MFC numerous features from HEAD.
[dragonfly.git] / sys / vfs / gnu / ext2fs / ext2_vfsops.c
blobfc72108d936c327edfb09486ad124f9fc6ea850f
1 /*
2 * modified for EXT2FS support in Lites 1.1
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.
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.
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.
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 * $DragonFly: src/sys/vfs/gnu/ext2fs/ext2_vfsops.c,v 1.56.4.1 2008/09/25 02:20:49 dillon Exp $
44 #include "opt_quota.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/nlookup.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/diskslice.h>
57 #include <sys/malloc.h>
58 #include <sys/stat.h>
59 #include <vm/vm_zone.h>
61 #include <sys/buf2.h>
62 #include <sys/thread2.h>
64 #include "quota.h"
65 #include "dinode.h"
66 #include "inode.h"
67 #include "ext2mount.h"
68 #include "ext2_extern.h"
70 #include "fs.h"
71 #include "ext2_fs.h"
72 #include "ext2_fs_sb.h"
74 extern struct vop_ops ext2_vnode_vops;
75 extern struct vop_ops ext2_spec_vops;
76 extern struct vop_ops ext2_fifo_vops;
78 static int ext2_fhtovp (struct mount *, struct vnode *,
79 struct fid *, struct vnode **);
80 static int ext2_flushfiles (struct mount *mp, int flags);
81 static int ext2_mount (struct mount *, char *, caddr_t, struct ucred *);
82 static int ext2_mountfs (struct vnode *, struct mount *, struct ucred *);
83 static int ext2_root(struct mount *, struct vnode **);
84 static int ext2_reload (struct mount *mountp, struct ucred *cred);
85 static int ext2_sbupdate (struct ext2mount *, int);
86 static int ext2_statfs (struct mount *, struct statfs *, struct ucred *);
87 static int ext2_sync (struct mount *, int);
88 static int ext2_unmount (struct mount *, int);
89 static int ext2_vget (struct mount *, ino_t, struct vnode **);
90 static int ext2_init(struct vfsconf *);
91 static int ext2_vptofh (struct vnode *, struct fid *);
93 static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part");
94 MALLOC_DEFINE(M_EXT2MNT, "EXT2 mount", "EXT2 mount structure");
96 static struct vfsops ext2fs_vfsops = {
97 .vfs_mount = ext2_mount,
98 .vfs_unmount = ext2_unmount,
99 .vfs_root = ext2_root, /* root inode via vget */
100 .vfs_quotactl = ext2_quotactl, /* quota operations */
101 .vfs_statfs = ext2_statfs,
102 .vfs_sync = ext2_sync,
103 .vfs_vget = ext2_vget,
104 .vfs_fhtovp = ext2_fhtovp,
105 .vfs_checkexp = ext2_check_export,
106 .vfs_vptofh = ext2_vptofh,
107 .vfs_init = ext2_init,
108 .vfs_uninit = ext2_uninit
111 VFS_SET(ext2fs_vfsops, ext2fs, 0);
112 #define bsd_malloc kmalloc
113 #define bsd_free kfree
115 static int ext2fs_inode_hash_lock;
117 static int ext2_check_sb_compat (struct ext2_super_block *es,
118 cdev_t dev, int ronly);
119 static int compute_sb_data (struct vnode * devvp,
120 struct ext2_super_block * es,
121 struct ext2_sb_info * fs);
123 static int
124 ext2_root(struct mount *mp, struct vnode **vpp)
126 struct vnode *nvp;
127 int error;
129 error = VFS_VGET(mp, (ino_t)ROOTINO, &nvp);
130 if (error)
131 return (error);
132 *vpp = nvp;
133 return (0);
137 * Do operations associated with quotas
140 ext2_quotactl(struct mount *mp, int cmds, uid_t uid, caddr_t arg,
141 struct ucred *cred)
143 #ifndef QUOTA
144 return (EOPNOTSUPP);
145 #else
146 int cmd, type, error;
148 type = cmds & SUBCMDMASK;
149 cmd = cmds >> SUBCMDSHIFT;
151 if (uid == -1) {
152 switch(type) {
153 case USRQUOTA:
154 uid = cred->cr_ruid;
155 break;
156 case GRPQUOTA:
157 uid = cred->cr_rgid;
158 break;
159 default:
160 return (EINVAL);
164 switch (cmd) {
165 case Q_SYNC:
166 break;
167 case Q_GETQUOTA:
168 if (uid == cred->cr_ruid)
169 break;
170 /* fall through */
171 default:
172 if ((error = suser_cred(cred, PRISON_ROOT)) != 0)
173 return (error);
176 type = cmds & SUBCMDMASK;
177 if ((uint)type >= MAXQUOTAS)
178 return (EINVAL);
179 if (vfs_busy(mp, LK_NOWAIT))
180 return (0);
182 switch (cmd) {
184 case Q_QUOTAON:
185 error = ext2_quotaon(cred, mp, type, arg);
186 break;
188 case Q_QUOTAOFF:
189 error = ext2_quotaoff(mp, type);
190 break;
192 case Q_SETQUOTA:
193 error = ext2_setquota(mp, uid, type, arg);
194 break;
196 case Q_SETUSE:
197 error = ext2_setuse(mp, uid, type, arg);
198 break;
200 case Q_GETQUOTA:
201 error = ext2_getquota(mp, uid, type, arg);
202 break;
204 case Q_SYNC:
205 error = ext2_qsync(mp);
206 break;
208 default:
209 error = EINVAL;
210 break;
212 vfs_unbusy(mp);
213 return (error);
214 #endif
218 * Initial UFS filesystems, done only once.
221 ext2_init(struct vfsconf *vfsp)
223 static int done;
225 if (done)
226 return (0);
227 done = 1;
228 ext2_ihashinit();
229 #ifdef QUOTA
230 ext2_dqinit();
231 #endif
232 return (0);
236 * VFS Operations.
238 * mount system call
240 * Parameters:
241 * data: this is actually a (struct ext2_args *)
243 static int
244 ext2_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
246 struct vnode *devvp;
247 struct ext2_args args;
248 struct ext2mount *ump = 0;
249 struct ext2_sb_info *fs;
250 size_t size;
251 int error, flags;
252 mode_t accessmode;
253 struct nlookupdata nd;
255 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ext2_args))) != 0)
256 return (error);
259 * If updating, check whether changing from read-only to
260 * read/write; if there is no device name, that's all we do.
262 if (mp->mnt_flag & MNT_UPDATE) {
263 ump = VFSTOEXT2(mp);
264 fs = ump->um_e2fs;
265 devvp = ump->um_devvp;
266 error = 0;
267 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
268 flags = WRITECLOSE;
269 if (mp->mnt_flag & MNT_FORCE)
270 flags |= FORCECLOSE;
271 if (vfs_busy(mp, LK_NOWAIT))
272 return (EBUSY);
273 error = ext2_flushfiles(mp, flags);
274 vfs_unbusy(mp);
275 if (!error && fs->s_wasvalid) {
276 fs->s_es->s_state |= EXT2_VALID_FS;
277 ext2_sbupdate(ump, MNT_WAIT);
279 fs->s_rd_only = 1;
280 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
281 VOP_OPEN(devvp, FREAD, FSCRED, NULL);
282 VOP_CLOSE(devvp, FREAD|FWRITE);
283 vn_unlock(devvp);
285 if (!error && (mp->mnt_flag & MNT_RELOAD))
286 error = ext2_reload(mp, cred);
287 if (error)
288 return (error);
289 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
290 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
291 return (EPERM);
292 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
294 * If upgrade to read-write by non-root, then verify
295 * that user has necessary permissions on the device.
297 if (cred->cr_uid != 0) {
298 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
299 error = VOP_ACCESS(devvp, VREAD | VWRITE, cred);
300 if (error) {
301 vn_unlock(devvp);
302 return (error);
304 vn_unlock(devvp);
307 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
308 (fs->s_es->s_state & EXT2_ERROR_FS)) {
309 if (mp->mnt_flag & MNT_FORCE) {
310 kprintf(
311 "WARNING: %s was not properly dismounted\n",
312 fs->fs_fsmnt);
313 } else {
314 kprintf(
315 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
316 fs->fs_fsmnt);
317 return (EPERM);
320 fs->s_es->s_state &= ~EXT2_VALID_FS;
321 ext2_sbupdate(ump, MNT_WAIT);
322 fs->s_rd_only = 0;
323 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
324 VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL);
325 VOP_CLOSE(devvp, FREAD);
326 vn_unlock(devvp);
328 if (args.fspec == 0) {
330 * Process export requests.
332 return (vfs_export(mp, &ump->um_export, &args.export));
336 * Not an update, or updating the name: look up the name
337 * and verify that it refers to a sensible block device.
339 devvp = NULL;
340 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
341 if (error == 0)
342 error = nlookup(&nd);
343 if (error == 0)
344 error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
345 nlookup_done(&nd);
346 if (error)
347 return (error);
349 if (!vn_isdisk(devvp, &error)) {
350 vrele(devvp);
351 return (error);
355 * If mount by non-root, then verify that user has necessary
356 * permissions on the device.
358 if (cred->cr_uid != 0) {
359 accessmode = VREAD;
360 if ((mp->mnt_flag & MNT_RDONLY) == 0)
361 accessmode |= VWRITE;
362 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
363 if ((error = VOP_ACCESS(devvp, accessmode, cred)) != 0) {
364 vput(devvp);
365 return (error);
367 vn_unlock(devvp);
370 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
371 error = ext2_mountfs(devvp, mp, cred);
372 } else {
373 if (devvp != ump->um_devvp)
374 error = EINVAL; /* needs translation */
375 else
376 vrele(devvp);
378 if (error) {
379 vrele(devvp);
380 return (error);
382 ump = VFSTOEXT2(mp);
383 fs = ump->um_e2fs;
384 copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
385 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
386 copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size);
387 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
388 ext2_statfs(mp, &mp->mnt_stat, cred);
389 return (0);
393 * checks that the data in the descriptor blocks make sense
394 * this is taken from ext2/super.c
396 static int
397 ext2_check_descriptors(struct ext2_sb_info *sb)
399 int i;
400 int desc_block = 0;
401 unsigned long block = sb->s_es->s_first_data_block;
402 struct ext2_group_desc * gdp = NULL;
404 /* ext2_debug ("Checking group descriptors"); */
406 for (i = 0; i < sb->s_groups_count; i++)
408 /* examine next descriptor block */
409 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
410 gdp = (struct ext2_group_desc *)
411 sb->s_group_desc[desc_block++]->b_data;
412 if (gdp->bg_block_bitmap < block ||
413 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
415 kprintf ("ext2_check_descriptors: "
416 "Block bitmap for group %d"
417 " not in group (block %lu)!\n",
418 i, (unsigned long) gdp->bg_block_bitmap);
419 return 0;
421 if (gdp->bg_inode_bitmap < block ||
422 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
424 kprintf ("ext2_check_descriptors: "
425 "Inode bitmap for group %d"
426 " not in group (block %lu)!\n",
427 i, (unsigned long) gdp->bg_inode_bitmap);
428 return 0;
430 if (gdp->bg_inode_table < block ||
431 gdp->bg_inode_table + sb->s_itb_per_group >=
432 block + EXT2_BLOCKS_PER_GROUP(sb))
434 kprintf ("ext2_check_descriptors: "
435 "Inode table for group %d"
436 " not in group (block %lu)!\n",
437 i, (unsigned long) gdp->bg_inode_table);
438 return 0;
440 block += EXT2_BLOCKS_PER_GROUP(sb);
441 gdp++;
443 return 1;
446 static int
447 ext2_check_sb_compat(struct ext2_super_block *es, cdev_t dev, int ronly)
449 if (es->s_magic != EXT2_SUPER_MAGIC) {
450 kprintf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
451 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
452 return (1);
454 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
455 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
456 kprintf(
457 "WARNING: mount of %s denied due to unsupported optional features\n",
458 devtoname(dev));
459 return (1);
461 if (!ronly &&
462 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
463 kprintf(
464 "WARNING: R/W mount of %s denied due to unsupported optional features\n",
465 devtoname(dev));
466 return (1);
469 return (0);
473 * this computes the fields of the ext2_sb_info structure from the
474 * data in the ext2_super_block structure read in
476 static int
477 compute_sb_data(struct vnode *devvp, struct ext2_super_block *es,
478 struct ext2_sb_info *fs)
480 int db_count, error;
481 int i, j;
482 int logic_sb_block = 1; /* XXX for now */
484 #if 1
485 #define V(v)
486 #else
487 #define V(v) kprintf(#v"= %d\n", fs->v);
488 #endif
490 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
491 V(s_blocksize)
492 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
493 V(s_bshift)
494 fs->s_fsbtodb = es->s_log_block_size + 1;
495 V(s_fsbtodb)
496 fs->s_qbmask = fs->s_blocksize - 1;
497 V(s_bmask)
498 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
499 V(s_blocksize_bits)
500 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
501 V(s_frag_size)
502 if (fs->s_frag_size)
503 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
504 V(s_frags_per_block)
505 fs->s_blocks_per_group = es->s_blocks_per_group;
506 V(s_blocks_per_group)
507 fs->s_frags_per_group = es->s_frags_per_group;
508 V(s_frags_per_group)
509 fs->s_inodes_per_group = es->s_inodes_per_group;
510 V(s_inodes_per_group)
511 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE;
512 V(s_inodes_per_block)
513 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
514 V(s_itb_per_group)
515 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
516 V(s_desc_per_block)
517 /* s_resuid / s_resgid ? */
518 fs->s_groups_count = (es->s_blocks_count -
519 es->s_first_data_block +
520 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
521 EXT2_BLOCKS_PER_GROUP(fs);
522 V(s_groups_count)
523 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
524 EXT2_DESC_PER_BLOCK(fs);
525 fs->s_db_per_group = db_count;
526 V(s_db_per_group)
528 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
529 M_EXT2MNT, M_WAITOK);
531 /* adjust logic_sb_block */
532 if(fs->s_blocksize > SBSIZE)
533 /* Godmar thinks: if the blocksize is greater than 1024, then
534 the superblock is logically part of block zero.
536 logic_sb_block = 0;
538 for (i = 0; i < db_count; i++) {
539 error = bread(devvp, fsbtodoff(fs, logic_sb_block + i + 1),
540 fs->s_blocksize, &fs->s_group_desc[i]);
541 if(error) {
542 for (j = 0; j < i; j++)
543 brelse(fs->s_group_desc[j]);
544 bsd_free(fs->s_group_desc, M_EXT2MNT);
545 kprintf("EXT2-fs: unable to read group descriptors (%d)\n", error);
546 return EIO;
548 /* Set the B_LOCKED flag on the buffer, then brelse() it */
549 LCK_BUF(fs->s_group_desc[i])
551 if(!ext2_check_descriptors(fs)) {
552 for (j = 0; j < db_count; j++)
553 ULCK_BUF(fs->s_group_desc[j])
554 bsd_free(fs->s_group_desc, M_EXT2MNT);
555 kprintf("EXT2-fs: (ext2_check_descriptors failure) "
556 "unable to read group descriptors\n");
557 return EIO;
560 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
561 fs->s_inode_bitmap_number[i] = 0;
562 fs->s_inode_bitmap[i] = NULL;
563 fs->s_block_bitmap_number[i] = 0;
564 fs->s_block_bitmap[i] = NULL;
566 fs->s_loaded_inode_bitmaps = 0;
567 fs->s_loaded_block_bitmaps = 0;
568 return 0;
572 * Reload all incore data for a filesystem (used after running fsck on
573 * the root filesystem and finding things to fix). The filesystem must
574 * be mounted read-only.
576 * Things to do to update the mount:
577 * 1) invalidate all cached meta-data.
578 * 2) re-read superblock from disk.
579 * 3) re-read summary information from disk.
580 * 4) invalidate all inactive vnodes.
581 * 5) invalidate all cached file data.
582 * 6) re-read inode data for all active vnodes.
584 static int ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *rescan);
585 static int ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *rescan);
587 struct scaninfo {
588 int rescan;
589 int allerror;
590 int waitfor;
591 struct vnode *devvp;
592 struct ext2_sb_info *fs;
595 static int
596 ext2_reload(struct mount *mountp, struct ucred *cred)
598 struct vnode *devvp;
599 struct buf *bp;
600 struct ext2_super_block * es;
601 struct ext2_sb_info *fs;
602 int error;
603 struct scaninfo scaninfo;
605 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
606 return (EINVAL);
608 * Step 1: invalidate all cached meta-data.
610 devvp = VFSTOEXT2(mountp)->um_devvp;
611 if (vinvalbuf(devvp, 0, 0, 0))
612 panic("ext2_reload: dirty1");
614 * Step 2: re-read superblock from disk.
615 * constants have been adjusted for ext2
617 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
618 return (error);
619 es = (struct ext2_super_block *)bp->b_data;
620 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
621 brelse(bp);
622 return (EIO); /* XXX needs translation */
624 fs = VFSTOEXT2(mountp)->um_e2fs;
625 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
627 if((error = compute_sb_data(devvp, es, fs)) != 0) {
628 brelse(bp);
629 return error;
631 #ifdef UNKLAR
632 if (fs->fs_sbsize < SBSIZE)
633 bp->b_flags |= B_INVAL;
634 #endif
635 brelse(bp);
637 scaninfo.rescan = 1;
638 scaninfo.devvp = devvp;
639 scaninfo.fs = fs;
640 while (error == 0 && scaninfo.rescan) {
641 scaninfo.rescan = 0;
642 error = vmntvnodescan(mountp, VMSC_GETVX, ext2_reload_scan1,
643 ext2_reload_scan2, &scaninfo);
645 return(error);
648 static int
649 ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
651 /*struct scaninfo *info = data;*/
653 return(0);
656 static int
657 ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
659 struct scaninfo *info = data;
660 struct inode *ip;
661 struct buf *bp;
662 int error;
665 * Try to recycle
667 if (vrecycle(vp))
668 return(0);
671 * Step 5: invalidate all cached file data.
673 if (vinvalbuf(vp, 0, 0, 0))
674 panic("ext2_reload: dirty2");
676 * Step 6: re-read inode data for all active vnodes.
678 ip = VTOI(vp);
679 error = bread(info->devvp,
680 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
681 (int)info->fs->s_blocksize, &bp);
682 if (error)
683 return (error);
684 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
685 EXT2_INODE_SIZE * ino_to_fsbo(info->fs, ip->i_number)),
686 &ip->i_din);
687 brelse(bp);
688 return(0);
692 * Common code for mount and mountroot
694 static int
695 ext2_mountfs(struct vnode *devvp, struct mount *mp, struct ucred *cred)
697 struct ext2mount *ump;
698 struct buf *bp;
699 struct ext2_sb_info *fs;
700 struct ext2_super_block * es;
701 cdev_t dev;
702 struct partinfo dpart;
703 int error, i, size;
704 int ronly;
707 * Disallow multiple mounts of the same device.
708 * Disallow mounting of a device that is currently in use
709 * (except for root, which might share swap device for miniroot).
710 * Flush out any old buffers remaining from a previous use.
712 if ((error = vfs_mountedon(devvp)) != 0)
713 return (error);
714 if (count_udev(devvp->v_umajor, devvp->v_uminor) > 0)
715 return (EBUSY);
716 if ((error = vinvalbuf(devvp, V_SAVE, 0, 0)) != 0)
717 return (error);
718 #ifdef READONLY
719 /* turn on this to force it to be read-only */
720 mp->mnt_flag |= MNT_RDONLY;
721 #endif
723 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
724 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
725 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
726 vn_unlock(devvp);
727 if (error)
728 return (error);
729 dev = devvp->v_rdev;
730 if (dev->si_iosize_max != 0)
731 mp->mnt_iosize_max = dev->si_iosize_max;
732 if (mp->mnt_iosize_max > MAXPHYS)
733 mp->mnt_iosize_max = MAXPHYS;
734 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred) != 0) {
735 size = DEV_BSIZE;
736 } else {
737 size = dpart.media_blksize;
740 bp = NULL;
741 ump = NULL;
742 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
743 goto out;
744 es = (struct ext2_super_block *)bp->b_data;
745 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
746 error = EINVAL; /* XXX needs translation */
747 goto out;
749 if ((es->s_state & EXT2_VALID_FS) == 0 ||
750 (es->s_state & EXT2_ERROR_FS)) {
751 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
752 kprintf(
753 "WARNING: Filesystem was not properly dismounted\n");
754 } else {
755 kprintf(
756 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
757 error = EPERM;
758 goto out;
761 ump = bsd_malloc(sizeof *ump, M_EXT2MNT, M_WAITOK | M_ZERO);
762 ump->um_malloctype = M_EXT2NODE;
763 ump->um_blkatoff = ext2_blkatoff;
764 ump->um_truncate = ext2_truncate;
765 ump->um_update = ext2_update;
766 ump->um_valloc = ext2_valloc;
767 ump->um_vfree = ext2_vfree;
768 /* I don't know whether this is the right strategy. Note that
769 we dynamically allocate both a ext2_sb_info and a ext2_super_block
770 while Linux keeps the super block in a locked buffer
772 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
773 M_EXT2MNT, M_WAITOK);
774 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
775 M_EXT2MNT, M_WAITOK);
776 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
777 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
778 goto out;
780 * We don't free the group descriptors allocated by compute_sb_data()
781 * until ext2_unmount(). This is OK since the mount will succeed.
783 brelse(bp);
784 bp = NULL;
785 fs = ump->um_e2fs;
786 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
787 /* if the fs is not mounted read-only, make sure the super block is
788 always written back on a sync()
790 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
791 if (ronly == 0) {
792 fs->s_dirt = 1; /* mark it modified */
793 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
795 mp->mnt_data = (qaddr_t)ump;
796 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
797 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
798 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
799 mp->mnt_flag |= MNT_LOCAL;
800 ump->um_mountp = mp;
801 ump->um_dev = dev;
802 ump->um_devvp = devvp;
803 /* setting those two parameters allows us to use
804 ext2_bmap w/o changse !
806 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
807 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
808 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
809 for (i = 0; i < MAXQUOTAS; i++)
810 ump->um_quotas[i] = NULLVP;
811 dev->si_mountpoint = mp;
813 vfs_add_vnodeops(mp, &ext2_vnode_vops, &mp->mnt_vn_norm_ops);
814 vfs_add_vnodeops(mp, &ext2_spec_vops, &mp->mnt_vn_spec_ops);
815 vfs_add_vnodeops(mp, &ext2_fifo_vops, &mp->mnt_vn_fifo_ops);
817 if (ronly == 0)
818 ext2_sbupdate(ump, MNT_WAIT);
819 return (0);
820 out:
821 if (bp)
822 brelse(bp);
823 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE);
824 if (ump) {
825 bsd_free(ump->um_e2fs->s_es, M_EXT2MNT);
826 bsd_free(ump->um_e2fs, M_EXT2MNT);
827 bsd_free(ump, M_EXT2MNT);
828 mp->mnt_data = (qaddr_t)0;
830 return (error);
834 * unmount system call
836 static int
837 ext2_unmount(struct mount *mp, int mntflags)
839 struct ext2mount *ump;
840 struct ext2_sb_info *fs;
841 int error, flags, ronly, i;
843 flags = 0;
844 if (mntflags & MNT_FORCE) {
845 if (mp->mnt_flag & MNT_ROOTFS)
846 return (EINVAL);
847 flags |= FORCECLOSE;
849 if ((error = ext2_flushfiles(mp, flags)) != 0)
850 return (error);
851 ump = VFSTOEXT2(mp);
852 fs = ump->um_e2fs;
853 ronly = fs->s_rd_only;
854 if (ronly == 0) {
855 if (fs->s_wasvalid)
856 fs->s_es->s_state |= EXT2_VALID_FS;
857 ext2_sbupdate(ump, MNT_WAIT);
860 /* release buffers containing group descriptors */
861 for(i = 0; i < fs->s_db_per_group; i++)
862 ULCK_BUF(fs->s_group_desc[i])
863 bsd_free(fs->s_group_desc, M_EXT2MNT);
865 /* release cached inode/block bitmaps */
866 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
867 if (fs->s_inode_bitmap[i])
868 ULCK_BUF(fs->s_inode_bitmap[i])
870 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
871 if (fs->s_block_bitmap[i])
872 ULCK_BUF(fs->s_block_bitmap[i])
874 ump->um_devvp->v_rdev->si_mountpoint = NULL;
875 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE);
876 vrele(ump->um_devvp);
877 bsd_free(fs->s_es, M_EXT2MNT);
878 bsd_free(fs, M_EXT2MNT);
879 bsd_free(ump, M_EXT2MNT);
880 mp->mnt_data = (qaddr_t)0;
881 mp->mnt_flag &= ~MNT_LOCAL;
882 return (error);
886 * Flush out all the files in a filesystem.
888 static int
889 ext2_flushfiles(struct mount *mp, int flags)
891 struct ext2mount *ump;
892 int error;
893 #if QUOTA
894 int i;
895 #endif
897 ump = VFSTOEXT2(mp);
898 #if QUOTA
899 if (mp->mnt_flag & MNT_QUOTA) {
900 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
901 return (error);
902 for (i = 0; i < MAXQUOTAS; i++) {
903 if (ump->um_quotas[i] == NULLVP)
904 continue;
905 ext2_quotaoff(mp, i);
908 * Here we fall through to vflush again to ensure
909 * that we have gotten rid of all the system vnodes.
912 #endif
913 error = vflush(mp, 0, flags);
914 return (error);
918 * Get file system statistics.
919 * taken from ext2/super.c ext2_statfs
921 static int
922 ext2_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
924 unsigned long overhead;
925 struct ext2mount *ump;
926 struct ext2_sb_info *fs;
927 struct ext2_super_block *es;
928 int i, nsb;
930 ump = VFSTOEXT2(mp);
931 fs = ump->um_e2fs;
932 es = fs->s_es;
934 if (es->s_magic != EXT2_SUPER_MAGIC)
935 panic("ext2_statfs - magic number spoiled");
938 * Compute the overhead (FS structures)
940 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
941 nsb = 0;
942 for (i = 0 ; i < fs->s_groups_count; i++)
943 if (ext2_group_sparse(i))
944 nsb++;
945 } else
946 nsb = fs->s_groups_count;
947 overhead = es->s_first_data_block +
948 /* Superblocks and block group descriptors: */
949 nsb * (1 + fs->s_db_per_group) +
950 /* Inode bitmap, block bitmap, and inode table: */
951 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
953 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
954 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
955 sbp->f_blocks = es->s_blocks_count - overhead;
956 sbp->f_bfree = es->s_free_blocks_count;
957 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
958 sbp->f_files = es->s_inodes_count;
959 sbp->f_ffree = es->s_free_inodes_count;
960 if (sbp != &mp->mnt_stat) {
961 sbp->f_type = mp->mnt_vfc->vfc_typenum;
962 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
963 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
965 return (0);
969 * Go through the disk queues to initiate sandbagged IO;
970 * go through the inodes to write those that have been modified;
971 * initiate the writing of the super block if it has been modified.
973 * Note: we are always called with the filesystem marked `MPBUSY'.
976 static int ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data);
978 static int
979 ext2_sync(struct mount *mp, int waitfor)
981 struct ext2mount *ump = VFSTOEXT2(mp);
982 struct ext2_sb_info *fs;
983 struct scaninfo scaninfo;
984 int error;
986 fs = ump->um_e2fs;
987 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
988 kprintf("fs = %s\n", fs->fs_fsmnt);
989 panic("ext2_sync: rofs mod");
993 * Write back each (modified) inode.
995 scaninfo.allerror = 0;
996 scaninfo.rescan = 1;
997 scaninfo.waitfor = waitfor;
998 while (scaninfo.rescan) {
999 scaninfo.rescan = 0;
1000 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
1001 NULL, ext2_sync_scan, &scaninfo);
1005 * Force stale file system control information to be flushed.
1007 if (waitfor != MNT_LAZY) {
1008 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1009 if ((error = VOP_FSYNC(ump->um_devvp, waitfor)) != 0)
1010 scaninfo.allerror = error;
1011 vn_unlock(ump->um_devvp);
1013 #if QUOTA
1014 ext2_qsync(mp);
1015 #endif
1017 * Write back modified superblock.
1019 if (fs->s_dirt != 0) {
1020 fs->s_dirt = 0;
1021 fs->s_es->s_wtime = time_second;
1022 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
1023 scaninfo.allerror = error;
1025 return (scaninfo.allerror);
1028 static int
1029 ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data)
1031 struct scaninfo *info = data;
1032 struct inode *ip;
1033 int error;
1035 ip = VTOI(vp);
1036 if (vp->v_type == VNON ||
1037 ((ip->i_flag &
1038 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1039 (RB_EMPTY(&vp->v_rbdirty_tree) || info->waitfor == MNT_LAZY))) {
1040 return(0);
1042 if ((error = VOP_FSYNC(vp, info->waitfor)) != 0)
1043 info->allerror = error;
1044 return(0);
1048 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
1049 * in from disk. If it is in core, wait for the lock bit to clear, then
1050 * return the inode locked. Detection and handling of mount points must be
1051 * done by the calling routine.
1053 static int
1054 ext2_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1056 struct ext2_sb_info *fs;
1057 struct inode *ip;
1058 struct ext2mount *ump;
1059 struct buf *bp;
1060 struct vnode *vp;
1061 cdev_t dev;
1062 int i, error;
1063 int used_blocks;
1065 ump = VFSTOEXT2(mp);
1066 dev = ump->um_dev;
1067 restart:
1068 if ((*vpp = ext2_ihashget(dev, ino)) != NULL)
1069 return (0);
1072 * Lock out the creation of new entries in the FFS hash table in
1073 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1074 * may occur!
1076 if (ext2fs_inode_hash_lock) {
1077 while (ext2fs_inode_hash_lock) {
1078 ext2fs_inode_hash_lock = -1;
1079 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
1081 goto restart;
1083 ext2fs_inode_hash_lock = 1;
1086 * If this MALLOC() is performed after the getnewvnode()
1087 * it might block, leaving a vnode with a NULL v_data to be
1088 * found by ext2_sync() if a sync happens to fire right then,
1089 * which will cause a panic because ext2_sync() blindly
1090 * dereferences vp->v_data (as well it should).
1092 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1094 /* Allocate a new vnode/inode. */
1095 if ((error = getnewvnode(VT_EXT2FS, mp, &vp, 0, LK_CANRECURSE)) != 0) {
1096 if (ext2fs_inode_hash_lock < 0)
1097 wakeup(&ext2fs_inode_hash_lock);
1098 ext2fs_inode_hash_lock = 0;
1099 *vpp = NULL;
1100 FREE(ip, M_EXT2NODE);
1101 return (error);
1103 bzero((caddr_t)ip, sizeof(struct inode));
1104 vp->v_data = ip;
1105 ip->i_vnode = vp;
1106 ip->i_e2fs = fs = ump->um_e2fs;
1107 ip->i_dev = dev;
1108 ip->i_number = ino;
1109 #if QUOTA
1110 for (i = 0; i < MAXQUOTAS; i++)
1111 ip->i_dquot[i] = NODQUOT;
1112 #endif
1114 * Put it onto its hash chain. Since our vnode is locked, other
1115 * requests for this inode will block if they arrive while we are
1116 * sleeping waiting for old data structures to be purged or for the
1117 * contents of the disk portion of this inode to be read.
1119 ext2_ihashins(ip);
1121 if (ext2fs_inode_hash_lock < 0)
1122 wakeup(&ext2fs_inode_hash_lock);
1123 ext2fs_inode_hash_lock = 0;
1125 /* Read in the disk contents for the inode, copy into the inode. */
1126 #if 0
1127 kprintf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1128 #endif
1129 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1130 (int)fs->s_blocksize, &bp);
1131 if (error) {
1133 * The inode does not contain anything useful, so it would
1134 * be misleading to leave it on its hash chain. With mode
1135 * still zero, it will be unlinked and returned to the free
1136 * list by vput().
1138 vx_put(vp);
1139 brelse(bp);
1140 *vpp = NULL;
1141 return (error);
1143 /* convert ext2 inode to dinode */
1144 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE *
1145 ino_to_fsbo(fs, ino)), &ip->i_din);
1146 ip->i_block_group = ino_to_cg(fs, ino);
1147 ip->i_next_alloc_block = 0;
1148 ip->i_next_alloc_goal = 0;
1149 ip->i_prealloc_count = 0;
1150 ip->i_prealloc_block = 0;
1151 /* now we want to make sure that block pointers for unused
1152 blocks are zeroed out - ext2_balloc depends on this
1153 although for regular files and directories only
1155 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1156 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1157 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1158 ip->i_db[i] = 0;
1160 #if 0
1161 ext2_print_inode(ip);
1162 #endif
1163 brelse(bp);
1166 * Initialize the vnode from the inode, check for aliases.
1167 * Note that the underlying vnode may have changed.
1169 if ((error = ext2_vinit(mp, &vp)) != 0) {
1170 vx_put(vp);
1171 *vpp = NULL;
1172 return (error);
1176 * Finish inode initialization now that aliasing has been resolved.
1178 ip->i_devvp = ump->um_devvp;
1179 vref(ip->i_devvp);
1181 * Set up a generation number for this inode if it does not
1182 * already have one. This should only happen on old filesystems.
1184 if (ip->i_gen == 0) {
1185 ip->i_gen = krandom() / 2 + 1;
1186 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1187 ip->i_flag |= IN_MODIFIED;
1190 * Return the locked and refd vnode.
1192 *vpp = vp;
1193 return (0);
1197 * File handle to vnode
1199 * Have to be really careful about stale file handles:
1200 * - check that the inode number is valid
1201 * - call ext2_vget() to get the locked inode
1202 * - check for an unallocated inode (i_mode == 0)
1203 * - check that the given client host has export rights and return
1204 * those rights via. exflagsp and credanonp
1206 static int
1207 ext2_fhtovp(struct mount *mp, struct vnode *rootvp,
1208 struct fid *fhp, struct vnode **vpp)
1210 struct ufid *ufhp;
1211 struct ext2_sb_info *fs;
1212 struct inode *ip;
1213 struct vnode *nvp;
1214 int error;
1216 ufhp = (struct ufid *)fhp;
1217 fs = VFSTOEXT2(mp)->um_e2fs;
1218 if (ufhp->ufid_ino < ROOTINO ||
1219 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1220 return (ESTALE);
1222 error = VFS_VGET(mp, ufhp->ufid_ino, &nvp);
1223 if (error) {
1224 *vpp = NULLVP;
1225 return (error);
1227 ip = VTOI(nvp);
1228 if (ip->i_mode == 0 ||
1229 ip->i_gen != ufhp->ufid_gen ||
1230 (VFSTOEXT2(mp)->um_i_effnlink_valid ? ip->i_effnlink :
1231 ip->i_nlink) <= 0) {
1232 vput(nvp);
1233 *vpp = NULLVP;
1234 return (ESTALE);
1236 *vpp = nvp;
1237 return (0);
1241 * Vnode pointer to File handle
1243 /* ARGSUSED */
1244 static int
1245 ext2_vptofh(struct vnode *vp, struct fid *fhp)
1247 struct inode *ip;
1248 struct ufid *ufhp;
1250 ip = VTOI(vp);
1251 ufhp = (struct ufid *)fhp;
1252 ufhp->ufid_len = sizeof(struct ufid);
1253 ufhp->ufid_ino = ip->i_number;
1254 ufhp->ufid_gen = ip->i_gen;
1255 return (0);
1259 * This is the generic part of fhtovp called after the underlying
1260 * filesystem has validated the file handle.
1262 * Verify that a host should have access to a filesystem.
1265 ext2_check_export(struct mount *mp, struct sockaddr *nam, int *exflagsp,
1266 struct ucred **credanonp)
1268 struct netcred *np;
1269 struct ext2mount *ump;
1271 ump = VFSTOEXT2(mp);
1273 * Get the export permission structure for this <mp, client> tuple.
1275 np = vfs_export_lookup(mp, &ump->um_export, nam);
1276 if (np == NULL)
1277 return (EACCES);
1279 *exflagsp = np->netc_exflags;
1280 *credanonp = &np->netc_anon;
1281 return (0);
1285 * Write a superblock and associated information back to disk.
1287 static int
1288 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1290 struct ext2_sb_info *fs = mp->um_e2fs;
1291 struct ext2_super_block *es = fs->s_es;
1292 struct buf *bp;
1293 int error = 0;
1295 kprintf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
1297 bp = getblk(mp->um_devvp, SBOFF, SBSIZE, 0, 0);
1298 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1299 if (waitfor == MNT_WAIT)
1300 error = bwrite(bp);
1301 else
1302 bawrite(bp);
1305 * The buffers for group descriptors, inode bitmaps and block bitmaps
1306 * are not busy at this point and are (hopefully) written by the
1307 * usual sync mechanism. No need to write them here
1310 return (error);