kernel - Do a better job with the filesystem background sync
[dragonfly.git] / sys / vfs / gnu / ext2fs / ext2_vfsops.c
CommitLineData
984263bc
MD
1/*
2 * modified for EXT2FS support in Lites 1.1
3 *
4 * Aug 1995, Godmar Back (gback@cs.utah.edu)
5 * University of Utah, Department of Computer Science
6 */
7/*
8 * Copyright (c) 1989, 1991, 1993, 1994
9 * The Regents of the University of California. All rights reserved.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. All advertising materials mentioning features or use of this software
20 * must display the following acknowledgement:
21 * This product includes software developed by the University of
22 * California, Berkeley and its contributors.
23 * 4. Neither the name of the University nor the names of its contributors
24 * may be used to endorse or promote products derived from this software
25 * without specific prior written permission.
26 *
27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * SUCH DAMAGE.
38 *
39 * @(#)ffs_vfsops.c 8.8 (Berkeley) 4/18/94
40 * $FreeBSD: src/sys/gnu/ext2fs/ext2_vfsops.c,v 1.63.2.7 2002/07/01 00:18:51 iedowse Exp $
67863d04 41 * $DragonFly: src/sys/vfs/gnu/ext2fs/ext2_vfsops.c,v 1.57 2008/09/17 21:44:21 dillon Exp $
984263bc
MD
42 */
43
44#include "opt_quota.h"
45
46#include <sys/param.h>
47#include <sys/systm.h>
fad57d0e 48#include <sys/nlookup.h>
984263bc 49#include <sys/proc.h>
895c1f85 50#include <sys/priv.h>
984263bc
MD
51#include <sys/kernel.h>
52#include <sys/vnode.h>
53#include <sys/mount.h>
54#include <sys/buf.h>
55#include <sys/conf.h>
56#include <sys/fcntl.h>
154b688d 57#include <sys/diskslice.h>
984263bc
MD
58#include <sys/malloc.h>
59#include <sys/stat.h>
1f1db49f
MD
60#include <vm/vm_zone.h>
61
7b95be2a 62#include <sys/buf2.h>
61670a01 63#include <sys/thread2.h>
984263bc 64
1f1db49f
MD
65#include "quota.h"
66#include "dinode.h"
67#include "inode.h"
68#include "ext2mount.h"
69#include "ext2_extern.h"
984263bc 70
1f2de5d4 71#include "fs.h"
1f2de5d4
MD
72#include "ext2_fs.h"
73#include "ext2_fs_sb.h"
984263bc 74
66a1ddf5
MD
75extern struct vop_ops ext2_vnode_vops;
76extern struct vop_ops ext2_spec_vops;
77extern struct vop_ops ext2_fifo_vops;
0961aa92 78
67863d04
MD
79static int ext2_fhtovp (struct mount *, struct vnode *,
80 struct fid *, struct vnode **);
f94bf21c 81static int ext2_flushfiles (struct mount *mp, int flags);
acde96db 82static int ext2_mount (struct mount *, char *, caddr_t, struct ucred *);
fef8985e 83static int ext2_mountfs (struct vnode *, struct mount *, struct ucred *);
1f1db49f 84static int ext2_root(struct mount *, struct vnode **);
f94bf21c 85static int ext2_reload (struct mount *mountp, struct ucred *cred);
1f1db49f 86static int ext2_sbupdate (struct ext2mount *, int);
acde96db 87static int ext2_statfs (struct mount *, struct statfs *, struct ucred *);
87de5057 88static int ext2_sync (struct mount *, int);
acde96db 89static int ext2_unmount (struct mount *, int);
b9b0a6d0 90static int ext2_vget (struct mount *, struct vnode *, ino_t, struct vnode **);
1f1db49f 91static int ext2_init(struct vfsconf *);
a6ee311a 92static int ext2_vptofh (struct vnode *, struct fid *);
984263bc
MD
93
94static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part");
1f1db49f 95MALLOC_DEFINE(M_EXT2MNT, "EXT2 mount", "EXT2 mount structure");
984263bc
MD
96
97static struct vfsops ext2fs_vfsops = {
43c45e8f
HP
98 .vfs_mount = ext2_mount,
99 .vfs_unmount = ext2_unmount,
1f1db49f
MD
100 .vfs_root = ext2_root, /* root inode via vget */
101 .vfs_quotactl = ext2_quotactl, /* quota operations */
43c45e8f
HP
102 .vfs_statfs = ext2_statfs,
103 .vfs_sync = ext2_sync,
104 .vfs_vget = ext2_vget,
105 .vfs_fhtovp = ext2_fhtovp,
1f1db49f 106 .vfs_checkexp = ext2_check_export,
43c45e8f 107 .vfs_vptofh = ext2_vptofh,
1f1db49f
MD
108 .vfs_init = ext2_init,
109 .vfs_uninit = ext2_uninit
984263bc
MD
110};
111
112VFS_SET(ext2fs_vfsops, ext2fs, 0);
77652cad
MD
113#define bsd_malloc kmalloc
114#define bsd_free kfree
984263bc
MD
115
116static int ext2fs_inode_hash_lock;
117
a6ee311a 118static int ext2_check_sb_compat (struct ext2_super_block *es,
b13267a5 119 cdev_t dev, int ronly);
a6ee311a 120static int compute_sb_data (struct vnode * devvp,
984263bc 121 struct ext2_super_block * es,
a6ee311a 122 struct ext2_sb_info * fs);
984263bc 123
1f1db49f
MD
124static int
125ext2_root(struct mount *mp, struct vnode **vpp)
126{
127 struct vnode *nvp;
128 int error;
129
b9b0a6d0 130 error = VFS_VGET(mp, NULL, (ino_t)ROOTINO, &nvp);
1f1db49f
MD
131 if (error)
132 return (error);
133 *vpp = nvp;
134 return (0);
135}
136
137/*
138 * Do operations associated with quotas
139 */
140int
141ext2_quotactl(struct mount *mp, int cmds, uid_t uid, caddr_t arg,
acde96db 142 struct ucred *cred)
1f1db49f
MD
143{
144#ifndef QUOTA
145 return (EOPNOTSUPP);
146#else
1f1db49f
MD
147 int cmd, type, error;
148
1f1db49f
MD
149 type = cmds & SUBCMDMASK;
150 cmd = cmds >> SUBCMDSHIFT;
151
152 if (uid == -1) {
153 switch(type) {
154 case USRQUOTA:
acde96db 155 uid = cred->cr_ruid;
1f1db49f
MD
156 break;
157 case GRPQUOTA:
acde96db 158 uid = cred->cr_rgid;
1f1db49f
MD
159 break;
160 default:
161 return (EINVAL);
162 }
163 }
98b480de
MN
164
165 /*
166 * Check permissions.
167 */
1f1db49f 168 switch (cmd) {
98b480de
MN
169
170 case Q_QUOTAON:
171 error = priv_check_cred(cred, PRIV_UFS_QUOTAON, 0);
172 break;
173
174 case Q_QUOTAOFF:
175 error = priv_check_cred(cred, PRIV_UFS_QUOTAOFF, 0);
176 break;
177
178 case Q_SETQUOTA:
179 error = priv_check_cred(cred, PRIV_VFS_SETQUOTA, 0);
180 break;
181
182 case Q_SETUSE:
183 error = priv_check_cred(cred, PRIV_UFS_SETUSE, 0);
1f1db49f 184 break;
98b480de 185
1f1db49f 186 case Q_GETQUOTA:
acde96db 187 if (uid == cred->cr_ruid)
98b480de
MN
188 error = 0;
189 else
190 error = priv_check_cred(cred, PRIV_VFS_GETQUOTA, 0);
191 break;
192
193 case Q_SYNC:
194 error = 0;
195 break;
196
1f1db49f 197 default:
98b480de
MN
198 error = EINVAL;
199 break;
1f1db49f
MD
200 }
201
98b480de
MN
202 if (error)
203 return (error);
204
205
1f1db49f
MD
206 if ((uint)type >= MAXQUOTAS)
207 return (EINVAL);
f9642f56 208 if (vfs_busy(mp, LK_NOWAIT))
1f1db49f
MD
209 return (0);
210
211 switch (cmd) {
212
213 case Q_QUOTAON:
acde96db 214 error = ext2_quotaon(cred, mp, type, arg);
1f1db49f
MD
215 break;
216
217 case Q_QUOTAOFF:
f94bf21c 218 error = ext2_quotaoff(mp, type);
1f1db49f
MD
219 break;
220
221 case Q_SETQUOTA:
222 error = ext2_setquota(mp, uid, type, arg);
223 break;
224
225 case Q_SETUSE:
226 error = ext2_setuse(mp, uid, type, arg);
227 break;
228
229 case Q_GETQUOTA:
230 error = ext2_getquota(mp, uid, type, arg);
231 break;
232
233 case Q_SYNC:
234 error = ext2_qsync(mp);
235 break;
236
237 default:
238 error = EINVAL;
239 break;
240 }
f9642f56 241 vfs_unbusy(mp);
1f1db49f
MD
242 return (error);
243#endif
244}
245
246/*
247 * Initial UFS filesystems, done only once.
248 */
249int
250ext2_init(struct vfsconf *vfsp)
251{
252 static int done;
253
254 if (done)
255 return (0);
256 done = 1;
257 ext2_ihashinit();
258#ifdef QUOTA
259 ext2_dqinit();
260#endif
261 return (0);
262}
263
984263bc
MD
264/*
265 * VFS Operations.
266 *
267 * mount system call
b1ce5639
SW
268 *
269 * Parameters:
1f1db49f 270 * data: this is actually a (struct ext2_args *)
984263bc
MD
271 */
272static int
acde96db 273ext2_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
984263bc
MD
274{
275 struct vnode *devvp;
1f1db49f
MD
276 struct ext2_args args;
277 struct ext2mount *ump = 0;
f7aae92f 278 struct ext2_sb_info *fs;
984263bc
MD
279 size_t size;
280 int error, flags;
281 mode_t accessmode;
fad57d0e 282 struct nlookupdata nd;
984263bc 283
1f1db49f 284 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ext2_args))) != 0)
984263bc 285 return (error);
7b95be2a 286
984263bc
MD
287 /*
288 * If updating, check whether changing from read-only to
289 * read/write; if there is no device name, that's all we do.
290 */
291 if (mp->mnt_flag & MNT_UPDATE) {
1f1db49f 292 ump = VFSTOEXT2(mp);
984263bc 293 fs = ump->um_e2fs;
8ddc6004 294 devvp = ump->um_devvp;
984263bc
MD
295 error = 0;
296 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
297 flags = WRITECLOSE;
298 if (mp->mnt_flag & MNT_FORCE)
299 flags |= FORCECLOSE;
f9642f56 300 if (vfs_busy(mp, LK_NOWAIT))
984263bc 301 return (EBUSY);
f94bf21c 302 error = ext2_flushfiles(mp, flags);
f9642f56 303 vfs_unbusy(mp);
984263bc
MD
304 if (!error && fs->s_wasvalid) {
305 fs->s_es->s_state |= EXT2_VALID_FS;
306 ext2_sbupdate(ump, MNT_WAIT);
307 }
308 fs->s_rd_only = 1;
ca466bae 309 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
87de5057
MD
310 VOP_OPEN(devvp, FREAD, FSCRED, NULL);
311 VOP_CLOSE(devvp, FREAD|FWRITE);
a11aaa81 312 vn_unlock(devvp);
984263bc
MD
313 }
314 if (!error && (mp->mnt_flag & MNT_RELOAD))
acde96db 315 error = ext2_reload(mp, cred);
984263bc
MD
316 if (error)
317 return (error);
984263bc
MD
318 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
319 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
320 return (EPERM);
321 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
322 /*
323 * If upgrade to read-write by non-root, then verify
324 * that user has necessary permissions on the device.
325 */
7b95be2a 326 if (cred->cr_uid != 0) {
ca466bae 327 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
cb66845a 328 error = VOP_EACCESS(devvp, VREAD | VWRITE, cred);
acde96db 329 if (error) {
a11aaa81 330 vn_unlock(devvp);
984263bc
MD
331 return (error);
332 }
a11aaa81 333 vn_unlock(devvp);
984263bc
MD
334 }
335
336 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
337 (fs->s_es->s_state & EXT2_ERROR_FS)) {
338 if (mp->mnt_flag & MNT_FORCE) {
086c1d7e 339 kprintf(
984263bc
MD
340"WARNING: %s was not properly dismounted\n",
341 fs->fs_fsmnt);
342 } else {
086c1d7e 343 kprintf(
984263bc
MD
344"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
345 fs->fs_fsmnt);
346 return (EPERM);
347 }
348 }
349 fs->s_es->s_state &= ~EXT2_VALID_FS;
350 ext2_sbupdate(ump, MNT_WAIT);
351 fs->s_rd_only = 0;
ca466bae 352 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
87de5057
MD
353 VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL);
354 VOP_CLOSE(devvp, FREAD);
a11aaa81 355 vn_unlock(devvp);
984263bc
MD
356 }
357 if (args.fspec == 0) {
358 /*
359 * Process export requests.
360 */
361 return (vfs_export(mp, &ump->um_export, &args.export));
362 }
363 }
364 /*
365 * Not an update, or updating the name: look up the name
366 * and verify that it refers to a sensible block device.
367 */
fad57d0e
MD
368 devvp = NULL;
369 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
370 if (error == 0)
371 error = nlookup(&nd);
372 if (error == 0)
28623bf9 373 error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
fad57d0e
MD
374 nlookup_done(&nd);
375 if (error)
984263bc 376 return (error);
984263bc
MD
377
378 if (!vn_isdisk(devvp, &error)) {
379 vrele(devvp);
380 return (error);
381 }
382
383 /*
384 * If mount by non-root, then verify that user has necessary
385 * permissions on the device.
386 */
7b95be2a 387 if (cred->cr_uid != 0) {
984263bc
MD
388 accessmode = VREAD;
389 if ((mp->mnt_flag & MNT_RDONLY) == 0)
390 accessmode |= VWRITE;
ca466bae 391 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
cb66845a 392 if ((error = VOP_EACCESS(devvp, accessmode, cred)) != 0) {
984263bc
MD
393 vput(devvp);
394 return (error);
395 }
a11aaa81 396 vn_unlock(devvp);
984263bc
MD
397 }
398
399 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
fef8985e 400 error = ext2_mountfs(devvp, mp, cred);
984263bc
MD
401 } else {
402 if (devvp != ump->um_devvp)
403 error = EINVAL; /* needs translation */
404 else
405 vrele(devvp);
406 }
407 if (error) {
408 vrele(devvp);
409 return (error);
410 }
1f1db49f 411 ump = VFSTOEXT2(mp);
984263bc 412 fs = ump->um_e2fs;
fad57d0e 413 copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
984263bc 414 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
fad57d0e 415 copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size);
984263bc 416 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
acde96db 417 ext2_statfs(mp, &mp->mnt_stat, cred);
984263bc
MD
418 return (0);
419}
420
421/*
422 * checks that the data in the descriptor blocks make sense
423 * this is taken from ext2/super.c
424 */
b1ce5639
SW
425static int
426ext2_check_descriptors(struct ext2_sb_info *sb)
984263bc
MD
427{
428 int i;
429 int desc_block = 0;
430 unsigned long block = sb->s_es->s_first_data_block;
431 struct ext2_group_desc * gdp = NULL;
432
433 /* ext2_debug ("Checking group descriptors"); */
434
435 for (i = 0; i < sb->s_groups_count; i++)
436 {
437 /* examine next descriptor block */
438 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
439 gdp = (struct ext2_group_desc *)
440 sb->s_group_desc[desc_block++]->b_data;
441 if (gdp->bg_block_bitmap < block ||
442 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
443 {
086c1d7e 444 kprintf ("ext2_check_descriptors: "
984263bc
MD
445 "Block bitmap for group %d"
446 " not in group (block %lu)!\n",
447 i, (unsigned long) gdp->bg_block_bitmap);
448 return 0;
449 }
450 if (gdp->bg_inode_bitmap < block ||
451 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
452 {
086c1d7e 453 kprintf ("ext2_check_descriptors: "
984263bc
MD
454 "Inode bitmap for group %d"
455 " not in group (block %lu)!\n",
456 i, (unsigned long) gdp->bg_inode_bitmap);
457 return 0;
458 }
459 if (gdp->bg_inode_table < block ||
460 gdp->bg_inode_table + sb->s_itb_per_group >=
461 block + EXT2_BLOCKS_PER_GROUP(sb))
462 {
086c1d7e 463 kprintf ("ext2_check_descriptors: "
984263bc
MD
464 "Inode table for group %d"
465 " not in group (block %lu)!\n",
466 i, (unsigned long) gdp->bg_inode_table);
467 return 0;
468 }
469 block += EXT2_BLOCKS_PER_GROUP(sb);
470 gdp++;
471 }
472 return 1;
473}
474
475static int
b13267a5 476ext2_check_sb_compat(struct ext2_super_block *es, cdev_t dev, int ronly)
984263bc 477{
984263bc 478 if (es->s_magic != EXT2_SUPER_MAGIC) {
086c1d7e 479 kprintf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
984263bc
MD
480 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
481 return (1);
482 }
483 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
484 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
086c1d7e 485 kprintf(
984263bc
MD
486"WARNING: mount of %s denied due to unsupported optional features\n",
487 devtoname(dev));
488 return (1);
489 }
490 if (!ronly &&
491 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
086c1d7e 492 kprintf(
984263bc
MD
493"WARNING: R/W mount of %s denied due to unsupported optional features\n",
494 devtoname(dev));
495 return (1);
496 }
497 }
498 return (0);
499}
500
501/*
502 * this computes the fields of the ext2_sb_info structure from the
503 * data in the ext2_super_block structure read in
504 */
0f7f7a49
CP
505static int
506compute_sb_data(struct vnode *devvp, struct ext2_super_block *es,
507 struct ext2_sb_info *fs)
984263bc
MD
508{
509 int db_count, error;
510 int i, j;
511 int logic_sb_block = 1; /* XXX for now */
512
513#if 1
514#define V(v)
515#else
086c1d7e 516#define V(v) kprintf(#v"= %d\n", fs->v);
984263bc
MD
517#endif
518
519 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
520 V(s_blocksize)
521 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
522 V(s_bshift)
523 fs->s_fsbtodb = es->s_log_block_size + 1;
524 V(s_fsbtodb)
525 fs->s_qbmask = fs->s_blocksize - 1;
526 V(s_bmask)
527 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
528 V(s_blocksize_bits)
529 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
530 V(s_frag_size)
531 if (fs->s_frag_size)
532 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
533 V(s_frags_per_block)
534 fs->s_blocks_per_group = es->s_blocks_per_group;
535 V(s_blocks_per_group)
536 fs->s_frags_per_group = es->s_frags_per_group;
537 V(s_frags_per_group)
538 fs->s_inodes_per_group = es->s_inodes_per_group;
539 V(s_inodes_per_group)
46932790
MD
540 if (es->s_rev_level == EXT2_GOOD_OLD_REV) {
541 fs->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
542 fs->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
543 } else {
544 fs->s_first_ino = es->s_first_ino;
545 fs->s_inode_size = es->s_inode_size;
546 /*
547 * Simple sanity check for superblock inode size value.
548 */
549 if (fs->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE ||
550 fs->s_inode_size > fs->s_blocksize ||
551 (fs->s_inode_size & (fs->s_inode_size - 1)) != 0) {
552 kprintf("EXT2-fs: invalid inode size %d\n",
553 fs->s_inode_size);
554 return (EIO);
555 }
556 }
557 V(s_first_ino)
558 V(s_inode_size)
559 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE(fs);
984263bc
MD
560 V(s_inodes_per_block)
561 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
562 V(s_itb_per_group)
563 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
564 V(s_desc_per_block)
565 /* s_resuid / s_resgid ? */
566 fs->s_groups_count = (es->s_blocks_count -
567 es->s_first_data_block +
568 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
569 EXT2_BLOCKS_PER_GROUP(fs);
570 V(s_groups_count)
571 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
572 EXT2_DESC_PER_BLOCK(fs);
573 fs->s_db_per_group = db_count;
574 V(s_db_per_group)
575
576 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
1f1db49f 577 M_EXT2MNT, M_WAITOK);
984263bc
MD
578
579 /* adjust logic_sb_block */
580 if(fs->s_blocksize > SBSIZE)
581 /* Godmar thinks: if the blocksize is greater than 1024, then
582 the superblock is logically part of block zero.
583 */
584 logic_sb_block = 0;
585
586 for (i = 0; i < db_count; i++) {
54078292
MD
587 error = bread(devvp, fsbtodoff(fs, logic_sb_block + i + 1),
588 fs->s_blocksize, &fs->s_group_desc[i]);
984263bc
MD
589 if(error) {
590 for (j = 0; j < i; j++)
591 brelse(fs->s_group_desc[j]);
1f1db49f 592 bsd_free(fs->s_group_desc, M_EXT2MNT);
086c1d7e 593 kprintf("EXT2-fs: unable to read group descriptors (%d)\n", error);
984263bc
MD
594 return EIO;
595 }
596 /* Set the B_LOCKED flag on the buffer, then brelse() it */
597 LCK_BUF(fs->s_group_desc[i])
598 }
599 if(!ext2_check_descriptors(fs)) {
600 for (j = 0; j < db_count; j++)
601 ULCK_BUF(fs->s_group_desc[j])
1f1db49f 602 bsd_free(fs->s_group_desc, M_EXT2MNT);
086c1d7e 603 kprintf("EXT2-fs: (ext2_check_descriptors failure) "
984263bc
MD
604 "unable to read group descriptors\n");
605 return EIO;
606 }
607
608 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
609 fs->s_inode_bitmap_number[i] = 0;
610 fs->s_inode_bitmap[i] = NULL;
611 fs->s_block_bitmap_number[i] = 0;
612 fs->s_block_bitmap[i] = NULL;
613 }
614 fs->s_loaded_inode_bitmaps = 0;
615 fs->s_loaded_block_bitmaps = 0;
616 return 0;
617}
618
619/*
620 * Reload all incore data for a filesystem (used after running fsck on
621 * the root filesystem and finding things to fix). The filesystem must
622 * be mounted read-only.
623 *
624 * Things to do to update the mount:
625 * 1) invalidate all cached meta-data.
626 * 2) re-read superblock from disk.
627 * 3) re-read summary information from disk.
628 * 4) invalidate all inactive vnodes.
629 * 5) invalidate all cached file data.
630 * 6) re-read inode data for all active vnodes.
631 */
41a01a4d 632static int ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *rescan);
5fd012e0 633static int ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *rescan);
41a01a4d
MD
634
635struct scaninfo {
636 int rescan;
637 int allerror;
638 int waitfor;
41a01a4d
MD
639 struct vnode *devvp;
640 struct ext2_sb_info *fs;
641};
642
984263bc 643static int
f94bf21c 644ext2_reload(struct mount *mountp, struct ucred *cred)
984263bc 645{
41a01a4d 646 struct vnode *devvp;
984263bc
MD
647 struct buf *bp;
648 struct ext2_super_block * es;
649 struct ext2_sb_info *fs;
650 int error;
41a01a4d 651 struct scaninfo scaninfo;
984263bc
MD
652
653 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
654 return (EINVAL);
655 /*
656 * Step 1: invalidate all cached meta-data.
657 */
1f1db49f 658 devvp = VFSTOEXT2(mountp)->um_devvp;
87de5057 659 if (vinvalbuf(devvp, 0, 0, 0))
984263bc
MD
660 panic("ext2_reload: dirty1");
661 /*
662 * Step 2: re-read superblock from disk.
663 * constants have been adjusted for ext2
664 */
2d5c9276 665 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
984263bc
MD
666 return (error);
667 es = (struct ext2_super_block *)bp->b_data;
668 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
669 brelse(bp);
670 return (EIO); /* XXX needs translation */
671 }
1f1db49f 672 fs = VFSTOEXT2(mountp)->um_e2fs;
984263bc
MD
673 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
674
675 if((error = compute_sb_data(devvp, es, fs)) != 0) {
676 brelse(bp);
677 return error;
678 }
679#ifdef UNKLAR
680 if (fs->fs_sbsize < SBSIZE)
681 bp->b_flags |= B_INVAL;
682#endif
683 brelse(bp);
684
41a01a4d 685 scaninfo.rescan = 1;
41a01a4d
MD
686 scaninfo.devvp = devvp;
687 scaninfo.fs = fs;
688 while (error == 0 && scaninfo.rescan) {
689 scaninfo.rescan = 0;
5fd012e0 690 error = vmntvnodescan(mountp, VMSC_GETVX, ext2_reload_scan1,
41a01a4d
MD
691 ext2_reload_scan2, &scaninfo);
692 }
693 return(error);
694}
695
696static int
697ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
698{
5fd012e0 699 /*struct scaninfo *info = data;*/
41a01a4d 700
41a01a4d
MD
701 return(0);
702}
703
704static int
5fd012e0 705ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
41a01a4d
MD
706{
707 struct scaninfo *info = data;
708 struct inode *ip;
709 struct buf *bp;
710 int error;
711
712 /*
5fd012e0 713 * Try to recycle
41a01a4d 714 */
87de5057 715 if (vrecycle(vp))
41a01a4d 716 return(0);
5fd012e0
MD
717
718 /*
719 * Step 5: invalidate all cached file data.
720 */
87de5057 721 if (vinvalbuf(vp, 0, 0, 0))
41a01a4d
MD
722 panic("ext2_reload: dirty2");
723 /*
724 * Step 6: re-read inode data for all active vnodes.
725 */
726 ip = VTOI(vp);
54078292
MD
727 error = bread(info->devvp,
728 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
729 (int)info->fs->s_blocksize, &bp);
5fd012e0 730 if (error)
41a01a4d 731 return (error);
41a01a4d 732 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
46932790 733 EXT2_INODE_SIZE(info->fs) * ino_to_fsbo(info->fs, ip->i_number)),
41a01a4d
MD
734 &ip->i_din);
735 brelse(bp);
41a01a4d 736 return(0);
984263bc
MD
737}
738
739/*
740 * Common code for mount and mountroot
741 */
742static int
fef8985e 743ext2_mountfs(struct vnode *devvp, struct mount *mp, struct ucred *cred)
984263bc 744{
1f1db49f 745 struct ext2mount *ump;
984263bc 746 struct buf *bp;
f7aae92f 747 struct ext2_sb_info *fs;
984263bc 748 struct ext2_super_block * es;
b13267a5 749 cdev_t dev;
984263bc 750 struct partinfo dpart;
984263bc
MD
751 int error, i, size;
752 int ronly;
753
754 /*
755 * Disallow multiple mounts of the same device.
756 * Disallow mounting of a device that is currently in use
757 * (except for root, which might share swap device for miniroot).
758 * Flush out any old buffers remaining from a previous use.
759 */
760 if ((error = vfs_mountedon(devvp)) != 0)
761 return (error);
8be7edad 762 if (vcount(devvp) > 0)
984263bc 763 return (EBUSY);
87de5057 764 if ((error = vinvalbuf(devvp, V_SAVE, 0, 0)) != 0)
984263bc
MD
765 return (error);
766#ifdef READONLY
767/* turn on this to force it to be read-only */
768 mp->mnt_flag |= MNT_RDONLY;
769#endif
770
771 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
ca466bae 772 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
87de5057 773 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
a11aaa81 774 vn_unlock(devvp);
984263bc
MD
775 if (error)
776 return (error);
e4c9c0c8
MD
777 dev = devvp->v_rdev;
778 if (dev->si_iosize_max != 0)
779 mp->mnt_iosize_max = dev->si_iosize_max;
984263bc
MD
780 if (mp->mnt_iosize_max > MAXPHYS)
781 mp->mnt_iosize_max = MAXPHYS;
87baaf0c
MD
782 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD,
783 cred, NULL) != 0) {
984263bc 784 size = DEV_BSIZE;
2ec8fb79
MD
785 } else {
786 size = dpart.media_blksize;
984263bc
MD
787 }
788
789 bp = NULL;
790 ump = NULL;
2d5c9276 791 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
984263bc
MD
792 goto out;
793 es = (struct ext2_super_block *)bp->b_data;
794 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
795 error = EINVAL; /* XXX needs translation */
796 goto out;
797 }
798 if ((es->s_state & EXT2_VALID_FS) == 0 ||
799 (es->s_state & EXT2_ERROR_FS)) {
800 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
086c1d7e 801 kprintf(
984263bc
MD
802"WARNING: Filesystem was not properly dismounted\n");
803 } else {
086c1d7e 804 kprintf(
984263bc
MD
805"WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
806 error = EPERM;
807 goto out;
808 }
809 }
e7b4468c 810 ump = bsd_malloc(sizeof *ump, M_EXT2MNT, M_WAITOK | M_ZERO);
984263bc
MD
811 ump->um_malloctype = M_EXT2NODE;
812 ump->um_blkatoff = ext2_blkatoff;
813 ump->um_truncate = ext2_truncate;
814 ump->um_update = ext2_update;
815 ump->um_valloc = ext2_valloc;
816 ump->um_vfree = ext2_vfree;
817 /* I don't know whether this is the right strategy. Note that
818 we dynamically allocate both a ext2_sb_info and a ext2_super_block
819 while Linux keeps the super block in a locked buffer
820 */
821 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
1f1db49f 822 M_EXT2MNT, M_WAITOK);
984263bc 823 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
1f1db49f 824 M_EXT2MNT, M_WAITOK);
984263bc
MD
825 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
826 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
827 goto out;
828 /*
829 * We don't free the group descriptors allocated by compute_sb_data()
830 * until ext2_unmount(). This is OK since the mount will succeed.
831 */
832 brelse(bp);
833 bp = NULL;
834 fs = ump->um_e2fs;
835 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
836 /* if the fs is not mounted read-only, make sure the super block is
837 always written back on a sync()
838 */
839 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
840 if (ronly == 0) {
841 fs->s_dirt = 1; /* mark it modified */
842 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
843 }
844 mp->mnt_data = (qaddr_t)ump;
845 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
846 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
847 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
848 mp->mnt_flag |= MNT_LOCAL;
849 ump->um_mountp = mp;
850 ump->um_dev = dev;
851 ump->um_devvp = devvp;
852 /* setting those two parameters allows us to use
1f1db49f 853 ext2_bmap w/o changse !
984263bc
MD
854 */
855 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
856 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
857 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
858 for (i = 0; i < MAXQUOTAS; i++)
859 ump->um_quotas[i] = NULLVP;
e4c9c0c8 860 dev->si_mountpoint = mp;
0961aa92 861
66a1ddf5
MD
862 vfs_add_vnodeops(mp, &ext2_vnode_vops, &mp->mnt_vn_norm_ops);
863 vfs_add_vnodeops(mp, &ext2_spec_vops, &mp->mnt_vn_spec_ops);
864 vfs_add_vnodeops(mp, &ext2_fifo_vops, &mp->mnt_vn_fifo_ops);
0961aa92 865
984263bc
MD
866 if (ronly == 0)
867 ext2_sbupdate(ump, MNT_WAIT);
868 return (0);
869out:
870 if (bp)
871 brelse(bp);
87de5057 872 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE);
984263bc 873 if (ump) {
1f1db49f
MD
874 bsd_free(ump->um_e2fs->s_es, M_EXT2MNT);
875 bsd_free(ump->um_e2fs, M_EXT2MNT);
876 bsd_free(ump, M_EXT2MNT);
984263bc
MD
877 mp->mnt_data = (qaddr_t)0;
878 }
879 return (error);
880}
881
882/*
883 * unmount system call
884 */
885static int
acde96db 886ext2_unmount(struct mount *mp, int mntflags)
984263bc 887{
1f1db49f 888 struct ext2mount *ump;
f7aae92f 889 struct ext2_sb_info *fs;
984263bc
MD
890 int error, flags, ronly, i;
891
892 flags = 0;
893 if (mntflags & MNT_FORCE) {
894 if (mp->mnt_flag & MNT_ROOTFS)
895 return (EINVAL);
896 flags |= FORCECLOSE;
897 }
f94bf21c 898 if ((error = ext2_flushfiles(mp, flags)) != 0)
984263bc 899 return (error);
1f1db49f 900 ump = VFSTOEXT2(mp);
984263bc
MD
901 fs = ump->um_e2fs;
902 ronly = fs->s_rd_only;
903 if (ronly == 0) {
904 if (fs->s_wasvalid)
905 fs->s_es->s_state |= EXT2_VALID_FS;
906 ext2_sbupdate(ump, MNT_WAIT);
907 }
908
909 /* release buffers containing group descriptors */
910 for(i = 0; i < fs->s_db_per_group; i++)
911 ULCK_BUF(fs->s_group_desc[i])
1f1db49f 912 bsd_free(fs->s_group_desc, M_EXT2MNT);
984263bc
MD
913
914 /* release cached inode/block bitmaps */
915 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
916 if (fs->s_inode_bitmap[i])
917 ULCK_BUF(fs->s_inode_bitmap[i])
918
919 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
920 if (fs->s_block_bitmap[i])
921 ULCK_BUF(fs->s_block_bitmap[i])
922
e4c9c0c8 923 ump->um_devvp->v_rdev->si_mountpoint = NULL;
87de5057 924 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE);
984263bc 925 vrele(ump->um_devvp);
1f1db49f
MD
926 bsd_free(fs->s_es, M_EXT2MNT);
927 bsd_free(fs, M_EXT2MNT);
928 bsd_free(ump, M_EXT2MNT);
984263bc
MD
929 mp->mnt_data = (qaddr_t)0;
930 mp->mnt_flag &= ~MNT_LOCAL;
931 return (error);
932}
933
934/*
935 * Flush out all the files in a filesystem.
936 */
937static int
f94bf21c 938ext2_flushfiles(struct mount *mp, int flags)
984263bc 939{
1f1db49f 940 struct ext2mount *ump;
984263bc
MD
941 int error;
942#if QUOTA
943 int i;
944#endif
945
1f1db49f 946 ump = VFSTOEXT2(mp);
984263bc
MD
947#if QUOTA
948 if (mp->mnt_flag & MNT_QUOTA) {
949 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
950 return (error);
951 for (i = 0; i < MAXQUOTAS; i++) {
952 if (ump->um_quotas[i] == NULLVP)
953 continue;
f94bf21c 954 ext2_quotaoff(mp, i);
984263bc
MD
955 }
956 /*
957 * Here we fall through to vflush again to ensure
958 * that we have gotten rid of all the system vnodes.
959 */
960 }
961#endif
962 error = vflush(mp, 0, flags);
963 return (error);
964}
965
966/*
967 * Get file system statistics.
968 * taken from ext2/super.c ext2_statfs
969 */
970static int
acde96db 971ext2_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
984263bc
MD
972{
973 unsigned long overhead;
1f1db49f 974 struct ext2mount *ump;
f7aae92f
RG
975 struct ext2_sb_info *fs;
976 struct ext2_super_block *es;
984263bc
MD
977 int i, nsb;
978
1f1db49f 979 ump = VFSTOEXT2(mp);
984263bc
MD
980 fs = ump->um_e2fs;
981 es = fs->s_es;
982
983 if (es->s_magic != EXT2_SUPER_MAGIC)
984 panic("ext2_statfs - magic number spoiled");
985
986 /*
987 * Compute the overhead (FS structures)
988 */
989 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
990 nsb = 0;
991 for (i = 0 ; i < fs->s_groups_count; i++)
992 if (ext2_group_sparse(i))
993 nsb++;
994 } else
995 nsb = fs->s_groups_count;
996 overhead = es->s_first_data_block +
997 /* Superblocks and block group descriptors: */
998 nsb * (1 + fs->s_db_per_group) +
999 /* Inode bitmap, block bitmap, and inode table: */
1000 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
1001
1002 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
1003 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
1004 sbp->f_blocks = es->s_blocks_count - overhead;
1005 sbp->f_bfree = es->s_free_blocks_count;
1006 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
1007 sbp->f_files = es->s_inodes_count;
1008 sbp->f_ffree = es->s_free_inodes_count;
1009 if (sbp != &mp->mnt_stat) {
1010 sbp->f_type = mp->mnt_vfc->vfc_typenum;
984263bc
MD
1011 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
1012 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
1013 }
1014 return (0);
1015}
1016
1017/*
1018 * Go through the disk queues to initiate sandbagged IO;
1019 * go through the inodes to write those that have been modified;
1020 * initiate the writing of the super block if it has been modified.
1021 *
1022 * Note: we are always called with the filesystem marked `MPBUSY'.
1023 */
41a01a4d 1024
5fd012e0 1025static int ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data);
41a01a4d 1026
984263bc 1027static int
87de5057 1028ext2_sync(struct mount *mp, int waitfor)
984263bc 1029{
1f1db49f 1030 struct ext2mount *ump = VFSTOEXT2(mp);
984263bc 1031 struct ext2_sb_info *fs;
41a01a4d
MD
1032 struct scaninfo scaninfo;
1033 int error;
984263bc
MD
1034
1035 fs = ump->um_e2fs;
1036 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
086c1d7e 1037 kprintf("fs = %s\n", fs->fs_fsmnt);
984263bc
MD
1038 panic("ext2_sync: rofs mod");
1039 }
41a01a4d 1040
984263bc
MD
1041 /*
1042 * Write back each (modified) inode.
1043 */
41a01a4d
MD
1044 scaninfo.allerror = 0;
1045 scaninfo.rescan = 1;
1046 scaninfo.waitfor = waitfor;
41a01a4d
MD
1047 while (scaninfo.rescan) {
1048 scaninfo.rescan = 0;
5fd012e0
MD
1049 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
1050 NULL, ext2_sync_scan, &scaninfo);
984263bc 1051 }
41a01a4d 1052
984263bc
MD
1053 /*
1054 * Force stale file system control information to be flushed.
1055 */
28271622 1056 if ((waitfor & MNT_LAZY) == 0) {
ca466bae 1057 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
52174f71 1058 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, 0)) != 0)
41a01a4d 1059 scaninfo.allerror = error;
a11aaa81 1060 vn_unlock(ump->um_devvp);
984263bc
MD
1061 }
1062#if QUOTA
1f1db49f 1063 ext2_qsync(mp);
984263bc
MD
1064#endif
1065 /*
1066 * Write back modified superblock.
1067 */
1068 if (fs->s_dirt != 0) {
1069 fs->s_dirt = 0;
1070 fs->s_es->s_wtime = time_second;
1071 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
41a01a4d
MD
1072 scaninfo.allerror = error;
1073 }
1074 return (scaninfo.allerror);
1075}
1076
1077static int
5fd012e0 1078ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data)
41a01a4d
MD
1079{
1080 struct scaninfo *info = data;
1081 struct inode *ip;
1082 int error;
1083
1084 ip = VTOI(vp);
1085 if (vp->v_type == VNON ||
1086 ((ip->i_flag &
1087 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
28271622 1088 (RB_EMPTY(&vp->v_rbdirty_tree) || (info->waitfor & MNT_LAZY)))) {
41a01a4d 1089 return(0);
984263bc 1090 }
52174f71 1091 if ((error = VOP_FSYNC(vp, info->waitfor, 0)) != 0)
41a01a4d 1092 info->allerror = error;
41a01a4d 1093 return(0);
984263bc
MD
1094}
1095
1096/*
1097 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
1098 * in from disk. If it is in core, wait for the lock bit to clear, then
1099 * return the inode locked. Detection and handling of mount points must be
1100 * done by the calling routine.
1101 */
1102static int
b9b0a6d0 1103ext2_vget(struct mount *mp, struct vnode *dvp, ino_t ino, struct vnode **vpp)
984263bc 1104{
f7aae92f
RG
1105 struct ext2_sb_info *fs;
1106 struct inode *ip;
1f1db49f 1107 struct ext2mount *ump;
984263bc
MD
1108 struct buf *bp;
1109 struct vnode *vp;
b13267a5 1110 cdev_t dev;
984263bc
MD
1111 int i, error;
1112 int used_blocks;
1113
1f1db49f 1114 ump = VFSTOEXT2(mp);
984263bc
MD
1115 dev = ump->um_dev;
1116restart:
1f1db49f 1117 if ((*vpp = ext2_ihashget(dev, ino)) != NULL)
984263bc
MD
1118 return (0);
1119
1120 /*
1121 * Lock out the creation of new entries in the FFS hash table in
1122 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1123 * may occur!
1124 */
1125 if (ext2fs_inode_hash_lock) {
1126 while (ext2fs_inode_hash_lock) {
1127 ext2fs_inode_hash_lock = -1;
377d4740 1128 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
984263bc
MD
1129 }
1130 goto restart;
1131 }
1132 ext2fs_inode_hash_lock = 1;
1133
1134 /*
1135 * If this MALLOC() is performed after the getnewvnode()
1136 * it might block, leaving a vnode with a NULL v_data to be
1137 * found by ext2_sync() if a sync happens to fire right then,
1138 * which will cause a panic because ext2_sync() blindly
1139 * dereferences vp->v_data (as well it should).
1140 */
1141 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1142
1143 /* Allocate a new vnode/inode. */
1f1db49f 1144 if ((error = getnewvnode(VT_EXT2FS, mp, &vp, 0, LK_CANRECURSE)) != 0) {
984263bc
MD
1145 if (ext2fs_inode_hash_lock < 0)
1146 wakeup(&ext2fs_inode_hash_lock);
1147 ext2fs_inode_hash_lock = 0;
1148 *vpp = NULL;
1149 FREE(ip, M_EXT2NODE);
1150 return (error);
1151 }
1152 bzero((caddr_t)ip, sizeof(struct inode));
984263bc
MD
1153 vp->v_data = ip;
1154 ip->i_vnode = vp;
1155 ip->i_e2fs = fs = ump->um_e2fs;
1156 ip->i_dev = dev;
1157 ip->i_number = ino;
1158#if QUOTA
1159 for (i = 0; i < MAXQUOTAS; i++)
1160 ip->i_dquot[i] = NODQUOT;
1161#endif
1162 /*
5fd012e0
MD
1163 * Put it onto its hash chain. Since our vnode is locked, other
1164 * requests for this inode will block if they arrive while we are
1165 * sleeping waiting for old data structures to be purged or for the
1166 * contents of the disk portion of this inode to be read.
984263bc 1167 */
1f1db49f 1168 ext2_ihashins(ip);
984263bc
MD
1169
1170 if (ext2fs_inode_hash_lock < 0)
1171 wakeup(&ext2fs_inode_hash_lock);
1172 ext2fs_inode_hash_lock = 0;
1173
1174 /* Read in the disk contents for the inode, copy into the inode. */
1175#if 0
086c1d7e 1176kprintf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
984263bc 1177#endif
54078292
MD
1178 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1179 (int)fs->s_blocksize, &bp);
1180 if (error) {
984263bc
MD
1181 /*
1182 * The inode does not contain anything useful, so it would
1183 * be misleading to leave it on its hash chain. With mode
1184 * still zero, it will be unlinked and returned to the free
1185 * list by vput().
1186 */
984263bc 1187 brelse(bp);
c081fe84 1188 vx_put(vp);
984263bc
MD
1189 *vpp = NULL;
1190 return (error);
1191 }
1192 /* convert ext2 inode to dinode */
46932790 1193 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
984263bc
MD
1194 ino_to_fsbo(fs, ino)), &ip->i_din);
1195 ip->i_block_group = ino_to_cg(fs, ino);
1196 ip->i_next_alloc_block = 0;
1197 ip->i_next_alloc_goal = 0;
1198 ip->i_prealloc_count = 0;
1199 ip->i_prealloc_block = 0;
1200 /* now we want to make sure that block pointers for unused
1201 blocks are zeroed out - ext2_balloc depends on this
1202 although for regular files and directories only
1203 */
1204 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1205 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1206 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1207 ip->i_db[i] = 0;
1208 }
933a3a0d 1209#if 0
984263bc 1210 ext2_print_inode(ip);
933a3a0d 1211#endif
c081fe84 1212 bqrelse(bp);
984263bc
MD
1213
1214 /*
1215 * Initialize the vnode from the inode, check for aliases.
1216 * Note that the underlying vnode may have changed.
1217 */
1f1db49f 1218 if ((error = ext2_vinit(mp, &vp)) != 0) {
5fd012e0 1219 vx_put(vp);
984263bc
MD
1220 *vpp = NULL;
1221 return (error);
1222 }
93f50d60 1223
984263bc
MD
1224 /*
1225 * Finish inode initialization now that aliasing has been resolved.
1226 */
1227 ip->i_devvp = ump->um_devvp;
597aea93 1228 vref(ip->i_devvp);
984263bc
MD
1229 /*
1230 * Set up a generation number for this inode if it does not
1231 * already have one. This should only happen on old filesystems.
1232 */
1233 if (ip->i_gen == 0) {
cddfb7bb 1234 ip->i_gen = krandom() / 2 + 1;
984263bc
MD
1235 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1236 ip->i_flag |= IN_MODIFIED;
1237 }
5fd012e0
MD
1238 /*
1239 * Return the locked and refd vnode.
1240 */
984263bc
MD
1241 *vpp = vp;
1242 return (0);
1243}
1244
1245/*
1246 * File handle to vnode
1247 *
1248 * Have to be really careful about stale file handles:
1249 * - check that the inode number is valid
1250 * - call ext2_vget() to get the locked inode
1251 * - check for an unallocated inode (i_mode == 0)
1252 * - check that the given client host has export rights and return
1253 * those rights via. exflagsp and credanonp
1254 */
1255static int
67863d04
MD
1256ext2_fhtovp(struct mount *mp, struct vnode *rootvp,
1257 struct fid *fhp, struct vnode **vpp)
984263bc 1258{
f7aae92f 1259 struct ufid *ufhp;
984263bc 1260 struct ext2_sb_info *fs;
1f1db49f
MD
1261 struct inode *ip;
1262 struct vnode *nvp;
1263 int error;
984263bc
MD
1264
1265 ufhp = (struct ufid *)fhp;
1f1db49f 1266 fs = VFSTOEXT2(mp)->um_e2fs;
984263bc
MD
1267 if (ufhp->ufid_ino < ROOTINO ||
1268 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1269 return (ESTALE);
1f1db49f 1270
b9b0a6d0 1271 error = VFS_VGET(mp, rootvp, ufhp->ufid_ino, &nvp);
1f1db49f
MD
1272 if (error) {
1273 *vpp = NULLVP;
1274 return (error);
1275 }
1276 ip = VTOI(nvp);
1277 if (ip->i_mode == 0 ||
1278 ip->i_gen != ufhp->ufid_gen ||
1279 (VFSTOEXT2(mp)->um_i_effnlink_valid ? ip->i_effnlink :
1280 ip->i_nlink) <= 0) {
1281 vput(nvp);
1282 *vpp = NULLVP;
1283 return (ESTALE);
1284 }
1285 *vpp = nvp;
1286 return (0);
984263bc
MD
1287}
1288
1289/*
1290 * Vnode pointer to File handle
1291 */
1292/* ARGSUSED */
1293static int
0f7f7a49 1294ext2_vptofh(struct vnode *vp, struct fid *fhp)
984263bc 1295{
f7aae92f
RG
1296 struct inode *ip;
1297 struct ufid *ufhp;
984263bc
MD
1298
1299 ip = VTOI(vp);
1300 ufhp = (struct ufid *)fhp;
1301 ufhp->ufid_len = sizeof(struct ufid);
1302 ufhp->ufid_ino = ip->i_number;
1303 ufhp->ufid_gen = ip->i_gen;
1304 return (0);
1305}
1306
1307/*
1f1db49f
MD
1308 * This is the generic part of fhtovp called after the underlying
1309 * filesystem has validated the file handle.
1310 *
1311 * Verify that a host should have access to a filesystem.
1312 */
1313int
1314ext2_check_export(struct mount *mp, struct sockaddr *nam, int *exflagsp,
1315 struct ucred **credanonp)
1316{
1317 struct netcred *np;
fc6d0222 1318 struct ext2mount *ump;
1f1db49f
MD
1319
1320 ump = VFSTOEXT2(mp);
1321 /*
1322 * Get the export permission structure for this <mp, client> tuple.
1323 */
1324 np = vfs_export_lookup(mp, &ump->um_export, nam);
1325 if (np == NULL)
1326 return (EACCES);
1327
1328 *exflagsp = np->netc_exflags;
1329 *credanonp = &np->netc_anon;
1330 return (0);
1331}
1332
1333/*
984263bc
MD
1334 * Write a superblock and associated information back to disk.
1335 */
1336static int
1f1db49f 1337ext2_sbupdate(struct ext2mount *mp, int waitfor)
984263bc 1338{
f7aae92f
RG
1339 struct ext2_sb_info *fs = mp->um_e2fs;
1340 struct ext2_super_block *es = fs->s_es;
1341 struct buf *bp;
984263bc
MD
1342 int error = 0;
1343/*
086c1d7e 1344kprintf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
984263bc 1345*/
2d5c9276 1346 bp = getblk(mp->um_devvp, SBOFF, SBSIZE, 0, 0);
984263bc
MD
1347 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1348 if (waitfor == MNT_WAIT)
1349 error = bwrite(bp);
1350 else
1351 bawrite(bp);
1352
1353 /*
1354 * The buffers for group descriptors, inode bitmaps and block bitmaps
1355 * are not busy at this point and are (hopefully) written by the
1356 * usual sync mechanism. No need to write them here
0f7f7a49 1357 */
984263bc
MD
1358
1359 return (error);
1360}