59#include <sys/disklabel.h>
60#include <sys/ioctl.h>
61#include <sys/errno.h>
62#include <sys/malloc.h>
63#include <sys/stat.h>
64
65#include <miscfs/specfs/specdev.h>
66
67#include <ufs/ufs/quota.h>
68#include <ufs/ufs/ufsmount.h>
69#include <ufs/ufs/inode.h>
70#include <ufs/ufs/ufs_extern.h>
71
72#include <gnu/ext2fs/fs.h>
73#include <gnu/ext2fs/ext2_extern.h>
74#include <gnu/ext2fs/ext2_fs.h>
75#include <gnu/ext2fs/ext2_fs_sb.h>
76
77static int ext2_fhtovp __P((struct mount *, struct fid *, struct mbuf *,
78 struct vnode **, int *, struct ucred **));
79static int ext2_flushfiles __P((struct mount *mp, int flags, struct proc *p));
80static int ext2_mount __P((struct mount *,
81 char *, caddr_t, struct nameidata *, struct proc *));
82static int ext2_mountfs __P((struct vnode *, struct mount *, struct proc *));
83static int ext2_reload __P((struct mount *mountp, struct ucred *cred,
84 struct proc *p));
85static int ext2_sbupdate __P((struct ufsmount *, int));
86static int ext2_statfs __P((struct mount *, struct statfs *, struct proc *));
87static int ext2_sync __P((struct mount *, int, struct ucred *, struct proc *));
88static int ext2_unmount __P((struct mount *, int, struct proc *));
89static int ext2_vget __P((struct mount *, ino_t, struct vnode **));
90static int ext2_vptofh __P((struct vnode *, struct fid *));
91
92static struct vfsops ext2fs_vfsops = {
93 ext2_mount,
94 ufs_start, /* empty function */
95 ext2_unmount,
96 ufs_root, /* root inode via vget */
97 ufs_quotactl, /* does operations associated with quotas */
98 ext2_statfs,
99 ext2_sync,
100 ext2_vget,
101 ext2_fhtovp,
102 ext2_vptofh,
103 ext2_init,
104};
105
106#if defined(__FreeBSD__)
107VFS_SET(ext2fs_vfsops, ext2fs, MOUNT_EXT2FS, 0);
108#define bsd_malloc malloc
109#define bsd_free free
110#endif
111
112extern u_long nextgennumber;
113#ifdef __FreeBSD__
114static int ext2fs_inode_hash_lock;
115#endif
116
117static int compute_sb_data __P((struct vnode * devvp,
118 struct ext2_super_block * es,
119 struct ext2_sb_info * fs));
120
121#ifdef notyet
122static int ext2_mountroot __P((void));
123
124/*
125 * Called by main() when ext2fs is going to be mounted as root.
126 *
127 * Name is updated by mount(8) after booting.
128 */
129#define ROOTNAME "root_device"
130
131static int
132ext2_mountroot()
133{
134#if !defined(__FreeBSD__)
135 extern struct vnode *rootvp;
136#endif
137 register struct ext2_sb_info *fs;
138 register struct mount *mp;
139#if defined(__FreeBSD__)
140 struct proc *p = curproc;
141#else
142 struct proc *p = get_proc(); /* XXX */
143#endif
144 struct ufsmount *ump;
145 u_int size;
146 int error;
147
148 /*
149 * Get vnodes for swapdev and rootdev.
150 */
151 if (bdevvp(swapdev, &swapdev_vp) || bdevvp(rootdev, &rootvp))
152 panic("ext2_mountroot: can't setup bdevvp's");
153
154 mp = bsd_malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
155 bzero((char *)mp, (u_long)sizeof(struct mount));
156 mp->mnt_op = &ext2fs_vfsops;
157 mp->mnt_flag = MNT_RDONLY;
158 if (error = ext2_mountfs(rootvp, mp, p)) {
159 bsd_free(mp, M_MOUNT);
160 return (error);
161 }
162 if (error = vfs_lock(mp)) {
163 (void)ext2_unmount(mp, 0, p);
164 bsd_free(mp, M_MOUNT);
165 return (error);
166 }
167#if defined(__FreeBSD__)
168 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
169#else
170 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
171#endif
172 mp->mnt_flag |= MNT_ROOTFS;
173 mp->mnt_vnodecovered = NULLVP;
174 ump = VFSTOUFS(mp);
175 fs = ump->um_e2fs;
176 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
177 fs->fs_fsmnt[0] = '/';
178 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
179 MNAMELEN);
180 (void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
181 &size);
182 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
183 (void)ext2_statfs(mp, &mp->mnt_stat, p);
184 vfs_unlock(mp);
185 inittodr(fs->s_es->s_wtime); /* this helps to set the time */
186 return (0);
187}
188#endif
189
190/*
191 * VFS Operations.
192 *
193 * mount system call
194 */
195static int
196ext2_mount(mp, path, data, ndp, p)
197 register struct mount *mp;
198 char *path;
199 caddr_t data; /* this is actually a (struct ufs_args *) */
200 struct nameidata *ndp;
201 struct proc *p;
202{
203 struct vnode *devvp;
204 struct ufs_args args;
205 struct ufsmount *ump = 0;
206 register struct ext2_sb_info *fs;
207 u_int size;
208 int error, flags;
209
210 if (error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args)))
211 return (error);
212 /*
213 * If updating, check whether changing from read-only to
214 * read/write; if there is no device name, that's all we do.
215 */
216 if (mp->mnt_flag & MNT_UPDATE) {
217 ump = VFSTOUFS(mp);
218 fs = ump->um_e2fs;
219 error = 0;
220 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
221 flags = WRITECLOSE;
222 if (mp->mnt_flag & MNT_FORCE)
223 flags |= FORCECLOSE;
224 if (vfs_busy(mp, LK_NOWAIT, 0, p))
225 return (EBUSY);
226 error = ext2_flushfiles(mp, flags, p);
227 vfs_unbusy(mp, p);
228 }
229 if (!error && (mp->mnt_flag & MNT_RELOAD))
230 error = ext2_reload(mp, ndp->ni_cnd.cn_cred, p);
231 if (error)
232 return (error);
233 if (fs->s_rd_only && (mp->mnt_flag & MNT_WANTRDWR))
234 fs->s_rd_only = 0;
235 if (fs->s_rd_only == 0) {
236 /* don't say it's clean */
237 fs->s_es->s_state &= ~EXT2_VALID_FS;
238 ext2_sbupdate(ump, MNT_WAIT);
239 }
240 if (args.fspec == 0) {
241 /*
242 * Process export requests.
243 */
244 return (vfs_export(mp, &ump->um_export, &args.export));
245 }
246 }
247 /*
248 * Not an update, or updating the name: look up the name
249 * and verify that it refers to a sensible block device.
250 */
251 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
252 if (error = namei(ndp))
253 return (error);
254 devvp = ndp->ni_vp;
255
256 if (devvp->v_type != VBLK) {
257 vrele(devvp);
258 return (ENOTBLK);
259 }
260 if (major(devvp->v_rdev) >= nblkdev) {
261 vrele(devvp);
262 return (ENXIO);
263 }
264 if ((mp->mnt_flag & MNT_UPDATE) == 0)
265 error = ext2_mountfs(devvp, mp, p);
266 else {
267 if (devvp != ump->um_devvp)
268 error = EINVAL; /* needs translation */
269 else
270 vrele(devvp);
271 }
272 if (error) {
273 vrele(devvp);
274 return (error);
275 }
276 ump = VFSTOUFS(mp);
277 fs = ump->um_e2fs;
278 (void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
279 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
280 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
281 MNAMELEN);
282 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
283 &size);
284 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
285 (void)ext2_statfs(mp, &mp->mnt_stat, p);
286 return (0);
287}
288
289/*
290 * checks that the data in the descriptor blocks make sense
291 * this is taken from ext2/super.c
292 */
293static int ext2_check_descriptors (struct ext2_sb_info * sb)
294{
295 int i;
296 int desc_block = 0;
297 unsigned long block = sb->s_es->s_first_data_block;
298 struct ext2_group_desc * gdp = NULL;
299
300 /* ext2_debug ("Checking group descriptors"); */
301
302 for (i = 0; i < sb->s_groups_count; i++)
303 {
304 /* examine next descriptor block */
305 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
306 gdp = (struct ext2_group_desc *)
307 sb->s_group_desc[desc_block++]->b_data;
308 if (gdp->bg_block_bitmap < block ||
309 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
310 {
311 printf ("ext2_check_descriptors: "
312 "Block bitmap for group %d"
313 " not in group (block %lu)!",
314 i, (unsigned long) gdp->bg_block_bitmap);
315 return 0;
316 }
317 if (gdp->bg_inode_bitmap < block ||
318 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
319 {
320 printf ("ext2_check_descriptors: "
321 "Inode bitmap for group %d"
322 " not in group (block %lu)!",
323 i, (unsigned long) gdp->bg_inode_bitmap);
324 return 0;
325 }
326 if (gdp->bg_inode_table < block ||
327 gdp->bg_inode_table + sb->s_itb_per_group >=
328 block + EXT2_BLOCKS_PER_GROUP(sb))
329 {
330 printf ("ext2_check_descriptors: "
331 "Inode table for group %d"
332 " not in group (block %lu)!",
333 i, (unsigned long) gdp->bg_inode_table);
334 return 0;
335 }
336 block += EXT2_BLOCKS_PER_GROUP(sb);
337 gdp++;
338 }
339 return 1;
340}
341
342/*
343 * this computes the fields of the ext2_sb_info structure from the
344 * data in the ext2_super_block structure read in
345 */
346static int compute_sb_data(devvp, es, fs)
347 struct vnode * devvp;
348 struct ext2_super_block * es;
349 struct ext2_sb_info * fs;
350{
351 int db_count, error;
352 int i, j;
353 int logic_sb_block = 1; /* XXX for now */
354
355#if 1
356#define V(v)
357#else
358#define V(v) printf(#v"= %d\n", fs->v);
359#endif
360
361 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
362 V(s_blocksize)
363 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
364 V(s_bshift)
365 fs->s_fsbtodb = es->s_log_block_size + 1;
366 V(s_fsbtodb)
367 fs->s_qbmask = fs->s_blocksize - 1;
368 V(s_bmask)
369 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
370 V(s_blocksize_bits)
371 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
372 V(s_frag_size)
373 if (fs->s_frag_size)
374 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
375 V(s_frags_per_block)
376 fs->s_blocks_per_group = es->s_blocks_per_group;
377 V(s_blocks_per_group)
378 fs->s_frags_per_group = es->s_frags_per_group;
379 V(s_frags_per_group)
380 fs->s_inodes_per_group = es->s_inodes_per_group;
381 V(s_inodes_per_group)
382 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE;
383 V(s_inodes_per_block)
384 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
385 V(s_itb_per_group)
386 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
387 V(s_desc_per_block)
388 /* s_resuid / s_resgid ? */
389 fs->s_groups_count = (es->s_blocks_count -
390 es->s_first_data_block +
391 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
392 EXT2_BLOCKS_PER_GROUP(fs);
393 V(s_groups_count)
394 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
395 EXT2_DESC_PER_BLOCK(fs);
396 fs->s_db_per_group = db_count;
397 V(s_db_per_group)
398
399 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
400 M_UFSMNT, M_WAITOK);
401
402 /* adjust logic_sb_block */
403 if(fs->s_blocksize > SBSIZE)
404 /* Godmar thinks: if the blocksize is greater than 1024, then
405 the superblock is logically part of block zero.
406 */
407 logic_sb_block = 0;
408
409 for (i = 0; i < db_count; i++) {
410 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1),
411 fs->s_blocksize, NOCRED, &fs->s_group_desc[i]);
412 if(error) {
413 for (j = 0; j < i; j++)
414 brelse(fs->s_group_desc[j]);
415 bsd_free(fs->s_group_desc, M_UFSMNT);
416 printf("EXT2-fs: unable to read group descriptors (%d)\n", error);
417 return EIO;
418 }
419 }
420 if(!ext2_check_descriptors(fs)) {
421 for (j = 0; j < db_count; j++)
422 brelse(fs->s_group_desc[j]);
423 bsd_free(fs->s_group_desc, M_UFSMNT);
424 printf("EXT2-fs: (ext2_check_descriptors failure) "
425 "unable to read group descriptors\n");
426 return EIO;
427 }
428
429 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
430 fs->s_inode_bitmap_number[i] = 0;
431 fs->s_inode_bitmap[i] = NULL;
432 fs->s_block_bitmap_number[i] = 0;
433 fs->s_block_bitmap[i] = NULL;
434 }
435 fs->s_loaded_inode_bitmaps = 0;
436 fs->s_loaded_block_bitmaps = 0;
437 return 0;
438}
439
440/*
441 * Reload all incore data for a filesystem (used after running fsck on
442 * the root filesystem and finding things to fix). The filesystem must
443 * be mounted read-only.
444 *
445 * Things to do to update the mount:
446 * 1) invalidate all cached meta-data.
447 * 2) re-read superblock from disk.
448 * 3) re-read summary information from disk.
449 * 4) invalidate all inactive vnodes.
450 * 5) invalidate all cached file data.
451 * 6) re-read inode data for all active vnodes.
452 */
453static int
454ext2_reload(mountp, cred, p)
455 register struct mount *mountp;
456 struct ucred *cred;
457 struct proc *p;
458{
459 register struct vnode *vp, *nvp, *devvp;
460 struct inode *ip;
461 struct buf *bp;
462 struct ext2_super_block * es;
463 struct ext2_sb_info *fs;
464 int error;
465
466 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
467 return (EINVAL);
468 /*
469 * Step 1: invalidate all cached meta-data.
470 */
471 devvp = VFSTOUFS(mountp)->um_devvp;
472 if (vinvalbuf(devvp, 0, cred, p, 0, 0))
473 panic("ext2_reload: dirty1");
474 /*
475 * Step 2: re-read superblock from disk.
476 * constants have been adjusted for ext2
477 */
478 if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp))
479 return (error);
480 es = (struct ext2_super_block *)bp->b_data;
481 if (es->s_magic != EXT2_SUPER_MAGIC) {
482 if(es->s_magic == EXT2_PRE_02B_MAGIC)
483 printf("This filesystem bears the magic number of a pre "
484 "0.2b version of ext2. This is not supported by "
485 "Lites.\n");
486 else
487 printf("Wrong magic number: %x (expected %x for ext2 fs\n",
488 es->s_magic, EXT2_SUPER_MAGIC);
489 brelse(bp);
490 return (EIO); /* XXX needs translation */
491 }
492 fs = VFSTOUFS(mountp)->um_e2fs;
493 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
494
495 if(error = compute_sb_data(devvp, es, fs)) {
496 brelse(bp);
497 return error;
498 }
499#ifdef UNKLAR
500 if (fs->fs_sbsize < SBSIZE)
501 bp->b_flags |= B_INVAL;
502#endif
503 brelse(bp);
504
505loop:
506 for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
507 nvp = vp->v_mntvnodes.le_next;
508 /*
509 * Step 4: invalidate all inactive vnodes.
510 */
511 if (vp->v_usecount == 0) {
512 vgone(vp);
513 continue;
514 }
515 /*
516 * Step 5: invalidate all cached file data.
517 */
518 if (vget(vp, LK_EXCLUSIVE, p))
519 goto loop;
520 if (vinvalbuf(vp, 0, cred, p, 0, 0))
521 panic("ext2_reload: dirty2");
522 /*
523 * Step 6: re-read inode data for all active vnodes.
524 */
525 ip = VTOI(vp);
526 if (error =
527 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
528 (int)fs->s_blocksize, NOCRED, &bp)) {
529 vput(vp);
530 return (error);
531 }
532 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
533 EXT2_INODE_SIZE * ino_to_fsbo(fs, ip->i_number)),
534 &ip->i_din);
535 brelse(bp);
536 vput(vp);
537 if (vp->v_mount != mountp)
538 goto loop;
539 }
540 return (0);
541}
542
543/*
544 * Common code for mount and mountroot
545 */
546static int
547ext2_mountfs(devvp, mp, p)
548 register struct vnode *devvp;
549 struct mount *mp;
550 struct proc *p;
551{
552 register struct ufsmount *ump;
553 struct buf *bp;
554 register struct ext2_sb_info *fs;
555 struct ext2_super_block * es;
556 dev_t dev = devvp->v_rdev;
557 struct partinfo dpart;
558 int havepart = 0;
559 int error, i, size;
560 int ronly;
561#if !defined(__FreeBSD__)
562 extern struct vnode *rootvp;
563#endif
564
565 /*
566 * Disallow multiple mounts of the same device.
567 * Disallow mounting of a device that is currently in use
568 * (except for root, which might share swap device for miniroot).
569 * Flush out any old buffers remaining from a previous use.
570 */
571 if (error = vfs_mountedon(devvp))
572 return (error);
573 if (vcount(devvp) > 1 && devvp != rootvp)
574 return (EBUSY);
575 if (error = vinvalbuf(devvp, V_SAVE, p->p_ucred, p, 0, 0))
576 return (error);
577#ifdef READONLY
578/* turn on this to force it to be read-only */
579 mp->mnt_flag |= MNT_RDONLY;
580#endif
581
582 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
583 if (error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p))
584 return (error);
585 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
586 size = DEV_BSIZE;
587 else {
588 havepart = 1;
589 size = dpart.disklab->d_secsize;
590 }
591
592 bp = NULL;
593 ump = NULL;
594 if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp))
595 goto out;
596 es = (struct ext2_super_block *)bp->b_data;
597 if (es->s_magic != EXT2_SUPER_MAGIC) {
598 if(es->s_magic == EXT2_PRE_02B_MAGIC)
599 printf("This filesystem bears the magic number of a pre "
600 "0.2b version of ext2. This is not supported by "
601 "Lites.\n");
602 else
603 printf("Wrong magic number: %x (expected %x for EXT2FS)\n",
604 es->s_magic, EXT2_SUPER_MAGIC);
605 error = EINVAL; /* XXX needs translation */
606 goto out;
607 }
608 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
609 bzero((caddr_t)ump, sizeof *ump);
610 /* I don't know whether this is the right strategy. Note that
611 we dynamically allocate both a ext2_sb_info and a ext2_super_block
612 while Linux keeps the super block in a locked buffer
613 */
614 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
615 M_UFSMNT, M_WAITOK);
616 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
617 M_UFSMNT, M_WAITOK);
618 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
619 if(error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)) {
620 brelse(bp);
621 return error;
622 }
623 brelse(bp);
624 bp = NULL;
625 fs = ump->um_e2fs;
626 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
627 if (!(fs->s_es->s_state & EXT2_VALID_FS)) {
628 printf("WARNING: %s was not properly dismounted\n",
629 fs->fs_fsmnt);
630 }
631 /* if the fs is not mounted read-only, make sure the super block is
632 always written back on a sync()
633 */
634 if (ronly == 0) {
635 fs->s_dirt = 1; /* mark it modified */
636 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
637 }
638 mp->mnt_data = (qaddr_t)ump;
639 mp->mnt_stat.f_fsid.val[0] = (long)dev;
640 mp->mnt_stat.f_fsid.val[1] = MOUNT_EXT2FS;
641 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
642 mp->mnt_flag |= MNT_LOCAL;
643 ump->um_mountp = mp;
644 ump->um_dev = dev;
645 ump->um_devvp = devvp;
646 /* setting those two parameters allows us to use
647 ufs_bmap w/o changse !
648 */
649 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
650 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
651 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
652 for (i = 0; i < MAXQUOTAS; i++)
653 ump->um_quotas[i] = NULLVP;
654 devvp->v_specflags |= SI_MOUNTEDON;
655 if (ronly == 0)
656 ext2_sbupdate(ump, MNT_WAIT);
657 return (0);
658out:
659 if (bp)
660 brelse(bp);
661 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
662 if (ump) {
663 bsd_free(ump->um_fs, M_UFSMNT);
664 bsd_free(ump, M_UFSMNT);
665 mp->mnt_data = (qaddr_t)0;
666 }
667 return (error);
668}
669
670/*
671 * unmount system call
672 */
673static int
674ext2_unmount(mp, mntflags, p)
675 struct mount *mp;
676 int mntflags;
677 struct proc *p;
678{
679 register struct ufsmount *ump;
680 register struct ext2_sb_info *fs;
681 int error, flags, ronly, i;
682
683 flags = 0;
684 if (mntflags & MNT_FORCE) {
685 if (mp->mnt_flag & MNT_ROOTFS)
686 return (EINVAL);
687 flags |= FORCECLOSE;
688 }
689 if (error = ext2_flushfiles(mp, flags, p))
690 return (error);
691 ump = VFSTOUFS(mp);
692 fs = ump->um_e2fs;
693 ronly = fs->s_rd_only;
694 if (!ronly) {
695 fs->s_es->s_state |= EXT2_VALID_FS; /* was fs_clean = 1 */
696 ext2_sbupdate(ump, MNT_WAIT);
697 }
698 /* release buffers containing group descriptors */
699 for(i = 0; i < fs->s_db_per_group; i++)
700 brelse(fs->s_group_desc[i]);
701 /* release cached inode/block bitmaps */
702 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
703 if (fs->s_inode_bitmap[i])
704 brelse (fs->s_inode_bitmap[i]);
705 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
706 if (fs->s_block_bitmap[i])
707 brelse (fs->s_block_bitmap[i]);
708
709 ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
710 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE,
711 NOCRED, p);
712 vrele(ump->um_devvp);
713 bsd_free(fs->s_es, M_UFSMNT);
714 bsd_free(fs, M_UFSMNT);
715 bsd_free(ump, M_UFSMNT);
716 mp->mnt_data = (qaddr_t)0;
717 mp->mnt_flag &= ~MNT_LOCAL;
718 return (error);
719}
720
721/*
722 * Flush out all the files in a filesystem.
723 */
724static int
725ext2_flushfiles(mp, flags, p)
726 register struct mount *mp;
727 int flags;
728 struct proc *p;
729{
730 register struct ufsmount *ump;
731 int error;
732#if QUOTA
733 int i;
734#endif
735
736 ump = VFSTOUFS(mp);
737#if QUOTA
738 if (mp->mnt_flag & MNT_QUOTA) {
739 if (error = vflush(mp, NULLVP, SKIPSYSTEM|flags))
740 return (error);
741 for (i = 0; i < MAXQUOTAS; i++) {
742 if (ump->um_quotas[i] == NULLVP)
743 continue;
744 quotaoff(p, mp, i);
745 }
746 /*
747 * Here we fall through to vflush again to ensure
748 * that we have gotten rid of all the system vnodes.
749 */
750 }
751#endif
752 error = vflush(mp, NULLVP, flags);
753 return (error);
754}
755
756/*
757 * Get file system statistics.
758 * taken from ext2/super.c ext2_statfs
759 */
760static int
761ext2_statfs(mp, sbp, p)
762 struct mount *mp;
763 register struct statfs *sbp;
764 struct proc *p;
765{
766 unsigned long overhead;
767 unsigned long overhead_per_group;
768
769 register struct ufsmount *ump;
770 register struct ext2_sb_info *fs;
771 register struct ext2_super_block *es;
772
773 ump = VFSTOUFS(mp);
774 fs = ump->um_e2fs;
775 es = fs->s_es;
776
777 if (es->s_magic != EXT2_SUPER_MAGIC)
778 panic("ext2_statfs - magic number spoiled");
779
780 /*
781 * Compute the overhead (FS structures)
782 */
783 overhead_per_group = 1 /* super block */ +
784 fs->s_db_per_group +
785 1 /* block bitmap */ +
786 1 /* inode bitmap */ +
787 fs->s_itb_per_group;
788 overhead = es->s_first_data_block +
789 fs->s_groups_count * overhead_per_group;
790
791 sbp->f_type = MOUNT_EXT2FS;
792 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
793 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
794 sbp->f_blocks = es->s_blocks_count - overhead;
795 sbp->f_bfree = es->s_free_blocks_count;
796 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
797 sbp->f_files = es->s_inodes_count;
798 sbp->f_ffree = es->s_free_inodes_count;
799 if (sbp != &mp->mnt_stat) {
800 bcopy((caddr_t)mp->mnt_stat.f_mntonname,
801 (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
802 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
803 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
804 }
805 return (0);
806}
807
808/*
809 * Go through the disk queues to initiate sandbagged IO;
810 * go through the inodes to write those that have been modified;
811 * initiate the writing of the super block if it has been modified.
812 *
813 * Note: we are always called with the filesystem marked `MPBUSY'.
814 */
815static int
816ext2_sync(mp, waitfor, cred, p)
817 struct mount *mp;
818 int waitfor;
819 struct ucred *cred;
820 struct proc *p;
821{
822 register struct vnode *vp;
823 register struct inode *ip;
824 register struct ufsmount *ump = VFSTOUFS(mp);
825 register struct ext2_sb_info *fs;
826 int error, allerror = 0;
827
828 fs = ump->um_e2fs;
829 /*
830 * Write back modified superblock.
831 * Consistency check that the superblock
832 * is still in the buffer cache.
833 */
834 if (fs->s_dirt) {
835#if !defined(__FreeBSD__)
836 struct timeval time;
837#endif
838
839 if (fs->s_rd_only != 0) { /* XXX */
840 printf("fs = %s\n", fs->fs_fsmnt);
841 panic("update: rofs mod");
842 }
843 fs->s_dirt = 0;
844#if !defined(__FreeBSD__)
845 get_time(&time);
846#endif
847 fs->s_es->s_wtime = time.tv_sec;
848 allerror = ext2_sbupdate(ump, waitfor);
849 }
850 /*
851 * Write back each (modified) inode.
852 */
853loop:
854 for (vp = mp->mnt_vnodelist.lh_first;
855 vp != NULL;
856 vp = vp->v_mntvnodes.le_next) {
857 /*
858 * If the vnode that we are about to sync is no longer
859 * associated with this mount point, start over.
860 */
861 if (vp->v_mount != mp)
862 goto loop;
863 if (VOP_ISLOCKED(vp))
864 continue;
865 ip = VTOI(vp);
866 if ((ip->i_flag &
867 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
868 vp->v_dirtyblkhd.lh_first == NULL)
869 continue;
870 if (vget(vp, LK_EXCLUSIVE, p))
871 goto loop;
872 if (error = VOP_FSYNC(vp, cred, waitfor, p))
873 allerror = error;
874 vput(vp);
875 }
876 /*
877 * Force stale file system control information to be flushed.
878 */
879 if (error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p))
880 allerror = error;
881#if QUOTA
882 qsync(mp);
883#endif
884 return (allerror);
885}
886
887/*
888 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
889 * in from disk. If it is in core, wait for the lock bit to clear, then
890 * return the inode locked. Detection and handling of mount points must be
891 * done by the calling routine.
892 */
893static int
894ext2_vget(mp, ino, vpp)
895 struct mount *mp;
896 ino_t ino;
897 struct vnode **vpp;
898{
899 register struct ext2_sb_info *fs;
900 register struct inode *ip;
901 struct ufsmount *ump;
902 struct buf *bp;
903 struct vnode *vp;
904 dev_t dev;
905 int i, type, error;
906 int used_blocks;
907
908 ump = VFSTOUFS(mp);
909 dev = ump->um_dev;
910restart:
911 if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
912 return (0);
913
914#ifdef __FreeBSD__
915 /*
916 * Lock out the creation of new entries in the FFS hash table in
917 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
918 * may occur!
919 */
920 if (ext2fs_inode_hash_lock) {
921 while (ext2fs_inode_hash_lock) {
922 ext2fs_inode_hash_lock = -1;
923 tsleep(&ext2fs_inode_hash_lock, PVM, "ffsvgt", 0);
924 }
925 goto restart;
926 }
927 ext2fs_inode_hash_lock = 1;
928#endif
929
930 /* Allocate a new vnode/inode. */
931 if (error = getnewvnode(VT_UFS, mp, ext2_vnodeop_p, &vp)) {
932 *vpp = NULL;
933 return (error);
934 }
935 /* I don't really know what this 'type' does. I suppose it's some kind
936 * of memory accounting. Let's just book this memory on FFS's account
937 * If I'm not mistaken, this stuff isn't implemented anyway in Lites
938 */
939 type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */
940 MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK);
941#ifndef __FreeBSD__
942 insmntque(vp, mp);
943#endif
944 bzero((caddr_t)ip, sizeof(struct inode));
945 vp->v_data = ip;
946 ip->i_vnode = vp;
947 ip->i_e2fs = fs = ump->um_e2fs;
948 ip->i_dev = dev;
949 ip->i_number = ino;
950#if QUOTA
951 for (i = 0; i < MAXQUOTAS; i++)
952 ip->i_dquot[i] = NODQUOT;
953#endif
954 /*
955 * Put it onto its hash chain and lock it so that other requests for
956 * this inode will block if they arrive while we are sleeping waiting
957 * for old data structures to be purged or for the contents of the
958 * disk portion of this inode to be read.
959 */
960 ufs_ihashins(ip);
961
962#ifdef __FreeBSD__
963 if (ext2fs_inode_hash_lock < 0)
964 wakeup(&ext2fs_inode_hash_lock);
965 ext2fs_inode_hash_lock = 0;
966#endif
967
968 /* Read in the disk contents for the inode, copy into the inode. */
969 /* Read in the disk contents for the inode, copy into the inode. */
970#if 0
971printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
972#endif
973 if (error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
974 (int)fs->s_blocksize, NOCRED, &bp)) {
975 /*
976 * The inode does not contain anything useful, so it would
977 * be misleading to leave it on its hash chain. With mode
978 * still zero, it will be unlinked and returned to the free
979 * list by vput().
980 */
981 vput(vp);
982 brelse(bp);
983 *vpp = NULL;
984 return (error);
985 }
986 /* convert ext2 inode to dinode */
987 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE *
988 ino_to_fsbo(fs, ino)), &ip->i_din);
989 ip->i_block_group = ino_to_cg(fs, ino);
990 ip->i_next_alloc_block = 0;
991 ip->i_next_alloc_goal = 0;
992 ip->i_prealloc_count = 0;
993 ip->i_prealloc_block = 0;
994 /* now we want to make sure that block pointers for unused
995 blocks are zeroed out - ext2_balloc depends on this
996 although for regular files and directories only
997 */
998 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
999 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1000 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1001 ip->i_db[i] = 0;
1002 }
1003/*
1004 ext2_print_inode(ip);
1005*/
1006 brelse(bp);
1007
1008 /*
1009 * Initialize the vnode from the inode, check for aliases.
1010 * Note that the underlying vnode may have changed.
1011 */
1012 if (error = ufs_vinit(mp, ext2_specop_p, ext2_fifoop_p, &vp)) {
1013 vput(vp);
1014 *vpp = NULL;
1015 return (error);
1016 }
1017 /*
1018 * Finish inode initialization now that aliasing has been resolved.
1019 */
1020 ip->i_devvp = ump->um_devvp;
1021 VREF(ip->i_devvp);
1022 /*
1023 * Set up a generation number for this inode if it does not
1024 * already have one. This should only happen on old filesystems.
1025 */
1026 if (ip->i_gen == 0) {
1027#if !defined(__FreeBSD__)
1028 struct timeval time;
1029 get_time(&time);
1030#endif
1031 if (++nextgennumber < (u_long)time.tv_sec)
1032 nextgennumber = time.tv_sec;
1033 ip->i_gen = nextgennumber;
1034 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1035 ip->i_flag |= IN_MODIFIED;
1036 }
1037 *vpp = vp;
1038 return (0);
1039}
1040
1041/*
1042 * File handle to vnode
1043 *
1044 * Have to be really careful about stale file handles:
1045 * - check that the inode number is valid
1046 * - call ext2_vget() to get the locked inode
1047 * - check for an unallocated inode (i_mode == 0)
1048 * - check that the given client host has export rights and return
1049 * those rights via. exflagsp and credanonp
1050 */
1051static int
1052ext2_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
1053 register struct mount *mp;
1054 struct fid *fhp;
1055 struct mbuf *nam;
1056 struct vnode **vpp;
1057 int *exflagsp;
1058 struct ucred **credanonp;
1059{
1060 register struct ufid *ufhp;
1061 struct ext2_sb_info *fs;
1062
1063 ufhp = (struct ufid *)fhp;
1064 fs = VFSTOUFS(mp)->um_e2fs;
1065 if (ufhp->ufid_ino < ROOTINO ||
1066 ufhp->ufid_ino >= fs->s_groups_count * fs->s_es->s_inodes_per_group)
1067 return (ESTALE);
1068 return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
1069}
1070
1071/*
1072 * Vnode pointer to File handle
1073 */
1074/* ARGSUSED */
1075static int
1076ext2_vptofh(vp, fhp)
1077 struct vnode *vp;
1078 struct fid *fhp;
1079{
1080 register struct inode *ip;
1081 register struct ufid *ufhp;
1082
1083 ip = VTOI(vp);
1084 ufhp = (struct ufid *)fhp;
1085 ufhp->ufid_len = sizeof(struct ufid);
1086 ufhp->ufid_ino = ip->i_number;
1087 ufhp->ufid_gen = ip->i_gen;
1088 return (0);
1089}
1090
1091/*
1092 * Write a superblock and associated information back to disk.
1093 */
1094static int
1095ext2_sbupdate(mp, waitfor)
1096 struct ufsmount *mp;
1097 int waitfor;
1098{
1099 register struct ext2_sb_info *fs = mp->um_e2fs;
1100 register struct ext2_super_block *es = fs->s_es;
1101 register struct buf *bp;
1102 int i, error = 0;
1103/*
1104printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
1105*/
1106 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1107 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1108 if (waitfor == MNT_WAIT)
1109 error = bwrite(bp);
1110 else
1111 bawrite(bp);
1112
1113 /* write group descriptors back on disk */
1114 for(i = 0; i < fs->s_db_per_group; i++)
1115 /* Godmar thinks: we must avoid using any of the b*write
1116 * functions here: we want to keep the buffer locked
1117 * so we use my 'housemade' write routine:
1118 */
1119 error |= ll_w_block(fs->s_group_desc[i], waitfor == MNT_WAIT);
1120
1121 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
1122 if (fs->s_inode_bitmap[i])
1123 ll_w_block (fs->s_inode_bitmap[i], 1);
1124 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
1125 if (fs->s_block_bitmap[i])
1126 ll_w_block (fs->s_block_bitmap[i], 1);
1127
1128 return (error);
1129}
|