xref: /netbsd-src/sys/fs/nilfs/nilfs_vfsops.c (revision b7b7574d3bf8eeb51a1fa3977b59142ec6434a55)
1 /* $NetBSD: nilfs_vfsops.c,v 1.16 2014/04/16 18:55:18 maxv 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 1.16 2014/04/16 18:55:18 maxv 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 	.vfs_name = MOUNT_NILFS,
115 	.vfs_min_mount_data = sizeof (struct nilfs_args),
116 	.vfs_mount = nilfs_mount,
117 	.vfs_start = nilfs_start,
118 	.vfs_unmount = nilfs_unmount,
119 	.vfs_root = nilfs_root,
120 	.vfs_quotactl = (void *)eopnotsupp,
121 	.vfs_statvfs = nilfs_statvfs,
122 	.vfs_sync = nilfs_sync,
123 	.vfs_vget = nilfs_vget,
124 	.vfs_fhtovp = nilfs_fhtovp,
125 	.vfs_vptofh = nilfs_vptofh,
126 	.vfs_init = nilfs_init,
127 	.vfs_reinit = nilfs_reinit,
128 	.vfs_done = nilfs_done,
129 	.vfs_mountroot = nilfs_mountroot,
130 	.vfs_snapshot = nilfs_snapshot,
131 	.vfs_extattrctl = vfs_stdextattrctl,
132 	.vfs_suspendctl = (void *)eopnotsupp,
133 	.vfs_renamelock_enter = genfs_renamelock_enter,
134 	.vfs_renamelock_exit = genfs_renamelock_exit,
135 	.vfs_fsync = (void *)eopnotsupp,
136 	.vfs_opv_descs = nilfs_vnodeopv_descs
137 };
138 
139 /* --------------------------------------------------------------------- */
140 
141 /* file system starts here */
142 void
143 nilfs_init(void)
144 {
145 	size_t size;
146 
147 	/* setup memory types */
148 	malloc_type_attach(M_NILFSMNT);
149 	malloc_type_attach(M_NILFSTEMP);
150 
151 	/* init device lists */
152 	SLIST_INIT(&nilfs_devices);
153 
154 	/* init node pools */
155 	size = sizeof(struct nilfs_node);
156 	pool_init(&nilfs_node_pool, size, 0, 0, 0,
157 		"nilfs_node_pool", NULL, IPL_NONE);
158 }
159 
160 
161 void
162 nilfs_reinit(void)
163 {
164 	/* nothing to do */
165 }
166 
167 
168 void
169 nilfs_done(void)
170 {
171 	/* remove pools */
172 	pool_destroy(&nilfs_node_pool);
173 
174 	malloc_type_detach(M_NILFSMNT);
175 	malloc_type_detach(M_NILFSTEMP);
176 }
177 
178 /*
179  * If running a DEBUG kernel, provide an easy way to set the debug flags when
180  * running into a problem.
181  */
182 #define NILFS_VERBOSE_SYSCTLOPT        1
183 
184 static int
185 nilfs_modcmd(modcmd_t cmd, void *arg)
186 {
187 	const struct sysctlnode *node;
188 	int error;
189 
190 	switch (cmd) {
191 	case MODULE_CMD_INIT:
192 		error = vfs_attach(&nilfs_vfsops);
193 		if (error != 0)
194 			break;
195 		/*
196 		 * XXX the "30" below could be dynamic, thereby eliminating one
197 		 * more instance of the "number to vfs" mapping problem, but
198 		 * "30" is the order as taken from sys/mount.h
199 		 */
200 		sysctl_createv(&nilfs_sysctl_log, 0, NULL, &node,
201 			       CTLFLAG_PERMANENT,
202 			       CTLTYPE_NODE, "nilfs",
203 			       SYSCTL_DESCR("NTT's NILFSv2"),
204 			       NULL, 0, NULL, 0,
205 			       CTL_VFS, 30, CTL_EOL);
206 #ifdef DEBUG
207 		sysctl_createv(&nilfs_sysctl_log, 0, NULL, &node,
208 			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
209 			       CTLTYPE_INT, "verbose",
210 			       SYSCTL_DESCR("Bitmask for filesystem debugging"),
211 			       NULL, 0, &nilfs_verbose, 0,
212 			       CTL_VFS, 30, NILFS_VERBOSE_SYSCTLOPT, CTL_EOL);
213 #endif
214 		break;
215 	case MODULE_CMD_FINI:
216 		error = vfs_detach(&nilfs_vfsops);
217 		if (error != 0)
218 			break;
219 		sysctl_teardown(&nilfs_sysctl_log);
220 		break;
221 	default:
222 		error = ENOTTY;
223 		break;
224 	}
225 
226 	return (error);
227 }
228 
229 /* --------------------------------------------------------------------- */
230 
231 int
232 nilfs_mountroot(void)
233 {
234 	return EOPNOTSUPP;
235 }
236 
237 /* --------------------------------------------------------------------- */
238 
239 /* system nodes */
240 static int
241 nilfs_create_system_nodes(struct nilfs_device *nilfsdev)
242 {
243 	int error;
244 
245 	error = nilfs_get_node_raw(nilfsdev, NULL, NILFS_DAT_INO,
246 		&nilfsdev->super_root.sr_dat, &nilfsdev->dat_node);
247 	if (error)
248 		goto errorout;
249 
250 	error = nilfs_get_node_raw(nilfsdev, NULL, NILFS_CPFILE_INO,
251 		&nilfsdev->super_root.sr_cpfile, &nilfsdev->cp_node);
252 	if (error)
253 		goto errorout;
254 
255 	error = nilfs_get_node_raw(nilfsdev, NULL, NILFS_SUFILE_INO,
256 		&nilfsdev->super_root.sr_sufile, &nilfsdev->su_node);
257 	if (error)
258 		goto errorout;
259 
260 	NILFS_SET_SYSTEMFILE(nilfsdev->dat_node->vnode);
261 	NILFS_SET_SYSTEMFILE(nilfsdev->cp_node->vnode);
262 	NILFS_SET_SYSTEMFILE(nilfsdev->su_node->vnode);
263 
264 	return 0;
265 errorout:
266 	nilfs_dispose_node(&nilfsdev->dat_node);
267 	nilfs_dispose_node(&nilfsdev->cp_node);
268 	nilfs_dispose_node(&nilfsdev->su_node);
269 
270 	return error;
271 }
272 
273 
274 static void
275 nilfs_release_system_nodes(struct nilfs_device *nilfsdev)
276 {
277 	if (!nilfsdev)
278 		return;
279 	if (nilfsdev->refcnt > 0)
280 		return;
281 
282 	if (nilfsdev->dat_node)
283 		NILFS_UNSET_SYSTEMFILE(nilfsdev->dat_node->vnode);
284 	if (nilfsdev->cp_node)
285 		NILFS_UNSET_SYSTEMFILE(nilfsdev->cp_node->vnode);
286 	if (nilfsdev->su_node)
287 		NILFS_UNSET_SYSTEMFILE(nilfsdev->su_node->vnode);
288 
289 	nilfs_dispose_node(&nilfsdev->dat_node);
290 	nilfs_dispose_node(&nilfsdev->cp_node);
291 	nilfs_dispose_node(&nilfsdev->su_node);
292 }
293 
294 
295 /* --------------------------------------------------------------------- */
296 
297 static int
298 nilfs_check_superblock_crc(struct nilfs_super_block *super)
299 {
300 	uint32_t super_crc, comp_crc;
301 
302 	/* check super block magic */
303 	if (nilfs_rw16(super->s_magic) != NILFS_SUPER_MAGIC)
304 		return 0;
305 
306 	/* preserve crc */
307 	super_crc  = nilfs_rw32(super->s_sum);
308 
309 	/* calculate */
310 	super->s_sum = 0;
311 	comp_crc = crc32_le(nilfs_rw32(super->s_crc_seed),
312 		(uint8_t *) super, nilfs_rw16(super->s_bytes));
313 
314 	/* restore */
315 	super->s_sum = nilfs_rw32(super_crc);
316 
317 	/* check CRC */
318 	return (super_crc == comp_crc);
319 }
320 
321 
322 
323 static int
324 nilfs_read_superblock(struct nilfs_device *nilfsdev)
325 {
326 	struct nilfs_super_block *super, tmp_super;
327 	struct buf *bp;
328 	uint64_t sb1off, sb2off;
329 	uint64_t last_cno1, last_cno2;
330 	uint64_t dev_blk;
331 	int dev_bsize, dev_blks;
332 	int sb1ok, sb2ok, swp;
333 	int error;
334 
335 	sb1off = NILFS_SB_OFFSET_BYTES;
336 	sb2off = NILFS_SB2_OFFSET_BYTES(nilfsdev->devsize);
337 
338 	dev_bsize = 1 << nilfsdev->devvp->v_mount->mnt_fs_bshift;
339 
340 	/* read our superblock regardless of backing device blocksize */
341 	dev_blk   = 0;
342 	dev_blks  = (sb1off + dev_bsize -1)/dev_bsize;
343 	error = bread(nilfsdev->devvp, dev_blk, dev_blks * dev_bsize, NOCRED, 0, &bp);
344 	if (error) {
345 		return error;
346 	}
347 
348 	/* copy read-in super block at the offset */
349 	super = &nilfsdev->super;
350 	memcpy(super, (uint8_t *) bp->b_data + NILFS_SB_OFFSET_BYTES,
351 		sizeof(struct nilfs_super_block));
352 	brelse(bp, BC_AGE);
353 
354 	/* read our 2nd superblock regardless of backing device blocksize */
355 	dev_blk   = sb2off / dev_bsize;
356 	dev_blks  = 2;		/* assumption max one dev_bsize */
357 	error = bread(nilfsdev->devvp, dev_blk, dev_blks * dev_bsize, NOCRED, 0, &bp);
358 	if (error) {
359 		return error;
360 	}
361 
362 	/* copy read-in superblock2 at the offset */
363 	super = &nilfsdev->super2;
364 	memcpy(super, (uint8_t *) bp->b_data + NILFS_SB_OFFSET_BYTES,
365 		sizeof(struct nilfs_super_block));
366 	brelse(bp, BC_AGE);
367 
368 	sb1ok = nilfs_check_superblock_crc(&nilfsdev->super);
369 	sb2ok = nilfs_check_superblock_crc(&nilfsdev->super2);
370 
371 	last_cno1 = nilfs_rw64(nilfsdev->super.s_last_cno);
372 	last_cno2 = nilfs_rw64(nilfsdev->super2.s_last_cno);
373 	swp = sb2ok && (last_cno2 > last_cno1);
374 
375 	if (swp) {
376 		printf("nilfs warning: broken superblock, using spare\n");
377 		tmp_super = nilfsdev->super2;
378 		nilfsdev->super2 = nilfsdev->super;	/* why preserve? */
379 		nilfsdev->super  = tmp_super;
380 	}
381 
382 	if (!sb1ok && !sb2ok) {
383 		printf("nilfs: no valid superblocks found\n");
384 		return EINVAL;
385 	}
386 
387 	return 0;
388 }
389 
390 
391 /* XXX NOTHING from the system nodes should need to be written here */
392 static void
393 nilfs_unmount_base(struct nilfs_device *nilfsdev)
394 {
395 	int error __diagused;
396 
397 	if (!nilfsdev)
398 		return;
399 
400 	/* remove all our information */
401 	error = vinvalbuf(nilfsdev->devvp, 0, FSCRED, curlwp, 0, 0);
402 	KASSERT(error == 0);
403 
404 	/* release the device's system nodes */
405 	nilfs_release_system_nodes(nilfsdev);
406 
407 	/* TODO writeout super_block? */
408 }
409 
410 
411 static int
412 nilfs_mount_base(struct nilfs_device *nilfsdev,
413 		struct mount *mp, struct nilfs_args *args)
414 {
415 	struct lwp *l = curlwp;
416 	uint64_t last_pseg, last_cno, last_seq;
417 	uint32_t log_blocksize;
418 	int error;
419 
420 	/* flush out any old buffers remaining from a previous use. */
421 	if ((error = vinvalbuf(nilfsdev->devvp, V_SAVE, l->l_cred, l, 0, 0)))
422 		return error;
423 
424 	/* read in our superblock */
425 	error = nilfs_read_superblock(nilfsdev);
426 	if (error) {
427 		printf("nilfs_mount: can't read in super block : %d\n", error);
428 		return error;
429 	}
430 
431 	/* get our blocksize */
432 	log_blocksize = nilfs_rw32(nilfsdev->super.s_log_block_size);
433 	nilfsdev->blocksize   = (uint64_t) 1 << (log_blocksize + 10);
434 	/* TODO check superblock's blocksize limits */
435 
436 	/* calculate dat structure parameters */
437 	nilfs_calc_mdt_consts(nilfsdev, &nilfsdev->dat_mdt,
438 			nilfs_rw16(nilfsdev->super.s_dat_entry_size));
439 	nilfs_calc_mdt_consts(nilfsdev, &nilfsdev->ifile_mdt,
440 			nilfs_rw16(nilfsdev->super.s_inode_size));
441 
442 	DPRINTF(VOLUMES, ("nilfs_mount: accepted super block\n"));
443 
444 	/* search for the super root and roll forward when needed */
445 	nilfs_search_super_root(nilfsdev);
446 
447 	nilfsdev->mount_state = nilfs_rw16(nilfsdev->super.s_state);
448 	if (nilfsdev->mount_state != NILFS_VALID_FS) {
449 		printf("FS is seriously damaged, needs repairing\n");
450 		printf("aborting mount\n");
451 		return EINVAL;
452 	}
453 
454 	/*
455 	 * FS should be ok now. The superblock and the last segsum could be
456 	 * updated from the repair so extract running values again.
457 	 */
458 	last_pseg = nilfs_rw64(nilfsdev->super.s_last_pseg); /*blknr */
459 	last_cno  = nilfs_rw64(nilfsdev->super.s_last_cno);
460 	last_seq  = nilfs_rw64(nilfsdev->super.s_last_seq);
461 
462 	nilfsdev->last_seg_seq = last_seq;
463 	nilfsdev->last_seg_num = nilfs_get_segnum_of_block(nilfsdev, last_pseg);
464 	nilfsdev->next_seg_num = nilfs_get_segnum_of_block(nilfsdev,
465 		nilfs_rw64(nilfsdev->last_segsum.ss_next));
466 	nilfsdev->last_cno     = last_cno;
467 
468 	DPRINTF(VOLUMES, ("nilfs_mount: accepted super root\n"));
469 
470 	/* create system vnodes for DAT, CP and SEGSUM */
471 	error = nilfs_create_system_nodes(nilfsdev);
472 	if (error)
473 		nilfs_unmount_base(nilfsdev);
474 	return error;
475 }
476 
477 
478 static void
479 nilfs_unmount_device(struct nilfs_device *nilfsdev)
480 {
481 	int error;
482 
483 	/* is there anything? */
484 	if (nilfsdev == NULL)
485 		return;
486 
487 	/* remove the device only if we're the last reference */
488 	nilfsdev->refcnt--;
489 	if (nilfsdev->refcnt >= 1)
490 		return;
491 
492 	/* unmount our base */
493 	nilfs_unmount_base(nilfsdev);
494 
495 	/* remove from our device list */
496 	SLIST_REMOVE(&nilfs_devices, nilfsdev, nilfs_device, next_device);
497 
498 	/* close device */
499 	DPRINTF(VOLUMES, ("closing device\n"));
500 
501 	/* remove our mount reference before closing device */
502 	spec_node_setmountedfs(nilfsdev->devvp, NULL);
503 
504 	/* devvp is still locked by us */
505 	vn_lock(nilfsdev->devvp, LK_EXCLUSIVE | LK_RETRY);
506 	error = VOP_CLOSE(nilfsdev->devvp, FREAD | FWRITE, NOCRED);
507 	if (error)
508 		printf("Error during closure of device! error %d, "
509 		       "device might stay locked\n", error);
510 	DPRINTF(VOLUMES, ("device close ok\n"));
511 
512 	/* clear our mount reference and release device node */
513 	vput(nilfsdev->devvp);
514 
515 	/* free our device info */
516 	free(nilfsdev, M_NILFSMNT);
517 }
518 
519 
520 static int
521 nilfs_check_mounts(struct nilfs_device *nilfsdev, struct mount *mp,
522 	struct nilfs_args *args)
523 {
524 	struct nilfs_mount  *ump;
525 	uint64_t last_cno;
526 
527 	/* no double-mounting of the same checkpoint */
528 	STAILQ_FOREACH(ump, &nilfsdev->mounts, next_mount) {
529 		if (ump->mount_args.cpno == args->cpno)
530 			return EBUSY;
531 	}
532 
533 	/* allow readonly mounts without questioning here */
534 	if (mp->mnt_flag & MNT_RDONLY)
535 		return 0;
536 
537 	/* readwrite mount you want */
538 	STAILQ_FOREACH(ump, &nilfsdev->mounts, next_mount) {
539 		/* only one RW mount on this device! */
540 		if ((ump->vfs_mountp->mnt_flag & MNT_RDONLY)==0)
541 			return EROFS;
542 		/* RDONLY on last mountpoint is device busy */
543 		last_cno = nilfs_rw64(ump->nilfsdev->super.s_last_cno);
544 		if (ump->mount_args.cpno == last_cno)
545 			return EBUSY;
546 	}
547 
548 	/* OK for now */
549 	return 0;
550 }
551 
552 
553 static int
554 nilfs_mount_device(struct vnode *devvp, struct mount *mp, struct nilfs_args *args,
555 	struct nilfs_device **nilfsdev_p)
556 {
557 	uint64_t psize;
558 	unsigned secsize;
559 	struct nilfs_device *nilfsdev;
560 	struct lwp *l = curlwp;
561 	int openflags, accessmode, error;
562 
563 	DPRINTF(VOLUMES, ("Mounting NILFS device\n"));
564 
565 	/* lookup device in our nilfs_mountpoints */
566 	*nilfsdev_p = NULL;
567 	SLIST_FOREACH(nilfsdev, &nilfs_devices, next_device)
568 		if (nilfsdev->devvp == devvp)
569 			break;
570 
571 	if (nilfsdev) {
572 		DPRINTF(VOLUMES, ("device already mounted\n"));
573 		error = nilfs_check_mounts(nilfsdev, mp, args);
574 		if (error)
575 			return error;
576 		nilfsdev->refcnt++;
577 		*nilfsdev_p = nilfsdev;
578 		return 0;
579 	}
580 
581 	DPRINTF(VOLUMES, ("no previous mounts on this device, mounting device\n"));
582 
583 	/* check if its a block device specified */
584 	if (devvp->v_type != VBLK) {
585 		vrele(devvp);
586 		return ENOTBLK;
587 	}
588 	if (bdevsw_lookup(devvp->v_rdev) == NULL) {
589 		vrele(devvp);
590 		return ENXIO;
591 	}
592 
593 	/*
594 	 * If mount by non-root, then verify that user has necessary
595 	 * permissions on the device.
596 	 */
597 	accessmode = VREAD;
598 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
599 		accessmode |= VWRITE;
600 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
601 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
602 	    KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, KAUTH_ARG(accessmode));
603 	VOP_UNLOCK(devvp);
604 	if (error) {
605 		vrele(devvp);
606 		return error;
607 	}
608 
609 	/*
610 	 * Open device read-write; TODO how about upgrading later when needed?
611 	 */
612 	openflags = FREAD | FWRITE;
613 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
614 	error = VOP_OPEN(devvp, openflags, FSCRED);
615 	VOP_UNLOCK(devvp);
616 	if (error) {
617 		vrele(devvp);
618 		return error;
619 	}
620 
621 	/* opened ok, try mounting */
622 	nilfsdev = malloc(sizeof(*nilfsdev), M_NILFSMNT, M_WAITOK | M_ZERO);
623 
624 	/* initialise */
625 	nilfsdev->refcnt        = 1;
626 	nilfsdev->devvp         = devvp;
627 	nilfsdev->uncomitted_bl = 0;
628 	cv_init(&nilfsdev->sync_cv, "nilfssyn");
629 	STAILQ_INIT(&nilfsdev->mounts);
630 
631 	/* register nilfs_device in list */
632 	SLIST_INSERT_HEAD(&nilfs_devices, nilfsdev, next_device);
633 
634 	/* get our device's size */
635 	error = getdisksize(devvp, &psize, &secsize);
636 	if (error) {
637 		/* remove all our information */
638 		nilfs_unmount_device(nilfsdev);
639 		return EINVAL;
640 	}
641 
642 	nilfsdev->devsize = psize * secsize;
643 
644 	/* connect to the head for most recent files XXX really pass mp and args? */
645 	error = nilfs_mount_base(nilfsdev, mp, args);
646 	if (error) {
647 		/* remove all our information */
648 		nilfs_unmount_device(nilfsdev);
649 		return EINVAL;
650 	}
651 
652 	*nilfsdev_p = nilfsdev;
653 	DPRINTF(VOLUMES, ("NILFS device mounted ok\n"));
654 
655 	return 0;
656 }
657 
658 
659 static int
660 nilfs_mount_checkpoint(struct nilfs_mount *ump)
661 {
662 	struct nilfs_cpfile_header *cphdr;
663 	struct nilfs_checkpoint *cp;
664 	struct nilfs_inode  ifile_inode;
665 	struct nilfs_node  *cp_node;
666 	struct buf *bp;
667 	uint64_t ncp, nsn, fcpno, blocknr, last_cno;
668 	uint32_t off, dlen;
669 	int cp_per_block, error;
670 
671 	DPRINTF(VOLUMES, ("mount_nilfs: trying to mount checkpoint number "
672 		"%"PRIu64"\n", ump->mount_args.cpno));
673 
674 	cp_node = ump->nilfsdev->cp_node;
675 
676 	/* get cpfile header from 1st block of cp file */
677 	error = nilfs_bread(cp_node, 0, NOCRED, 0, &bp);
678 	if (error)
679 		return error;
680 	cphdr = (struct nilfs_cpfile_header *) bp->b_data;
681 	ncp = nilfs_rw64(cphdr->ch_ncheckpoints);
682 	nsn = nilfs_rw64(cphdr->ch_nsnapshots);
683 
684 	brelse(bp, BC_AGE);
685 
686 	DPRINTF(VOLUMES, ("mount_nilfs: checkpoint header read in\n"));
687 	DPRINTF(VOLUMES, ("\tNumber of checkpoints %"PRIu64"\n", ncp));
688 	DPRINTF(VOLUMES, ("\tNumber of snapshots   %"PRIu64"\n", nsn));
689 #ifndef NILFS_DEBUG
690 	__USE(ncp);
691 	__USE(nsn);
692 #endif
693 
694 	/* read in our specified checkpoint */
695 	dlen = nilfs_rw16(ump->nilfsdev->super.s_checkpoint_size);
696 	cp_per_block = ump->nilfsdev->blocksize / dlen;
697 
698 	fcpno = ump->mount_args.cpno + NILFS_CPFILE_FIRST_CHECKPOINT_OFFSET -1;
699 	blocknr =  fcpno / cp_per_block;
700 	off     = (fcpno % cp_per_block) * dlen;
701 
702 	error = nilfs_bread(cp_node, blocknr, NOCRED, 0, &bp);
703 	if (error) {
704 		printf("mount_nilfs: couldn't read cp block %"PRIu64"\n",
705 			fcpno);
706 		return EINVAL;
707 	}
708 
709 	/* needs to be a valid checkpoint */
710 	cp = (struct nilfs_checkpoint *) ((uint8_t *) bp->b_data + off);
711 	if (cp->cp_flags & NILFS_CHECKPOINT_INVALID) {
712 		printf("mount_nilfs: checkpoint marked invalid\n");
713 		brelse(bp, BC_AGE);
714 		return EINVAL;
715 	}
716 
717 	/* is this really the checkpoint we want? */
718 	if (nilfs_rw64(cp->cp_cno) != ump->mount_args.cpno) {
719 		printf("mount_nilfs: checkpoint file corrupt? "
720 			"expected cpno %"PRIu64", found cpno %"PRIu64"\n",
721 			ump->mount_args.cpno, nilfs_rw64(cp->cp_cno));
722 		brelse(bp, BC_AGE);
723 		return EINVAL;
724 	}
725 
726 	/* check if its a snapshot ! */
727 	last_cno = nilfs_rw64(ump->nilfsdev->super.s_last_cno);
728 	if (ump->mount_args.cpno != last_cno) {
729 		/* only allow snapshots if not mounting on the last cp */
730 		if ((cp->cp_flags & NILFS_CHECKPOINT_SNAPSHOT) == 0) {
731 			printf( "mount_nilfs: checkpoint %"PRIu64" is not a "
732 				"snapshot\n", ump->mount_args.cpno);
733 			brelse(bp, BC_AGE);
734 			return EINVAL;
735 		}
736 	}
737 
738 	ifile_inode = cp->cp_ifile_inode;
739 	brelse(bp, BC_AGE);
740 
741 	/* get ifile inode */
742 	error = nilfs_get_node_raw(ump->nilfsdev, NULL, NILFS_IFILE_INO,
743 		&ifile_inode, &ump->ifile_node);
744 	if (error) {
745 		printf("mount_nilfs: can't read ifile node\n");
746 		return EINVAL;
747 	}
748 	NILFS_SET_SYSTEMFILE(ump->ifile_node->vnode);
749 
750 	/* get root node? */
751 
752 	return 0;
753 }
754 
755 
756 static int
757 nilfs_stop_writing(struct nilfs_mount *ump)
758 {
759 	/* readonly mounts won't write */
760 	if (ump->vfs_mountp->mnt_flag & MNT_RDONLY)
761 		return 0;
762 
763 	DPRINTF(CALL, ("nilfs_stop_writing called for RW mount\n"));
764 
765 	/* TODO writeout super_block? */
766 	/* XXX no support for writing yet anyway */
767 	return 0;
768 }
769 
770 
771 /* --------------------------------------------------------------------- */
772 
773 
774 
775 #define MPFREE(a, lst) \
776 	if ((a)) free((a), lst);
777 static void
778 free_nilfs_mountinfo(struct mount *mp)
779 {
780 	struct nilfs_mount *ump = VFSTONILFS(mp);
781 
782 	if (ump == NULL)
783 		return;
784 
785 	mutex_destroy(&ump->ihash_lock);
786 	mutex_destroy(&ump->get_node_lock);
787 	MPFREE(ump, M_NILFSMNT);
788 }
789 #undef MPFREE
790 
791 int
792 nilfs_mount(struct mount *mp, const char *path,
793 	  void *data, size_t *data_len)
794 {
795 	struct nilfs_args   *args = data;
796 	struct nilfs_device *nilfsdev;
797 	struct nilfs_mount  *ump;
798 	struct vnode *devvp;
799 	int lst, error;
800 
801 	DPRINTF(VFSCALL, ("nilfs_mount called\n"));
802 
803 	if (args == NULL)
804 		return EINVAL;
805 	if (*data_len < sizeof *args)
806 		return EINVAL;
807 
808 	if (mp->mnt_flag & MNT_GETARGS) {
809 		/* request for the mount arguments */
810 		ump = VFSTONILFS(mp);
811 		if (ump == NULL)
812 			return EINVAL;
813 		*args = ump->mount_args;
814 		*data_len = sizeof *args;
815 		return 0;
816 	}
817 
818 	/* check/translate struct version */
819 	if (args->version != 1) {
820 		printf("mount_nilfs: unrecognized argument structure version\n");
821 		return EINVAL;
822 	}
823 	/* TODO sanity checking other mount arguments */
824 
825 	/* handle request for updating mount parameters */
826 	if (mp->mnt_flag & MNT_UPDATE) {
827 		/* TODO can't update my mountpoint yet */
828 		return EOPNOTSUPP;
829 	}
830 
831 	/* lookup name to get its vnode */
832 	error = namei_simple_user(args->fspec, NSM_FOLLOW_NOEMULROOT, &devvp);
833 	if (error)
834 		return error;
835 
836 #ifdef DEBUG
837 	if (nilfs_verbose & NILFS_DEBUG_VOLUMES)
838 		vprint("NILFS mount, trying to mount \n", devvp);
839 #endif
840 
841 	error = nilfs_mount_device(devvp, mp, args, &nilfsdev);
842 	if (error)
843 		return error;
844 
845 	/*
846 	 * Create a nilfs_mount on the specified checkpoint. Note that only
847 	 * ONE RW mount point can exist and it needs to have the highest
848 	 * checkpoint nr. If mounting RW and its not on the last checkpoint we
849 	 * need to invalidate all checkpoints that follow!!! This is an
850 	 * advanced option.
851 	 */
852 
853 	/* setup basic mountpoint structure */
854 	mp->mnt_data = NULL;
855 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (uint32_t) devvp->v_rdev;
856 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_NILFS);
857 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
858 	mp->mnt_stat.f_namemax = NILFS_NAME_LEN;
859 	mp->mnt_flag  |= MNT_LOCAL;
860 
861 	/* XXX can't enable MPSAFE yet since genfs barfs on bad CV */
862 	// mp->mnt_iflag |= IMNT_MPSAFE;
863 
864 	/* set our dev and fs units */
865 	mp->mnt_dev_bshift = nilfs_rw32(nilfsdev->super.s_log_block_size) + 10;
866 	mp->mnt_fs_bshift  = mp->mnt_dev_bshift;
867 
868 	/* allocate nilfs part of mount structure; malloc always succeeds */
869 	ump = malloc(sizeof(struct nilfs_mount), M_NILFSMNT, M_WAITOK | M_ZERO);
870 
871 	/* init locks */
872 	mutex_init(&ump->ihash_lock, MUTEX_DEFAULT, IPL_NONE);
873 	mutex_init(&ump->get_node_lock, MUTEX_DEFAULT, IPL_NONE);
874 
875 	/* init our hash table for inode lookup */
876 	for (lst = 0; lst < NILFS_INODE_HASHSIZE; lst++) {
877 		LIST_INIT(&ump->nilfs_nodes[lst]);
878 	}
879 
880 	/* set up linkage */
881 	mp->mnt_data    =  ump;
882 	ump->vfs_mountp =  mp;
883 	ump->nilfsdev   =  nilfsdev;
884 
885 #if 0
886 #ifndef NILFS_READWRITE
887 	/* force read-only for now */
888 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
889 		printf( "Enable kernel/module option NILFS_READWRITE for "
890 			"writing, downgrading access to read-only\n");
891 		mp->mnt_flag |= MNT_RDONLY;
892 	}
893 #endif
894 #endif
895 
896 	/* DONT register our nilfs mountpoint on our vfs mountpoint */
897 	spec_node_setmountedfs(devvp, NULL);
898 #if 0
899 	if (spec_node_getmountedfs(devvp) == NULL)
900 		spec_node_setmountedfs(devvp, mp);
901 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
902 		spec_node_setmountedfs(devvp, mp);
903 #endif
904 
905 	/* add our mountpoint */
906 	STAILQ_INSERT_TAIL(&nilfsdev->mounts, ump, next_mount);
907 
908 	/* get our selected checkpoint */
909 	if (args->cpno == 0)
910 		args->cpno = nilfsdev->last_cno;
911 	args->cpno = MIN(args->cpno, nilfsdev->last_cno);
912 
913 	/* setting up other parameters */
914 	ump->mount_args = *args;
915 	error = nilfs_mount_checkpoint(ump);
916 	if (error) {
917 		nilfs_unmount(mp, MNT_FORCE);
918 		return error;
919 	}
920 
921 	/* set VFS info */
922 	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec, UIO_USERSPACE,
923 			mp->mnt_op->vfs_name, mp, curlwp);
924 	if (error) {
925 		nilfs_unmount(mp, MNT_FORCE);
926 		return error;
927 	}
928 
929 	/* successfully mounted */
930 	DPRINTF(VOLUMES, ("nilfs_mount() successfull\n"));
931 
932 	return 0;
933 }
934 
935 /* --------------------------------------------------------------------- */
936 
937 
938 /* remove our mountpoint and if its the last reference, remove our device */
939 int
940 nilfs_unmount(struct mount *mp, int mntflags)
941 {
942 	struct nilfs_device *nilfsdev;
943 	struct nilfs_mount  *ump;
944 	int error, flags;
945 
946 	DPRINTF(VFSCALL, ("nilfs_umount called\n"));
947 
948 	ump = VFSTONILFS(mp);
949 	if (!ump)
950 		panic("NILFS unmount: empty ump\n");
951 	nilfsdev = ump->nilfsdev;
952 
953 	/*
954 	 * Flush all nodes associated to this mountpoint. By specifying
955 	 * SKIPSYSTEM we can skip vnodes marked with VV_SYSTEM. This hardly
956 	 * documented feature allows us to exempt certain files from being
957 	 * flushed.
958 	 */
959 	flags = (mntflags & MNT_FORCE) ? FORCECLOSE : 0;
960 	if ((error = vflush(mp, NULLVP, flags | SKIPSYSTEM)) != 0)
961 		return error;
962 
963 	/* if we're the write mount, we ought to close the writing session */
964 	error = nilfs_stop_writing(ump);
965 	if (error)
966 		return error;
967 
968 	if (ump->ifile_node)
969 		NILFS_UNSET_SYSTEMFILE(ump->ifile_node->vnode);
970 	nilfs_dispose_node(&ump->ifile_node);
971 
972 	/* remove our mount point */
973 	STAILQ_REMOVE(&nilfsdev->mounts, ump, nilfs_mount, next_mount);
974 	free_nilfs_mountinfo(mp);
975 
976 	/* free ump struct references */
977 	mp->mnt_data = NULL;
978 	mp->mnt_flag &= ~MNT_LOCAL;
979 
980 	/* unmount the device itself when we're the last one */
981 	nilfs_unmount_device(nilfsdev);
982 
983 	DPRINTF(VOLUMES, ("Fin unmount\n"));
984 	return error;
985 }
986 
987 /* --------------------------------------------------------------------- */
988 
989 int
990 nilfs_start(struct mount *mp, int flags)
991 {
992 	/* do we have to do something here? */
993 	return 0;
994 }
995 
996 /* --------------------------------------------------------------------- */
997 
998 int
999 nilfs_root(struct mount *mp, struct vnode **vpp)
1000 {
1001 	struct nilfs_mount *ump = VFSTONILFS(mp);
1002 	struct nilfs_node  *node;
1003 	int error;
1004 
1005 	DPRINTF(NODE, ("nilfs_root called\n"));
1006 
1007 	error = nilfs_get_node(ump, NILFS_ROOT_INO, &node);
1008 	if (node)  {
1009 		*vpp = node->vnode;
1010 		KASSERT(node->vnode->v_vflag & VV_ROOT);
1011 	}
1012 
1013 	DPRINTF(NODE, ("nilfs_root finished\n"));
1014 	return error;
1015 }
1016 
1017 /* --------------------------------------------------------------------- */
1018 
1019 int
1020 nilfs_statvfs(struct mount *mp, struct statvfs *sbp)
1021 {
1022 	struct nilfs_mount *ump = VFSTONILFS(mp);
1023 	uint32_t blocksize;
1024 
1025 	DPRINTF(VFSCALL, ("nilfs_statvfs called\n"));
1026 
1027 	blocksize = ump->nilfsdev->blocksize;
1028 	sbp->f_flag   = mp->mnt_flag;
1029 	sbp->f_bsize  = blocksize;
1030 	sbp->f_frsize = blocksize;
1031 	sbp->f_iosize = blocksize;
1032 
1033 	copy_statvfs_info(sbp, mp);
1034 	return 0;
1035 }
1036 
1037 /* --------------------------------------------------------------------- */
1038 
1039 int
1040 nilfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
1041 {
1042 //	struct nilfs_mount *ump = VFSTONILFS(mp);
1043 
1044 	DPRINTF(VFSCALL, ("nilfs_sync called\n"));
1045 	/* if called when mounted readonly, just ignore */
1046 	if (mp->mnt_flag & MNT_RDONLY)
1047 		return 0;
1048 
1049 	DPRINTF(VFSCALL, ("end of nilfs_sync()\n"));
1050 
1051 	return 0;
1052 }
1053 
1054 /* --------------------------------------------------------------------- */
1055 
1056 /*
1057  * Get vnode for the file system type specific file id ino for the fs. Its
1058  * used for reference to files by unique ID and for NFSv3.
1059  * (optional) TODO lookup why some sources state NFSv3
1060  */
1061 int
1062 nilfs_vget(struct mount *mp, ino_t ino,
1063     struct vnode **vpp)
1064 {
1065 	DPRINTF(NOTIMPL, ("nilfs_vget called\n"));
1066 	return EOPNOTSUPP;
1067 }
1068 
1069 /* --------------------------------------------------------------------- */
1070 
1071 /*
1072  * Lookup vnode for file handle specified
1073  */
1074 int
1075 nilfs_fhtovp(struct mount *mp, struct fid *fhp,
1076     struct vnode **vpp)
1077 {
1078 	DPRINTF(NOTIMPL, ("nilfs_fhtovp called\n"));
1079 	return EOPNOTSUPP;
1080 }
1081 
1082 /* --------------------------------------------------------------------- */
1083 
1084 /*
1085  * Create an unique file handle. Its structure is opaque and won't be used by
1086  * other subsystems. It should uniquely identify the file in the filingsystem
1087  * and enough information to know if a file has been removed and/or resources
1088  * have been recycled.
1089  */
1090 int
1091 nilfs_vptofh(struct vnode *vp, struct fid *fid,
1092     size_t *fh_size)
1093 {
1094 	DPRINTF(NOTIMPL, ("nilfs_vptofh called\n"));
1095 	return EOPNOTSUPP;
1096 }
1097 
1098 /* --------------------------------------------------------------------- */
1099 
1100 /*
1101  * Create a file system snapshot at the specified timestamp.
1102  */
1103 int
1104 nilfs_snapshot(struct mount *mp, struct vnode *vp,
1105     struct timespec *tm)
1106 {
1107 	DPRINTF(NOTIMPL, ("nilfs_snapshot called\n"));
1108 	return EOPNOTSUPP;
1109 }
1110 
1111 /* --------------------------------------------------------------------- */
1112