xref: /netbsd/sys/fs/nilfs/nilfs_subr.c (revision 1d7e7b25)
1 /* $NetBSD: nilfs_subr.c,v 1.16 2021/12/05 07:47:40 msaitoh 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 1.16 2021/12/05 07:47:40 msaitoh 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 */
nilfs_get_segnum_of_block(struct nilfs_device * nilfsdev,uint64_t blocknr)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
nilfs_get_segment_range(struct nilfs_device * nilfsdev,uint64_t segnum,uint64_t * seg_start,uint64_t * seg_end)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 
nilfs_calc_mdt_consts(struct nilfs_device * nilfsdev,struct nilfs_mdt * mdt,int entry_size)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
crc32_le(uint32_t crc,const uint8_t * buf,size_t len)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 /* dev reading */
129 static int
nilfs_dev_bread(struct nilfs_device * nilfsdev,uint64_t blocknr,int flags,struct buf ** bpp)130 nilfs_dev_bread(struct nilfs_device *nilfsdev, uint64_t blocknr,
131 	int flags, struct buf **bpp)
132 {
133 	int blk2dev = nilfsdev->blocksize / DEV_BSIZE;
134 
135 	return bread(nilfsdev->devvp, blocknr * blk2dev, nilfsdev->blocksize,
136 		0, bpp);
137 }
138 
139 
140 /* read on a node */
141 int
nilfs_bread(struct nilfs_node * node,uint64_t blocknr,int flags,struct buf ** bpp)142 nilfs_bread(struct nilfs_node *node, uint64_t blocknr,
143 	int flags, struct buf **bpp)
144 {
145 	struct nilfs_device *nilfsdev = node->nilfsdev;
146 	uint64_t vblocknr, pblockno;
147 	int error;
148 
149 	error = nilfs_btree_lookup(node, blocknr, &vblocknr);
150 	if (error)
151 		return error;
152 
153 	/* Read special files through devvp as they have no vnode attached. */
154 	if (node->ino < NILFS_USER_INO && node->ino != NILFS_ROOT_INO) {
155 		error = nilfs_nvtop(node, 1, &vblocknr, &pblockno);
156 		if (error)
157 			return error;
158 		return nilfs_dev_bread(nilfsdev, pblockno, flags, bpp);
159 	}
160 
161 	return bread(node->vnode, vblocknr, node->nilfsdev->blocksize,
162 		flags, bpp);
163 }
164 
165 
166 /* segment-log reading */
167 int
nilfs_get_segment_log(struct nilfs_device * nilfsdev,uint64_t * blocknr,uint64_t * offset,struct buf ** bpp,int len,void * blob)168 nilfs_get_segment_log(struct nilfs_device *nilfsdev, uint64_t *blocknr,
169 	uint64_t *offset, struct buf **bpp, int len, void *blob)
170 {
171 	int blocksize = nilfsdev->blocksize;
172 	int error;
173 
174 	KASSERT(len <= blocksize);
175 
176 	if (*offset + len > blocksize) {
177 		*blocknr = *blocknr + 1;
178 		*offset = 0;
179 	}
180 	if (*offset == 0) {
181 		if (*bpp)
182 			brelse(*bpp, BC_AGE);
183 		/* read in block */
184 		error = nilfs_dev_bread(nilfsdev, *blocknr, 0, bpp);
185 		if (error)
186 			return error;
187 	}
188 	memcpy(blob, ((uint8_t *) (*bpp)->b_data) + *offset, len);
189 	*offset += len;
190 
191 	return 0;
192 }
193 
194 /* -------------------------------------------------------------------------- */
195 
196 /* btree operations */
197 
198 static int
nilfs_btree_lookup_level(struct nilfs_node * node,uint64_t lblocknr,uint64_t btree_vblknr,int level,uint64_t * vblocknr)199 nilfs_btree_lookup_level(struct nilfs_node *node, uint64_t lblocknr,
200 		uint64_t btree_vblknr, int level, uint64_t *vblocknr)
201 {
202 	struct nilfs_device *nilfsdev = node->nilfsdev;
203 	struct nilfs_btree_node *btree_hdr;
204 	struct buf *bp;
205 	uint64_t btree_blknr;
206 	uint64_t *dkeys, *dptrs, child_btree_blk;
207 	uint8_t  *pos;
208 	int i, error, selected;
209 
210 	DPRINTF(TRANSLATE, ("nilfs_btree_lookup_level ino %"PRIu64", "
211 		"lblocknr %"PRIu64", btree_vblknr %"PRIu64", level %d\n",
212 		node->ino, lblocknr, btree_vblknr, level));
213 
214 	/* translate btree_vblknr */
215 	error = nilfs_nvtop(node, 1, &btree_vblknr, &btree_blknr);
216 	if (error)
217 		return error;
218 
219 	/* get our block */
220 	error = nilfs_dev_bread(nilfsdev, btree_blknr, 0, &bp);
221 	if (error) {
222 		return error;
223 	}
224 
225 	btree_hdr = (struct nilfs_btree_node *) bp->b_data;
226 	pos =   (uint8_t *) bp->b_data +
227 		sizeof(struct nilfs_btree_node) +
228 		NILFS_BTREE_NODE_EXTRA_PAD_SIZE;
229 	dkeys = (uint64_t *) pos;
230 	dptrs = dkeys + NILFS_BTREE_NODE_NCHILDREN_MAX(nilfsdev->blocksize);
231 
232 	assert((btree_hdr->bn_flags & NILFS_BTREE_NODE_ROOT) == 0);
233 	assert((btree_hdr->bn_level == level));
234 
235 	/* select matching child XXX could use binary search */
236 	selected = 0;
237 	for (i = 0; i < nilfs_rw16(btree_hdr->bn_nchildren); i++) {
238 		if (dkeys[i] > lblocknr)
239 			break;
240 		selected = i;
241 	}
242 
243 	if (level == 1) {
244 		/* if found it mapped */
245 		if (dkeys[selected] == lblocknr)
246 			*vblocknr = dptrs[selected];
247 		brelse(bp, BC_AGE);
248 		return 0;
249 	}
250 
251 	/* lookup in selected child */
252 	assert(dkeys[selected] <= lblocknr);
253 	child_btree_blk = dptrs[selected];
254 	brelse(bp, BC_AGE);
255 
256 	return nilfs_btree_lookup_level(node, lblocknr,
257 			child_btree_blk, level-1, vblocknr);
258 }
259 
260 
261 /* internal function */
262 static int
nilfs_btree_lookup(struct nilfs_node * node,uint64_t lblocknr,uint64_t * vblocknr)263 nilfs_btree_lookup(struct nilfs_node *node, uint64_t lblocknr,
264 		uint64_t *vblocknr)
265 {
266 	struct nilfs_inode  *inode    = &node->inode;
267 	struct nilfs_btree_node  *btree_hdr;
268 	uint64_t *dkeys, *dptrs, *dtrans;
269 	int i, selected;
270 	int error;
271 
272 	DPRINTF(TRANSLATE, ("nilfs_btree_lookup ino %"PRIu64", "
273 		"lblocknr %"PRIu64"\n", node->ino, lblocknr));
274 
275 	btree_hdr  = (struct nilfs_btree_node *) &inode->i_bmap[0];
276 	dkeys  = &inode->i_bmap[1];
277 	dptrs  = dkeys + NILFS_BTREE_ROOT_NCHILDREN_MAX;
278 	dtrans = &inode->i_bmap[1];
279 
280 	/* SMALL, direct lookup */
281 	*vblocknr = 0;
282 	if ((btree_hdr->bn_flags & NILFS_BMAP_LARGE) == 0) {
283 		if (lblocknr < NILFS_DIRECT_NBLOCKS) {
284 			*vblocknr = dtrans[lblocknr];
285 			return 0;
286 		}
287 		/* not mapped XXX could be considered error here */
288 		return 0;
289 	}
290 
291 	/* LARGE, select matching child; XXX could use binary search */
292 	dtrans = NULL;
293 	error = 0;
294 	selected = 0;
295 	for (i = 0; i < nilfs_rw16(btree_hdr->bn_nchildren); i++) {
296 		if (dkeys[i] > lblocknr)
297 			break;
298 		selected = i;
299 	}
300 
301 	/* if selected key > lblocknr, its not mapped */
302 	if (dkeys[selected] > lblocknr)
303 		return 0;
304 
305 	/* overshooting? then not mapped */
306 	if (selected == nilfs_rw16(btree_hdr->bn_nchildren))
307 		return 0;
308 
309 	/* level should be > 1 or otherwise it should be a direct one */
310 	assert(btree_hdr->bn_level > 1);
311 
312 	/* lookup in selected child */
313 	assert(dkeys[selected] <= lblocknr);
314 	error = nilfs_btree_lookup_level(node, lblocknr,
315 			dptrs[selected], btree_hdr->bn_level-1, vblocknr);
316 
317 	return error;
318 }
319 
320 
321 /* node should be locked on entry to prevent btree changes (unlikely) */
322 int
nilfs_btree_nlookup(struct nilfs_node * node,uint64_t from,uint64_t blks,uint64_t * l2vmap)323 nilfs_btree_nlookup(struct nilfs_node *node, uint64_t from, uint64_t blks,
324 		uint64_t *l2vmap)
325 {
326 	uint64_t lblocknr, *vblocknr;
327 	int i, error;
328 
329 	/* TODO / OPTI multiple translations in one go possible */
330 	error = EINVAL;
331 	for (i = 0; i < blks; i++) {
332 		lblocknr  = from + i;
333 		vblocknr  = l2vmap + i;
334 		error = nilfs_btree_lookup(node, lblocknr, vblocknr);
335 
336 		DPRINTF(TRANSLATE, ("btree_nlookup ino %"PRIu64", "
337 			"lblocknr %"PRIu64" -> %"PRIu64"\n",
338 			node->ino, lblocknr, *vblocknr));
339 		if (error)
340 			break;
341 	}
342 
343 	return error;
344 }
345 
346 /* --------------------------------------------------------------------- */
347 
348 /* vtop operations */
349 
350 /* translate index to a file block number and an entry */
351 void
nilfs_mdt_trans(struct nilfs_mdt * mdt,uint64_t index,uint64_t * blocknr,uint32_t * entry_in_block)352 nilfs_mdt_trans(struct nilfs_mdt *mdt, uint64_t index,
353 	uint64_t *blocknr, uint32_t *entry_in_block)
354 {
355 	uint64_t blknr;
356 	uint64_t group, group_offset, blocknr_in_group;
357 	uint64_t desc_block, desc_offset;
358 
359 	/* calculate our offset in the file */
360 	group             = index / mdt->entries_per_group;
361 	group_offset      = index % mdt->entries_per_group;
362 	desc_block        = group / mdt->groups_per_desc_block;
363 	desc_offset       = group % mdt->groups_per_desc_block;
364 	blocknr_in_group  = group_offset / mdt->entries_per_block;
365 
366 	/* to descgroup offset */
367 	blknr = 1 + desc_block * mdt->blocks_per_desc_block;
368 
369 	/* to group offset */
370 	blknr += desc_offset * mdt->blocks_per_group;
371 
372 	/* to actual file block */
373 	blknr += 1 + blocknr_in_group;
374 
375 	*blocknr        = blknr;
376 	*entry_in_block = group_offset % mdt->entries_per_block;
377 }
378 
379 
380 static int
nilfs_vtop(struct nilfs_device * nilfsdev,uint64_t vblocknr,uint64_t * pblocknr)381 nilfs_vtop(struct nilfs_device *nilfsdev, uint64_t vblocknr, uint64_t *pblocknr)
382 {
383 	struct nilfs_dat_entry *entry;
384 	struct buf *bp;
385 	uint64_t  ldatblknr;
386 	uint32_t  entry_in_block;
387 	int error;
388 
389 	nilfs_mdt_trans(&nilfsdev->dat_mdt, vblocknr,
390 		&ldatblknr, &entry_in_block);
391 
392 	error = nilfs_bread(nilfsdev->dat_node, ldatblknr, 0, &bp);
393 	if (error) {
394 		printf("vtop: can't read in DAT block %"PRIu64"!\n", ldatblknr);
395 		return error;
396 	}
397 
398 	/* get our translation */
399 	entry = ((struct nilfs_dat_entry *) bp->b_data) + entry_in_block;
400 #if 0
401 	printf("\tvblk %4"PRIu64" -> %"PRIu64" for "
402 		"checkpoint %"PRIu64" to %"PRIu64"\n",
403 		vblocknr,
404 		nilfs_rw64(entry->de_blocknr),
405 		nilfs_rw64(entry->de_start),
406 		nilfs_rw64(entry->de_end));
407 #endif
408 
409 	*pblocknr = nilfs_rw64(entry->de_blocknr);
410 	brelse(bp, BC_AGE);
411 
412 	return 0;
413 }
414 
415 
416 int
nilfs_nvtop(struct nilfs_node * node,uint64_t blks,uint64_t * l2vmap,uint64_t * v2pmap)417 nilfs_nvtop(struct nilfs_node *node, uint64_t blks, uint64_t *l2vmap,
418 		uint64_t *v2pmap)
419 {
420 	uint64_t vblocknr, *pblocknr;
421 	int i, error;
422 
423 	/* the DAT inode is the only one not mapped virtual */
424 	if (node->ino == NILFS_DAT_INO) {
425 		memcpy(v2pmap, l2vmap, blks * sizeof(uint64_t));
426 		return 0;
427 	}
428 
429 	/* TODO / OPTI more translations in one go */
430 	error = 0;
431 	for (i = 0; i < blks; i++) {
432 		vblocknr  = l2vmap[i];
433 		pblocknr  = v2pmap + i;
434 		*pblocknr = 0;
435 
436 		/* only translate valid vblocknrs */
437 		if (vblocknr == 0)
438 			continue;
439 		error = nilfs_vtop(node->nilfsdev, vblocknr, pblocknr);
440 		if (error)
441 			break;
442 	}
443 
444 	return error;
445 }
446 
447 /* --------------------------------------------------------------------- */
448 
449 struct nilfs_recover_info {
450 	uint64_t segnum;
451 	uint64_t pseg;
452 
453 	struct nilfs_segment_summary segsum;
454 	struct nilfs_super_root      super_root;
455 	STAILQ_ENTRY(nilfs_recover_info) next;
456 };
457 
458 
459 /*
460  * Helper functions of nilfs_mount() that actually mounts the disc.
461  */
462 static int
nilfs_load_segsum(struct nilfs_device * nilfsdev,struct nilfs_recover_info * ri)463 nilfs_load_segsum(struct nilfs_device *nilfsdev,
464 	struct nilfs_recover_info *ri)
465 {
466 	struct buf *bp;
467 	uint64_t blocknr, offset;
468 	uint32_t segsum_struct_size;
469 	uint32_t magic;
470 	int error;
471 
472 	segsum_struct_size = sizeof(struct nilfs_segment_summary);
473 
474 	/* read in segsum structure */
475 	bp      = NULL;
476 	blocknr = ri->pseg;
477 	offset  = 0;
478 	error = nilfs_get_segment_log(nilfsdev,
479 			&blocknr, &offset, &bp,
480 			segsum_struct_size, (void *) &ri->segsum);
481 	if (error)
482 		goto out;
483 
484 	/* sanity checks */
485 	magic = nilfs_rw32(ri->segsum.ss_magic);
486 	if (magic != NILFS_SEGSUM_MAGIC) {
487 		DPRINTF(VOLUMES, ("nilfs: bad magic in pseg %"PRIu64"\n",
488 			ri->pseg));
489 		error = EINVAL;
490 		goto out;
491 	}
492 
493 	/* TODO check segment summary checksum */
494 	/* TODO check data checksum */
495 
496 out:
497 	if (bp)
498 		brelse(bp, BC_AGE);
499 
500 	return error;
501 }
502 
503 
504 static int
nilfs_load_super_root(struct nilfs_device * nilfsdev,struct nilfs_recover_info * ri)505 nilfs_load_super_root(struct nilfs_device *nilfsdev,
506 	struct nilfs_recover_info *ri)
507 {
508 	struct nilfs_segment_summary *segsum = &ri->segsum;
509 	struct nilfs_super_root *super_root;
510 	struct buf *bp;
511 	uint64_t blocknr, offset;
512 	uint32_t segsum_size, size;
513 	uint32_t nsumblk, nfileblk;
514 	uint32_t super_root_crc, comp_crc;
515 	int off, error;
516 
517 	/* process segment summary */
518 	segsum_size = nilfs_rw32(segsum->ss_sumbytes);
519 	nsumblk     = (segsum_size - 1) / nilfsdev->blocksize + 1;
520 	nfileblk    = nilfs_rw32(segsum->ss_nblocks) - nsumblk;
521 
522 	/* check if there is a superroot */
523 	if ((nilfs_rw16(segsum->ss_flags) & NILFS_SS_SR) == 0) {
524 		DPRINTF(VOLUMES, ("nilfs: no super root in pseg %"PRIu64"\n",
525 			ri->pseg));
526 		return ENOENT;
527 	}
528 
529 	/* get our super root, located at the end of the pseg */
530 	blocknr = ri->pseg + nsumblk + nfileblk - 1;
531 	offset = 0;
532 	size = sizeof(struct nilfs_super_root);
533 	bp = NULL;
534 	error = nilfs_get_segment_log(nilfsdev,
535 			&blocknr, &offset, &bp,
536 			size, (void *) &nilfsdev->super_root);
537 	if (bp)
538 		brelse(bp, BC_AGE);
539 	if (error) {
540 		printf("read in of superroot failed\n");
541 		return EIO;
542 	}
543 
544 	/* check super root crc */
545 	super_root = &nilfsdev->super_root;
546 	super_root_crc = nilfs_rw32(super_root->sr_sum);
547 	off = sizeof(super_root->sr_sum);
548 	comp_crc = crc32_le(nilfs_rw32(nilfsdev->super.s_crc_seed),
549 		(uint8_t *) super_root + off,
550 		NILFS_SR_BYTES - off);
551 	if (super_root_crc != comp_crc) {
552 		DPRINTF(VOLUMES, ("    invalid superroot, likely from old format\n"));
553 		return EINVAL;
554 	}
555 
556 	DPRINTF(VOLUMES, ("    got valid superroot\n"));
557 
558 	return 0;
559 }
560 
561 /*
562  * Search for the last super root recorded.
563  */
564 void
nilfs_search_super_root(struct nilfs_device * nilfsdev)565 nilfs_search_super_root(struct nilfs_device *nilfsdev)
566 {
567 	struct nilfs_super_block *super;
568 	struct nilfs_segment_summary *segsum;
569 	struct nilfs_recover_info *ri, *ori, *i_ri;
570 	STAILQ_HEAD(,nilfs_recover_info) ri_list;
571 	uint64_t seg_start, seg_end, cno;
572 	uint32_t segsum_size;
573 	uint32_t nsumblk, nfileblk;
574 	int error;
575 
576 	STAILQ_INIT(&ri_list);
577 
578 	/* search for last super root */
579 	ri = malloc(sizeof(struct nilfs_recover_info), M_NILFSTEMP, M_WAITOK);
580 	memset(ri, 0, sizeof(struct nilfs_recover_info));
581 
582 	/* if enabled, start from the specified position */
583 	if (0) {
584 		/* start from set start */
585 		nilfsdev->super.s_last_pseg = nilfsdev->super.s_first_data_block;
586 		nilfsdev->super.s_last_cno  = nilfs_rw64(1);
587 	}
588 
589 	ri->pseg   = nilfs_rw64(nilfsdev->super.s_last_pseg); /* blknr */
590 	ri->segnum = nilfs_get_segnum_of_block(nilfsdev, ri->pseg);
591 
592 	error = 0;
593 	cno = nilfs_rw64(nilfsdev->super.s_last_cno);
594 	DPRINTF(VOLUMES, ("nilfs: search_super_root start in pseg %"PRIu64"\n",
595 			ri->pseg));
596 	for (;;) {
597 		DPRINTF(VOLUMES, (" at pseg %"PRIu64"\n", ri->pseg));
598 		error = nilfs_load_segsum(nilfsdev, ri);
599 		if (error)
600 			break;
601 
602 		segsum = &ri->segsum;
603 
604 		/* try to load super root */
605 		if (nilfs_rw16(segsum->ss_flags) & NILFS_SS_SR) {
606 			DPRINTF(VOLUMES, (" try super root\n"));
607 			error = nilfs_load_super_root(nilfsdev, ri);
608 			if (error)
609 				break;	/* confused */
610 			/* wipe current list of ri */
611 			while (!STAILQ_EMPTY(&ri_list)) {
612 				i_ri = STAILQ_FIRST(&ri_list);
613 				STAILQ_REMOVE_HEAD(&ri_list, next);
614 				free(i_ri, M_NILFSTEMP);
615 			}
616 			super = &nilfsdev->super;
617 
618 			super->s_last_pseg = nilfs_rw64(ri->pseg);
619 			super->s_last_cno  = cno++;
620 			super->s_last_seq  = segsum->ss_seq;
621 			super->s_state     = nilfs_rw16(NILFS_VALID_FS);
622 		} else {
623 			STAILQ_INSERT_TAIL(&ri_list, ri, next);
624 			ori = ri;
625 			ri = malloc(sizeof(struct nilfs_recover_info),
626 				M_NILFSTEMP, M_WAITOK);
627 			memset(ri, 0, sizeof(struct nilfs_recover_info));
628 			ri->segnum = ori->segnum;
629 			ri->pseg   = ori->pseg;
630 			/* segsum keeps pointing to the `old' ri */
631 		}
632 
633 		/* continue to the next pseg */
634 		segsum_size = nilfs_rw32(segsum->ss_sumbytes);
635 		nsumblk     = (segsum_size - 1) / nilfsdev->blocksize + 1;
636 		nfileblk    = nilfs_rw32(segsum->ss_nblocks) - nsumblk;
637 
638 		/* calculate next partial segment location */
639 		ri->pseg += nsumblk + nfileblk;
640 
641 		/* did we reach the end of the segment? if so, go to the next */
642 		nilfs_get_segment_range(nilfsdev, ri->segnum, &seg_start, &seg_end);
643 		if (ri->pseg >= seg_end)
644 			ri->pseg = nilfs_rw64(segsum->ss_next);
645 		ri->segnum = nilfs_get_segnum_of_block(nilfsdev, ri->pseg);
646 	}
647 
648 	/*
649 	 * XXX No roll-forward yet of the remaining partial segments.
650 	 */
651 
652 	/* wipe current list of ri */
653 	while (!STAILQ_EMPTY(&ri_list)) {
654 		i_ri = STAILQ_FIRST(&ri_list);
655 		STAILQ_REMOVE_HEAD(&ri_list, next);
656 		printf("nilfs: ignoring pseg at %"PRIu64"\n", i_ri->pseg);
657 		free(i_ri, M_NILFSTEMP);
658 	}
659 	free(ri, M_NILFSTEMP);
660 }
661 
662 /* --------------------------------------------------------------------- */
663 
664 int
nilfs_get_node_raw(struct nilfs_device * nilfsdev,struct nilfs_mount * ump,uint64_t ino,struct nilfs_inode * inode,struct nilfs_node ** nodep)665 nilfs_get_node_raw(struct nilfs_device *nilfsdev, struct nilfs_mount *ump,
666 	uint64_t ino, struct nilfs_inode *inode, struct nilfs_node **nodep)
667 {
668 	struct nilfs_node *node;
669 
670 	*nodep = NULL;
671 
672 	node = pool_get(&nilfs_node_pool, PR_WAITOK);
673 	memset(node, 0, sizeof(struct nilfs_node));
674 
675 	/* crosslink */
676 	node->ump      = ump;
677 	node->nilfsdev = nilfsdev;
678 
679 	/* initiase nilfs node */
680 	node->ino   = ino;
681 	node->inode = *inode;
682 	node->lockf = NULL;
683 
684 	/* initialise locks */
685 	mutex_init(&node->node_mutex, MUTEX_DEFAULT, IPL_NONE);
686 	cv_init(&node->node_lock, "nilfsnlk");
687 
688 	/* fixup inode size for system nodes */
689 	if ((ino < NILFS_USER_INO) && (ino != NILFS_ROOT_INO)) {
690 		DPRINTF(VOLUMES, ("NEED TO GET my size for inode %"PRIu64"?\n",
691 			ino));
692 		/* for now set it to maximum, -1 is illegal */
693 		DPRINTF(VOLUMES, ("  current size of inode is %"PRIu64"\n", inode->i_size));
694 		inode->i_size = nilfs_rw64(((uint64_t) -2));
695 	}
696 
697 	/* return node */
698 	*nodep = node;
699 	return 0;
700 }
701 
702 void
nilfs_dispose_node(struct nilfs_node ** nodep)703 nilfs_dispose_node(struct nilfs_node **nodep)
704 {
705 	struct nilfs_node *node;
706 
707 	/* protect against rogue values */
708 	if (!*nodep)
709 		return;
710 
711 	node = *nodep;
712 
713 	/* remove dirhash if present */
714 	dirhash_purge(&node->dir_hash);
715 
716 	/* destroy our locks */
717 	mutex_destroy(&node->node_mutex);
718 	cv_destroy(&node->node_lock);
719 
720 	/* free our associated memory */
721 	pool_put(&nilfs_node_pool, node);
722 
723 	*nodep = NULL;
724 }
725 
726 
727 void
nilfs_itimes(struct nilfs_node * node,struct timespec * acc,struct timespec * mod,struct timespec * birth)728 nilfs_itimes(struct nilfs_node *node, struct timespec *acc,
729 	struct timespec *mod, struct timespec *birth)
730 {
731 }
732 
733 
734 int
nilfs_update(struct vnode * node,struct timespec * acc,struct timespec * mod,struct timespec * birth,int updflags)735 nilfs_update(struct vnode *node, struct timespec *acc,
736 	struct timespec *mod, struct timespec *birth, int updflags)
737 {
738 	return EROFS;
739 }
740 
741 
742 int
nilfs_chsize(struct vnode * vp,u_quad_t newsize,kauth_cred_t cred)743 nilfs_chsize(struct vnode *vp, u_quad_t newsize, kauth_cred_t cred)
744 {
745 	return EROFS;
746 }
747 
748 
749 
750 int
nilfs_grow_node(struct nilfs_node * node,uint64_t new_size)751 nilfs_grow_node(struct nilfs_node *node, uint64_t new_size)
752 {
753 	return EROFS;
754 }
755 
756 
757 int
nilfs_shrink_node(struct nilfs_node * node,uint64_t new_size)758 nilfs_shrink_node(struct nilfs_node *node, uint64_t new_size)
759 {
760 	return EROFS;
761 }
762 
763 
764 static int
dirhash_fill(struct nilfs_node * dir_node)765 dirhash_fill(struct nilfs_node *dir_node)
766 {
767 	struct vnode *dvp = dir_node->vnode;
768 	struct dirhash *dirh;
769 	struct nilfs_dir_entry *ndirent;
770 	struct dirent dirent;
771 	struct buf *bp;
772 	uint64_t file_size, diroffset, blkoff;
773 	uint64_t blocknr;
774 	uint32_t blocksize = dir_node->nilfsdev->blocksize;
775 	uint8_t *pos, name_len;
776 	int error;
777 
778 	DPRINTF(CALL, ("dirhash_fill called\n"));
779 
780 	if (dvp->v_type != VDIR)
781 		return ENOTDIR;
782 
783 	/* make sure we have a dirhash to work on */
784 	dirh = dir_node->dir_hash;
785 	KASSERT(dirh);
786 	KASSERT(dirh->refcnt > 0);
787 
788 	if (dirh->flags & DIRH_BROKEN)
789 		return EIO;
790 
791 	if (dirh->flags & DIRH_COMPLETE)
792 		return 0;
793 
794 	DPRINTF(DIRHASH, ("Filling directory hash\n"));
795 
796 	/* make sure we have a clean dirhash to add to */
797 	dirhash_purge_entries(dirh);
798 
799 	/* get directory filesize */
800 	file_size = nilfs_rw64(dir_node->inode.i_size);
801 
802 	/* walk the directory */
803 	error = 0;
804 	diroffset = 0;
805 
806 	blocknr = diroffset / blocksize;
807 	blkoff  = diroffset % blocksize;
808 	error = nilfs_bread(dir_node, blocknr, 0, &bp);
809 	if (error) {
810 		dirh->flags |= DIRH_BROKEN;
811 		dirhash_purge_entries(dirh);
812 		return EIO;
813 	}
814 	while (diroffset < file_size) {
815 		DPRINTF(READDIR, ("filldir : offset = %"PRIu64"\n",
816 			diroffset));
817 		if (blkoff >= blocksize) {
818 			blkoff = 0; blocknr++;
819 			brelse(bp, BC_AGE);
820 			error = nilfs_bread(dir_node, blocknr, 0, &bp);
821 			if (error) {
822 				dirh->flags |= DIRH_BROKEN;
823 				dirhash_purge_entries(dirh);
824 				return EIO;
825 			}
826 		}
827 
828 		/* read in one dirent */
829 		pos = (uint8_t *) bp->b_data + blkoff;
830 		ndirent = (struct nilfs_dir_entry *) pos;
831 		name_len = ndirent->name_len;
832 
833 		memset(&dirent, 0, sizeof(struct dirent));
834 		dirent.d_fileno = nilfs_rw64(ndirent->inode);
835 		dirent.d_type   = ndirent->file_type;	/* 1:1 ? */
836 		dirent.d_namlen = name_len;
837 		strncpy(dirent.d_name, ndirent->name, name_len);
838 		dirent.d_reclen = _DIRENT_SIZE(&dirent);
839 		DPRINTF(DIRHASH, ("copying `%*.*s`\n", name_len,
840 			name_len, dirent.d_name));
841 
842 		/* XXX is it deleted? extra free space? */
843 		dirhash_enter(dirh, &dirent, diroffset,
844 			nilfs_rw16(ndirent->rec_len), 0);
845 
846 		/* advance */
847 		diroffset += nilfs_rw16(ndirent->rec_len);
848 		blkoff    += nilfs_rw16(ndirent->rec_len);
849 	}
850 	brelse(bp, BC_AGE);
851 
852 	dirh->flags |= DIRH_COMPLETE;
853 
854 	return 0;
855 }
856 
857 
858 int
nilfs_lookup_name_in_dir(struct vnode * dvp,const char * name,int namelen,uint64_t * ino,int * found)859 nilfs_lookup_name_in_dir(struct vnode *dvp, const char *name, int namelen,
860 		uint64_t *ino, int *found)
861 {
862 	struct nilfs_node	*dir_node = VTOI(dvp);
863 	struct nilfs_dir_entry *ndirent;
864 	struct dirhash		*dirh;
865 	struct dirhash_entry	*dirh_ep;
866 	struct buf *bp;
867 	uint64_t diroffset, blkoff;
868 	uint64_t blocknr;
869 	uint32_t blocksize = dir_node->nilfsdev->blocksize;
870 	uint8_t *pos;
871 	int hit, error;
872 
873 	/* set default return */
874 	*found = 0;
875 
876 	/* get our dirhash and make sure its read in */
877 	dirhash_get(&dir_node->dir_hash);
878 	error = dirhash_fill(dir_node);
879 	if (error) {
880 		dirhash_put(dir_node->dir_hash);
881 		return error;
882 	}
883 	dirh = dir_node->dir_hash;
884 
885 	/* allocate temporary space for fid */
886 
887 	DPRINTF(DIRHASH, ("dirhash_lookup looking for `%*.*s`\n",
888 		namelen, namelen, name));
889 
890 	/* search our dirhash hits */
891 	*ino = 0;
892 	dirh_ep = NULL;
893 	for (;;) {
894 		hit = dirhash_lookup(dirh, name, namelen, &dirh_ep);
895 		/* if no hit, abort the search */
896 		if (!hit)
897 			break;
898 
899 		/* check this hit */
900 		diroffset = dirh_ep->offset;
901 
902 		blocknr = diroffset / blocksize;
903 		blkoff  = diroffset % blocksize;
904 		error = nilfs_bread(dir_node, blocknr, 0, &bp);
905 		if (error)
906 			return EIO;
907 
908 		/* read in one dirent */
909 		pos = (uint8_t *) bp->b_data + blkoff;
910 		ndirent = (struct nilfs_dir_entry *) pos;
911 
912 		DPRINTF(DIRHASH, ("dirhash_lookup\tchecking `%*.*s`\n",
913 			ndirent->name_len, ndirent->name_len, ndirent->name));
914 
915 		/* see if its our entry */
916 		KASSERT(ndirent->name_len == namelen);
917 		if (strncmp(ndirent->name, name, namelen) == 0) {
918 			*found = 1;
919 			*ino = nilfs_rw64(ndirent->inode);
920 			brelse(bp, BC_AGE);
921 			break;
922 		}
923 		brelse(bp, BC_AGE);
924 	}
925 
926 	dirhash_put(dir_node->dir_hash);
927 
928 	return error;
929 }
930 
931 
932 int
nilfs_dir_detach(struct nilfs_mount * ump,struct nilfs_node * dir_node,struct nilfs_node * node,struct componentname * cnp)933 nilfs_dir_detach(struct nilfs_mount *ump, struct nilfs_node *dir_node, struct nilfs_node *node, struct componentname *cnp)
934 {
935 	return EROFS;
936 }
937 
938 
939 int
nilfs_dir_attach(struct nilfs_mount * ump,struct nilfs_node * dir_node,struct nilfs_node * node,struct vattr * vap,struct componentname * cnp)940 nilfs_dir_attach(struct nilfs_mount *ump, struct nilfs_node *dir_node, struct nilfs_node *node, struct vattr *vap, struct componentname *cnp)
941 {
942 	return EROFS;
943 }
944 
945 
946 /* XXX return vnode? */
947 int
nilfs_create_node(struct vnode * dvp,struct vnode ** vpp,struct vattr * vap,struct componentname * cnp)948 nilfs_create_node(struct vnode *dvp, struct vnode **vpp, struct vattr *vap, struct componentname *cnp)
949 {
950 	return EROFS;
951 }
952 
953 
954 void
nilfs_delete_node(struct nilfs_node * node)955 nilfs_delete_node(struct nilfs_node *node)
956 {
957 }
958 
959 
960