proc->thread stage 5: BUF/VFS clearance! Remove the ucred argument from
[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 $
3b568787 41 * $DragonFly: src/sys/vfs/gnu/ext2fs/ext2_vfsops.c,v 1.3 2003/06/26 05:55:12 dillon Exp $
984263bc
MD
42 */
43
44#include "opt_quota.h"
45
46#include <sys/param.h>
47#include <sys/systm.h>
48#include <sys/namei.h>
49#include <sys/proc.h>
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>
56#include <sys/disklabel.h>
57#include <sys/malloc.h>
58#include <sys/stat.h>
59
60#include <ufs/ufs/quota.h>
61#include <ufs/ufs/ufsmount.h>
62#include <ufs/ufs/inode.h>
63#include <ufs/ufs/ufs_extern.h>
64
65#include <vm/vm_zone.h>
66
67#include <gnu/ext2fs/fs.h>
68#include <gnu/ext2fs/ext2_extern.h>
69#include <gnu/ext2fs/ext2_fs.h>
70#include <gnu/ext2fs/ext2_fs_sb.h>
71
72static int ext2_fhtovp __P((struct mount *, struct fid *, struct vnode **));
73static int ext2_flushfiles __P((struct mount *mp, int flags, struct proc *p));
74static int ext2_mount __P((struct mount *,
75 char *, caddr_t, struct nameidata *, struct proc *));
76static int ext2_mountfs __P((struct vnode *, struct mount *, struct proc *));
77static int ext2_reload __P((struct mount *mountp, struct ucred *cred,
78 struct proc *p));
79static int ext2_sbupdate __P((struct ufsmount *, int));
80static int ext2_statfs __P((struct mount *, struct statfs *, struct proc *));
81static int ext2_sync __P((struct mount *, int, struct ucred *, struct proc *));
82static int ext2_unmount __P((struct mount *, int, struct proc *));
83static int ext2_vget __P((struct mount *, ino_t, struct vnode **));
84static int ext2_vptofh __P((struct vnode *, struct fid *));
85
86static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part");
87
88static struct vfsops ext2fs_vfsops = {
89 ext2_mount,
90 ufs_start, /* empty function */
91 ext2_unmount,
92 ufs_root, /* root inode via vget */
93 ufs_quotactl, /* does operations associated with quotas */
94 ext2_statfs,
95 ext2_sync,
96 ext2_vget,
97 ext2_fhtovp,
98 ufs_check_export,
99 ext2_vptofh,
100 ext2_init,
101 vfs_stduninit,
102 vfs_stdextattrctl,
103};
104
105VFS_SET(ext2fs_vfsops, ext2fs, 0);
106#define bsd_malloc malloc
107#define bsd_free free
108
109static int ext2fs_inode_hash_lock;
110
111static int ext2_check_sb_compat __P((struct ext2_super_block *es,
112 dev_t dev, int ronly));
113static int compute_sb_data __P((struct vnode * devvp,
114 struct ext2_super_block * es,
115 struct ext2_sb_info * fs));
116
117#ifdef notyet
118static int ext2_mountroot __P((void));
119
120/*
121 * Called by main() when ext2fs is going to be mounted as root.
122 *
123 * Name is updated by mount(8) after booting.
124 */
125#define ROOTNAME "root_device"
126
127static int
128ext2_mountroot()
129{
130 register struct ext2_sb_info *fs;
131 register struct mount *mp;
132 struct proc *p = curproc;
133 struct ufsmount *ump;
134 u_int size;
135 int error;
136
137 if ((error = bdevvp(rootdev, &rootvp))) {
138 printf("ext2_mountroot: can't find rootvp\n");
139 return (error);
140 }
141 mp = bsd_malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
142 bzero((char *)mp, (u_long)sizeof(struct mount));
143 TAILQ_INIT(&mp->mnt_nvnodelist);
144 TAILQ_INIT(&mp->mnt_reservedvnlist);
145 mp->mnt_op = &ext2fs_vfsops;
146 mp->mnt_flag = MNT_RDONLY;
147 if (error = ext2_mountfs(rootvp, mp, p)) {
148 bsd_free(mp, M_MOUNT);
149 return (error);
150 }
151 if (error = vfs_lock(mp)) {
152 (void)ext2_unmount(mp, 0, p);
153 bsd_free(mp, M_MOUNT);
154 return (error);
155 }
156 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list);
157 mp->mnt_flag |= MNT_ROOTFS;
158 mp->mnt_vnodecovered = NULLVP;
159 ump = VFSTOUFS(mp);
160 fs = ump->um_e2fs;
161 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
162 fs->fs_fsmnt[0] = '/';
163 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
164 MNAMELEN);
165 (void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
166 &size);
167 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
168 (void)ext2_statfs(mp, &mp->mnt_stat, p);
169 vfs_unlock(mp);
170 inittodr(fs->s_es->s_wtime); /* this helps to set the time */
171 return (0);
172}
173#endif
174
175/*
176 * VFS Operations.
177 *
178 * mount system call
179 */
180static int
181ext2_mount(mp, path, data, ndp, p)
182 register struct mount *mp;
183 char *path;
184 caddr_t data; /* this is actually a (struct ufs_args *) */
185 struct nameidata *ndp;
186 struct proc *p;
187{
188 struct vnode *devvp;
189 struct ufs_args args;
190 struct ufsmount *ump = 0;
191 register struct ext2_sb_info *fs;
192 size_t size;
193 int error, flags;
194 mode_t accessmode;
195
196 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args))) != 0)
197 return (error);
198 /*
199 * If updating, check whether changing from read-only to
200 * read/write; if there is no device name, that's all we do.
201 */
202 if (mp->mnt_flag & MNT_UPDATE) {
203 ump = VFSTOUFS(mp);
204 fs = ump->um_e2fs;
205 error = 0;
206 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
207 flags = WRITECLOSE;
208 if (mp->mnt_flag & MNT_FORCE)
209 flags |= FORCECLOSE;
210 if (vfs_busy(mp, LK_NOWAIT, 0, p))
211 return (EBUSY);
212 error = ext2_flushfiles(mp, flags, p);
213 vfs_unbusy(mp, p);
214 if (!error && fs->s_wasvalid) {
215 fs->s_es->s_state |= EXT2_VALID_FS;
216 ext2_sbupdate(ump, MNT_WAIT);
217 }
218 fs->s_rd_only = 1;
219 }
220 if (!error && (mp->mnt_flag & MNT_RELOAD))
221 error = ext2_reload(mp, ndp->ni_cnd.cn_cred, p);
222 if (error)
223 return (error);
224 devvp = ump->um_devvp;
225 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
226 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
227 return (EPERM);
228 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
229 /*
230 * If upgrade to read-write by non-root, then verify
231 * that user has necessary permissions on the device.
232 */
233 if (p->p_ucred->cr_uid != 0) {
234 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
235 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
236 p->p_ucred, p)) != 0) {
237 VOP_UNLOCK(devvp, 0, p);
238 return (error);
239 }
240 VOP_UNLOCK(devvp, 0, p);
241 }
242
243 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
244 (fs->s_es->s_state & EXT2_ERROR_FS)) {
245 if (mp->mnt_flag & MNT_FORCE) {
246 printf(
247"WARNING: %s was not properly dismounted\n",
248 fs->fs_fsmnt);
249 } else {
250 printf(
251"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
252 fs->fs_fsmnt);
253 return (EPERM);
254 }
255 }
256 fs->s_es->s_state &= ~EXT2_VALID_FS;
257 ext2_sbupdate(ump, MNT_WAIT);
258 fs->s_rd_only = 0;
259 }
260 if (args.fspec == 0) {
261 /*
262 * Process export requests.
263 */
264 return (vfs_export(mp, &ump->um_export, &args.export));
265 }
266 }
267 /*
268 * Not an update, or updating the name: look up the name
269 * and verify that it refers to a sensible block device.
270 */
271 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
272 if ((error = namei(ndp)) != 0)
273 return (error);
274 NDFREE(ndp, NDF_ONLY_PNBUF);
275 devvp = ndp->ni_vp;
276
277 if (!vn_isdisk(devvp, &error)) {
278 vrele(devvp);
279 return (error);
280 }
281
282 /*
283 * If mount by non-root, then verify that user has necessary
284 * permissions on the device.
285 */
286 if (p->p_ucred->cr_uid != 0) {
287 accessmode = VREAD;
288 if ((mp->mnt_flag & MNT_RDONLY) == 0)
289 accessmode |= VWRITE;
290 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
291 if ((error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p)) != 0) {
292 vput(devvp);
293 return (error);
294 }
295 VOP_UNLOCK(devvp, 0, p);
296 }
297
298 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
299 error = ext2_mountfs(devvp, mp, p);
300 } else {
301 if (devvp != ump->um_devvp)
302 error = EINVAL; /* needs translation */
303 else
304 vrele(devvp);
305 }
306 if (error) {
307 vrele(devvp);
308 return (error);
309 }
310 ump = VFSTOUFS(mp);
311 fs = ump->um_e2fs;
312 (void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
313 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
314 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
315 MNAMELEN);
316 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
317 &size);
318 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
319 (void)ext2_statfs(mp, &mp->mnt_stat, p);
320 return (0);
321}
322
323/*
324 * checks that the data in the descriptor blocks make sense
325 * this is taken from ext2/super.c
326 */
327static int ext2_check_descriptors (struct ext2_sb_info * sb)
328{
329 int i;
330 int desc_block = 0;
331 unsigned long block = sb->s_es->s_first_data_block;
332 struct ext2_group_desc * gdp = NULL;
333
334 /* ext2_debug ("Checking group descriptors"); */
335
336 for (i = 0; i < sb->s_groups_count; i++)
337 {
338 /* examine next descriptor block */
339 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
340 gdp = (struct ext2_group_desc *)
341 sb->s_group_desc[desc_block++]->b_data;
342 if (gdp->bg_block_bitmap < block ||
343 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
344 {
345 printf ("ext2_check_descriptors: "
346 "Block bitmap for group %d"
347 " not in group (block %lu)!\n",
348 i, (unsigned long) gdp->bg_block_bitmap);
349 return 0;
350 }
351 if (gdp->bg_inode_bitmap < block ||
352 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
353 {
354 printf ("ext2_check_descriptors: "
355 "Inode bitmap for group %d"
356 " not in group (block %lu)!\n",
357 i, (unsigned long) gdp->bg_inode_bitmap);
358 return 0;
359 }
360 if (gdp->bg_inode_table < block ||
361 gdp->bg_inode_table + sb->s_itb_per_group >=
362 block + EXT2_BLOCKS_PER_GROUP(sb))
363 {
364 printf ("ext2_check_descriptors: "
365 "Inode table for group %d"
366 " not in group (block %lu)!\n",
367 i, (unsigned long) gdp->bg_inode_table);
368 return 0;
369 }
370 block += EXT2_BLOCKS_PER_GROUP(sb);
371 gdp++;
372 }
373 return 1;
374}
375
376static int
377ext2_check_sb_compat(es, dev, ronly)
378 struct ext2_super_block *es;
379 dev_t dev;
380 int ronly;
381{
382
383 if (es->s_magic != EXT2_SUPER_MAGIC) {
384 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
385 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
386 return (1);
387 }
388 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
389 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
390 printf(
391"WARNING: mount of %s denied due to unsupported optional features\n",
392 devtoname(dev));
393 return (1);
394 }
395 if (!ronly &&
396 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
397 printf(
398"WARNING: R/W mount of %s denied due to unsupported optional features\n",
399 devtoname(dev));
400 return (1);
401 }
402 }
403 return (0);
404}
405
406/*
407 * this computes the fields of the ext2_sb_info structure from the
408 * data in the ext2_super_block structure read in
409 */
410static int compute_sb_data(devvp, es, fs)
411 struct vnode * devvp;
412 struct ext2_super_block * es;
413 struct ext2_sb_info * fs;
414{
415 int db_count, error;
416 int i, j;
417 int logic_sb_block = 1; /* XXX for now */
418
419#if 1
420#define V(v)
421#else
422#define V(v) printf(#v"= %d\n", fs->v);
423#endif
424
425 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
426 V(s_blocksize)
427 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
428 V(s_bshift)
429 fs->s_fsbtodb = es->s_log_block_size + 1;
430 V(s_fsbtodb)
431 fs->s_qbmask = fs->s_blocksize - 1;
432 V(s_bmask)
433 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
434 V(s_blocksize_bits)
435 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
436 V(s_frag_size)
437 if (fs->s_frag_size)
438 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
439 V(s_frags_per_block)
440 fs->s_blocks_per_group = es->s_blocks_per_group;
441 V(s_blocks_per_group)
442 fs->s_frags_per_group = es->s_frags_per_group;
443 V(s_frags_per_group)
444 fs->s_inodes_per_group = es->s_inodes_per_group;
445 V(s_inodes_per_group)
446 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE;
447 V(s_inodes_per_block)
448 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
449 V(s_itb_per_group)
450 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
451 V(s_desc_per_block)
452 /* s_resuid / s_resgid ? */
453 fs->s_groups_count = (es->s_blocks_count -
454 es->s_first_data_block +
455 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
456 EXT2_BLOCKS_PER_GROUP(fs);
457 V(s_groups_count)
458 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
459 EXT2_DESC_PER_BLOCK(fs);
460 fs->s_db_per_group = db_count;
461 V(s_db_per_group)
462
463 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
464 M_UFSMNT, M_WAITOK);
465
466 /* adjust logic_sb_block */
467 if(fs->s_blocksize > SBSIZE)
468 /* Godmar thinks: if the blocksize is greater than 1024, then
469 the superblock is logically part of block zero.
470 */
471 logic_sb_block = 0;
472
473 for (i = 0; i < db_count; i++) {
474 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1),
475 fs->s_blocksize, NOCRED, &fs->s_group_desc[i]);
476 if(error) {
477 for (j = 0; j < i; j++)
478 brelse(fs->s_group_desc[j]);
479 bsd_free(fs->s_group_desc, M_UFSMNT);
480 printf("EXT2-fs: unable to read group descriptors (%d)\n", error);
481 return EIO;
482 }
483 /* Set the B_LOCKED flag on the buffer, then brelse() it */
484 LCK_BUF(fs->s_group_desc[i])
485 }
486 if(!ext2_check_descriptors(fs)) {
487 for (j = 0; j < db_count; j++)
488 ULCK_BUF(fs->s_group_desc[j])
489 bsd_free(fs->s_group_desc, M_UFSMNT);
490 printf("EXT2-fs: (ext2_check_descriptors failure) "
491 "unable to read group descriptors\n");
492 return EIO;
493 }
494
495 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
496 fs->s_inode_bitmap_number[i] = 0;
497 fs->s_inode_bitmap[i] = NULL;
498 fs->s_block_bitmap_number[i] = 0;
499 fs->s_block_bitmap[i] = NULL;
500 }
501 fs->s_loaded_inode_bitmaps = 0;
502 fs->s_loaded_block_bitmaps = 0;
503 return 0;
504}
505
506/*
507 * Reload all incore data for a filesystem (used after running fsck on
508 * the root filesystem and finding things to fix). The filesystem must
509 * be mounted read-only.
510 *
511 * Things to do to update the mount:
512 * 1) invalidate all cached meta-data.
513 * 2) re-read superblock from disk.
514 * 3) re-read summary information from disk.
515 * 4) invalidate all inactive vnodes.
516 * 5) invalidate all cached file data.
517 * 6) re-read inode data for all active vnodes.
518 */
519static int
520ext2_reload(mountp, cred, p)
521 register struct mount *mountp;
522 struct ucred *cred;
523 struct proc *p;
524{
525 register struct vnode *vp, *nvp, *devvp;
526 struct inode *ip;
527 struct buf *bp;
528 struct ext2_super_block * es;
529 struct ext2_sb_info *fs;
530 int error;
531
532 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
533 return (EINVAL);
534 /*
535 * Step 1: invalidate all cached meta-data.
536 */
537 devvp = VFSTOUFS(mountp)->um_devvp;
3b568787 538 if (vinvalbuf(devvp, 0, p, 0, 0))
984263bc
MD
539 panic("ext2_reload: dirty1");
540 /*
541 * Step 2: re-read superblock from disk.
542 * constants have been adjusted for ext2
543 */
544 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
545 return (error);
546 es = (struct ext2_super_block *)bp->b_data;
547 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
548 brelse(bp);
549 return (EIO); /* XXX needs translation */
550 }
551 fs = VFSTOUFS(mountp)->um_e2fs;
552 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
553
554 if((error = compute_sb_data(devvp, es, fs)) != 0) {
555 brelse(bp);
556 return error;
557 }
558#ifdef UNKLAR
559 if (fs->fs_sbsize < SBSIZE)
560 bp->b_flags |= B_INVAL;
561#endif
562 brelse(bp);
563
564loop:
565 simple_lock(&mntvnode_slock);
566 for (vp = TAILQ_FIRST(&mountp->mnt_nvnodelist); vp != NULL; vp = nvp) {
567 if (vp->v_mount != mountp) {
568 simple_unlock(&mntvnode_slock);
569 goto loop;
570 }
571 nvp = TAILQ_NEXT(vp, v_nmntvnodes);
572 /*
573 * Step 4: invalidate all inactive vnodes.
574 */
575 if (vrecycle(vp, &mntvnode_slock, p))
576 goto loop;
577 /*
578 * Step 5: invalidate all cached file data.
579 */
580 simple_lock(&vp->v_interlock);
581 simple_unlock(&mntvnode_slock);
582 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) {
583 goto loop;
584 }
3b568787 585 if (vinvalbuf(vp, 0, p, 0, 0))
984263bc
MD
586 panic("ext2_reload: dirty2");
587 /*
588 * Step 6: re-read inode data for all active vnodes.
589 */
590 ip = VTOI(vp);
591 error =
592 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
593 (int)fs->s_blocksize, NOCRED, &bp);
594 if (error) {
595 vput(vp);
596 return (error);
597 }
598 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
599 EXT2_INODE_SIZE * ino_to_fsbo(fs, ip->i_number)),
600 &ip->i_din);
601 brelse(bp);
602 vput(vp);
603 simple_lock(&mntvnode_slock);
604 }
605 simple_unlock(&mntvnode_slock);
606 return (0);
607}
608
609/*
610 * Common code for mount and mountroot
611 */
612static int
613ext2_mountfs(devvp, mp, p)
614 register struct vnode *devvp;
615 struct mount *mp;
616 struct proc *p;
617{
618 register struct ufsmount *ump;
619 struct buf *bp;
620 register struct ext2_sb_info *fs;
621 struct ext2_super_block * es;
622 dev_t dev = devvp->v_rdev;
623 struct partinfo dpart;
624 int havepart = 0;
625 int error, i, size;
626 int ronly;
627
628 /*
629 * Disallow multiple mounts of the same device.
630 * Disallow mounting of a device that is currently in use
631 * (except for root, which might share swap device for miniroot).
632 * Flush out any old buffers remaining from a previous use.
633 */
634 if ((error = vfs_mountedon(devvp)) != 0)
635 return (error);
636 if (vcount(devvp) > 1 && devvp != rootvp)
637 return (EBUSY);
3b568787 638 if ((error = vinvalbuf(devvp, V_SAVE, p, 0, 0)) != 0)
984263bc
MD
639 return (error);
640#ifdef READONLY
641/* turn on this to force it to be read-only */
642 mp->mnt_flag |= MNT_RDONLY;
643#endif
644
645 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
646 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
647 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
648 VOP_UNLOCK(devvp, 0, p);
649 if (error)
650 return (error);
651 if (devvp->v_rdev->si_iosize_max != 0)
652 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
653 if (mp->mnt_iosize_max > MAXPHYS)
654 mp->mnt_iosize_max = MAXPHYS;
655 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
656 size = DEV_BSIZE;
657 else {
658 havepart = 1;
659 size = dpart.disklab->d_secsize;
660 }
661
662 bp = NULL;
663 ump = NULL;
664 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
665 goto out;
666 es = (struct ext2_super_block *)bp->b_data;
667 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
668 error = EINVAL; /* XXX needs translation */
669 goto out;
670 }
671 if ((es->s_state & EXT2_VALID_FS) == 0 ||
672 (es->s_state & EXT2_ERROR_FS)) {
673 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
674 printf(
675"WARNING: Filesystem was not properly dismounted\n");
676 } else {
677 printf(
678"WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
679 error = EPERM;
680 goto out;
681 }
682 }
683 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
684 bzero((caddr_t)ump, sizeof *ump);
685 ump->um_malloctype = M_EXT2NODE;
686 ump->um_blkatoff = ext2_blkatoff;
687 ump->um_truncate = ext2_truncate;
688 ump->um_update = ext2_update;
689 ump->um_valloc = ext2_valloc;
690 ump->um_vfree = ext2_vfree;
691 /* I don't know whether this is the right strategy. Note that
692 we dynamically allocate both a ext2_sb_info and a ext2_super_block
693 while Linux keeps the super block in a locked buffer
694 */
695 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
696 M_UFSMNT, M_WAITOK);
697 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
698 M_UFSMNT, M_WAITOK);
699 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
700 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
701 goto out;
702 /*
703 * We don't free the group descriptors allocated by compute_sb_data()
704 * until ext2_unmount(). This is OK since the mount will succeed.
705 */
706 brelse(bp);
707 bp = NULL;
708 fs = ump->um_e2fs;
709 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
710 /* if the fs is not mounted read-only, make sure the super block is
711 always written back on a sync()
712 */
713 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
714 if (ronly == 0) {
715 fs->s_dirt = 1; /* mark it modified */
716 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
717 }
718 mp->mnt_data = (qaddr_t)ump;
719 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
720 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
721 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
722 mp->mnt_flag |= MNT_LOCAL;
723 ump->um_mountp = mp;
724 ump->um_dev = dev;
725 ump->um_devvp = devvp;
726 /* setting those two parameters allows us to use
727 ufs_bmap w/o changse !
728 */
729 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
730 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
731 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
732 for (i = 0; i < MAXQUOTAS; i++)
733 ump->um_quotas[i] = NULLVP;
734 devvp->v_specmountpoint = mp;
735 if (ronly == 0)
736 ext2_sbupdate(ump, MNT_WAIT);
737 return (0);
738out:
739 if (bp)
740 brelse(bp);
741 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
742 if (ump) {
743 bsd_free(ump->um_e2fs->s_es, M_UFSMNT);
744 bsd_free(ump->um_e2fs, M_UFSMNT);
745 bsd_free(ump, M_UFSMNT);
746 mp->mnt_data = (qaddr_t)0;
747 }
748 return (error);
749}
750
751/*
752 * unmount system call
753 */
754static int
755ext2_unmount(mp, mntflags, p)
756 struct mount *mp;
757 int mntflags;
758 struct proc *p;
759{
760 register struct ufsmount *ump;
761 register struct ext2_sb_info *fs;
762 int error, flags, ronly, i;
763
764 flags = 0;
765 if (mntflags & MNT_FORCE) {
766 if (mp->mnt_flag & MNT_ROOTFS)
767 return (EINVAL);
768 flags |= FORCECLOSE;
769 }
770 if ((error = ext2_flushfiles(mp, flags, p)) != 0)
771 return (error);
772 ump = VFSTOUFS(mp);
773 fs = ump->um_e2fs;
774 ronly = fs->s_rd_only;
775 if (ronly == 0) {
776 if (fs->s_wasvalid)
777 fs->s_es->s_state |= EXT2_VALID_FS;
778 ext2_sbupdate(ump, MNT_WAIT);
779 }
780
781 /* release buffers containing group descriptors */
782 for(i = 0; i < fs->s_db_per_group; i++)
783 ULCK_BUF(fs->s_group_desc[i])
784 bsd_free(fs->s_group_desc, M_UFSMNT);
785
786 /* release cached inode/block bitmaps */
787 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
788 if (fs->s_inode_bitmap[i])
789 ULCK_BUF(fs->s_inode_bitmap[i])
790
791 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
792 if (fs->s_block_bitmap[i])
793 ULCK_BUF(fs->s_block_bitmap[i])
794
795 ump->um_devvp->v_specmountpoint = NULL;
796 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE,
797 NOCRED, p);
798 vrele(ump->um_devvp);
799 bsd_free(fs->s_es, M_UFSMNT);
800 bsd_free(fs, M_UFSMNT);
801 bsd_free(ump, M_UFSMNT);
802 mp->mnt_data = (qaddr_t)0;
803 mp->mnt_flag &= ~MNT_LOCAL;
804 return (error);
805}
806
807/*
808 * Flush out all the files in a filesystem.
809 */
810static int
811ext2_flushfiles(mp, flags, p)
812 register struct mount *mp;
813 int flags;
814 struct proc *p;
815{
816 register struct ufsmount *ump;
817 int error;
818#if QUOTA
819 int i;
820#endif
821
822 ump = VFSTOUFS(mp);
823#if QUOTA
824 if (mp->mnt_flag & MNT_QUOTA) {
825 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
826 return (error);
827 for (i = 0; i < MAXQUOTAS; i++) {
828 if (ump->um_quotas[i] == NULLVP)
829 continue;
830 quotaoff(p, mp, i);
831 }
832 /*
833 * Here we fall through to vflush again to ensure
834 * that we have gotten rid of all the system vnodes.
835 */
836 }
837#endif
838 error = vflush(mp, 0, flags);
839 return (error);
840}
841
842/*
843 * Get file system statistics.
844 * taken from ext2/super.c ext2_statfs
845 */
846static int
847ext2_statfs(mp, sbp, p)
848 struct mount *mp;
849 register struct statfs *sbp;
850 struct proc *p;
851{
852 unsigned long overhead;
853 register struct ufsmount *ump;
854 register struct ext2_sb_info *fs;
855 register struct ext2_super_block *es;
856 int i, nsb;
857
858 ump = VFSTOUFS(mp);
859 fs = ump->um_e2fs;
860 es = fs->s_es;
861
862 if (es->s_magic != EXT2_SUPER_MAGIC)
863 panic("ext2_statfs - magic number spoiled");
864
865 /*
866 * Compute the overhead (FS structures)
867 */
868 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
869 nsb = 0;
870 for (i = 0 ; i < fs->s_groups_count; i++)
871 if (ext2_group_sparse(i))
872 nsb++;
873 } else
874 nsb = fs->s_groups_count;
875 overhead = es->s_first_data_block +
876 /* Superblocks and block group descriptors: */
877 nsb * (1 + fs->s_db_per_group) +
878 /* Inode bitmap, block bitmap, and inode table: */
879 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
880
881 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
882 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
883 sbp->f_blocks = es->s_blocks_count - overhead;
884 sbp->f_bfree = es->s_free_blocks_count;
885 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
886 sbp->f_files = es->s_inodes_count;
887 sbp->f_ffree = es->s_free_inodes_count;
888 if (sbp != &mp->mnt_stat) {
889 sbp->f_type = mp->mnt_vfc->vfc_typenum;
890 bcopy((caddr_t)mp->mnt_stat.f_mntonname,
891 (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
892 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
893 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
894 }
895 return (0);
896}
897
898/*
899 * Go through the disk queues to initiate sandbagged IO;
900 * go through the inodes to write those that have been modified;
901 * initiate the writing of the super block if it has been modified.
902 *
903 * Note: we are always called with the filesystem marked `MPBUSY'.
904 */
905static int
906ext2_sync(mp, waitfor, cred, p)
907 struct mount *mp;
908 int waitfor;
909 struct ucred *cred;
910 struct proc *p;
911{
912 struct vnode *nvp, *vp;
913 struct inode *ip;
914 struct ufsmount *ump = VFSTOUFS(mp);
915 struct ext2_sb_info *fs;
916 int error, allerror = 0;
917
918 fs = ump->um_e2fs;
919 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
920 printf("fs = %s\n", fs->fs_fsmnt);
921 panic("ext2_sync: rofs mod");
922 }
923 /*
924 * Write back each (modified) inode.
925 */
926 simple_lock(&mntvnode_slock);
927loop:
928 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp != NULL; vp = nvp) {
929 /*
930 * If the vnode that we are about to sync is no longer
931 * associated with this mount point, start over.
932 */
933 if (vp->v_mount != mp)
934 goto loop;
935 simple_lock(&vp->v_interlock);
936 nvp = TAILQ_NEXT(vp, v_nmntvnodes);
937 ip = VTOI(vp);
938 if (vp->v_type == VNON ||
939 ((ip->i_flag &
940 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
941 (TAILQ_EMPTY(&vp->v_dirtyblkhd) || waitfor == MNT_LAZY))) {
942 simple_unlock(&vp->v_interlock);
943 continue;
944 }
945 simple_unlock(&mntvnode_slock);
946 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, p);
947 if (error) {
948 simple_lock(&mntvnode_slock);
949 if (error == ENOENT)
950 goto loop;
951 continue;
952 }
953 if ((error = VOP_FSYNC(vp, cred, waitfor, p)) != 0)
954 allerror = error;
955 VOP_UNLOCK(vp, 0, p);
956 vrele(vp);
957 simple_lock(&mntvnode_slock);
958 }
959 simple_unlock(&mntvnode_slock);
960 /*
961 * Force stale file system control information to be flushed.
962 */
963 if (waitfor != MNT_LAZY) {
964 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, p);
965 if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) != 0)
966 allerror = error;
967 VOP_UNLOCK(ump->um_devvp, 0, p);
968 }
969#if QUOTA
970 qsync(mp);
971#endif
972 /*
973 * Write back modified superblock.
974 */
975 if (fs->s_dirt != 0) {
976 fs->s_dirt = 0;
977 fs->s_es->s_wtime = time_second;
978 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
979 allerror = error;
980 }
981 return (allerror);
982}
983
984/*
985 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
986 * in from disk. If it is in core, wait for the lock bit to clear, then
987 * return the inode locked. Detection and handling of mount points must be
988 * done by the calling routine.
989 */
990static int
991ext2_vget(mp, ino, vpp)
992 struct mount *mp;
993 ino_t ino;
994 struct vnode **vpp;
995{
996 register struct ext2_sb_info *fs;
997 register struct inode *ip;
998 struct ufsmount *ump;
999 struct buf *bp;
1000 struct vnode *vp;
1001 dev_t dev;
1002 int i, error;
1003 int used_blocks;
1004
1005 ump = VFSTOUFS(mp);
1006 dev = ump->um_dev;
1007restart:
1008 if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
1009 return (0);
1010
1011 /*
1012 * Lock out the creation of new entries in the FFS hash table in
1013 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1014 * may occur!
1015 */
1016 if (ext2fs_inode_hash_lock) {
1017 while (ext2fs_inode_hash_lock) {
1018 ext2fs_inode_hash_lock = -1;
1019 tsleep(&ext2fs_inode_hash_lock, PVM, "e2vget", 0);
1020 }
1021 goto restart;
1022 }
1023 ext2fs_inode_hash_lock = 1;
1024
1025 /*
1026 * If this MALLOC() is performed after the getnewvnode()
1027 * it might block, leaving a vnode with a NULL v_data to be
1028 * found by ext2_sync() if a sync happens to fire right then,
1029 * which will cause a panic because ext2_sync() blindly
1030 * dereferences vp->v_data (as well it should).
1031 */
1032 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1033
1034 /* Allocate a new vnode/inode. */
1035 if ((error = getnewvnode(VT_UFS, mp, ext2_vnodeop_p, &vp)) != 0) {
1036 if (ext2fs_inode_hash_lock < 0)
1037 wakeup(&ext2fs_inode_hash_lock);
1038 ext2fs_inode_hash_lock = 0;
1039 *vpp = NULL;
1040 FREE(ip, M_EXT2NODE);
1041 return (error);
1042 }
1043 bzero((caddr_t)ip, sizeof(struct inode));
1044 lockinit(&ip->i_lock, PINOD, "ext2in", 0, 0);
1045 vp->v_data = ip;
1046 ip->i_vnode = vp;
1047 ip->i_e2fs = fs = ump->um_e2fs;
1048 ip->i_dev = dev;
1049 ip->i_number = ino;
1050#if QUOTA
1051 for (i = 0; i < MAXQUOTAS; i++)
1052 ip->i_dquot[i] = NODQUOT;
1053#endif
1054 /*
1055 * Put it onto its hash chain and lock it so that other requests for
1056 * this inode will block if they arrive while we are sleeping waiting
1057 * for old data structures to be purged or for the contents of the
1058 * disk portion of this inode to be read.
1059 */
1060 ufs_ihashins(ip);
1061
1062 if (ext2fs_inode_hash_lock < 0)
1063 wakeup(&ext2fs_inode_hash_lock);
1064 ext2fs_inode_hash_lock = 0;
1065
1066 /* Read in the disk contents for the inode, copy into the inode. */
1067#if 0
1068printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1069#endif
1070 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1071 (int)fs->s_blocksize, NOCRED, &bp)) != 0) {
1072 /*
1073 * The inode does not contain anything useful, so it would
1074 * be misleading to leave it on its hash chain. With mode
1075 * still zero, it will be unlinked and returned to the free
1076 * list by vput().
1077 */
1078 vput(vp);
1079 brelse(bp);
1080 *vpp = NULL;
1081 return (error);
1082 }
1083 /* convert ext2 inode to dinode */
1084 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE *
1085 ino_to_fsbo(fs, ino)), &ip->i_din);
1086 ip->i_block_group = ino_to_cg(fs, ino);
1087 ip->i_next_alloc_block = 0;
1088 ip->i_next_alloc_goal = 0;
1089 ip->i_prealloc_count = 0;
1090 ip->i_prealloc_block = 0;
1091 /* now we want to make sure that block pointers for unused
1092 blocks are zeroed out - ext2_balloc depends on this
1093 although for regular files and directories only
1094 */
1095 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1096 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1097 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1098 ip->i_db[i] = 0;
1099 }
1100/*
1101 ext2_print_inode(ip);
1102*/
1103 brelse(bp);
1104
1105 /*
1106 * Initialize the vnode from the inode, check for aliases.
1107 * Note that the underlying vnode may have changed.
1108 */
1109 if ((error = ufs_vinit(mp, ext2_specop_p, ext2_fifoop_p, &vp)) != 0) {
1110 vput(vp);
1111 *vpp = NULL;
1112 return (error);
1113 }
1114 /*
1115 * Finish inode initialization now that aliasing has been resolved.
1116 */
1117 ip->i_devvp = ump->um_devvp;
1118 VREF(ip->i_devvp);
1119 /*
1120 * Set up a generation number for this inode if it does not
1121 * already have one. This should only happen on old filesystems.
1122 */
1123 if (ip->i_gen == 0) {
1124 ip->i_gen = random() / 2 + 1;
1125 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1126 ip->i_flag |= IN_MODIFIED;
1127 }
1128 *vpp = vp;
1129 return (0);
1130}
1131
1132/*
1133 * File handle to vnode
1134 *
1135 * Have to be really careful about stale file handles:
1136 * - check that the inode number is valid
1137 * - call ext2_vget() to get the locked inode
1138 * - check for an unallocated inode (i_mode == 0)
1139 * - check that the given client host has export rights and return
1140 * those rights via. exflagsp and credanonp
1141 */
1142static int
1143ext2_fhtovp(mp, fhp, vpp)
1144 register struct mount *mp;
1145 struct fid *fhp;
1146 struct vnode **vpp;
1147{
1148 register struct ufid *ufhp;
1149 struct ext2_sb_info *fs;
1150
1151 ufhp = (struct ufid *)fhp;
1152 fs = VFSTOUFS(mp)->um_e2fs;
1153 if (ufhp->ufid_ino < ROOTINO ||
1154 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1155 return (ESTALE);
1156 return (ufs_fhtovp(mp, ufhp, vpp));
1157}
1158
1159/*
1160 * Vnode pointer to File handle
1161 */
1162/* ARGSUSED */
1163static int
1164ext2_vptofh(vp, fhp)
1165 struct vnode *vp;
1166 struct fid *fhp;
1167{
1168 register struct inode *ip;
1169 register struct ufid *ufhp;
1170
1171 ip = VTOI(vp);
1172 ufhp = (struct ufid *)fhp;
1173 ufhp->ufid_len = sizeof(struct ufid);
1174 ufhp->ufid_ino = ip->i_number;
1175 ufhp->ufid_gen = ip->i_gen;
1176 return (0);
1177}
1178
1179/*
1180 * Write a superblock and associated information back to disk.
1181 */
1182static int
1183ext2_sbupdate(mp, waitfor)
1184 struct ufsmount *mp;
1185 int waitfor;
1186{
1187 register struct ext2_sb_info *fs = mp->um_e2fs;
1188 register struct ext2_super_block *es = fs->s_es;
1189 register struct buf *bp;
1190 int error = 0;
1191/*
1192printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
1193*/
1194 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1195 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1196 if (waitfor == MNT_WAIT)
1197 error = bwrite(bp);
1198 else
1199 bawrite(bp);
1200
1201 /*
1202 * The buffers for group descriptors, inode bitmaps and block bitmaps
1203 * are not busy at this point and are (hopefully) written by the
1204 * usual sync mechanism. No need to write them here
1205 */
1206
1207 return (error);
1208}