1/*-
2 * Copyright 2000 Hans Reiser
3 * See README for licensing and copyright details
4 *
5 * Ported to FreeBSD by Jean-S��bastien P��dron <jspedron@club-internet.fr>
6 *
7 * $FreeBSD$
8 */
9
10#include <gnu/fs/reiserfs/reiserfs_fs.h>
11
12const char reiserfs_3_5_magic_string[] = REISERFS_SUPER_MAGIC_STRING;
13const char reiserfs_3_6_magic_string[] = REISER2FS_SUPER_MAGIC_STRING;
14const char reiserfs_jr_magic_string[]  = REISER2FS_JR_SUPER_MAGIC_STRING;
15
16/*
17 * Default recommended I/O size is 128k. There might be broken
18 * applications that are confused by this. Use nolargeio mount option to
19 * get usual i/o size = PAGE_SIZE.
20 */
21int reiserfs_default_io_size = 128 * 1024;
22
23static vfs_cmount_t	reiserfs_cmount;
24static vfs_fhtovp_t	reiserfs_fhtovp;
25static vfs_mount_t	reiserfs_mount;
26static vfs_root_t	reiserfs_root;
27static vfs_statfs_t	reiserfs_statfs;
28static vfs_unmount_t	reiserfs_unmount;
29
30static int	reiserfs_mountfs(struct vnode *devvp, struct mount *mp,
31		    struct thread *td);
32static void	load_bitmap_info_data(struct reiserfs_sb_info *sbi,
33		    struct reiserfs_bitmap_info *bi);
34static int	read_bitmaps(struct reiserfs_mount *rmp);
35static int	read_old_bitmaps(struct reiserfs_mount *rmp);
36static int	read_super_block(struct reiserfs_mount *rmp, int offset);
37static hashf_t	hash_function(struct reiserfs_mount *rmp);
38
39static int	get_root_node(struct reiserfs_mount *rmp,
40		    struct reiserfs_node **root);
41uint32_t	find_hash_out(struct reiserfs_mount *rmp);
42
43MALLOC_DEFINE(M_REISERFSMNT, "reiserfs_mount", "ReiserFS mount structure");
44MALLOC_DEFINE(M_REISERFSPATH, "reiserfs_path", "ReiserFS path structure");
45MALLOC_DEFINE(M_REISERFSNODE, "reiserfs_node", "ReiserFS vnode private part");
46
47/* -------------------------------------------------------------------
48 * VFS operations
49 * -------------------------------------------------------------------*/
50
51static int
52reiserfs_cmount(struct mntarg *ma, void *data, uint64_t flags)
53{
54	struct reiserfs_args args;
55	struct export_args exp;
56	int error;
57
58	error = copyin(data, &args, sizeof(args));
59	if (error)
60		return (error);
61	vfs_oexport_conv(&args.export, &exp);
62
63	ma = mount_argsu(ma, "from", args.fspec, MAXPATHLEN);
64	ma = mount_arg(ma, "export", &exp, sizeof(exp));
65
66	error = kernel_mount(ma, flags);
67
68	return (error);
69}
70
71/*
72 * Mount system call
73 */
74static int
75reiserfs_mount(struct mount *mp)
76{
77	size_t size;
78	int error, len;
79	accmode_t accmode;
80	char *path, *fspec;
81	struct vnode *devvp;
82	struct vfsoptlist *opts;
83	struct reiserfs_mount *rmp;
84	struct reiserfs_sb_info *sbi;
85	struct nameidata nd, *ndp = &nd;
86	struct thread *td;
87
88	td = curthread;
89	if (!(mp->mnt_flag & MNT_RDONLY))
90		return EROFS;
91
92	/* Get the new options passed to mount */
93	opts = mp->mnt_optnew;
94
95	/* `fspath' contains the mount point (eg. /mnt/linux); REQUIRED */
96	vfs_getopt(opts, "fspath", (void **)&path, NULL);
97	reiserfs_log(LOG_INFO, "mount point is `%s'\n", path);
98
99	/* `from' contains the device name (eg. /dev/ad0s1); REQUIRED */
100	fspec = NULL;
101	error = vfs_getopt(opts, "from", (void **)&fspec, &len);
102	if (!error && fspec[len - 1] != '\0')
103		return (EINVAL);
104	reiserfs_log(LOG_INFO, "device is `%s'\n", fspec);
105
106	/* Handle MNT_UPDATE (mp->mnt_flag) */
107	if (mp->mnt_flag & MNT_UPDATE) {
108		/* For now, only NFS export is supported. */
109		if (vfs_flagopt(opts, "export", NULL, 0))
110			return (0);
111	}
112
113	/* Not an update, or updating the name: look up the name
114	 * and verify that it refers to a sensible disk device. */
115	if (fspec == NULL)
116		return (EINVAL);
117
118	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspec, td);
119	if ((error = namei(ndp)) != 0)
120		return (error);
121	NDFREE(ndp, NDF_ONLY_PNBUF);
122	devvp = ndp->ni_vp;
123
124	if (!vn_isdisk(devvp, &error)) {
125		vput(devvp);
126		return (error);
127	}
128
129	/* If mount by non-root, then verify that user has necessary
130	 * permissions on the device. */
131	accmode = VREAD;
132	if ((mp->mnt_flag & MNT_RDONLY) == 0)
133		accmode |= VWRITE;
134	error = VOP_ACCESS(devvp, accmode, td->td_ucred, td);
135	if (error)
136		error = priv_check(td, PRIV_VFS_MOUNT_PERM);
137	if (error) {
138		vput(devvp);
139		return (error);
140	}
141
142	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
143		error = reiserfs_mountfs(devvp, mp, td);
144	} else {
145		/* TODO Handle MNT_UPDATE */
146		vput(devvp);
147		return (EOPNOTSUPP);
148	}
149
150	if (error) {
151		vrele(devvp);
152		return (error);
153	}
154
155	rmp = VFSTOREISERFS(mp);
156	sbi = rmp->rm_reiserfs;
157
158	/*
159	 * Note that this strncpy() is ok because of a check at the start
160	 * of reiserfs_mount().
161	 */
162	reiserfs_log(LOG_DEBUG, "prepare statfs data\n");
163	(void)copystr(fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size);
164	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
165	(void)reiserfs_statfs(mp, &mp->mnt_stat);
166
167	reiserfs_log(LOG_DEBUG, "done\n");
168	return (0);
169}
170
171/*
172 * Unmount system call
173 */
174static int
175reiserfs_unmount(struct mount *mp, int mntflags)
176{
177	int error, flags = 0;
178	struct reiserfs_mount *rmp;
179	struct reiserfs_sb_info *sbi;
180
181	reiserfs_log(LOG_DEBUG, "get private data\n");
182	rmp = VFSTOREISERFS(mp);
183	sbi = rmp->rm_reiserfs;
184
185	/* Flangs handling */
186	reiserfs_log(LOG_DEBUG, "handle mntflags\n");
187	if (mntflags & MNT_FORCE)
188		flags |= FORCECLOSE;
189
190	/* Flush files -> vflush */
191	reiserfs_log(LOG_DEBUG, "flush vnodes\n");
192	if ((error = vflush(mp, 0, flags, curthread)))
193		return (error);
194
195	/* XXX Super block update */
196
197	if (sbi) {
198		if (SB_AP_BITMAP(sbi)) {
199			int i;
200			reiserfs_log(LOG_DEBUG,
201			    "release bitmap buffers (total: %d)\n",
202			    SB_BMAP_NR(sbi));
203			for (i = 0; i < SB_BMAP_NR(sbi); i++) {
204				if (SB_AP_BITMAP(sbi)[i].bp_data) {
205					free(SB_AP_BITMAP(sbi)[i].bp_data,
206					    M_REISERFSMNT);
207					SB_AP_BITMAP(sbi)[i].bp_data = NULL;
208				}
209			}
210
211			reiserfs_log(LOG_DEBUG, "free bitmaps structure\n");
212			free(SB_AP_BITMAP(sbi), M_REISERFSMNT);
213			SB_AP_BITMAP(sbi) = NULL;
214		}
215
216		if (sbi->s_rs) {
217			reiserfs_log(LOG_DEBUG, "free super block data\n");
218			free(sbi->s_rs, M_REISERFSMNT);
219			sbi->s_rs = NULL;
220		}
221	}
222
223	reiserfs_log(LOG_DEBUG, "close device\n");
224#if defined(si_mountpoint)
225	rmp->rm_devvp->v_rdev->si_mountpoint = NULL;
226#endif
227
228	DROP_GIANT();
229	g_topology_lock();
230	g_vfs_close(rmp->rm_cp);
231	g_topology_unlock();
232	PICKUP_GIANT();
233	vrele(rmp->rm_devvp);
234	dev_rel(rmp->rm_dev);
235
236	if (sbi) {
237		reiserfs_log(LOG_DEBUG, "free sbi\n");
238		free(sbi, M_REISERFSMNT);
239		sbi = rmp->rm_reiserfs = NULL;
240	}
241	if (rmp) {
242		reiserfs_log(LOG_DEBUG, "free rmp\n");
243		free(rmp, M_REISERFSMNT);
244		rmp = NULL;
245	}
246
247	mp->mnt_data  = 0;
248	MNT_ILOCK(mp);
249	mp->mnt_flag &= ~MNT_LOCAL;
250	MNT_IUNLOCK(mp);
251
252	reiserfs_log(LOG_DEBUG, "done\n");
253	return (error);
254}
255
256/*
257 * Return the root of a filesystem.
258 */
259static int
260reiserfs_root(struct mount *mp, int flags, struct vnode **vpp)
261{
262	int error;
263	struct vnode *vp;
264	struct cpu_key rootkey;
265
266	rootkey.on_disk_key.k_dir_id = REISERFS_ROOT_PARENT_OBJECTID;
267	rootkey.on_disk_key.k_objectid = REISERFS_ROOT_OBJECTID;
268
269	error = reiserfs_iget(mp, &rootkey, &vp, curthread);
270
271	if (error == 0)
272		*vpp = vp;
273	return (error);
274}
275
276/*
277 * The statfs syscall
278 */
279static int
280reiserfs_statfs(struct mount *mp, struct statfs *sbp)
281{
282	struct reiserfs_mount *rmp;
283	struct reiserfs_sb_info *sbi;
284	struct reiserfs_super_block *rs;
285
286	reiserfs_log(LOG_DEBUG, "get private data\n");
287	rmp = VFSTOREISERFS(mp);
288	sbi = rmp->rm_reiserfs;
289	rs  = sbi->s_rs;
290
291	reiserfs_log(LOG_DEBUG, "fill statfs structure\n");
292	sbp->f_bsize  = sbi->s_blocksize;
293	sbp->f_iosize = sbp->f_bsize;
294	sbp->f_blocks = sb_block_count(rs) - sb_bmap_nr(rs) - 1;
295	sbp->f_bfree  = sb_free_blocks(rs);
296	sbp->f_bavail = sbp->f_bfree;
297	sbp->f_files  = 0;
298	sbp->f_ffree  = 0;
299	reiserfs_log(LOG_DEBUG, "  block size   = %ju\n",
300	    (intmax_t)sbp->f_bsize);
301	reiserfs_log(LOG_DEBUG, "  IO size      = %ju\n",
302	    (intmax_t)sbp->f_iosize);
303	reiserfs_log(LOG_DEBUG, "  block count  = %ju\n",
304	    (intmax_t)sbp->f_blocks);
305	reiserfs_log(LOG_DEBUG, "  free blocks  = %ju\n",
306	    (intmax_t)sbp->f_bfree);
307	reiserfs_log(LOG_DEBUG, "  avail blocks = %ju\n",
308	    (intmax_t)sbp->f_bavail);
309	reiserfs_log(LOG_DEBUG, "...done\n");
310
311	if (sbp != &mp->mnt_stat) {
312		reiserfs_log(LOG_DEBUG, "copying monut point info\n");
313		sbp->f_type = mp->mnt_vfc->vfc_typenum;
314		bcopy((caddr_t)mp->mnt_stat.f_mntonname,
315		    (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
316		bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
317		    (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
318		reiserfs_log(LOG_DEBUG, "  mount from: %s\n",
319		    sbp->f_mntfromname);
320		reiserfs_log(LOG_DEBUG, "  mount on:   %s\n",
321		    sbp->f_mntonname);
322		reiserfs_log(LOG_DEBUG, "...done\n");
323	}
324
325	return (0);
326}
327
328/*
329 * File handle to vnode
330 *
331 * Have to be really careful about stale file handles:
332 * - check that the inode key is valid
333 * - call ffs_vget() to get the locked inode
334 * - check for an unallocated inode (i_mode == 0)
335 * - check that the given client host has export rights and return
336 *   those rights via. exflagsp and credanonp
337 */
338static int
339reiserfs_fhtovp(struct mount *mp, struct fid *fhp, int flags,
340    struct vnode **vpp)
341{
342	int error;
343	struct rfid *rfhp;
344	struct vnode *nvp;
345	struct cpu_key key;
346	struct reiserfs_node *ip;
347	struct reiserfs_sb_info *sbi;
348	struct thread *td = curthread;
349
350	rfhp = (struct rfid *)fhp;
351	sbi  = VFSTOREISERFS(mp)->rm_reiserfs;
352
353	/* Check that the key is valid */
354	if (rfhp->rfid_dirid < REISERFS_ROOT_PARENT_OBJECTID &&
355	    rfhp->rfid_objectid < REISERFS_ROOT_OBJECTID)
356		return (ESTALE);
357
358	reiserfs_log(LOG_DEBUG,
359	    "file handle key is (dirid=%d, objectid=%d)\n",
360	    rfhp->rfid_dirid, rfhp->rfid_objectid);
361	key.on_disk_key.k_dir_id   = rfhp->rfid_dirid;
362	key.on_disk_key.k_objectid = rfhp->rfid_objectid;
363
364	reiserfs_log(LOG_DEBUG, "read this inode\n");
365	error = reiserfs_iget(mp, &key, &nvp, td);
366	if (error) {
367		*vpp = NULLVP;
368		return (error);
369	}
370
371	reiserfs_log(LOG_DEBUG, "check validity\n");
372	ip = VTOI(nvp);
373	if (ip->i_mode == 0 || ip->i_generation != rfhp->rfid_gen) {
374		vput(nvp);
375		*vpp = NULLVP;
376		return (ESTALE);
377	}
378
379	reiserfs_log(LOG_DEBUG, "return it\n");
380	*vpp = nvp;
381	return (0);
382}
383
384/* -------------------------------------------------------------------
385 * Functions for the journal
386 * -------------------------------------------------------------------*/
387
388int
389is_reiserfs_3_5(struct reiserfs_super_block *rs)
390{
391
392	return (!strncmp(rs->s_v1.s_magic, reiserfs_3_5_magic_string,
393	    strlen(reiserfs_3_5_magic_string)));
394}
395
396int
397is_reiserfs_3_6(struct reiserfs_super_block *rs)
398{
399
400	return (!strncmp(rs->s_v1.s_magic, reiserfs_3_6_magic_string,
401	    strlen(reiserfs_3_6_magic_string)));
402}
403
404int
405is_reiserfs_jr(struct reiserfs_super_block *rs)
406{
407
408	return (!strncmp(rs->s_v1.s_magic, reiserfs_jr_magic_string,
409	    strlen(reiserfs_jr_magic_string)));
410}
411
412static int
413is_any_reiserfs_magic_string(struct reiserfs_super_block *rs)
414{
415
416	return ((is_reiserfs_3_5(rs) || is_reiserfs_3_6(rs) ||
417	    is_reiserfs_jr(rs)));
418}
419
420/* -------------------------------------------------------------------
421 * Internal functions
422 * -------------------------------------------------------------------*/
423
424/*
425 * Common code for mount and mountroot
426 */
427static int
428reiserfs_mountfs(struct vnode *devvp, struct mount *mp, struct thread *td)
429{
430	int error, old_format = 0;
431	struct reiserfs_mount *rmp;
432	struct reiserfs_sb_info *sbi;
433	struct reiserfs_super_block *rs;
434	struct cdev *dev;
435
436	struct g_consumer *cp;
437	struct bufobj *bo;
438
439	//ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
440
441	dev = devvp->v_rdev;
442	dev_ref(dev);
443	DROP_GIANT();
444	g_topology_lock();
445	error = g_vfs_open(devvp, &cp, "reiserfs", /* read-only */ 0);
446	g_topology_unlock();
447	PICKUP_GIANT();
448	VOP_UNLOCK(devvp, 0);
449	if (error) {
450		dev_rel(dev);
451		return (error);
452	}
453
454	bo = &devvp->v_bufobj;
455	bo->bo_private = cp;
456	bo->bo_ops = g_vfs_bufops;
457
458	if (devvp->v_rdev->si_iosize_max != 0)
459		mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
460	if (mp->mnt_iosize_max > MAXPHYS)
461		mp->mnt_iosize_max = MAXPHYS;
462
463	rmp = NULL;
464	sbi = NULL;
465
466	/* rmp contains any information about this specific mount */
467	rmp = malloc(sizeof *rmp, M_REISERFSMNT, M_WAITOK | M_ZERO);
468	if (!rmp) {
469		error = (ENOMEM);
470		goto out;
471	}
472	sbi = malloc(sizeof *sbi, M_REISERFSMNT, M_WAITOK | M_ZERO);
473	if (!sbi) {
474		error = (ENOMEM);
475		goto out;
476	}
477	rmp->rm_reiserfs = sbi;
478	rmp->rm_mountp   = mp;
479	rmp->rm_devvp    = devvp;
480	rmp->rm_dev      = dev;
481	rmp->rm_bo       = &devvp->v_bufobj;
482	rmp->rm_cp       = cp;
483
484	/* Set default values for options: non-aggressive tails */
485	REISERFS_SB(sbi)->s_mount_opt = (1 << REISERFS_SMALLTAIL);
486	REISERFS_SB(sbi)->s_rd_only   = 1;
487	REISERFS_SB(sbi)->s_devvp     = devvp;
488
489	/* Read the super block */
490	if ((error = read_super_block(rmp, REISERFS_OLD_DISK_OFFSET)) == 0) {
491		/* The read process succeeded, it's an old format */
492		old_format = 1;
493	} else if ((error = read_super_block(rmp, REISERFS_DISK_OFFSET)) != 0) {
494		reiserfs_log(LOG_ERR, "can not find a ReiserFS filesystem\n");
495		goto out;
496	}
497
498	rs = SB_DISK_SUPER_BLOCK(sbi);
499
500	/*
501	 * Let's do basic sanity check to verify that underlying device is
502	 * not smaller than the filesystem. If the check fails then abort and
503	 * scream, because bad stuff will happen otherwise.
504	 */
505#if 0
506	if (s->s_bdev && s->s_bdev->bd_inode &&
507	    i_size_read(s->s_bdev->bd_inode) <
508	    sb_block_count(rs) * sb_blocksize(rs)) {
509		reiserfs_log(LOG_ERR,
510		    "reiserfs: filesystem cannot be mounted because it is "
511		    "bigger than the device.\n");
512		reiserfs_log(LOG_ERR, "reiserfs: you may need to run fsck "
513		    "rr may be you forgot to reboot after fdisk when it "
514		    "told you to.\n");
515		goto out;
516	}
517#endif
518
519	/*
520	 * XXX This is from the original Linux code, but why affecting 2 values
521	 * to the same variable?
522	 */
523	sbi->s_mount_state = SB_REISERFS_STATE(sbi);
524	sbi->s_mount_state = REISERFS_VALID_FS;
525
526	if ((error = (old_format ?
527	    read_old_bitmaps(rmp) : read_bitmaps(rmp)))) {
528		reiserfs_log(LOG_ERR, "unable to read bitmap\n");
529		goto out;
530	}
531
532	/* Make data=ordered the default */
533	if (!reiserfs_data_log(sbi) && !reiserfs_data_ordered(sbi) &&
534	    !reiserfs_data_writeback(sbi)) {
535		REISERFS_SB(sbi)->s_mount_opt |= (1 << REISERFS_DATA_ORDERED);
536	}
537
538	if (reiserfs_data_log(sbi)) {
539		reiserfs_log(LOG_INFO, "using journaled data mode\n");
540	} else if (reiserfs_data_ordered(sbi)) {
541		reiserfs_log(LOG_INFO, "using ordered data mode\n");
542	} else {
543		reiserfs_log(LOG_INFO, "using writeback data mode\n");
544	}
545
546	/* TODO Not yet supported */
547#if 0
548	if(journal_init(sbi, jdev_name, old_format, commit_max_age)) {
549		reiserfs_log(LOG_ERR, "unable to initialize journal space\n");
550		goto out;
551	} else {
552		jinit_done = 1 ; /* once this is set, journal_release must
553				    be called if we error out of the mount */
554	}
555
556	if (reread_meta_blocks(sbi)) {
557		reiserfs_log(LOG_ERR,
558		    "unable to reread meta blocks after journal init\n");
559		goto out;
560	}
561#endif
562
563	/* Define and initialize hash function */
564	sbi->s_hash_function = hash_function(rmp);
565
566	if (sbi->s_hash_function == NULL) {
567		reiserfs_log(LOG_ERR, "couldn't determined hash function\n");
568		error = (EINVAL);
569		goto out;
570	}
571
572	if (is_reiserfs_3_5(rs) ||
573	    (is_reiserfs_jr(rs) && SB_VERSION(sbi) == REISERFS_VERSION_1))
574		bit_set(&(sbi->s_properties), REISERFS_3_5);
575	else
576		bit_set(&(sbi->s_properties), REISERFS_3_6);
577
578	mp->mnt_data = rmp;
579	mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
580	mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
581	MNT_ILOCK(mp);
582	mp->mnt_flag |= MNT_LOCAL;
583	MNT_IUNLOCK(mp);
584#if defined(si_mountpoint)
585	devvp->v_rdev->si_mountpoint = mp;
586#endif
587
588	return (0);
589
590out:
591	reiserfs_log(LOG_INFO, "*** error during mount ***\n");
592	if (sbi) {
593		if (SB_AP_BITMAP(sbi)) {
594			int i;
595			for (i = 0; i < SB_BMAP_NR(sbi); i++) {
596				if (!SB_AP_BITMAP(sbi)[i].bp_data)
597					break;
598				free(SB_AP_BITMAP(sbi)[i].bp_data,
599				    M_REISERFSMNT);
600			}
601			free(SB_AP_BITMAP(sbi), M_REISERFSMNT);
602		}
603
604		if (sbi->s_rs) {
605			free(sbi->s_rs, M_REISERFSMNT);
606			sbi->s_rs = NULL;
607		}
608	}
609
610	if (cp != NULL) {
611		DROP_GIANT();
612		g_topology_lock();
613		g_vfs_close(cp);
614		g_topology_unlock();
615		PICKUP_GIANT();
616	}
617
618	if (sbi)
619		free(sbi, M_REISERFSMNT);
620	if (rmp)
621		free(rmp, M_REISERFSMNT);
622	dev_rel(dev);
623	return (error);
624}
625
626/*
627 * Read the super block
628 */
629static int
630read_super_block(struct reiserfs_mount *rmp, int offset)
631{
632	struct buf *bp;
633	int error, bits;
634	struct reiserfs_super_block *rs;
635	struct reiserfs_sb_info *sbi;
636	uint16_t fs_blocksize;
637
638	if (offset == REISERFS_OLD_DISK_OFFSET) {
639		reiserfs_log(LOG_DEBUG,
640		    "reiserfs/super: read old format super block\n");
641	} else {
642		reiserfs_log(LOG_DEBUG,
643		    "reiserfs/super: read new format super block\n");
644	}
645
646	/* Read the super block */
647	if ((error = bread(rmp->rm_devvp, offset * btodb(REISERFS_BSIZE),
648	    REISERFS_BSIZE, NOCRED, &bp)) != 0) {
649		reiserfs_log(LOG_ERR, "can't read device\n");
650		return (error);
651	}
652
653	/* Get it from the buffer data */
654	rs = (struct reiserfs_super_block *)bp->b_data;
655	if (!is_any_reiserfs_magic_string(rs)) {
656		brelse(bp);
657		return (EINVAL);
658	}
659
660	fs_blocksize = sb_blocksize(rs);
661	brelse(bp);
662	bp = NULL;
663
664	if (fs_blocksize <= 0) {
665		reiserfs_log(LOG_ERR, "unexpected null block size");
666		return (EINVAL);
667	}
668
669	/* Read the super block (for double check)
670	 * We can't read the same blkno with a different size: it causes
671	 * panic() if INVARIANTS is set. So we keep REISERFS_BSIZE */
672	if ((error = bread(rmp->rm_devvp,
673	    offset * REISERFS_BSIZE / fs_blocksize * btodb(fs_blocksize),
674	    REISERFS_BSIZE, NOCRED, &bp)) != 0) {
675		reiserfs_log(LOG_ERR, "can't reread the super block\n");
676		return (error);
677	}
678
679	rs = (struct reiserfs_super_block *)bp->b_data;
680	if (sb_blocksize(rs) != fs_blocksize) {
681		reiserfs_log(LOG_ERR, "unexpected block size "
682		    "(found=%u, expected=%u)\n",
683		    sb_blocksize(rs), fs_blocksize);
684		brelse(bp);
685		return (EINVAL);
686	}
687
688	reiserfs_log(LOG_DEBUG, "magic: `%s'\n", rs->s_v1.s_magic);
689	reiserfs_log(LOG_DEBUG, "label: `%s'\n", rs->s_label);
690	reiserfs_log(LOG_DEBUG, "block size:     %6d\n", sb_blocksize(rs));
691	reiserfs_log(LOG_DEBUG, "block count:    %6u\n",
692	    rs->s_v1.s_block_count);
693	reiserfs_log(LOG_DEBUG, "bitmaps number: %6u\n",
694	    rs->s_v1.s_bmap_nr);
695
696	if (rs->s_v1.s_root_block == -1) {
697		log(LOG_ERR,
698		    "reiserfs: Unfinished reiserfsck --rebuild-tree run "
699		    "detected. Please\n"
700		    "run reiserfsck --rebuild-tree and wait for a "
701		    "completion. If that\n"
702		    "fails, get newer reiserfsprogs package");
703		brelse(bp);
704		return (EINVAL);
705	}
706
707	sbi = rmp->rm_reiserfs;
708	sbi->s_blocksize = fs_blocksize;
709
710	for (bits = 9, fs_blocksize >>= 9; fs_blocksize >>= 1; bits++)
711		;
712	sbi->s_blocksize_bits = bits;
713
714	/* Copy the buffer and release it */
715	sbi->s_rs = malloc(sizeof *rs, M_REISERFSMNT, M_WAITOK | M_ZERO);
716	if (!sbi->s_rs) {
717		reiserfs_log(LOG_ERR, "can not read the super block\n");
718		brelse(bp);
719		return (ENOMEM);
720	}
721	bcopy(rs, sbi->s_rs, sizeof(struct reiserfs_super_block));
722	brelse(bp);
723
724	if (is_reiserfs_jr(rs)) {
725		if (sb_version(rs) == REISERFS_VERSION_2)
726			reiserfs_log(LOG_INFO, "found reiserfs format \"3.6\""
727			    " with non-standard journal");
728		else if (sb_version(rs) == REISERFS_VERSION_1)
729			reiserfs_log(LOG_INFO, "found reiserfs format \"3.5\""
730			    " with non-standard journal");
731		else {
732			reiserfs_log(LOG_ERR, "found unknown "
733			    "format \"%u\" of reiserfs with non-standard magic",
734			    sb_version(rs));
735			return (EINVAL);
736		}
737	} else {
738		/*
739		 * s_version of standard format may contain incorrect
740		 * information, so we just look at the magic string
741		 */
742		reiserfs_log(LOG_INFO,
743		    "found reiserfs format \"%s\" with standard journal\n",
744		    is_reiserfs_3_5(rs) ? "3.5" : "3.6");
745	}
746
747	return (0);
748}
749
750/*
751 * load_bitmap_info_data - Sets up the reiserfs_bitmap_info structure
752 * from disk.
753 * @sbi - superblock info for this filesystem
754 * @bi  - the bitmap info to be loaded. Requires that bi->bp is valid.
755 *
756 * This routine counts how many free bits there are, finding the first
757 * zero as a side effect. Could also be implemented as a loop of
758 * test_bit() calls, or a loop of find_first_zero_bit() calls. This
759 * implementation is similar to find_first_zero_bit(), but doesn't
760 * return after it finds the first bit. Should only be called on fs
761 * mount, but should be fairly efficient anyways.
762 *
763 * bi->first_zero_hint is considered unset if it == 0, since the bitmap
764 * itself will invariably occupt block 0 represented in the bitmap. The
765 * only exception to this is when free_count also == 0, since there will
766 * be no free blocks at all.
767 */
768static void
769load_bitmap_info_data(struct reiserfs_sb_info *sbi,
770    struct reiserfs_bitmap_info *bi)
771{
772	unsigned long *cur;
773
774	cur = (unsigned long *)bi->bp_data;
775	while ((char *)cur < (bi->bp_data + sbi->s_blocksize)) {
776		/*
777		 * No need to scan if all 0's or all 1's.
778		 * Since we're only counting 0's, we can simply ignore
779		 * all 1's
780		 */
781		if (*cur == 0) {
782			if (bi->first_zero_hint == 0) {
783				bi->first_zero_hint =
784				    ((char *)cur - bi->bp_data) << 3;
785			}
786			bi->free_count += sizeof(unsigned long) * 8;
787		} else if (*cur != ~0L) {
788			int b;
789
790			for (b = 0; b < sizeof(unsigned long) * 8; b++) {
791				if (!reiserfs_test_le_bit(b, cur)) {
792					bi->free_count++;
793					if (bi->first_zero_hint == 0)
794						bi->first_zero_hint =
795						    (((char *)cur -
796						      bi->bp_data) << 3) + b;
797				}
798			}
799		}
800		cur++;
801	}
802}
803
804/*
805 * Read the bitmaps
806 */
807static int
808read_bitmaps(struct reiserfs_mount *rmp)
809{
810	int i, bmap_nr;
811	struct buf *bp = NULL;
812	struct reiserfs_sb_info *sbi = rmp->rm_reiserfs;
813
814	/* Allocate memory for the table of bitmaps */
815	SB_AP_BITMAP(sbi) =
816	    malloc(sizeof(struct reiserfs_bitmap_info) * SB_BMAP_NR(sbi),
817		M_REISERFSMNT, M_WAITOK | M_ZERO);
818	if (!SB_AP_BITMAP(sbi))
819		return (ENOMEM);
820
821	/* Read all the bitmaps */
822	for (i = 0,
823	    bmap_nr = (REISERFS_DISK_OFFSET_IN_BYTES / sbi->s_blocksize + 1) *
824	    btodb(sbi->s_blocksize);
825	    i < SB_BMAP_NR(sbi); i++, bmap_nr = sbi->s_blocksize * 8 * i) {
826		SB_AP_BITMAP(sbi)[i].bp_data = malloc(sbi->s_blocksize,
827		    M_REISERFSMNT, M_WAITOK | M_ZERO);
828		if (!SB_AP_BITMAP(sbi)[i].bp_data)
829			return (ENOMEM);
830		bread(rmp->rm_devvp, bmap_nr, sbi->s_blocksize, NOCRED, &bp);
831		bcopy(bp->b_data, SB_AP_BITMAP(sbi)[i].bp_data,
832		    sbi->s_blocksize);
833		brelse(bp);
834		bp = NULL;
835
836		/*if (!buffer_uptodate(SB_AP_BITMAP(s)[i].bh))
837			ll_rw_block(READ, 1, &SB_AP_BITMAP(s)[i].bh);*/
838	}
839
840	for (i = 0; i < SB_BMAP_NR(sbi); i++) {
841		/*if (!buffer_uptodate(SB_AP_BITMAP(s)[i].bh)) {
842		  reiserfs_warning(s,"sh-2029: reiserfs read_bitmaps: "
843		  "bitmap block (#%lu) reading failed",
844		  SB_AP_BITMAP(s)[i].bh->b_blocknr);
845		  for (i = 0; i < SB_BMAP_NR(s); i++)
846		  brelse(SB_AP_BITMAP(s)[i].bh);
847		  vfree(SB_AP_BITMAP(s));
848		  SB_AP_BITMAP(s) = NULL;
849		  return 1;
850		  }*/
851		load_bitmap_info_data(sbi, SB_AP_BITMAP(sbi) + i);
852		reiserfs_log(LOG_DEBUG,
853		    "%d free blocks (starting at block %ld)\n",
854		    SB_AP_BITMAP(sbi)[i].free_count,
855		    (long)SB_AP_BITMAP(sbi)[i].first_zero_hint);
856	}
857
858	return (0);
859}
860
861// TODO Not supported
862static int
863read_old_bitmaps(struct reiserfs_mount *rmp)
864{
865
866	return (EOPNOTSUPP);
867#if 0
868	int i;
869	struct reiserfs_sb_info *sbi = rmp->rm_reiserfs;
870	struct reiserfs_super_block *rs = SB_DISK_SUPER_BLOCK(sbi);
871
872	/* First of bitmap blocks */
873	int bmp1 = (REISERFS_OLD_DISK_OFFSET / sbi->s_blocksize) *
874	    btodb(sbi->s_blocksize);
875
876	/* Read true bitmap */
877	SB_AP_BITMAP(sbi) =
878	    malloc(sizeof (struct reiserfs_buffer_info *) * sb_bmap_nr(rs),
879		M_REISERFSMNT, M_WAITOK | M_ZERO);
880	if (!SB_AP_BITMAP(sbi))
881		return 1;
882
883	for (i = 0; i < sb_bmap_nr(rs); i ++) {
884		SB_AP_BITMAP(sbi)[i].bp = getblk(rmp->rm_devvp,
885		    (bmp1 + i) * btodb(sbi->s_blocksize), sbi->s_blocksize, 0, 0, 0);
886		if (!SB_AP_BITMAP(sbi)[i].bp)
887			return 1;
888		load_bitmap_info_data(sbi, SB_AP_BITMAP(sbi) + i);
889	}
890
891	return 0;
892#endif
893}
894
895/* -------------------------------------------------------------------
896 * Hash detection stuff
897 * -------------------------------------------------------------------*/
898
899static int
900get_root_node(struct reiserfs_mount *rmp, struct reiserfs_node **root)
901{
902	struct reiserfs_node *ip;
903	struct reiserfs_iget_args args;
904
905	/* Allocate the node structure */
906	reiserfs_log(LOG_DEBUG, "malloc(struct reiserfs_node)\n");
907	ip = malloc(sizeof(struct reiserfs_node),
908	    M_REISERFSNODE, M_WAITOK | M_ZERO);
909
910	/* Fill the structure */
911	reiserfs_log(LOG_DEBUG, "filling *ip\n");
912	ip->i_dev      = rmp->rm_dev;
913	ip->i_number   = REISERFS_ROOT_OBJECTID;
914	ip->i_ino      = REISERFS_ROOT_PARENT_OBJECTID;
915	ip->i_reiserfs = rmp->rm_reiserfs;
916
917	/* Read the inode */
918	args.objectid = ip->i_number;
919	args.dirid    = ip->i_ino;
920	reiserfs_log(LOG_DEBUG, "call reiserfs_read_locked_inode("
921	    "objectid=%d,dirid=%d)\n", args.objectid, args.dirid);
922	reiserfs_read_locked_inode(ip, &args);
923
924	ip->i_devvp = rmp->rm_devvp;
925	//XXX VREF(ip->i_devvp); Is it necessary ?
926
927	*root = ip;
928	return (0);
929}
930
931/*
932 * If root directory is empty - we set default - Yura's - hash and warn
933 * about it.
934 * FIXME: we look for only one name in a directory. If tea and yura both
935 * have the same value - we ask user to send report to the mailing list
936 */
937uint32_t find_hash_out(struct reiserfs_mount *rmp)
938{
939	int retval;
940	struct cpu_key key;
941	INITIALIZE_PATH(path);
942	struct reiserfs_node *ip;
943	struct reiserfs_sb_info *sbi;
944	struct reiserfs_dir_entry de;
945	uint32_t hash = DEFAULT_HASH;
946
947	get_root_node(rmp, &ip);
948	if (!ip)
949		return (UNSET_HASH);
950
951	sbi = rmp->rm_reiserfs;
952
953	do {
954		uint32_t teahash, r5hash, yurahash;
955
956		reiserfs_log(LOG_DEBUG, "make_cpu_key\n");
957		make_cpu_key(&key, ip, ~0, TYPE_DIRENTRY, 3);
958		reiserfs_log(LOG_DEBUG, "search_by_entry_key for "
959		    "key(objectid=%d,dirid=%d)\n",
960		    key.on_disk_key.k_objectid, key.on_disk_key.k_dir_id);
961		retval = search_by_entry_key(sbi, &key, &path, &de);
962		if (retval == IO_ERROR) {
963			pathrelse(&path);
964			return (UNSET_HASH);
965		}
966		if (retval == NAME_NOT_FOUND)
967			de.de_entry_num--;
968
969		reiserfs_log(LOG_DEBUG, "name found\n");
970
971		set_de_name_and_namelen(&de);
972
973		if (deh_offset(&(de.de_deh[de.de_entry_num])) == DOT_DOT_OFFSET) {
974			/* Allow override in this case */
975			if (reiserfs_rupasov_hash(sbi)) {
976				hash = YURA_HASH;
977			}
978			reiserfs_log(LOG_DEBUG,
979			    "FS seems to be empty, autodetect "
980			    "is using the default hash");
981			break;
982		}
983
984		r5hash   = GET_HASH_VALUE(r5_hash(de.de_name, de.de_namelen));
985		teahash  = GET_HASH_VALUE(keyed_hash(de.de_name,
986		    de.de_namelen));
987		yurahash = GET_HASH_VALUE(yura_hash(de.de_name, de.de_namelen));
988		if (((teahash == r5hash) &&
989		    (GET_HASH_VALUE(
990		     deh_offset(&(de.de_deh[de.de_entry_num]))) == r5hash)) ||
991		    ((teahash == yurahash) &&
992		     (yurahash ==
993		      GET_HASH_VALUE(
994		      deh_offset(&(de.de_deh[de.de_entry_num]))))) ||
995		    ((r5hash == yurahash) &&
996		     (yurahash ==
997		      GET_HASH_VALUE(
998		      deh_offset(&(de.de_deh[de.de_entry_num])))))) {
999			reiserfs_log(LOG_ERR,
1000			    "unable to automatically detect hash "
1001			    "function. Please mount with -o "
1002			    "hash={tea,rupasov,r5}");
1003			hash = UNSET_HASH;
1004			break;
1005		}
1006
1007		if (GET_HASH_VALUE(
1008		    deh_offset(&(de.de_deh[de.de_entry_num]))) == yurahash) {
1009			reiserfs_log(LOG_DEBUG, "detected YURA hash\n");
1010			hash = YURA_HASH;
1011		} else if (GET_HASH_VALUE(
1012		    deh_offset(&(de.de_deh[de.de_entry_num]))) == teahash) {
1013			reiserfs_log(LOG_DEBUG, "detected TEA hash\n");
1014			hash = TEA_HASH;
1015		} else if (GET_HASH_VALUE(
1016		    deh_offset(&(de.de_deh[de.de_entry_num]))) == r5hash) {
1017			reiserfs_log(LOG_DEBUG, "detected R5 hash\n");
1018			hash = R5_HASH;
1019		} else {
1020			reiserfs_log(LOG_WARNING, "unrecognised hash function");
1021			hash = UNSET_HASH;
1022		}
1023	} while (0);
1024
1025	pathrelse(&path);
1026	return (hash);
1027}
1028
1029/* Finds out which hash names are sorted with */
1030static int
1031what_hash(struct reiserfs_mount *rmp)
1032{
1033	uint32_t code;
1034	struct reiserfs_sb_info *sbi = rmp->rm_reiserfs;
1035
1036	find_hash_out(rmp);
1037	code = sb_hash_function_code(SB_DISK_SUPER_BLOCK(sbi));
1038
1039	/*
1040	 * reiserfs_hash_detect() == true if any of the hash mount options
1041	 * were used. We must check them to make sure the user isn't using a
1042	 * bad hash value
1043	 */
1044	if (code == UNSET_HASH || reiserfs_hash_detect(sbi))
1045		code = find_hash_out(rmp);
1046
1047	if (code != UNSET_HASH && reiserfs_hash_detect(sbi)) {
1048		/*
1049		 * Detection has found the hash, and we must check against
1050		 * the mount options
1051		 */
1052		if (reiserfs_rupasov_hash(sbi) && code != YURA_HASH) {
1053			reiserfs_log(LOG_ERR, "error, %s hash detected, "
1054			    "unable to force rupasov hash",
1055			    reiserfs_hashname(code));
1056			code = UNSET_HASH;
1057		} else if (reiserfs_tea_hash(sbi) && code != TEA_HASH) {
1058			reiserfs_log(LOG_ERR, "error, %s hash detected, "
1059			    "unable to force tea hash",
1060			    reiserfs_hashname(code));
1061			code = UNSET_HASH;
1062		} else if (reiserfs_r5_hash(sbi) && code != R5_HASH) {
1063			reiserfs_log(LOG_ERR, "error, %s hash detected, "
1064			    "unable to force r5 hash",
1065			    reiserfs_hashname(code));
1066			code = UNSET_HASH;
1067		}
1068	} else {
1069		/*
1070		 * Find_hash_out was not called or could not determine
1071		 * the hash
1072		 */
1073		if (reiserfs_rupasov_hash(sbi)) {
1074			code = YURA_HASH;
1075		} else if (reiserfs_tea_hash(sbi)) {
1076			code = TEA_HASH;
1077		} else if (reiserfs_r5_hash(sbi)) {
1078			code = R5_HASH;
1079		}
1080	}
1081
1082	/* TODO Not supported yet */
1083#if 0
1084	/* If we are mounted RW, and we have a new valid hash code, update
1085	 * the super */
1086	if (code != UNSET_HASH &&
1087	    !(s->s_flags & MS_RDONLY) &&
1088	    code != sb_hash_function_code(SB_DISK_SUPER_BLOCK(s))) {
1089		set_sb_hash_function_code(SB_DISK_SUPER_BLOCK(s), code);
1090	}
1091#endif
1092
1093	return (code);
1094}
1095
1096/* Return pointer to appropriate function */
1097static hashf_t
1098hash_function(struct reiserfs_mount *rmp)
1099{
1100
1101	switch (what_hash(rmp)) {
1102	case TEA_HASH:
1103		reiserfs_log(LOG_INFO, "using tea hash to sort names\n");
1104		return (keyed_hash);
1105	case YURA_HASH:
1106		reiserfs_log(LOG_INFO, "using rupasov hash to sort names\n");
1107		return (yura_hash);
1108	case R5_HASH:
1109		reiserfs_log(LOG_INFO, "using r5 hash to sort names\n");
1110		return (r5_hash);
1111	}
1112
1113	return (NULL);
1114}
1115
1116/* -------------------------------------------------------------------
1117 * VFS registration
1118 * -------------------------------------------------------------------*/
1119
1120static struct vfsops reiser_vfsops = {
1121	.vfs_cmount	= reiserfs_cmount,
1122	.vfs_mount	= reiserfs_mount,
1123	.vfs_unmount	= reiserfs_unmount,
1124	//.vfs_checkexp	= reiserfs_checkexp,
1125	//.vfs_extattrctl = reiserfs_extattrctl,
1126	.vfs_fhtovp	= reiserfs_fhtovp,
1127	//.vfs_quotactl	= reiserfs_quotactl,
1128	.vfs_root	= reiserfs_root,
1129	//.vfs_start	= reiserfs_start,
1130	.vfs_statfs	= reiserfs_statfs,
1131	//.vfs_sync	= reiserfs_sync,
1132	//.vfs_vget	= reiserfs_vget,
1133};
1134
1135VFS_SET(reiser_vfsops, reiserfs, VFCF_READONLY);
1136