EXT2FS: support variable inode size.
[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);
a6ee311a 90static int ext2_vget (struct mount *, 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
130 error = VFS_VGET(mp, (ino_t)ROOTINO, &nvp);
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);
acde96db
MD
328 error = VOP_ACCESS(devvp, VREAD | VWRITE, cred);
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);
87de5057 392 if ((error = VOP_ACCESS(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);
0e9b9130 762 if (count_udev(devvp->v_umajor, devvp->v_uminor) > 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;
2ec8fb79 782 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred) != 0) {
984263bc 783 size = DEV_BSIZE;
2ec8fb79
MD
784 } else {
785 size = dpart.media_blksize;
984263bc
MD
786 }
787
788 bp = NULL;
789 ump = NULL;
2d5c9276 790 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
984263bc
MD
791 goto out;
792 es = (struct ext2_super_block *)bp->b_data;
793 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
794 error = EINVAL; /* XXX needs translation */
795 goto out;
796 }
797 if ((es->s_state & EXT2_VALID_FS) == 0 ||
798 (es->s_state & EXT2_ERROR_FS)) {
799 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
086c1d7e 800 kprintf(
984263bc
MD
801"WARNING: Filesystem was not properly dismounted\n");
802 } else {
086c1d7e 803 kprintf(
984263bc
MD
804"WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
805 error = EPERM;
806 goto out;
807 }
808 }
e7b4468c 809 ump = bsd_malloc(sizeof *ump, M_EXT2MNT, M_WAITOK | M_ZERO);
984263bc
MD
810 ump->um_malloctype = M_EXT2NODE;
811 ump->um_blkatoff = ext2_blkatoff;
812 ump->um_truncate = ext2_truncate;
813 ump->um_update = ext2_update;
814 ump->um_valloc = ext2_valloc;
815 ump->um_vfree = ext2_vfree;
816 /* I don't know whether this is the right strategy. Note that
817 we dynamically allocate both a ext2_sb_info and a ext2_super_block
818 while Linux keeps the super block in a locked buffer
819 */
820 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
1f1db49f 821 M_EXT2MNT, M_WAITOK);
984263bc 822 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
1f1db49f 823 M_EXT2MNT, M_WAITOK);
984263bc
MD
824 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
825 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
826 goto out;
827 /*
828 * We don't free the group descriptors allocated by compute_sb_data()
829 * until ext2_unmount(). This is OK since the mount will succeed.
830 */
831 brelse(bp);
832 bp = NULL;
833 fs = ump->um_e2fs;
834 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
835 /* if the fs is not mounted read-only, make sure the super block is
836 always written back on a sync()
837 */
838 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
839 if (ronly == 0) {
840 fs->s_dirt = 1; /* mark it modified */
841 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
842 }
843 mp->mnt_data = (qaddr_t)ump;
844 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
845 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
846 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
847 mp->mnt_flag |= MNT_LOCAL;
848 ump->um_mountp = mp;
849 ump->um_dev = dev;
850 ump->um_devvp = devvp;
851 /* setting those two parameters allows us to use
1f1db49f 852 ext2_bmap w/o changse !
984263bc
MD
853 */
854 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
855 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
856 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
857 for (i = 0; i < MAXQUOTAS; i++)
858 ump->um_quotas[i] = NULLVP;
e4c9c0c8 859 dev->si_mountpoint = mp;
0961aa92 860
66a1ddf5
MD
861 vfs_add_vnodeops(mp, &ext2_vnode_vops, &mp->mnt_vn_norm_ops);
862 vfs_add_vnodeops(mp, &ext2_spec_vops, &mp->mnt_vn_spec_ops);
863 vfs_add_vnodeops(mp, &ext2_fifo_vops, &mp->mnt_vn_fifo_ops);
0961aa92 864
984263bc
MD
865 if (ronly == 0)
866 ext2_sbupdate(ump, MNT_WAIT);
867 return (0);
868out:
869 if (bp)
870 brelse(bp);
87de5057 871 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE);
984263bc 872 if (ump) {
1f1db49f
MD
873 bsd_free(ump->um_e2fs->s_es, M_EXT2MNT);
874 bsd_free(ump->um_e2fs, M_EXT2MNT);
875 bsd_free(ump, M_EXT2MNT);
984263bc
MD
876 mp->mnt_data = (qaddr_t)0;
877 }
878 return (error);
879}
880
881/*
882 * unmount system call
883 */
884static int
acde96db 885ext2_unmount(struct mount *mp, int mntflags)
984263bc 886{
1f1db49f 887 struct ext2mount *ump;
f7aae92f 888 struct ext2_sb_info *fs;
984263bc
MD
889 int error, flags, ronly, i;
890
891 flags = 0;
892 if (mntflags & MNT_FORCE) {
893 if (mp->mnt_flag & MNT_ROOTFS)
894 return (EINVAL);
895 flags |= FORCECLOSE;
896 }
f94bf21c 897 if ((error = ext2_flushfiles(mp, flags)) != 0)
984263bc 898 return (error);
1f1db49f 899 ump = VFSTOEXT2(mp);
984263bc
MD
900 fs = ump->um_e2fs;
901 ronly = fs->s_rd_only;
902 if (ronly == 0) {
903 if (fs->s_wasvalid)
904 fs->s_es->s_state |= EXT2_VALID_FS;
905 ext2_sbupdate(ump, MNT_WAIT);
906 }
907
908 /* release buffers containing group descriptors */
909 for(i = 0; i < fs->s_db_per_group; i++)
910 ULCK_BUF(fs->s_group_desc[i])
1f1db49f 911 bsd_free(fs->s_group_desc, M_EXT2MNT);
984263bc
MD
912
913 /* release cached inode/block bitmaps */
914 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
915 if (fs->s_inode_bitmap[i])
916 ULCK_BUF(fs->s_inode_bitmap[i])
917
918 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
919 if (fs->s_block_bitmap[i])
920 ULCK_BUF(fs->s_block_bitmap[i])
921
e4c9c0c8 922 ump->um_devvp->v_rdev->si_mountpoint = NULL;
87de5057 923 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE);
984263bc 924 vrele(ump->um_devvp);
1f1db49f
MD
925 bsd_free(fs->s_es, M_EXT2MNT);
926 bsd_free(fs, M_EXT2MNT);
927 bsd_free(ump, M_EXT2MNT);
984263bc
MD
928 mp->mnt_data = (qaddr_t)0;
929 mp->mnt_flag &= ~MNT_LOCAL;
930 return (error);
931}
932
933/*
934 * Flush out all the files in a filesystem.
935 */
936static int
f94bf21c 937ext2_flushfiles(struct mount *mp, int flags)
984263bc 938{
1f1db49f 939 struct ext2mount *ump;
984263bc
MD
940 int error;
941#if QUOTA
942 int i;
943#endif
944
1f1db49f 945 ump = VFSTOEXT2(mp);
984263bc
MD
946#if QUOTA
947 if (mp->mnt_flag & MNT_QUOTA) {
948 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
949 return (error);
950 for (i = 0; i < MAXQUOTAS; i++) {
951 if (ump->um_quotas[i] == NULLVP)
952 continue;
f94bf21c 953 ext2_quotaoff(mp, i);
984263bc
MD
954 }
955 /*
956 * Here we fall through to vflush again to ensure
957 * that we have gotten rid of all the system vnodes.
958 */
959 }
960#endif
961 error = vflush(mp, 0, flags);
962 return (error);
963}
964
965/*
966 * Get file system statistics.
967 * taken from ext2/super.c ext2_statfs
968 */
969static int
acde96db 970ext2_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
984263bc
MD
971{
972 unsigned long overhead;
1f1db49f 973 struct ext2mount *ump;
f7aae92f
RG
974 struct ext2_sb_info *fs;
975 struct ext2_super_block *es;
984263bc
MD
976 int i, nsb;
977
1f1db49f 978 ump = VFSTOEXT2(mp);
984263bc
MD
979 fs = ump->um_e2fs;
980 es = fs->s_es;
981
982 if (es->s_magic != EXT2_SUPER_MAGIC)
983 panic("ext2_statfs - magic number spoiled");
984
985 /*
986 * Compute the overhead (FS structures)
987 */
988 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
989 nsb = 0;
990 for (i = 0 ; i < fs->s_groups_count; i++)
991 if (ext2_group_sparse(i))
992 nsb++;
993 } else
994 nsb = fs->s_groups_count;
995 overhead = es->s_first_data_block +
996 /* Superblocks and block group descriptors: */
997 nsb * (1 + fs->s_db_per_group) +
998 /* Inode bitmap, block bitmap, and inode table: */
999 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
1000
1001 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
1002 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
1003 sbp->f_blocks = es->s_blocks_count - overhead;
1004 sbp->f_bfree = es->s_free_blocks_count;
1005 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
1006 sbp->f_files = es->s_inodes_count;
1007 sbp->f_ffree = es->s_free_inodes_count;
1008 if (sbp != &mp->mnt_stat) {
1009 sbp->f_type = mp->mnt_vfc->vfc_typenum;
984263bc
MD
1010 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
1011 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
1012 }
1013 return (0);
1014}
1015
1016/*
1017 * Go through the disk queues to initiate sandbagged IO;
1018 * go through the inodes to write those that have been modified;
1019 * initiate the writing of the super block if it has been modified.
1020 *
1021 * Note: we are always called with the filesystem marked `MPBUSY'.
1022 */
41a01a4d 1023
5fd012e0 1024static int ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data);
41a01a4d 1025
984263bc 1026static int
87de5057 1027ext2_sync(struct mount *mp, int waitfor)
984263bc 1028{
1f1db49f 1029 struct ext2mount *ump = VFSTOEXT2(mp);
984263bc 1030 struct ext2_sb_info *fs;
41a01a4d
MD
1031 struct scaninfo scaninfo;
1032 int error;
984263bc
MD
1033
1034 fs = ump->um_e2fs;
1035 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
086c1d7e 1036 kprintf("fs = %s\n", fs->fs_fsmnt);
984263bc
MD
1037 panic("ext2_sync: rofs mod");
1038 }
41a01a4d 1039
984263bc
MD
1040 /*
1041 * Write back each (modified) inode.
1042 */
41a01a4d
MD
1043 scaninfo.allerror = 0;
1044 scaninfo.rescan = 1;
1045 scaninfo.waitfor = waitfor;
41a01a4d
MD
1046 while (scaninfo.rescan) {
1047 scaninfo.rescan = 0;
5fd012e0
MD
1048 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
1049 NULL, ext2_sync_scan, &scaninfo);
984263bc 1050 }
41a01a4d 1051
984263bc
MD
1052 /*
1053 * Force stale file system control information to be flushed.
1054 */
1055 if (waitfor != MNT_LAZY) {
ca466bae 1056 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
87de5057 1057 if ((error = VOP_FSYNC(ump->um_devvp, waitfor)) != 0)
41a01a4d 1058 scaninfo.allerror = error;
a11aaa81 1059 vn_unlock(ump->um_devvp);
984263bc
MD
1060 }
1061#if QUOTA
1f1db49f 1062 ext2_qsync(mp);
984263bc
MD
1063#endif
1064 /*
1065 * Write back modified superblock.
1066 */
1067 if (fs->s_dirt != 0) {
1068 fs->s_dirt = 0;
1069 fs->s_es->s_wtime = time_second;
1070 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
41a01a4d
MD
1071 scaninfo.allerror = error;
1072 }
1073 return (scaninfo.allerror);
1074}
1075
1076static int
5fd012e0 1077ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data)
41a01a4d
MD
1078{
1079 struct scaninfo *info = data;
1080 struct inode *ip;
1081 int error;
1082
1083 ip = VTOI(vp);
1084 if (vp->v_type == VNON ||
1085 ((ip->i_flag &
1086 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
6bae6177 1087 (RB_EMPTY(&vp->v_rbdirty_tree) || info->waitfor == MNT_LAZY))) {
41a01a4d 1088 return(0);
984263bc 1089 }
87de5057 1090 if ((error = VOP_FSYNC(vp, info->waitfor)) != 0)
41a01a4d 1091 info->allerror = error;
41a01a4d 1092 return(0);
984263bc
MD
1093}
1094
1095/*
1096 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
1097 * in from disk. If it is in core, wait for the lock bit to clear, then
1098 * return the inode locked. Detection and handling of mount points must be
1099 * done by the calling routine.
1100 */
1101static int
0f7f7a49 1102ext2_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
984263bc 1103{
f7aae92f
RG
1104 struct ext2_sb_info *fs;
1105 struct inode *ip;
1f1db49f 1106 struct ext2mount *ump;
984263bc
MD
1107 struct buf *bp;
1108 struct vnode *vp;
b13267a5 1109 cdev_t dev;
984263bc
MD
1110 int i, error;
1111 int used_blocks;
1112
1f1db49f 1113 ump = VFSTOEXT2(mp);
984263bc
MD
1114 dev = ump->um_dev;
1115restart:
1f1db49f 1116 if ((*vpp = ext2_ihashget(dev, ino)) != NULL)
984263bc
MD
1117 return (0);
1118
1119 /*
1120 * Lock out the creation of new entries in the FFS hash table in
1121 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1122 * may occur!
1123 */
1124 if (ext2fs_inode_hash_lock) {
1125 while (ext2fs_inode_hash_lock) {
1126 ext2fs_inode_hash_lock = -1;
377d4740 1127 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
984263bc
MD
1128 }
1129 goto restart;
1130 }
1131 ext2fs_inode_hash_lock = 1;
1132
1133 /*
1134 * If this MALLOC() is performed after the getnewvnode()
1135 * it might block, leaving a vnode with a NULL v_data to be
1136 * found by ext2_sync() if a sync happens to fire right then,
1137 * which will cause a panic because ext2_sync() blindly
1138 * dereferences vp->v_data (as well it should).
1139 */
1140 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1141
1142 /* Allocate a new vnode/inode. */
1f1db49f 1143 if ((error = getnewvnode(VT_EXT2FS, mp, &vp, 0, LK_CANRECURSE)) != 0) {
984263bc
MD
1144 if (ext2fs_inode_hash_lock < 0)
1145 wakeup(&ext2fs_inode_hash_lock);
1146 ext2fs_inode_hash_lock = 0;
1147 *vpp = NULL;
1148 FREE(ip, M_EXT2NODE);
1149 return (error);
1150 }
1151 bzero((caddr_t)ip, sizeof(struct inode));
984263bc
MD
1152 vp->v_data = ip;
1153 ip->i_vnode = vp;
1154 ip->i_e2fs = fs = ump->um_e2fs;
1155 ip->i_dev = dev;
1156 ip->i_number = ino;
1157#if QUOTA
1158 for (i = 0; i < MAXQUOTAS; i++)
1159 ip->i_dquot[i] = NODQUOT;
1160#endif
1161 /*
5fd012e0
MD
1162 * Put it onto its hash chain. Since our vnode is locked, other
1163 * requests for this inode will block if they arrive while we are
1164 * sleeping waiting for old data structures to be purged or for the
1165 * contents of the disk portion of this inode to be read.
984263bc 1166 */
1f1db49f 1167 ext2_ihashins(ip);
984263bc
MD
1168
1169 if (ext2fs_inode_hash_lock < 0)
1170 wakeup(&ext2fs_inode_hash_lock);
1171 ext2fs_inode_hash_lock = 0;
1172
1173 /* Read in the disk contents for the inode, copy into the inode. */
1174#if 0
086c1d7e 1175kprintf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
984263bc 1176#endif
54078292
MD
1177 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1178 (int)fs->s_blocksize, &bp);
1179 if (error) {
984263bc
MD
1180 /*
1181 * The inode does not contain anything useful, so it would
1182 * be misleading to leave it on its hash chain. With mode
1183 * still zero, it will be unlinked and returned to the free
1184 * list by vput().
1185 */
5fd012e0 1186 vx_put(vp);
984263bc
MD
1187 brelse(bp);
1188 *vpp = NULL;
1189 return (error);
1190 }
1191 /* convert ext2 inode to dinode */
46932790 1192 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
984263bc
MD
1193 ino_to_fsbo(fs, ino)), &ip->i_din);
1194 ip->i_block_group = ino_to_cg(fs, ino);
1195 ip->i_next_alloc_block = 0;
1196 ip->i_next_alloc_goal = 0;
1197 ip->i_prealloc_count = 0;
1198 ip->i_prealloc_block = 0;
1199 /* now we want to make sure that block pointers for unused
1200 blocks are zeroed out - ext2_balloc depends on this
1201 although for regular files and directories only
1202 */
1203 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1204 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1205 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1206 ip->i_db[i] = 0;
1207 }
933a3a0d 1208#if 0
984263bc 1209 ext2_print_inode(ip);
933a3a0d 1210#endif
984263bc
MD
1211 brelse(bp);
1212
1213 /*
1214 * Initialize the vnode from the inode, check for aliases.
1215 * Note that the underlying vnode may have changed.
1216 */
1f1db49f 1217 if ((error = ext2_vinit(mp, &vp)) != 0) {
5fd012e0 1218 vx_put(vp);
984263bc
MD
1219 *vpp = NULL;
1220 return (error);
1221 }
93f50d60 1222
984263bc
MD
1223 /*
1224 * Finish inode initialization now that aliasing has been resolved.
1225 */
1226 ip->i_devvp = ump->um_devvp;
597aea93 1227 vref(ip->i_devvp);
984263bc
MD
1228 /*
1229 * Set up a generation number for this inode if it does not
1230 * already have one. This should only happen on old filesystems.
1231 */
1232 if (ip->i_gen == 0) {
cddfb7bb 1233 ip->i_gen = krandom() / 2 + 1;
984263bc
MD
1234 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1235 ip->i_flag |= IN_MODIFIED;
1236 }
5fd012e0
MD
1237 /*
1238 * Return the locked and refd vnode.
1239 */
984263bc
MD
1240 *vpp = vp;
1241 return (0);
1242}
1243
1244/*
1245 * File handle to vnode
1246 *
1247 * Have to be really careful about stale file handles:
1248 * - check that the inode number is valid
1249 * - call ext2_vget() to get the locked inode
1250 * - check for an unallocated inode (i_mode == 0)
1251 * - check that the given client host has export rights and return
1252 * those rights via. exflagsp and credanonp
1253 */
1254static int
67863d04
MD
1255ext2_fhtovp(struct mount *mp, struct vnode *rootvp,
1256 struct fid *fhp, struct vnode **vpp)
984263bc 1257{
f7aae92f 1258 struct ufid *ufhp;
984263bc 1259 struct ext2_sb_info *fs;
1f1db49f
MD
1260 struct inode *ip;
1261 struct vnode *nvp;
1262 int error;
984263bc
MD
1263
1264 ufhp = (struct ufid *)fhp;
1f1db49f 1265 fs = VFSTOEXT2(mp)->um_e2fs;
984263bc
MD
1266 if (ufhp->ufid_ino < ROOTINO ||
1267 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1268 return (ESTALE);
1f1db49f
MD
1269
1270 error = VFS_VGET(mp, ufhp->ufid_ino, &nvp);
1271 if (error) {
1272 *vpp = NULLVP;
1273 return (error);
1274 }
1275 ip = VTOI(nvp);
1276 if (ip->i_mode == 0 ||
1277 ip->i_gen != ufhp->ufid_gen ||
1278 (VFSTOEXT2(mp)->um_i_effnlink_valid ? ip->i_effnlink :
1279 ip->i_nlink) <= 0) {
1280 vput(nvp);
1281 *vpp = NULLVP;
1282 return (ESTALE);
1283 }
1284 *vpp = nvp;
1285 return (0);
984263bc
MD
1286}
1287
1288/*
1289 * Vnode pointer to File handle
1290 */
1291/* ARGSUSED */
1292static int
0f7f7a49 1293ext2_vptofh(struct vnode *vp, struct fid *fhp)
984263bc 1294{
f7aae92f
RG
1295 struct inode *ip;
1296 struct ufid *ufhp;
984263bc
MD
1297
1298 ip = VTOI(vp);
1299 ufhp = (struct ufid *)fhp;
1300 ufhp->ufid_len = sizeof(struct ufid);
1301 ufhp->ufid_ino = ip->i_number;
1302 ufhp->ufid_gen = ip->i_gen;
1303 return (0);
1304}
1305
1f1db49f
MD
1306/*
1307 * This is the generic part of fhtovp called after the underlying
1308 * filesystem has validated the file handle.
1309 *
1310 * Verify that a host should have access to a filesystem.
1311 */
1312int
1313ext2_check_export(struct mount *mp, struct sockaddr *nam, int *exflagsp,
1314 struct ucred **credanonp)
1315{
1316 struct netcred *np;
fc6d0222 1317 struct ext2mount *ump;
1f1db49f
MD
1318
1319 ump = VFSTOEXT2(mp);
1320 /*
1321 * Get the export permission structure for this <mp, client> tuple.
1322 */
1323 np = vfs_export_lookup(mp, &ump->um_export, nam);
1324 if (np == NULL)
1325 return (EACCES);
1326
1327 *exflagsp = np->netc_exflags;
1328 *credanonp = &np->netc_anon;
1329 return (0);
1330}
1331
984263bc
MD
1332/*
1333 * Write a superblock and associated information back to disk.
1334 */
1335static int
1f1db49f 1336ext2_sbupdate(struct ext2mount *mp, int waitfor)
984263bc 1337{
f7aae92f
RG
1338 struct ext2_sb_info *fs = mp->um_e2fs;
1339 struct ext2_super_block *es = fs->s_es;
1340 struct buf *bp;
984263bc
MD
1341 int error = 0;
1342/*
086c1d7e 1343kprintf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
984263bc 1344*/
2d5c9276 1345 bp = getblk(mp->um_devvp, SBOFF, SBSIZE, 0, 0);
984263bc
MD
1346 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1347 if (waitfor == MNT_WAIT)
1348 error = bwrite(bp);
1349 else
1350 bawrite(bp);
1351
1352 /*
1353 * The buffers for group descriptors, inode bitmaps and block bitmaps
1354 * are not busy at this point and are (hopefully) written by the
1355 * usual sync mechanism. No need to write them here
0f7f7a49 1356 */
984263bc
MD
1357
1358 return (error);
1359}