xref: /netbsd-src/sys/fs/nilfs/nilfs_vfsops.c (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
1 /* $NetBSD: nilfs_vfsops.c,v 1.5 2010/08/11 13:26:25 pgoyette 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 1.5 2010/08/11 13:26:25 pgoyette 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 	struct partinfo dpart;
567 	struct nilfs_device *nilfsdev;
568 	struct lwp *l = curlwp;
569 	int openflags, accessmode, error;
570 
571 	DPRINTF(VOLUMES, ("Mounting NILFS device\n"));
572 
573 	/* lookup device in our nilfs_mountpoints */
574 	*nilfsdev_p = NULL;
575 	SLIST_FOREACH(nilfsdev, &nilfs_devices, next_device)
576 		if (nilfsdev->devvp == devvp)
577 			break;
578 
579 	if (nilfsdev) {
580 		DPRINTF(VOLUMES, ("device already mounted\n"));
581 		error = nilfs_check_mounts(nilfsdev, mp, args);
582 		if (error)
583 			return error;
584 		nilfsdev->refcnt++;
585 		*nilfsdev_p = nilfsdev;
586 		return 0;
587 	}
588 
589 	DPRINTF(VOLUMES, ("no previous mounts on this device, mounting device\n"));
590 
591 	/* check if its a block device specified */
592 	if (devvp->v_type != VBLK) {
593 		vrele(devvp);
594 		return ENOTBLK;
595 	}
596 	if (bdevsw_lookup(devvp->v_rdev) == NULL) {
597 		vrele(devvp);
598 		return ENXIO;
599 	}
600 
601 	/*
602 	 * If mount by non-root, then verify that user has necessary
603 	 * permissions on the device.
604 	 */
605 	accessmode = VREAD;
606 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
607 		accessmode |= VWRITE;
608 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
609 	error = genfs_can_mount(devvp, accessmode, l->l_cred);
610 	VOP_UNLOCK(devvp);
611 	if (error) {
612 		vrele(devvp);
613 		return error;
614 	}
615 
616 	/*
617 	 * Open device read-write; TODO how about upgrading later when needed?
618 	 */
619 	openflags = FREAD | FWRITE;
620 	error = VOP_OPEN(devvp, openflags, FSCRED);
621 	if (error) {
622 		vrele(devvp);
623 		return error;
624 	}
625 
626 	/* opened ok, try mounting */
627 	nilfsdev = malloc(sizeof(struct nilfs_device), M_NILFSMNT, M_WAITOK);
628 	KASSERT(nilfsdev);
629 
630 	/* initialise */
631 	memset(nilfsdev, 0, sizeof(struct nilfs_device));
632 	nilfsdev->refcnt        = 1;
633 	nilfsdev->devvp         = devvp;
634 	nilfsdev->uncomitted_bl = 0;
635 	cv_init(&nilfsdev->sync_cv, "nilfssyn");
636 	STAILQ_INIT(&nilfsdev->mounts);
637 
638 	/* get our device's size */
639 	error = VOP_IOCTL(devvp, DIOCGPART, &dpart, FREAD, NOCRED);
640 	if (error) {
641 		/* remove all our information */
642 		nilfs_unmount_device(nilfsdev);
643 		return EINVAL;
644 	}
645 	nilfsdev->devsize = dpart.part->p_size * dpart.disklab->d_secsize;
646 
647 	/* register nilfs_device in list */
648 	SLIST_INSERT_HEAD(&nilfs_devices, nilfsdev, next_device);
649 
650 	/* connect to the head for most recent files XXX really pass mp and args? */
651 	error = nilfs_mount_base(nilfsdev, mp, args);
652 	if (error) {
653 		/* remove all our information */
654 		nilfs_unmount_device(nilfsdev);
655 		return EINVAL;
656 	}
657 
658 	*nilfsdev_p = nilfsdev;
659 	DPRINTF(VOLUMES, ("NILFS device mounted ok\n"));
660 
661 	return 0;
662 }
663 
664 
665 static int
666 nilfs_mount_checkpoint(struct nilfs_mount *ump)
667 {
668 	struct nilfs_cpfile_header *cphdr;
669 	struct nilfs_checkpoint *cp;
670 	struct nilfs_inode  ifile_inode;
671 	struct nilfs_node  *cp_node;
672 	struct buf *bp;
673 	uint64_t ncp, nsn, fcpno, blocknr, last_cno;
674 	uint32_t off, dlen;
675 	int cp_per_block, error;
676 
677 	DPRINTF(VOLUMES, ("mount_nilfs: trying to mount checkpoint number "
678 		"%"PRIu64"\n", ump->mount_args.cpno));
679 
680 	cp_node = ump->nilfsdev->cp_node;
681 
682 	/* get cpfile header from 1st block of cp file */
683 	error = nilfs_bread(cp_node, 0, NOCRED, 0, &bp);
684 	if (error)
685 		return error;
686 	cphdr = (struct nilfs_cpfile_header *) bp->b_data;
687 	ncp = nilfs_rw64(cphdr->ch_ncheckpoints);
688 	nsn = nilfs_rw64(cphdr->ch_nsnapshots);
689 
690 	brelse(bp, BC_AGE);
691 
692 	DPRINTF(VOLUMES, ("mount_nilfs: checkpoint header read in\n"));
693 	DPRINTF(VOLUMES, ("\tNumber of checkpoints %"PRIu64"\n", ncp));
694 	DPRINTF(VOLUMES, ("\tNumber of snapshots   %"PRIu64"\n", nsn));
695 
696 	/* read in our specified checkpoint */
697 	dlen = nilfs_rw16(ump->nilfsdev->super.s_checkpoint_size);
698 	cp_per_block = ump->nilfsdev->blocksize / dlen;
699 
700 	fcpno = ump->mount_args.cpno + NILFS_CPFILE_FIRST_CHECKPOINT_OFFSET -1;
701 	blocknr =  fcpno / cp_per_block;
702 	off     = (fcpno % cp_per_block) * dlen;
703 
704 	error = nilfs_bread(cp_node, blocknr, NOCRED, 0, &bp);
705 	if (error) {
706 		printf("mount_nilfs: couldn't read cp block %"PRIu64"\n",
707 			fcpno);
708 		return EINVAL;
709 	}
710 
711 	/* needs to be a valid checkpoint */
712 	cp = (struct nilfs_checkpoint *) ((uint8_t *) bp->b_data + off);
713 	if (cp->cp_flags & NILFS_CHECKPOINT_INVALID) {
714 		printf("mount_nilfs: checkpoint marked invalid\n");
715 		brelse(bp, BC_AGE);
716 		return EINVAL;
717 	}
718 
719 	/* is this really the checkpoint we want? */
720 	if (nilfs_rw64(cp->cp_cno) != ump->mount_args.cpno) {
721 		printf("mount_nilfs: checkpoint file corrupt? "
722 			"expected cpno %"PRIu64", found cpno %"PRIu64"\n",
723 			ump->mount_args.cpno, nilfs_rw64(cp->cp_cno));
724 		brelse(bp, BC_AGE);
725 		return EINVAL;
726 	}
727 
728 	/* check if its a snapshot ! */
729 	last_cno = nilfs_rw64(ump->nilfsdev->super.s_last_cno);
730 	if (ump->mount_args.cpno != last_cno) {
731 		/* only allow snapshots if not mounting on the last cp */
732 		if ((cp->cp_flags & NILFS_CHECKPOINT_SNAPSHOT) == 0) {
733 			printf( "mount_nilfs: checkpoint %"PRIu64" is not a "
734 				"snapshot\n", ump->mount_args.cpno);
735 			brelse(bp, BC_AGE);
736 			return EINVAL;
737 		}
738 	}
739 
740 	ifile_inode = cp->cp_ifile_inode;
741 	brelse(bp, BC_AGE);
742 
743 	/* get ifile inode */
744 	error = nilfs_get_node_raw(ump->nilfsdev, NULL, NILFS_IFILE_INO,
745 		&ifile_inode, &ump->ifile_node);
746 	if (error) {
747 		printf("mount_nilfs: can't read ifile node\n");
748 		return EINVAL;
749 	}
750 	NILFS_SET_SYSTEMFILE(ump->ifile_node->vnode);
751 
752 	/* get root node? */
753 
754 	return 0;
755 }
756 
757 
758 static int
759 nilfs_stop_writing(struct nilfs_mount *ump)
760 {
761 	/* readonly mounts won't write */
762 	if (ump->vfs_mountp->mnt_flag & MNT_RDONLY)
763 		return 0;
764 
765 	DPRINTF(CALL, ("nilfs_stop_writing called for RW mount\n"));
766 
767 	/* TODO writeout super_block? */
768 	/* XXX no support for writing yet anyway */
769 	return 0;
770 }
771 
772 
773 /* --------------------------------------------------------------------- */
774 
775 
776 
777 #define MPFREE(a, lst) \
778 	if ((a)) free((a), lst);
779 static void
780 free_nilfs_mountinfo(struct mount *mp)
781 {
782 	struct nilfs_mount *ump = VFSTONILFS(mp);
783 
784 	if (ump == NULL)
785 		return;
786 
787 	mutex_destroy(&ump->ihash_lock);
788 	mutex_destroy(&ump->get_node_lock);
789 	MPFREE(ump, M_NILFSMNT);
790 }
791 #undef MPFREE
792 
793 int
794 nilfs_mount(struct mount *mp, const char *path,
795 	  void *data, size_t *data_len)
796 {
797 	struct nilfs_args   *args = data;
798 	struct nilfs_device *nilfsdev;
799 	struct nilfs_mount  *ump;
800 	struct vnode *devvp;
801 	int lst, error;
802 
803 	DPRINTF(VFSCALL, ("nilfs_mount called\n"));
804 
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 	devvp->v_specmountpoint = NULL;
898 #if 0
899 	if (devvp->v_specmountpoint == NULL)
900 		devvp->v_specmountpoint = mp;
901 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
902 		devvp->v_specmountpoint = 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