c35492 375 }
732a1697
MD
376
377 /*
378 * master-id validation. The master id may not be changed by a
379 * mount update.
380 */
381 if (info.hflags & HMNT_MASTERID) {
382 if (hmp && hmp->master_id != info.master_id) {
383 kprintf("hammer: cannot change master id "
384 "with mount update\n");
385 return(EINVAL);
386 }
387 master_id = info.master_id;
388 if (master_id < -1 || master_id >= HAMMER_MAX_MASTERS)
5de0c0e5 389 return (EINVAL);
732a1697
MD
390 } else {
391 if (hmp)
392 master_id = hmp->master_id;
393 else
394 master_id = 0;
5de0c0e5 395 }
427e5fc6
MD
396
397 /*
398 * Interal mount data structure
399 */
732a1697 400 if (hmp == NULL) {
195c19a1
MD
401 hmp = kmalloc(sizeof(*hmp), M_HAMMER, M_WAITOK | M_ZERO);
402 mp->mnt_data = (qaddr_t)hmp;
403 hmp->mp = mp;
46fe7ae1 404 /*TAILQ_INIT(&hmp->recycle_list);*/
47197d71 405
bac808fe 406 /*
7c457ac8
MD
407 * Make sure kmalloc type limits are set appropriately.
408 *
409 * Our inode kmalloc group is sized based on maxvnodes
410 * (controlled by the system, not us).
bac808fe
MD
411 */
412 kmalloc_create(&hmp->m_misc, "HAMMER-others");
413 kmalloc_create(&hmp->m_inodes, "HAMMER-inodes");
414
7c457ac8 415 kmalloc_raise_limit(hmp->m_inodes, 0); /* unlimited */
bac808fe 416
dd94f1b1 417 hmp->root_btree_beg.localization = 0x00000000U;
47197d71
MD
418 hmp->root_btree_beg.obj_id = -0x8000000000000000LL;
419 hmp->root_btree_beg.key = -0x8000000000000000LL;
420 hmp->root_btree_beg.create_tid = 1;
421 hmp->root_btree_beg.delete_tid = 1;
422 hmp->root_btree_beg.rec_type = 0;
423 hmp->root_btree_beg.obj_type = 0;
424
dd94f1b1 425 hmp->root_btree_end.localization = 0xFFFFFFFFU;
47197d71
MD
426 hmp->root_btree_end.obj_id = 0x7FFFFFFFFFFFFFFFLL;
427 hmp->root_btree_end.key = 0x7FFFFFFFFFFFFFFFLL;
428 hmp->root_btree_end.create_tid = 0xFFFFFFFFFFFFFFFFULL;
429 hmp->root_btree_end.delete_tid = 0; /* special case */
430 hmp->root_btree_end.rec_type = 0xFFFFU;
431 hmp->root_btree_end.obj_type = 0;
f03c9cf4 432
cdb6e4e6
MD
433 hmp->krate.freq = 1; /* maximum reporting rate (hz) */
434 hmp->krate.count = -16; /* initial burst */
435
9480ff55 436 hmp->sync_lock.refs = 1;
c9b9e29d 437 hmp->free_lock.refs = 1;
d99d6bf5
MD
438 hmp->undo_lock.refs = 1;
439 hmp->blkmap_lock.refs = 1;
83f2a3aa 440 hmp->snapshot_lock.refs = 1;
52e547e3 441 hmp->volume_lock.refs = 1;
9480ff55 442
cebe9493 443 TAILQ_INIT(&hmp->delay_list);
7a61b85d 444 TAILQ_INIT(&hmp->flush_group_list);
0729c8c8 445 TAILQ_INIT(&hmp->objid_cache_list);
e8599db1 446 TAILQ_INIT(&hmp->undo_lru_list);
7bc5b8c2 447 TAILQ_INIT(&hmp->reclaim_list);
195c19a1 448 }
51c35492
MD
449 hmp->hflags &= ~HMNT_USERFLAGS;
450 hmp->hflags |= info.hflags & HMNT_USERFLAGS;
732a1697
MD
451
452 hmp->master_id = master_id;
453
7f7c1f84
MD
454 if (info.asof) {
455 mp->mnt_flag |= MNT_RDONLY;
456 hmp->asof = info.asof;
457 } else {
458 hmp->asof = HAMMER_MAX_TID;
459 }
195c19a1 460
865c9609
MN
461 hmp->volume_to_remove = -1;
462
195c19a1 463 /*
51c35492 464 * Re-open read-write if originally read-only, or vise-versa.
02428fb6
MD
465 *
466 * When going from read-only to read-write execute the stage2
467 * recovery if it has not already been run.
195c19a1
MD
468 */
469 if (mp->mnt_flag & MNT_UPDATE) {
51c35492
MD
470 error = 0;
471 if (hmp->ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
472 kprintf("HAMMER read-only -> read-write\n");
473 hmp->ronly = 0;
474 RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
475 hammer_adjust_volume_mode, NULL);
476 rootvol = hammer_get_root_volume(hmp, &error);
477 if (rootvol) {
06ad81ff 478 hammer_recover_flush_buffers(hmp, rootvol, 1);
02428fb6 479 error = hammer_recover_stage2(hmp, rootvol);
9f5097dc
MD
480 bcopy(rootvol->ondisk->vol0_blockmap,
481 hmp->blockmap,
482 sizeof(hmp->blockmap));
51c35492
MD
483 hammer_rel_volume(rootvol, 0);
484 }
485 RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
486 hammer_reload_inode, NULL);
487 /* kernel clears MNT_RDONLY */
488 } else if (hmp->ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
489 kprintf("HAMMER read-write -> read-only\n");
490 hmp->ronly = 1; /* messy */
491 RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
492 hammer_reload_inode, NULL);
195c19a1 493 hmp->ronly = 0;
51c35492
MD
494 hammer_flusher_sync(hmp);
495 hammer_flusher_sync(hmp);
496 hammer_flusher_sync(hmp);
497 hmp->ronly = 1;
498 RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
499 hammer_adjust_volume_mode, NULL);
195c19a1 500 }
51c35492 501 return(error);
195c19a1
MD
502 }
503
427e5fc6
MD
504 RB_INIT(&hmp->rb_vols_root);
505 RB_INIT(&hmp->rb_inos_root);
73896937 506 RB_INIT(&hmp->rb_redo_root);
40043e7f 507 RB_INIT(&hmp->rb_nods_root);
e8599db1 508 RB_INIT(&hmp->rb_undo_root);
0832c9bb
MD
509 RB_INIT(&hmp->rb_resv_root);
510 RB_INIT(&hmp->rb_bufs_root);
5fa5c92f 511 RB_INIT(&hmp->rb_pfsm_root);
0832c9bb 512
195c19a1 513 hmp->ronly = ((mp->mnt_flag & MNT_RDONLY) != 0);
427e5fc6 514
10a5d1ba
MD
515 TAILQ_INIT(&hmp->volu_list);
516 TAILQ_INIT(&hmp->undo_list);
517 TAILQ_INIT(&hmp->data_list);
518 TAILQ_INIT(&hmp->meta_list);
519 TAILQ_INIT(&hmp->lose_list);
eddadaee 520 TAILQ_INIT(&hmp->iorun_list);
10a5d1ba 521
427e5fc6
MD
522 /*
523 * Load volumes
524 */
525 path = objcache_get(namei_oc, M_WAITOK);
1b0ab2c3 526 hmp->nvolumes = -1;
427e5fc6 527 for (i = 0; i < info.nvolumes; ++i) {
7c19b529
MN
528 if (mntpt == NULL) {
529 /*
530 * Root mount.
7c19b529 531 */
104cb849 532 KKASSERT(next_volume_ptr != NULL);
a407819f
MN
533 strcpy(path, "");
534 if (*next_volume_ptr != '/') {
535 /* relative path */
536 strcpy(path, "/dev/");
537 }
104cb849
MN
538 int k;
539 for (k = strlen(path); k < MAXPATHLEN-1; ++k) {
540 if (*next_volume_ptr == '\0') {
541 break;
542 } else if (*next_volume_ptr == ':') {
543 ++next_volume_ptr;
544 break;
545 } else {
546 path[k] = *next_volume_ptr;
547 ++next_volume_ptr;
548 }
549 }
550 path[k] = '\0';
551
7c19b529 552 error = 0;
104cb849
MN
553 cdev_t dev = kgetdiskbyname(path);
554 error = bdevvp(dev, &devvp);
555 if (error) {
556 kprintf("hammer_mountroot: can't find devvp\n");
557 }
7c19b529
MN
558 } else {
559 error = copyin(&info.volumes[i], &upath,
560 sizeof(char *));
561 if (error == 0)
562 error = copyinstr(upath, path,
563 MAXPATHLEN, NULL);
564 }
427e5fc6 565 if (error == 0)
7c19b529 566 error = hammer_install_volume(hmp, path, devvp);
427e5fc6
MD
567 if (error)
568 break;
569 }
570 objcache_put(namei_oc, path);
571
572 /*
573 * Make sure we found a root volume
574 */
575 if (error == 0 && hmp->rootvol == NULL) {
576 kprintf("hammer_mount: No root volume found!\n");
577 error = EINVAL;
578 }
1b0ab2c3
MD
579
580 /*
581 * Check that all required volumes are available
582 */
583 if (error == 0 && hammer_mountcheck_volumes(hmp)) {
584 kprintf("hammer_mount: Missing volumes, cannot mount!\n");
585 error = EINVAL;
586 }
587
427e5fc6
MD
588 if (error) {
589 hammer_free_hmp(mp);
590 return (error);
591 }
592
593 /*
27ea2398
MD
594 * No errors, setup enough of the mount point so we can lookup the
595 * root vnode.
427e5fc6 596 */
27ea2398
MD
597 mp->mnt_iosize_max = MAXPHYS;
598 mp->mnt_kern_flag |= MNTK_FSMID;
c0ade690 599
aac0aabd
MD
600 /*
601 * MPSAFE code. Note that VOPs and VFSops which are not MPSAFE
602 * will acquire a per-mount token prior to entry and release it
603 * on return, so even if we do not specify it we no longer get
604 * the BGL regardlless of how we are flagged.
605 */
2247fe02
MD
606 mp->mnt_kern_flag |= MNTK_RD_MPSAFE | MNTK_GA_MPSAFE |
607 MNTK_IN_MPSAFE;
aac0aabd 608
c0ade690
MD
609 /*
610 * note: f_iosize is used by vnode_pager_haspage() when constructing
611 * its VOP_BMAP call.
612 */
613 mp->mnt_stat.f_iosize = HAMMER_BUFSIZE;
fbc6e32a 614 mp->mnt_stat.f_bsize = HAMMER_BUFSIZE;
6f97fce3
MD
615
616 mp->mnt_vstat.f_frsize = HAMMER_BUFSIZE;
617 mp->mnt_vstat.f_bsize = HAMMER_BUFSIZE;
618
27ea2398
MD
619 mp->mnt_maxsymlinklen = 255;
620 mp->mnt_flag |= MNT_LOCAL;
621
622 vfs_add_vnodeops(mp, &hammer_vnode_vops, &mp->mnt_vn_norm_ops);
7a04d74f
MD
623 vfs_add_vnodeops(mp, &hammer_spec_vops, &mp->mnt_vn_spec_ops);
624 vfs_add_vnodeops(mp, &hammer_fifo_vops, &mp->mnt_vn_fifo_ops);
27ea2398
MD
625
626 /*
a89aec1b
MD
627 * The root volume's ondisk pointer is only valid if we hold a
628 * reference to it.
629 */
630 rootvol = hammer_get_root_volume(hmp, &error);
631 if (error)
f90dde4c
MD
632 goto failed;
633
634 /*
9f5097dc 635 * Perform any necessary UNDO operations. The recovery code does
0729c8c8
MD
636 * call hammer_undo_lookup() so we have to pre-cache the blockmap,
637 * and then re-copy it again after recovery is complete.
c9b9e29d 638 *
51c35492
MD
639 * If this is a read-only mount the UNDO information is retained
640 * in memory in the form of dirty buffer cache buffers, and not
641 * written back to the media.
f90dde4c 642 */
0729c8c8
MD
643 bcopy(rootvol->ondisk->vol0_blockmap, hmp->blockmap,
644 sizeof(hmp->blockmap));
c9b9e29d 645
7a61b85d 646 /*
44a83111
MD
647 * Check filesystem version
648 */
649 hmp->version = rootvol->ondisk->vol_version;
650 if (hmp->version < HAMMER_VOL_VERSION_MIN ||
651 hmp->version > HAMMER_VOL_VERSION_MAX) {
652 kprintf("HAMMER: mount unsupported fs version %d\n",
653 hmp->version);
654 error = ERANGE;
655 goto done;
656 }
657
658 /*
7a61b85d
MD
659 * The undo_rec_limit limits the size of flush groups to avoid
660 * blowing out the UNDO FIFO. This calculation is typically in
661 * the tens of thousands and is designed primarily when small
662 * HAMMER filesystems are created.
663 */
664 hmp->undo_rec_limit = hammer_undo_max(hmp) / 8192 + 100;
665 if (hammer_debug_general & 0x0001)
666 kprintf("HAMMER: undo_rec_limit %d\n", hmp->undo_rec_limit);
667
02428fb6
MD
668 /*
669 * NOTE: Recover stage1 not only handles meta-data recovery, it
670 * also sets hmp->undo_seqno for HAMMER VERSION 4+ filesystems.
671 */
672 error = hammer_recover_stage1(hmp, rootvol);
f90dde4c
MD
673 if (error) {
674 kprintf("Failed to recover HAMMER filesystem on mount\n");
a89aec1b 675 goto done;
f90dde4c
MD
676 }
677
678 /*
adf01747
MD
679 * Finish setup now that we have a good root volume.
680 *
681 * The top 16 bits of fsid.val[1] is a pfs id.
f90dde4c 682 */
a89aec1b
MD
683 ksnprintf(mp->mnt_stat.f_mntfromname,
684 sizeof(mp->mnt_stat.f_mntfromname), "%s",
685 rootvol->ondisk->vol_name);
513ca7d7
MD
686 mp->mnt_stat.f_fsid.val[0] =
687 crc32((char *)&rootvol->ondisk->vol_fsid + 0, 8);
688 mp->mnt_stat.f_fsid.val[1] =
689 crc32((char *)&rootvol->ondisk->vol_fsid + 8, 8);
adf01747 690 mp->mnt_stat.f_fsid.val[1] &= 0x0000FFFF;
b84de5af 691
6f97fce3
MD
692 mp->mnt_vstat.f_fsid_uuid = rootvol->ondisk->vol_fsid;
693 mp->mnt_vstat.f_fsid = crc32(&mp->mnt_vstat.f_fsid_uuid,
694 sizeof(mp->mnt_vstat.f_fsid_uuid));
695
0729c8c8
MD
696 /*
697 * Certain often-modified fields in the root volume are cached in
698 * the hammer_mount structure so we do not have to generate lots
699 * of little UNDO structures for them.
c9b9e29d 700 *
9f5097dc
MD
701 * Recopy after recovery. This also has the side effect of
702 * setting our cached undo FIFO's first_offset, which serves to
703 * placemark the FIFO start for the NEXT flush cycle while the
704 * on-disk first_offset represents the LAST flush cycle.
0729c8c8 705 */
b84de5af 706 hmp->next_tid = rootvol->ondisk->vol0_next_tid;
4889cbd4
MD
707 hmp->flush_tid1 = hmp->next_tid;
708 hmp->flush_tid2 = hmp->next_tid;
0729c8c8
MD
709 bcopy(rootvol->ondisk->vol0_blockmap, hmp->blockmap,
710 sizeof(hmp->blockmap));
e63644f0 711 hmp->copy_stat_freebigblocks = rootvol->ondisk->vol0_stat_freebigblocks;
0729c8c8 712
059819e3
MD
713 hammer_flusher_create(hmp);
714
a89aec1b 715 /*
27ea2398
MD
716 * Locate the root directory using the root cluster's B-Tree as a
717 * starting point. The root directory uses an obj_id of 1.
718 *
719 * FUTURE: Leave the root directory cached referenced but unlocked
720 * in hmp->rootvp (need to flush it on unmount).
721 */
b9b0a6d0 722 error = hammer_vfs_vget(mp, NULL, 1, &rootvp);
a89aec1b
MD
723 if (error)
724 goto done;
725 vput(rootvp);
27ea2398 726 /*vn_unlock(hmp->rootvp);*/
88c39f64
TN
727 if (hmp->ronly == 0)
728 error = hammer_recover_stage2(hmp, rootvol);
427e5fc6 729
9fa352a2
MD
730 /*
731 * If the stage2 recovery fails be sure to clean out all cached
732 * vnodes before throwing away the mount structure or bad things
733 * will happen.
734 */
735 if (error)
736 vflush(mp, 0, 0);
737
a89aec1b 738done:
f90dde4c
MD
739 hammer_rel_volume(rootvol, 0);
740failed:
427e5fc6 741 /*
27ea2398 742 * Cleanup and return.
427e5fc6 743 */
27ea2398 744 if (error)
427e5fc6 745 hammer_free_hmp(mp);
427e5fc6
MD
746 return (error);
747}
748
749static int
750hammer_vfs_unmount(struct mount *mp, int mntflags)
751{
752#if 0
753 struct hammer_mount *hmp = (void *)mp->mnt_data;
754#endif
755 int flags;
66325755 756 int error;
27ea2398
MD
757
758 /*
427e5fc6
MD
759 * Clean out the vnodes
760 */
66325755
MD
761 flags = 0;
762 if (mntflags & MNT_FORCE)
763 flags |= FORCECLOSE;
764 if ((error = vflush(mp, 0, flags)) != 0)
765 return (error);
427e5fc6
MD
766
767 /*
768 * Clean up the internal mount structure and related entities. This
769 * may issue I/O.
770 */
771 hammer_free_hmp(mp);
772 return(0);
773}
774
775/*
776 * Clean up the internal mount structure and disassociate it from the mount.
777 * This may issue I/O.
778 */
779static void
780hammer_free_hmp(struct mount *mp)
781{
782 struct hammer_mount *hmp = (void *)mp->mnt_data;
cdb6e4e6 783 hammer_flush_group_t flg;
1b0ab2c3 784 int count;
d40bfeca 785 int dummy;
427e5fc6
MD
786
787 /*
cdb6e4e6
MD
788 * Flush anything dirty. This won't even run if the
789 * filesystem errored-out.
427e5fc6 790 */
1b0ab2c3
MD
791 count = 0;
792 while (hammer_flusher_haswork(hmp)) {
793 hammer_flusher_sync(hmp);
794 ++count;
795 if (count >= 5) {
796 if (count == 5)
797 kprintf("HAMMER: umount flushing.");
798 else
799 kprintf(".");
d40bfeca 800 tsleep(&dummy, 0, "hmrufl", hz);
1b0ab2c3
MD
801 }
802 if (count == 30) {
803 kprintf("giving up\n");
804 break;
805 }
806 }
807 if (count >= 5 && count < 30)
808 kprintf("\n");
427e5fc6 809
cdb6e4e6
MD
810 /*
811 * If the mount had a critical error we have to destroy any
812 * remaining inodes before we can finish cleaning up the flusher.
813 */
814 if (hmp->flags & HAMMER_MOUNT_CRITICAL_ERROR) {
815 RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
816 hammer_destroy_inode_callback, NULL);
817 }
b84de5af 818
427e5fc6 819 /*
cdb6e4e6
MD
820 * There shouldn't be any inodes left now and any left over
821 * flush groups should now be empty.
427e5fc6 822 */
cdb6e4e6
MD
823 KKASSERT(RB_EMPTY(&hmp->rb_inos_root));
824 while ((flg = TAILQ_FIRST(&hmp->flush_group_list)) != NULL) {
825 TAILQ_REMOVE(&hmp->flush_group_list, flg, flush_entry);
ff003b11 826 KKASSERT(RB_EMPTY(&flg->flush_tree));
cdb6e4e6
MD
827 if (flg->refs) {
828 kprintf("HAMMER: Warning, flush_group %p was "
829 "not empty on umount!\n", flg);
830 }
bac808fe 831 kfree(flg, hmp->m_misc);
cdb6e4e6 832 }
427e5fc6
MD
833
834 /*
cdb6e4e6 835 * We can finally destroy the flusher
427e5fc6 836 */
cdb6e4e6
MD
837 hammer_flusher_destroy(hmp);
838
b84de5af 839 /*
00f16fad
MD
840 * We may have held recovered buffers due to a read-only mount.
841 * These must be discarded.
842 */
843 if (hmp->ronly)
844 hammer_recover_flush_buffers(hmp, NULL, -1);
845
846 /*
0832c9bb 847 * Unload buffers and then volumes
b84de5af 848 */
0832c9bb
MD
849 RB_SCAN(hammer_buf_rb_tree, &hmp->rb_bufs_root, NULL,
850 hammer_unload_buffer, NULL);
427e5fc6
MD
851 RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
852 hammer_unload_volume, NULL);
853
854 mp->mnt_data = NULL;
66325755 855 mp->mnt_flag &= ~MNT_LOCAL;
427e5fc6 856 hmp->mp = NULL;
0729c8c8 857 hammer_destroy_objid_cache(hmp);
bac808fe
MD
858 kmalloc_destroy(&hmp->m_misc);
859 kmalloc_destroy(&hmp->m_inodes);
427e5fc6
MD
860 kfree(hmp, M_HAMMER);
861}
862
863/*
cdb6e4e6
MD
864 * Report critical errors. ip may be NULL.
865 */
866void
867hammer_critical_error(hammer_mount_t hmp, hammer_inode_t ip,
868 int error, const char *msg)
869{
870 hmp->flags |= HAMMER_MOUNT_CRITICAL_ERROR;
c9ce54d6 871
cdb6e4e6 872 krateprintf(&hmp->krate,
c9ce54d6
MD
873 "HAMMER(%s): Critical error inode=%jd error=%d %s\n",
874 hmp->mp->mnt_stat.f_mntfromname,
875 (intmax_t)(ip ? ip->obj_id : -1),
876 error, msg);
877
cdb6e4e6
MD
878 if (hmp->ronly == 0) {
879 hmp->ronly = 2; /* special errored read-only mode */
880 hmp->mp->mnt_flag |= MNT_RDONLY;
881 kprintf("HAMMER(%s): Forcing read-only mode\n",
882 hmp->mp->mnt_stat.f_mntfromname);
883 }
884 hmp->error = error;
fc73edd8 885 if (hammer_debug_critical)
c9ce54d6 886 Debugger("Entering debugger");
cdb6e4e6
MD
887}
888
889
890/*
513ca7d7
MD
891 * Obtain a vnode for the specified inode number. An exclusively locked
892 * vnode is returned.
893 */
894int
b9b0a6d0
MD
895hammer_vfs_vget(struct mount *mp, struct vnode *dvp,
896 ino_t ino, struct vnode **vpp)
513ca7d7 897{
36f82b23 898 struct hammer_transaction trans;
513ca7d7
MD
899 struct hammer_mount *hmp = (void *)mp->mnt_data;
900 struct hammer_inode *ip;
901 int error;
b9b0a6d0 902 u_int32_t localization;
513ca7d7 903
36f82b23
MD
904 hammer_simple_transaction(&trans, hmp);
905
513ca7d7 906 /*
b9b0a6d0
MD
907 * If a directory vnode is supplied (mainly NFS) then we can acquire
908 * the PFS domain from it. Otherwise we would only be able to vget
909 * inodes in the root PFS.
910 */
911 if (dvp) {
912 localization = HAMMER_DEF_LOCALIZATION +
913 VTOI(dvp)->obj_localization;
914 } else {
915 localization = HAMMER_DEF_LOCALIZATION;
916 }
917
918 /*
36f82b23
MD
919 * Lookup the requested HAMMER inode. The structure must be
920 * left unlocked while we manipulate the related vnode to avoid
921 * a deadlock.
513ca7d7 922 */
ddfdf542 923 ip = hammer_get_inode(&trans, NULL, ino,
b9b0a6d0 924 hmp->asof, localization,
ddfdf542 925 0, &error);
513ca7d7
MD
926 if (ip == NULL) {
927 *vpp = NULL;
00f16fad 928 hammer_done_transaction(&trans);
513ca7d7
MD
929 return(error);
930 }
e8599db1 931 error = hammer_get_vnode(ip, vpp);
513ca7d7 932 hammer_rel_inode(ip, 0);
b84de5af 933 hammer_done_transaction(&trans);
513ca7d7
MD
934 return (error);
935}
936
937/*
427e5fc6
MD
938 * Return the root vnode for the filesystem.
939 *
940 * HAMMER stores the root vnode in the hammer_mount structure so
941 * getting it is easy.
942 */
943static int
944hammer_vfs_root(struct mount *mp, struct vnode **vpp)
945{
47197d71 946#if 0
427e5fc6 947 struct hammer_mount *hmp = (void *)mp->mnt_data;
47197d71 948#endif
27ea2398 949 int error;
427e5fc6 950
b9b0a6d0 951 error = hammer_vfs_vget(mp, NULL, 1, vpp);
27ea2398 952 return (error);
427e5fc6
MD
953}
954
955static int
956hammer_vfs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
957{
fbc6e32a
MD
958 struct hammer_mount *hmp = (void *)mp->mnt_data;
959 hammer_volume_t volume;
960 hammer_volume_ondisk_t ondisk;
961 int error;
47197d71 962 int64_t bfree;
31a56ce2 963 int64_t breserved;
fbc6e32a
MD
964
965 volume = hammer_get_root_volume(hmp, &error);
966 if (error)
967 return(error);
fbc6e32a
MD
968 ondisk = volume->ondisk;
969
47197d71
MD
970 /*
971 * Basic stats
972 */
31a56ce2 973 _hammer_checkspace(hmp, HAMMER_CHKSPC_WRITE, &breserved);
fbc6e32a 974 mp->mnt_stat.f_files = ondisk->vol0_stat_inodes;
c3be93f2 975 bfree = ondisk->vol0_stat_freebigblocks * HAMMER_LARGEBLOCK_SIZE;
40043e7f 976 hammer_rel_volume(volume, 0);
47197d71 977
31a56ce2 978 mp->mnt_stat.f_bfree = (bfree - breserved) / HAMMER_BUFSIZE;
47197d71 979 mp->mnt_stat.f_bavail = mp->mnt_stat.f_bfree;
fbc6e32a
MD
980 if (mp->mnt_stat.f_files < 0)
981 mp->mnt_stat.f_files = 0;
982
27ea2398
MD
983 *sbp = mp->mnt_stat;
984 return(0);
427e5fc6
MD
985}
986
6f97fce3
MD
987static int
988hammer_vfs_statvfs(struct mount *mp, struct statvfs *sbp, struct ucred *cred)
989{
990 struct hammer_mount *hmp = (void *)mp->mnt_data;
991 hammer_volume_t volume;
992 hammer_volume_ondisk_t ondisk;
993 int error;
994 int64_t bfree;
0f65be10 995 int64_t breserved;
6f97fce3
MD
996
997 volume = hammer_get_root_volume(hmp, &error);
998 if (error)
999 return(error);
1000 ondisk = volume->ondisk;
1001
1002 /*
1003 * Basic stats
1004 */
0f65be10 1005 _hammer_checkspace(hmp, HAMMER_CHKSPC_WRITE, &breserved);
6f97fce3
MD
1006 mp->mnt_vstat.f_files = ondisk->vol0_stat_inodes;
1007 bfree = ondisk->vol0_stat_freebigblocks * HAMMER_LARGEBLOCK_SIZE;
1008 hammer_rel_volume(volume, 0);
1009
0f65be10 1010 mp->mnt_vstat.f_bfree = (bfree - breserved) / HAMMER_BUFSIZE;
c0763659 1011 mp->mnt_vstat.f_bavail = mp->mnt_vstat.f_bfree;
6f97fce3
MD
1012 if (mp->mnt_vstat.f_files < 0)
1013 mp->mnt_vstat.f_files = 0;
1014 *sbp = mp->mnt_vstat;
1015 return(0);
1016}
1017
0729c8c8
MD
1018/*
1019 * Sync the filesystem. Currently we have to run it twice, the second
1020 * one will advance the undo start index to the end index, so if a crash
1021 * occurs no undos will be run on mount.
77062c8a
MD
1022 *
1023 * We do not sync the filesystem if we are called from a panic. If we did
1024 * we might end up blowing up a sync that was already in progress.
0729c8c8 1025 */
427e5fc6
MD
1026static int
1027hammer_vfs_sync(struct mount *mp, int waitfor)
1028{
fbc6e32a 1029 struct hammer_mount *hmp = (void *)mp->mnt_data;
0729c8c8
MD
1030 int error;
1031
77062c8a 1032 if (panicstr == NULL) {
0729c8c8 1033 error = hammer_sync_hmp(hmp, waitfor);
77062c8a
MD
1034 } else {
1035 error = EIO;
77062c8a 1036 }
0729c8c8 1037 return (error);
427e5fc6
MD
1038}
1039
513ca7d7
MD
1040/*
1041 * Convert a vnode to a file handle.
1042 */
1043static int
1044hammer_vfs_vptofh(struct vnode *vp, struct fid *fhp)
1045{
1046 hammer_inode_t ip;
1047
1048 KKASSERT(MAXFIDSZ >= 16);
1049 ip = VTOI(vp);
1050 fhp->fid_len = offsetof(struct fid, fid_data[16]);
adf01747 1051 fhp->fid_ext = ip->obj_localization >> 16;
513ca7d7
MD
1052 bcopy(&ip->obj_id, fhp->fid_data + 0, sizeof(ip->obj_id));
1053 bcopy(&ip->obj_asof, fhp->fid_data + 8, sizeof(ip->obj_asof));
1054 return(0);
1055}
1056
1057
1058/*
1059 * Convert a file handle back to a vnode.
67863d04
MD
1060 *
1061 * Use rootvp to enforce PFS isolation when a PFS is exported via a
1062 * null mount.
513ca7d7
MD
1063 */
1064static int
67863d04
MD
1065hammer_vfs_fhtovp(struct mount *mp, struct vnode *rootvp,
1066 struct fid *fhp, struct vnode **vpp)
513ca7d7 1067{
36f82b23 1068 struct hammer_transaction trans;
513ca7d7
MD
1069 struct hammer_inode *ip;
1070 struct hammer_inode_info info;
1071 int error;
adf01747 1072 u_int32_t localization;
513ca7d7
MD
1073
1074 bcopy(fhp->fid_data + 0, &info.obj_id, sizeof(info.obj_id));
1075 bcopy(fhp->fid_data + 8, &info.obj_asof, sizeof(info.obj_asof));
67863d04
MD
1076 if (rootvp)
1077 localization = VTOI(rootvp)->obj_localization;
1078 else
1079 localization = (u_int32_t)fhp->fid_ext << 16;
513ca7d7 1080
36f82b23
MD
1081 hammer_simple_transaction(&trans, (void *)mp->mnt_data);
1082
513ca7d7
MD
1083 /*
1084 * Get/allocate the hammer_inode structure. The structure must be
1085 * unlocked while we manipulate the related vnode to avoid a
1086 * deadlock.
1087 */
ddfdf542 1088 ip = hammer_get_inode(&trans, NULL, info.obj_id,
adf01747 1089 info.obj_asof, localization, 0, &error);
d978e7cf
MD
1090 if (ip) {
1091 error = hammer_get_vnode(ip, vpp);
1092 hammer_rel_inode(ip, 0);
1093 } else {
513ca7d7 1094 *vpp = NULL;
513ca7d7 1095 }
b84de5af 1096 hammer_done_transaction(&trans);
513ca7d7
MD
1097 return (error);
1098}
1099
1100static int
1101hammer_vfs_checkexp(struct mount *mp, struct sockaddr *nam,
1102 int *exflagsp, struct ucred **credanonp)
1103{
1104 hammer_mount_t hmp = (void *)mp->mnt_data;
1105 struct netcred *np;
1106 int error;
1107
1108 np = vfs_export_lookup(mp, &hmp->export, nam);
1109 if (np) {
1110 *exflagsp = np->netc_exflags;
1111 *credanonp = &np->netc_anon;
1112 error = 0;
1113 } else {
1114 error = EACCES;
1115 }
1116 return (error);
1117
1118}
1119
1120int
1121hammer_vfs_export(struct mount *mp, int op, const struct export_args *export)
1122{
1123 hammer_mount_t hmp = (void *)mp->mnt_data;
1124 int error;
1125
1126 switch(op) {
1127 case MOUNTCTL_SET_EXPORT:
1128 error = vfs_export(mp, &hmp->export, export);
1129 break;
1130 default:
1131 error = EOPNOTSUPP;
1132 break;
1133 }
1134 return(error);
1135}
1136