xref: /netbsd-src/sys/fs/nilfs/nilfs_vfsops.c (revision 5bbd2a12505d72a8177929a37b5cee489d0a1cfd)
1 /* $NetBSD: nilfs_vfsops.c,v 1.9 2012/03/13 18:40:38 elad Exp $ */
2 
3 /*
4  * Copyright (c) 2008, 2009 Reinoud Zandijk
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
17  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
18  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
19  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
20  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
21  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
22  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
23  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
25  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26  *
27  */
28 
29 #include <sys/cdefs.h>
30 #ifndef lint
31 __KERNEL_RCSID(0, "$NetBSD: nilfs_vfsops.c,v 1.9 2012/03/13 18:40:38 elad Exp $");
32 #endif /* not lint */
33 
34 
35 #if defined(_KERNEL_OPT)
36 #include "opt_compat_netbsd.h"
37 #endif
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/sysctl.h>
42 #include <sys/namei.h>
43 #include <sys/proc.h>
44 #include <sys/kernel.h>
45 #include <sys/vnode.h>
46 #include <miscfs/genfs/genfs.h>
47 #include <miscfs/specfs/specdev.h>
48 #include <sys/mount.h>
49 #include <sys/buf.h>
50 #include <sys/file.h>
51 #include <sys/device.h>
52 #include <sys/disklabel.h>
53 #include <sys/ioctl.h>
54 #include <sys/malloc.h>
55 #include <sys/dirent.h>
56 #include <sys/stat.h>
57 #include <sys/conf.h>
58 #include <sys/kauth.h>
59 #include <sys/module.h>
60 
61 #include <fs/nilfs/nilfs_mount.h>
62 #include <sys/dirhash.h>
63 
64 
65 #include "nilfs.h"
66 #include "nilfs_subr.h"
67 #include "nilfs_bswap.h"
68 
69 MODULE(MODULE_CLASS_VFS, nilfs, NULL);
70 
71 #define VTOI(vnode) ((struct nilfs_node *) vnode->v_data)
72 
73 #define NILFS_SET_SYSTEMFILE(vp) { \
74 	/* XXXAD Is the vnode locked? */	\
75 	(vp)->v_vflag |= VV_SYSTEM;		\
76 	vref(vp);				\
77 	vput(vp); }
78 
79 #define NILFS_UNSET_SYSTEMFILE(vp) { \
80 	/* XXXAD Is the vnode locked? */	\
81 	(vp)->v_vflag &= ~VV_SYSTEM;		\
82 	vrele(vp); }
83 
84 
85 /* verbose levels of the nilfs filingsystem */
86 int nilfs_verbose = NILFS_DEBUGGING;
87 
88 /* malloc regions */
89 MALLOC_JUSTDEFINE(M_NILFSMNT,   "NILFS mount",	"NILFS mount structures");
90 MALLOC_JUSTDEFINE(M_NILFSTEMP,  "NILFS temp",	"NILFS scrap space");
91 struct pool nilfs_node_pool;
92 
93 /* globals */
94 struct _nilfs_devices nilfs_devices;
95 static struct sysctllog *nilfs_sysctl_log;
96 
97 /* supported functions predefined */
98 VFS_PROTOS(nilfs);
99 
100 
101 /* --------------------------------------------------------------------- */
102 
103 /* predefine vnode-op list descriptor */
104 extern const struct vnodeopv_desc nilfs_vnodeop_opv_desc;
105 
106 const struct vnodeopv_desc * const nilfs_vnodeopv_descs[] = {
107 	&nilfs_vnodeop_opv_desc,
108 	NULL,
109 };
110 
111 
112 /* vfsops descriptor linked in as anchor point for the filingsystem */
113 struct vfsops nilfs_vfsops = {
114 	MOUNT_NILFS,			/* vfs_name */
115 	sizeof (struct nilfs_args),
116 	nilfs_mount,
117 	nilfs_start,
118 	nilfs_unmount,
119 	nilfs_root,
120 	(void *)eopnotsupp,		/* vfs_quotactl */
121 	nilfs_statvfs,
122 	nilfs_sync,
123 	nilfs_vget,
124 	nilfs_fhtovp,
125 	nilfs_vptofh,
126 	nilfs_init,
127 	nilfs_reinit,
128 	nilfs_done,
129 	nilfs_mountroot,
130 	nilfs_snapshot,
131 	vfs_stdextattrctl,
132 	(void *)eopnotsupp,		/* vfs_suspendctl */
133 	genfs_renamelock_enter,
134 	genfs_renamelock_exit,
135 	(void *)eopnotsupp,		/* vfs_full_fsync */
136 	nilfs_vnodeopv_descs,
137 	0, /* int vfs_refcount   */
138 	{ NULL, NULL, }, /* LIST_ENTRY(vfsops) */
139 };
140 
141 /* --------------------------------------------------------------------- */
142 
143 /* file system starts here */
144 void
145 nilfs_init(void)
146 {
147 	size_t size;
148 
149 	/* setup memory types */
150 	malloc_type_attach(M_NILFSMNT);
151 	malloc_type_attach(M_NILFSTEMP);
152 
153 	/* init device lists */
154 	SLIST_INIT(&nilfs_devices);
155 
156 	/* init node pools */
157 	size = sizeof(struct nilfs_node);
158 	pool_init(&nilfs_node_pool, size, 0, 0, 0,
159 		"nilfs_node_pool", NULL, IPL_NONE);
160 }
161 
162 
163 void
164 nilfs_reinit(void)
165 {
166 	/* nothing to do */
167 }
168 
169 
170 void
171 nilfs_done(void)
172 {
173 	/* remove pools */
174 	pool_destroy(&nilfs_node_pool);
175 
176 	malloc_type_detach(M_NILFSMNT);
177 	malloc_type_detach(M_NILFSTEMP);
178 }
179 
180 /*
181  * If running a DEBUG kernel, provide an easy way to set the debug flags when
182  * running into a problem.
183  */
184 #define NILFS_VERBOSE_SYSCTLOPT        1
185 
186 static int
187 nilfs_modcmd(modcmd_t cmd, void *arg)
188 {
189 	const struct sysctlnode *node;
190 	int error;
191 
192 	switch (cmd) {
193 	case MODULE_CMD_INIT:
194 		error = vfs_attach(&nilfs_vfsops);
195 		if (error != 0)
196 			break;
197 		/*
198 		 * XXX the "30" below could be dynamic, thereby eliminating one
199 		 * more instance of the "number to vfs" mapping problem, but
200 		 * "30" is the order as taken from sys/mount.h
201 		 */
202 		sysctl_createv(&nilfs_sysctl_log, 0, NULL, NULL,
203 			       CTLFLAG_PERMANENT,
204 			       CTLTYPE_NODE, "vfs", NULL,
205 			       NULL, 0, NULL, 0,
206 			       CTL_VFS, CTL_EOL);
207 		sysctl_createv(&nilfs_sysctl_log, 0, NULL, &node,
208 			       CTLFLAG_PERMANENT,
209 			       CTLTYPE_NODE, "nilfs",
210 			       SYSCTL_DESCR("NTT's NILFSv2"),
211 			       NULL, 0, NULL, 0,
212 			       CTL_VFS, 30, CTL_EOL);
213 #ifdef DEBUG
214 		sysctl_createv(&nilfs_sysctl_log, 0, NULL, &node,
215 			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
216 			       CTLTYPE_INT, "verbose",
217 			       SYSCTL_DESCR("Bitmask for filesystem debugging"),
218 			       NULL, 0, &nilfs_verbose, 0,
219 			       CTL_VFS, 30, NILFS_VERBOSE_SYSCTLOPT, CTL_EOL);
220 #endif
221 		break;
222 	case MODULE_CMD_FINI:
223 		error = vfs_detach(&nilfs_vfsops);
224 		if (error != 0)
225 			break;
226 		sysctl_teardown(&nilfs_sysctl_log);
227 		break;
228 	default:
229 		error = ENOTTY;
230 		break;
231 	}
232 
233 	return (error);
234 }
235 
236 /* --------------------------------------------------------------------- */
237 
238 int
239 nilfs_mountroot(void)
240 {
241 	return EOPNOTSUPP;
242 }
243 
244 /* --------------------------------------------------------------------- */
245 
246 /* system nodes */
247 static int
248 nilfs_create_system_nodes(struct nilfs_device *nilfsdev)
249 {
250 	int error;
251 
252 	error = nilfs_get_node_raw(nilfsdev, NULL, NILFS_DAT_INO,
253 		&nilfsdev->super_root.sr_dat, &nilfsdev->dat_node);
254 	if (error)
255 		goto errorout;
256 
257 	error = nilfs_get_node_raw(nilfsdev, NULL, NILFS_CPFILE_INO,
258 		&nilfsdev->super_root.sr_cpfile, &nilfsdev->cp_node);
259 	if (error)
260 		goto errorout;
261 
262 	error = nilfs_get_node_raw(nilfsdev, NULL, NILFS_SUFILE_INO,
263 		&nilfsdev->super_root.sr_sufile, &nilfsdev->su_node);
264 	if (error)
265 		goto errorout;
266 
267 	NILFS_SET_SYSTEMFILE(nilfsdev->dat_node->vnode);
268 	NILFS_SET_SYSTEMFILE(nilfsdev->cp_node->vnode);
269 	NILFS_SET_SYSTEMFILE(nilfsdev->su_node->vnode);
270 
271 	return 0;
272 errorout:
273 	nilfs_dispose_node(&nilfsdev->dat_node);
274 	nilfs_dispose_node(&nilfsdev->cp_node);
275 	nilfs_dispose_node(&nilfsdev->su_node);
276 
277 	return error;
278 }
279 
280 
281 static void
282 nilfs_release_system_nodes(struct nilfs_device *nilfsdev)
283 {
284 	if (!nilfsdev)
285 		return;
286 	if (nilfsdev->refcnt > 0)
287 		return;
288 
289 	if (nilfsdev->dat_node)
290 		NILFS_UNSET_SYSTEMFILE(nilfsdev->dat_node->vnode);
291 	if (nilfsdev->cp_node)
292 		NILFS_UNSET_SYSTEMFILE(nilfsdev->cp_node->vnode);
293 	if (nilfsdev->su_node)
294 		NILFS_UNSET_SYSTEMFILE(nilfsdev->su_node->vnode);
295 
296 	nilfs_dispose_node(&nilfsdev->dat_node);
297 	nilfs_dispose_node(&nilfsdev->cp_node);
298 	nilfs_dispose_node(&nilfsdev->su_node);
299 }
300 
301 
302 /* --------------------------------------------------------------------- */
303 
304 static int
305 nilfs_check_superblock_crc(struct nilfs_super_block *super)
306 {
307 	uint32_t super_crc, comp_crc;
308 
309 	/* check super block magic */
310 	if (nilfs_rw16(super->s_magic) != NILFS_SUPER_MAGIC)
311 		return 0;
312 
313 	/* preserve crc */
314 	super_crc  = nilfs_rw32(super->s_sum);
315 
316 	/* calculate */
317 	super->s_sum = 0;
318 	comp_crc = crc32_le(nilfs_rw32(super->s_crc_seed),
319 		(uint8_t *) super, nilfs_rw16(super->s_bytes));
320 
321 	/* restore */
322 	super->s_sum = nilfs_rw32(super_crc);
323 
324 	/* check CRC */
325 	return (super_crc == comp_crc);
326 }
327 
328 
329 
330 static int
331 nilfs_read_superblock(struct nilfs_device *nilfsdev)
332 {
333 	struct nilfs_super_block *super, tmp_super;
334 	struct buf *bp;
335 	uint64_t sb1off, sb2off;
336 	uint64_t last_cno1, last_cno2;
337 	uint64_t dev_blk;
338 	int dev_bsize, dev_blks;
339 	int sb1ok, sb2ok, swp;
340 	int error;
341 
342 	sb1off = NILFS_SB_OFFSET_BYTES;
343 	sb2off = NILFS_SB2_OFFSET_BYTES(nilfsdev->devsize);
344 
345 	dev_bsize = 1 << nilfsdev->devvp->v_mount->mnt_fs_bshift;
346 
347 	/* read our superblock regardless of backing device blocksize */
348 	dev_blk   = 0;
349 	dev_blks  = (sb1off + dev_bsize -1)/dev_bsize;
350 	error = bread(nilfsdev->devvp, dev_blk, dev_blks * dev_bsize, NOCRED, 0, &bp);
351 	if (error) {
352 		brelse(bp, BC_AGE);
353 		return error;
354 	}
355 
356 	/* copy read-in super block at the offset */
357 	super = &nilfsdev->super;
358 	memcpy(super, (uint8_t *) bp->b_data + NILFS_SB_OFFSET_BYTES,
359 		sizeof(struct nilfs_super_block));
360 	brelse(bp, BC_AGE);
361 
362 	/* read our 2nd superblock regardless of backing device blocksize */
363 	dev_blk   = sb2off / dev_bsize;
364 	dev_blks  = 2;		/* assumption max one dev_bsize */
365 	error = bread(nilfsdev->devvp, dev_blk, dev_blks * dev_bsize, NOCRED, 0, &bp);
366 	if (error) {
367 		brelse(bp, BC_AGE);
368 		return error;
369 	}
370 
371 	/* copy read-in superblock2 at the offset */
372 	super = &nilfsdev->super2;
373 	memcpy(super, (uint8_t *) bp->b_data + NILFS_SB_OFFSET_BYTES,
374 		sizeof(struct nilfs_super_block));
375 	brelse(bp, BC_AGE);
376 
377 	sb1ok = nilfs_check_superblock_crc(&nilfsdev->super);
378 	sb2ok = nilfs_check_superblock_crc(&nilfsdev->super2);
379 
380 	last_cno1 = nilfs_rw64(nilfsdev->super.s_last_cno);
381 	last_cno2 = nilfs_rw64(nilfsdev->super2.s_last_cno);
382 	swp = sb2ok && (last_cno2 > last_cno1);
383 
384 	if (swp) {
385 		printf("nilfs warning: broken superblock, using spare\n");
386 		tmp_super = nilfsdev->super2;
387 		nilfsdev->super2 = nilfsdev->super;	/* why preserve? */
388 		nilfsdev->super  = tmp_super;
389 	}
390 
391 	if (!sb1ok && !sb2ok) {
392 		printf("nilfs: no valid superblocks found\n");
393 		return EINVAL;
394 	}
395 
396 	return 0;
397 }
398 
399 
400 /* XXX NOTHING from the system nodes should need to be written here */
401 static void
402 nilfs_unmount_base(struct nilfs_device *nilfsdev)
403 {
404 	int error;
405 
406 	if (!nilfsdev)
407 		return;
408 
409 	/* remove all our information */
410 	error = vinvalbuf(nilfsdev->devvp, 0, FSCRED, curlwp, 0, 0);
411 	KASSERT(error == 0);
412 
413 	/* release the device's system nodes */
414 	nilfs_release_system_nodes(nilfsdev);
415 
416 	/* TODO writeout super_block? */
417 }
418 
419 
420 static int
421 nilfs_mount_base(struct nilfs_device *nilfsdev,
422 		struct mount *mp, struct nilfs_args *args)
423 {
424 	struct lwp *l = curlwp;
425 	uint64_t last_pseg, last_cno, last_seq;
426 	uint32_t log_blocksize;
427 	int error;
428 
429 	/* flush out any old buffers remaining from a previous use. */
430 	if ((error = vinvalbuf(nilfsdev->devvp, V_SAVE, l->l_cred, l, 0, 0)))
431 		return error;
432 
433 	/* read in our superblock */
434 	error = nilfs_read_superblock(nilfsdev);
435 	if (error) {
436 		printf("nilfs_mount: can't read in super block : %d\n", error);
437 		return error;
438 	}
439 
440 	/* get our blocksize */
441 	log_blocksize = nilfs_rw32(nilfsdev->super.s_log_block_size);
442 	nilfsdev->blocksize   = (uint64_t) 1 << (log_blocksize + 10);
443 	/* TODO check superblock's blocksize limits */
444 
445 	/* calculate dat structure parameters */
446 	nilfs_calc_mdt_consts(nilfsdev, &nilfsdev->dat_mdt,
447 			nilfs_rw16(nilfsdev->super.s_dat_entry_size));
448 	nilfs_calc_mdt_consts(nilfsdev, &nilfsdev->ifile_mdt,
449 			nilfs_rw16(nilfsdev->super.s_inode_size));
450 
451 	DPRINTF(VOLUMES, ("nilfs_mount: accepted super block\n"));
452 
453 	/* search for the super root and roll forward when needed */
454 	nilfs_search_super_root(nilfsdev);
455 
456 	nilfsdev->mount_state = nilfs_rw16(nilfsdev->super.s_state);
457 	if (nilfsdev->mount_state != NILFS_VALID_FS) {
458 		printf("FS is seriously damaged, needs repairing\n");
459 		printf("aborting mount\n");
460 		return EINVAL;
461 	}
462 
463 	/*
464 	 * FS should be ok now. The superblock and the last segsum could be
465 	 * updated from the repair so extract running values again.
466 	 */
467 	last_pseg = nilfs_rw64(nilfsdev->super.s_last_pseg); /*blknr */
468 	last_cno  = nilfs_rw64(nilfsdev->super.s_last_cno);
469 	last_seq  = nilfs_rw64(nilfsdev->super.s_last_seq);
470 
471 	nilfsdev->last_seg_seq = last_seq;
472 	nilfsdev->last_seg_num = nilfs_get_segnum_of_block(nilfsdev, last_pseg);
473 	nilfsdev->next_seg_num = nilfs_get_segnum_of_block(nilfsdev,
474 		nilfs_rw64(nilfsdev->last_segsum.ss_next));
475 	nilfsdev->last_cno     = last_cno;
476 
477 	DPRINTF(VOLUMES, ("nilfs_mount: accepted super root\n"));
478 
479 	/* create system vnodes for DAT, CP and SEGSUM */
480 	error = nilfs_create_system_nodes(nilfsdev);
481 	if (error)
482 		nilfs_unmount_base(nilfsdev);
483 	return error;
484 }
485 
486 
487 static void
488 nilfs_unmount_device(struct nilfs_device *nilfsdev)
489 {
490 	int error;
491 
492 	/* is there anything? */
493 	if (nilfsdev == NULL)
494 		return;
495 
496 	/* remove the device only if we're the last reference */
497 	nilfsdev->refcnt--;
498 	if (nilfsdev->refcnt >= 1)
499 		return;
500 
501 	/* unmount our base */
502 	nilfs_unmount_base(nilfsdev);
503 
504 	/* remove from our device list */
505 	SLIST_REMOVE(&nilfs_devices, nilfsdev, nilfs_device, next_device);
506 
507 	/* close device */
508 	DPRINTF(VOLUMES, ("closing device\n"));
509 
510 	/* remove our mount reference before closing device */
511 	nilfsdev->devvp->v_specmountpoint = NULL;
512 
513 	/* devvp is still locked by us */
514 	vn_lock(nilfsdev->devvp, LK_EXCLUSIVE | LK_RETRY);
515 	error = VOP_CLOSE(nilfsdev->devvp, FREAD | FWRITE, NOCRED);
516 	if (error)
517 		printf("Error during closure of device! error %d, "
518 		       "device might stay locked\n", error);
519 	DPRINTF(VOLUMES, ("device close ok\n"));
520 
521 	/* clear our mount reference and release device node */
522 	vput(nilfsdev->devvp);
523 
524 	/* free our device info */
525 	free(nilfsdev, M_NILFSMNT);
526 }
527 
528 
529 static int
530 nilfs_check_mounts(struct nilfs_device *nilfsdev, struct mount *mp,
531 	struct nilfs_args *args)
532 {
533 	struct nilfs_mount  *ump;
534 	uint64_t last_cno;
535 
536 	/* no double-mounting of the same checkpoint */
537 	STAILQ_FOREACH(ump, &nilfsdev->mounts, next_mount) {
538 		if (ump->mount_args.cpno == args->cpno)
539 			return EBUSY;
540 	}
541 
542 	/* allow readonly mounts without questioning here */
543 	if (mp->mnt_flag & MNT_RDONLY)
544 		return 0;
545 
546 	/* readwrite mount you want */
547 	STAILQ_FOREACH(ump, &nilfsdev->mounts, next_mount) {
548 		/* only one RW mount on this device! */
549 		if ((ump->vfs_mountp->mnt_flag & MNT_RDONLY)==0)
550 			return EROFS;
551 		/* RDONLY on last mountpoint is device busy */
552 		last_cno = nilfs_rw64(ump->nilfsdev->super.s_last_cno);
553 		if (ump->mount_args.cpno == last_cno)
554 			return EBUSY;
555 	}
556 
557 	/* OK for now */
558 	return 0;
559 }
560 
561 
562 static int
563 nilfs_mount_device(struct vnode *devvp, struct mount *mp, struct nilfs_args *args,
564 	struct nilfs_device **nilfsdev_p)
565 {
566 	uint64_t psize;
567 	unsigned secsize;
568 	struct nilfs_device *nilfsdev;
569 	struct lwp *l = curlwp;
570 	int openflags, accessmode, error;
571 
572 	DPRINTF(VOLUMES, ("Mounting NILFS device\n"));
573 
574 	/* lookup device in our nilfs_mountpoints */
575 	*nilfsdev_p = NULL;
576 	SLIST_FOREACH(nilfsdev, &nilfs_devices, next_device)
577 		if (nilfsdev->devvp == devvp)
578 			break;
579 
580 	if (nilfsdev) {
581 		DPRINTF(VOLUMES, ("device already mounted\n"));
582 		error = nilfs_check_mounts(nilfsdev, mp, args);
583 		if (error)
584 			return error;
585 		nilfsdev->refcnt++;
586 		*nilfsdev_p = nilfsdev;
587 		return 0;
588 	}
589 
590 	DPRINTF(VOLUMES, ("no previous mounts on this device, mounting device\n"));
591 
592 	/* check if its a block device specified */
593 	if (devvp->v_type != VBLK) {
594 		vrele(devvp);
595 		return ENOTBLK;
596 	}
597 	if (bdevsw_lookup(devvp->v_rdev) == NULL) {
598 		vrele(devvp);
599 		return ENXIO;
600 	}
601 
602 	/*
603 	 * If mount by non-root, then verify that user has necessary
604 	 * permissions on the device.
605 	 */
606 	accessmode = VREAD;
607 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
608 		accessmode |= VWRITE;
609 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
610 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
611 	    KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, KAUTH_ARG(accessmode));
612 	VOP_UNLOCK(devvp);
613 	if (error) {
614 		vrele(devvp);
615 		return error;
616 	}
617 
618 	/*
619 	 * Open device read-write; TODO how about upgrading later when needed?
620 	 */
621 	openflags = FREAD | FWRITE;
622 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
623 	error = VOP_OPEN(devvp, openflags, FSCRED);
624 	VOP_UNLOCK(devvp);
625 	if (error) {
626 		vrele(devvp);
627 		return error;
628 	}
629 
630 	/* opened ok, try mounting */
631 	nilfsdev = malloc(sizeof(*nilfsdev), M_NILFSMNT, M_WAITOK | M_ZERO);
632 
633 	/* initialise */
634 	nilfsdev->refcnt        = 1;
635 	nilfsdev->devvp         = devvp;
636 	nilfsdev->uncomitted_bl = 0;
637 	cv_init(&nilfsdev->sync_cv, "nilfssyn");
638 	STAILQ_INIT(&nilfsdev->mounts);
639 
640 	/* register nilfs_device in list */
641 	SLIST_INSERT_HEAD(&nilfs_devices, nilfsdev, next_device);
642 
643 	/* get our device's size */
644 	error = getdisksize(devvp, &psize, &secsize);
645 	if (error) {
646 		/* remove all our information */
647 		nilfs_unmount_device(nilfsdev);
648 		return EINVAL;
649 	}
650 
651 	nilfsdev->devsize = psize * secsize;
652 
653 	/* connect to the head for most recent files XXX really pass mp and args? */
654 	error = nilfs_mount_base(nilfsdev, mp, args);
655 	if (error) {
656 		/* remove all our information */
657 		nilfs_unmount_device(nilfsdev);
658 		return EINVAL;
659 	}
660 
661 	*nilfsdev_p = nilfsdev;
662 	DPRINTF(VOLUMES, ("NILFS device mounted ok\n"));
663 
664 	return 0;
665 }
666 
667 
668 static int
669 nilfs_mount_checkpoint(struct nilfs_mount *ump)
670 {
671 	struct nilfs_cpfile_header *cphdr;
672 	struct nilfs_checkpoint *cp;
673 	struct nilfs_inode  ifile_inode;
674 	struct nilfs_node  *cp_node;
675 	struct buf *bp;
676 	uint64_t ncp, nsn, fcpno, blocknr, last_cno;
677 	uint32_t off, dlen;
678 	int cp_per_block, error;
679 
680 	DPRINTF(VOLUMES, ("mount_nilfs: trying to mount checkpoint number "
681 		"%"PRIu64"\n", ump->mount_args.cpno));
682 
683 	cp_node = ump->nilfsdev->cp_node;
684 
685 	/* get cpfile header from 1st block of cp file */
686 	error = nilfs_bread(cp_node, 0, NOCRED, 0, &bp);
687 	if (error)
688 		return error;
689 	cphdr = (struct nilfs_cpfile_header *) bp->b_data;
690 	ncp = nilfs_rw64(cphdr->ch_ncheckpoints);
691 	nsn = nilfs_rw64(cphdr->ch_nsnapshots);
692 
693 	brelse(bp, BC_AGE);
694 
695 	DPRINTF(VOLUMES, ("mount_nilfs: checkpoint header read in\n"));
696 	DPRINTF(VOLUMES, ("\tNumber of checkpoints %"PRIu64"\n", ncp));
697 	DPRINTF(VOLUMES, ("\tNumber of snapshots   %"PRIu64"\n", nsn));
698 
699 	/* read in our specified checkpoint */
700 	dlen = nilfs_rw16(ump->nilfsdev->super.s_checkpoint_size);
701 	cp_per_block = ump->nilfsdev->blocksize / dlen;
702 
703 	fcpno = ump->mount_args.cpno + NILFS_CPFILE_FIRST_CHECKPOINT_OFFSET -1;
704 	blocknr =  fcpno / cp_per_block;
705 	off     = (fcpno % cp_per_block) * dlen;
706 
707 	error = nilfs_bread(cp_node, blocknr, NOCRED, 0, &bp);
708 	if (error) {
709 		printf("mount_nilfs: couldn't read cp block %"PRIu64"\n",
710 			fcpno);
711 		return EINVAL;
712 	}
713 
714 	/* needs to be a valid checkpoint */
715 	cp = (struct nilfs_checkpoint *) ((uint8_t *) bp->b_data + off);
716 	if (cp->cp_flags & NILFS_CHECKPOINT_INVALID) {
717 		printf("mount_nilfs: checkpoint marked invalid\n");
718 		brelse(bp, BC_AGE);
719 		return EINVAL;
720 	}
721 
722 	/* is this really the checkpoint we want? */
723 	if (nilfs_rw64(cp->cp_cno) != ump->mount_args.cpno) {
724 		printf("mount_nilfs: checkpoint file corrupt? "
725 			"expected cpno %"PRIu64", found cpno %"PRIu64"\n",
726 			ump->mount_args.cpno, nilfs_rw64(cp->cp_cno));
727 		brelse(bp, BC_AGE);
728 		return EINVAL;
729 	}
730 
731 	/* check if its a snapshot ! */
732 	last_cno = nilfs_rw64(ump->nilfsdev->super.s_last_cno);
733 	if (ump->mount_args.cpno != last_cno) {
734 		/* only allow snapshots if not mounting on the last cp */
735 		if ((cp->cp_flags & NILFS_CHECKPOINT_SNAPSHOT) == 0) {
736 			printf( "mount_nilfs: checkpoint %"PRIu64" is not a "
737 				"snapshot\n", ump->mount_args.cpno);
738 			brelse(bp, BC_AGE);
739 			return EINVAL;
740 		}
741 	}
742 
743 	ifile_inode = cp->cp_ifile_inode;
744 	brelse(bp, BC_AGE);
745 
746 	/* get ifile inode */
747 	error = nilfs_get_node_raw(ump->nilfsdev, NULL, NILFS_IFILE_INO,
748 		&ifile_inode, &ump->ifile_node);
749 	if (error) {
750 		printf("mount_nilfs: can't read ifile node\n");
751 		return EINVAL;
752 	}
753 	NILFS_SET_SYSTEMFILE(ump->ifile_node->vnode);
754 
755 	/* get root node? */
756 
757 	return 0;
758 }
759 
760 
761 static int
762 nilfs_stop_writing(struct nilfs_mount *ump)
763 {
764 	/* readonly mounts won't write */
765 	if (ump->vfs_mountp->mnt_flag & MNT_RDONLY)
766 		return 0;
767 
768 	DPRINTF(CALL, ("nilfs_stop_writing called for RW mount\n"));
769 
770 	/* TODO writeout super_block? */
771 	/* XXX no support for writing yet anyway */
772 	return 0;
773 }
774 
775 
776 /* --------------------------------------------------------------------- */
777 
778 
779 
780 #define MPFREE(a, lst) \
781 	if ((a)) free((a), lst);
782 static void
783 free_nilfs_mountinfo(struct mount *mp)
784 {
785 	struct nilfs_mount *ump = VFSTONILFS(mp);
786 
787 	if (ump == NULL)
788 		return;
789 
790 	mutex_destroy(&ump->ihash_lock);
791 	mutex_destroy(&ump->get_node_lock);
792 	MPFREE(ump, M_NILFSMNT);
793 }
794 #undef MPFREE
795 
796 int
797 nilfs_mount(struct mount *mp, const char *path,
798 	  void *data, size_t *data_len)
799 {
800 	struct nilfs_args   *args = data;
801 	struct nilfs_device *nilfsdev;
802 	struct nilfs_mount  *ump;
803 	struct vnode *devvp;
804 	int lst, error;
805 
806 	DPRINTF(VFSCALL, ("nilfs_mount called\n"));
807 
808 	if (*data_len < sizeof *args)
809 		return EINVAL;
810 
811 	if (mp->mnt_flag & MNT_GETARGS) {
812 		/* request for the mount arguments */
813 		ump = VFSTONILFS(mp);
814 		if (ump == NULL)
815 			return EINVAL;
816 		*args = ump->mount_args;
817 		*data_len = sizeof *args;
818 		return 0;
819 	}
820 
821 	/* check/translate struct version */
822 	if (args->version != 1) {
823 		printf("mount_nilfs: unrecognized argument structure version\n");
824 		return EINVAL;
825 	}
826 	/* TODO sanity checking other mount arguments */
827 
828 	/* handle request for updating mount parameters */
829 	if (mp->mnt_flag & MNT_UPDATE) {
830 		/* TODO can't update my mountpoint yet */
831 		return EOPNOTSUPP;
832 	}
833 
834 	/* lookup name to get its vnode */
835 	error = namei_simple_user(args->fspec, NSM_FOLLOW_NOEMULROOT, &devvp);
836 	if (error)
837 		return error;
838 
839 #ifdef DEBUG
840 	if (nilfs_verbose & NILFS_DEBUG_VOLUMES)
841 		vprint("NILFS mount, trying to mount \n", devvp);
842 #endif
843 
844 	error = nilfs_mount_device(devvp, mp, args, &nilfsdev);
845 	if (error)
846 		return error;
847 
848 	/*
849 	 * Create a nilfs_mount on the specified checkpoint. Note that only
850 	 * ONE RW mount point can exist and it needs to have the highest
851 	 * checkpoint nr. If mounting RW and its not on the last checkpoint we
852 	 * need to invalidate all checkpoints that follow!!! This is an
853 	 * advanced option.
854 	 */
855 
856 	/* setup basic mountpoint structure */
857 	mp->mnt_data = NULL;
858 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (uint32_t) devvp->v_rdev;
859 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_NILFS);
860 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
861 	mp->mnt_stat.f_namemax = NILFS_NAME_LEN;
862 	mp->mnt_flag  |= MNT_LOCAL;
863 
864 	/* XXX can't enable MPSAFE yet since genfs barfs on bad CV */
865 	// mp->mnt_iflag |= IMNT_MPSAFE;
866 
867 	/* set our dev and fs units */
868 	mp->mnt_dev_bshift = nilfs_rw32(nilfsdev->super.s_log_block_size) + 10;
869 	mp->mnt_fs_bshift  = mp->mnt_dev_bshift;
870 
871 	/* allocate nilfs part of mount structure; malloc always succeeds */
872 	ump = malloc(sizeof(struct nilfs_mount), M_NILFSMNT, M_WAITOK | M_ZERO);
873 
874 	/* init locks */
875 	mutex_init(&ump->ihash_lock, MUTEX_DEFAULT, IPL_NONE);
876 	mutex_init(&ump->get_node_lock, MUTEX_DEFAULT, IPL_NONE);
877 
878 	/* init our hash table for inode lookup */
879 	for (lst = 0; lst < NILFS_INODE_HASHSIZE; lst++) {
880 		LIST_INIT(&ump->nilfs_nodes[lst]);
881 	}
882 
883 	/* set up linkage */
884 	mp->mnt_data    =  ump;
885 	ump->vfs_mountp =  mp;
886 	ump->nilfsdev   =  nilfsdev;
887 
888 #if 0
889 #ifndef NILFS_READWRITE
890 	/* force read-only for now */
891 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
892 		printf( "Enable kernel/module option NILFS_READWRITE for "
893 			"writing, downgrading access to read-only\n");
894 		mp->mnt_flag |= MNT_RDONLY;
895 	}
896 #endif
897 #endif
898 
899 	/* DONT register our nilfs mountpoint on our vfs mountpoint */
900 	devvp->v_specmountpoint = NULL;
901 #if 0
902 	if (devvp->v_specmountpoint == NULL)
903 		devvp->v_specmountpoint = mp;
904 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
905 		devvp->v_specmountpoint = mp;
906 #endif
907 
908 	/* add our mountpoint */
909 	STAILQ_INSERT_TAIL(&nilfsdev->mounts, ump, next_mount);
910 
911 	/* get our selected checkpoint */
912 	if (args->cpno == 0)
913 		args->cpno = nilfsdev->last_cno;
914 	args->cpno = MIN(args->cpno, nilfsdev->last_cno);
915 
916 	/* setting up other parameters */
917 	ump->mount_args = *args;
918 	error = nilfs_mount_checkpoint(ump);
919 	if (error) {
920 		nilfs_unmount(mp, MNT_FORCE);
921 		return error;
922 	}
923 
924 	/* set VFS info */
925 	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec, UIO_USERSPACE,
926 			mp->mnt_op->vfs_name, mp, curlwp);
927 	if (error) {
928 		nilfs_unmount(mp, MNT_FORCE);
929 		return error;
930 	}
931 
932 	/* successfully mounted */
933 	DPRINTF(VOLUMES, ("nilfs_mount() successfull\n"));
934 
935 	return 0;
936 }
937 
938 /* --------------------------------------------------------------------- */
939 
940 
941 /* remove our mountpoint and if its the last reference, remove our device */
942 int
943 nilfs_unmount(struct mount *mp, int mntflags)
944 {
945 	struct nilfs_device *nilfsdev;
946 	struct nilfs_mount  *ump;
947 	int error, flags;
948 
949 	DPRINTF(VFSCALL, ("nilfs_umount called\n"));
950 
951 	ump = VFSTONILFS(mp);
952 	if (!ump)
953 		panic("NILFS unmount: empty ump\n");
954 	nilfsdev = ump->nilfsdev;
955 
956 	/*
957 	 * Flush all nodes associated to this mountpoint. By specifying
958 	 * SKIPSYSTEM we can skip vnodes marked with VV_SYSTEM. This hardly
959 	 * documented feature allows us to exempt certain files from being
960 	 * flushed.
961 	 */
962 	flags = (mntflags & MNT_FORCE) ? FORCECLOSE : 0;
963 	if ((error = vflush(mp, NULLVP, flags | SKIPSYSTEM)) != 0)
964 		return error;
965 
966 	/* if we're the write mount, we ought to close the writing session */
967 	error = nilfs_stop_writing(ump);
968 	if (error)
969 		return error;
970 
971 	if (ump->ifile_node)
972 		NILFS_UNSET_SYSTEMFILE(ump->ifile_node->vnode);
973 	nilfs_dispose_node(&ump->ifile_node);
974 
975 	/* remove our mount point */
976 	STAILQ_REMOVE(&nilfsdev->mounts, ump, nilfs_mount, next_mount);
977 	free_nilfs_mountinfo(mp);
978 
979 	/* free ump struct references */
980 	mp->mnt_data = NULL;
981 	mp->mnt_flag &= ~MNT_LOCAL;
982 
983 	/* unmount the device itself when we're the last one */
984 	nilfs_unmount_device(nilfsdev);
985 
986 	DPRINTF(VOLUMES, ("Fin unmount\n"));
987 	return error;
988 }
989 
990 /* --------------------------------------------------------------------- */
991 
992 int
993 nilfs_start(struct mount *mp, int flags)
994 {
995 	/* do we have to do something here? */
996 	return 0;
997 }
998 
999 /* --------------------------------------------------------------------- */
1000 
1001 int
1002 nilfs_root(struct mount *mp, struct vnode **vpp)
1003 {
1004 	struct nilfs_mount *ump = VFSTONILFS(mp);
1005 	struct nilfs_node  *node;
1006 	int error;
1007 
1008 	DPRINTF(NODE, ("nilfs_root called\n"));
1009 
1010 	error = nilfs_get_node(ump, NILFS_ROOT_INO, &node);
1011 	if (node)  {
1012 		*vpp = node->vnode;
1013 		KASSERT(node->vnode->v_vflag & VV_ROOT);
1014 	}
1015 
1016 	DPRINTF(NODE, ("nilfs_root finished\n"));
1017 	return error;
1018 }
1019 
1020 /* --------------------------------------------------------------------- */
1021 
1022 int
1023 nilfs_statvfs(struct mount *mp, struct statvfs *sbp)
1024 {
1025 	struct nilfs_mount *ump = VFSTONILFS(mp);
1026 	uint32_t blocksize;
1027 
1028 	DPRINTF(VFSCALL, ("nilfs_statvfs called\n"));
1029 
1030 	blocksize = ump->nilfsdev->blocksize;
1031 	sbp->f_flag   = mp->mnt_flag;
1032 	sbp->f_bsize  = blocksize;
1033 	sbp->f_frsize = blocksize;
1034 	sbp->f_iosize = blocksize;
1035 
1036 	copy_statvfs_info(sbp, mp);
1037 	return 0;
1038 }
1039 
1040 /* --------------------------------------------------------------------- */
1041 
1042 int
1043 nilfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
1044 {
1045 //	struct nilfs_mount *ump = VFSTONILFS(mp);
1046 
1047 	DPRINTF(VFSCALL, ("nilfs_sync called\n"));
1048 	/* if called when mounted readonly, just ignore */
1049 	if (mp->mnt_flag & MNT_RDONLY)
1050 		return 0;
1051 
1052 	DPRINTF(VFSCALL, ("end of nilfs_sync()\n"));
1053 
1054 	return 0;
1055 }
1056 
1057 /* --------------------------------------------------------------------- */
1058 
1059 /*
1060  * Get vnode for the file system type specific file id ino for the fs. Its
1061  * used for reference to files by unique ID and for NFSv3.
1062  * (optional) TODO lookup why some sources state NFSv3
1063  */
1064 int
1065 nilfs_vget(struct mount *mp, ino_t ino,
1066     struct vnode **vpp)
1067 {
1068 	DPRINTF(NOTIMPL, ("nilfs_vget called\n"));
1069 	return EOPNOTSUPP;
1070 }
1071 
1072 /* --------------------------------------------------------------------- */
1073 
1074 /*
1075  * Lookup vnode for file handle specified
1076  */
1077 int
1078 nilfs_fhtovp(struct mount *mp, struct fid *fhp,
1079     struct vnode **vpp)
1080 {
1081 	DPRINTF(NOTIMPL, ("nilfs_fhtovp called\n"));
1082 	return EOPNOTSUPP;
1083 }
1084 
1085 /* --------------------------------------------------------------------- */
1086 
1087 /*
1088  * Create an unique file handle. Its structure is opaque and won't be used by
1089  * other subsystems. It should uniquely identify the file in the filingsystem
1090  * and enough information to know if a file has been removed and/or resources
1091  * have been recycled.
1092  */
1093 int
1094 nilfs_vptofh(struct vnode *vp, struct fid *fid,
1095     size_t *fh_size)
1096 {
1097 	DPRINTF(NOTIMPL, ("nilfs_vptofh called\n"));
1098 	return EOPNOTSUPP;
1099 }
1100 
1101 /* --------------------------------------------------------------------- */
1102 
1103 /*
1104  * Create a file system snapshot at the specified timestamp.
1105  */
1106 int
1107 nilfs_snapshot(struct mount *mp, struct vnode *vp,
1108     struct timespec *tm)
1109 {
1110 	DPRINTF(NOTIMPL, ("nilfs_snapshot called\n"));
1111 	return EOPNOTSUPP;
1112 }
1113 
1114 /* --------------------------------------------------------------------- */
1115