xref: /netbsd-src/sys/fs/nilfs/nilfs_subr.c (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
1 /* $NetBSD: nilfs_subr.c,v 1.7 2010/08/11 11:43:13 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_subr.c,v 1.7 2010/08/11 11:43:13 pgoyette Exp $");
32 #endif /* not lint */
33 
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/namei.h>
37 #include <sys/resourcevar.h>	/* defines plimit structure in proc struct */
38 #include <sys/kernel.h>
39 #include <sys/file.h>		/* define FWRITE ... */
40 #include <sys/stat.h>
41 #include <sys/buf.h>
42 #include <sys/proc.h>
43 #include <sys/mount.h>
44 #include <sys/vnode.h>
45 #include <sys/signalvar.h>
46 #include <sys/malloc.h>
47 #include <sys/dirent.h>
48 #include <sys/lockf.h>
49 #include <sys/kauth.h>
50 #include <sys/dirhash.h>
51 
52 #include <miscfs/genfs/genfs.h>
53 #include <uvm/uvm_extern.h>
54 
55 #include <fs/nilfs/nilfs_mount.h>
56 #include "nilfs.h"
57 #include "nilfs_subr.h"
58 #include "nilfs_bswap.h"
59 
60 
61 #define VTOI(vnode) ((struct nilfs_node *) (vnode)->v_data)
62 
63 /* forwards */
64 static int nilfs_btree_lookup(struct nilfs_node *node, uint64_t lblocknr,
65 	uint64_t *vblocknr);
66 
67 /* basic calculators */
68 uint64_t nilfs_get_segnum_of_block(struct nilfs_device *nilfsdev,
69 	uint64_t blocknr)
70 {
71 	return blocknr / nilfs_rw32(nilfsdev->super.s_blocks_per_segment);
72 }
73 
74 
75 void
76 nilfs_get_segment_range(struct nilfs_device *nilfsdev, uint64_t segnum,
77         uint64_t *seg_start, uint64_t *seg_end)
78 {
79         uint64_t blks_per_seg;
80 
81         blks_per_seg = nilfs_rw64(nilfsdev->super.s_blocks_per_segment);
82         *seg_start = blks_per_seg * segnum;
83         *seg_end   = *seg_start + blks_per_seg -1;
84         if (segnum == 0)
85                 *seg_start = nilfs_rw64(nilfsdev->super.s_first_data_block);
86 }
87 
88 
89 void nilfs_calc_mdt_consts(struct nilfs_device *nilfsdev,
90 	struct nilfs_mdt *mdt, int entry_size)
91 {
92 	uint32_t blocksize = nilfsdev->blocksize;
93 
94 	mdt->entries_per_group = blocksize * 8;	   /* bits in sector */
95 	mdt->entries_per_block = blocksize / entry_size;
96 
97 	mdt->blocks_per_group  =
98 		(mdt->entries_per_group -1) / mdt->entries_per_block + 1 + 1;
99 	mdt->groups_per_desc_block =
100 		blocksize / sizeof(struct nilfs_block_group_desc);
101 	mdt->blocks_per_desc_block =
102 		mdt->groups_per_desc_block * mdt->blocks_per_group + 1;
103 }
104 
105 
106 /* from NetBSD's src/sys/net/if_ethersubr.c */
107 uint32_t
108 crc32_le(uint32_t crc, const uint8_t *buf, size_t len)
109 {
110         static const uint32_t crctab[] = {
111                 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
112                 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
113                 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
114                 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
115         };
116         size_t i;
117 
118         for (i = 0; i < len; i++) {
119                 crc ^= buf[i];
120                 crc = (crc >> 4) ^ crctab[crc & 0xf];
121                 crc = (crc >> 4) ^ crctab[crc & 0xf];
122         }
123 
124         return (crc);
125 }
126 
127 
128 static int
129 nilfs_calchash(uint64_t ino)
130 {
131 	return (int) ino;
132 }
133 
134 
135 /* dev reading */
136 static int
137 nilfs_dev_bread(struct nilfs_device *nilfsdev, uint64_t blocknr,
138 	struct kauth_cred *cred, int flags, struct buf **bpp)
139 {
140 	int blk2dev = nilfsdev->blocksize / DEV_BSIZE;
141 
142 	return bread(nilfsdev->devvp, blocknr * blk2dev, nilfsdev->blocksize,
143 		NOCRED, 0, bpp);
144 }
145 
146 
147 /* read on a node */
148 int
149 nilfs_bread(struct nilfs_node *node, uint64_t blocknr,
150 	struct kauth_cred *cred, int flags, struct buf **bpp)
151 {
152 	uint64_t vblocknr;
153 	int error;
154 
155 	error = nilfs_btree_lookup(node, blocknr, &vblocknr);
156 	if (error)
157 		return error;
158 	return bread(node->vnode, vblocknr, node->nilfsdev->blocksize,
159 		cred, flags, bpp);
160 }
161 
162 
163 /* segment-log reading */
164 int
165 nilfs_get_segment_log(struct nilfs_device *nilfsdev, uint64_t *blocknr,
166 	uint64_t *offset, struct buf **bpp, int len, void *blob)
167 {
168 	int blocksize = nilfsdev->blocksize;
169 	int error;
170 
171 	KASSERT(len <= blocksize);
172 
173 	if (*offset + len > blocksize) {
174 		*blocknr = *blocknr + 1;
175 		*offset = 0;
176 	}
177 	if (*offset == 0) {
178 		if (*bpp)
179 			brelse(*bpp, BC_AGE);
180 		/* read in block */
181 		error = nilfs_dev_bread(nilfsdev, *blocknr, NOCRED, 0, bpp);
182 		if (error)
183 			return error;
184 	}
185 	memcpy(blob, ((uint8_t *) (*bpp)->b_data) + *offset, len);
186 	*offset += len;
187 
188 	return 0;
189 }
190 
191 /* -------------------------------------------------------------------------- */
192 
193 /* btree operations */
194 
195 static int
196 nilfs_btree_lookup_level(struct nilfs_node *node, uint64_t lblocknr,
197 		uint64_t btree_vblknr, int level, uint64_t *vblocknr)
198 {
199 	struct nilfs_device *nilfsdev = node->nilfsdev;
200 	struct nilfs_btree_node *btree_hdr;
201 	struct buf *bp;
202 	uint64_t btree_blknr;
203 	uint64_t *dkeys, *dptrs, child_btree_blk;
204 	uint8_t  *pos;
205 	int i, error, selected;
206 
207 	DPRINTF(TRANSLATE, ("nilfs_btree_lookup_level ino %"PRIu64", "
208 		"lblocknr %"PRIu64", btree_vblknr %"PRIu64", level %d\n",
209 		node->ino, lblocknr, btree_vblknr, level));
210 
211 	/* translate btree_vblknr */
212 	error = nilfs_nvtop(node, 1, &btree_vblknr, &btree_blknr);
213 	if (error)
214 		return error;
215 
216 	/* get our block */
217 	error = nilfs_dev_bread(nilfsdev, btree_blknr, NOCRED, 0, &bp);
218 	if (error) {
219 		brelse(bp, BC_AGE);
220 		return error;
221 	}
222 
223 	btree_hdr = (struct nilfs_btree_node *) bp->b_data;
224 	pos =   (uint8_t *) bp->b_data +
225 		sizeof(struct nilfs_btree_node) +
226 		NILFS_BTREE_NODE_EXTRA_PAD_SIZE;
227 	dkeys = (uint64_t *) pos;
228 	dptrs = dkeys + NILFS_BTREE_NODE_NCHILDREN_MAX(nilfsdev->blocksize);
229 
230 	assert((btree_hdr->bn_flags & NILFS_BTREE_NODE_ROOT) == 0);
231 
232 	/* select matching child XXX could use binary search */
233 	selected = 0;
234 	for (i = 0; i < nilfs_rw16(btree_hdr->bn_nchildren); i++) {
235 		if (dkeys[i] > lblocknr)
236 			break;
237 		selected = i;
238 	}
239 
240 	if (level == 1) {
241 		/* if found it mapped */
242 		if (dkeys[selected] == lblocknr)
243 			*vblocknr = dptrs[selected];
244 		brelse(bp, BC_AGE);
245 		return 0;
246 	}
247 
248 	/* lookup in selected child */
249 	assert(dkeys[selected] <= lblocknr);
250 	child_btree_blk = dptrs[selected];
251 	brelse(bp, BC_AGE);
252 
253 	return nilfs_btree_lookup_level(node, lblocknr,
254 			child_btree_blk, btree_hdr->bn_level-1, vblocknr);
255 }
256 
257 
258 /* internal function */
259 static int
260 nilfs_btree_lookup(struct nilfs_node *node, uint64_t lblocknr,
261 		uint64_t *vblocknr)
262 {
263 	struct nilfs_inode  *inode    = &node->inode;
264 	struct nilfs_btree_node  *btree_hdr;
265 	uint64_t *dkeys, *dptrs, *dtrans;
266 	int i, selected;
267 	int error;
268 
269 	DPRINTF(TRANSLATE, ("nilfs_btree_lookup ino %"PRIu64", "
270 		"lblocknr %"PRIu64"\n", node->ino, lblocknr));
271 
272 	btree_hdr  = (struct nilfs_btree_node *) &inode->i_bmap[0];
273 	dkeys  = &inode->i_bmap[1];
274 	dptrs  = dkeys + NILFS_BTREE_ROOT_NCHILDREN_MAX;
275 	dtrans = &inode->i_bmap[1];
276 
277 	/* SMALL, direct lookup */
278 	*vblocknr = 0;
279 	if ((btree_hdr->bn_flags & NILFS_BMAP_LARGE) == 0) {
280 		if (lblocknr < NILFS_DIRECT_NBLOCKS) {
281 			*vblocknr = dtrans[lblocknr];
282 			return 0;
283 		}
284 		/* not mapped XXX could be considered error here */
285 		return 0;
286 	}
287 
288 	/* LARGE, select matching child; XXX could use binary search */
289 	dtrans = NULL;
290 	error = 0;
291 	selected = 0;
292 	for (i = 0; i < nilfs_rw16(btree_hdr->bn_nchildren); i++) {
293 		if (dkeys[i] > lblocknr)
294 			break;
295 		selected = i;
296 	}
297 
298 	/* if selected key > lblocknr, its not mapped */
299 	if (dkeys[selected] > lblocknr)
300 		return 0;
301 
302 	/* overshooting? then not mapped */
303 	if (selected == nilfs_rw16(btree_hdr->bn_nchildren))
304 		return 0;
305 
306 	/* level should be > 1 or otherwise it should be a direct one */
307 	assert(btree_hdr->bn_level > 1);
308 
309 	/* lookup in selected child */
310 	assert(dkeys[selected] <= lblocknr);
311 	error = nilfs_btree_lookup_level(node, lblocknr,
312 			dptrs[selected], btree_hdr->bn_level-1, vblocknr);
313 
314 	return error;
315 }
316 
317 
318 /* node should be locked on entry to prevent btree changes (unlikely) */
319 int
320 nilfs_btree_nlookup(struct nilfs_node *node, uint64_t from, uint64_t blks,
321 		uint64_t *l2vmap)
322 {
323 	uint64_t lblocknr, *vblocknr;
324 	int i, error;
325 
326 	/* TODO / OPTI multiple translations in one go possible */
327 	error = EINVAL;
328 	for (i = 0; i < blks; i++) {
329 		lblocknr  = from + i;
330 		vblocknr  = l2vmap + i;
331 		error = nilfs_btree_lookup(node, lblocknr, vblocknr);
332 
333 		DPRINTF(TRANSLATE, ("btree_nlookup ino %"PRIu64", "
334 			"lblocknr %"PRIu64" -> %"PRIu64"\n",
335 			node->ino, lblocknr, *vblocknr));
336 		if (error)
337 			break;
338 	}
339 
340 	return error;
341 }
342 
343 /* --------------------------------------------------------------------- */
344 
345 /* vtop operations */
346 
347 /* translate index to a file block number and an entry */
348 static void
349 nilfs_mdt_trans(struct nilfs_mdt *mdt, uint64_t index,
350 	uint64_t *blocknr, uint32_t *entry_in_block)
351 {
352 	uint64_t blknr;
353 	uint64_t group, group_offset, blocknr_in_group;
354 	uint64_t desc_block, desc_offset;
355 
356 	/* calculate our offset in the file */
357 	group             = index / mdt->entries_per_group;
358 	group_offset      = index % mdt->entries_per_group;
359 	desc_block        = group / mdt->groups_per_desc_block;
360 	desc_offset       = group % mdt->groups_per_desc_block;
361 	blocknr_in_group  = group_offset / mdt->entries_per_block;
362 
363 	/* to descgroup offset */
364 	blknr = 1 + desc_block * mdt->blocks_per_desc_block;
365 
366 	/* to group offset */
367 	blknr += desc_offset * mdt->blocks_per_group;
368 
369 	/* to actual file block */
370 	blknr += 1 + blocknr_in_group;
371 
372 	*blocknr        = blknr;
373 	*entry_in_block = group_offset % mdt->entries_per_block;
374 }
375 
376 
377 static int
378 nilfs_vtop(struct nilfs_device *nilfsdev, uint64_t vblocknr, uint64_t *pblocknr)
379 {
380 	struct nilfs_dat_entry *entry;
381 	struct buf *bp;
382 	uint64_t  ldatblknr;
383 	uint32_t  entry_in_block;
384 	int error;
385 
386 	nilfs_mdt_trans(&nilfsdev->dat_mdt, vblocknr,
387 		&ldatblknr, &entry_in_block);
388 
389 	error = nilfs_bread(nilfsdev->dat_node, ldatblknr, NOCRED, 0, &bp);
390 	if (error) {
391 		printf("vtop: can't read in DAT block %"PRIu64"!\n", ldatblknr);
392 		brelse(bp, BC_AGE);
393 		return error;
394 	}
395 
396 	/* get our translation */
397 	entry = ((struct nilfs_dat_entry *) bp->b_data) + entry_in_block;
398 #if 0
399 	printf("\tvblk %4"PRIu64" -> %"PRIu64" for "
400 		"checkpoint %"PRIu64" to %"PRIu64"\n",
401 		vblocknr,
402 		nilfs_rw64(entry->de_blocknr),
403 		nilfs_rw64(entry->de_start),
404 		nilfs_rw64(entry->de_end));
405 #endif
406 
407 	*pblocknr = nilfs_rw64(entry->de_blocknr);
408 	brelse(bp, BC_AGE);
409 
410 	return 0;
411 }
412 
413 
414 int
415 nilfs_nvtop(struct nilfs_node *node, uint64_t blks, uint64_t *l2vmap,
416 		uint64_t *v2pmap)
417 {
418 	uint64_t vblocknr, *pblocknr;
419 	int i, error;
420 
421 	/* the DAT inode is the only one not mapped virtual */
422 	if (node->ino == NILFS_DAT_INO) {
423 		memcpy(v2pmap, l2vmap, blks * sizeof(uint64_t));
424 		return 0;
425 	}
426 
427 	/* TODO / OPTI more translations in one go */
428 	error = 0;
429 	for (i = 0; i < blks; i++) {
430 		vblocknr  = l2vmap[i];
431 		pblocknr  = v2pmap + i;
432 		*pblocknr = 0;
433 
434 		/* only translate valid vblocknrs */
435 		if (vblocknr == 0)
436 			continue;
437 		error = nilfs_vtop(node->nilfsdev, vblocknr, pblocknr);
438 		if (error)
439 			break;
440 	}
441 
442 	return error;
443 }
444 
445 /* --------------------------------------------------------------------- */
446 
447 struct nilfs_recover_info {
448 	uint64_t segnum;
449 	uint64_t pseg;
450 
451 	struct nilfs_segment_summary segsum;
452 	struct nilfs_super_root      super_root;
453 	STAILQ_ENTRY(nilfs_recover_info) next;
454 };
455 
456 
457 /*
458  * Helper functions of nilfs_mount() that actually mounts the disc.
459  */
460 static int
461 nilfs_load_segsum(struct nilfs_device *nilfsdev,
462 	struct nilfs_recover_info *ri)
463 {
464 	struct buf *bp;
465 	uint64_t blocknr, offset;
466 	uint32_t segsum_struct_size;
467 	uint32_t magic;
468 	int error;
469 
470 	segsum_struct_size = sizeof(struct nilfs_segment_summary);
471 
472 	/* read in segsum structure */
473 	bp      = NULL;
474 	blocknr = ri->pseg;
475 	offset  = 0;
476 	error = nilfs_get_segment_log(nilfsdev,
477 			&blocknr, &offset, &bp,
478 			segsum_struct_size, (void *) &ri->segsum);
479 	if (error)
480 		goto out;
481 
482 	/* sanity checks */
483 	magic = nilfs_rw32(ri->segsum.ss_magic);
484 	if (magic != NILFS_SEGSUM_MAGIC) {
485 		DPRINTF(VOLUMES, ("nilfs: bad magic in pseg %"PRIu64"\n",
486 			ri->pseg));
487 		error = EINVAL;
488 		goto out;
489 	}
490 
491 	/* TODO check segment summary checksum */
492 	/* TODO check data checksum */
493 
494 out:
495 	if (bp)
496 		brelse(bp, BC_AGE);
497 
498 	return error;
499 }
500 
501 
502 static int
503 nilfs_load_super_root(struct nilfs_device *nilfsdev,
504 	struct nilfs_recover_info *ri)
505 {
506 	struct nilfs_segment_summary *segsum = &ri->segsum;
507 	struct nilfs_super_root *super_root;
508 	struct buf *bp;
509 	uint64_t blocknr, offset;
510 	uint32_t segsum_size, size;
511 	uint32_t nsumblk, nfileblk;
512 	uint32_t super_root_crc, comp_crc;
513 	int off, error;
514 
515 	/* process segment summary */
516 	segsum_size = nilfs_rw32(segsum->ss_sumbytes);
517 	nsumblk     = (segsum_size - 1) / nilfsdev->blocksize + 1;
518 	nfileblk    = nilfs_rw32(segsum->ss_nblocks) - nsumblk;
519 
520 	/* check if there is a superroot */
521 	if ((nilfs_rw16(segsum->ss_flags) & NILFS_SS_SR) == 0) {
522 		DPRINTF(VOLUMES, ("nilfs: no super root in pseg %"PRIu64"\n",
523 			ri->pseg));
524 		return ENOENT;
525 	}
526 
527 	/* get our super root, located at the end of the pseg */
528 	blocknr = ri->pseg + nsumblk + nfileblk - 1;
529 	offset = 0;
530 	size = sizeof(struct nilfs_super_root);
531 	bp = NULL;
532 	error = nilfs_get_segment_log(nilfsdev,
533 			&blocknr, &offset, &bp,
534 			size, (void *) &nilfsdev->super_root);
535 	if (bp)
536 		brelse(bp, BC_AGE);
537 	if (error) {
538 		printf("read in of superroot failed\n");
539 		return EIO;
540 	}
541 
542 	/* check super root crc */
543 	super_root = &nilfsdev->super_root;
544 	super_root_crc = nilfs_rw32(super_root->sr_sum);
545 	off = sizeof(super_root->sr_sum);
546 	comp_crc = crc32_le(nilfs_rw32(nilfsdev->super.s_crc_seed),
547 		(uint8_t *) super_root + off,
548 		NILFS_SR_BYTES - off);
549 	if (super_root_crc != comp_crc) {
550 		DPRINTF(VOLUMES, ("    invalid superroot, likely from old format\n"));
551 		return EINVAL;
552 	}
553 
554 	DPRINTF(VOLUMES, ("    got valid superroot\n"));
555 
556 	return 0;
557 }
558 
559 /*
560  * Search for the last super root recorded.
561  */
562 void
563 nilfs_search_super_root(struct nilfs_device *nilfsdev)
564 {
565 	struct nilfs_super_block *super;
566 	struct nilfs_segment_summary *segsum;
567 	struct nilfs_recover_info *ri, *ori, *i_ri;
568 	STAILQ_HEAD(,nilfs_recover_info) ri_list;
569 	uint64_t seg_start, seg_end, cno;
570 	uint32_t segsum_size;
571 	uint32_t nsumblk, nfileblk;
572 	int error;
573 
574 	STAILQ_INIT(&ri_list);
575 
576 	/* search for last super root */
577 	ri = malloc(sizeof(struct nilfs_recover_info), M_NILFSTEMP, M_WAITOK);
578 	memset(ri, 0, sizeof(struct nilfs_recover_info));
579 
580 	/* if enabled, start from the specified position */
581 	if (0) {
582 		/* start from set start */
583 		nilfsdev->super.s_last_pseg = nilfsdev->super.s_first_data_block;
584 		nilfsdev->super.s_last_cno  = nilfs_rw64(1);
585 	}
586 
587 	ri->pseg   = nilfs_rw64(nilfsdev->super.s_last_pseg); /* blknr */
588 	ri->segnum = nilfs_get_segnum_of_block(nilfsdev, ri->pseg);
589 
590 	error = 0;
591 	cno = nilfs_rw64(nilfsdev->super.s_last_cno);
592 	DPRINTF(VOLUMES, ("nilfs: seach_super_root start in pseg %"PRIu64"\n",
593 			ri->pseg));
594 	for (;;) {
595 		DPRINTF(VOLUMES, (" at pseg %"PRIu64"\n", ri->pseg));
596 		error = nilfs_load_segsum(nilfsdev, ri);
597 		if (error)
598 			break;
599 
600 		segsum = &ri->segsum;
601 
602 		/* try to load super root */
603 		if (nilfs_rw16(segsum->ss_flags) & NILFS_SS_SR) {
604 			DPRINTF(VOLUMES, (" try super root\n"));
605 			error = nilfs_load_super_root(nilfsdev, ri);
606 			if (error)
607 				break;	/* confused */
608 			/* wipe current list of ri */
609 			while (!STAILQ_EMPTY(&ri_list)) {
610 				i_ri = STAILQ_FIRST(&ri_list);
611 				STAILQ_REMOVE_HEAD(&ri_list, next);
612 				free(i_ri, M_NILFSTEMP);
613 			}
614 			super = &nilfsdev->super;
615 
616 			super->s_last_pseg = nilfs_rw64(ri->pseg);
617 			super->s_last_cno  = cno++;
618 			super->s_last_seq  = segsum->ss_seq;
619 			super->s_state     = nilfs_rw16(NILFS_VALID_FS);
620 		} else {
621 			STAILQ_INSERT_TAIL(&ri_list, ri, next);
622 			ori = ri;
623 			ri = malloc(sizeof(struct nilfs_recover_info),
624 				M_NILFSTEMP, M_WAITOK);
625 			memset(ri, 0, sizeof(struct nilfs_recover_info));
626 			ri->segnum = ori->segnum;
627 			ri->pseg   = ori->pseg;
628 			/* segsum keeps pointing to the `old' ri */
629 		}
630 
631 		/* continue to the next pseg */
632 		segsum_size = nilfs_rw32(segsum->ss_sumbytes);
633 		nsumblk     = (segsum_size - 1) / nilfsdev->blocksize + 1;
634 		nfileblk    = nilfs_rw32(segsum->ss_nblocks) - nsumblk;
635 
636 		/* calculate next partial segment location */
637 		ri->pseg += nsumblk + nfileblk;
638 
639 		/* did we reach the end of the segment? if so, go to the next */
640 		nilfs_get_segment_range(nilfsdev, ri->segnum, &seg_start, &seg_end);
641 		if (ri->pseg >= seg_end)
642 			ri->pseg = nilfs_rw64(segsum->ss_next);
643 		ri->segnum = nilfs_get_segnum_of_block(nilfsdev, ri->pseg);
644 	}
645 
646 	/*
647 	 * XXX No roll-forward yet of the remaining partial segments.
648 	 */
649 
650 	/* wipe current list of ri */
651 	while (!STAILQ_EMPTY(&ri_list)) {
652 		i_ri = STAILQ_FIRST(&ri_list);
653 		STAILQ_REMOVE_HEAD(&ri_list, next);
654 		printf("nilfs: ignoring pseg at %"PRIu64"\n", i_ri->pseg);
655 		free(i_ri, M_NILFSTEMP);
656 	}
657 	free(ri, M_NILFSTEMP);
658 }
659 
660 /* --------------------------------------------------------------------- */
661 
662 /*
663  * Genfs interfacing
664  *
665  * static const struct genfs_ops nilfs_genfsops = {
666  * 	.gop_size = genfs_size,
667  * 		size of transfers
668  * 	.gop_alloc = nilfs_gop_alloc,
669  * 		allocate len bytes at offset
670  * 	.gop_write = genfs_gop_write,
671  * 		putpages interface code
672  * 	.gop_markupdate = nilfs_gop_markupdate,
673  * 		set update/modify flags etc.
674  * }
675  */
676 
677 /*
678  * Callback from genfs to allocate len bytes at offset off; only called when
679  * filling up gaps in the allocation.
680  */
681 static int
682 nilfs_gop_alloc(struct vnode *vp, off_t off,
683     off_t len, int flags, kauth_cred_t cred)
684 {
685 	DPRINTF(NOTIMPL, ("nilfs_gop_alloc not implemented\n"));
686 	DPRINTF(ALLOC, ("nilfs_gop_alloc called for %"PRIu64" bytes\n", len));
687 
688 	return 0;
689 }
690 
691 
692 /*
693  * callback from genfs to update our flags
694  */
695 static void
696 nilfs_gop_markupdate(struct vnode *vp, int flags)
697 {
698 	struct nilfs_node *nilfs_node = VTOI(vp);
699 	u_long mask = 0;
700 
701 	if ((flags & GOP_UPDATE_ACCESSED) != 0) {
702 		mask = IN_ACCESS;
703 	}
704 	if ((flags & GOP_UPDATE_MODIFIED) != 0) {
705 		if (vp->v_type == VREG) {
706 			mask |= IN_CHANGE | IN_UPDATE;
707 		} else {
708 			mask |= IN_MODIFY;
709 		}
710 	}
711 	if (mask) {
712 		nilfs_node->i_flags |= mask;
713 	}
714 }
715 
716 
717 static const struct genfs_ops nilfs_genfsops = {
718 	.gop_size = genfs_size,
719 	.gop_alloc = nilfs_gop_alloc,
720 	.gop_write = genfs_gop_write_rwmap,
721 	.gop_markupdate = nilfs_gop_markupdate,
722 };
723 
724 /* --------------------------------------------------------------------- */
725 
726 static void
727 nilfs_register_node(struct nilfs_node *node)
728 {
729 	struct nilfs_mount *ump;
730 	struct nilfs_node *chk;
731 	uint32_t hashline;
732 
733 	ump = node->ump;
734 	mutex_enter(&ump->ihash_lock);
735 
736 	/* add to our hash table */
737 	hashline = nilfs_calchash(node->ino) & NILFS_INODE_HASHMASK;
738 #ifdef DEBUG
739 	LIST_FOREACH(chk, &ump->nilfs_nodes[hashline], hashchain) {
740 		assert(chk);
741 		if (chk->ino == node->ino)
742 			panic("Double node entered\n");
743 	}
744 #else
745 	chk = NULL;
746 #endif
747 	LIST_INSERT_HEAD(&ump->nilfs_nodes[hashline], node, hashchain);
748 
749 	mutex_exit(&ump->ihash_lock);
750 }
751 
752 
753 static void
754 nilfs_deregister_node(struct nilfs_node *node)
755 {
756 	struct nilfs_mount *ump;
757 
758 	ump = node->ump;
759 	mutex_enter(&ump->ihash_lock);
760 
761 	/* remove from hash list */
762 	LIST_REMOVE(node, hashchain);
763 
764 	mutex_exit(&ump->ihash_lock);
765 }
766 
767 
768 static struct nilfs_node *
769 nilfs_hash_lookup(struct nilfs_mount *ump, ino_t ino)
770 {
771 	struct nilfs_node *node;
772 	struct vnode *vp;
773 	uint32_t hashline;
774 
775 loop:
776 	mutex_enter(&ump->ihash_lock);
777 
778 	/* search our hash table */
779 	hashline = nilfs_calchash(ino) & NILFS_INODE_HASHMASK;
780 	LIST_FOREACH(node, &ump->nilfs_nodes[hashline], hashchain) {
781 		assert(node);
782 		if (node->ino == ino) {
783 			vp = node->vnode;
784 			assert(vp);
785 			mutex_enter(&vp->v_interlock);
786 			mutex_exit(&ump->ihash_lock);
787 			if (vget(vp, LK_EXCLUSIVE))
788 				goto loop;
789 			return node;
790 		}
791 	}
792 	mutex_exit(&ump->ihash_lock);
793 
794 	return NULL;
795 }
796 
797 
798 /* node action implementators */
799 extern int (**nilfs_vnodeop_p)(void *);
800 
801 int
802 nilfs_get_node_raw(struct nilfs_device *nilfsdev, struct nilfs_mount *ump,
803 	uint64_t ino, struct nilfs_inode *inode, struct nilfs_node **nodep)
804 {
805 	struct nilfs_node *node;
806 	struct vnode *nvp;
807 	struct mount *mp;
808 	int (**vnodeops)(void *);
809 	int error;
810 
811 	*nodep = NULL;
812 	vnodeops = nilfs_vnodeop_p;
813 
814 	/* associate with mountpoint if present*/
815 	mp = ump? ump->vfs_mountp : NULL;
816 	error = getnewvnode(VT_NILFS, mp, vnodeops, &nvp);
817 	if (error)
818 		return error;
819 
820 	/* lock node */
821 	error = vn_lock(nvp, LK_EXCLUSIVE | LK_RETRY);
822 	if (error) {
823 		nvp->v_data = NULL;
824 		ungetnewvnode(nvp);
825 		return error;
826 	}
827 
828 	node = pool_get(&nilfs_node_pool, PR_WAITOK);
829 	memset(node, 0, sizeof(struct nilfs_node));
830 
831 	/* crosslink */
832 	node->vnode    = nvp;
833 	node->ump      = ump;
834 	node->nilfsdev = nilfsdev;
835 	nvp->v_data    = node;
836 
837 	/* initiase nilfs node */
838 	node->ino   = ino;
839 	node->inode = *inode;
840 	node->lockf = NULL;
841 
842 	/* needed? */
843 	mutex_init(&node->node_mutex, MUTEX_DEFAULT, IPL_NONE);
844 	cv_init(&node->node_lock, "nilfsnlk");
845 
846 	/* initialise genfs */
847 	genfs_node_init(nvp, &nilfs_genfsops);
848 
849 	/* check if we're fetching the root */
850 	if (ino == NILFS_ROOT_INO)
851 		nvp->v_vflag |= VV_ROOT;
852 
853 	/* update vnode's file type XXX is there a function for this? */
854 	nvp->v_type = VREG;
855 	if (S_ISDIR(inode->i_mode))
856 		nvp->v_type = VDIR;
857 	if (S_ISLNK(inode->i_mode))
858 		nvp->v_type = VLNK;
859 #if 0
860 	if (S_ISCHR(inode->i_mode))
861 		nvp->v_type = VCHR;
862 	if (S_ISBLK(inode->i_mode))
863 		nvp->v_type = VBLK;
864 #endif
865 	/* XXX what else? */
866 
867 	/* fixup inode size for system nodes */
868 	if ((ino < NILFS_USER_INO) && (ino != NILFS_ROOT_INO)) {
869 		DPRINTF(VOLUMES, ("NEED TO GET my size for inode %"PRIu64"\n",
870 			ino));
871 		/* for now set it to maximum, -1 is illegal */
872 		inode->i_size = nilfs_rw64(((uint64_t) -2));
873 	}
874 
875 	uvm_vnp_setsize(nvp, nilfs_rw64(inode->i_size));
876 
877 	if (ump)
878 		nilfs_register_node(node);
879 
880 	/* return node */
881 	*nodep = node;
882 	return 0;
883 }
884 
885 
886 int
887 nilfs_get_node(struct nilfs_mount *ump, uint64_t ino, struct nilfs_node **nodep)
888 {
889 	struct nilfs_device *nilfsdev;
890 	struct nilfs_inode   inode, *entry;
891 	struct buf *bp;
892 	uint64_t ivblocknr;
893 	uint32_t entry_in_block;
894 	int error;
895 
896 	/* lookup node in hash table */
897 	*nodep = nilfs_hash_lookup(ump, ino);
898 	if (*nodep)
899 		return 0;
900 
901 	/* lock to disallow simultanious creation of same udf_node */
902 	mutex_enter(&ump->get_node_lock);
903 
904 	/* relookup since it could be created while waiting for the mutex */
905 	*nodep = nilfs_hash_lookup(ump, ino);
906 	if (*nodep) {
907 		mutex_exit(&ump->get_node_lock);
908 		return 0;
909 	}
910 
911 	/* create new inode; XXX check could be handier */
912 	if ((ino < NILFS_ATIME_INO) && (ino != NILFS_ROOT_INO)) {
913 		printf("nilfs_get_node: system ino %"PRIu64" not in mount "
914 			"point!\n", ino);
915 		mutex_exit(&ump->get_node_lock);
916 		return ENOENT;
917 	}
918 
919 	/* lookup inode in the ifile */
920 	DPRINTF(NODE, ("lookup ino %"PRIu64"\n", ino));
921 
922 	/* lookup inode structure in mountpoints ifile */
923 	nilfsdev = ump->nilfsdev;
924 	nilfs_mdt_trans(&nilfsdev->ifile_mdt, ino, &ivblocknr, &entry_in_block);
925 
926 	error = nilfs_bread(ump->ifile_node, ivblocknr, NOCRED, 0, &bp);
927 	if (error) {
928 		mutex_exit(&ump->get_node_lock);
929 		return ENOENT;
930 	}
931 
932 	/* get inode entry */
933 	entry =  (struct nilfs_inode *) bp->b_data + entry_in_block;
934 	inode = *entry;
935 	brelse(bp, BC_AGE);
936 
937 	/* get node */
938 	error = nilfs_get_node_raw(ump->nilfsdev, ump, ino, &inode, nodep);
939 	mutex_exit(&ump->get_node_lock);
940 
941 	return error;
942 }
943 
944 
945 void
946 nilfs_dispose_node(struct nilfs_node **nodep)
947 {
948 	struct vnode *vp;
949 	struct nilfs_node *node;
950 
951 	/* protect against rogue values */
952 	if (!*nodep)
953 		return;
954 
955 	node = *nodep;
956 	vp = node->vnode;
957 
958 	/* remove dirhash if present */
959 	dirhash_purge(&node->dir_hash);
960 
961 	/* remove from our hash lookup table */
962 	if (node->ump)
963 		nilfs_deregister_node(node);
964 
965 	/* destroy our locks */
966 	mutex_destroy(&node->node_mutex);
967 	cv_destroy(&node->node_lock);
968 
969 	/* dissociate from our vnode */
970 	genfs_node_destroy(node->vnode);
971 	vp->v_data = NULL;
972 
973 	/* free our associated memory */
974 	pool_put(&nilfs_node_pool, node);
975 
976 	*nodep = NULL;
977 }
978 
979 
980 void
981 nilfs_itimes(struct nilfs_node *node, struct timespec *acc,
982 	struct timespec *mod, struct timespec *birth)
983 {
984 }
985 
986 
987 int
988 nilfs_update(struct vnode *node, struct timespec *acc,
989 	struct timespec *mod, struct timespec *birth, int updflags)
990 {
991 	return EROFS;
992 }
993 
994 
995 int
996 nilfs_chsize(struct vnode *vp, u_quad_t newsize, kauth_cred_t cred)
997 {
998 	return EROFS;
999 }
1000 
1001 
1002 
1003 int
1004 nilfs_grow_node(struct nilfs_node *node, uint64_t new_size)
1005 {
1006 	return EROFS;
1007 }
1008 
1009 
1010 int
1011 nilfs_shrink_node(struct nilfs_node *node, uint64_t new_size)
1012 {
1013 	return EROFS;
1014 }
1015 
1016 
1017 static int
1018 dirhash_fill(struct nilfs_node *dir_node)
1019 {
1020 	struct vnode *dvp = dir_node->vnode;
1021 	struct dirhash *dirh;
1022 	struct nilfs_dir_entry *ndirent;
1023 	struct dirent dirent;
1024 	struct buf *bp;
1025 	uint64_t file_size, diroffset, blkoff;
1026 	uint64_t blocknr;
1027 	uint32_t blocksize = dir_node->nilfsdev->blocksize;
1028 	uint8_t *pos, name_len;
1029 	int error;
1030 
1031 	DPRINTF(CALL, ("dirhash_fill called\n"));
1032 
1033 	if (dvp->v_type != VDIR)
1034 		return ENOTDIR;
1035 
1036 	/* make sure we have a dirhash to work on */
1037 	dirh = dir_node->dir_hash;
1038 	KASSERT(dirh);
1039 	KASSERT(dirh->refcnt > 0);
1040 
1041 	if (dirh->flags & DIRH_BROKEN)
1042 		return EIO;
1043 
1044 	if (dirh->flags & DIRH_COMPLETE)
1045 		return 0;
1046 
1047 	DPRINTF(DIRHASH, ("Filling directory hash\n"));
1048 
1049 	/* make sure we have a clean dirhash to add to */
1050 	dirhash_purge_entries(dirh);
1051 
1052 	/* get directory filesize */
1053 	file_size = nilfs_rw64(dir_node->inode.i_size);
1054 
1055 	/* walk the directory */
1056 	error = 0;
1057 	diroffset = 0;
1058 
1059 	blocknr = diroffset / blocksize;
1060 	blkoff  = diroffset % blocksize;
1061 	error = nilfs_bread(dir_node, blocknr, NOCRED, 0, &bp);
1062 	if (error) {
1063 		dirh->flags |= DIRH_BROKEN;
1064 		dirhash_purge_entries(dirh);
1065 		return EIO;
1066 	}
1067 	while (diroffset < file_size) {
1068 		DPRINTF(READDIR, ("filldir : offset = %"PRIu64"\n",
1069 			diroffset));
1070 		if (blkoff >= blocksize) {
1071 			blkoff = 0; blocknr++;
1072 			brelse(bp, BC_AGE);
1073 			error = nilfs_bread(dir_node, blocknr, NOCRED, 0,
1074 					&bp);
1075 			if (error) {
1076 				dirh->flags |= DIRH_BROKEN;
1077 				dirhash_purge_entries(dirh);
1078 				return EIO;
1079 			}
1080 		}
1081 
1082 		/* read in one dirent */
1083 		pos = (uint8_t *) bp->b_data + blkoff;
1084 		ndirent = (struct nilfs_dir_entry *) pos;
1085 		name_len = ndirent->name_len;
1086 
1087 		memset(&dirent, 0, sizeof(struct dirent));
1088 		dirent.d_fileno = nilfs_rw64(ndirent->inode);
1089 		dirent.d_type   = ndirent->file_type;	/* 1:1 ? */
1090 		dirent.d_namlen = name_len;
1091 		strncpy(dirent.d_name, ndirent->name, name_len);
1092 		dirent.d_reclen = _DIRENT_SIZE(&dirent);
1093 		DPRINTF(DIRHASH, ("copying `%*.*s`\n", name_len,
1094 			name_len, dirent.d_name));
1095 
1096 		/* XXX is it deleted? extra free space? */
1097 		dirhash_enter(dirh, &dirent, diroffset,
1098 			nilfs_rw16(ndirent->rec_len), 0);
1099 
1100 		/* advance */
1101 		diroffset += nilfs_rw16(ndirent->rec_len);
1102 		blkoff    += nilfs_rw16(ndirent->rec_len);
1103 	}
1104 	brelse(bp, BC_AGE);
1105 
1106 	dirh->flags |= DIRH_COMPLETE;
1107 
1108 	return 0;
1109 }
1110 
1111 
1112 int
1113 nilfs_lookup_name_in_dir(struct vnode *dvp, const char *name, int namelen,
1114 		uint64_t *ino, int *found)
1115 {
1116 	struct nilfs_node	*dir_node = VTOI(dvp);
1117 	struct nilfs_dir_entry *ndirent;
1118 	struct dirhash		*dirh;
1119 	struct dirhash_entry	*dirh_ep;
1120 	struct buf *bp;
1121 	uint64_t diroffset, blkoff;
1122 	uint64_t blocknr;
1123 	uint32_t blocksize = dir_node->nilfsdev->blocksize;
1124 	uint8_t *pos;
1125 	int hit, error;
1126 
1127 	/* set default return */
1128 	*found = 0;
1129 
1130 	/* get our dirhash and make sure its read in */
1131 	dirhash_get(&dir_node->dir_hash);
1132 	error = dirhash_fill(dir_node);
1133 	if (error) {
1134 		dirhash_put(dir_node->dir_hash);
1135 		return error;
1136 	}
1137 	dirh = dir_node->dir_hash;
1138 
1139 	/* allocate temporary space for fid */
1140 
1141 	DPRINTF(DIRHASH, ("dirhash_lookup looking for `%*.*s`\n",
1142 		namelen, namelen, name));
1143 
1144 	/* search our dirhash hits */
1145 	*ino = 0;
1146 	dirh_ep = NULL;
1147 	for (;;) {
1148 		hit = dirhash_lookup(dirh, name, namelen, &dirh_ep);
1149 		/* if no hit, abort the search */
1150 		if (!hit)
1151 			break;
1152 
1153 		/* check this hit */
1154 		diroffset = dirh_ep->offset;
1155 
1156 		blocknr = diroffset / blocksize;
1157 		blkoff  = diroffset % blocksize;
1158 		error = nilfs_bread(dir_node, blocknr, NOCRED, 0, &bp);
1159 		if (error)
1160 			return EIO;
1161 
1162 		/* read in one dirent */
1163 		pos = (uint8_t *) bp->b_data + blkoff;
1164 		ndirent = (struct nilfs_dir_entry *) pos;
1165 
1166 		DPRINTF(DIRHASH, ("dirhash_lookup\tchecking `%*.*s`\n",
1167 			ndirent->name_len, ndirent->name_len, ndirent->name));
1168 
1169 		/* see if its our entry */
1170 		KASSERT(ndirent->name_len == namelen);
1171 		if (strncmp(ndirent->name, name, namelen) == 0) {
1172 			*found = 1;
1173 			*ino = nilfs_rw64(ndirent->inode);
1174 			brelse(bp, BC_AGE);
1175 			break;
1176 		}
1177 		brelse(bp, BC_AGE);
1178 	}
1179 
1180 	dirhash_put(dir_node->dir_hash);
1181 
1182 	return error;
1183 }
1184 
1185 
1186 int
1187 nilfs_dir_detach(struct nilfs_mount *ump, struct nilfs_node *dir_node, struct nilfs_node *node, struct componentname *cnp)
1188 {
1189 	return EROFS;
1190 }
1191 
1192 
1193 int
1194 nilfs_dir_attach(struct nilfs_mount *ump, struct nilfs_node *dir_node, struct nilfs_node *node, struct vattr *vap, struct componentname *cnp)
1195 {
1196 	return EROFS;
1197 }
1198 
1199 
1200 /* XXX return vnode? */
1201 int
1202 nilfs_create_node(struct vnode *dvp, struct vnode **vpp, struct vattr *vap, struct componentname *cnp)
1203 {
1204 	return EROFS;
1205 }
1206 
1207 
1208 void
1209 nilfs_delete_node(struct nilfs_node *node)
1210 {
1211 }
1212 
1213 
1214