xref: /netbsd/sys/fs/nilfs/nilfs_vfsops.c (revision 3efe214b)
1 /* $NetBSD: nilfs_vfsops.c,v 1.27 2021/09/16 22:19:11 andvar 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 1.27 2021/09/16 22:19:11 andvar 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 /* verbose levels of the nilfs filingsystem */
74 int nilfs_verbose = NILFS_DEBUGGING;
75 
76 /* malloc regions */
77 MALLOC_JUSTDEFINE(M_NILFSMNT,   "NILFS mount",	"NILFS mount structures");
78 MALLOC_JUSTDEFINE(M_NILFSTEMP,  "NILFS temp",	"NILFS scrap space");
79 struct pool nilfs_node_pool;
80 
81 /* globals */
82 struct _nilfs_devices nilfs_devices;
83 
84 /* supported functions predefined */
85 VFS_PROTOS(nilfs);
86 
87 
88 /* --------------------------------------------------------------------- */
89 
90 /*
91  * Genfs interfacing
92  *
93  * static const struct genfs_ops nilfs_genfsops = {
94  * 	.gop_size = genfs_size,
95  * 		size of transfers
96  * 	.gop_alloc = nilfs_gop_alloc,
97  * 		allocate len bytes at offset
98  * 	.gop_write = genfs_gop_write,
99  * 		putpages interface code
100  * 	.gop_markupdate = nilfs_gop_markupdate,
101  * 		set update/modify flags etc.
102  * }
103  */
104 
105 /*
106  * Callback from genfs to allocate len bytes at offset off; only called when
107  * filling up gaps in the allocation.
108  */
109 static int
nilfs_gop_alloc(struct vnode * vp,off_t off,off_t len,int flags,kauth_cred_t cred)110 nilfs_gop_alloc(struct vnode *vp, off_t off,
111     off_t len, int flags, kauth_cred_t cred)
112 {
113 	DPRINTF(NOTIMPL, ("nilfs_gop_alloc not implemented\n"));
114 	DPRINTF(ALLOC, ("nilfs_gop_alloc called for %"PRIu64" bytes\n", len));
115 
116 	return 0;
117 }
118 
119 
120 /*
121  * callback from genfs to update our flags
122  */
123 static void
nilfs_gop_markupdate(struct vnode * vp,int flags)124 nilfs_gop_markupdate(struct vnode *vp, int flags)
125 {
126 	struct nilfs_node *nilfs_node = VTOI(vp);
127 	u_long mask = 0;
128 
129 	if ((flags & GOP_UPDATE_ACCESSED) != 0) {
130 		mask = IN_ACCESS;
131 	}
132 	if ((flags & GOP_UPDATE_MODIFIED) != 0) {
133 		if (vp->v_type == VREG) {
134 			mask |= IN_CHANGE | IN_UPDATE;
135 		} else {
136 			mask |= IN_MODIFY;
137 		}
138 	}
139 	if (mask) {
140 		nilfs_node->i_flags |= mask;
141 	}
142 }
143 
144 
145 static const struct genfs_ops nilfs_genfsops = {
146 	.gop_size = genfs_size,
147 	.gop_alloc = nilfs_gop_alloc,
148 	.gop_write = genfs_gop_write_rwmap,
149 	.gop_markupdate = nilfs_gop_markupdate,
150 	.gop_putrange = genfs_gop_putrange,
151 };
152 
153 /* --------------------------------------------------------------------- */
154 
155 /* predefine vnode-op list descriptor */
156 extern const struct vnodeopv_desc nilfs_vnodeop_opv_desc;
157 
158 const struct vnodeopv_desc * const nilfs_vnodeopv_descs[] = {
159 	&nilfs_vnodeop_opv_desc,
160 	NULL,
161 };
162 
163 
164 /* vfsops descriptor linked in as anchor point for the filingsystem */
165 struct vfsops nilfs_vfsops = {
166 	.vfs_name = MOUNT_NILFS,
167 	.vfs_min_mount_data = sizeof (struct nilfs_args),
168 	.vfs_mount = nilfs_mount,
169 	.vfs_start = nilfs_start,
170 	.vfs_unmount = nilfs_unmount,
171 	.vfs_root = nilfs_root,
172 	.vfs_quotactl = (void *)eopnotsupp,
173 	.vfs_statvfs = nilfs_statvfs,
174 	.vfs_sync = nilfs_sync,
175 	.vfs_vget = nilfs_vget,
176 	.vfs_loadvnode = nilfs_loadvnode,
177 	.vfs_fhtovp = nilfs_fhtovp,
178 	.vfs_vptofh = nilfs_vptofh,
179 	.vfs_init = nilfs_init,
180 	.vfs_reinit = nilfs_reinit,
181 	.vfs_done = nilfs_done,
182 	.vfs_mountroot = nilfs_mountroot,
183 	.vfs_snapshot = nilfs_snapshot,
184 	.vfs_extattrctl = vfs_stdextattrctl,
185 	.vfs_suspendctl = genfs_suspendctl,
186 	.vfs_renamelock_enter = genfs_renamelock_enter,
187 	.vfs_renamelock_exit = genfs_renamelock_exit,
188 	.vfs_fsync = (void *)eopnotsupp,
189 	.vfs_opv_descs = nilfs_vnodeopv_descs
190 };
191 
192 /* --------------------------------------------------------------------- */
193 
194 /* file system starts here */
195 void
nilfs_init(void)196 nilfs_init(void)
197 {
198 	size_t size;
199 
200 	/* setup memory types */
201 	malloc_type_attach(M_NILFSMNT);
202 	malloc_type_attach(M_NILFSTEMP);
203 
204 	/* init device lists */
205 	SLIST_INIT(&nilfs_devices);
206 
207 	/* init node pools */
208 	size = sizeof(struct nilfs_node);
209 	pool_init(&nilfs_node_pool, size, 0, 0, 0,
210 		"nilfs_node_pool", NULL, IPL_NONE);
211 }
212 
213 
214 void
nilfs_reinit(void)215 nilfs_reinit(void)
216 {
217 	/* nothing to do */
218 }
219 
220 
221 void
nilfs_done(void)222 nilfs_done(void)
223 {
224 	/* remove pools */
225 	pool_destroy(&nilfs_node_pool);
226 
227 	malloc_type_detach(M_NILFSMNT);
228 	malloc_type_detach(M_NILFSTEMP);
229 }
230 
231 /*
232  * If running a DEBUG kernel, provide an easy way to set the debug flags when
233  * running into a problem.
234  */
235 #define NILFS_VERBOSE_SYSCTLOPT        1
236 
237 SYSCTL_SETUP(nilfs_sysctl_setup, "nilfs sysctl")
238 {
239 	const struct sysctlnode *node;
240 
241 	/*
242 	 * XXX the "30" below could be dynamic, thereby eliminating one
243 	 * more instance of the "number to vfs" mapping problem, but
244 	 * "30" is the order as taken from sys/mount.h
245 	 */
246 	sysctl_createv(clog, 0, NULL, &node,
247 		       CTLFLAG_PERMANENT,
248 		       CTLTYPE_NODE, "nilfs",
249 		       SYSCTL_DESCR("NTT's NILFSv2"),
250 		       NULL, 0, NULL, 0,
251 		       CTL_VFS, 30, CTL_EOL);
252 #ifdef DEBUG
253 	sysctl_createv(clog, 0, NULL, &node,
254 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
255 		       CTLTYPE_INT, "verbose",
256 		       SYSCTL_DESCR("Bitmask for filesystem debugging"),
257 		       NULL, 0, &nilfs_verbose, 0,
258 		       CTL_VFS, 30, NILFS_VERBOSE_SYSCTLOPT, CTL_EOL);
259 #endif
260 }
261 
262 static int
nilfs_modcmd(modcmd_t cmd,void * arg)263 nilfs_modcmd(modcmd_t cmd, void *arg)
264 {
265 	int error;
266 
267 	switch (cmd) {
268 	case MODULE_CMD_INIT:
269 		error = vfs_attach(&nilfs_vfsops);
270 		if (error != 0)
271 			break;
272 		break;
273 	case MODULE_CMD_FINI:
274 		error = vfs_detach(&nilfs_vfsops);
275 		if (error != 0)
276 			break;
277 		break;
278 	default:
279 		error = ENOTTY;
280 		break;
281 	}
282 
283 	return (error);
284 }
285 
286 /* --------------------------------------------------------------------- */
287 
288 int
nilfs_mountroot(void)289 nilfs_mountroot(void)
290 {
291 	return EOPNOTSUPP;
292 }
293 
294 /* --------------------------------------------------------------------- */
295 
296 /* system nodes */
297 static int
nilfs_create_system_nodes(struct nilfs_device * nilfsdev)298 nilfs_create_system_nodes(struct nilfs_device *nilfsdev)
299 {
300 	int error;
301 
302 	error = nilfs_get_node_raw(nilfsdev, NULL, NILFS_DAT_INO,
303 		&nilfsdev->super_root.sr_dat, &nilfsdev->dat_node);
304 	if (error)
305 		goto errorout;
306 
307 	error = nilfs_get_node_raw(nilfsdev, NULL, NILFS_CPFILE_INO,
308 		&nilfsdev->super_root.sr_cpfile, &nilfsdev->cp_node);
309 	if (error)
310 		goto errorout;
311 
312 	error = nilfs_get_node_raw(nilfsdev, NULL, NILFS_SUFILE_INO,
313 		&nilfsdev->super_root.sr_sufile, &nilfsdev->su_node);
314 	if (error)
315 		goto errorout;
316 
317 	return 0;
318 errorout:
319 	nilfs_dispose_node(&nilfsdev->dat_node);
320 	nilfs_dispose_node(&nilfsdev->cp_node);
321 	nilfs_dispose_node(&nilfsdev->su_node);
322 
323 	return error;
324 }
325 
326 
327 static void
nilfs_release_system_nodes(struct nilfs_device * nilfsdev)328 nilfs_release_system_nodes(struct nilfs_device *nilfsdev)
329 {
330 	if (!nilfsdev)
331 		return;
332 	if (nilfsdev->refcnt > 0)
333 		return;
334 
335 	nilfs_dispose_node(&nilfsdev->dat_node);
336 	nilfs_dispose_node(&nilfsdev->cp_node);
337 	nilfs_dispose_node(&nilfsdev->su_node);
338 }
339 
340 
341 /* --------------------------------------------------------------------- */
342 
343 static int
nilfs_check_superblock_crc(struct nilfs_super_block * super)344 nilfs_check_superblock_crc(struct nilfs_super_block *super)
345 {
346 	uint32_t super_crc, comp_crc;
347 
348 	/* check super block magic */
349 	if (nilfs_rw16(super->s_magic) != NILFS_SUPER_MAGIC)
350 		return 0;
351 
352 	/* preserve crc */
353 	super_crc  = nilfs_rw32(super->s_sum);
354 
355 	/* calculate */
356 	super->s_sum = 0;
357 	comp_crc = crc32_le(nilfs_rw32(super->s_crc_seed),
358 		(uint8_t *) super, nilfs_rw16(super->s_bytes));
359 
360 	/* restore */
361 	super->s_sum = nilfs_rw32(super_crc);
362 
363 	/* check CRC */
364 	return (super_crc == comp_crc);
365 }
366 
367 
368 
369 static int
nilfs_read_superblock(struct nilfs_device * nilfsdev)370 nilfs_read_superblock(struct nilfs_device *nilfsdev)
371 {
372 	struct nilfs_super_block *super, tmp_super;
373 	struct buf *bp;
374 	uint64_t sb1off, sb2off;
375 	uint64_t last_cno1, last_cno2;
376 	uint64_t dev_blk;
377 	int dev_bsize, dev_blks;
378 	int sb1ok, sb2ok, swp;
379 	int error;
380 
381 	sb1off = NILFS_SB_OFFSET_BYTES;
382 	sb2off = NILFS_SB2_OFFSET_BYTES(nilfsdev->devsize);
383 
384 	dev_bsize = 1 << nilfsdev->devvp->v_mount->mnt_fs_bshift;
385 
386 	/* read our superblock regardless of backing device blocksize */
387 	dev_blk   = 0;
388 	dev_blks  = (sb1off + dev_bsize -1)/dev_bsize;
389 	error = bread(nilfsdev->devvp, dev_blk, dev_blks * dev_bsize, 0, &bp);
390 	if (error) {
391 		return error;
392 	}
393 
394 	/* copy read-in super block at the offset */
395 	super = &nilfsdev->super;
396 	memcpy(super, (uint8_t *) bp->b_data + NILFS_SB_OFFSET_BYTES,
397 		sizeof(struct nilfs_super_block));
398 	brelse(bp, BC_AGE);
399 
400 	/* read our 2nd superblock regardless of backing device blocksize */
401 	dev_blk   = sb2off / dev_bsize;
402 	dev_blks  = 2;		/* assumption max one dev_bsize */
403 	error = bread(nilfsdev->devvp, dev_blk, dev_blks * dev_bsize, 0, &bp);
404 	if (error) {
405 		return error;
406 	}
407 
408 	/* copy read-in superblock2 at the offset */
409 	super = &nilfsdev->super2;
410 	memcpy(super, (uint8_t *) bp->b_data + NILFS_SB_OFFSET_BYTES,
411 		sizeof(struct nilfs_super_block));
412 	brelse(bp, BC_AGE);
413 
414 	sb1ok = nilfs_check_superblock_crc(&nilfsdev->super);
415 	sb2ok = nilfs_check_superblock_crc(&nilfsdev->super2);
416 
417 	last_cno1 = nilfs_rw64(nilfsdev->super.s_last_cno);
418 	last_cno2 = nilfs_rw64(nilfsdev->super2.s_last_cno);
419 	swp = sb2ok && (last_cno2 > last_cno1);
420 
421 	if (swp) {
422 		printf("nilfs warning: broken superblock, using spare\n");
423 		tmp_super = nilfsdev->super2;
424 		nilfsdev->super2 = nilfsdev->super;	/* why preserve? */
425 		nilfsdev->super  = tmp_super;
426 	}
427 
428 	if (!sb1ok && !sb2ok) {
429 		printf("nilfs: no valid superblocks found\n");
430 		return EINVAL;
431 	}
432 
433 	return 0;
434 }
435 
436 
437 /* XXX NOTHING from the system nodes should need to be written here */
438 static void
nilfs_unmount_base(struct nilfs_device * nilfsdev)439 nilfs_unmount_base(struct nilfs_device *nilfsdev)
440 {
441 	int error __diagused;
442 
443 	if (!nilfsdev)
444 		return;
445 
446 	/* remove all our information */
447 	error = vinvalbuf(nilfsdev->devvp, 0, FSCRED, curlwp, 0, 0);
448 	KASSERT(error == 0);
449 
450 	/* release the device's system nodes */
451 	nilfs_release_system_nodes(nilfsdev);
452 
453 	/* TODO writeout super_block? */
454 }
455 
456 
457 static int
nilfs_mount_base(struct nilfs_device * nilfsdev,struct mount * mp,struct nilfs_args * args)458 nilfs_mount_base(struct nilfs_device *nilfsdev,
459 		struct mount *mp, struct nilfs_args *args)
460 {
461 	struct lwp *l = curlwp;
462 	uint64_t last_pseg, last_cno, last_seq;
463 	uint32_t log_blocksize;
464 	int error;
465 
466 	/* flush out any old buffers remaining from a previous use. */
467 	if ((error = vinvalbuf(nilfsdev->devvp, V_SAVE, l->l_cred, l, 0, 0)))
468 		return error;
469 
470 	/* read in our superblock */
471 	error = nilfs_read_superblock(nilfsdev);
472 	if (error) {
473 		printf("nilfs_mount: can't read in super block : %d\n", error);
474 		return error;
475 	}
476 
477 	/* get our blocksize */
478 	log_blocksize = nilfs_rw32(nilfsdev->super.s_log_block_size);
479 	nilfsdev->blocksize   = (uint64_t) 1 << (log_blocksize + 10);
480 	/* TODO check superblock's blocksize limits */
481 
482 	/* calculate dat structure parameters */
483 	nilfs_calc_mdt_consts(nilfsdev, &nilfsdev->dat_mdt,
484 			nilfs_rw16(nilfsdev->super.s_dat_entry_size));
485 	nilfs_calc_mdt_consts(nilfsdev, &nilfsdev->ifile_mdt,
486 			nilfs_rw16(nilfsdev->super.s_inode_size));
487 
488 	DPRINTF(VOLUMES, ("nilfs_mount: accepted super block\n"));
489 
490 	/* search for the super root and roll forward when needed */
491 	nilfs_search_super_root(nilfsdev);
492 
493 	nilfsdev->mount_state = nilfs_rw16(nilfsdev->super.s_state);
494 	if (nilfsdev->mount_state != NILFS_VALID_FS) {
495 		printf("FS is seriously damaged, needs repairing\n");
496 		printf("aborting mount\n");
497 		return EINVAL;
498 	}
499 
500 	/*
501 	 * FS should be ok now. The superblock and the last segsum could be
502 	 * updated from the repair so extract running values again.
503 	 */
504 	last_pseg = nilfs_rw64(nilfsdev->super.s_last_pseg); /*blknr */
505 	last_cno  = nilfs_rw64(nilfsdev->super.s_last_cno);
506 	last_seq  = nilfs_rw64(nilfsdev->super.s_last_seq);
507 
508 	nilfsdev->last_seg_seq = last_seq;
509 	nilfsdev->last_seg_num = nilfs_get_segnum_of_block(nilfsdev, last_pseg);
510 	nilfsdev->next_seg_num = nilfs_get_segnum_of_block(nilfsdev,
511 		nilfs_rw64(nilfsdev->last_segsum.ss_next));
512 	nilfsdev->last_cno     = last_cno;
513 
514 	DPRINTF(VOLUMES, ("nilfs_mount: accepted super root\n"));
515 
516 	/* create system vnodes for DAT, CP and SEGSUM */
517 	error = nilfs_create_system_nodes(nilfsdev);
518 	if (error)
519 		nilfs_unmount_base(nilfsdev);
520 	return error;
521 }
522 
523 
524 static void
nilfs_unmount_device(struct nilfs_device * nilfsdev)525 nilfs_unmount_device(struct nilfs_device *nilfsdev)
526 {
527 	int error;
528 
529 	/* is there anything? */
530 	if (nilfsdev == NULL)
531 		return;
532 
533 	/* remove the device only if we're the last reference */
534 	nilfsdev->refcnt--;
535 	if (nilfsdev->refcnt >= 1)
536 		return;
537 
538 	/* unmount our base */
539 	nilfs_unmount_base(nilfsdev);
540 
541 	/* remove from our device list */
542 	SLIST_REMOVE(&nilfs_devices, nilfsdev, nilfs_device, next_device);
543 
544 	/* close device */
545 	DPRINTF(VOLUMES, ("closing device\n"));
546 
547 	/* remove our mount reference before closing device */
548 	spec_node_setmountedfs(nilfsdev->devvp, NULL);
549 
550 	/* devvp is still locked by us */
551 	vn_lock(nilfsdev->devvp, LK_EXCLUSIVE | LK_RETRY);
552 	error = VOP_CLOSE(nilfsdev->devvp, FREAD | FWRITE, NOCRED);
553 	if (error)
554 		printf("Error during closure of device! error %d, "
555 		       "device might stay locked\n", error);
556 	DPRINTF(VOLUMES, ("device close ok\n"));
557 
558 	/* clear our mount reference and release device node */
559 	vput(nilfsdev->devvp);
560 
561 	/* free our device info */
562 	cv_destroy(&nilfsdev->sync_cv);
563 	free(nilfsdev, M_NILFSMNT);
564 }
565 
566 
567 static int
nilfs_check_mounts(struct nilfs_device * nilfsdev,struct mount * mp,struct nilfs_args * args)568 nilfs_check_mounts(struct nilfs_device *nilfsdev, struct mount *mp,
569 	struct nilfs_args *args)
570 {
571 	struct nilfs_mount  *ump;
572 	uint64_t last_cno;
573 
574 	/* no double-mounting of the same checkpoint */
575 	STAILQ_FOREACH(ump, &nilfsdev->mounts, next_mount) {
576 		if (ump->mount_args.cpno == args->cpno)
577 			return EBUSY;
578 	}
579 
580 	/* allow readonly mounts without questioning here */
581 	if (mp->mnt_flag & MNT_RDONLY)
582 		return 0;
583 
584 	/* readwrite mount you want */
585 	STAILQ_FOREACH(ump, &nilfsdev->mounts, next_mount) {
586 		/* only one RW mount on this device! */
587 		if ((ump->vfs_mountp->mnt_flag & MNT_RDONLY)==0)
588 			return EROFS;
589 		/* RDONLY on last mountpoint is device busy */
590 		last_cno = nilfs_rw64(ump->nilfsdev->super.s_last_cno);
591 		if (ump->mount_args.cpno == last_cno)
592 			return EBUSY;
593 	}
594 
595 	/* OK for now */
596 	return 0;
597 }
598 
599 
600 static int
nilfs_mount_device(struct vnode * devvp,struct mount * mp,struct nilfs_args * args,struct nilfs_device ** nilfsdev_p)601 nilfs_mount_device(struct vnode *devvp, struct mount *mp, struct nilfs_args *args,
602 	struct nilfs_device **nilfsdev_p)
603 {
604 	uint64_t psize;
605 	unsigned secsize;
606 	struct nilfs_device *nilfsdev;
607 	struct lwp *l = curlwp;
608 	int openflags, accessmode, error;
609 
610 	DPRINTF(VOLUMES, ("Mounting NILFS device\n"));
611 
612 	/* lookup device in our nilfs_mountpoints */
613 	*nilfsdev_p = NULL;
614 	SLIST_FOREACH(nilfsdev, &nilfs_devices, next_device)
615 		if (nilfsdev->devvp == devvp)
616 			break;
617 
618 	if (nilfsdev) {
619 		DPRINTF(VOLUMES, ("device already mounted\n"));
620 		error = nilfs_check_mounts(nilfsdev, mp, args);
621 		if (error)
622 			return error;
623 		nilfsdev->refcnt++;
624 		*nilfsdev_p = nilfsdev;
625 		return 0;
626 	}
627 
628 	DPRINTF(VOLUMES, ("no previous mounts on this device, mounting device\n"));
629 
630 	/* check if its a block device specified */
631 	if (devvp->v_type != VBLK) {
632 		vrele(devvp);
633 		return ENOTBLK;
634 	}
635 	if (bdevsw_lookup(devvp->v_rdev) == NULL) {
636 		vrele(devvp);
637 		return ENXIO;
638 	}
639 
640 	/*
641 	 * If mount by non-root, then verify that user has necessary
642 	 * permissions on the device.
643 	 */
644 	accessmode = VREAD;
645 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
646 		accessmode |= VWRITE;
647 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
648 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
649 	    KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, KAUTH_ARG(accessmode));
650 	VOP_UNLOCK(devvp);
651 	if (error) {
652 		vrele(devvp);
653 		return error;
654 	}
655 
656 	/*
657 	 * Open device read-write; TODO how about upgrading later when needed?
658 	 */
659 	openflags = FREAD | FWRITE;
660 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
661 	error = VOP_OPEN(devvp, openflags, FSCRED);
662 	VOP_UNLOCK(devvp);
663 	if (error) {
664 		vrele(devvp);
665 		return error;
666 	}
667 
668 	/* opened ok, try mounting */
669 	nilfsdev = malloc(sizeof(*nilfsdev), M_NILFSMNT, M_WAITOK | M_ZERO);
670 
671 	/* initialise */
672 	nilfsdev->refcnt        = 1;
673 	nilfsdev->devvp         = devvp;
674 	nilfsdev->uncomitted_bl = 0;
675 	cv_init(&nilfsdev->sync_cv, "nilfssyn");
676 	STAILQ_INIT(&nilfsdev->mounts);
677 
678 	/* register nilfs_device in list */
679 	SLIST_INSERT_HEAD(&nilfs_devices, nilfsdev, next_device);
680 
681 	/* get our device's size */
682 	error = getdisksize(devvp, &psize, &secsize);
683 	if (error) {
684 		/* remove all our information */
685 		nilfs_unmount_device(nilfsdev);
686 		return EINVAL;
687 	}
688 
689 	nilfsdev->devsize = psize * secsize;
690 
691 	/* connect to the head for most recent files XXX really pass mp and args? */
692 	error = nilfs_mount_base(nilfsdev, mp, args);
693 	if (error) {
694 		/* remove all our information */
695 		nilfs_unmount_device(nilfsdev);
696 		return EINVAL;
697 	}
698 
699 	*nilfsdev_p = nilfsdev;
700 	DPRINTF(VOLUMES, ("NILFS device mounted ok\n"));
701 
702 	return 0;
703 }
704 
705 
706 static int
nilfs_mount_checkpoint(struct nilfs_mount * ump)707 nilfs_mount_checkpoint(struct nilfs_mount *ump)
708 {
709 	struct nilfs_cpfile_header *cphdr;
710 	struct nilfs_checkpoint *cp;
711 	struct nilfs_inode  ifile_inode;
712 	struct nilfs_node  *cp_node;
713 	struct buf *bp;
714 	uint64_t ncp, nsn, fcpno, blocknr, last_cno;
715 	uint32_t off, dlen;
716 	int cp_per_block, error;
717 
718 	DPRINTF(VOLUMES, ("mount_nilfs: trying to mount checkpoint number "
719 		"%"PRIu64"\n", ump->mount_args.cpno));
720 
721 	cp_node = ump->nilfsdev->cp_node;
722 
723 	/* get cpfile header from 1st block of cp file */
724 	error = nilfs_bread(cp_node, 0, 0, &bp);
725 	if (error)
726 		return error;
727 	cphdr = (struct nilfs_cpfile_header *) bp->b_data;
728 	ncp = nilfs_rw64(cphdr->ch_ncheckpoints);
729 	nsn = nilfs_rw64(cphdr->ch_nsnapshots);
730 
731 	brelse(bp, BC_AGE);
732 
733 	DPRINTF(VOLUMES, ("mount_nilfs: checkpoint header read in\n"));
734 	DPRINTF(VOLUMES, ("\tNumber of checkpoints %"PRIu64"\n", ncp));
735 	DPRINTF(VOLUMES, ("\tNumber of snapshots   %"PRIu64"\n", nsn));
736 #ifndef NILFS_DEBUG
737 	__USE(ncp);
738 	__USE(nsn);
739 #endif
740 
741 	/* read in our specified checkpoint */
742 	dlen = nilfs_rw16(ump->nilfsdev->super.s_checkpoint_size);
743 	cp_per_block = ump->nilfsdev->blocksize / dlen;
744 
745 	fcpno = ump->mount_args.cpno + NILFS_CPFILE_FIRST_CHECKPOINT_OFFSET -1;
746 	blocknr =  fcpno / cp_per_block;
747 	off     = (fcpno % cp_per_block) * dlen;
748 
749 	error = nilfs_bread(cp_node, blocknr, 0, &bp);
750 	if (error) {
751 		printf("mount_nilfs: couldn't read cp block %"PRIu64"\n",
752 			fcpno);
753 		return EINVAL;
754 	}
755 
756 	/* needs to be a valid checkpoint */
757 	cp = (struct nilfs_checkpoint *) ((uint8_t *) bp->b_data + off);
758 	if (cp->cp_flags & NILFS_CHECKPOINT_INVALID) {
759 		printf("mount_nilfs: checkpoint marked invalid\n");
760 		brelse(bp, BC_AGE);
761 		return EINVAL;
762 	}
763 
764 	/* is this really the checkpoint we want? */
765 	if (nilfs_rw64(cp->cp_cno) != ump->mount_args.cpno) {
766 		printf("mount_nilfs: checkpoint file corrupt? "
767 			"expected cpno %"PRIu64", found cpno %"PRIu64"\n",
768 			ump->mount_args.cpno, nilfs_rw64(cp->cp_cno));
769 		brelse(bp, BC_AGE);
770 		return EINVAL;
771 	}
772 
773 	/* check if its a snapshot ! */
774 	last_cno = nilfs_rw64(ump->nilfsdev->super.s_last_cno);
775 	if (ump->mount_args.cpno != last_cno) {
776 		/* only allow snapshots if not mounting on the last cp */
777 		if ((cp->cp_flags & NILFS_CHECKPOINT_SNAPSHOT) == 0) {
778 			printf( "mount_nilfs: checkpoint %"PRIu64" is not a "
779 				"snapshot\n", ump->mount_args.cpno);
780 			brelse(bp, BC_AGE);
781 			return EINVAL;
782 		}
783 	}
784 
785 	ifile_inode = cp->cp_ifile_inode;
786 	brelse(bp, BC_AGE);
787 
788 	/* get ifile inode */
789 	error = nilfs_get_node_raw(ump->nilfsdev, NULL, NILFS_IFILE_INO,
790 		&ifile_inode, &ump->ifile_node);
791 	if (error) {
792 		printf("mount_nilfs: can't read ifile node\n");
793 		return EINVAL;
794 	}
795 
796 	/* get root node? */
797 
798 	return 0;
799 }
800 
801 
802 static int
nilfs_stop_writing(struct nilfs_mount * ump)803 nilfs_stop_writing(struct nilfs_mount *ump)
804 {
805 	/* readonly mounts won't write */
806 	if (ump->vfs_mountp->mnt_flag & MNT_RDONLY)
807 		return 0;
808 
809 	DPRINTF(CALL, ("nilfs_stop_writing called for RW mount\n"));
810 
811 	/* TODO writeout super_block? */
812 	/* XXX no support for writing yet anyway */
813 	return 0;
814 }
815 
816 
817 /* --------------------------------------------------------------------- */
818 
819 
820 
821 #define MPFREE(a, lst) \
822 	if ((a)) free((a), lst);
823 static void
free_nilfs_mountinfo(struct mount * mp)824 free_nilfs_mountinfo(struct mount *mp)
825 {
826 	struct nilfs_mount *ump = VFSTONILFS(mp);
827 
828 	if (ump == NULL)
829 		return;
830 
831 	MPFREE(ump, M_NILFSMNT);
832 }
833 #undef MPFREE
834 
835 int
nilfs_mount(struct mount * mp,const char * path,void * data,size_t * data_len)836 nilfs_mount(struct mount *mp, const char *path,
837 	  void *data, size_t *data_len)
838 {
839 	struct nilfs_args   *args = data;
840 	struct nilfs_device *nilfsdev;
841 	struct nilfs_mount  *ump;
842 	struct vnode *devvp;
843 	int error;
844 
845 	DPRINTF(VFSCALL, ("nilfs_mount called\n"));
846 
847 	if (args == NULL)
848 		return EINVAL;
849 	if (*data_len < sizeof *args)
850 		return EINVAL;
851 
852 	if (mp->mnt_flag & MNT_GETARGS) {
853 		/* request for the mount arguments */
854 		ump = VFSTONILFS(mp);
855 		if (ump == NULL)
856 			return EINVAL;
857 		*args = ump->mount_args;
858 		*data_len = sizeof *args;
859 		return 0;
860 	}
861 
862 	/* check/translate struct version */
863 	if (args->version != 1) {
864 		printf("mount_nilfs: unrecognized argument structure version\n");
865 		return EINVAL;
866 	}
867 	/* TODO sanity checking other mount arguments */
868 
869 	/* handle request for updating mount parameters */
870 	if (mp->mnt_flag & MNT_UPDATE) {
871 		/* TODO can't update my mountpoint yet */
872 		return EOPNOTSUPP;
873 	}
874 
875 	/* lookup name to get its vnode */
876 	error = namei_simple_user(args->fspec, NSM_FOLLOW_NOEMULROOT, &devvp);
877 	if (error)
878 		return error;
879 
880 #ifdef DEBUG
881 	if (nilfs_verbose & NILFS_DEBUG_VOLUMES)
882 		vprint("NILFS mount, trying to mount \n", devvp);
883 #endif
884 
885 	error = nilfs_mount_device(devvp, mp, args, &nilfsdev);
886 	if (error)
887 		return error;
888 
889 	/*
890 	 * Create a nilfs_mount on the specified checkpoint. Note that only
891 	 * ONE RW mount point can exist and it needs to have the highest
892 	 * checkpoint nr. If mounting RW and its not on the last checkpoint we
893 	 * need to invalidate all checkpoints that follow!!! This is an
894 	 * advanced option.
895 	 */
896 
897 	/* setup basic mountpoint structure */
898 	mp->mnt_data = NULL;
899 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (uint32_t) devvp->v_rdev;
900 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_NILFS);
901 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
902 	mp->mnt_stat.f_namemax = NILFS_NAME_LEN;
903 	mp->mnt_flag  |= MNT_LOCAL;
904 
905 	/* XXX can't enable MPSAFE yet since genfs barfs on bad CV */
906 	// mp->mnt_iflag |= IMNT_MPSAFE;
907 
908 	/* set our dev and fs units */
909 	mp->mnt_dev_bshift = nilfs_rw32(nilfsdev->super.s_log_block_size) + 10;
910 	mp->mnt_fs_bshift  = mp->mnt_dev_bshift;
911 
912 	/* allocate nilfs part of mount structure; malloc always succeeds */
913 	ump = malloc(sizeof(struct nilfs_mount), M_NILFSMNT, M_WAITOK | M_ZERO);
914 
915 	/* set up linkage */
916 	mp->mnt_data    =  ump;
917 	ump->vfs_mountp =  mp;
918 	ump->nilfsdev   =  nilfsdev;
919 
920 #if 0
921 #ifndef NILFS_READWRITE
922 	/* force read-only for now */
923 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
924 		printf( "Enable kernel/module option NILFS_READWRITE for "
925 			"writing, downgrading access to read-only\n");
926 		mp->mnt_flag |= MNT_RDONLY;
927 	}
928 #endif
929 #endif
930 
931 	/* DONT register our nilfs mountpoint on our vfs mountpoint */
932 	spec_node_setmountedfs(devvp, NULL);
933 #if 0
934 	if (spec_node_getmountedfs(devvp) == NULL)
935 		spec_node_setmountedfs(devvp, mp);
936 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
937 		spec_node_setmountedfs(devvp, mp);
938 #endif
939 
940 	/* add our mountpoint */
941 	STAILQ_INSERT_TAIL(&nilfsdev->mounts, ump, next_mount);
942 
943 	/* get our selected checkpoint */
944 	if (args->cpno == 0)
945 		args->cpno = nilfsdev->last_cno;
946 	args->cpno = MIN(args->cpno, nilfsdev->last_cno);
947 
948 	/* setting up other parameters */
949 	ump->mount_args = *args;
950 	error = nilfs_mount_checkpoint(ump);
951 	if (error) {
952 		nilfs_unmount(mp, MNT_FORCE);
953 		return error;
954 	}
955 
956 	/* set VFS info */
957 	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec, UIO_USERSPACE,
958 			mp->mnt_op->vfs_name, mp, curlwp);
959 	if (error) {
960 		nilfs_unmount(mp, MNT_FORCE);
961 		return error;
962 	}
963 
964 	/* successfully mounted */
965 	DPRINTF(VOLUMES, ("nilfs_mount() successful\n"));
966 
967 	return 0;
968 }
969 
970 /* --------------------------------------------------------------------- */
971 
972 
973 /* remove our mountpoint and if its the last reference, remove our device */
974 int
nilfs_unmount(struct mount * mp,int mntflags)975 nilfs_unmount(struct mount *mp, int mntflags)
976 {
977 	struct nilfs_device *nilfsdev;
978 	struct nilfs_mount  *ump;
979 	int error, flags;
980 
981 	DPRINTF(VFSCALL, ("nilfs_umount called\n"));
982 
983 	ump = VFSTONILFS(mp);
984 	if (!ump)
985 		panic("NILFS unmount: empty ump\n");
986 	nilfsdev = ump->nilfsdev;
987 
988 	/*
989 	 * Flush all nodes associated to this mountpoint.
990 	 */
991 	flags = (mntflags & MNT_FORCE) ? FORCECLOSE : 0;
992 	if ((error = vflush(mp, NULLVP, flags)) != 0)
993 		return error;
994 
995 	/* if we're the write mount, we ought to close the writing session */
996 	error = nilfs_stop_writing(ump);
997 	if (error)
998 		return error;
999 
1000 	nilfs_dispose_node(&ump->ifile_node);
1001 
1002 	/* remove our mount point */
1003 	STAILQ_REMOVE(&nilfsdev->mounts, ump, nilfs_mount, next_mount);
1004 	free_nilfs_mountinfo(mp);
1005 
1006 	/* free ump struct references */
1007 	mp->mnt_data = NULL;
1008 	mp->mnt_flag &= ~MNT_LOCAL;
1009 
1010 	/* unmount the device itself when we're the last one */
1011 	nilfs_unmount_device(nilfsdev);
1012 
1013 	DPRINTF(VOLUMES, ("Fin unmount\n"));
1014 	return error;
1015 }
1016 
1017 /* --------------------------------------------------------------------- */
1018 
1019 int
nilfs_start(struct mount * mp,int flags)1020 nilfs_start(struct mount *mp, int flags)
1021 {
1022 	/* do we have to do something here? */
1023 	return 0;
1024 }
1025 
1026 /* --------------------------------------------------------------------- */
1027 
1028 int
nilfs_root(struct mount * mp,int lktype,struct vnode ** vpp)1029 nilfs_root(struct mount *mp, int lktype, struct vnode **vpp)
1030 {
1031 	uint64_t ino = NILFS_ROOT_INO;
1032 	int error;
1033 
1034 	DPRINTF(NODE, ("nilfs_root called\n"));
1035 
1036 	error = vcache_get(mp, &ino, sizeof(ino), vpp);
1037 	if (error == 0) {
1038 		error = vn_lock(*vpp, lktype);
1039 		if (error) {
1040 			vrele(*vpp);
1041 			*vpp = NULL;
1042 			return error;
1043 		}
1044 	}
1045 	KASSERT(error != 0 || ((*vpp)->v_vflag & VV_ROOT));
1046 
1047 	DPRINTF(NODE, ("nilfs_root finished\n"));
1048 	return error;
1049 }
1050 
1051 /* --------------------------------------------------------------------- */
1052 
1053 int
nilfs_statvfs(struct mount * mp,struct statvfs * sbp)1054 nilfs_statvfs(struct mount *mp, struct statvfs *sbp)
1055 {
1056 	struct nilfs_mount *ump = VFSTONILFS(mp);
1057 	uint32_t blocksize;
1058 
1059 	DPRINTF(VFSCALL, ("nilfs_statvfs called\n"));
1060 
1061 	blocksize = ump->nilfsdev->blocksize;
1062 	sbp->f_flag   = mp->mnt_flag;
1063 	sbp->f_bsize  = blocksize;
1064 	sbp->f_frsize = blocksize;
1065 	sbp->f_iosize = blocksize;
1066 
1067 	copy_statvfs_info(sbp, mp);
1068 	return 0;
1069 }
1070 
1071 /* --------------------------------------------------------------------- */
1072 
1073 int
nilfs_sync(struct mount * mp,int waitfor,kauth_cred_t cred)1074 nilfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
1075 {
1076 //	struct nilfs_mount *ump = VFSTONILFS(mp);
1077 
1078 	DPRINTF(VFSCALL, ("nilfs_sync called\n"));
1079 	/* if called when mounted readonly, just ignore */
1080 	if (mp->mnt_flag & MNT_RDONLY)
1081 		return 0;
1082 
1083 	DPRINTF(VFSCALL, ("end of nilfs_sync()\n"));
1084 
1085 	return 0;
1086 }
1087 
1088 /* --------------------------------------------------------------------- */
1089 
1090 /*
1091  * Get vnode for the file system type specific file id ino for the fs. Its
1092  * used for reference to files by unique ID and for NFSv3.
1093  * (optional) TODO lookup why some sources state NFSv3
1094  */
1095 int
nilfs_vget(struct mount * mp,ino_t ino,int lktype,struct vnode ** vpp)1096 nilfs_vget(struct mount *mp, ino_t ino, int lktype,
1097     struct vnode **vpp)
1098 {
1099 	DPRINTF(NOTIMPL, ("nilfs_vget called\n"));
1100 	return EOPNOTSUPP;
1101 }
1102 
1103 /* --------------------------------------------------------------------- */
1104 
1105 /*
1106  * Read an inode from disk and initialize this vnode / inode pair.
1107  * Caller assures no other thread will try to load this inode.
1108  */
1109 int
nilfs_loadvnode(struct mount * mp,struct vnode * vp,const void * key,size_t key_len,const void ** new_key)1110 nilfs_loadvnode(struct mount *mp, struct vnode *vp,
1111     const void *key, size_t key_len, const void **new_key)
1112 {
1113 	uint64_t ino;
1114 	struct nilfs_device *nilfsdev;
1115 	struct nilfs_inode   inode, *entry;
1116 	struct nilfs_node *node;
1117 	struct nilfs_mount *ump;
1118 	struct buf *bp;
1119 	uint64_t ivblocknr;
1120 	uint32_t entry_in_block;
1121 	int error;
1122 	extern int (**nilfs_vnodeop_p)(void *);
1123 
1124 	KASSERT(key_len == sizeof(ino));
1125 	memcpy(&ino, key, key_len);
1126 
1127 	ump = VFSTONILFS(mp);
1128 
1129 	/* create new inode; XXX check could be handier */
1130 	if ((ino < NILFS_USER_INO) && (ino != NILFS_ROOT_INO)) {
1131 		printf("nilfs_get_node: system ino %"PRIu64" not in mount "
1132 			"point!\n", ino);
1133 		return ENOENT;
1134 	}
1135 
1136 	/* lookup inode in the ifile */
1137 	DPRINTF(NODE, ("lookup ino %"PRIu64"\n", ino));
1138 
1139 	/* lookup inode structure in mountpoints ifile */
1140 	nilfsdev = ump->nilfsdev;
1141 	nilfs_mdt_trans(&nilfsdev->ifile_mdt, ino, &ivblocknr, &entry_in_block);
1142 
1143 	error = nilfs_bread(ump->ifile_node, ivblocknr, 0, &bp);
1144 	if (error)
1145 		return ENOENT;
1146 
1147 	/* get inode entry */
1148 	entry =  (struct nilfs_inode *) bp->b_data + entry_in_block;
1149 	inode = *entry;
1150 	brelse(bp, BC_AGE);
1151 
1152 	/* get node */
1153 	error = nilfs_get_node_raw(ump->nilfsdev, ump, ino, &inode, &node);
1154 	if (error)
1155 		return error;
1156 
1157 	vp->v_type = IFTOVT(inode.i_mode);
1158 	switch (vp->v_type) {
1159 	case VREG:
1160 	case VDIR:
1161 	case VLNK:
1162 		break;
1163 	/* other types not yet supported. */
1164 	default:
1165 		vp->v_type = VNON;
1166 		nilfs_dispose_node(&node);
1167 		return ENXIO;
1168 	}
1169 
1170 	vp->v_tag = VT_NILFS;
1171 	vp->v_op = nilfs_vnodeop_p;
1172 	vp->v_data = node;
1173 	node->vnode = vp;
1174 
1175 	/* initialise genfs */
1176 	genfs_node_init(vp, &nilfs_genfsops);
1177 
1178 	/* check if we're fetching the root */
1179 	if (ino == NILFS_ROOT_INO)
1180 		vp->v_vflag |= VV_ROOT;
1181 
1182 	uvm_vnp_setsize(vp, nilfs_rw64(inode.i_size));
1183 	*new_key = &node->ino;
1184 
1185 	return 0;
1186 
1187 }
1188 
1189 /* --------------------------------------------------------------------- */
1190 
1191 /*
1192  * Lookup vnode for file handle specified
1193  */
1194 int
nilfs_fhtovp(struct mount * mp,struct fid * fhp,int lktype,struct vnode ** vpp)1195 nilfs_fhtovp(struct mount *mp, struct fid *fhp, int lktype,
1196     struct vnode **vpp)
1197 {
1198 	DPRINTF(NOTIMPL, ("nilfs_fhtovp called\n"));
1199 	return EOPNOTSUPP;
1200 }
1201 
1202 /* --------------------------------------------------------------------- */
1203 
1204 /*
1205  * Create an unique file handle. Its structure is opaque and won't be used by
1206  * other subsystems. It should uniquely identify the file in the filingsystem
1207  * and enough information to know if a file has been removed and/or resources
1208  * have been recycled.
1209  */
1210 int
nilfs_vptofh(struct vnode * vp,struct fid * fid,size_t * fh_size)1211 nilfs_vptofh(struct vnode *vp, struct fid *fid,
1212     size_t *fh_size)
1213 {
1214 	DPRINTF(NOTIMPL, ("nilfs_vptofh called\n"));
1215 	return EOPNOTSUPP;
1216 }
1217 
1218 /* --------------------------------------------------------------------- */
1219 
1220 /*
1221  * Create a file system snapshot at the specified timestamp.
1222  */
1223 int
nilfs_snapshot(struct mount * mp,struct vnode * vp,struct timespec * tm)1224 nilfs_snapshot(struct mount *mp, struct vnode *vp,
1225     struct timespec *tm)
1226 {
1227 	DPRINTF(NOTIMPL, ("nilfs_snapshot called\n"));
1228 	return EOPNOTSUPP;
1229 }
1230 
1231 /* --------------------------------------------------------------------- */
1232