kernel: Move GPL'd kernel files to sys/gnu to have them all in one place.
[dragonfly.git] / sys / gnu / vfs / 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 */
b993bb87
SW
7/*
8 * Copyright (c) 1989, 1991, 1993, 1994
984263bc
MD
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 $
41 */
42
43#include "opt_quota.h"
44
45#include <sys/param.h>
46#include <sys/systm.h>
fad57d0e 47#include <sys/nlookup.h>
984263bc 48#include <sys/proc.h>
895c1f85 49#include <sys/priv.h>
984263bc
MD
50#include <sys/kernel.h>
51#include <sys/vnode.h>
52#include <sys/mount.h>
53#include <sys/buf.h>
54#include <sys/conf.h>
55#include <sys/fcntl.h>
154b688d 56#include <sys/diskslice.h>
984263bc
MD
57#include <sys/malloc.h>
58#include <sys/stat.h>
1f1db49f
MD
59#include <vm/vm_zone.h>
60
7b95be2a 61#include <sys/buf2.h>
61670a01 62#include <sys/thread2.h>
984263bc 63
1f1db49f
MD
64#include "quota.h"
65#include "dinode.h"
66#include "inode.h"
67#include "ext2mount.h"
68#include "ext2_extern.h"
984263bc 69
1f2de5d4 70#include "fs.h"
1f2de5d4
MD
71#include "ext2_fs.h"
72#include "ext2_fs_sb.h"
984263bc 73
66a1ddf5
MD
74extern struct vop_ops ext2_vnode_vops;
75extern struct vop_ops ext2_spec_vops;
76extern struct vop_ops ext2_fifo_vops;
0961aa92 77
67863d04
MD
78static int ext2_fhtovp (struct mount *, struct vnode *,
79 struct fid *, struct vnode **);
f94bf21c 80static int ext2_flushfiles (struct mount *mp, int flags);
acde96db 81static int ext2_mount (struct mount *, char *, caddr_t, struct ucred *);
fef8985e 82static int ext2_mountfs (struct vnode *, struct mount *, struct ucred *);
1f1db49f 83static int ext2_root(struct mount *, struct vnode **);
f94bf21c 84static int ext2_reload (struct mount *mountp, struct ucred *cred);
1f1db49f 85static int ext2_sbupdate (struct ext2mount *, int);
acde96db 86static int ext2_statfs (struct mount *, struct statfs *, struct ucred *);
87de5057 87static int ext2_sync (struct mount *, int);
acde96db 88static int ext2_unmount (struct mount *, int);
b9b0a6d0 89static int ext2_vget (struct mount *, struct vnode *, ino_t, struct vnode **);
1f1db49f 90static int ext2_init(struct vfsconf *);
a6ee311a 91static int ext2_vptofh (struct vnode *, struct fid *);
984263bc
MD
92
93static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part");
1f1db49f 94MALLOC_DEFINE(M_EXT2MNT, "EXT2 mount", "EXT2 mount structure");
984263bc
MD
95
96static struct vfsops ext2fs_vfsops = {
43c45e8f
HP
97 .vfs_mount = ext2_mount,
98 .vfs_unmount = ext2_unmount,
1f1db49f
MD
99 .vfs_root = ext2_root, /* root inode via vget */
100 .vfs_quotactl = ext2_quotactl, /* quota operations */
43c45e8f
HP
101 .vfs_statfs = ext2_statfs,
102 .vfs_sync = ext2_sync,
103 .vfs_vget = ext2_vget,
104 .vfs_fhtovp = ext2_fhtovp,
1f1db49f 105 .vfs_checkexp = ext2_check_export,
43c45e8f 106 .vfs_vptofh = ext2_vptofh,
1f1db49f
MD
107 .vfs_init = ext2_init,
108 .vfs_uninit = ext2_uninit
984263bc
MD
109};
110
111VFS_SET(ext2fs_vfsops, ext2fs, 0);
77652cad
MD
112#define bsd_malloc kmalloc
113#define bsd_free kfree
984263bc
MD
114
115static int ext2fs_inode_hash_lock;
116
a6ee311a 117static int ext2_check_sb_compat (struct ext2_super_block *es,
b13267a5 118 cdev_t dev, int ronly);
a6ee311a 119static int compute_sb_data (struct vnode * devvp,
984263bc 120 struct ext2_super_block * es,
a6ee311a 121 struct ext2_sb_info * fs);
984263bc 122
1f1db49f
MD
123static int
124ext2_root(struct mount *mp, struct vnode **vpp)
125{
126 struct vnode *nvp;
127 int error;
128
b9b0a6d0 129 error = VFS_VGET(mp, NULL, (ino_t)ROOTINO, &nvp);
1f1db49f
MD
130 if (error)
131 return (error);
132 *vpp = nvp;
133 return (0);
134}
135
136/*
137 * Do operations associated with quotas
138 */
139int
140ext2_quotactl(struct mount *mp, int cmds, uid_t uid, caddr_t arg,
acde96db 141 struct ucred *cred)
1f1db49f
MD
142{
143#ifndef QUOTA
144 return (EOPNOTSUPP);
145#else
1f1db49f
MD
146 int cmd, type, error;
147
1f1db49f
MD
148 type = cmds & SUBCMDMASK;
149 cmd = cmds >> SUBCMDSHIFT;
150
151 if (uid == -1) {
152 switch(type) {
153 case USRQUOTA:
acde96db 154 uid = cred->cr_ruid;
1f1db49f
MD
155 break;
156 case GRPQUOTA:
acde96db 157 uid = cred->cr_rgid;
1f1db49f
MD
158 break;
159 default:
160 return (EINVAL);
161 }
162 }
98b480de
MN
163
164 /*
165 * Check permissions.
166 */
1f1db49f 167 switch (cmd) {
98b480de
MN
168
169 case Q_QUOTAON:
170 error = priv_check_cred(cred, PRIV_UFS_QUOTAON, 0);
171 break;
172
173 case Q_QUOTAOFF:
174 error = priv_check_cred(cred, PRIV_UFS_QUOTAOFF, 0);
175 break;
176
177 case Q_SETQUOTA:
178 error = priv_check_cred(cred, PRIV_VFS_SETQUOTA, 0);
179 break;
180
181 case Q_SETUSE:
182 error = priv_check_cred(cred, PRIV_UFS_SETUSE, 0);
1f1db49f 183 break;
98b480de 184
1f1db49f 185 case Q_GETQUOTA:
acde96db 186 if (uid == cred->cr_ruid)
98b480de
MN
187 error = 0;
188 else
189 error = priv_check_cred(cred, PRIV_VFS_GETQUOTA, 0);
190 break;
191
192 case Q_SYNC:
193 error = 0;
194 break;
195
1f1db49f 196 default:
98b480de
MN
197 error = EINVAL;
198 break;
1f1db49f
MD
199 }
200
98b480de
MN
201 if (error)
202 return (error);
203
204
1f1db49f
MD
205 if ((uint)type >= MAXQUOTAS)
206 return (EINVAL);
f9642f56 207 if (vfs_busy(mp, LK_NOWAIT))
1f1db49f
MD
208 return (0);
209
210 switch (cmd) {
211
212 case Q_QUOTAON:
acde96db 213 error = ext2_quotaon(cred, mp, type, arg);
1f1db49f
MD
214 break;
215
216 case Q_QUOTAOFF:
f94bf21c 217 error = ext2_quotaoff(mp, type);
1f1db49f
MD
218 break;
219
220 case Q_SETQUOTA:
221 error = ext2_setquota(mp, uid, type, arg);
222 break;
223
224 case Q_SETUSE:
225 error = ext2_setuse(mp, uid, type, arg);
226 break;
227
228 case Q_GETQUOTA:
229 error = ext2_getquota(mp, uid, type, arg);
230 break;
231
232 case Q_SYNC:
233 error = ext2_qsync(mp);
234 break;
235
236 default:
237 error = EINVAL;
238 break;
239 }
f9642f56 240 vfs_unbusy(mp);
1f1db49f
MD
241 return (error);
242#endif
243}
244
245/*
246 * Initial UFS filesystems, done only once.
247 */
248int
249ext2_init(struct vfsconf *vfsp)
250{
251 static int done;
252
253 if (done)
254 return (0);
255 done = 1;
256 ext2_ihashinit();
257#ifdef QUOTA
258 ext2_dqinit();
259#endif
260 return (0);
261}
262
984263bc
MD
263/*
264 * VFS Operations.
265 *
266 * mount system call
b1ce5639
SW
267 *
268 * Parameters:
1f1db49f 269 * data: this is actually a (struct ext2_args *)
984263bc
MD
270 */
271static int
acde96db 272ext2_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
984263bc
MD
273{
274 struct vnode *devvp;
1f1db49f
MD
275 struct ext2_args args;
276 struct ext2mount *ump = 0;
f7aae92f 277 struct ext2_sb_info *fs;
984263bc
MD
278 size_t size;
279 int error, flags;
280 mode_t accessmode;
fad57d0e 281 struct nlookupdata nd;
984263bc 282
1f1db49f 283 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ext2_args))) != 0)
984263bc 284 return (error);
7b95be2a 285
984263bc
MD
286 /*
287 * If updating, check whether changing from read-only to
288 * read/write; if there is no device name, that's all we do.
289 */
290 if (mp->mnt_flag & MNT_UPDATE) {
1f1db49f 291 ump = VFSTOEXT2(mp);
984263bc 292 fs = ump->um_e2fs;
8ddc6004 293 devvp = ump->um_devvp;
984263bc
MD
294 error = 0;
295 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
296 flags = WRITECLOSE;
297 if (mp->mnt_flag & MNT_FORCE)
298 flags |= FORCECLOSE;
f9642f56 299 if (vfs_busy(mp, LK_NOWAIT))
984263bc 300 return (EBUSY);
f94bf21c 301 error = ext2_flushfiles(mp, flags);
f9642f56 302 vfs_unbusy(mp);
984263bc
MD
303 if (!error && fs->s_wasvalid) {
304 fs->s_es->s_state |= EXT2_VALID_FS;
305 ext2_sbupdate(ump, MNT_WAIT);
306 }
307 fs->s_rd_only = 1;
ca466bae 308 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
87de5057
MD
309 VOP_OPEN(devvp, FREAD, FSCRED, NULL);
310 VOP_CLOSE(devvp, FREAD|FWRITE);
a11aaa81 311 vn_unlock(devvp);
984263bc
MD
312 }
313 if (!error && (mp->mnt_flag & MNT_RELOAD))
acde96db 314 error = ext2_reload(mp, cred);
984263bc
MD
315 if (error)
316 return (error);
984263bc
MD
317 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
318 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
319 return (EPERM);
320 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
321 /*
322 * If upgrade to read-write by non-root, then verify
323 * that user has necessary permissions on the device.
324 */
7b95be2a 325 if (cred->cr_uid != 0) {
ca466bae 326 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
cb66845a 327 error = VOP_EACCESS(devvp, VREAD | VWRITE, cred);
acde96db 328 if (error) {
a11aaa81 329 vn_unlock(devvp);
984263bc
MD
330 return (error);
331 }
a11aaa81 332 vn_unlock(devvp);
984263bc
MD
333 }
334
335 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
336 (fs->s_es->s_state & EXT2_ERROR_FS)) {
337 if (mp->mnt_flag & MNT_FORCE) {
086c1d7e 338 kprintf(
984263bc
MD
339"WARNING: %s was not properly dismounted\n",
340 fs->fs_fsmnt);
341 } else {
086c1d7e 342 kprintf(
984263bc
MD
343"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
344 fs->fs_fsmnt);
345 return (EPERM);
346 }
347 }
348 fs->s_es->s_state &= ~EXT2_VALID_FS;
349 ext2_sbupdate(ump, MNT_WAIT);
350 fs->s_rd_only = 0;
ca466bae 351 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
87de5057
MD
352 VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL);
353 VOP_CLOSE(devvp, FREAD);
a11aaa81 354 vn_unlock(devvp);
984263bc
MD
355 }
356 if (args.fspec == 0) {
357 /*
358 * Process export requests.
359 */
360 return (vfs_export(mp, &ump->um_export, &args.export));
361 }
362 }
363 /*
364 * Not an update, or updating the name: look up the name
365 * and verify that it refers to a sensible block device.
366 */
fad57d0e
MD
367 devvp = NULL;
368 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
369 if (error == 0)
370 error = nlookup(&nd);
371 if (error == 0)
28623bf9 372 error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
fad57d0e
MD
373 nlookup_done(&nd);
374 if (error)
984263bc 375 return (error);
984263bc
MD
376
377 if (!vn_isdisk(devvp, &error)) {
378 vrele(devvp);
379 return (error);
380 }
381
382 /*
383 * If mount by non-root, then verify that user has necessary
384 * permissions on the device.
385 */
7b95be2a 386 if (cred->cr_uid != 0) {
984263bc
MD
387 accessmode = VREAD;
388 if ((mp->mnt_flag & MNT_RDONLY) == 0)
389 accessmode |= VWRITE;
ca466bae 390 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
cb66845a 391 if ((error = VOP_EACCESS(devvp, accessmode, cred)) != 0) {
984263bc
MD
392 vput(devvp);
393 return (error);
394 }
a11aaa81 395 vn_unlock(devvp);
984263bc
MD
396 }
397
398 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
fef8985e 399 error = ext2_mountfs(devvp, mp, cred);
984263bc
MD
400 } else {
401 if (devvp != ump->um_devvp)
402 error = EINVAL; /* needs translation */
403 else
404 vrele(devvp);
405 }
406 if (error) {
407 vrele(devvp);
408 return (error);
409 }
1f1db49f 410 ump = VFSTOEXT2(mp);
984263bc 411 fs = ump->um_e2fs;
fad57d0e 412 copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
984263bc 413 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
fad57d0e 414 copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size);
984263bc 415 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
acde96db 416 ext2_statfs(mp, &mp->mnt_stat, cred);
984263bc
MD
417 return (0);
418}
419
420/*
421 * checks that the data in the descriptor blocks make sense
422 * this is taken from ext2/super.c
423 */
b1ce5639
SW
424static int
425ext2_check_descriptors(struct ext2_sb_info *sb)
984263bc
MD
426{
427 int i;
428 int desc_block = 0;
429 unsigned long block = sb->s_es->s_first_data_block;
430 struct ext2_group_desc * gdp = NULL;
431
432 /* ext2_debug ("Checking group descriptors"); */
433
434 for (i = 0; i < sb->s_groups_count; i++)
435 {
436 /* examine next descriptor block */
437 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
b993bb87 438 gdp = (struct ext2_group_desc *)
984263bc
MD
439 sb->s_group_desc[desc_block++]->b_data;
440 if (gdp->bg_block_bitmap < block ||
441 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
442 {
086c1d7e 443 kprintf ("ext2_check_descriptors: "
984263bc
MD
444 "Block bitmap for group %d"
445 " not in group (block %lu)!\n",
446 i, (unsigned long) gdp->bg_block_bitmap);
447 return 0;
448 }
449 if (gdp->bg_inode_bitmap < block ||
450 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
451 {
086c1d7e 452 kprintf ("ext2_check_descriptors: "
984263bc
MD
453 "Inode bitmap for group %d"
454 " not in group (block %lu)!\n",
455 i, (unsigned long) gdp->bg_inode_bitmap);
456 return 0;
457 }
458 if (gdp->bg_inode_table < block ||
459 gdp->bg_inode_table + sb->s_itb_per_group >=
460 block + EXT2_BLOCKS_PER_GROUP(sb))
461 {
086c1d7e 462 kprintf ("ext2_check_descriptors: "
984263bc
MD
463 "Inode table for group %d"
464 " not in group (block %lu)!\n",
465 i, (unsigned long) gdp->bg_inode_table);
466 return 0;
467 }
468 block += EXT2_BLOCKS_PER_GROUP(sb);
469 gdp++;
470 }
471 return 1;
472}
473
474static int
b13267a5 475ext2_check_sb_compat(struct ext2_super_block *es, cdev_t dev, int ronly)
984263bc 476{
984263bc 477 if (es->s_magic != EXT2_SUPER_MAGIC) {
086c1d7e 478 kprintf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
984263bc
MD
479 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
480 return (1);
481 }
482 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
483 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
086c1d7e 484 kprintf(
984263bc
MD
485"WARNING: mount of %s denied due to unsupported optional features\n",
486 devtoname(dev));
487 return (1);
488 }
489 if (!ronly &&
490 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
086c1d7e 491 kprintf(
984263bc
MD
492"WARNING: R/W mount of %s denied due to unsupported optional features\n",
493 devtoname(dev));
494 return (1);
495 }
496 }
497 return (0);
498}
499
500/*
501 * this computes the fields of the ext2_sb_info structure from the
502 * data in the ext2_super_block structure read in
503 */
0f7f7a49
CP
504static int
505compute_sb_data(struct vnode *devvp, struct ext2_super_block *es,
506 struct ext2_sb_info *fs)
984263bc
MD
507{
508 int db_count, error;
509 int i, j;
510 int logic_sb_block = 1; /* XXX for now */
511
512#if 1
b993bb87 513#define V(v)
984263bc 514#else
086c1d7e 515#define V(v) kprintf(#v"= %d\n", fs->v);
984263bc
MD
516#endif
517
b993bb87 518 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
984263bc
MD
519 V(s_blocksize)
520 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
521 V(s_bshift)
522 fs->s_fsbtodb = es->s_log_block_size + 1;
523 V(s_fsbtodb)
524 fs->s_qbmask = fs->s_blocksize - 1;
525 V(s_bmask)
526 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
527 V(s_blocksize_bits)
528 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
529 V(s_frag_size)
530 if (fs->s_frag_size)
531 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
532 V(s_frags_per_block)
533 fs->s_blocks_per_group = es->s_blocks_per_group;
534 V(s_blocks_per_group)
535 fs->s_frags_per_group = es->s_frags_per_group;
536 V(s_frags_per_group)
537 fs->s_inodes_per_group = es->s_inodes_per_group;
538 V(s_inodes_per_group)
46932790
MD
539 if (es->s_rev_level == EXT2_GOOD_OLD_REV) {
540 fs->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
541 fs->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
542 } else {
543 fs->s_first_ino = es->s_first_ino;
544 fs->s_inode_size = es->s_inode_size;
545 /*
546 * Simple sanity check for superblock inode size value.
547 */
548 if (fs->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE ||
549 fs->s_inode_size > fs->s_blocksize ||
550 (fs->s_inode_size & (fs->s_inode_size - 1)) != 0) {
551 kprintf("EXT2-fs: invalid inode size %d\n",
552 fs->s_inode_size);
553 return (EIO);
554 }
555 }
556 V(s_first_ino)
557 V(s_inode_size)
558 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE(fs);
984263bc
MD
559 V(s_inodes_per_block)
560 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
561 V(s_itb_per_group)
562 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
563 V(s_desc_per_block)
564 /* s_resuid / s_resgid ? */
565 fs->s_groups_count = (es->s_blocks_count -
566 es->s_first_data_block +
567 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
568 EXT2_BLOCKS_PER_GROUP(fs);
569 V(s_groups_count)
570 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
571 EXT2_DESC_PER_BLOCK(fs);
572 fs->s_db_per_group = db_count;
573 V(s_db_per_group)
574
575 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
1f1db49f 576 M_EXT2MNT, M_WAITOK);
984263bc
MD
577
578 /* adjust logic_sb_block */
b993bb87 579 if(fs->s_blocksize > SBSIZE)
984263bc 580 /* Godmar thinks: if the blocksize is greater than 1024, then
b993bb87 581 the superblock is logically part of block zero.
984263bc
MD
582 */
583 logic_sb_block = 0;
b993bb87 584
984263bc 585 for (i = 0; i < db_count; i++) {
b993bb87 586 error = bread(devvp, fsbtodoff(fs, logic_sb_block + i + 1),
54078292 587 fs->s_blocksize, &fs->s_group_desc[i]);
984263bc
MD
588 if(error) {
589 for (j = 0; j < i; j++)
590 brelse(fs->s_group_desc[j]);
1f1db49f 591 bsd_free(fs->s_group_desc, M_EXT2MNT);
086c1d7e 592 kprintf("EXT2-fs: unable to read group descriptors (%d)\n", error);
984263bc
MD
593 return EIO;
594 }
595 /* Set the B_LOCKED flag on the buffer, then brelse() it */
596 LCK_BUF(fs->s_group_desc[i])
597 }
598 if(!ext2_check_descriptors(fs)) {
599 for (j = 0; j < db_count; j++)
600 ULCK_BUF(fs->s_group_desc[j])
1f1db49f 601 bsd_free(fs->s_group_desc, M_EXT2MNT);
086c1d7e 602 kprintf("EXT2-fs: (ext2_check_descriptors failure) "
984263bc
MD
603 "unable to read group descriptors\n");
604 return EIO;
605 }
606
607 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
608 fs->s_inode_bitmap_number[i] = 0;
609 fs->s_inode_bitmap[i] = NULL;
610 fs->s_block_bitmap_number[i] = 0;
611 fs->s_block_bitmap[i] = NULL;
612 }
613 fs->s_loaded_inode_bitmaps = 0;
614 fs->s_loaded_block_bitmaps = 0;
615 return 0;
616}
617
618/*
619 * Reload all incore data for a filesystem (used after running fsck on
620 * the root filesystem and finding things to fix). The filesystem must
621 * be mounted read-only.
622 *
623 * Things to do to update the mount:
624 * 1) invalidate all cached meta-data.
625 * 2) re-read superblock from disk.
626 * 3) re-read summary information from disk.
627 * 4) invalidate all inactive vnodes.
628 * 5) invalidate all cached file data.
629 * 6) re-read inode data for all active vnodes.
630 */
41a01a4d 631static int ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *rescan);
5fd012e0 632static int ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *rescan);
41a01a4d
MD
633
634struct scaninfo {
635 int rescan;
636 int allerror;
637 int waitfor;
41a01a4d
MD
638 struct vnode *devvp;
639 struct ext2_sb_info *fs;
640};
641
984263bc 642static int
f94bf21c 643ext2_reload(struct mount *mountp, struct ucred *cred)
984263bc 644{
41a01a4d 645 struct vnode *devvp;
984263bc
MD
646 struct buf *bp;
647 struct ext2_super_block * es;
648 struct ext2_sb_info *fs;
649 int error;
41a01a4d 650 struct scaninfo scaninfo;
984263bc
MD
651
652 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
653 return (EINVAL);
654 /*
655 * Step 1: invalidate all cached meta-data.
656 */
1f1db49f 657 devvp = VFSTOEXT2(mountp)->um_devvp;
87de5057 658 if (vinvalbuf(devvp, 0, 0, 0))
984263bc
MD
659 panic("ext2_reload: dirty1");
660 /*
661 * Step 2: re-read superblock from disk.
662 * constants have been adjusted for ext2
663 */
2d5c9276 664 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
984263bc
MD
665 return (error);
666 es = (struct ext2_super_block *)bp->b_data;
667 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
668 brelse(bp);
669 return (EIO); /* XXX needs translation */
670 }
1f1db49f 671 fs = VFSTOEXT2(mountp)->um_e2fs;
984263bc
MD
672 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
673
674 if((error = compute_sb_data(devvp, es, fs)) != 0) {
675 brelse(bp);
676 return error;
677 }
678#ifdef UNKLAR
679 if (fs->fs_sbsize < SBSIZE)
680 bp->b_flags |= B_INVAL;
681#endif
682 brelse(bp);
683
41a01a4d 684 scaninfo.rescan = 1;
41a01a4d
MD
685 scaninfo.devvp = devvp;
686 scaninfo.fs = fs;
687 while (error == 0 && scaninfo.rescan) {
688 scaninfo.rescan = 0;
b993bb87 689 error = vmntvnodescan(mountp, VMSC_GETVX, ext2_reload_scan1,
41a01a4d
MD
690 ext2_reload_scan2, &scaninfo);
691 }
692 return(error);
693}
694
695static int
696ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
697{
5fd012e0 698 /*struct scaninfo *info = data;*/
41a01a4d 699
41a01a4d
MD
700 return(0);
701}
702
703static int
5fd012e0 704ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
41a01a4d
MD
705{
706 struct scaninfo *info = data;
707 struct inode *ip;
708 struct buf *bp;
709 int error;
710
711 /*
5fd012e0 712 * Try to recycle
41a01a4d 713 */
87de5057 714 if (vrecycle(vp))
41a01a4d 715 return(0);
5fd012e0
MD
716
717 /*
718 * Step 5: invalidate all cached file data.
719 */
87de5057 720 if (vinvalbuf(vp, 0, 0, 0))
41a01a4d
MD
721 panic("ext2_reload: dirty2");
722 /*
723 * Step 6: re-read inode data for all active vnodes.
724 */
725 ip = VTOI(vp);
b993bb87 726 error = bread(info->devvp,
54078292
MD
727 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
728 (int)info->fs->s_blocksize, &bp);
5fd012e0 729 if (error)
41a01a4d 730 return (error);
b993bb87 731 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
46932790 732 EXT2_INODE_SIZE(info->fs) * ino_to_fsbo(info->fs, ip->i_number)),
41a01a4d
MD
733 &ip->i_din);
734 brelse(bp);
41a01a4d 735 return(0);
984263bc
MD
736}
737
738/*
739 * Common code for mount and mountroot
740 */
741static int
fef8985e 742ext2_mountfs(struct vnode *devvp, struct mount *mp, struct ucred *cred)
984263bc 743{
1f1db49f 744 struct ext2mount *ump;
984263bc 745 struct buf *bp;
f7aae92f 746 struct ext2_sb_info *fs;
984263bc 747 struct ext2_super_block * es;
b13267a5 748 cdev_t dev;
984263bc 749 struct partinfo dpart;
30b14d35 750 int error, i;
984263bc
MD
751 int ronly;
752
753 /*
754 * Disallow multiple mounts of the same device.
755 * Disallow mounting of a device that is currently in use
756 * (except for root, which might share swap device for miniroot).
757 * Flush out any old buffers remaining from a previous use.
758 */
759 if ((error = vfs_mountedon(devvp)) != 0)
760 return (error);
8be7edad 761 if (vcount(devvp) > 0)
984263bc 762 return (EBUSY);
87de5057 763 if ((error = vinvalbuf(devvp, V_SAVE, 0, 0)) != 0)
984263bc
MD
764 return (error);
765#ifdef READONLY
766/* turn on this to force it to be read-only */
767 mp->mnt_flag |= MNT_RDONLY;
768#endif
769
770 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
ca466bae 771 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
87de5057 772 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
a11aaa81 773 vn_unlock(devvp);
984263bc
MD
774 if (error)
775 return (error);
e4c9c0c8
MD
776 dev = devvp->v_rdev;
777 if (dev->si_iosize_max != 0)
778 mp->mnt_iosize_max = dev->si_iosize_max;
984263bc
MD
779 if (mp->mnt_iosize_max > MAXPHYS)
780 mp->mnt_iosize_max = MAXPHYS;
30b14d35 781 VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, NULL);
984263bc
MD
782
783 bp = NULL;
784 ump = NULL;
2d5c9276 785 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
984263bc
MD
786 goto out;
787 es = (struct ext2_super_block *)bp->b_data;
788 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
789 error = EINVAL; /* XXX needs translation */
790 goto out;
791 }
792 if ((es->s_state & EXT2_VALID_FS) == 0 ||
793 (es->s_state & EXT2_ERROR_FS)) {
794 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
086c1d7e 795 kprintf(
984263bc
MD
796"WARNING: Filesystem was not properly dismounted\n");
797 } else {
086c1d7e 798 kprintf(
984263bc
MD
799"WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
800 error = EPERM;
801 goto out;
802 }
803 }
e7b4468c 804 ump = bsd_malloc(sizeof *ump, M_EXT2MNT, M_WAITOK | M_ZERO);
984263bc
MD
805 ump->um_malloctype = M_EXT2NODE;
806 ump->um_blkatoff = ext2_blkatoff;
807 ump->um_truncate = ext2_truncate;
808 ump->um_update = ext2_update;
809 ump->um_valloc = ext2_valloc;
810 ump->um_vfree = ext2_vfree;
811 /* I don't know whether this is the right strategy. Note that
812 we dynamically allocate both a ext2_sb_info and a ext2_super_block
813 while Linux keeps the super block in a locked buffer
814 */
b993bb87 815 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
1f1db49f 816 M_EXT2MNT, M_WAITOK);
b993bb87 817 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
1f1db49f 818 M_EXT2MNT, M_WAITOK);
984263bc
MD
819 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
820 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
821 goto out;
822 /*
823 * We don't free the group descriptors allocated by compute_sb_data()
824 * until ext2_unmount(). This is OK since the mount will succeed.
825 */
826 brelse(bp);
827 bp = NULL;
828 fs = ump->um_e2fs;
829 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
b993bb87 830 /* if the fs is not mounted read-only, make sure the super block is
984263bc
MD
831 always written back on a sync()
832 */
833 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
834 if (ronly == 0) {
835 fs->s_dirt = 1; /* mark it modified */
836 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
837 }
838 mp->mnt_data = (qaddr_t)ump;
839 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
840 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
841 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
842 mp->mnt_flag |= MNT_LOCAL;
843 ump->um_mountp = mp;
844 ump->um_dev = dev;
845 ump->um_devvp = devvp;
b993bb87 846 /* setting those two parameters allows us to use
1f1db49f 847 ext2_bmap w/o changse !
984263bc
MD
848 */
849 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
850 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
851 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
852 for (i = 0; i < MAXQUOTAS; i++)
b993bb87 853 ump->um_quotas[i] = NULLVP;
e4c9c0c8 854 dev->si_mountpoint = mp;
0961aa92 855
66a1ddf5
MD
856 vfs_add_vnodeops(mp, &ext2_vnode_vops, &mp->mnt_vn_norm_ops);
857 vfs_add_vnodeops(mp, &ext2_spec_vops, &mp->mnt_vn_spec_ops);
858 vfs_add_vnodeops(mp, &ext2_fifo_vops, &mp->mnt_vn_fifo_ops);
0961aa92 859
b993bb87 860 if (ronly == 0)
984263bc
MD
861 ext2_sbupdate(ump, MNT_WAIT);
862 return (0);
863out:
864 if (bp)
865 brelse(bp);
87de5057 866 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE);
984263bc 867 if (ump) {
1f1db49f
MD
868 bsd_free(ump->um_e2fs->s_es, M_EXT2MNT);
869 bsd_free(ump->um_e2fs, M_EXT2MNT);
870 bsd_free(ump, M_EXT2MNT);
984263bc
MD
871 mp->mnt_data = (qaddr_t)0;
872 }
873 return (error);
874}
875
876/*
877 * unmount system call
878 */
879static int
acde96db 880ext2_unmount(struct mount *mp, int mntflags)
984263bc 881{
1f1db49f 882 struct ext2mount *ump;
f7aae92f 883 struct ext2_sb_info *fs;
984263bc
MD
884 int error, flags, ronly, i;
885
886 flags = 0;
887 if (mntflags & MNT_FORCE) {
888 if (mp->mnt_flag & MNT_ROOTFS)
889 return (EINVAL);
890 flags |= FORCECLOSE;
891 }
f94bf21c 892 if ((error = ext2_flushfiles(mp, flags)) != 0)
984263bc 893 return (error);
1f1db49f 894 ump = VFSTOEXT2(mp);
984263bc
MD
895 fs = ump->um_e2fs;
896 ronly = fs->s_rd_only;
897 if (ronly == 0) {
898 if (fs->s_wasvalid)
899 fs->s_es->s_state |= EXT2_VALID_FS;
900 ext2_sbupdate(ump, MNT_WAIT);
901 }
902
903 /* release buffers containing group descriptors */
b993bb87 904 for(i = 0; i < fs->s_db_per_group; i++)
984263bc 905 ULCK_BUF(fs->s_group_desc[i])
1f1db49f 906 bsd_free(fs->s_group_desc, M_EXT2MNT);
984263bc
MD
907
908 /* release cached inode/block bitmaps */
909 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
910 if (fs->s_inode_bitmap[i])
911 ULCK_BUF(fs->s_inode_bitmap[i])
912
913 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
914 if (fs->s_block_bitmap[i])
915 ULCK_BUF(fs->s_block_bitmap[i])
916
e4c9c0c8 917 ump->um_devvp->v_rdev->si_mountpoint = NULL;
87de5057 918 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE);
984263bc 919 vrele(ump->um_devvp);
1f1db49f
MD
920 bsd_free(fs->s_es, M_EXT2MNT);
921 bsd_free(fs, M_EXT2MNT);
922 bsd_free(ump, M_EXT2MNT);
984263bc
MD
923 mp->mnt_data = (qaddr_t)0;
924 mp->mnt_flag &= ~MNT_LOCAL;
925 return (error);
926}
927
928/*
929 * Flush out all the files in a filesystem.
930 */
931static int
f94bf21c 932ext2_flushfiles(struct mount *mp, int flags)
984263bc 933{
1f1db49f 934 struct ext2mount *ump;
984263bc
MD
935 int error;
936#if QUOTA
937 int i;
938#endif
939
1f1db49f 940 ump = VFSTOEXT2(mp);
984263bc
MD
941#if QUOTA
942 if (mp->mnt_flag & MNT_QUOTA) {
943 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
944 return (error);
945 for (i = 0; i < MAXQUOTAS; i++) {
946 if (ump->um_quotas[i] == NULLVP)
947 continue;
f94bf21c 948 ext2_quotaoff(mp, i);
984263bc
MD
949 }
950 /*
951 * Here we fall through to vflush again to ensure
952 * that we have gotten rid of all the system vnodes.
953 */
954 }
955#endif
956 error = vflush(mp, 0, flags);
957 return (error);
958}
959
960/*
961 * Get file system statistics.
962 * taken from ext2/super.c ext2_statfs
963 */
964static int
acde96db 965ext2_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
984263bc
MD
966{
967 unsigned long overhead;
1f1db49f 968 struct ext2mount *ump;
f7aae92f
RG
969 struct ext2_sb_info *fs;
970 struct ext2_super_block *es;
984263bc
MD
971 int i, nsb;
972
1f1db49f 973 ump = VFSTOEXT2(mp);
984263bc
MD
974 fs = ump->um_e2fs;
975 es = fs->s_es;
976
977 if (es->s_magic != EXT2_SUPER_MAGIC)
978 panic("ext2_statfs - magic number spoiled");
979
980 /*
981 * Compute the overhead (FS structures)
982 */
983 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
984 nsb = 0;
985 for (i = 0 ; i < fs->s_groups_count; i++)
986 if (ext2_group_sparse(i))
987 nsb++;
988 } else
989 nsb = fs->s_groups_count;
b993bb87 990 overhead = es->s_first_data_block +
984263bc
MD
991 /* Superblocks and block group descriptors: */
992 nsb * (1 + fs->s_db_per_group) +
993 /* Inode bitmap, block bitmap, and inode table: */
994 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
995
b993bb87 996 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
984263bc
MD
997 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
998 sbp->f_blocks = es->s_blocks_count - overhead;
b993bb87
SW
999 sbp->f_bfree = es->s_free_blocks_count;
1000 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
1001 sbp->f_files = es->s_inodes_count;
1002 sbp->f_ffree = es->s_free_inodes_count;
984263bc
MD
1003 if (sbp != &mp->mnt_stat) {
1004 sbp->f_type = mp->mnt_vfc->vfc_typenum;
984263bc
MD
1005 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
1006 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
1007 }
1008 return (0);
1009}
1010
1011/*
1012 * Go through the disk queues to initiate sandbagged IO;
1013 * go through the inodes to write those that have been modified;
1014 * initiate the writing of the super block if it has been modified.
1015 *
1016 * Note: we are always called with the filesystem marked `MPBUSY'.
1017 */
41a01a4d 1018
5fd012e0 1019static int ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data);
41a01a4d 1020
984263bc 1021static int
87de5057 1022ext2_sync(struct mount *mp, int waitfor)
984263bc 1023{
1f1db49f 1024 struct ext2mount *ump = VFSTOEXT2(mp);
984263bc 1025 struct ext2_sb_info *fs;
41a01a4d
MD
1026 struct scaninfo scaninfo;
1027 int error;
984263bc
MD
1028
1029 fs = ump->um_e2fs;
1030 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
086c1d7e 1031 kprintf("fs = %s\n", fs->fs_fsmnt);
984263bc
MD
1032 panic("ext2_sync: rofs mod");
1033 }
41a01a4d 1034
984263bc
MD
1035 /*
1036 * Write back each (modified) inode.
1037 */
41a01a4d
MD
1038 scaninfo.allerror = 0;
1039 scaninfo.rescan = 1;
1040 scaninfo.waitfor = waitfor;
41a01a4d
MD
1041 while (scaninfo.rescan) {
1042 scaninfo.rescan = 0;
b993bb87 1043 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
5fd012e0 1044 NULL, ext2_sync_scan, &scaninfo);
984263bc 1045 }
41a01a4d 1046
984263bc
MD
1047 /*
1048 * Force stale file system control information to be flushed.
1049 */
28271622 1050 if ((waitfor & MNT_LAZY) == 0) {
ca466bae 1051 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
52174f71 1052 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, 0)) != 0)
41a01a4d 1053 scaninfo.allerror = error;
a11aaa81 1054 vn_unlock(ump->um_devvp);
984263bc
MD
1055 }
1056#if QUOTA
1f1db49f 1057 ext2_qsync(mp);
984263bc
MD
1058#endif
1059 /*
1060 * Write back modified superblock.
1061 */
1062 if (fs->s_dirt != 0) {
1063 fs->s_dirt = 0;
1064 fs->s_es->s_wtime = time_second;
1065 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
41a01a4d
MD
1066 scaninfo.allerror = error;
1067 }
1068 return (scaninfo.allerror);
1069}
1070
1071static int
5fd012e0 1072ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data)
41a01a4d
MD
1073{
1074 struct scaninfo *info = data;
1075 struct inode *ip;
1076 int error;
1077
1078 ip = VTOI(vp);
1079 if (vp->v_type == VNON ||
1080 ((ip->i_flag &
1081 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
28271622 1082 (RB_EMPTY(&vp->v_rbdirty_tree) || (info->waitfor & MNT_LAZY)))) {
41a01a4d 1083 return(0);
984263bc 1084 }
52174f71 1085 if ((error = VOP_FSYNC(vp, info->waitfor, 0)) != 0)
41a01a4d 1086 info->allerror = error;
41a01a4d 1087 return(0);
984263bc
MD
1088}
1089
1090/*
1091 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
1092 * in from disk. If it is in core, wait for the lock bit to clear, then
1093 * return the inode locked. Detection and handling of mount points must be
1094 * done by the calling routine.
1095 */
1096static int
b9b0a6d0 1097ext2_vget(struct mount *mp, struct vnode *dvp, ino_t ino, struct vnode **vpp)
984263bc 1098{
f7aae92f
RG
1099 struct ext2_sb_info *fs;
1100 struct inode *ip;
1f1db49f 1101 struct ext2mount *ump;
984263bc
MD
1102 struct buf *bp;
1103 struct vnode *vp;
b13267a5 1104 cdev_t dev;
984263bc
MD
1105 int i, error;
1106 int used_blocks;
1107
1f1db49f 1108 ump = VFSTOEXT2(mp);
984263bc
MD
1109 dev = ump->um_dev;
1110restart:
1f1db49f 1111 if ((*vpp = ext2_ihashget(dev, ino)) != NULL)
984263bc
MD
1112 return (0);
1113
1114 /*
1115 * Lock out the creation of new entries in the FFS hash table in
1116 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1117 * may occur!
1118 */
1119 if (ext2fs_inode_hash_lock) {
1120 while (ext2fs_inode_hash_lock) {
1121 ext2fs_inode_hash_lock = -1;
377d4740 1122 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
984263bc
MD
1123 }
1124 goto restart;
1125 }
1126 ext2fs_inode_hash_lock = 1;
1127
1128 /*
1129 * If this MALLOC() is performed after the getnewvnode()
1130 * it might block, leaving a vnode with a NULL v_data to be
1131 * found by ext2_sync() if a sync happens to fire right then,
1132 * which will cause a panic because ext2_sync() blindly
1133 * dereferences vp->v_data (as well it should).
1134 */
1135 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1136
1137 /* Allocate a new vnode/inode. */
1f1db49f 1138 if ((error = getnewvnode(VT_EXT2FS, mp, &vp, 0, LK_CANRECURSE)) != 0) {
984263bc
MD
1139 if (ext2fs_inode_hash_lock < 0)
1140 wakeup(&ext2fs_inode_hash_lock);
1141 ext2fs_inode_hash_lock = 0;
1142 *vpp = NULL;
1143 FREE(ip, M_EXT2NODE);
1144 return (error);
1145 }
1146 bzero((caddr_t)ip, sizeof(struct inode));
984263bc
MD
1147 vp->v_data = ip;
1148 ip->i_vnode = vp;
1149 ip->i_e2fs = fs = ump->um_e2fs;
1150 ip->i_dev = dev;
1151 ip->i_number = ino;
1152#if QUOTA
1153 for (i = 0; i < MAXQUOTAS; i++)
1154 ip->i_dquot[i] = NODQUOT;
1155#endif
1156 /*
5fd012e0
MD
1157 * Put it onto its hash chain. Since our vnode is locked, other
1158 * requests for this inode will block if they arrive while we are
1159 * sleeping waiting for old data structures to be purged or for the
1160 * contents of the disk portion of this inode to be read.
984263bc 1161 */
1f1db49f 1162 ext2_ihashins(ip);
984263bc
MD
1163
1164 if (ext2fs_inode_hash_lock < 0)
1165 wakeup(&ext2fs_inode_hash_lock);
1166 ext2fs_inode_hash_lock = 0;
1167
1168 /* Read in the disk contents for the inode, copy into the inode. */
1169#if 0
086c1d7e 1170kprintf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
984263bc 1171#endif
54078292
MD
1172 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1173 (int)fs->s_blocksize, &bp);
1174 if (error) {
984263bc
MD
1175 /*
1176 * The inode does not contain anything useful, so it would
1177 * be misleading to leave it on its hash chain. With mode
1178 * still zero, it will be unlinked and returned to the free
1179 * list by vput().
1180 */
984263bc 1181 brelse(bp);
c081fe84 1182 vx_put(vp);
984263bc
MD
1183 *vpp = NULL;
1184 return (error);
1185 }
1186 /* convert ext2 inode to dinode */
46932790 1187 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
984263bc
MD
1188 ino_to_fsbo(fs, ino)), &ip->i_din);
1189 ip->i_block_group = ino_to_cg(fs, ino);
1190 ip->i_next_alloc_block = 0;
1191 ip->i_next_alloc_goal = 0;
1192 ip->i_prealloc_count = 0;
1193 ip->i_prealloc_block = 0;
1194 /* now we want to make sure that block pointers for unused
b993bb87 1195 blocks are zeroed out - ext2_balloc depends on this
984263bc
MD
1196 although for regular files and directories only
1197 */
1198 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1199 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1200 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1201 ip->i_db[i] = 0;
1202 }
933a3a0d 1203#if 0
984263bc 1204 ext2_print_inode(ip);
933a3a0d 1205#endif
c081fe84 1206 bqrelse(bp);
984263bc
MD
1207
1208 /*
1209 * Initialize the vnode from the inode, check for aliases.
1210 * Note that the underlying vnode may have changed.
1211 */
1f1db49f 1212 if ((error = ext2_vinit(mp, &vp)) != 0) {
5fd012e0 1213 vx_put(vp);
984263bc
MD
1214 *vpp = NULL;
1215 return (error);
1216 }
93f50d60 1217
984263bc
MD
1218 /*
1219 * Finish inode initialization now that aliasing has been resolved.
1220 */
1221 ip->i_devvp = ump->um_devvp;
597aea93 1222 vref(ip->i_devvp);
984263bc
MD
1223 /*
1224 * Set up a generation number for this inode if it does not
1225 * already have one. This should only happen on old filesystems.
1226 */
1227 if (ip->i_gen == 0) {
cddfb7bb 1228 ip->i_gen = krandom() / 2 + 1;
984263bc
MD
1229 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1230 ip->i_flag |= IN_MODIFIED;
1231 }
5fd012e0
MD
1232 /*
1233 * Return the locked and refd vnode.
1234 */
984263bc
MD
1235 *vpp = vp;
1236 return (0);
1237}
1238
1239/*
1240 * File handle to vnode
1241 *
1242 * Have to be really careful about stale file handles:
1243 * - check that the inode number is valid
1244 * - call ext2_vget() to get the locked inode
1245 * - check for an unallocated inode (i_mode == 0)
1246 * - check that the given client host has export rights and return
1247 * those rights via. exflagsp and credanonp
1248 */
1249static int
67863d04
MD
1250ext2_fhtovp(struct mount *mp, struct vnode *rootvp,
1251 struct fid *fhp, struct vnode **vpp)
984263bc 1252{
f7aae92f 1253 struct ufid *ufhp;
984263bc 1254 struct ext2_sb_info *fs;
1f1db49f
MD
1255 struct inode *ip;
1256 struct vnode *nvp;
1257 int error;
984263bc
MD
1258
1259 ufhp = (struct ufid *)fhp;
1f1db49f 1260 fs = VFSTOEXT2(mp)->um_e2fs;
984263bc
MD
1261 if (ufhp->ufid_ino < ROOTINO ||
1262 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1263 return (ESTALE);
1f1db49f 1264
b9b0a6d0 1265 error = VFS_VGET(mp, rootvp, ufhp->ufid_ino, &nvp);
1f1db49f
MD
1266 if (error) {
1267 *vpp = NULLVP;
1268 return (error);
1269 }
1270 ip = VTOI(nvp);
1271 if (ip->i_mode == 0 ||
1272 ip->i_gen != ufhp->ufid_gen ||
1273 (VFSTOEXT2(mp)->um_i_effnlink_valid ? ip->i_effnlink :
1274 ip->i_nlink) <= 0) {
1275 vput(nvp);
1276 *vpp = NULLVP;
1277 return (ESTALE);
1278 }
1279 *vpp = nvp;
1280 return (0);
984263bc
MD
1281}
1282
1283/*
1284 * Vnode pointer to File handle
1285 */
1286/* ARGSUSED */
1287static int
0f7f7a49 1288ext2_vptofh(struct vnode *vp, struct fid *fhp)
984263bc 1289{
f7aae92f
RG
1290 struct inode *ip;
1291 struct ufid *ufhp;
984263bc
MD
1292
1293 ip = VTOI(vp);
1294 ufhp = (struct ufid *)fhp;
1295 ufhp->ufid_len = sizeof(struct ufid);
1296 ufhp->ufid_ino = ip->i_number;
1297 ufhp->ufid_gen = ip->i_gen;
1298 return (0);
1299}
1300
1f1db49f
MD
1301/*
1302 * This is the generic part of fhtovp called after the underlying
1303 * filesystem has validated the file handle.
1304 *
1305 * Verify that a host should have access to a filesystem.
1306 */
1307int
1308ext2_check_export(struct mount *mp, struct sockaddr *nam, int *exflagsp,
1309 struct ucred **credanonp)
1310{
1311 struct netcred *np;
fc6d0222 1312 struct ext2mount *ump;
b993bb87 1313
1f1db49f
MD
1314 ump = VFSTOEXT2(mp);
1315 /*
1316 * Get the export permission structure for this <mp, client> tuple.
1317 */
1318 np = vfs_export_lookup(mp, &ump->um_export, nam);
1319 if (np == NULL)
1320 return (EACCES);
1321
1322 *exflagsp = np->netc_exflags;
1323 *credanonp = &np->netc_anon;
1324 return (0);
1325}
1326
984263bc
MD
1327/*
1328 * Write a superblock and associated information back to disk.
1329 */
1330static int
1f1db49f 1331ext2_sbupdate(struct ext2mount *mp, int waitfor)
984263bc 1332{
f7aae92f
RG
1333 struct ext2_sb_info *fs = mp->um_e2fs;
1334 struct ext2_super_block *es = fs->s_es;
1335 struct buf *bp;
984263bc
MD
1336 int error = 0;
1337/*
086c1d7e 1338kprintf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
984263bc 1339*/
2d5c9276 1340 bp = getblk(mp->um_devvp, SBOFF, SBSIZE, 0, 0);
984263bc
MD
1341 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1342 if (waitfor == MNT_WAIT)
1343 error = bwrite(bp);
1344 else
1345 bawrite(bp);
1346
1347 /*
1348 * The buffers for group descriptors, inode bitmaps and block bitmaps
1349 * are not busy at this point and are (hopefully) written by the
1350 * usual sync mechanism. No need to write them here
0f7f7a49 1351 */
984263bc
MD
1352
1353 return (error);
1354}