xref: /netbsd-src/sys/fs/udf/udf_subr.c (revision fff57c5525bbe431aee7bdb3983954f0627a42cb)
1 /* $NetBSD: udf_subr.c,v 1.49 2008/05/19 23:48:04 christos Exp $ */
2 
3 /*
4  * Copyright (c) 2006, 2008 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 
30 #include <sys/cdefs.h>
31 #ifndef lint
32 __KERNEL_RCSID(0, "$NetBSD: udf_subr.c,v 1.49 2008/05/19 23:48:04 christos Exp $");
33 #endif /* not lint */
34 
35 
36 #if defined(_KERNEL_OPT)
37 #include "opt_quota.h"
38 #include "opt_compat_netbsd.h"
39 #endif
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/sysctl.h>
44 #include <sys/namei.h>
45 #include <sys/proc.h>
46 #include <sys/kernel.h>
47 #include <sys/vnode.h>
48 #include <miscfs/genfs/genfs_node.h>
49 #include <sys/mount.h>
50 #include <sys/buf.h>
51 #include <sys/file.h>
52 #include <sys/device.h>
53 #include <sys/disklabel.h>
54 #include <sys/ioctl.h>
55 #include <sys/malloc.h>
56 #include <sys/dirent.h>
57 #include <sys/stat.h>
58 #include <sys/conf.h>
59 #include <sys/kauth.h>
60 #include <fs/unicode.h>
61 #include <dev/clock_subr.h>
62 
63 #include <fs/udf/ecma167-udf.h>
64 #include <fs/udf/udf_mount.h>
65 
66 #if defined(_KERNEL_OPT)
67 #include "opt_udf.h"
68 #endif
69 
70 #include "udf.h"
71 #include "udf_subr.h"
72 #include "udf_bswap.h"
73 
74 
75 #define VTOI(vnode) ((struct udf_node *) (vnode)->v_data)
76 
77 #define UDF_SET_SYSTEMFILE(vp) \
78 	/* XXXAD Is the vnode locked? */	\
79 	(vp)->v_vflag |= VV_SYSTEM;		\
80 	vref(vp);			\
81 	vput(vp);			\
82 
83 extern int syncer_maxdelay;     /* maximum delay time */
84 extern int (**udf_vnodeop_p)(void *);
85 
86 /* --------------------------------------------------------------------- */
87 
88 //#ifdef DEBUG
89 #if 1
90 
91 #if 0
92 static void
93 udf_dumpblob(boid *blob, uint32_t dlen)
94 {
95 	int i, j;
96 
97 	printf("blob = %p\n", blob);
98 	printf("dump of %d bytes\n", dlen);
99 
100 	for (i = 0; i < dlen; i+ = 16) {
101 		printf("%04x ", i);
102 		for (j = 0; j < 16; j++) {
103 			if (i+j < dlen) {
104 				printf("%02x ", blob[i+j]);
105 			} else {
106 				printf("   ");
107 			}
108 		}
109 		for (j = 0; j < 16; j++) {
110 			if (i+j < dlen) {
111 				if (blob[i+j]>32 && blob[i+j]! = 127) {
112 					printf("%c", blob[i+j]);
113 				} else {
114 					printf(".");
115 				}
116 			}
117 		}
118 		printf("\n");
119 	}
120 	printf("\n");
121 	Debugger();
122 }
123 #endif
124 
125 static void
126 udf_dump_discinfo(struct udf_mount *ump)
127 {
128 	char   bits[128];
129 	struct mmc_discinfo *di = &ump->discinfo;
130 
131 	if ((udf_verbose & UDF_DEBUG_VOLUMES) == 0)
132 		return;
133 
134 	printf("Device/media info  :\n");
135 	printf("\tMMC profile        0x%02x\n", di->mmc_profile);
136 	printf("\tderived class      %d\n", di->mmc_class);
137 	printf("\tsector size        %d\n", di->sector_size);
138 	printf("\tdisc state         %d\n", di->disc_state);
139 	printf("\tlast ses state     %d\n", di->last_session_state);
140 	printf("\tbg format state    %d\n", di->bg_format_state);
141 	printf("\tfrst track         %d\n", di->first_track);
142 	printf("\tfst on last ses    %d\n", di->first_track_last_session);
143 	printf("\tlst on last ses    %d\n", di->last_track_last_session);
144 	printf("\tlink block penalty %d\n", di->link_block_penalty);
145 	bitmask_snprintf(di->disc_flags, MMC_DFLAGS_FLAGBITS, bits,
146 		sizeof(bits));
147 	printf("\tdisc flags         %s\n", bits);
148 	printf("\tdisc id            %x\n", di->disc_id);
149 	printf("\tdisc barcode       %"PRIx64"\n", di->disc_barcode);
150 
151 	printf("\tnum sessions       %d\n", di->num_sessions);
152 	printf("\tnum tracks         %d\n", di->num_tracks);
153 
154 	bitmask_snprintf(di->mmc_cur, MMC_CAP_FLAGBITS, bits, sizeof(bits));
155 	printf("\tcapabilities cur   %s\n", bits);
156 	bitmask_snprintf(di->mmc_cap, MMC_CAP_FLAGBITS, bits, sizeof(bits));
157 	printf("\tcapabilities cap   %s\n", bits);
158 }
159 #else
160 #define udf_dump_discinfo(a);
161 #endif
162 
163 
164 /* --------------------------------------------------------------------- */
165 
166 /* not called often */
167 int
168 udf_update_discinfo(struct udf_mount *ump)
169 {
170 	struct vnode *devvp = ump->devvp;
171 	struct partinfo dpart;
172 	struct mmc_discinfo *di;
173 	int error;
174 
175 	DPRINTF(VOLUMES, ("read/update disc info\n"));
176 	di = &ump->discinfo;
177 	memset(di, 0, sizeof(struct mmc_discinfo));
178 
179 	/* check if we're on a MMC capable device, i.e. CD/DVD */
180 	error = VOP_IOCTL(devvp, MMCGETDISCINFO, di, FKIOCTL, NOCRED);
181 	if (error == 0) {
182 		udf_dump_discinfo(ump);
183 		return 0;
184 	}
185 
186 	/* disc partition support */
187 	error = VOP_IOCTL(devvp, DIOCGPART, &dpart, FREAD, NOCRED);
188 	if (error)
189 		return ENODEV;
190 
191 	/* set up a disc info profile for partitions */
192 	di->mmc_profile		= 0x01;	/* disc type */
193 	di->mmc_class		= MMC_CLASS_DISC;
194 	di->disc_state		= MMC_STATE_CLOSED;
195 	di->last_session_state	= MMC_STATE_CLOSED;
196 	di->bg_format_state	= MMC_BGFSTATE_COMPLETED;
197 	di->link_block_penalty	= 0;
198 
199 	di->mmc_cur     = MMC_CAP_RECORDABLE | MMC_CAP_REWRITABLE |
200 		MMC_CAP_ZEROLINKBLK | MMC_CAP_HW_DEFECTFREE;
201 	di->mmc_cap    = di->mmc_cur;
202 	di->disc_flags = MMC_DFLAGS_UNRESTRICTED;
203 
204 	/* TODO problem with last_possible_lba on resizable VND; request */
205 	di->last_possible_lba = dpart.part->p_size;
206 	di->sector_size       = dpart.disklab->d_secsize;
207 
208 	di->num_sessions = 1;
209 	di->num_tracks   = 1;
210 
211 	di->first_track  = 1;
212 	di->first_track_last_session = di->last_track_last_session = 1;
213 
214 	udf_dump_discinfo(ump);
215 	return 0;
216 }
217 
218 
219 int
220 udf_update_trackinfo(struct udf_mount *ump, struct mmc_trackinfo *ti)
221 {
222 	struct vnode *devvp = ump->devvp;
223 	struct mmc_discinfo *di = &ump->discinfo;
224 	int error, class;
225 
226 	DPRINTF(VOLUMES, ("read track info\n"));
227 
228 	class = di->mmc_class;
229 	if (class != MMC_CLASS_DISC) {
230 		/* tracknr specified in struct ti */
231 		error = VOP_IOCTL(devvp, MMCGETTRACKINFO, ti, FKIOCTL, NOCRED);
232 		return error;
233 	}
234 
235 	/* disc partition support */
236 	if (ti->tracknr != 1)
237 		return EIO;
238 
239 	/* create fake ti (TODO check for resized vnds) */
240 	ti->sessionnr  = 1;
241 
242 	ti->track_mode = 0;	/* XXX */
243 	ti->data_mode  = 0;	/* XXX */
244 	ti->flags = MMC_TRACKINFO_LRA_VALID | MMC_TRACKINFO_NWA_VALID;
245 
246 	ti->track_start    = 0;
247 	ti->packet_size    = 1;
248 
249 	/* TODO support for resizable vnd */
250 	ti->track_size    = di->last_possible_lba;
251 	ti->next_writable = di->last_possible_lba;
252 	ti->last_recorded = ti->next_writable;
253 	ti->free_blocks   = 0;
254 
255 	return 0;
256 }
257 
258 
259 int
260 udf_setup_writeparams(struct udf_mount *ump)
261 {
262 	struct mmc_writeparams mmc_writeparams;
263 	int error;
264 
265 	if (ump->discinfo.mmc_class == MMC_CLASS_DISC)
266 		return 0;
267 
268 	/*
269 	 * only CD burning normally needs setting up, but other disc types
270 	 * might need other settings to be made. The MMC framework will set up
271 	 * the nessisary recording parameters according to the disc
272 	 * characteristics read in. Modifications can be made in the discinfo
273 	 * structure passed to change the nature of the disc.
274 	 */
275 
276 	memset(&mmc_writeparams, 0, sizeof(struct mmc_writeparams));
277 	mmc_writeparams.mmc_class  = ump->discinfo.mmc_class;
278 	mmc_writeparams.mmc_cur    = ump->discinfo.mmc_cur;
279 
280 	/*
281 	 * UDF dictates first track to determine track mode for the whole
282 	 * disc. [UDF 1.50/6.10.1.1, UDF 1.50/6.10.2.1]
283 	 * To prevent problems with a `reserved' track in front we start with
284 	 * the 2nd track and if that is not valid, go for the 1st.
285 	 */
286 	mmc_writeparams.tracknr = 2;
287 	mmc_writeparams.data_mode  = MMC_DATAMODE_DEFAULT;	/* XA disc */
288 	mmc_writeparams.track_mode = MMC_TRACKMODE_DEFAULT;	/* data */
289 
290 	error = VOP_IOCTL(ump->devvp, MMCSETUPWRITEPARAMS, &mmc_writeparams,
291 			FKIOCTL, NOCRED);
292 	if (error) {
293 		mmc_writeparams.tracknr = 1;
294 		error = VOP_IOCTL(ump->devvp, MMCSETUPWRITEPARAMS,
295 				&mmc_writeparams, FKIOCTL, NOCRED);
296 	}
297 	return error;
298 }
299 
300 
301 int
302 udf_synchronise_caches(struct udf_mount *ump)
303 {
304 	struct mmc_op mmc_op;
305 
306 	DPRINTF(CALL, ("udf_synchronise_caches()\n"));
307 
308 	if (ump->vfs_mountp->mnt_flag & MNT_RDONLY)
309 		return 0;
310 
311 	/* discs are done now */
312 	if (ump->discinfo.mmc_class == MMC_CLASS_DISC)
313 		return 0;
314 
315 	bzero(&mmc_op, sizeof(struct mmc_op));
316 	mmc_op.operation = MMC_OP_SYNCHRONISECACHE;
317 
318 	/* ignore return code */
319 	(void) VOP_IOCTL(ump->devvp, MMCOP, &mmc_op, FKIOCTL, NOCRED);
320 
321 	return 0;
322 }
323 
324 /* --------------------------------------------------------------------- */
325 
326 /* track/session searching for mounting */
327 int
328 udf_search_tracks(struct udf_mount *ump, struct udf_args *args,
329 		  int *first_tracknr, int *last_tracknr)
330 {
331 	struct mmc_trackinfo trackinfo;
332 	uint32_t tracknr, start_track, num_tracks;
333 	int error;
334 
335 	/* if negative, sessionnr is relative to last session */
336 	if (args->sessionnr < 0) {
337 		args->sessionnr += ump->discinfo.num_sessions;
338 	}
339 
340 	/* sanity */
341 	if (args->sessionnr < 0)
342 		args->sessionnr = 0;
343 	if (args->sessionnr > ump->discinfo.num_sessions)
344 		args->sessionnr = ump->discinfo.num_sessions;
345 
346 	/* search the tracks for this session, zero session nr indicates last */
347 	if (args->sessionnr == 0)
348 		args->sessionnr = ump->discinfo.num_sessions;
349 	if (ump->discinfo.last_session_state == MMC_STATE_EMPTY)
350 		args->sessionnr--;
351 
352 	/* sanity again */
353 	if (args->sessionnr < 0)
354 		args->sessionnr = 0;
355 
356 	/* search the first and last track of the specified session */
357 	num_tracks  = ump->discinfo.num_tracks;
358 	start_track = ump->discinfo.first_track;
359 
360 	/* search for first track of this session */
361 	for (tracknr = start_track; tracknr <= num_tracks; tracknr++) {
362 		/* get track info */
363 		trackinfo.tracknr = tracknr;
364 		error = udf_update_trackinfo(ump, &trackinfo);
365 		if (error)
366 			return error;
367 
368 		if (trackinfo.sessionnr == args->sessionnr)
369 			break;
370 	}
371 	*first_tracknr = tracknr;
372 
373 	/* search for last track of this session */
374 	for (;tracknr <= num_tracks; tracknr++) {
375 		/* get track info */
376 		trackinfo.tracknr = tracknr;
377 		error = udf_update_trackinfo(ump, &trackinfo);
378 		if (error || (trackinfo.sessionnr != args->sessionnr)) {
379 			tracknr--;
380 			break;
381 		}
382 	}
383 	if (tracknr > num_tracks)
384 		tracknr--;
385 
386 	*last_tracknr = tracknr;
387 
388 	if (*last_tracknr < *first_tracknr) {
389 		printf( "udf_search_tracks: sanity check on drive+disc failed, "
390 			"drive returned garbage\n");
391 		return EINVAL;
392 	}
393 
394 	assert(*last_tracknr >= *first_tracknr);
395 	return 0;
396 }
397 
398 
399 /*
400  * NOTE: this is the only routine in this file that directly peeks into the
401  * metadata file but since its at a larval state of the mount it can't hurt.
402  *
403  * XXX candidate for udf_allocation.c
404  * XXX clean me up!, change to new node reading code.
405  */
406 
407 static void
408 udf_check_track_metadata_overlap(struct udf_mount *ump,
409 	struct mmc_trackinfo *trackinfo)
410 {
411 	struct part_desc *part;
412 	struct file_entry      *fe;
413 	struct extfile_entry   *efe;
414 	struct short_ad        *s_ad;
415 	struct long_ad         *l_ad;
416 	uint32_t track_start, track_end;
417 	uint32_t phys_part_start, phys_part_end, part_start, part_end;
418 	uint32_t sector_size, len, alloclen, plb_num;
419 	uint8_t *pos;
420 	int addr_type, icblen, icbflags, flags;
421 
422 	/* get our track extents */
423 	track_start = trackinfo->track_start;
424 	track_end   = track_start + trackinfo->track_size;
425 
426 	/* get our base partition extent */
427 	part = ump->partitions[ump->metadata_part];
428 	phys_part_start = udf_rw32(part->start_loc);
429 	phys_part_end   = phys_part_start + udf_rw32(part->part_len);
430 
431 	/* no use if its outside the physical partition */
432 	if ((phys_part_start >= track_end) || (phys_part_end < track_start))
433 		return;
434 
435 	/*
436 	 * now follow all extents in the fe/efe to see if they refer to this
437 	 * track
438 	 */
439 
440 	sector_size = ump->discinfo.sector_size;
441 
442 	/* XXX should we claim exclusive access to the metafile ? */
443 	/* TODO: move to new node read code */
444 	fe  = ump->metadata_node->fe;
445 	efe = ump->metadata_node->efe;
446 	if (fe) {
447 		alloclen = udf_rw32(fe->l_ad);
448 		pos      = &fe->data[0] + udf_rw32(fe->l_ea);
449 		icbflags = udf_rw16(fe->icbtag.flags);
450 	} else {
451 		assert(efe);
452 		alloclen = udf_rw32(efe->l_ad);
453 		pos      = &efe->data[0] + udf_rw32(efe->l_ea);
454 		icbflags = udf_rw16(efe->icbtag.flags);
455 	}
456 	addr_type = icbflags & UDF_ICB_TAG_FLAGS_ALLOC_MASK;
457 
458 	while (alloclen) {
459 		if (addr_type == UDF_ICB_SHORT_ALLOC) {
460 			icblen = sizeof(struct short_ad);
461 			s_ad   = (struct short_ad *) pos;
462 			len        = udf_rw32(s_ad->len);
463 			plb_num    = udf_rw32(s_ad->lb_num);
464 		} else {
465 			/* should not be present, but why not */
466 			icblen = sizeof(struct long_ad);
467 			l_ad   = (struct long_ad *) pos;
468 			len        = udf_rw32(l_ad->len);
469 			plb_num    = udf_rw32(l_ad->loc.lb_num);
470 			/* pvpart_num = udf_rw16(l_ad->loc.part_num); */
471 		}
472 		/* process extent */
473 		flags   = UDF_EXT_FLAGS(len);
474 		len     = UDF_EXT_LEN(len);
475 
476 		part_start = phys_part_start + plb_num;
477 		part_end   = part_start + (len / sector_size);
478 
479 		if ((part_start >= track_start) && (part_end <= track_end)) {
480 			/* extent is enclosed within this track */
481 			ump->metadata_track = *trackinfo;
482 			return;
483 		}
484 
485 		pos        += icblen;
486 		alloclen   -= icblen;
487 	}
488 }
489 
490 
491 int
492 udf_search_writing_tracks(struct udf_mount *ump)
493 {
494 	struct mmc_trackinfo trackinfo;
495 	struct part_desc *part;
496 	uint32_t tracknr, start_track, num_tracks;
497 	uint32_t track_start, track_end, part_start, part_end;
498 	int error;
499 
500 	/*
501 	 * in the CD/(HD)DVD/BD recordable device model a few tracks within
502 	 * the last session might be open but in the UDF device model at most
503 	 * three tracks can be open: a reserved track for delayed ISO VRS
504 	 * writing, a data track and a metadata track. We search here for the
505 	 * data track and the metadata track. Note that the reserved track is
506 	 * troublesome but can be detected by its small size of < 512 sectors.
507 	 */
508 
509 	num_tracks  = ump->discinfo.num_tracks;
510 	start_track = ump->discinfo.first_track;
511 
512 	/* fetch info on first and possibly only track */
513 	trackinfo.tracknr = start_track;
514 	error = udf_update_trackinfo(ump, &trackinfo);
515 	if (error)
516 		return error;
517 
518 	/* copy results to our mount point */
519 	ump->data_track     = trackinfo;
520 	ump->metadata_track = trackinfo;
521 
522 	/* if not sequential, we're done */
523 	if (num_tracks == 1)
524 		return 0;
525 
526 	for (tracknr = start_track;tracknr <= num_tracks; tracknr++) {
527 		/* get track info */
528 		trackinfo.tracknr = tracknr;
529 		error = udf_update_trackinfo(ump, &trackinfo);
530 		if (error)
531 			return error;
532 
533 		if ((trackinfo.flags & MMC_TRACKINFO_NWA_VALID) == 0)
534 			continue;
535 
536 		track_start = trackinfo.track_start;
537 		track_end   = track_start + trackinfo.track_size;
538 
539 		/* check for overlap on data partition */
540 		part = ump->partitions[ump->data_part];
541 		part_start = udf_rw32(part->start_loc);
542 		part_end   = part_start + udf_rw32(part->part_len);
543 		if ((part_start < track_end) && (part_end > track_start)) {
544 			ump->data_track = trackinfo;
545 			/* TODO check if UDF partition data_part is writable */
546 		}
547 
548 		/* check for overlap on metadata partition */
549 		if ((ump->meta_alloc == UDF_ALLOC_METASEQUENTIAL) ||
550 		    (ump->meta_alloc == UDF_ALLOC_METABITMAP)) {
551 			udf_check_track_metadata_overlap(ump, &trackinfo);
552 		} else {
553 			ump->metadata_track = trackinfo;
554 		}
555 	}
556 
557 	if ((ump->data_track.flags & MMC_TRACKINFO_NWA_VALID) == 0)
558 		return EROFS;
559 
560 	if ((ump->metadata_track.flags & MMC_TRACKINFO_NWA_VALID) == 0)
561 		return EROFS;
562 
563 	return 0;
564 }
565 
566 /* --------------------------------------------------------------------- */
567 
568 /*
569  * Check if the blob starts with a good UDF tag. Tags are protected by a
570  * checksum over the reader except one byte at position 4 that is the checksum
571  * itself.
572  */
573 
574 int
575 udf_check_tag(void *blob)
576 {
577 	struct desc_tag *tag = blob;
578 	uint8_t *pos, sum, cnt;
579 
580 	/* check TAG header checksum */
581 	pos = (uint8_t *) tag;
582 	sum = 0;
583 
584 	for(cnt = 0; cnt < 16; cnt++) {
585 		if (cnt != 4)
586 			sum += *pos;
587 		pos++;
588 	}
589 	if (sum != tag->cksum) {
590 		/* bad tag header checksum; this is not a valid tag */
591 		return EINVAL;
592 	}
593 
594 	return 0;
595 }
596 
597 
598 /*
599  * check tag payload will check descriptor CRC as specified.
600  * If the descriptor is too long, it will return EIO otherwise EINVAL.
601  */
602 
603 int
604 udf_check_tag_payload(void *blob, uint32_t max_length)
605 {
606 	struct desc_tag *tag = blob;
607 	uint16_t crc, crc_len;
608 
609 	crc_len = udf_rw16(tag->desc_crc_len);
610 
611 	/* check payload CRC if applicable */
612 	if (crc_len == 0)
613 		return 0;
614 
615 	if (crc_len > max_length)
616 		return EIO;
617 
618 	crc = udf_cksum(((uint8_t *) tag) + UDF_DESC_TAG_LENGTH, crc_len);
619 	if (crc != udf_rw16(tag->desc_crc)) {
620 		/* bad payload CRC; this is a broken tag */
621 		return EINVAL;
622 	}
623 
624 	return 0;
625 }
626 
627 
628 void
629 udf_validate_tag_sum(void *blob)
630 {
631 	struct desc_tag *tag = blob;
632 	uint8_t *pos, sum, cnt;
633 
634 	/* calculate TAG header checksum */
635 	pos = (uint8_t *) tag;
636 	sum = 0;
637 
638 	for(cnt = 0; cnt < 16; cnt++) {
639 		if (cnt != 4) sum += *pos;
640 		pos++;
641 	}
642 	tag->cksum = sum;	/* 8 bit */
643 }
644 
645 
646 /* assumes sector number of descriptor to be saved already present */
647 void
648 udf_validate_tag_and_crc_sums(void *blob)
649 {
650 	struct desc_tag *tag  = blob;
651 	uint8_t         *btag = (uint8_t *) tag;
652 	uint16_t crc, crc_len;
653 
654 	crc_len = udf_rw16(tag->desc_crc_len);
655 
656 	/* check payload CRC if applicable */
657 	if (crc_len > 0) {
658 		crc = udf_cksum(btag + UDF_DESC_TAG_LENGTH, crc_len);
659 		tag->desc_crc = udf_rw16(crc);
660 	}
661 
662 	/* calculate TAG header checksum */
663 	udf_validate_tag_sum(blob);
664 }
665 
666 /* --------------------------------------------------------------------- */
667 
668 /*
669  * XXX note the different semantics from udfclient: for FIDs it still rounds
670  * up to sectors. Use udf_fidsize() for a correct length.
671  */
672 
673 int
674 udf_tagsize(union dscrptr *dscr, uint32_t lb_size)
675 {
676 	uint32_t size, tag_id, num_lb, elmsz;
677 
678 	tag_id = udf_rw16(dscr->tag.id);
679 
680 	switch (tag_id) {
681 	case TAGID_LOGVOL :
682 		size  = sizeof(struct logvol_desc) - 1;
683 		size += udf_rw32(dscr->lvd.mt_l);
684 		break;
685 	case TAGID_UNALLOC_SPACE :
686 		elmsz = sizeof(struct extent_ad);
687 		size  = sizeof(struct unalloc_sp_desc) - elmsz;
688 		size += udf_rw32(dscr->usd.alloc_desc_num) * elmsz;
689 		break;
690 	case TAGID_FID :
691 		size = UDF_FID_SIZE + dscr->fid.l_fi + udf_rw16(dscr->fid.l_iu);
692 		size = (size + 3) & ~3;
693 		break;
694 	case TAGID_LOGVOL_INTEGRITY :
695 		size  = sizeof(struct logvol_int_desc) - sizeof(uint32_t);
696 		size += udf_rw32(dscr->lvid.l_iu);
697 		size += (2 * udf_rw32(dscr->lvid.num_part) * sizeof(uint32_t));
698 		break;
699 	case TAGID_SPACE_BITMAP :
700 		size  = sizeof(struct space_bitmap_desc) - 1;
701 		size += udf_rw32(dscr->sbd.num_bytes);
702 		break;
703 	case TAGID_SPARING_TABLE :
704 		elmsz = sizeof(struct spare_map_entry);
705 		size  = sizeof(struct udf_sparing_table) - elmsz;
706 		size += udf_rw16(dscr->spt.rt_l) * elmsz;
707 		break;
708 	case TAGID_FENTRY :
709 		size  = sizeof(struct file_entry);
710 		size += udf_rw32(dscr->fe.l_ea) + udf_rw32(dscr->fe.l_ad)-1;
711 		break;
712 	case TAGID_EXTFENTRY :
713 		size  = sizeof(struct extfile_entry);
714 		size += udf_rw32(dscr->efe.l_ea) + udf_rw32(dscr->efe.l_ad)-1;
715 		break;
716 	case TAGID_FSD :
717 		size  = sizeof(struct fileset_desc);
718 		break;
719 	default :
720 		size = sizeof(union dscrptr);
721 		break;
722 	}
723 
724 	if ((size == 0) || (lb_size == 0)) return 0;
725 
726 	/* round up in sectors */
727 	num_lb = (size + lb_size -1) / lb_size;
728 	return num_lb * lb_size;
729 }
730 
731 
732 int
733 udf_fidsize(struct fileid_desc *fid)
734 {
735 	uint32_t size;
736 
737 	if (udf_rw16(fid->tag.id) != TAGID_FID)
738 		panic("got udf_fidsize on non FID\n");
739 
740 	size = UDF_FID_SIZE + fid->l_fi + udf_rw16(fid->l_iu);
741 	size = (size + 3) & ~3;
742 
743 	return size;
744 }
745 
746 /* --------------------------------------------------------------------- */
747 
748 void
749 udf_lock_node(struct udf_node *udf_node, int flag, char const *fname, const int lineno)
750 {
751 	int ret;
752 
753 	mutex_enter(&udf_node->node_mutex);
754 	/* wait until free */
755 	while (udf_node->i_flags & IN_LOCKED) {
756 		ret = cv_timedwait(&udf_node->node_lock, &udf_node->node_mutex, hz/8);
757 		/* TODO check if we should return error; abort */
758 		if (ret == EWOULDBLOCK) {
759 			DPRINTF(LOCKING, ( "udf_lock_node: udf_node %p would block "
760 				"wanted at %s:%d, previously locked at %s:%d\n",
761 				udf_node, fname, lineno,
762 				udf_node->lock_fname, udf_node->lock_lineno));
763 		}
764 	}
765 	/* grab */
766 	udf_node->i_flags |= IN_LOCKED | flag;
767 	/* debug */
768 	udf_node->lock_fname  = fname;
769 	udf_node->lock_lineno = lineno;
770 
771 	mutex_exit(&udf_node->node_mutex);
772 }
773 
774 
775 void
776 udf_unlock_node(struct udf_node *udf_node, int flag)
777 {
778 	mutex_enter(&udf_node->node_mutex);
779 	udf_node->i_flags &= ~(IN_LOCKED | flag);
780 	cv_broadcast(&udf_node->node_lock);
781 	mutex_exit(&udf_node->node_mutex);
782 }
783 
784 
785 /* --------------------------------------------------------------------- */
786 
787 static int
788 udf_read_anchor(struct udf_mount *ump, uint32_t sector, struct anchor_vdp **dst)
789 {
790 	int error;
791 
792 	error = udf_read_phys_dscr(ump, sector, M_UDFVOLD,
793 			(union dscrptr **) dst);
794 	if (!error) {
795 		/* blank terminator blocks are not allowed here */
796 		if (*dst == NULL)
797 			return ENOENT;
798 		if (udf_rw16((*dst)->tag.id) != TAGID_ANCHOR) {
799 			error = ENOENT;
800 			free(*dst, M_UDFVOLD);
801 			*dst = NULL;
802 			DPRINTF(VOLUMES, ("Not an anchor\n"));
803 		}
804 	}
805 
806 	return error;
807 }
808 
809 
810 int
811 udf_read_anchors(struct udf_mount *ump)
812 {
813 	struct udf_args *args = &ump->mount_args;
814 	struct mmc_trackinfo first_track;
815 	struct mmc_trackinfo second_track;
816 	struct mmc_trackinfo last_track;
817 	struct anchor_vdp **anchorsp;
818 	uint32_t track_start;
819 	uint32_t track_end;
820 	uint32_t positions[4];
821 	int first_tracknr, last_tracknr;
822 	int error, anch, ok, first_anchor;
823 
824 	/* search the first and last track of the specified session */
825 	error = udf_search_tracks(ump, args, &first_tracknr, &last_tracknr);
826 	if (!error) {
827 		first_track.tracknr = first_tracknr;
828 		error = udf_update_trackinfo(ump, &first_track);
829 	}
830 	if (!error) {
831 		last_track.tracknr = last_tracknr;
832 		error = udf_update_trackinfo(ump, &last_track);
833 	}
834 	if ((!error) && (first_tracknr != last_tracknr)) {
835 		second_track.tracknr = first_tracknr+1;
836 		error = udf_update_trackinfo(ump, &second_track);
837 	}
838 	if (error) {
839 		printf("UDF mount: reading disc geometry failed\n");
840 		return 0;
841 	}
842 
843 	track_start = first_track.track_start;
844 
845 	/* `end' is not as straitforward as start. */
846 	track_end =   last_track.track_start
847 		    + last_track.track_size - last_track.free_blocks - 1;
848 
849 	if (ump->discinfo.mmc_cur & MMC_CAP_SEQUENTIAL) {
850 		/* end of track is not straitforward here */
851 		if (last_track.flags & MMC_TRACKINFO_LRA_VALID)
852 			track_end = last_track.last_recorded;
853 		else if (last_track.flags & MMC_TRACKINFO_NWA_VALID)
854 			track_end = last_track.next_writable
855 				    - ump->discinfo.link_block_penalty;
856 	}
857 
858 	/* its no use reading a blank track */
859 	first_anchor = 0;
860 	if (first_track.flags & MMC_TRACKINFO_BLANK)
861 		first_anchor = 1;
862 
863 	/* get our packet size */
864 	ump->packet_size = first_track.packet_size;
865 	if (first_track.flags & MMC_TRACKINFO_BLANK)
866 		ump->packet_size = second_track.packet_size;
867 
868 	if (ump->packet_size <= 1) {
869 		/* take max, but not bigger than 64 */
870 		ump->packet_size = MAXPHYS / ump->discinfo.sector_size;
871 		ump->packet_size = MIN(ump->packet_size, 64);
872 	}
873 	KASSERT(ump->packet_size >= 1);
874 
875 	/* read anchors start+256, start+512, end-256, end */
876 	positions[0] = track_start+256;
877 	positions[1] =   track_end-256;
878 	positions[2] =   track_end;
879 	positions[3] = track_start+512;	/* [UDF 2.60/6.11.2] */
880 	/* XXX shouldn't +512 be prefered above +256 for compat with Roxio CD */
881 
882 	ok = 0;
883 	anchorsp = ump->anchors;
884 	for (anch = first_anchor; anch < 4; anch++) {
885 		DPRINTF(VOLUMES, ("Read anchor %d at sector %d\n", anch,
886 		    positions[anch]));
887 		error = udf_read_anchor(ump, positions[anch], anchorsp);
888 		if (!error) {
889 			anchorsp++;
890 			ok++;
891 		}
892 	}
893 
894 	/* VATs are only recorded on sequential media, but initialise */
895 	ump->first_possible_vat_location = track_start + 2;
896 	ump->last_possible_vat_location  = track_end + last_track.packet_size;
897 
898 	return ok;
899 }
900 
901 /* --------------------------------------------------------------------- */
902 
903 /* we dont try to be smart; we just record the parts */
904 #define UDF_UPDATE_DSCR(name, dscr) \
905 	if (name) \
906 		free(name, M_UDFVOLD); \
907 	name = dscr;
908 
909 static int
910 udf_process_vds_descriptor(struct udf_mount *ump, union dscrptr *dscr)
911 {
912 	struct part_desc *part;
913 	uint16_t phys_part, raw_phys_part;
914 
915 	DPRINTF(VOLUMES, ("\tprocessing VDS descr %d\n",
916 	    udf_rw16(dscr->tag.id)));
917 	switch (udf_rw16(dscr->tag.id)) {
918 	case TAGID_PRI_VOL :		/* primary partition		*/
919 		UDF_UPDATE_DSCR(ump->primary_vol, &dscr->pvd);
920 		break;
921 	case TAGID_LOGVOL :		/* logical volume		*/
922 		UDF_UPDATE_DSCR(ump->logical_vol, &dscr->lvd);
923 		break;
924 	case TAGID_UNALLOC_SPACE :	/* unallocated space		*/
925 		UDF_UPDATE_DSCR(ump->unallocated, &dscr->usd);
926 		break;
927 	case TAGID_IMP_VOL :		/* implementation		*/
928 		/* XXX do we care about multiple impl. descr ? */
929 		UDF_UPDATE_DSCR(ump->implementation, &dscr->ivd);
930 		break;
931 	case TAGID_PARTITION :		/* physical partition		*/
932 		/* not much use if its not allocated */
933 		if ((udf_rw16(dscr->pd.flags) & UDF_PART_FLAG_ALLOCATED) == 0) {
934 			free(dscr, M_UDFVOLD);
935 			break;
936 		}
937 
938 		/*
939 		 * BUGALERT: some rogue implementations use random physical
940 		 * partion numbers to break other implementations so lookup
941 		 * the number.
942 		 */
943 		raw_phys_part = udf_rw16(dscr->pd.part_num);
944 		for (phys_part = 0; phys_part < UDF_PARTITIONS; phys_part++) {
945 			part = ump->partitions[phys_part];
946 			if (part == NULL)
947 				break;
948 			if (udf_rw16(part->part_num) == raw_phys_part)
949 				break;
950 		}
951 		if (phys_part == UDF_PARTITIONS) {
952 			free(dscr, M_UDFVOLD);
953 			return EINVAL;
954 		}
955 
956 		UDF_UPDATE_DSCR(ump->partitions[phys_part], &dscr->pd);
957 		break;
958 	case TAGID_VOL :		/* volume space extender; rare	*/
959 		DPRINTF(VOLUMES, ("VDS extender ignored\n"));
960 		free(dscr, M_UDFVOLD);
961 		break;
962 	default :
963 		DPRINTF(VOLUMES, ("Unhandled VDS type %d\n",
964 		    udf_rw16(dscr->tag.id)));
965 		free(dscr, M_UDFVOLD);
966 	}
967 
968 	return 0;
969 }
970 #undef UDF_UPDATE_DSCR
971 
972 /* --------------------------------------------------------------------- */
973 
974 static int
975 udf_read_vds_extent(struct udf_mount *ump, uint32_t loc, uint32_t len)
976 {
977 	union dscrptr *dscr;
978 	uint32_t sector_size, dscr_size;
979 	int error;
980 
981 	sector_size = ump->discinfo.sector_size;
982 
983 	/* loc is sectornr, len is in bytes */
984 	error = EIO;
985 	while (len) {
986 		error = udf_read_phys_dscr(ump, loc, M_UDFVOLD, &dscr);
987 		if (error)
988 			return error;
989 
990 		/* blank block is a terminator */
991 		if (dscr == NULL)
992 			return 0;
993 
994 		/* TERM descriptor is a terminator */
995 		if (udf_rw16(dscr->tag.id) == TAGID_TERM) {
996 			free(dscr, M_UDFVOLD);
997 			return 0;
998 		}
999 
1000 		/* process all others */
1001 		dscr_size = udf_tagsize(dscr, sector_size);
1002 		error = udf_process_vds_descriptor(ump, dscr);
1003 		if (error) {
1004 			free(dscr, M_UDFVOLD);
1005 			break;
1006 		}
1007 		assert((dscr_size % sector_size) == 0);
1008 
1009 		len -= dscr_size;
1010 		loc += dscr_size / sector_size;
1011 	}
1012 
1013 	return error;
1014 }
1015 
1016 
1017 int
1018 udf_read_vds_space(struct udf_mount *ump)
1019 {
1020 	/* struct udf_args *args = &ump->mount_args; */
1021 	struct anchor_vdp *anchor, *anchor2;
1022 	size_t size;
1023 	uint32_t main_loc, main_len;
1024 	uint32_t reserve_loc, reserve_len;
1025 	int error;
1026 
1027 	/*
1028 	 * read in VDS space provided by the anchors; if one descriptor read
1029 	 * fails, try the mirror sector.
1030 	 *
1031 	 * check if 2nd anchor is different from 1st; if so, go for 2nd. This
1032 	 * avoids the `compatibility features' of DirectCD that may confuse
1033 	 * stuff completely.
1034 	 */
1035 
1036 	anchor  = ump->anchors[0];
1037 	anchor2 = ump->anchors[1];
1038 	assert(anchor);
1039 
1040 	if (anchor2) {
1041 		size = sizeof(struct extent_ad);
1042 		if (memcmp(&anchor->main_vds_ex, &anchor2->main_vds_ex, size))
1043 			anchor = anchor2;
1044 		/* reserve is specified to be a literal copy of main */
1045 	}
1046 
1047 	main_loc    = udf_rw32(anchor->main_vds_ex.loc);
1048 	main_len    = udf_rw32(anchor->main_vds_ex.len);
1049 
1050 	reserve_loc = udf_rw32(anchor->reserve_vds_ex.loc);
1051 	reserve_len = udf_rw32(anchor->reserve_vds_ex.len);
1052 
1053 	error = udf_read_vds_extent(ump, main_loc, main_len);
1054 	if (error) {
1055 		printf("UDF mount: reading in reserve VDS extent\n");
1056 		error = udf_read_vds_extent(ump, reserve_loc, reserve_len);
1057 	}
1058 
1059 	return error;
1060 }
1061 
1062 /* --------------------------------------------------------------------- */
1063 
1064 /*
1065  * Read in the logical volume integrity sequence pointed to by our logical
1066  * volume descriptor. Its a sequence that can be extended using fields in the
1067  * integrity descriptor itself. On sequential media only one is found, on
1068  * rewritable media a sequence of descriptors can be found as a form of
1069  * history keeping and on non sequential write-once media the chain is vital
1070  * to allow more and more descriptors to be written. The last descriptor
1071  * written in an extent needs to claim space for a new extent.
1072  */
1073 
1074 static int
1075 udf_retrieve_lvint(struct udf_mount *ump)
1076 {
1077 	union dscrptr *dscr;
1078 	struct logvol_int_desc *lvint;
1079 	struct udf_lvintq *trace;
1080 	uint32_t lb_size, lbnum, len;
1081 	int dscr_type, error, trace_len;
1082 
1083 	lb_size = udf_rw32(ump->logical_vol->lb_size);
1084 	len     = udf_rw32(ump->logical_vol->integrity_seq_loc.len);
1085 	lbnum   = udf_rw32(ump->logical_vol->integrity_seq_loc.loc);
1086 
1087 	/* clean trace */
1088 	memset(ump->lvint_trace, 0,
1089 	    UDF_LVDINT_SEGMENTS * sizeof(struct udf_lvintq));
1090 
1091 	trace_len    = 0;
1092 	trace        = ump->lvint_trace;
1093 	trace->start = lbnum;
1094 	trace->end   = lbnum + len/lb_size;
1095 	trace->pos   = 0;
1096 	trace->wpos  = 0;
1097 
1098 	lvint = NULL;
1099 	dscr  = NULL;
1100 	error = 0;
1101 	while (len) {
1102 		trace->pos  = lbnum - trace->start;
1103 		trace->wpos = trace->pos + 1;
1104 
1105 		/* read in our integrity descriptor */
1106 		error = udf_read_phys_dscr(ump, lbnum, M_UDFVOLD, &dscr);
1107 		if (!error) {
1108 			if (dscr == NULL) {
1109 				trace->wpos = trace->pos;
1110 				break;		/* empty terminates */
1111 			}
1112 			dscr_type = udf_rw16(dscr->tag.id);
1113 			if (dscr_type == TAGID_TERM) {
1114 				trace->wpos = trace->pos;
1115 				break;		/* clean terminator */
1116 			}
1117 			if (dscr_type != TAGID_LOGVOL_INTEGRITY) {
1118 				/* fatal... corrupt disc */
1119 				error = ENOENT;
1120 				break;
1121 			}
1122 			if (lvint)
1123 				free(lvint, M_UDFVOLD);
1124 			lvint = &dscr->lvid;
1125 			dscr = NULL;
1126 		} /* else hope for the best... maybe the next is ok */
1127 
1128 		DPRINTFIF(VOLUMES, lvint, ("logvol integrity read, state %s\n",
1129 		    udf_rw32(lvint->integrity_type) ? "CLOSED" : "OPEN"));
1130 
1131 		/* proceed sequential */
1132 		lbnum += 1;
1133 		len    -= lb_size;
1134 
1135 		/* are we linking to a new piece? */
1136 		if (dscr && lvint->next_extent.len) {
1137 			len    = udf_rw32(lvint->next_extent.len);
1138 			lbnum = udf_rw32(lvint->next_extent.loc);
1139 
1140 			if (trace_len >= UDF_LVDINT_SEGMENTS-1) {
1141 				/* IEK! segment link full... */
1142 				DPRINTF(VOLUMES, ("lvdint segments full\n"));
1143 				error = EINVAL;
1144 			} else {
1145 				trace++;
1146 				trace_len++;
1147 
1148 				trace->start = lbnum;
1149 				trace->end   = lbnum + len/lb_size;
1150 				trace->pos   = 0;
1151 				trace->wpos  = 0;
1152 			}
1153 		}
1154 	}
1155 
1156 	/* clean up the mess, esp. when there is an error */
1157 	if (dscr)
1158 		free(dscr, M_UDFVOLD);
1159 
1160 	if (error && lvint) {
1161 		free(lvint, M_UDFVOLD);
1162 		lvint = NULL;
1163 	}
1164 
1165 	if (!lvint)
1166 		error = ENOENT;
1167 
1168 	ump->logvol_integrity = lvint;
1169 	return error;
1170 }
1171 
1172 
1173 static int
1174 udf_loose_lvint_history(struct udf_mount *ump)
1175 {
1176 	union dscrptr **bufs, *dscr, *last_dscr;
1177 	struct udf_lvintq *trace, *in_trace, *out_trace;
1178 	struct logvol_int_desc *lvint;
1179 	uint32_t in_ext, in_pos, in_len;
1180 	uint32_t out_ext, out_wpos, out_len;
1181 	uint32_t lb_size, packet_size, lb_num;
1182 	uint32_t len, start;
1183 	int ext, minext, extlen, cnt, cpy_len, dscr_type;
1184 	int losing;
1185 	int error;
1186 
1187 	DPRINTF(VOLUMES, ("need to lose some lvint history\n"));
1188 
1189 	lb_size = udf_rw32(ump->logical_vol->lb_size);
1190 	packet_size = ump->data_track.packet_size;	/* XXX data track */
1191 
1192 	/* search smallest extent */
1193 	trace = &ump->lvint_trace[0];
1194 	minext = trace->end - trace->start;
1195 	for (ext = 1; ext < UDF_LVDINT_SEGMENTS; ext++) {
1196 		trace = &ump->lvint_trace[ext];
1197 		extlen = trace->end - trace->start;
1198 		if (extlen == 0)
1199 			break;
1200 		minext = MIN(minext, extlen);
1201 	}
1202 	losing = MIN(minext, UDF_LVINT_LOSSAGE);
1203 	/* no sense wiping all */
1204 	if (losing == minext)
1205 		losing--;
1206 
1207 	DPRINTF(VOLUMES, ("\tlosing %d entries\n", losing));
1208 
1209 	/* get buffer for pieces */
1210 	bufs = malloc(UDF_LVDINT_SEGMENTS * sizeof(void *), M_TEMP, M_WAITOK);
1211 
1212 	in_ext    = 0;
1213 	in_pos    = losing;
1214 	in_trace  = &ump->lvint_trace[in_ext];
1215 	in_len    = in_trace->end - in_trace->start;
1216 	out_ext   = 0;
1217 	out_wpos  = 0;
1218 	out_trace = &ump->lvint_trace[out_ext];
1219 	out_len   = out_trace->end - out_trace->start;
1220 
1221 	last_dscr = NULL;
1222 	for(;;) {
1223 		out_trace->pos  = out_wpos;
1224 		out_trace->wpos = out_trace->pos;
1225 		if (in_pos >= in_len) {
1226 			in_ext++;
1227 			in_pos = 0;
1228 			in_trace = &ump->lvint_trace[in_ext];
1229 			in_len   = in_trace->end - in_trace->start;
1230 		}
1231 		if (out_wpos >= out_len) {
1232 			out_ext++;
1233 			out_wpos = 0;
1234 			out_trace = &ump->lvint_trace[out_ext];
1235 			out_len   = out_trace->end - out_trace->start;
1236 		}
1237 		/* copy overlap contents */
1238 		cpy_len = MIN(in_len - in_pos, out_len - out_wpos);
1239 		cpy_len = MIN(cpy_len, in_len - in_trace->pos);
1240 		if (cpy_len == 0)
1241 			break;
1242 
1243 		/* copy */
1244 		DPRINTF(VOLUMES, ("\treading %d lvid descriptors\n", cpy_len));
1245 		for (cnt = 0; cnt < cpy_len; cnt++) {
1246 			/* read in our integrity descriptor */
1247 			lb_num = in_trace->start + in_pos + cnt;
1248 			error = udf_read_phys_dscr(ump, lb_num, M_UDFVOLD,
1249 				&dscr);
1250 			if (error) {
1251 				/* copy last one */
1252 				dscr = last_dscr;
1253 			}
1254 			bufs[cnt] = dscr;
1255 			if (!error) {
1256 				if (dscr == NULL) {
1257 					out_trace->pos  = out_wpos + cnt;
1258 					out_trace->wpos = out_trace->pos;
1259 					break;		/* empty terminates */
1260 				}
1261 				dscr_type = udf_rw16(dscr->tag.id);
1262 				if (dscr_type == TAGID_TERM) {
1263 					out_trace->pos  = out_wpos + cnt;
1264 					out_trace->wpos = out_trace->pos;
1265 					break;		/* clean terminator */
1266 				}
1267 				if (dscr_type != TAGID_LOGVOL_INTEGRITY) {
1268 					panic(  "UDF integrity sequence "
1269 						"corrupted while mounted!\n");
1270 				}
1271 				last_dscr = dscr;
1272 			}
1273 		}
1274 
1275 		/* patch up if first entry was on error */
1276 		if (bufs[0] == NULL) {
1277 			for (cnt = 0; cnt < cpy_len; cnt++)
1278 				if (bufs[cnt] != NULL)
1279 					break;
1280 			last_dscr = bufs[cnt];
1281 			for (; cnt > 0; cnt--) {
1282 				bufs[cnt] = last_dscr;
1283 			}
1284 		}
1285 
1286 		/* glue + write out */
1287 		DPRINTF(VOLUMES, ("\twriting %d lvid descriptors\n", cpy_len));
1288 		for (cnt = 0; cnt < cpy_len; cnt++) {
1289 			lb_num = out_trace->start + out_wpos + cnt;
1290 			lvint  = &bufs[cnt]->lvid;
1291 
1292 			/* set continuation */
1293 			len = 0;
1294 			start = 0;
1295 			if (out_wpos + cnt == out_len) {
1296 				/* get continuation */
1297 				trace = &ump->lvint_trace[out_ext+1];
1298 				len   = trace->end - trace->start;
1299 				start = trace->start;
1300 			}
1301 			lvint->next_extent.len = udf_rw32(len);
1302 			lvint->next_extent.loc = udf_rw32(start);
1303 
1304 			lb_num = trace->start + trace->wpos;
1305 			error = udf_write_phys_dscr_sync(ump, NULL, UDF_C_DSCR,
1306 				bufs[cnt], lb_num, lb_num);
1307 			DPRINTFIF(VOLUMES, error,
1308 				("error writing lvint lb_num\n"));
1309 		}
1310 
1311 		/* free non repeating descriptors */
1312 		last_dscr = NULL;
1313 		for (cnt = 0; cnt < cpy_len; cnt++) {
1314 			if (bufs[cnt] != last_dscr)
1315 				free(bufs[cnt], M_UDFVOLD);
1316 			last_dscr = bufs[cnt];
1317 		}
1318 
1319 		/* advance */
1320 		in_pos   += cpy_len;
1321 		out_wpos += cpy_len;
1322 	}
1323 
1324 	free(bufs, M_TEMP);
1325 
1326 	return 0;
1327 }
1328 
1329 
1330 static int
1331 udf_writeout_lvint(struct udf_mount *ump, int lvflag)
1332 {
1333 	struct udf_lvintq *trace;
1334 	struct timeval  now_v;
1335 	struct timespec now_s;
1336 	uint32_t sector;
1337 	int logvol_integrity;
1338 	int space, error;
1339 
1340 	DPRINTF(VOLUMES, ("writing out logvol integrity descriptor\n"));
1341 
1342 again:
1343 	/* get free space in last chunk */
1344 	trace = ump->lvint_trace;
1345 	while (trace->wpos > (trace->end - trace->start)) {
1346 		DPRINTF(VOLUMES, ("skip : start = %d, end = %d, pos = %d, "
1347 				  "wpos = %d\n", trace->start, trace->end,
1348 				  trace->pos, trace->wpos));
1349 		trace++;
1350 	}
1351 
1352 	/* check if there is space to append */
1353 	space = (trace->end - trace->start) - trace->wpos;
1354 	DPRINTF(VOLUMES, ("write start = %d, end = %d, pos = %d, wpos = %d, "
1355 			  "space = %d\n", trace->start, trace->end, trace->pos,
1356 			  trace->wpos, space));
1357 
1358 	/* get state */
1359 	logvol_integrity = udf_rw32(ump->logvol_integrity->integrity_type);
1360 	if (logvol_integrity == UDF_INTEGRITY_CLOSED) {
1361 		if ((space < 3) && (lvflag & UDF_APPENDONLY_LVINT)) {
1362 			/* don't allow this logvol to be opened */
1363 			/* TODO extent LVINT space if possible */
1364 			return EROFS;
1365 		}
1366 	}
1367 
1368 	if (space < 1) {
1369 		if (lvflag & UDF_APPENDONLY_LVINT)
1370 			return EROFS;
1371 		/* loose history by re-writing extents */
1372 		error = udf_loose_lvint_history(ump);
1373 		if (error)
1374 			return error;
1375 		goto again;
1376 	}
1377 
1378 	/* update our integrity descriptor to identify us and timestamp it */
1379 	DPRINTF(VOLUMES, ("updating integrity descriptor\n"));
1380 	microtime(&now_v);
1381 	TIMEVAL_TO_TIMESPEC(&now_v, &now_s);
1382 	udf_timespec_to_timestamp(&now_s, &ump->logvol_integrity->time);
1383 	udf_set_regid(&ump->logvol_info->impl_id, IMPL_NAME);
1384 	udf_add_impl_regid(ump, &ump->logvol_info->impl_id);
1385 
1386 	/* writeout integrity descriptor */
1387 	sector = trace->start + trace->wpos;
1388 	error = udf_write_phys_dscr_sync(ump, NULL, UDF_C_DSCR,
1389 			(union dscrptr *) ump->logvol_integrity,
1390 			sector, sector);
1391 	DPRINTF(VOLUMES, ("writeout lvint : error = %d\n", error));
1392 	if (error)
1393 		return error;
1394 
1395 	/* advance write position */
1396 	trace->wpos++; space--;
1397 	if (space >= 1) {
1398 		/* append terminator */
1399 		sector = trace->start + trace->wpos;
1400 		error = udf_write_terminator(ump, sector);
1401 
1402 		DPRINTF(VOLUMES, ("write terminator : error = %d\n", error));
1403 	}
1404 
1405 	space = (trace->end - trace->start) - trace->wpos;
1406 	DPRINTF(VOLUMES, ("write start = %d, end = %d, pos = %d, wpos = %d, "
1407 		"space = %d\n", trace->start, trace->end, trace->pos,
1408 		trace->wpos, space));
1409 	DPRINTF(VOLUMES, ("finished writing out logvol integrity descriptor "
1410 		"successfull\n"));
1411 
1412 	return error;
1413 }
1414 
1415 /* --------------------------------------------------------------------- */
1416 
1417 static int
1418 udf_read_partition_spacetables(struct udf_mount *ump)
1419 {
1420 	union dscrptr        *dscr;
1421 	/* struct udf_args *args = &ump->mount_args; */
1422 	struct part_desc     *partd;
1423 	struct part_hdr_desc *parthdr;
1424 	struct udf_bitmap    *bitmap;
1425 	uint32_t phys_part;
1426 	uint32_t lb_num, len;
1427 	int error, dscr_type;
1428 
1429 	/* unallocated space map */
1430 	for (phys_part = 0; phys_part < UDF_PARTITIONS; phys_part++) {
1431 		partd = ump->partitions[phys_part];
1432 		if (partd == NULL)
1433 			continue;
1434 		parthdr = &partd->_impl_use.part_hdr;
1435 
1436 		lb_num  = udf_rw32(partd->start_loc);
1437 		lb_num += udf_rw32(parthdr->unalloc_space_bitmap.lb_num);
1438 		len     = udf_rw32(parthdr->unalloc_space_bitmap.len);
1439 		if (len == 0)
1440 			continue;
1441 
1442 		DPRINTF(VOLUMES, ("Read unalloc. space bitmap %d\n", lb_num));
1443 		error = udf_read_phys_dscr(ump, lb_num, M_UDFVOLD, &dscr);
1444 		if (!error && dscr) {
1445 			/* analyse */
1446 			dscr_type = udf_rw16(dscr->tag.id);
1447 			if (dscr_type == TAGID_SPACE_BITMAP) {
1448 				DPRINTF(VOLUMES, ("Accepting space bitmap\n"));
1449 				ump->part_unalloc_dscr[phys_part] = &dscr->sbd;
1450 
1451 				/* fill in ump->part_unalloc_bits */
1452 				bitmap = &ump->part_unalloc_bits[phys_part];
1453 				bitmap->blob  = (uint8_t *) dscr;
1454 				bitmap->bits  = dscr->sbd.data;
1455 				bitmap->max_offset = udf_rw32(dscr->sbd.num_bits);
1456 				bitmap->pages = NULL;	/* TODO */
1457 				bitmap->data_pos     = 0;
1458 				bitmap->metadata_pos = 0;
1459 			} else {
1460 				free(dscr, M_UDFVOLD);
1461 
1462 				printf( "UDF mount: error reading unallocated "
1463 					"space bitmap\n");
1464 				return EROFS;
1465 			}
1466 		} else {
1467 			/* blank not allowed */
1468 			printf("UDF mount: blank unallocated space bitmap\n");
1469 			return EROFS;
1470 		}
1471 	}
1472 
1473 	/* unallocated space table (not supported) */
1474 	for (phys_part = 0; phys_part < UDF_PARTITIONS; phys_part++) {
1475 		partd = ump->partitions[phys_part];
1476 		if (partd == NULL)
1477 			continue;
1478 		parthdr = &partd->_impl_use.part_hdr;
1479 
1480 		len     = udf_rw32(parthdr->unalloc_space_table.len);
1481 		if (len) {
1482 			printf("UDF mount: space tables not supported\n");
1483 			return EROFS;
1484 		}
1485 	}
1486 
1487 	/* freed space map */
1488 	for (phys_part = 0; phys_part < UDF_PARTITIONS; phys_part++) {
1489 		partd = ump->partitions[phys_part];
1490 		if (partd == NULL)
1491 			continue;
1492 		parthdr = &partd->_impl_use.part_hdr;
1493 
1494 		/* freed space map */
1495 		lb_num  = udf_rw32(partd->start_loc);
1496 		lb_num += udf_rw32(parthdr->freed_space_bitmap.lb_num);
1497 		len     = udf_rw32(parthdr->freed_space_bitmap.len);
1498 		if (len == 0)
1499 			continue;
1500 
1501 		DPRINTF(VOLUMES, ("Read unalloc. space bitmap %d\n", lb_num));
1502 		error = udf_read_phys_dscr(ump, lb_num, M_UDFVOLD, &dscr);
1503 		if (!error && dscr) {
1504 			/* analyse */
1505 			dscr_type = udf_rw16(dscr->tag.id);
1506 			if (dscr_type == TAGID_SPACE_BITMAP) {
1507 				DPRINTF(VOLUMES, ("Accepting space bitmap\n"));
1508 				ump->part_freed_dscr[phys_part] = &dscr->sbd;
1509 
1510 				/* fill in ump->part_freed_bits */
1511 				bitmap = &ump->part_unalloc_bits[phys_part];
1512 				bitmap->blob  = (uint8_t *) dscr;
1513 				bitmap->bits  = dscr->sbd.data;
1514 				bitmap->max_offset = udf_rw32(dscr->sbd.num_bits);
1515 				bitmap->pages = NULL;	/* TODO */
1516 				bitmap->data_pos     = 0;
1517 				bitmap->metadata_pos = 0;
1518 			} else {
1519 				free(dscr, M_UDFVOLD);
1520 
1521 				printf( "UDF mount: error reading freed  "
1522 					"space bitmap\n");
1523 				return EROFS;
1524 			}
1525 		} else {
1526 			/* blank not allowed */
1527 			printf("UDF mount: blank freed space bitmap\n");
1528 			return EROFS;
1529 		}
1530 	}
1531 
1532 	/* freed space table (not supported) */
1533 	for (phys_part = 0; phys_part < UDF_PARTITIONS; phys_part++) {
1534 		partd = ump->partitions[phys_part];
1535 		if (partd == NULL)
1536 			continue;
1537 		parthdr = &partd->_impl_use.part_hdr;
1538 
1539 		len     = udf_rw32(parthdr->freed_space_table.len);
1540 		if (len) {
1541 			printf("UDF mount: space tables not supported\n");
1542 			return EROFS;
1543 		}
1544 	}
1545 
1546 	return 0;
1547 }
1548 
1549 
1550 /* TODO implement async writeout */
1551 int
1552 udf_write_partition_spacetables(struct udf_mount *ump, int waitfor)
1553 {
1554 	union dscrptr        *dscr;
1555 	/* struct udf_args *args = &ump->mount_args; */
1556 	struct part_desc     *partd;
1557 	struct part_hdr_desc *parthdr;
1558 	uint32_t phys_part;
1559 	uint32_t lb_num, len, ptov;
1560 	int error_all, error;
1561 
1562 	error_all = 0;
1563 	/* unallocated space map */
1564 	for (phys_part = 0; phys_part < UDF_PARTITIONS; phys_part++) {
1565 		partd = ump->partitions[phys_part];
1566 		if (partd == NULL)
1567 			continue;
1568 		parthdr = &partd->_impl_use.part_hdr;
1569 
1570 		ptov   = udf_rw32(partd->start_loc);
1571 		lb_num = udf_rw32(parthdr->unalloc_space_bitmap.lb_num);
1572 		len    = udf_rw32(parthdr->unalloc_space_bitmap.len);
1573 		if (len == 0)
1574 			continue;
1575 
1576 		DPRINTF(VOLUMES, ("Write unalloc. space bitmap %d\n",
1577 			lb_num + ptov));
1578 		dscr = (union dscrptr *) ump->part_unalloc_dscr[phys_part];
1579 		error = udf_write_phys_dscr_sync(ump, NULL, UDF_C_DSCR,
1580 				(union dscrptr *) dscr,
1581 				ptov + lb_num, lb_num);
1582 		if (error) {
1583 			DPRINTF(VOLUMES, ("\tfailed!! (error %d)\n", error));
1584 			error_all = error;
1585 		}
1586 	}
1587 
1588 	/* freed space map */
1589 	for (phys_part = 0; phys_part < UDF_PARTITIONS; phys_part++) {
1590 		partd = ump->partitions[phys_part];
1591 		if (partd == NULL)
1592 			continue;
1593 		parthdr = &partd->_impl_use.part_hdr;
1594 
1595 		/* freed space map */
1596 		ptov   = udf_rw32(partd->start_loc);
1597 		lb_num = udf_rw32(parthdr->freed_space_bitmap.lb_num);
1598 		len    = udf_rw32(parthdr->freed_space_bitmap.len);
1599 		if (len == 0)
1600 			continue;
1601 
1602 		DPRINTF(VOLUMES, ("Write freed space bitmap %d\n",
1603 			lb_num + ptov));
1604 		dscr = (union dscrptr *) ump->part_freed_dscr[phys_part];
1605 		error = udf_write_phys_dscr_sync(ump, NULL, UDF_C_DSCR,
1606 				(union dscrptr *) dscr,
1607 				ptov + lb_num, lb_num);
1608 		if (error) {
1609 			DPRINTF(VOLUMES, ("\tfailed!! (error %d)\n", error));
1610 			error_all = error;
1611 		}
1612 	}
1613 
1614 	return error_all;
1615 }
1616 
1617 /*
1618  * Checks if ump's vds information is correct and complete
1619  */
1620 
1621 int
1622 udf_process_vds(struct udf_mount *ump) {
1623 	union udf_pmap *mapping;
1624 	/* struct udf_args *args = &ump->mount_args; */
1625 	struct logvol_int_desc *lvint;
1626 	struct udf_logvol_info *lvinfo;
1627 	struct part_desc *part;
1628 	uint32_t n_pm, mt_l;
1629 	uint8_t *pmap_pos;
1630 	char *domain_name, *map_name;
1631 	const char *check_name;
1632 	char bits[128];
1633 	int pmap_stype, pmap_size;
1634 	int pmap_type, log_part, phys_part, raw_phys_part;
1635 	int n_phys, n_virt, n_spar, n_meta;
1636 	int len, error;
1637 
1638 	if (ump == NULL)
1639 		return ENOENT;
1640 
1641 	/* we need at least an anchor (trivial, but for safety) */
1642 	if (ump->anchors[0] == NULL)
1643 		return EINVAL;
1644 
1645 	/* we need at least one primary and one logical volume descriptor */
1646 	if ((ump->primary_vol == NULL) || (ump->logical_vol) == NULL)
1647 		return EINVAL;
1648 
1649 	/* we need at least one partition descriptor */
1650 	if (ump->partitions[0] == NULL)
1651 		return EINVAL;
1652 
1653 	/* check logical volume sector size verses device sector size */
1654 	if (udf_rw32(ump->logical_vol->lb_size) != ump->discinfo.sector_size) {
1655 		printf("UDF mount: format violation, lb_size != sector size\n");
1656 		return EINVAL;
1657 	}
1658 
1659 	/* check domain name */
1660 	domain_name = ump->logical_vol->domain_id.id;
1661 	if (strncmp(domain_name, "*OSTA UDF Compliant", 20)) {
1662 		printf("mount_udf: disc not OSTA UDF Compliant, aborting\n");
1663 		return EINVAL;
1664 	}
1665 
1666 	/* retrieve logical volume integrity sequence */
1667 	error = udf_retrieve_lvint(ump);
1668 
1669 	/*
1670 	 * We need at least one logvol integrity descriptor recorded.  Note
1671 	 * that its OK to have an open logical volume integrity here. The VAT
1672 	 * will close/update the integrity.
1673 	 */
1674 	if (ump->logvol_integrity == NULL)
1675 		return EINVAL;
1676 
1677 	/* read in and check unallocated and free space info if writing */
1678 	if ((ump->vfs_mountp->mnt_flag & MNT_RDONLY) == 0) {
1679 		error = udf_read_partition_spacetables(ump);
1680 		if (error)
1681 			return error;
1682 	}
1683 
1684 	/* process derived structures */
1685 	n_pm   = udf_rw32(ump->logical_vol->n_pm);   /* num partmaps         */
1686 	lvint  = ump->logvol_integrity;
1687 	lvinfo = (struct udf_logvol_info *) (&lvint->tables[2 * n_pm]);
1688 	ump->logvol_info = lvinfo;
1689 
1690 	/* TODO check udf versions? */
1691 
1692 	/*
1693 	 * check logvol mappings: effective virt->log partmap translation
1694 	 * check and recording of the mapping results. Saves expensive
1695 	 * strncmp() in tight places.
1696 	 */
1697 	DPRINTF(VOLUMES, ("checking logvol mappings\n"));
1698 	n_pm = udf_rw32(ump->logical_vol->n_pm);   /* num partmaps         */
1699 	mt_l = udf_rw32(ump->logical_vol->mt_l);   /* partmaps data length */
1700 	pmap_pos =  ump->logical_vol->maps;
1701 
1702 	if (n_pm > UDF_PMAPS) {
1703 		printf("UDF mount: too many mappings\n");
1704 		return EINVAL;
1705 	}
1706 
1707 	ump->data_part = ump->metadata_part = 0;
1708 	n_phys = n_virt = n_spar = n_meta = 0;
1709 	for (log_part = 0; log_part < n_pm; log_part++) {
1710 		mapping = (union udf_pmap *) pmap_pos;
1711 		pmap_stype = pmap_pos[0];
1712 		pmap_size  = pmap_pos[1];
1713 		switch (pmap_stype) {
1714 		case 1:	/* physical mapping */
1715 			/* volseq    = udf_rw16(mapping->pm1.vol_seq_num); */
1716 			raw_phys_part = udf_rw16(mapping->pm1.part_num);
1717 			pmap_type = UDF_VTOP_TYPE_PHYS;
1718 			n_phys++;
1719 			ump->data_part     = log_part;
1720 			ump->metadata_part = log_part;
1721 			break;
1722 		case 2: /* virtual/sparable/meta mapping */
1723 			map_name  = mapping->pm2.part_id.id;
1724 			/* volseq  = udf_rw16(mapping->pm2.vol_seq_num); */
1725 			raw_phys_part = udf_rw16(mapping->pm2.part_num);
1726 			pmap_type = UDF_VTOP_TYPE_UNKNOWN;
1727 			len = UDF_REGID_ID_SIZE;
1728 
1729 			check_name = "*UDF Virtual Partition";
1730 			if (strncmp(map_name, check_name, len) == 0) {
1731 				pmap_type = UDF_VTOP_TYPE_VIRT;
1732 				n_virt++;
1733 				ump->metadata_part = log_part;
1734 				break;
1735 			}
1736 			check_name = "*UDF Sparable Partition";
1737 			if (strncmp(map_name, check_name, len) == 0) {
1738 				pmap_type = UDF_VTOP_TYPE_SPARABLE;
1739 				n_spar++;
1740 				ump->data_part     = log_part;
1741 				ump->metadata_part = log_part;
1742 				break;
1743 			}
1744 			check_name = "*UDF Metadata Partition";
1745 			if (strncmp(map_name, check_name, len) == 0) {
1746 				pmap_type = UDF_VTOP_TYPE_META;
1747 				n_meta++;
1748 				ump->metadata_part = log_part;
1749 				break;
1750 			}
1751 			break;
1752 		default:
1753 			return EINVAL;
1754 		}
1755 
1756 		/*
1757 		 * BUGALERT: some rogue implementations use random physical
1758 		 * partion numbers to break other implementations so lookup
1759 		 * the number.
1760 		 */
1761 		for (phys_part = 0; phys_part < UDF_PARTITIONS; phys_part++) {
1762 			part = ump->partitions[phys_part];
1763 			if (part == NULL)
1764 				continue;
1765 			if (udf_rw16(part->part_num) == raw_phys_part)
1766 				break;
1767 		}
1768 
1769 		DPRINTF(VOLUMES, ("\t%d -> %d(%d) type %d\n", log_part,
1770 		    raw_phys_part, phys_part, pmap_type));
1771 
1772 		if (phys_part == UDF_PARTITIONS)
1773 			return EINVAL;
1774 		if (pmap_type == UDF_VTOP_TYPE_UNKNOWN)
1775 			return EINVAL;
1776 
1777 		ump->vtop   [log_part] = phys_part;
1778 		ump->vtop_tp[log_part] = pmap_type;
1779 
1780 		pmap_pos += pmap_size;
1781 	}
1782 	/* not winning the beauty contest */
1783 	ump->vtop_tp[UDF_VTOP_RAWPART] = UDF_VTOP_TYPE_RAW;
1784 
1785 	/* test some basic UDF assertions/requirements */
1786 	if ((n_virt > 1) || (n_spar > 1) || (n_meta > 1))
1787 		return EINVAL;
1788 
1789 	if (n_virt) {
1790 		if ((n_phys == 0) || n_spar || n_meta)
1791 			return EINVAL;
1792 	}
1793 	if (n_spar + n_phys == 0)
1794 		return EINVAL;
1795 
1796 	/* determine allocation scheme's based on disc format */
1797 	/* VAT's can only be on a sequential media */
1798 	ump->data_alloc = UDF_ALLOC_SPACEMAP;
1799 	if (n_virt)
1800 		ump->data_alloc = UDF_ALLOC_SEQUENTIAL;
1801 
1802 	ump->meta_alloc = UDF_ALLOC_SPACEMAP;
1803 	if (n_virt)
1804 		ump->meta_alloc = UDF_ALLOC_VAT;
1805 	if (n_meta)
1806 		ump->meta_alloc = UDF_ALLOC_METABITMAP;
1807 
1808 	/* special cases for pseudo-overwrite */
1809 	if (ump->discinfo.mmc_cur & MMC_CAP_PSEUDOOVERWRITE) {
1810 		ump->data_alloc = UDF_ALLOC_SEQUENTIAL;
1811 		if (n_meta) {
1812 			ump->meta_alloc = UDF_ALLOC_METASEQUENTIAL;
1813 		} else {
1814 			ump->meta_alloc = UDF_ALLOC_RELAXEDSEQUENTIAL;
1815 		}
1816 	}
1817 
1818 	/* determine logical volume open/closure actions */
1819 	if (n_virt) {
1820 		ump->lvopen  = 0;
1821 		if (ump->discinfo.last_session_state == MMC_STATE_CLOSED)
1822 			ump->lvopen |= UDF_OPEN_SESSION ;
1823 		ump->lvclose = UDF_WRITE_VAT;
1824 		if (ump->mount_args.udfmflags & UDFMNT_CLOSESESSION)
1825 			ump->lvclose |= UDF_CLOSE_SESSION;
1826 	} else {
1827 		/* `normal' rewritable or non sequential media */
1828 		ump->lvopen  = UDF_WRITE_LVINT;
1829 		ump->lvclose = UDF_WRITE_LVINT;
1830 		if ((ump->discinfo.mmc_cur & MMC_CAP_REWRITABLE) == 0)
1831 			ump->lvopen  |= UDF_APPENDONLY_LVINT;
1832 	}
1833 
1834 	/*
1835 	 * Determine sheduler error behaviour. For virtual partions, update
1836 	 * the trackinfo; for sparable partitions replace a whole block on the
1837 	 * sparable table. Allways requeue.
1838 	 */
1839 	ump->lvreadwrite = 0;
1840 	if (n_virt)
1841 		ump->lvreadwrite = UDF_UPDATE_TRACKINFO;
1842 	if (n_spar)
1843 		ump->lvreadwrite = UDF_REMAP_BLOCK;
1844 
1845 	/*
1846 	 * Select our sheduler
1847 	 */
1848 	ump->strategy = &udf_strat_rmw;
1849 	if (n_virt || (ump->discinfo.mmc_cur & MMC_CAP_PSEUDOOVERWRITE))
1850 		ump->strategy = &udf_strat_sequential;
1851 	if ((ump->discinfo.mmc_class == MMC_CLASS_DISC) ||
1852 		(ump->discinfo.mmc_class == MMC_CLASS_UNKN))
1853 			ump->strategy = &udf_strat_direct;
1854 	if (n_spar)
1855 		ump->strategy = &udf_strat_rmw;
1856 
1857 	/* print results */
1858 	DPRINTF(VOLUMES, ("\tdata alloc scheme %d, meta alloc scheme %d\n",
1859 	    ump->data_alloc, ump->meta_alloc));
1860 	DPRINTF(VOLUMES, ("\tdata partition %d, metadata partition %d\n",
1861 	    ump->data_part, ump->metadata_part));
1862 
1863 	bitmask_snprintf(ump->lvopen,  UDFLOGVOL_BITS, bits, sizeof(bits));
1864 	DPRINTF(VOLUMES, ("\tactions on logvol open  %s\n", bits));
1865 	bitmask_snprintf(ump->lvclose, UDFLOGVOL_BITS, bits, sizeof(bits));
1866 	DPRINTF(VOLUMES, ("\tactions on logvol close %s\n", bits));
1867 	bitmask_snprintf(ump->lvreadwrite, UDFONERROR_BITS, bits, sizeof(bits));
1868 	DPRINTF(VOLUMES, ("\tactions on logvol errors %s\n", bits));
1869 
1870 	DPRINTF(VOLUMES, ("\tselected sheduler `%s`\n",
1871 		(ump->strategy == &udf_strat_direct) ? "Direct" :
1872 		(ump->strategy == &udf_strat_sequential) ? "Sequential" :
1873 		(ump->strategy == &udf_strat_rmw) ? "RMW" : "UNKNOWN!"));
1874 
1875 	/* signal its OK for now */
1876 	return 0;
1877 }
1878 
1879 /* --------------------------------------------------------------------- */
1880 
1881 /*
1882  * Update logical volume name in all structures that keep a record of it. We
1883  * use memmove since each of them might be specified as a source.
1884  *
1885  * Note that it doesn't update the VAT structure!
1886  */
1887 
1888 static void
1889 udf_update_logvolname(struct udf_mount *ump, char *logvol_id)
1890 {
1891 	struct logvol_desc     *lvd = NULL;
1892 	struct fileset_desc    *fsd = NULL;
1893 	struct udf_lv_info     *lvi = NULL;
1894 
1895 	DPRINTF(VOLUMES, ("Updating logical volume name\n"));
1896 	lvd = ump->logical_vol;
1897 	fsd = ump->fileset_desc;
1898 	if (ump->implementation)
1899 		lvi = &ump->implementation->_impl_use.lv_info;
1900 
1901 	/* logvol's id might be specified as origional so use memmove here */
1902 	memmove(lvd->logvol_id, logvol_id, 128);
1903 	if (fsd)
1904 		memmove(fsd->logvol_id, logvol_id, 128);
1905 	if (lvi)
1906 		memmove(lvi->logvol_id, logvol_id, 128);
1907 }
1908 
1909 /* --------------------------------------------------------------------- */
1910 
1911 /*
1912  * Etended attribute support. UDF knows of 3 places for extended attributes:
1913  *
1914  * (a) inside the file's (e)fe in the length of the extended attriubute area
1915  * before the allocation desctriptors/filedata
1916  *
1917  * (b) in a file referenced by (e)fe->ext_attr_icb and
1918  *
1919  * (c) in the e(fe)'s associated stream directory that can hold various
1920 	struct part_desc *part;
1921  * sub-files. In the stream directory a few fixed named subfiles are reserved
1922  * for NT/Unix ACL's and OS/2 attributes.
1923  *
1924  * NOTE: Extended attributes are read randomly but allways written
1925  * *atomicaly*. For ACL's this interface is propably different but not known
1926  * to me yet.
1927  */
1928 
1929 static int
1930 udf_impl_extattr_check(struct impl_extattr_entry *implext)
1931 {
1932 	uint16_t   *spos;
1933 
1934 	if (strncmp(implext->imp_id.id, "*UDF", 4) == 0) {
1935 		/* checksum valid? */
1936 		DPRINTF(EXTATTR, ("checking UDF impl. attr checksum\n"));
1937 		spos = (uint16_t *) implext->data;
1938 		if (udf_rw16(*spos) != udf_ea_cksum((uint8_t *) implext))
1939 			return EINVAL;
1940 	}
1941 	return 0;
1942 }
1943 
1944 static void
1945 udf_calc_impl_extattr_checksum(struct impl_extattr_entry *implext)
1946 {
1947 	uint16_t   *spos;
1948 
1949 	if (strncmp(implext->imp_id.id, "*UDF", 4) == 0) {
1950 		/* set checksum */
1951 		spos = (uint16_t *) implext->data;
1952 		*spos = udf_rw16(udf_ea_cksum((uint8_t *) implext));
1953 	}
1954 }
1955 
1956 
1957 int
1958 udf_extattr_search_intern(struct udf_node *node,
1959 	uint32_t sattr, char const *sattrname,
1960 	uint32_t *offsetp, uint32_t *lengthp)
1961 {
1962 	struct extattrhdr_desc    *eahdr;
1963 	struct extattr_entry      *attrhdr;
1964 	struct impl_extattr_entry *implext;
1965 	uint32_t    offset, a_l, sector_size;
1966 	 int32_t    l_ea;
1967 	uint8_t    *pos;
1968 	int         error;
1969 
1970 	/* get mountpoint */
1971 	sector_size = node->ump->discinfo.sector_size;
1972 
1973 	/* get information from fe/efe */
1974 	if (node->fe) {
1975 		l_ea  = udf_rw32(node->fe->l_ea);
1976 		eahdr = (struct extattrhdr_desc *) node->fe->data;
1977 	} else {
1978 		assert(node->efe);
1979 		l_ea  = udf_rw32(node->efe->l_ea);
1980 		eahdr = (struct extattrhdr_desc *) node->efe->data;
1981 	}
1982 
1983 	/* something recorded here? */
1984 	if (l_ea == 0)
1985 		return ENOENT;
1986 
1987 	/* check extended attribute tag; what to do if it fails? */
1988 	error = udf_check_tag(eahdr);
1989 	if (error)
1990 		return EINVAL;
1991 	if (udf_rw16(eahdr->tag.id) != TAGID_EXTATTR_HDR)
1992 		return EINVAL;
1993 	error = udf_check_tag_payload(eahdr, sizeof(struct extattrhdr_desc));
1994 	if (error)
1995 		return EINVAL;
1996 
1997 	DPRINTF(EXTATTR, ("Found %d bytes of extended attributes\n", l_ea));
1998 
1999 	/* looking for Ecma-167 attributes? */
2000 	offset = sizeof(struct extattrhdr_desc);
2001 
2002 	/* looking for either implemenation use or application use */
2003 	if (sattr == 2048) {				/* [4/48.10.8] */
2004 		offset = udf_rw32(eahdr->impl_attr_loc);
2005 		if (offset == UDF_IMPL_ATTR_LOC_NOT_PRESENT)
2006 			return ENOENT;
2007 	}
2008 	if (sattr == 65536) {				/* [4/48.10.9] */
2009 		offset = udf_rw32(eahdr->appl_attr_loc);
2010 		if (offset == UDF_APPL_ATTR_LOC_NOT_PRESENT)
2011 			return ENOENT;
2012 	}
2013 
2014 	/* paranoia check offset and l_ea */
2015 	if (l_ea + offset >= sector_size - sizeof(struct extattr_entry))
2016 		return EINVAL;
2017 
2018 	DPRINTF(EXTATTR, ("Starting at offset %d\n", offset));
2019 
2020 	/* find our extended attribute  */
2021 	l_ea -= offset;
2022 	pos = (uint8_t *) eahdr + offset;
2023 
2024 	while (l_ea >= sizeof(struct extattr_entry)) {
2025 		DPRINTF(EXTATTR, ("%d extended attr bytes left\n", l_ea));
2026 		attrhdr = (struct extattr_entry *) pos;
2027 		implext = (struct impl_extattr_entry *) pos;
2028 
2029 		/* get complete attribute length and check for roque values */
2030 		a_l = udf_rw32(attrhdr->a_l);
2031 		DPRINTF(EXTATTR, ("attribute %d:%d, len %d/%d\n",
2032 				udf_rw32(attrhdr->type),
2033 				attrhdr->subtype, a_l, l_ea));
2034 		if ((a_l == 0) || (a_l > l_ea))
2035 			return EINVAL;
2036 
2037 		if (attrhdr->type != sattr)
2038 			goto next_attribute;
2039 
2040 		/* we might have found it! */
2041 		if (attrhdr->type < 2048) {	/* Ecma-167 attribute */
2042 			*offsetp = offset;
2043 			*lengthp = a_l;
2044 			return 0;		/* success */
2045 		}
2046 
2047 		/*
2048 		 * Implementation use and application use extended attributes
2049 		 * have a name to identify. They share the same structure only
2050 		 * UDF implementation use extended attributes have a checksum
2051 		 * we need to check
2052 		 */
2053 
2054 		DPRINTF(EXTATTR, ("named attribute %s\n", implext->imp_id.id));
2055 		if (strcmp(implext->imp_id.id, sattrname) == 0) {
2056 			/* we have found our appl/implementation attribute */
2057 			*offsetp = offset;
2058 			*lengthp = a_l;
2059 			return 0;		/* success */
2060 		}
2061 
2062 next_attribute:
2063 		/* next attribute */
2064 		pos    += a_l;
2065 		l_ea   -= a_l;
2066 		offset += a_l;
2067 	}
2068 	/* not found */
2069 	return ENOENT;
2070 }
2071 
2072 
2073 
2074 /* --------------------------------------------------------------------- */
2075 
2076 static int
2077 udf_update_lvid_from_vat_extattr(struct udf_node *vat_node)
2078 {
2079 	struct udf_mount       *ump;
2080 	struct udf_logvol_info *lvinfo;
2081 	struct impl_extattr_entry     *implext;
2082 	struct vatlvext_extattr_entry  lvext;
2083 	const char *extstr = "*UDF VAT LVExtension";
2084 	uint64_t    vat_uniqueid;
2085 	uint32_t    offset, a_l;
2086 	uint8_t    *ea_start, *lvextpos;
2087 	int         error;
2088 
2089 	/* get mountpoint and lvinfo */
2090 	ump    = vat_node->ump;
2091 	lvinfo = ump->logvol_info;
2092 
2093 	/* get information from fe/efe */
2094 	if (vat_node->fe) {
2095 		vat_uniqueid = udf_rw64(vat_node->fe->unique_id);
2096 		ea_start     = vat_node->fe->data;
2097 	} else {
2098 		vat_uniqueid = udf_rw64(vat_node->efe->unique_id);
2099 		ea_start     = vat_node->efe->data;
2100 	}
2101 
2102 	error = udf_extattr_search_intern(vat_node, 2048, extstr, &offset, &a_l);
2103 	if (error)
2104 		return error;
2105 
2106 	implext = (struct impl_extattr_entry *) (ea_start + offset);
2107 	error = udf_impl_extattr_check(implext);
2108 	if (error)
2109 		return error;
2110 
2111 	/* paranoia */
2112 	if (a_l != sizeof(*implext) -1 + udf_rw32(implext->iu_l) + sizeof(lvext)) {
2113 		DPRINTF(VOLUMES, ("VAT LVExtension size doesn't compute\n"));
2114 		return EINVAL;
2115 	}
2116 
2117 	/*
2118 	 * we have found our "VAT LVExtension attribute. BUT due to a
2119 	 * bug in the specification it might not be word aligned so
2120 	 * copy first to avoid panics on some machines (!!)
2121 	 */
2122 	DPRINTF(VOLUMES, ("Found VAT LVExtension attr\n"));
2123 	lvextpos = implext->data + udf_rw32(implext->iu_l);
2124 	memcpy(&lvext, lvextpos, sizeof(lvext));
2125 
2126 	/* check if it was updated the last time */
2127 	if (udf_rw64(lvext.unique_id_chk) == vat_uniqueid) {
2128 		lvinfo->num_files       = lvext.num_files;
2129 		lvinfo->num_directories = lvext.num_directories;
2130 		udf_update_logvolname(ump, lvext.logvol_id);
2131 	} else {
2132 		DPRINTF(VOLUMES, ("VAT LVExtension out of date\n"));
2133 		/* replace VAT LVExt by free space EA */
2134 		memset(implext->imp_id.id, 0, UDF_REGID_ID_SIZE);
2135 		strcpy(implext->imp_id.id, "*UDF FreeEASpace");
2136 		udf_calc_impl_extattr_checksum(implext);
2137 	}
2138 
2139 	return 0;
2140 }
2141 
2142 
2143 static int
2144 udf_update_vat_extattr_from_lvid(struct udf_node *vat_node)
2145 {
2146 	struct udf_mount       *ump;
2147 	struct udf_logvol_info *lvinfo;
2148 	struct impl_extattr_entry     *implext;
2149 	struct vatlvext_extattr_entry  lvext;
2150 	const char *extstr = "*UDF VAT LVExtension";
2151 	uint64_t    vat_uniqueid;
2152 	uint32_t    offset, a_l;
2153 	uint8_t    *ea_start, *lvextpos;
2154 	int         error;
2155 
2156 	/* get mountpoint and lvinfo */
2157 	ump    = vat_node->ump;
2158 	lvinfo = ump->logvol_info;
2159 
2160 	/* get information from fe/efe */
2161 	if (vat_node->fe) {
2162 		vat_uniqueid = udf_rw64(vat_node->fe->unique_id);
2163 		ea_start     = vat_node->fe->data;
2164 	} else {
2165 		vat_uniqueid = udf_rw64(vat_node->efe->unique_id);
2166 		ea_start     = vat_node->efe->data;
2167 	}
2168 
2169 	error = udf_extattr_search_intern(vat_node, 2048, extstr, &offset, &a_l);
2170 	if (error)
2171 		return error;
2172 	/* found, it existed */
2173 
2174 	/* paranoia */
2175 	implext = (struct impl_extattr_entry *) (ea_start + offset);
2176 	error = udf_impl_extattr_check(implext);
2177 	if (error) {
2178 		DPRINTF(VOLUMES, ("VAT LVExtension bad on update\n"));
2179 		return error;
2180 	}
2181 	/* it is correct */
2182 
2183 	/*
2184 	 * we have found our "VAT LVExtension attribute. BUT due to a
2185 	 * bug in the specification it might not be word aligned so
2186 	 * copy first to avoid panics on some machines (!!)
2187 	 */
2188 	DPRINTF(VOLUMES, ("Updating VAT LVExtension attr\n"));
2189 	lvextpos = implext->data + udf_rw32(implext->iu_l);
2190 
2191 	lvext.unique_id_chk   = vat_uniqueid;
2192 	lvext.num_files       = lvinfo->num_files;
2193 	lvext.num_directories = lvinfo->num_directories;
2194 	memmove(lvext.logvol_id, ump->logical_vol->logvol_id, 128);
2195 
2196 	memcpy(lvextpos, &lvext, sizeof(lvext));
2197 
2198 	return 0;
2199 }
2200 
2201 /* --------------------------------------------------------------------- */
2202 
2203 int
2204 udf_vat_read(struct udf_node *vat_node, uint8_t *blob, int size, uint32_t offset)
2205 {
2206 	struct udf_mount *ump = vat_node->ump;
2207 
2208 	if (offset + size > ump->vat_offset + ump->vat_entries * 4)
2209 		return EINVAL;
2210 
2211 	memcpy(blob, ump->vat_table + offset, size);
2212 	return 0;
2213 }
2214 
2215 int
2216 udf_vat_write(struct udf_node *vat_node, uint8_t *blob, int size, uint32_t offset)
2217 {
2218 	struct udf_mount *ump = vat_node->ump;
2219 	uint32_t offset_high;
2220 	uint8_t *new_vat_table;
2221 
2222 	/* extent VAT allocation if needed */
2223 	offset_high = offset + size;
2224 	if (offset_high >= ump->vat_table_alloc_len) {
2225 		/* realloc */
2226 		new_vat_table = realloc(ump->vat_table,
2227 			ump->vat_table_alloc_len + UDF_VAT_CHUNKSIZE,
2228 			M_UDFVOLD, M_WAITOK | M_CANFAIL);
2229 		if (!new_vat_table) {
2230 			printf("udf_vat_write: can't extent VAT, out of mem\n");
2231 			return ENOMEM;
2232 		}
2233 		ump->vat_table = new_vat_table;
2234 		ump->vat_table_alloc_len += UDF_VAT_CHUNKSIZE;
2235 	}
2236 	ump->vat_table_len = MAX(ump->vat_table_len, offset_high);
2237 
2238 	memcpy(ump->vat_table + offset, blob, size);
2239 	return 0;
2240 }
2241 
2242 /* --------------------------------------------------------------------- */
2243 
2244 /* TODO support previous VAT location writeout */
2245 static int
2246 udf_update_vat_descriptor(struct udf_mount *ump)
2247 {
2248 	struct udf_node *vat_node = ump->vat_node;
2249 	struct udf_logvol_info *lvinfo = ump->logvol_info;
2250 	struct icb_tag *icbtag;
2251 	struct udf_oldvat_tail *oldvat_tl;
2252 	struct udf_vat *vat;
2253 	uint64_t unique_id;
2254 	uint32_t lb_size;
2255 	uint8_t *raw_vat;
2256 	int filetype, error;
2257 
2258 	KASSERT(vat_node);
2259 	KASSERT(lvinfo);
2260 	lb_size = udf_rw32(ump->logical_vol->lb_size);
2261 
2262 	/* get our new unique_id */
2263 	unique_id = udf_advance_uniqueid(ump);
2264 
2265 	/* get information from fe/efe */
2266 	if (vat_node->fe) {
2267 		icbtag    = &vat_node->fe->icbtag;
2268 		vat_node->fe->unique_id = udf_rw64(unique_id);
2269 	} else {
2270 		icbtag = &vat_node->efe->icbtag;
2271 		vat_node->efe->unique_id = udf_rw64(unique_id);
2272 	}
2273 
2274 	/* Check icb filetype! it has to be 0 or UDF_ICB_FILETYPE_VAT */
2275 	filetype = icbtag->file_type;
2276 	KASSERT((filetype == 0) || (filetype == UDF_ICB_FILETYPE_VAT));
2277 
2278 	/* allocate piece to process head or tail of VAT file */
2279 	raw_vat = malloc(lb_size, M_TEMP, M_WAITOK);
2280 
2281 	if (filetype == 0) {
2282 		/*
2283 		 * Update "*UDF VAT LVExtension" extended attribute from the
2284 		 * lvint if present.
2285 		 */
2286 		udf_update_vat_extattr_from_lvid(vat_node);
2287 
2288 		/* setup identifying regid */
2289 		oldvat_tl = (struct udf_oldvat_tail *) raw_vat;
2290 		memset(oldvat_tl, 0, sizeof(struct udf_oldvat_tail));
2291 
2292 		udf_set_regid(&oldvat_tl->id, "*UDF Virtual Alloc Tbl");
2293 		udf_add_udf_regid(ump, &oldvat_tl->id);
2294 		oldvat_tl->prev_vat = udf_rw32(0xffffffff);
2295 
2296 		/* write out new tail of virtual allocation table file */
2297 		error = udf_vat_write(vat_node, raw_vat,
2298 			sizeof(struct udf_oldvat_tail), ump->vat_entries * 4);
2299 	} else {
2300 		/* compose the VAT2 header */
2301 		vat = (struct udf_vat *) raw_vat;
2302 		memset(vat, 0, sizeof(struct udf_vat));
2303 
2304 		vat->header_len       = udf_rw16(152);	/* as per spec */
2305 		vat->impl_use_len     = udf_rw16(0);
2306 		memmove(vat->logvol_id, ump->logical_vol->logvol_id, 128);
2307 		vat->prev_vat         = udf_rw32(0xffffffff);
2308 		vat->num_files        = lvinfo->num_files;
2309 		vat->num_directories  = lvinfo->num_directories;
2310 		vat->min_udf_readver  = lvinfo->min_udf_readver;
2311 		vat->min_udf_writever = lvinfo->min_udf_writever;
2312 		vat->max_udf_writever = lvinfo->max_udf_writever;
2313 
2314 		error = udf_vat_write(vat_node, raw_vat,
2315 			sizeof(struct udf_vat), 0);
2316 	}
2317 	free(raw_vat, M_TEMP);
2318 
2319 	return error;	/* success! */
2320 }
2321 
2322 
2323 int
2324 udf_writeout_vat(struct udf_mount *ump)
2325 {
2326 	struct udf_node *vat_node = ump->vat_node;
2327 	uint32_t vat_length;
2328 	int error;
2329 
2330 	KASSERT(vat_node);
2331 
2332 	DPRINTF(CALL, ("udf_writeout_vat\n"));
2333 
2334 	mutex_enter(&ump->allocate_mutex);
2335 	udf_update_vat_descriptor(ump);
2336 
2337 	/* write out the VAT contents ; TODO intelligent writing */
2338 	vat_length = ump->vat_table_len;
2339 	error = vn_rdwr(UIO_WRITE, vat_node->vnode,
2340 		ump->vat_table, ump->vat_table_len, 0,
2341 		UIO_SYSSPACE, IO_NODELOCKED, FSCRED, NULL, NULL);
2342 	if (error) {
2343 		printf("udf_writeout_vat: failed to write out VAT contents\n");
2344 		goto out;
2345 	}
2346 
2347 	mutex_exit(&ump->allocate_mutex);
2348 
2349 	vflushbuf(ump->vat_node->vnode, 1 /* sync */);
2350 	error = VOP_FSYNC(ump->vat_node->vnode,
2351 			FSCRED, FSYNC_WAIT, 0, 0);
2352 	if (error)
2353 		printf("udf_writeout_vat: error writing VAT node!\n");
2354 out:
2355 
2356 	return error;
2357 }
2358 
2359 /* --------------------------------------------------------------------- */
2360 
2361 /*
2362  * Read in relevant pieces of VAT file and check if its indeed a VAT file
2363  * descriptor. If OK, read in complete VAT file.
2364  */
2365 
2366 static int
2367 udf_check_for_vat(struct udf_node *vat_node)
2368 {
2369 	struct udf_mount *ump;
2370 	struct icb_tag   *icbtag;
2371 	struct timestamp *mtime;
2372 	struct udf_vat   *vat;
2373 	struct udf_oldvat_tail *oldvat_tl;
2374 	struct udf_logvol_info *lvinfo;
2375 	uint64_t  unique_id;
2376 	uint32_t  vat_length;
2377 	uint32_t  vat_offset, vat_entries, vat_table_alloc_len;
2378 	uint32_t  sector_size;
2379 	uint32_t *raw_vat;
2380 	uint8_t  *vat_table;
2381 	char     *regid_name;
2382 	int filetype;
2383 	int error;
2384 
2385 	/* vat_length is really 64 bits though impossible */
2386 
2387 	DPRINTF(VOLUMES, ("Checking for VAT\n"));
2388 	if (!vat_node)
2389 		return ENOENT;
2390 
2391 	/* get mount info */
2392 	ump = vat_node->ump;
2393 	sector_size = udf_rw32(ump->logical_vol->lb_size);
2394 
2395 	/* check assertions */
2396 	assert(vat_node->fe || vat_node->efe);
2397 	assert(ump->logvol_integrity);
2398 
2399 	/* set vnode type to regular file or we can't read from it! */
2400 	vat_node->vnode->v_type = VREG;
2401 
2402 	/* get information from fe/efe */
2403 	if (vat_node->fe) {
2404 		vat_length = udf_rw64(vat_node->fe->inf_len);
2405 		icbtag    = &vat_node->fe->icbtag;
2406 		mtime     = &vat_node->fe->mtime;
2407 		unique_id = udf_rw64(vat_node->fe->unique_id);
2408 	} else {
2409 		vat_length = udf_rw64(vat_node->efe->inf_len);
2410 		icbtag = &vat_node->efe->icbtag;
2411 		mtime  = &vat_node->efe->mtime;
2412 		unique_id = udf_rw64(vat_node->efe->unique_id);
2413 	}
2414 
2415 	/* Check icb filetype! it has to be 0 or UDF_ICB_FILETYPE_VAT */
2416 	filetype = icbtag->file_type;
2417 	if ((filetype != 0) && (filetype != UDF_ICB_FILETYPE_VAT))
2418 		return ENOENT;
2419 
2420 	DPRINTF(VOLUMES, ("\tPossible VAT length %d\n", vat_length));
2421 
2422 	vat_table_alloc_len =
2423 		((vat_length + UDF_VAT_CHUNKSIZE-1) / UDF_VAT_CHUNKSIZE)
2424 			* UDF_VAT_CHUNKSIZE;
2425 
2426 	vat_table = malloc(vat_table_alloc_len, M_UDFVOLD,
2427 		M_CANFAIL | M_WAITOK);
2428 	if (vat_table == NULL) {
2429 		printf("allocation of %d bytes failed for VAT\n",
2430 			vat_table_alloc_len);
2431 		return ENOMEM;
2432 	}
2433 
2434 	/* allocate piece to read in head or tail of VAT file */
2435 	raw_vat = malloc(sector_size, M_TEMP, M_WAITOK);
2436 
2437 	/*
2438 	 * check contents of the file if its the old 1.50 VAT table format.
2439 	 * Its notoriously broken and allthough some implementations support an
2440 	 * extention as defined in the UDF 1.50 errata document, its doubtfull
2441 	 * to be useable since a lot of implementations don't maintain it.
2442 	 */
2443 	lvinfo = ump->logvol_info;
2444 
2445 	if (filetype == 0) {
2446 		/* definition */
2447 		vat_offset  = 0;
2448 		vat_entries = (vat_length-36)/4;
2449 
2450 		/* read in tail of virtual allocation table file */
2451 		error = vn_rdwr(UIO_READ, vat_node->vnode,
2452 				(uint8_t *) raw_vat,
2453 				sizeof(struct udf_oldvat_tail),
2454 				vat_entries * 4,
2455 				UIO_SYSSPACE, IO_SYNC | IO_NODELOCKED, FSCRED,
2456 				NULL, NULL);
2457 		if (error)
2458 			goto out;
2459 
2460 		/* check 1.50 VAT */
2461 		oldvat_tl = (struct udf_oldvat_tail *) raw_vat;
2462 		regid_name = (char *) oldvat_tl->id.id;
2463 		error = strncmp(regid_name, "*UDF Virtual Alloc Tbl", 22);
2464 		if (error) {
2465 			DPRINTF(VOLUMES, ("VAT format 1.50 rejected\n"));
2466 			error = ENOENT;
2467 			goto out;
2468 		}
2469 
2470 		/*
2471 		 * update LVID from "*UDF VAT LVExtension" extended attribute
2472 		 * if present.
2473 		 */
2474 		udf_update_lvid_from_vat_extattr(vat_node);
2475 	} else {
2476 		/* read in head of virtual allocation table file */
2477 		error = vn_rdwr(UIO_READ, vat_node->vnode,
2478 				(uint8_t *) raw_vat,
2479 				sizeof(struct udf_vat), 0,
2480 				UIO_SYSSPACE, IO_SYNC | IO_NODELOCKED, FSCRED,
2481 				NULL, NULL);
2482 		if (error)
2483 			goto out;
2484 
2485 		/* definition */
2486 		vat = (struct udf_vat *) raw_vat;
2487 		vat_offset  = vat->header_len;
2488 		vat_entries = (vat_length - vat_offset)/4;
2489 
2490 		assert(lvinfo);
2491 		lvinfo->num_files        = vat->num_files;
2492 		lvinfo->num_directories  = vat->num_directories;
2493 		lvinfo->min_udf_readver  = vat->min_udf_readver;
2494 		lvinfo->min_udf_writever = vat->min_udf_writever;
2495 		lvinfo->max_udf_writever = vat->max_udf_writever;
2496 
2497 		udf_update_logvolname(ump, vat->logvol_id);
2498 	}
2499 
2500 	/* read in complete VAT file */
2501 	error = vn_rdwr(UIO_READ, vat_node->vnode,
2502 			vat_table,
2503 			vat_length, 0,
2504 			UIO_SYSSPACE, IO_SYNC | IO_NODELOCKED, FSCRED,
2505 			NULL, NULL);
2506 	if (error)
2507 		printf("read in of complete VAT file failed (error %d)\n",
2508 			error);
2509 	if (error)
2510 		goto out;
2511 
2512 	DPRINTF(VOLUMES, ("VAT format accepted, marking it closed\n"));
2513 	ump->logvol_integrity->lvint_next_unique_id = unique_id;
2514 	ump->logvol_integrity->integrity_type = udf_rw32(UDF_INTEGRITY_CLOSED);
2515 	ump->logvol_integrity->time           = *mtime;
2516 
2517 	ump->vat_table_len = vat_length;
2518 	ump->vat_table_alloc_len = vat_table_alloc_len;
2519 	ump->vat_table   = vat_table;
2520 	ump->vat_offset  = vat_offset;
2521 	ump->vat_entries = vat_entries;
2522 	ump->vat_last_free_lb = 0;		/* start at beginning */
2523 
2524 out:
2525 	if (error) {
2526 		if (vat_table)
2527 			free(vat_table, M_UDFVOLD);
2528 	}
2529 	free(raw_vat, M_TEMP);
2530 
2531 	return error;
2532 }
2533 
2534 /* --------------------------------------------------------------------- */
2535 
2536 static int
2537 udf_search_vat(struct udf_mount *ump, union udf_pmap *mapping)
2538 {
2539 	struct udf_node *vat_node;
2540 	struct long_ad	 icb_loc;
2541 	uint32_t early_vat_loc, late_vat_loc, vat_loc;
2542 	int error;
2543 
2544 	/* mapping info not needed */
2545 	mapping = mapping;
2546 
2547 	vat_loc = ump->last_possible_vat_location;
2548 	early_vat_loc = vat_loc - 256;	/* 8 blocks of 32 sectors */
2549 
2550 	DPRINTF(VOLUMES, ("1) last possible %d, early_vat_loc %d \n",
2551 		vat_loc, early_vat_loc));
2552 	early_vat_loc = MAX(early_vat_loc, ump->first_possible_vat_location);
2553 	late_vat_loc  = vat_loc + 1024;
2554 
2555 	DPRINTF(VOLUMES, ("2) last possible %d, early_vat_loc %d \n",
2556 		vat_loc, early_vat_loc));
2557 
2558 	/* start looking from the end of the range */
2559 	do {
2560 		DPRINTF(VOLUMES, ("Checking for VAT at sector %d\n", vat_loc));
2561 		icb_loc.loc.part_num = udf_rw16(UDF_VTOP_RAWPART);
2562 		icb_loc.loc.lb_num   = udf_rw32(vat_loc);
2563 
2564 		error = udf_get_node(ump, &icb_loc, &vat_node);
2565 		if (!error) {
2566 			error = udf_check_for_vat(vat_node);
2567 			DPRINTFIF(VOLUMES, !error,
2568 				("VAT accepted at %d\n", vat_loc));
2569 			if (!error)
2570 				break;
2571 		}
2572 		if (vat_node) {
2573 			vput(vat_node->vnode);
2574 			vat_node = NULL;
2575 		}
2576 		vat_loc--;	/* walk backwards */
2577 	} while (vat_loc >= early_vat_loc);
2578 
2579 	/* keep our VAT node around */
2580 	if (vat_node) {
2581 		UDF_SET_SYSTEMFILE(vat_node->vnode);
2582 		ump->vat_node = vat_node;
2583 	}
2584 
2585 	return error;
2586 }
2587 
2588 /* --------------------------------------------------------------------- */
2589 
2590 static int
2591 udf_read_sparables(struct udf_mount *ump, union udf_pmap *mapping)
2592 {
2593 	union dscrptr *dscr;
2594 	struct part_map_spare *pms = &mapping->pms;
2595 	uint32_t lb_num;
2596 	int spar, error;
2597 
2598 	/*
2599 	 * The partition mapping passed on to us specifies the information we
2600 	 * need to locate and initialise the sparable partition mapping
2601 	 * information we need.
2602 	 */
2603 
2604 	DPRINTF(VOLUMES, ("Read sparable table\n"));
2605 	ump->sparable_packet_size = udf_rw16(pms->packet_len);
2606 	KASSERT(ump->sparable_packet_size >= ump->packet_size);	/* XXX */
2607 
2608 	for (spar = 0; spar < pms->n_st; spar++) {
2609 		lb_num = pms->st_loc[spar];
2610 		DPRINTF(VOLUMES, ("Checking for sparing table %d\n", lb_num));
2611 		error = udf_read_phys_dscr(ump, lb_num, M_UDFVOLD, &dscr);
2612 		if (!error && dscr) {
2613 			if (udf_rw16(dscr->tag.id) == TAGID_SPARING_TABLE) {
2614 				if (ump->sparing_table)
2615 					free(ump->sparing_table, M_UDFVOLD);
2616 				ump->sparing_table = &dscr->spt;
2617 				dscr = NULL;
2618 				DPRINTF(VOLUMES,
2619 				    ("Sparing table accepted (%d entries)\n",
2620 				     udf_rw16(ump->sparing_table->rt_l)));
2621 				break;	/* we're done */
2622 			}
2623 		}
2624 		if (dscr)
2625 			free(dscr, M_UDFVOLD);
2626 	}
2627 
2628 	if (ump->sparing_table)
2629 		return 0;
2630 
2631 	return ENOENT;
2632 }
2633 
2634 /* --------------------------------------------------------------------- */
2635 
2636 static int
2637 udf_read_metadata_nodes(struct udf_mount *ump, union udf_pmap *mapping)
2638 {
2639 	struct part_map_meta *pmm = &mapping->pmm;
2640 	struct long_ad	 icb_loc;
2641 	struct vnode *vp;
2642 	int error;
2643 
2644 	DPRINTF(VOLUMES, ("Reading in Metadata files\n"));
2645 	icb_loc.loc.part_num = pmm->part_num;
2646 	icb_loc.loc.lb_num   = pmm->meta_file_lbn;
2647 	DPRINTF(VOLUMES, ("Metadata file\n"));
2648 	error = udf_get_node(ump, &icb_loc, &ump->metadata_node);
2649 	if (ump->metadata_node) {
2650 		vp = ump->metadata_node->vnode;
2651 		UDF_SET_SYSTEMFILE(vp);
2652 	}
2653 
2654 	icb_loc.loc.lb_num   = pmm->meta_mirror_file_lbn;
2655 	if (icb_loc.loc.lb_num != -1) {
2656 		DPRINTF(VOLUMES, ("Metadata copy file\n"));
2657 		error = udf_get_node(ump, &icb_loc, &ump->metadatamirror_node);
2658 		if (ump->metadatamirror_node) {
2659 			vp = ump->metadatamirror_node->vnode;
2660 			UDF_SET_SYSTEMFILE(vp);
2661 		}
2662 	}
2663 
2664 	icb_loc.loc.lb_num   = pmm->meta_bitmap_file_lbn;
2665 	if (icb_loc.loc.lb_num != -1) {
2666 		DPRINTF(VOLUMES, ("Metadata bitmap file\n"));
2667 		error = udf_get_node(ump, &icb_loc, &ump->metadatabitmap_node);
2668 		if (ump->metadatabitmap_node) {
2669 			vp = ump->metadatabitmap_node->vnode;
2670 			UDF_SET_SYSTEMFILE(vp);
2671 		}
2672 	}
2673 
2674 	/* if we're mounting read-only we relax the requirements */
2675 	if (ump->vfs_mountp->mnt_flag & MNT_RDONLY) {
2676 		error = EFAULT;
2677 		if (ump->metadata_node)
2678 			error = 0;
2679 		if ((ump->metadata_node == NULL) && (ump->metadatamirror_node)) {
2680 			printf( "udf mount: Metadata file not readable, "
2681 				"substituting Metadata copy file\n");
2682 			ump->metadata_node = ump->metadatamirror_node;
2683 			ump->metadatamirror_node = NULL;
2684 			error = 0;
2685 		}
2686 	} else {
2687 		/* mounting read/write */
2688 		DPRINTF(VOLUMES, ("udf mount: read only file system\n"));
2689 		error = EROFS;
2690 	}
2691 	DPRINTFIF(VOLUMES, error, ("udf mount: failed to read "
2692 				   "metadata files\n"));
2693 	return error;
2694 }
2695 
2696 /* --------------------------------------------------------------------- */
2697 
2698 int
2699 udf_read_vds_tables(struct udf_mount *ump)
2700 {
2701 	union udf_pmap *mapping;
2702 	/* struct udf_args *args = &ump->mount_args; */
2703 	uint32_t n_pm, mt_l;
2704 	uint32_t log_part;
2705 	uint8_t *pmap_pos;
2706 	int pmap_size;
2707 	int error;
2708 
2709 	/* Iterate again over the part mappings for locations   */
2710 	n_pm = udf_rw32(ump->logical_vol->n_pm);   /* num partmaps         */
2711 	mt_l = udf_rw32(ump->logical_vol->mt_l);   /* partmaps data length */
2712 	pmap_pos =  ump->logical_vol->maps;
2713 
2714 	for (log_part = 0; log_part < n_pm; log_part++) {
2715 		mapping = (union udf_pmap *) pmap_pos;
2716 		switch (ump->vtop_tp[log_part]) {
2717 		case UDF_VTOP_TYPE_PHYS :
2718 			/* nothing */
2719 			break;
2720 		case UDF_VTOP_TYPE_VIRT :
2721 			/* search and load VAT */
2722 			error = udf_search_vat(ump, mapping);
2723 			if (error)
2724 				return ENOENT;
2725 			break;
2726 		case UDF_VTOP_TYPE_SPARABLE :
2727 			/* load one of the sparable tables */
2728 			error = udf_read_sparables(ump, mapping);
2729 			if (error)
2730 				return ENOENT;
2731 			break;
2732 		case UDF_VTOP_TYPE_META :
2733 			/* load the associated file descriptors */
2734 			error = udf_read_metadata_nodes(ump, mapping);
2735 			if (error)
2736 				return ENOENT;
2737 			break;
2738 		default:
2739 			break;
2740 		}
2741 		pmap_size  = pmap_pos[1];
2742 		pmap_pos  += pmap_size;
2743 	}
2744 
2745 	return 0;
2746 }
2747 
2748 /* --------------------------------------------------------------------- */
2749 
2750 int
2751 udf_read_rootdirs(struct udf_mount *ump)
2752 {
2753 	union dscrptr *dscr;
2754 	/* struct udf_args *args = &ump->mount_args; */
2755 	struct udf_node *rootdir_node, *streamdir_node;
2756 	struct long_ad  fsd_loc, *dir_loc;
2757 	uint32_t lb_num, dummy;
2758 	uint32_t fsd_len;
2759 	int dscr_type;
2760 	int error;
2761 
2762 	/* TODO implement FSD reading in separate function like integrity? */
2763 	/* get fileset descriptor sequence */
2764 	fsd_loc = ump->logical_vol->lv_fsd_loc;
2765 	fsd_len = udf_rw32(fsd_loc.len);
2766 
2767 	dscr  = NULL;
2768 	error = 0;
2769 	while (fsd_len || error) {
2770 		DPRINTF(VOLUMES, ("fsd_len = %d\n", fsd_len));
2771 		/* translate fsd_loc to lb_num */
2772 		error = udf_translate_vtop(ump, &fsd_loc, &lb_num, &dummy);
2773 		if (error)
2774 			break;
2775 		DPRINTF(VOLUMES, ("Reading FSD at lb %d\n", lb_num));
2776 		error = udf_read_phys_dscr(ump, lb_num, M_UDFVOLD, &dscr);
2777 		/* end markers */
2778 		if (error || (dscr == NULL))
2779 			break;
2780 
2781 		/* analyse */
2782 		dscr_type = udf_rw16(dscr->tag.id);
2783 		if (dscr_type == TAGID_TERM)
2784 			break;
2785 		if (dscr_type != TAGID_FSD) {
2786 			free(dscr, M_UDFVOLD);
2787 			return ENOENT;
2788 		}
2789 
2790 		/*
2791 		 * TODO check for multiple fileset descriptors; its only
2792 		 * picking the last now. Also check for FSD
2793 		 * correctness/interpretability
2794 		 */
2795 
2796 		/* update */
2797 		if (ump->fileset_desc) {
2798 			free(ump->fileset_desc, M_UDFVOLD);
2799 		}
2800 		ump->fileset_desc = &dscr->fsd;
2801 		dscr = NULL;
2802 
2803 		/* continue to the next fsd */
2804 		fsd_len -= ump->discinfo.sector_size;
2805 		fsd_loc.loc.lb_num = udf_rw32(udf_rw32(fsd_loc.loc.lb_num)+1);
2806 
2807 		/* follow up to fsd->next_ex (long_ad) if its not null */
2808 		if (udf_rw32(ump->fileset_desc->next_ex.len)) {
2809 			DPRINTF(VOLUMES, ("follow up FSD extent\n"));
2810 			fsd_loc = ump->fileset_desc->next_ex;
2811 			fsd_len = udf_rw32(ump->fileset_desc->next_ex.len);
2812 		}
2813 	}
2814 	if (dscr)
2815 		free(dscr, M_UDFVOLD);
2816 
2817 	/* there has to be one */
2818 	if (ump->fileset_desc == NULL)
2819 		return ENOENT;
2820 
2821 	DPRINTF(VOLUMES, ("FSD read in fine\n"));
2822 	DPRINTF(VOLUMES, ("Updating fsd logical volume id\n"));
2823 	udf_update_logvolname(ump, ump->logical_vol->logvol_id);
2824 
2825 	/*
2826 	 * Now the FSD is known, read in the rootdirectory and if one exists,
2827 	 * the system stream dir. Some files in the system streamdir are not
2828 	 * wanted in this implementation since they are not maintained. If
2829 	 * writing is enabled we'll delete these files if they exist.
2830 	 */
2831 
2832 	rootdir_node = streamdir_node = NULL;
2833 	dir_loc = NULL;
2834 
2835 	/* try to read in the rootdir */
2836 	dir_loc = &ump->fileset_desc->rootdir_icb;
2837 	error = udf_get_node(ump, dir_loc, &rootdir_node);
2838 	if (error)
2839 		return ENOENT;
2840 
2841 	/* aparently it read in fine */
2842 
2843 	/*
2844 	 * Try the system stream directory; not very likely in the ones we
2845 	 * test, but for completeness.
2846 	 */
2847 	dir_loc = &ump->fileset_desc->streamdir_icb;
2848 	if (udf_rw32(dir_loc->len)) {
2849 		printf("udf_read_rootdirs: streamdir defined ");
2850 		error = udf_get_node(ump, dir_loc, &streamdir_node);
2851 		if (error) {
2852 			printf("but error in streamdir reading\n");
2853 		} else {
2854 			printf("but ignored\n");
2855 			/*
2856 			 * TODO process streamdir `baddies' i.e. files we dont
2857 			 * want if R/W
2858 			 */
2859 		}
2860 	}
2861 
2862 	DPRINTF(VOLUMES, ("Rootdir(s) read in fine\n"));
2863 
2864 	/* release the vnodes again; they'll be auto-recycled later */
2865 	if (streamdir_node) {
2866 		vput(streamdir_node->vnode);
2867 	}
2868 	if (rootdir_node) {
2869 		vput(rootdir_node->vnode);
2870 	}
2871 
2872 	return 0;
2873 }
2874 
2875 /* --------------------------------------------------------------------- */
2876 
2877 /* To make absolutely sure we are NOT returning zero, add one :) */
2878 
2879 long
2880 udf_calchash(struct long_ad *icbptr)
2881 {
2882 	/* ought to be enough since each mountpoint has its own chain */
2883 	return udf_rw32(icbptr->loc.lb_num) + 1;
2884 }
2885 
2886 
2887 static struct udf_node *
2888 udf_hash_lookup(struct udf_mount *ump, struct long_ad *icbptr)
2889 {
2890 	struct udf_node *node;
2891 	struct vnode *vp;
2892 	uint32_t hashline;
2893 
2894 loop:
2895 	mutex_enter(&ump->ihash_lock);
2896 
2897 	hashline = udf_calchash(icbptr) & UDF_INODE_HASHMASK;
2898 	LIST_FOREACH(node, &ump->udf_nodes[hashline], hashchain) {
2899 		assert(node);
2900 		if (node->loc.loc.lb_num   == icbptr->loc.lb_num &&
2901 		    node->loc.loc.part_num == icbptr->loc.part_num) {
2902 			vp = node->vnode;
2903 			assert(vp);
2904 			mutex_enter(&vp->v_interlock);
2905 			mutex_exit(&ump->ihash_lock);
2906 			if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK))
2907 				goto loop;
2908 			return node;
2909 		}
2910 	}
2911 	mutex_exit(&ump->ihash_lock);
2912 
2913 	return NULL;
2914 }
2915 
2916 
2917 static void
2918 udf_sorted_list_insert(struct udf_node *node)
2919 {
2920 	struct udf_mount *ump;
2921 	struct udf_node  *s_node, *last_node;
2922 	uint32_t loc, s_loc;
2923 
2924 	ump = node->ump;
2925 	last_node = NULL;	/* XXX gcc */
2926 
2927 	if (LIST_EMPTY(&ump->sorted_udf_nodes)) {
2928 		LIST_INSERT_HEAD(&ump->sorted_udf_nodes, node, sortchain);
2929 		return;
2930 	}
2931 
2932 	/*
2933 	 * We sort on logical block number here and not on physical block
2934 	 * number here. Ideally we should go for the physical block nr to get
2935 	 * better sync performance though this sort will ensure that packets
2936 	 * won't get spit up unnessisarily.
2937 	 */
2938 
2939 	loc = udf_rw32(node->loc.loc.lb_num);
2940 	LIST_FOREACH(s_node, &ump->sorted_udf_nodes, sortchain) {
2941 		s_loc = udf_rw32(s_node->loc.loc.lb_num);
2942 		if (s_loc > loc) {
2943 			LIST_INSERT_BEFORE(s_node, node, sortchain);
2944 			return;
2945 		}
2946 		last_node = s_node;
2947 	}
2948 	LIST_INSERT_AFTER(last_node, node, sortchain);
2949 }
2950 
2951 
2952 static void
2953 udf_register_node(struct udf_node *node)
2954 {
2955 	struct udf_mount *ump;
2956 	struct udf_node *chk;
2957 	uint32_t hashline;
2958 
2959 	ump = node->ump;
2960 	mutex_enter(&ump->ihash_lock);
2961 
2962 	/* add to our hash table */
2963 	hashline = udf_calchash(&node->loc) & UDF_INODE_HASHMASK;
2964 #ifdef DEBUG
2965 	LIST_FOREACH(chk, &ump->udf_nodes[hashline], hashchain) {
2966 		assert(chk);
2967 		if (chk->loc.loc.lb_num   == node->loc.loc.lb_num &&
2968 		    chk->loc.loc.part_num == node->loc.loc.part_num)
2969 			panic("Double node entered\n");
2970 	}
2971 #else
2972 	chk = NULL;
2973 #endif
2974 	LIST_INSERT_HEAD(&ump->udf_nodes[hashline], node, hashchain);
2975 
2976 	/* add to our sorted list */
2977 	udf_sorted_list_insert(node);
2978 
2979 	mutex_exit(&ump->ihash_lock);
2980 }
2981 
2982 
2983 static void
2984 udf_deregister_node(struct udf_node *node)
2985 {
2986 	struct udf_mount *ump;
2987 
2988 	ump = node->ump;
2989 	mutex_enter(&ump->ihash_lock);
2990 
2991 	/* from hash and sorted list */
2992 	LIST_REMOVE(node, hashchain);
2993 	LIST_REMOVE(node, sortchain);
2994 
2995 	mutex_exit(&ump->ihash_lock);
2996 }
2997 
2998 /* --------------------------------------------------------------------- */
2999 
3000 static void
3001 udf_inittag(struct udf_mount *ump, struct desc_tag *tag, int tagid,
3002 	uint32_t sector)
3003 {
3004 	assert(ump->logical_vol);
3005 
3006 	tag->id 		= udf_rw16(tagid);
3007 	tag->descriptor_ver	= ump->logical_vol->tag.descriptor_ver;
3008 	tag->cksum		= 0;
3009 	tag->reserved		= 0;
3010 	tag->serial_num		= ump->logical_vol->tag.serial_num;
3011 	tag->tag_loc            = udf_rw32(sector);
3012 }
3013 
3014 
3015 uint64_t
3016 udf_advance_uniqueid(struct udf_mount *ump)
3017 {
3018 	uint64_t unique_id;
3019 
3020 	mutex_enter(&ump->logvol_mutex);
3021 	unique_id = udf_rw64(ump->logvol_integrity->lvint_next_unique_id);
3022 	if (unique_id < 0x10)
3023 		unique_id = 0x10;
3024 	ump->logvol_integrity->lvint_next_unique_id = udf_rw64(unique_id + 1);
3025 	mutex_exit(&ump->logvol_mutex);
3026 
3027 	return unique_id;
3028 }
3029 
3030 
3031 static void
3032 udf_adjust_filecount(struct udf_node *udf_node, int sign)
3033 {
3034 	struct udf_mount *ump = udf_node->ump;
3035 	uint32_t num_dirs, num_files;
3036 	int udf_file_type;
3037 
3038 	/* get file type */
3039 	if (udf_node->fe) {
3040 		udf_file_type = udf_node->fe->icbtag.file_type;
3041 	} else {
3042 		udf_file_type = udf_node->efe->icbtag.file_type;
3043 	}
3044 
3045 	/* adjust file count */
3046 	mutex_enter(&ump->allocate_mutex);
3047 	if (udf_file_type == UDF_ICB_FILETYPE_DIRECTORY) {
3048 		num_dirs = udf_rw32(ump->logvol_info->num_directories);
3049 		ump->logvol_info->num_directories =
3050 			udf_rw32((num_dirs + sign));
3051 	} else {
3052 		num_files = udf_rw32(ump->logvol_info->num_files);
3053 		ump->logvol_info->num_files =
3054 			udf_rw32((num_files + sign));
3055 	}
3056 	mutex_exit(&ump->allocate_mutex);
3057 }
3058 
3059 
3060 void
3061 udf_osta_charset(struct charspec *charspec)
3062 {
3063 	bzero(charspec, sizeof(struct charspec));
3064 	charspec->type = 0;
3065 	strcpy((char *) charspec->inf, "OSTA Compressed Unicode");
3066 }
3067 
3068 
3069 /* first call udf_set_regid and then the suffix */
3070 void
3071 udf_set_regid(struct regid *regid, char const *name)
3072 {
3073 	bzero(regid, sizeof(struct regid));
3074 	regid->flags    = 0;		/* not dirty and not protected */
3075 	strcpy((char *) regid->id, name);
3076 }
3077 
3078 
3079 void
3080 udf_add_domain_regid(struct udf_mount *ump, struct regid *regid)
3081 {
3082 	uint16_t *ver;
3083 
3084 	ver  = (uint16_t *) regid->id_suffix;
3085 	*ver = ump->logvol_info->min_udf_readver;
3086 }
3087 
3088 
3089 void
3090 udf_add_udf_regid(struct udf_mount *ump, struct regid *regid)
3091 {
3092 	uint16_t *ver;
3093 
3094 	ver  = (uint16_t *) regid->id_suffix;
3095 	*ver = ump->logvol_info->min_udf_readver;
3096 
3097 	regid->id_suffix[2] = 4;	/* unix */
3098 	regid->id_suffix[3] = 8;	/* NetBSD */
3099 }
3100 
3101 
3102 void
3103 udf_add_impl_regid(struct udf_mount *ump, struct regid *regid)
3104 {
3105 	regid->id_suffix[0] = 4;	/* unix */
3106 	regid->id_suffix[1] = 8;	/* NetBSD */
3107 }
3108 
3109 
3110 void
3111 udf_add_app_regid(struct udf_mount *ump, struct regid *regid)
3112 {
3113 	regid->id_suffix[0] = APP_VERSION_MAIN;
3114 	regid->id_suffix[1] = APP_VERSION_SUB;
3115 }
3116 
3117 static int
3118 udf_create_parentfid(struct udf_mount *ump, struct fileid_desc *fid,
3119 	struct long_ad *parent, uint64_t unique_id)
3120 {
3121 	/* the size of an empty FID is 38 but needs to be a multiple of 4 */
3122 	int fidsize = 40;
3123 
3124 	udf_inittag(ump, &fid->tag, TAGID_FID, udf_rw32(parent->loc.lb_num));
3125 	fid->file_version_num = udf_rw16(1);	/* UDF 2.3.4.1 */
3126 	fid->file_char = UDF_FILE_CHAR_DIR | UDF_FILE_CHAR_PAR;
3127 	fid->icb = *parent;
3128 	fid->icb.longad_uniqueid = udf_rw32((uint32_t) unique_id);
3129 	fid->tag.desc_crc_len = fidsize - UDF_DESC_TAG_LENGTH;
3130 	(void) udf_validate_tag_and_crc_sums((union dscrptr *) fid);
3131 
3132 	return fidsize;
3133 }
3134 
3135 /* --------------------------------------------------------------------- */
3136 
3137 int
3138 udf_open_logvol(struct udf_mount *ump)
3139 {
3140 	int logvol_integrity;
3141 	int error;
3142 
3143 	/* already/still open? */
3144 	logvol_integrity = udf_rw32(ump->logvol_integrity->integrity_type);
3145 	if (logvol_integrity == UDF_INTEGRITY_OPEN)
3146 		return 0;
3147 
3148 	/* can we open it ? */
3149 	if (ump->vfs_mountp->mnt_flag & MNT_RDONLY)
3150 		return EROFS;
3151 
3152 	/* setup write parameters */
3153 	DPRINTF(VOLUMES, ("Setting up write parameters\n"));
3154 	if ((error = udf_setup_writeparams(ump)) != 0)
3155 		return error;
3156 
3157 	/* determine data and metadata tracks (most likely same) */
3158 	error = udf_search_writing_tracks(ump);
3159 	if (error) {
3160 		/* most likely lack of space */
3161 		printf("udf_open_logvol: error searching writing tracks\n");
3162 		return EROFS;
3163 	}
3164 
3165 	/* writeout/update lvint on disc or only in memory */
3166 	DPRINTF(VOLUMES, ("Opening logical volume\n"));
3167 	if (ump->lvopen & UDF_OPEN_SESSION) {
3168 		/* TODO implement writeout of VRS + VDS */
3169 		printf( "udf_open_logvol:Opening a closed session not yet "
3170 			"implemented\n");
3171 		return EROFS;
3172 
3173 		/* determine data and metadata tracks again */
3174 		error = udf_search_writing_tracks(ump);
3175 	}
3176 
3177 	/* mark it open */
3178 	ump->logvol_integrity->integrity_type = udf_rw32(UDF_INTEGRITY_OPEN);
3179 
3180 	/* do we need to write it out? */
3181 	if (ump->lvopen & UDF_WRITE_LVINT) {
3182 		error = udf_writeout_lvint(ump, ump->lvopen);
3183 		/* if we couldn't write it mark it closed again */
3184 		if (error) {
3185 			ump->logvol_integrity->integrity_type =
3186 						udf_rw32(UDF_INTEGRITY_CLOSED);
3187 			return error;
3188 		}
3189 	}
3190 
3191 	return 0;
3192 }
3193 
3194 
3195 int
3196 udf_close_logvol(struct udf_mount *ump, int mntflags)
3197 {
3198 	int logvol_integrity;
3199 	int error = 0;
3200 	int n;
3201 
3202 	/* already/still closed? */
3203 	logvol_integrity = udf_rw32(ump->logvol_integrity->integrity_type);
3204 	if (logvol_integrity == UDF_INTEGRITY_CLOSED)
3205 		return 0;
3206 
3207 	/* writeout/update lvint or write out VAT */
3208 	DPRINTF(VOLUMES, ("Closing logical volume\n"));
3209 	if (ump->lvclose & UDF_WRITE_VAT) {
3210 		DPRINTF(VOLUMES, ("lvclose & UDF_WRITE_VAT\n"));
3211 
3212 		/* preprocess the VAT node; its modified on every writeout */
3213 		DPRINTF(VOLUMES, ("writeout vat_node\n"));
3214 		udf_update_vat_descriptor(ump->vat_node->ump);
3215 
3216 		/* write out the VAT node */
3217 		vflushbuf(ump->vat_node->vnode, 1 /* sync */);
3218 		for (n = 0; n < 16; n++) {
3219 			ump->vat_node->i_flags |= IN_MODIFIED;
3220 			error = VOP_FSYNC(ump->vat_node->vnode,
3221 					FSCRED, FSYNC_WAIT, 0, 0);
3222 		}
3223 		if (error) {
3224 			printf("udf_close_logvol: writeout of VAT failed\n");
3225 			return error;
3226 		}
3227 	}
3228 
3229 	if (ump->lvclose & UDF_WRITE_PART_BITMAPS) {
3230 		error = udf_write_partition_spacetables(ump, 1 /* waitfor */);
3231 		if (error) {
3232 			printf( "udf_close_logvol: writeout of space tables "
3233 				"failed\n");
3234 			return error;
3235 		}
3236 		ump->lvclose &= ~UDF_WRITE_PART_BITMAPS;
3237 	}
3238 
3239 	if (ump->lvclose & UDF_CLOSE_SESSION) {
3240 		printf("TODO: Closing a session is not yet implemented\n");
3241 		return EROFS;
3242 		ump->lvopen |= UDF_OPEN_SESSION;
3243 	}
3244 
3245 	/* mark it closed */
3246 	ump->logvol_integrity->integrity_type = udf_rw32(UDF_INTEGRITY_CLOSED);
3247 
3248 	/* do we need to write out the logical volume integrity */
3249 	if (ump->lvclose & UDF_WRITE_LVINT)
3250 		error = udf_writeout_lvint(ump, ump->lvopen);
3251 	if (error) {
3252 		/* HELP now what? mark it open again for now */
3253 		ump->logvol_integrity->integrity_type =
3254 			udf_rw32(UDF_INTEGRITY_OPEN);
3255 		return error;
3256 	}
3257 
3258 	(void) udf_synchronise_caches(ump);
3259 
3260 	return 0;
3261 }
3262 
3263 /* --------------------------------------------------------------------- */
3264 
3265 /*
3266  * Genfs interfacing
3267  *
3268  * static const struct genfs_ops udf_genfsops = {
3269  * 	.gop_size = genfs_size,
3270  * 		size of transfers
3271  * 	.gop_alloc = udf_gop_alloc,
3272  * 		allocate len bytes at offset
3273  * 	.gop_write = genfs_gop_write,
3274  * 		putpages interface code
3275  * 	.gop_markupdate = udf_gop_markupdate,
3276  * 		set update/modify flags etc.
3277  * }
3278  */
3279 
3280 /*
3281  * Genfs interface. These four functions are the only ones defined though not
3282  * documented... great....
3283  */
3284 
3285 /*
3286  * Callback from genfs to allocate len bytes at offset off; only called when
3287  * filling up gaps in the allocation.
3288  */
3289 /* XXX should be check if there is space enough in udf_gop_alloc? */
3290 static int
3291 udf_gop_alloc(struct vnode *vp, off_t off,
3292     off_t len, int flags, kauth_cred_t cred)
3293 {
3294 #if 0
3295 	struct udf_node *udf_node = VTOI(vp);
3296 	struct udf_mount *ump = udf_node->ump;
3297 	uint32_t lb_size, num_lb;
3298 #endif
3299 
3300 	DPRINTF(NOTIMPL, ("udf_gop_alloc not implemented\n"));
3301 	DPRINTF(ALLOC, ("udf_gop_alloc called for %"PRIu64" bytes\n", len));
3302 
3303 	return 0;
3304 }
3305 
3306 
3307 /*
3308  * callback from genfs to update our flags
3309  */
3310 static void
3311 udf_gop_markupdate(struct vnode *vp, int flags)
3312 {
3313 	struct udf_node *udf_node = VTOI(vp);
3314 	u_long mask = 0;
3315 
3316 	if ((flags & GOP_UPDATE_ACCESSED) != 0) {
3317 		mask = IN_ACCESS;
3318 	}
3319 	if ((flags & GOP_UPDATE_MODIFIED) != 0) {
3320 		if (vp->v_type == VREG) {
3321 			mask |= IN_CHANGE | IN_UPDATE;
3322 		} else {
3323 			mask |= IN_MODIFY;
3324 		}
3325 	}
3326 	if (mask) {
3327 		udf_node->i_flags |= mask;
3328 	}
3329 }
3330 
3331 
3332 static const struct genfs_ops udf_genfsops = {
3333 	.gop_size = genfs_size,
3334 	.gop_alloc = udf_gop_alloc,
3335 	.gop_write = genfs_gop_write_rwmap,
3336 	.gop_markupdate = udf_gop_markupdate,
3337 };
3338 
3339 
3340 /* --------------------------------------------------------------------- */
3341 
3342 int
3343 udf_write_terminator(struct udf_mount *ump, uint32_t sector)
3344 {
3345 	union dscrptr *dscr;
3346 	int error;
3347 
3348 	dscr = malloc(ump->discinfo.sector_size, M_TEMP, M_WAITOK);
3349 	bzero(dscr, ump->discinfo.sector_size);
3350 	udf_inittag(ump, &dscr->tag, TAGID_TERM, sector);
3351 
3352 	/* CRC length for an anchor is 512 - tag length; defined in Ecma 167 */
3353 	dscr->tag.desc_crc_len = udf_rw16(512-UDF_DESC_TAG_LENGTH);
3354 	(void) udf_validate_tag_and_crc_sums(dscr);
3355 
3356 	error = udf_write_phys_dscr_sync(ump, NULL, UDF_C_DSCR,
3357 			dscr, sector, sector);
3358 
3359 	free(dscr, M_TEMP);
3360 
3361 	return error;
3362 }
3363 
3364 
3365 /* --------------------------------------------------------------------- */
3366 
3367 /* UDF<->unix converters */
3368 
3369 /* --------------------------------------------------------------------- */
3370 
3371 static mode_t
3372 udf_perm_to_unix_mode(uint32_t perm)
3373 {
3374 	mode_t mode;
3375 
3376 	mode  = ((perm & UDF_FENTRY_PERM_USER_MASK)      );
3377 	mode |= ((perm & UDF_FENTRY_PERM_GRP_MASK  ) >> 2);
3378 	mode |= ((perm & UDF_FENTRY_PERM_OWNER_MASK) >> 4);
3379 
3380 	return mode;
3381 }
3382 
3383 /* --------------------------------------------------------------------- */
3384 
3385 static uint32_t
3386 unix_mode_to_udf_perm(mode_t mode)
3387 {
3388 	uint32_t perm;
3389 
3390 	perm  = ((mode & S_IRWXO)     );
3391 	perm |= ((mode & S_IRWXG) << 2);
3392 	perm |= ((mode & S_IRWXU) << 4);
3393 	perm |= ((mode & S_IWOTH) << 3);
3394 	perm |= ((mode & S_IWGRP) << 5);
3395 	perm |= ((mode & S_IWUSR) << 7);
3396 
3397 	return perm;
3398 }
3399 
3400 /* --------------------------------------------------------------------- */
3401 
3402 static uint32_t
3403 udf_icb_to_unix_filetype(uint32_t icbftype)
3404 {
3405 	switch (icbftype) {
3406 	case UDF_ICB_FILETYPE_DIRECTORY :
3407 	case UDF_ICB_FILETYPE_STREAMDIR :
3408 		return S_IFDIR;
3409 	case UDF_ICB_FILETYPE_FIFO :
3410 		return S_IFIFO;
3411 	case UDF_ICB_FILETYPE_CHARDEVICE :
3412 		return S_IFCHR;
3413 	case UDF_ICB_FILETYPE_BLOCKDEVICE :
3414 		return S_IFBLK;
3415 	case UDF_ICB_FILETYPE_RANDOMACCESS :
3416 	case UDF_ICB_FILETYPE_REALTIME :
3417 		return S_IFREG;
3418 	case UDF_ICB_FILETYPE_SYMLINK :
3419 		return S_IFLNK;
3420 	case UDF_ICB_FILETYPE_SOCKET :
3421 		return S_IFSOCK;
3422 	}
3423 	/* no idea what this is */
3424 	return 0;
3425 }
3426 
3427 /* --------------------------------------------------------------------- */
3428 
3429 void
3430 udf_to_unix_name(char *result, int result_len, char *id, int len,
3431 	struct charspec *chsp)
3432 {
3433 	uint16_t   *raw_name, *unix_name;
3434 	uint16_t   *inchp, ch;
3435 	uint8_t	   *outchp;
3436 	const char *osta_id = "OSTA Compressed Unicode";
3437 	int         ucode_chars, nice_uchars, is_osta_typ0, nout;
3438 
3439 	raw_name = malloc(2048 * sizeof(uint16_t), M_UDFTEMP, M_WAITOK);
3440 	unix_name = raw_name + 1024;			/* split space in half */
3441 	assert(sizeof(char) == sizeof(uint8_t));
3442 	outchp = (uint8_t *) result;
3443 
3444 	is_osta_typ0  = (chsp->type == 0);
3445 	is_osta_typ0 &= (strcmp((char *) chsp->inf, osta_id) == 0);
3446 	if (is_osta_typ0) {
3447 		/* TODO clean up */
3448 		*raw_name = *unix_name = 0;
3449 		ucode_chars = udf_UncompressUnicode(len, (uint8_t *) id, raw_name);
3450 		ucode_chars = MIN(ucode_chars, UnicodeLength((unicode_t *) raw_name));
3451 		nice_uchars = UDFTransName(unix_name, raw_name, ucode_chars);
3452 		/* output UTF8 */
3453 		for (inchp = unix_name; nice_uchars>0; inchp++, nice_uchars--) {
3454 			ch = *inchp;
3455 			nout = wput_utf8(outchp, result_len, ch);
3456 			outchp += nout; result_len -= nout;
3457 			if (!ch) break;
3458 		}
3459 		*outchp++ = 0;
3460 	} else {
3461 		/* assume 8bit char length byte latin-1 */
3462 		assert(*id == 8);
3463 		assert(strlen((char *) (id+1)) <= MAXNAMLEN);
3464 		strncpy((char *) result, (char *) (id+1), strlen((char *) (id+1)));
3465 	}
3466 	free(raw_name, M_UDFTEMP);
3467 }
3468 
3469 /* --------------------------------------------------------------------- */
3470 
3471 void
3472 unix_to_udf_name(char *result, uint8_t *result_len, char const *name, int name_len,
3473 	struct charspec *chsp)
3474 {
3475 	uint16_t   *raw_name;
3476 	uint16_t   *outchp;
3477 	const char *inchp;
3478 	const char *osta_id = "OSTA Compressed Unicode";
3479 	int         udf_chars, is_osta_typ0, bits;
3480 	size_t      cnt;
3481 
3482 	/* allocate temporary unicode-16 buffer */
3483 	raw_name = malloc(1024, M_UDFTEMP, M_WAITOK);
3484 
3485 	/* convert utf8 to unicode-16 */
3486 	*raw_name = 0;
3487 	inchp  = name;
3488 	outchp = raw_name;
3489 	bits = 8;
3490 	for (cnt = name_len, udf_chars = 0; cnt;) {
3491 /*###3490 [cc] warning: passing argument 2 of 'wget_utf8' from incompatible pointer type%%%*/
3492 		*outchp = wget_utf8(&inchp, &cnt);
3493 		if (*outchp > 0xff)
3494 			bits=16;
3495 		outchp++;
3496 		udf_chars++;
3497 	}
3498 	/* null terminate just in case */
3499 	*outchp++ = 0;
3500 
3501 	is_osta_typ0  = (chsp->type == 0);
3502 	is_osta_typ0 &= (strcmp((char *) chsp->inf, osta_id) == 0);
3503 	if (is_osta_typ0) {
3504 		udf_chars = udf_CompressUnicode(udf_chars, bits,
3505 				(unicode_t *) raw_name,
3506 				(byte *) result);
3507 	} else {
3508 		printf("unix to udf name: no CHSP0 ?\n");
3509 		/* XXX assume 8bit char length byte latin-1 */
3510 		*result++ = 8; udf_chars = 1;
3511 		strncpy(result, name + 1, name_len);
3512 		udf_chars += name_len;
3513 	}
3514 	*result_len = udf_chars;
3515 	free(raw_name, M_UDFTEMP);
3516 }
3517 
3518 /* --------------------------------------------------------------------- */
3519 
3520 void
3521 udf_timestamp_to_timespec(struct udf_mount *ump,
3522 			  struct timestamp *timestamp,
3523 			  struct timespec  *timespec)
3524 {
3525 	struct clock_ymdhms ymdhms;
3526 	uint32_t usecs, secs, nsecs;
3527 	uint16_t tz;
3528 
3529 	/* fill in ymdhms structure from timestamp */
3530 	memset(&ymdhms, 0, sizeof(ymdhms));
3531 	ymdhms.dt_year = udf_rw16(timestamp->year);
3532 	ymdhms.dt_mon  = timestamp->month;
3533 	ymdhms.dt_day  = timestamp->day;
3534 	ymdhms.dt_wday = 0; /* ? */
3535 	ymdhms.dt_hour = timestamp->hour;
3536 	ymdhms.dt_min  = timestamp->minute;
3537 	ymdhms.dt_sec  = timestamp->second;
3538 
3539 	secs = clock_ymdhms_to_secs(&ymdhms);
3540 	usecs = timestamp->usec +
3541 		100*timestamp->hund_usec + 10000*timestamp->centisec;
3542 	nsecs = usecs * 1000;
3543 
3544 	/*
3545 	 * Calculate the time zone.  The timezone is 12 bit signed 2's
3546 	 * compliment, so we gotta do some extra magic to handle it right.
3547 	 */
3548 	tz  = udf_rw16(timestamp->type_tz);
3549 	tz &= 0x0fff;			/* only lower 12 bits are significant */
3550 	if (tz & 0x0800)		/* sign extention */
3551 		tz |= 0xf000;
3552 
3553 	/* TODO check timezone conversion */
3554 	/* check if we are specified a timezone to convert */
3555 	if (udf_rw16(timestamp->type_tz) & 0x1000) {
3556 		if ((int16_t) tz != -2047)
3557 			secs -= (int16_t) tz * 60;
3558 	} else {
3559 		secs -= ump->mount_args.gmtoff;
3560 	}
3561 
3562 	timespec->tv_sec  = secs;
3563 	timespec->tv_nsec = nsecs;
3564 }
3565 
3566 
3567 void
3568 udf_timespec_to_timestamp(struct timespec *timespec, struct timestamp *timestamp)
3569 {
3570 	struct clock_ymdhms ymdhms;
3571 	uint32_t husec, usec, csec;
3572 
3573 	(void) clock_secs_to_ymdhms(timespec->tv_sec, &ymdhms);
3574 
3575 	usec   = (timespec->tv_nsec + 500) / 1000;	/* round */
3576 	husec  =   usec / 100;
3577 	usec  -=  husec * 100;				/* only 0-99 in usec  */
3578 	csec   =  husec / 100;				/* only 0-99 in csec  */
3579 	husec -=   csec * 100;				/* only 0-99 in husec */
3580 
3581 	/* set method 1 for CUT/GMT */
3582 	timestamp->type_tz	= udf_rw16((1<<12) + 0);
3583 	timestamp->year		= udf_rw16(ymdhms.dt_year);
3584 	timestamp->month	= ymdhms.dt_mon;
3585 	timestamp->day		= ymdhms.dt_day;
3586 	timestamp->hour		= ymdhms.dt_hour;
3587 	timestamp->minute	= ymdhms.dt_min;
3588 	timestamp->second	= ymdhms.dt_sec;
3589 	timestamp->centisec	= csec;
3590 	timestamp->hund_usec	= husec;
3591 	timestamp->usec		= usec;
3592 }
3593 
3594 /* --------------------------------------------------------------------- */
3595 
3596 /*
3597  * Attribute and filetypes converters with get/set pairs
3598  */
3599 
3600 uint32_t
3601 udf_getaccessmode(struct udf_node *udf_node)
3602 {
3603 	struct file_entry     *fe = udf_node->fe;;
3604 	struct extfile_entry *efe = udf_node->efe;
3605 	uint32_t udf_perm, icbftype;
3606 	uint32_t mode, ftype;
3607 	uint16_t icbflags;
3608 
3609 	UDF_LOCK_NODE(udf_node, 0);
3610 	if (fe) {
3611 		udf_perm = udf_rw32(fe->perm);
3612 		icbftype = fe->icbtag.file_type;
3613 		icbflags = udf_rw16(fe->icbtag.flags);
3614 	} else {
3615 		assert(udf_node->efe);
3616 		udf_perm = udf_rw32(efe->perm);
3617 		icbftype = efe->icbtag.file_type;
3618 		icbflags = udf_rw16(efe->icbtag.flags);
3619 	}
3620 
3621 	mode  = udf_perm_to_unix_mode(udf_perm);
3622 	ftype = udf_icb_to_unix_filetype(icbftype);
3623 
3624 	/* set suid, sgid, sticky from flags in fe/efe */
3625 	if (icbflags & UDF_ICB_TAG_FLAGS_SETUID)
3626 		mode |= S_ISUID;
3627 	if (icbflags & UDF_ICB_TAG_FLAGS_SETGID)
3628 		mode |= S_ISGID;
3629 	if (icbflags & UDF_ICB_TAG_FLAGS_STICKY)
3630 		mode |= S_ISVTX;
3631 
3632 	UDF_UNLOCK_NODE(udf_node, 0);
3633 
3634 	return mode | ftype;
3635 }
3636 
3637 
3638 void
3639 udf_setaccessmode(struct udf_node *udf_node, mode_t mode)
3640 {
3641 	struct file_entry    *fe  = udf_node->fe;
3642 	struct extfile_entry *efe = udf_node->efe;
3643 	uint32_t udf_perm;
3644 	uint16_t icbflags;
3645 
3646 	UDF_LOCK_NODE(udf_node, 0);
3647 	udf_perm = unix_mode_to_udf_perm(mode & ALLPERMS);
3648 	if (fe) {
3649 		icbflags = udf_rw16(fe->icbtag.flags);
3650 	} else {
3651 		icbflags = udf_rw16(efe->icbtag.flags);
3652 	}
3653 
3654 	icbflags &= ~UDF_ICB_TAG_FLAGS_SETUID;
3655 	icbflags &= ~UDF_ICB_TAG_FLAGS_SETGID;
3656 	icbflags &= ~UDF_ICB_TAG_FLAGS_STICKY;
3657 	if (mode & S_ISUID)
3658 		icbflags |= UDF_ICB_TAG_FLAGS_SETUID;
3659 	if (mode & S_ISGID)
3660 		icbflags |= UDF_ICB_TAG_FLAGS_SETGID;
3661 	if (mode & S_ISVTX)
3662 		icbflags |= UDF_ICB_TAG_FLAGS_STICKY;
3663 
3664 	if (fe) {
3665 		fe->perm  = udf_rw32(udf_perm);
3666 		fe->icbtag.flags  = udf_rw16(icbflags);
3667 	} else {
3668 		efe->perm = udf_rw32(udf_perm);
3669 		efe->icbtag.flags = udf_rw16(icbflags);
3670 	}
3671 
3672 	UDF_UNLOCK_NODE(udf_node, 0);
3673 }
3674 
3675 
3676 void
3677 udf_getownership(struct udf_node *udf_node, uid_t *uidp, gid_t *gidp)
3678 {
3679 	struct udf_mount     *ump = udf_node->ump;
3680 	struct file_entry    *fe  = udf_node->fe;
3681 	struct extfile_entry *efe = udf_node->efe;
3682 	uid_t uid;
3683 	gid_t gid;
3684 
3685 	UDF_LOCK_NODE(udf_node, 0);
3686 	if (fe) {
3687 		uid = (uid_t)udf_rw32(fe->uid);
3688 		gid = (gid_t)udf_rw32(fe->gid);
3689 	} else {
3690 		assert(udf_node->efe);
3691 		uid = (uid_t)udf_rw32(efe->uid);
3692 		gid = (gid_t)udf_rw32(efe->gid);
3693 	}
3694 
3695 	/* do the uid/gid translation game */
3696 	if ((uid == (uid_t) -1) && (gid == (gid_t) -1)) {
3697 		uid = ump->mount_args.anon_uid;
3698 		gid = ump->mount_args.anon_gid;
3699 	}
3700 	*uidp = uid;
3701 	*gidp = gid;
3702 
3703 	UDF_UNLOCK_NODE(udf_node, 0);
3704 }
3705 
3706 
3707 void
3708 udf_setownership(struct udf_node *udf_node, uid_t uid, gid_t gid)
3709 {
3710 	struct udf_mount     *ump = udf_node->ump;
3711 	struct file_entry    *fe  = udf_node->fe;
3712 	struct extfile_entry *efe = udf_node->efe;
3713 	uid_t nobody_uid;
3714 	gid_t nobody_gid;
3715 
3716 	UDF_LOCK_NODE(udf_node, 0);
3717 
3718 	/* do the uid/gid translation game */
3719 	nobody_uid = ump->mount_args.nobody_uid;
3720 	nobody_gid = ump->mount_args.nobody_gid;
3721 	if ((uid == nobody_uid) && (gid == nobody_gid)) {
3722 		uid = (uid_t) -1;
3723 		gid = (gid_t) -1;
3724 	}
3725 
3726 	if (fe) {
3727 		fe->uid  = udf_rw32((uint32_t) uid);
3728 		fe->gid  = udf_rw32((uint32_t) gid);
3729 	} else {
3730 		efe->uid = udf_rw32((uint32_t) uid);
3731 		efe->gid = udf_rw32((uint32_t) gid);
3732 	}
3733 
3734 	UDF_UNLOCK_NODE(udf_node, 0);
3735 }
3736 
3737 
3738 /* --------------------------------------------------------------------- */
3739 
3740 /*
3741  * Directory read and manipulation functions.
3742  *
3743  * Note that if the file is found, the cached diroffset possition *before* the
3744  * advance is remembered. Thus if the same filename is lookup again just after
3745  * this lookup its immediately found.
3746  */
3747 
3748 int
3749 udf_lookup_name_in_dir(struct vnode *vp, const char *name, int namelen,
3750 		       struct long_ad *icb_loc)
3751 {
3752 	struct udf_node  *dir_node = VTOI(vp);
3753 	struct file_entry    *fe  = dir_node->fe;
3754 	struct extfile_entry *efe = dir_node->efe;
3755 	struct fileid_desc *fid;
3756 	struct dirent *dirent;
3757 	uint64_t file_size, diroffset, pre_diroffset;
3758 	uint32_t lb_size;
3759 	int found, error;
3760 
3761 	/* get directory filesize */
3762 	if (fe) {
3763 		file_size = udf_rw64(fe->inf_len);
3764 	} else {
3765 		assert(efe);
3766 		file_size = udf_rw64(efe->inf_len);
3767 	}
3768 
3769 	/* allocate temporary space for fid */
3770 	lb_size = udf_rw32(dir_node->ump->logical_vol->lb_size);
3771 	fid = malloc(lb_size, M_UDFTEMP, M_WAITOK);
3772 
3773 	found = 0;
3774 	diroffset = dir_node->last_diroffset;
3775 
3776 	/*
3777 	 * if the directory is trunced or if we have never visited it yet,
3778 	 * start at the end.
3779 	 */
3780 	if ((diroffset >= file_size) || (diroffset == 0)) {
3781 		diroffset = dir_node->last_diroffset = file_size;
3782 	}
3783 
3784 	dirent = malloc(sizeof(struct dirent), M_UDFTEMP, M_WAITOK);
3785 
3786 	while (!found) {
3787 		/* if at the end, go trough zero */
3788 		if (diroffset >= file_size)
3789 			diroffset = 0;
3790 
3791 		pre_diroffset = diroffset;
3792 
3793 		/* transfer a new fid/dirent */
3794 		error = udf_read_fid_stream(vp, &diroffset, fid, dirent);
3795 		if (error)
3796 			break;
3797 
3798 		/* skip deleted entries */
3799 		if ((fid->file_char & UDF_FILE_CHAR_DEL) == 0) {
3800 			if ((strlen(dirent->d_name) == namelen) &&
3801 			    (strncmp(dirent->d_name, name, namelen) == 0)) {
3802 				found = 1;
3803 				*icb_loc = fid->icb;
3804 				/* remember where we were before the advance */
3805 				diroffset = pre_diroffset;
3806 			}
3807 		}
3808 
3809 		if (diroffset == dir_node->last_diroffset) {
3810 			/* we have cycled */
3811 			break;
3812 		}
3813 	}
3814 	free(fid, M_UDFTEMP);
3815 	free(dirent, M_UDFTEMP);
3816 	dir_node->last_diroffset = diroffset;
3817 
3818 	return found;
3819 }
3820 
3821 /* --------------------------------------------------------------------- */
3822 
3823 static int
3824 udf_create_new_fe(struct udf_mount *ump, struct file_entry *fe, int file_type,
3825 	struct long_ad *node_icb, struct long_ad *parent_icb,
3826 	uint64_t parent_unique_id)
3827 {
3828 	struct timespec now;
3829 	struct icb_tag *icb;
3830 	uint64_t unique_id;
3831 	uint32_t fidsize, lb_num;
3832 	int crclen;
3833 
3834 	lb_num = udf_rw32(node_icb->loc.lb_num);
3835 	udf_inittag(ump, &fe->tag, TAGID_FENTRY, lb_num);
3836 	icb = &fe->icbtag;
3837 
3838 	/*
3839 	 * Always use strategy type 4 unless on WORM wich we don't support
3840 	 * (yet). Fill in defaults and set for internal allocation of data.
3841 	 */
3842 	icb->strat_type      = udf_rw16(4);
3843 	icb->max_num_entries = udf_rw16(1);
3844 	icb->file_type       = file_type;	/* 8 bit */
3845 	icb->flags           = udf_rw16(UDF_ICB_INTERN_ALLOC);
3846 
3847 	fe->perm     = udf_rw32(0x7fff);	/* all is allowed   */
3848 	fe->link_cnt = udf_rw16(0);		/* explicit setting */
3849 
3850 	fe->ckpoint  = udf_rw32(1);		/* user supplied file version */
3851 
3852 	vfs_timestamp(&now);
3853 	udf_timespec_to_timestamp(&now, &fe->atime);
3854 	udf_timespec_to_timestamp(&now, &fe->attrtime);
3855 	udf_timespec_to_timestamp(&now, &fe->mtime);
3856 
3857 	udf_set_regid(&fe->imp_id, IMPL_NAME);
3858 	udf_add_impl_regid(ump, &fe->imp_id);
3859 
3860 	unique_id = udf_advance_uniqueid(ump);
3861 	fe->unique_id = udf_rw64(unique_id);
3862 
3863 	fidsize = 0;
3864 	if (file_type == UDF_ICB_FILETYPE_DIRECTORY) {
3865 		fidsize = udf_create_parentfid(ump,
3866 			(struct fileid_desc *) fe->data, parent_icb,
3867 			parent_unique_id);
3868 	}
3869 
3870 	/* record fidlength information */
3871 	fe->inf_len = udf_rw64(fidsize);
3872 	fe->l_ea    = udf_rw32(0);
3873 	fe->l_ad    = udf_rw32(fidsize);
3874 	fe->logblks_rec = udf_rw64(0);		/* intern */
3875 
3876 	crclen  = sizeof(struct file_entry) - 1 - UDF_DESC_TAG_LENGTH;
3877 	crclen += fidsize;
3878 	fe->tag.desc_crc_len = udf_rw16(crclen);
3879 
3880 	(void) udf_validate_tag_and_crc_sums((union dscrptr *) fe);
3881 
3882 	return fidsize;
3883 }
3884 
3885 /* --------------------------------------------------------------------- */
3886 
3887 static int
3888 udf_create_new_efe(struct udf_mount *ump, struct extfile_entry *efe,
3889 	int file_type, struct long_ad *node_icb, struct long_ad *parent_icb,
3890 	uint64_t parent_unique_id)
3891 {
3892 	struct timespec now;
3893 	struct icb_tag *icb;
3894 	uint64_t unique_id;
3895 	uint32_t fidsize, lb_num;
3896 	int crclen;
3897 
3898 	lb_num = udf_rw32(node_icb->loc.lb_num);
3899 	udf_inittag(ump, &efe->tag, TAGID_EXTFENTRY, lb_num);
3900 	icb = &efe->icbtag;
3901 
3902 	/*
3903 	 * Always use strategy type 4 unless on WORM wich we don't support
3904 	 * (yet). Fill in defaults and set for internal allocation of data.
3905 	 */
3906 	icb->strat_type      = udf_rw16(4);
3907 	icb->max_num_entries = udf_rw16(1);
3908 	icb->file_type       = file_type;	/* 8 bit */
3909 	icb->flags           = udf_rw16(UDF_ICB_INTERN_ALLOC);
3910 
3911 	efe->perm     = udf_rw32(0x7fff);	/* all is allowed   */
3912 	efe->link_cnt = udf_rw16(0);		/* explicit setting */
3913 
3914 	efe->ckpoint  = udf_rw32(1);		/* user supplied file version */
3915 
3916 	vfs_timestamp(&now);
3917 	udf_timespec_to_timestamp(&now, &efe->ctime);
3918 	udf_timespec_to_timestamp(&now, &efe->atime);
3919 	udf_timespec_to_timestamp(&now, &efe->attrtime);
3920 	udf_timespec_to_timestamp(&now, &efe->mtime);
3921 
3922 	udf_set_regid(&efe->imp_id, IMPL_NAME);
3923 	udf_add_impl_regid(ump, &efe->imp_id);
3924 
3925 	unique_id = udf_advance_uniqueid(ump);
3926 	efe->unique_id = udf_rw64(unique_id);
3927 
3928 	fidsize = 0;
3929 	if (file_type == UDF_ICB_FILETYPE_DIRECTORY) {
3930 		fidsize = udf_create_parentfid(ump,
3931 			(struct fileid_desc *) efe->data, parent_icb,
3932 			parent_unique_id);
3933 	}
3934 
3935 	/* record fidlength information */
3936 	efe->obj_size = udf_rw64(fidsize);
3937 	efe->inf_len  = udf_rw64(fidsize);
3938 	efe->l_ea     = udf_rw32(0);
3939 	efe->l_ad     = udf_rw32(fidsize);
3940 	efe->logblks_rec = udf_rw64(0);		/* intern */
3941 
3942 	crclen  = sizeof(struct extfile_entry) - 1 - UDF_DESC_TAG_LENGTH;
3943 	crclen += fidsize;
3944 	efe->tag.desc_crc_len = udf_rw16(crclen);
3945 
3946 	(void) udf_validate_tag_and_crc_sums((union dscrptr *) efe);
3947 
3948 	return fidsize;
3949 }
3950 
3951 /* --------------------------------------------------------------------- */
3952 
3953 int
3954 udf_dir_detach(struct udf_mount *ump, struct udf_node *dir_node,
3955 	struct udf_node *udf_node, struct componentname *cnp)
3956 {
3957 	struct file_entry    *fe  = dir_node->fe;
3958 	struct extfile_entry *efe = dir_node->efe;
3959 	struct fileid_desc *fid;
3960 	struct dirent *dirent;
3961 	uint64_t file_size, diroffset;
3962 	uint32_t lb_size, fidsize;
3963 	int found, error;
3964 	char const *name  = cnp->cn_nameptr;
3965 	int namelen = cnp->cn_namelen;
3966 	int refcnt;
3967 
3968 	/* get directory filesize */
3969 	if (fe) {
3970 		file_size = udf_rw64(fe->inf_len);
3971 	} else {
3972 		assert(efe);
3973 		file_size = udf_rw64(efe->inf_len);
3974 	}
3975 
3976 	/* allocate temporary space for fid */
3977 	lb_size = udf_rw32(dir_node->ump->logical_vol->lb_size);
3978 	fid = malloc(lb_size, M_UDFTEMP, M_WAITOK);
3979 
3980 	found = 0;
3981 	diroffset = dir_node->last_diroffset;
3982 
3983 	/*
3984 	 * if the directory is trunced or if we have never visited it yet,
3985 	 * start at the end.
3986 	 */
3987 	if ((diroffset >= file_size) || (diroffset == 0)) {
3988 		diroffset = dir_node->last_diroffset = file_size;
3989 	}
3990 
3991 	dirent = malloc(sizeof(struct dirent), M_UDFTEMP, M_WAITOK);
3992 
3993 	while (!found) {
3994 		/* if at the end, go trough zero */
3995 		if (diroffset >= file_size)
3996 			diroffset = 0;
3997 
3998 		/* transfer a new fid/dirent */
3999 		error = udf_read_fid_stream(dir_node->vnode, &diroffset,
4000 				fid, dirent);
4001 		if (error)
4002 			break;
4003 
4004 		/* skip deleted entries */
4005 		if ((fid->file_char & UDF_FILE_CHAR_DEL) == 0) {
4006 			if ((strlen(dirent->d_name) == namelen) &&
4007 			    (strncmp(dirent->d_name, name, namelen) == 0)) {
4008 				found = 1;
4009 			}
4010 		}
4011 
4012 		if (diroffset == dir_node->last_diroffset) {
4013 			/* we have cycled */
4014 			break;
4015 		}
4016 	}
4017 	if (!found) {
4018 		free(fid, M_UDFTEMP);
4019 		free(dirent, M_UDFTEMP);
4020 		dir_node->last_diroffset = diroffset;
4021 		return ENOENT;
4022 	}
4023 
4024 	/* mark deleted */
4025 	fid->file_char |= UDF_FILE_CHAR_DEL;
4026 #ifdef UDF_COMPLETE_DELETE
4027 	memset(&fid->icb, 0, sizeof(fid->icb));
4028 #endif
4029 	(void) udf_validate_tag_and_crc_sums((union dscrptr *) fid);
4030 
4031 	/* roll back last advance from udf_read_fid_stream */
4032 	fidsize = udf_fidsize(fid);
4033 	diroffset -= fidsize;
4034 
4035 	/* write out */
4036 	error = vn_rdwr(UIO_WRITE, dir_node->vnode,
4037 			fid, fidsize, diroffset,
4038 			UIO_SYSSPACE, IO_ALTSEMANTICS | IO_NODELOCKED,
4039 			FSCRED, NULL, NULL);
4040 	if (error == 0) {
4041 		/* get reference count of attached node */
4042 		if (udf_node->fe) {
4043 			refcnt = udf_rw16(udf_node->fe->link_cnt);
4044 		} else {
4045 			KASSERT(udf_node->efe);
4046 			refcnt = udf_rw16(udf_node->efe->link_cnt);
4047 		}
4048 #ifdef UDF_COMPLETE_DELETE
4049 		/* substract reference counter in attached node */
4050 		refcnt -= 1;
4051 		if (udf_node->fe) {
4052 			udf_node->fe->link_cnt = udf_rw16(refcnt);
4053 		} else {
4054 			udf_node->efe->link_cnt = udf_rw16(refcnt);
4055 		}
4056 
4057 		/* prevent writeout when refcnt == 0 */
4058 		if (refcnt == 0)
4059 			udf_node->i_flags |= IN_DELETED;
4060 
4061 		if (fid->file_char & UDF_FILE_CHAR_DIR) {
4062 			int drefcnt;
4063 
4064 			/* substract reference counter in directory node */
4065 			/* note subtract 2 (?) for its was also backreferenced */
4066 			if (dir_node->fe) {
4067 				drefcnt  = udf_rw16(dir_node->fe->link_cnt);
4068 				drefcnt -= 1;
4069 				dir_node->fe->link_cnt = udf_rw16(drefcnt);
4070 			} else {
4071 				KASSERT(dir_node->efe);
4072 				drefcnt  = udf_rw16(dir_node->efe->link_cnt);
4073 				drefcnt -= 1;
4074 				dir_node->efe->link_cnt = udf_rw16(drefcnt);
4075 			}
4076 		}
4077 
4078 		udf_node->i_flags |= IN_MODIFIED;
4079 		dir_node->i_flags |= IN_MODIFIED;
4080 #endif
4081 		/* if it is/was a hardlink adjust the file count */
4082 		if (refcnt > 0)
4083 			udf_adjust_filecount(udf_node, -1);
4084 
4085 		/* XXX we could restart at the deleted entry */
4086 		diroffset = 0;
4087 	}
4088 
4089 	free(fid, M_UDFTEMP);
4090 	free(dirent, M_UDFTEMP);
4091 	dir_node->last_diroffset = diroffset;
4092 
4093 	return error;
4094 }
4095 
4096 /* --------------------------------------------------------------------- */
4097 
4098 /*
4099  * We are not allowed to split the fid tag itself over an logical block so
4100  * check the space remaining in the logical block.
4101  *
4102  * We try to select the smallest candidate for recycling or when none is
4103  * found, append a new one at the end of the directory.
4104  */
4105 
4106 int
4107 udf_dir_attach(struct udf_mount *ump, struct udf_node *dir_node,
4108 	struct udf_node *udf_node, struct vattr *vap, struct componentname *cnp)
4109 {
4110 	struct vnode *dvp = dir_node->vnode;
4111 	struct fileid_desc   *fid;
4112 	struct icb_tag       *icbtag;
4113 	struct charspec osta_charspec;
4114 	struct dirent   dirent;
4115 	uint64_t unique_id, dir_size, diroffset;
4116 	uint64_t fid_pos, end_fid_pos, chosen_fid_pos;
4117 	uint32_t chosen_size, chosen_size_diff;
4118 	int lb_size, lb_rest, fidsize, this_fidsize, size_diff;
4119 	int file_char, refcnt, icbflags, addr_type, error;
4120 
4121 	lb_size = udf_rw32(ump->logical_vol->lb_size);
4122 	udf_osta_charset(&osta_charspec);
4123 
4124 	if (dir_node->fe) {
4125 		dir_size = udf_rw64(dir_node->fe->inf_len);
4126 		icbtag  = &dir_node->fe->icbtag;
4127 	} else {
4128 		dir_size = udf_rw64(dir_node->efe->inf_len);
4129 		icbtag  = &dir_node->efe->icbtag;
4130 	}
4131 
4132 	icbflags   = udf_rw16(icbtag->flags);
4133 	addr_type  = icbflags & UDF_ICB_TAG_FLAGS_ALLOC_MASK;
4134 
4135 	if (udf_node->fe) {
4136 		unique_id = udf_rw64(udf_node->fe->unique_id);
4137 		refcnt   = udf_rw16(udf_node->fe->link_cnt);
4138 	} else {
4139 		unique_id = udf_rw64(udf_node->efe->unique_id);
4140 		refcnt   = udf_rw16(udf_node->efe->link_cnt);
4141 	}
4142 
4143 	if (refcnt > 0) {
4144 		unique_id = udf_advance_uniqueid(ump);
4145 		udf_adjust_filecount(udf_node, 1);
4146 	}
4147 
4148 
4149 	/* determine file characteristics */
4150 	file_char = 0;	/* visible non deleted file and not stream metadata */
4151 	if (vap->va_type == VDIR)
4152 		file_char = UDF_FILE_CHAR_DIR;
4153 
4154 	/* malloc scrap buffer */
4155 	fid = malloc(lb_size, M_TEMP, M_WAITOK);
4156 	bzero(fid, lb_size);
4157 
4158 	/* calculate _minimum_ fid size */
4159 	unix_to_udf_name((char *) fid->data, &fid->l_fi,
4160 		cnp->cn_nameptr, cnp->cn_namelen, &osta_charspec);
4161 	fidsize = UDF_FID_SIZE + fid->l_fi;
4162 	fidsize = (fidsize + 3) & ~3;		/* multiple of 4 */
4163 
4164 	/* find position that will fit the FID */
4165 	diroffset = dir_node->last_diroffset;
4166 
4167 	/*
4168 	 * if the directory is trunced or if we have never visited it yet,
4169 	 * start at the end.
4170 	 */
4171 	if ((diroffset >= dir_size) || (diroffset == 0)) {
4172 		diroffset = dir_node->last_diroffset = dir_size;
4173 	}
4174 
4175 	chosen_fid_pos   = diroffset;
4176 	chosen_size      = 0;
4177 	chosen_size_diff = UINT_MAX;
4178 
4179 	for (;;) {
4180 		/* if at the end, go trough zero */
4181 		if (diroffset >= dir_size)
4182 			diroffset = 0;
4183 
4184 		/* get fid/dirent */
4185 		fid_pos = diroffset;
4186 		error = udf_read_fid_stream(dvp, &diroffset, fid, &dirent);
4187 		if (error)
4188 			break;
4189 
4190 		this_fidsize = udf_fidsize(fid);
4191 
4192 		/* reuse deleted entries */
4193 		if ((fid->file_char & UDF_FILE_CHAR_DEL)) {
4194 			size_diff = this_fidsize - fidsize;
4195 			end_fid_pos = fid_pos + this_fidsize;
4196 			lb_rest = lb_size - (end_fid_pos % lb_size);
4197 
4198 #ifndef UDF_COMPLETE_DELETE
4199 			/* only reuse entries that are wiped */
4200 			/* check if the len + loc are marked zero */
4201 			if (udf_rw32(fid->icb.len != 0))
4202 				break;
4203 			if (udf_rw32(fid->icb.loc.lb_num) != 0)
4204 				break;
4205 			if (udf_rw16(fid->icb.loc.part_num != 0))
4206 				break;
4207 #endif
4208 			/* select if not splitting the tag and its smaller */
4209 			if ((size_diff >= 0)  &&
4210 				(size_diff < chosen_size_diff) &&
4211 				(lb_rest >= sizeof(struct desc_tag)))
4212 			{
4213 				/* UDF 2.3.4.2+3 specifies rules for iu size */
4214 				if ((size_diff == 0) || (size_diff >= 32)) {
4215 					chosen_fid_pos   = fid_pos;
4216 					chosen_size      = this_fidsize;
4217 					chosen_size_diff = size_diff;
4218 				}
4219 			}
4220 		}
4221 
4222 		if (diroffset == dir_node->last_diroffset) {
4223 			/* we have cycled */
4224 			break;
4225 		}
4226 	}
4227 	/* unlikely */
4228 	if (error) {
4229 		free(fid, M_TEMP);
4230 		return error;
4231 	}
4232 
4233 	/* extend directory if no other candidate found */
4234 	if (chosen_size == 0) {
4235 		chosen_fid_pos   = dir_size;
4236 		chosen_size      = fidsize;
4237 		chosen_size_diff = 0;
4238 
4239 		/* special case UDF 2.00+ 2.3.4.4, no splitting up fid tag */
4240 		if (addr_type == UDF_ICB_INTERN_ALLOC) {
4241 			/* pre-grow directory to see if we're to switch */
4242 			udf_grow_node(dir_node, dir_size + chosen_size);
4243 
4244 			icbflags   = udf_rw16(icbtag->flags);
4245 			addr_type  = icbflags & UDF_ICB_TAG_FLAGS_ALLOC_MASK;
4246 		}
4247 
4248 		/* make sure the next fid desc_tag won't be splitted */
4249 		if (addr_type != UDF_ICB_INTERN_ALLOC) {
4250 			end_fid_pos = chosen_fid_pos + chosen_size;
4251 			lb_rest = lb_size - (end_fid_pos % lb_size);
4252 
4253 			/* pad with implementation use regid if needed */
4254 			if (lb_rest < sizeof(struct desc_tag))
4255 				chosen_size += 32;
4256 		}
4257 	}
4258 	chosen_size_diff = chosen_size - fidsize;
4259 	diroffset = chosen_fid_pos + chosen_size;
4260 
4261 	/* populate the FID */
4262 	memset(fid, 0, lb_size);
4263 	udf_inittag(ump, &fid->tag, TAGID_FID, 0);
4264 	fid->file_version_num    = udf_rw16(1);	/* UDF 2.3.4.1 */
4265 	fid->file_char           = file_char;
4266 	fid->icb                 = udf_node->loc;
4267 	fid->icb.longad_uniqueid = udf_rw32((uint32_t) unique_id);
4268 	fid->l_iu                = udf_rw16(0);
4269 
4270 	if (chosen_size > fidsize) {
4271 		/* insert implementation-use regid to space it correctly */
4272 		fid->l_iu = udf_rw16(chosen_size_diff);
4273 
4274 		/* set implementation use */
4275 		udf_set_regid((struct regid *) fid->data, IMPL_NAME);
4276 		udf_add_impl_regid(ump, (struct regid *) fid->data);
4277 	}
4278 
4279 	/* fill in name */
4280 	unix_to_udf_name((char *) fid->data + udf_rw16(fid->l_iu),
4281 		&fid->l_fi, cnp->cn_nameptr, cnp->cn_namelen, &osta_charspec);
4282 
4283 	fid->tag.desc_crc_len = chosen_size - UDF_DESC_TAG_LENGTH;
4284 	(void) udf_validate_tag_and_crc_sums((union dscrptr *) fid);
4285 
4286 	/* writeout FID/update parent directory */
4287 	error = vn_rdwr(UIO_WRITE, dvp,
4288 			fid, chosen_size, chosen_fid_pos,
4289 			UIO_SYSSPACE, IO_ALTSEMANTICS | IO_NODELOCKED,
4290 			FSCRED, NULL, NULL);
4291 
4292 	if (error) {
4293 		free(fid, M_TEMP);
4294 		return error;
4295 	}
4296 
4297 	/* add reference counter in attached node */
4298 	if (udf_node->fe) {
4299 		refcnt = udf_rw16(udf_node->fe->link_cnt);
4300 		udf_node->fe->link_cnt = udf_rw16(refcnt+1);
4301 	} else {
4302 		KASSERT(udf_node->efe);
4303 		refcnt = udf_rw16(udf_node->efe->link_cnt);
4304 		udf_node->efe->link_cnt = udf_rw16(refcnt+1);
4305 	}
4306 
4307 	/* mark not deleted if it was... just in case, but do warn */
4308 	if (udf_node->i_flags & IN_DELETED) {
4309 		printf("udf: warning, marking a file undeleted\n");
4310 		udf_node->i_flags &= ~IN_DELETED;
4311 	}
4312 
4313 	if (file_char & UDF_FILE_CHAR_DIR) {
4314 		/* add reference counter in directory node for '..' */
4315 		if (dir_node->fe) {
4316 			refcnt = udf_rw16(dir_node->fe->link_cnt);
4317 			refcnt++;
4318 			dir_node->fe->link_cnt = udf_rw16(refcnt);
4319 		} else {
4320 			KASSERT(dir_node->efe);
4321 			refcnt = udf_rw16(dir_node->efe->link_cnt);
4322 			refcnt++;
4323 			dir_node->efe->link_cnt = udf_rw16(refcnt);
4324 		}
4325 	}
4326 
4327 	/* update our last position so we dont have to cycle again and again */
4328 	dir_node->last_diroffset = diroffset;
4329 
4330 	udf_node->i_flags |= IN_CHANGE | IN_MODIFY; /* | IN_CREATE? */
4331 	/* VN_KNOTE(udf_node,  ...) */
4332 	udf_update(udf_node->vnode, NULL, NULL, 0);
4333 
4334 	free(fid, M_TEMP);
4335 
4336 	return 0;
4337 }
4338 
4339 /* --------------------------------------------------------------------- */
4340 
4341 /*
4342  * Each node can have an attached streamdir node though not recursively. These
4343  * are otherwise known as named substreams/named extended attributes that have
4344  * no size limitations.
4345  *
4346  * `Normal' extended attributes are indicated with a number and are recorded
4347  * in either the fe/efe descriptor itself for small descriptors or recorded in
4348  * the attached extended attribute file. Since these spaces can get
4349  * fragmented, care ought to be taken.
4350  *
4351  * Since the size of the space reserved for allocation descriptors is limited,
4352  * there is a mechanim provided for extending this space; this is done by a
4353  * special extent to allow schrinking of the allocations without breaking the
4354  * linkage to the allocation extent descriptor.
4355  */
4356 
4357 int
4358 udf_get_node(struct udf_mount *ump, struct long_ad *node_icb_loc,
4359 	     struct udf_node **udf_noderes)
4360 {
4361 	union dscrptr   *dscr;
4362 	struct udf_node *udf_node;
4363 	struct vnode    *nvp;
4364 	struct long_ad   icb_loc, last_fe_icb_loc;
4365 	uint64_t file_size;
4366 	uint32_t lb_size, sector, dummy;
4367 	uint8_t  *file_data;
4368 	int udf_file_type, dscr_type, strat, strat4096, needs_indirect;
4369 	int slot, eof, error;
4370 
4371 	DPRINTF(NODE, ("udf_get_node called\n"));
4372 	*udf_noderes = udf_node = NULL;
4373 
4374 	/* lock to disallow simultanious creation of same udf_node */
4375 	mutex_enter(&ump->get_node_lock);
4376 
4377 	DPRINTF(NODE, ("\tlookup in hash table\n"));
4378 	/* lookup in hash table */
4379 	assert(ump);
4380 	assert(node_icb_loc);
4381 	udf_node = udf_hash_lookup(ump, node_icb_loc);
4382 	if (udf_node) {
4383 		DPRINTF(NODE, ("\tgot it from the hash!\n"));
4384 		/* vnode is returned locked */
4385 		*udf_noderes = udf_node;
4386 		mutex_exit(&ump->get_node_lock);
4387 		return 0;
4388 	}
4389 
4390 	/* garbage check: translate udf_node_icb_loc to sectornr */
4391 	error = udf_translate_vtop(ump, node_icb_loc, &sector, &dummy);
4392 	if (error) {
4393 		/* no use, this will fail anyway */
4394 		mutex_exit(&ump->get_node_lock);
4395 		return EINVAL;
4396 	}
4397 
4398 	/* build udf_node (do initialise!) */
4399 	udf_node = pool_get(&udf_node_pool, PR_WAITOK);
4400 	memset(udf_node, 0, sizeof(struct udf_node));
4401 
4402 	DPRINTF(NODE, ("\tget new vnode\n"));
4403 	/* give it a vnode */
4404 	error = getnewvnode(VT_UDF, ump->vfs_mountp, udf_vnodeop_p, &nvp);
4405         if (error) {
4406 		pool_put(&udf_node_pool, udf_node);
4407 		mutex_exit(&ump->get_node_lock);
4408 		return error;
4409 	}
4410 
4411 	/* always return locked vnode */
4412 	if ((error = vn_lock(nvp, LK_EXCLUSIVE | LK_RETRY))) {
4413 		/* recycle vnode and unlock; simultanious will fail too */
4414 		ungetnewvnode(nvp);
4415 		mutex_exit(&ump->get_node_lock);
4416 		return error;
4417 	}
4418 
4419 	/* initialise crosslinks, note location of fe/efe for hashing */
4420 	udf_node->ump    =  ump;
4421 	udf_node->vnode  =  nvp;
4422 	nvp->v_data      =  udf_node;
4423 	udf_node->loc    = *node_icb_loc;
4424 	udf_node->lockf  =  0;
4425 	mutex_init(&udf_node->node_mutex, MUTEX_DEFAULT, IPL_NONE);
4426 	cv_init(&udf_node->node_lock, "udf_nlk");
4427 	genfs_node_init(nvp, &udf_genfsops);	/* inititise genfs */
4428 	udf_node->outstanding_bufs = 0;
4429 	udf_node->outstanding_nodedscr = 0;
4430 
4431 	/* insert into the hash lookup */
4432 	udf_register_node(udf_node);
4433 
4434 	/* safe to unlock, the entry is in the hash table, vnode is locked */
4435 	mutex_exit(&ump->get_node_lock);
4436 
4437 	icb_loc = *node_icb_loc;
4438 	needs_indirect = 0;
4439 	strat4096 = 0;
4440 	udf_file_type = UDF_ICB_FILETYPE_UNKNOWN;
4441 	file_size = 0;
4442 	file_data = NULL;
4443 	lb_size = udf_rw32(ump->logical_vol->lb_size);
4444 
4445 	DPRINTF(NODE, ("\tstart reading descriptors\n"));
4446 	do {
4447 		/* try to read in fe/efe */
4448 		error = udf_read_logvol_dscr(ump, &icb_loc, &dscr);
4449 
4450 		/* blank sector marks end of sequence, check this */
4451 		if ((dscr == NULL) &&  (!strat4096))
4452 			error = ENOENT;
4453 
4454 		/* break if read error or blank sector */
4455 		if (error || (dscr == NULL))
4456 			break;
4457 
4458 		/* process descriptor based on the descriptor type */
4459 		dscr_type = udf_rw16(dscr->tag.id);
4460 		DPRINTF(NODE, ("\tread descriptor %d\n", dscr_type));
4461 
4462 		/* if dealing with an indirect entry, follow the link */
4463 		if (dscr_type == TAGID_INDIRECTENTRY) {
4464 			needs_indirect = 0;
4465 			udf_free_logvol_dscr(ump, &icb_loc, dscr);
4466 			icb_loc = dscr->inde.indirect_icb;
4467 			continue;
4468 		}
4469 
4470 		/* only file entries and extended file entries allowed here */
4471 		if ((dscr_type != TAGID_FENTRY) &&
4472 		    (dscr_type != TAGID_EXTFENTRY)) {
4473 			udf_free_logvol_dscr(ump, &icb_loc, dscr);
4474 			error = ENOENT;
4475 			break;
4476 		}
4477 
4478 		KASSERT(udf_tagsize(dscr, lb_size) == lb_size);
4479 
4480 		/* choose this one */
4481 		last_fe_icb_loc = icb_loc;
4482 
4483 		/* record and process/update (ext)fentry */
4484 		file_data = NULL;
4485 		if (dscr_type == TAGID_FENTRY) {
4486 			if (udf_node->fe)
4487 				udf_free_logvol_dscr(ump, &last_fe_icb_loc,
4488 					udf_node->fe);
4489 			udf_node->fe  = &dscr->fe;
4490 			strat = udf_rw16(udf_node->fe->icbtag.strat_type);
4491 			udf_file_type = udf_node->fe->icbtag.file_type;
4492 			file_size = udf_rw64(udf_node->fe->inf_len);
4493 			file_data = udf_node->fe->data;
4494 		} else {
4495 			if (udf_node->efe)
4496 				udf_free_logvol_dscr(ump, &last_fe_icb_loc,
4497 					udf_node->efe);
4498 			udf_node->efe = &dscr->efe;
4499 			strat = udf_rw16(udf_node->efe->icbtag.strat_type);
4500 			udf_file_type = udf_node->efe->icbtag.file_type;
4501 			file_size = udf_rw64(udf_node->efe->inf_len);
4502 			file_data = udf_node->efe->data;
4503 		}
4504 
4505 		/* check recording strategy (structure) */
4506 
4507 		/*
4508 		 * Strategy 4096 is a daisy linked chain terminating with an
4509 		 * unrecorded sector or a TERM descriptor. The next
4510 		 * descriptor is to be found in the sector that follows the
4511 		 * current sector.
4512 		 */
4513 		if (strat == 4096) {
4514 			strat4096 = 1;
4515 			needs_indirect = 1;
4516 
4517 			icb_loc.loc.lb_num = udf_rw32(icb_loc.loc.lb_num) + 1;
4518 		}
4519 
4520 		/*
4521 		 * Strategy 4 is the normal strategy and terminates, but if
4522 		 * we're in strategy 4096, we can't have strategy 4 mixed in
4523 		 */
4524 
4525 		if (strat == 4) {
4526 			if (strat4096) {
4527 				error = EINVAL;
4528 				break;
4529 			}
4530 			break;		/* done */
4531 		}
4532 	} while (!error);
4533 
4534 	/* first round of cleanup code */
4535 	if (error) {
4536 		DPRINTF(NODE, ("\tnode fe/efe failed!\n"));
4537 		/* recycle udf_node */
4538 		udf_dispose_node(udf_node);
4539 
4540 		vlockmgr(nvp->v_vnlock, LK_RELEASE);
4541 		nvp->v_data = NULL;
4542 		ungetnewvnode(nvp);
4543 
4544 		return EINVAL;		/* error code ok? */
4545 	}
4546 	DPRINTF(NODE, ("\tnode fe/efe read in fine\n"));
4547 
4548 	/* assert no references to dscr anymore beyong this point */
4549 	assert((udf_node->fe) || (udf_node->efe));
4550 	dscr = NULL;
4551 
4552 	/*
4553 	 * Remember where to record an updated version of the descriptor. If
4554 	 * there is a sequence of indirect entries, icb_loc will have been
4555 	 * updated. Its the write disipline to allocate new space and to make
4556 	 * sure the chain is maintained.
4557 	 *
4558 	 * `needs_indirect' flags if the next location is to be filled with
4559 	 * with an indirect entry.
4560 	 */
4561 	udf_node->write_loc = icb_loc;
4562 	udf_node->needs_indirect = needs_indirect;
4563 
4564 	/*
4565 	 * Go trough all allocations extents of this descriptor and when
4566 	 * encountering a redirect read in the allocation extension. These are
4567 	 * daisy-chained.
4568 	 */
4569 	UDF_LOCK_NODE(udf_node, 0);
4570 	udf_node->num_extensions = 0;
4571 
4572 	error   = 0;
4573 	slot    = 0;
4574 	for (;;) {
4575 		udf_get_adslot(udf_node, slot, &icb_loc, &eof);
4576 		if (eof)
4577 			break;
4578 
4579 		if (UDF_EXT_FLAGS(udf_rw32(icb_loc.len)) != UDF_EXT_REDIRECT) {
4580 			slot++;
4581 			continue;
4582 		}
4583 
4584 		DPRINTF(NODE, ("\tgot redirect extent\n"));
4585 		if (udf_node->num_extensions >= UDF_MAX_ALLOC_EXTENTS) {
4586 			DPRINTF(ALLOC, ("udf_get_node: implementation limit, "
4587 					"too many allocation extensions on "
4588 					"udf_node\n"));
4589 			error = EINVAL;
4590 			break;
4591 		}
4592 
4593 		/* length can only be *one* lb : UDF 2.50/2.3.7.1 */
4594 		if (udf_rw32(icb_loc.len) != lb_size) {
4595 			DPRINTF(ALLOC, ("udf_get_node: bad allocation "
4596 					"extension size in udf_node\n"));
4597 			error = EINVAL;
4598 			break;
4599 		}
4600 
4601 		/* load in allocation extent */
4602 		error = udf_read_logvol_dscr(ump, &icb_loc, &dscr);
4603 		if (error || (dscr == NULL))
4604 			break;
4605 
4606 		/* process read-in descriptor */
4607 		dscr_type = udf_rw16(dscr->tag.id);
4608 
4609 		if (dscr_type != TAGID_ALLOCEXTENT) {
4610 			udf_free_logvol_dscr(ump, &icb_loc, dscr);
4611 			error = ENOENT;
4612 			break;
4613 		}
4614 
4615 		DPRINTF(NODE, ("\trecording redirect extent\n"));
4616 		udf_node->ext[udf_node->num_extensions] = &dscr->aee;
4617 		udf_node->ext_loc[udf_node->num_extensions] = icb_loc;
4618 
4619 		udf_node->num_extensions++;
4620 
4621 	} /* while */
4622 	UDF_UNLOCK_NODE(udf_node, 0);
4623 
4624 	/* second round of cleanup code */
4625 	if (error) {
4626 		/* recycle udf_node */
4627 		udf_dispose_node(udf_node);
4628 
4629 		vlockmgr(nvp->v_vnlock, LK_RELEASE);
4630 		nvp->v_data = NULL;
4631 		ungetnewvnode(nvp);
4632 
4633 		return EINVAL;		/* error code ok? */
4634 	}
4635 
4636 	DPRINTF(NODE, ("\tnode read in fine\n"));
4637 
4638 	/*
4639 	 * Translate UDF filetypes into vnode types.
4640 	 *
4641 	 * Systemfiles like the meta main and mirror files are not treated as
4642 	 * normal files, so we type them as having no type. UDF dictates that
4643 	 * they are not allowed to be visible.
4644 	 */
4645 
4646 	switch (udf_file_type) {
4647 	case UDF_ICB_FILETYPE_DIRECTORY :
4648 	case UDF_ICB_FILETYPE_STREAMDIR :
4649 		nvp->v_type = VDIR;
4650 		break;
4651 	case UDF_ICB_FILETYPE_BLOCKDEVICE :
4652 		nvp->v_type = VBLK;
4653 		break;
4654 	case UDF_ICB_FILETYPE_CHARDEVICE :
4655 		nvp->v_type = VCHR;
4656 		break;
4657 	case UDF_ICB_FILETYPE_SOCKET :
4658 		nvp->v_type = VSOCK;
4659 		break;
4660 	case UDF_ICB_FILETYPE_FIFO :
4661 		nvp->v_type = VFIFO;
4662 		break;
4663 	case UDF_ICB_FILETYPE_SYMLINK :
4664 		nvp->v_type = VLNK;
4665 		break;
4666 	case UDF_ICB_FILETYPE_VAT :
4667 	case UDF_ICB_FILETYPE_META_MAIN :
4668 	case UDF_ICB_FILETYPE_META_MIRROR :
4669 		nvp->v_type = VNON;
4670 		break;
4671 	case UDF_ICB_FILETYPE_RANDOMACCESS :
4672 	case UDF_ICB_FILETYPE_REALTIME :
4673 		nvp->v_type = VREG;
4674 		break;
4675 	default:
4676 		/* YIKES, something else */
4677 		nvp->v_type = VNON;
4678 	}
4679 
4680 	/* TODO specfs, fifofs etc etc. vnops setting */
4681 
4682 	/* don't forget to set vnode's v_size */
4683 	uvm_vnp_setsize(nvp, file_size);
4684 
4685 	/* TODO ext attr and streamdir udf_nodes */
4686 
4687 	*udf_noderes = udf_node;
4688 
4689 	return 0;
4690 }
4691 
4692 /* --------------------------------------------------------------------- */
4693 
4694 
4695 /* TODO !!!!! writeout alloc_ext_entry's!!! */
4696 int
4697 udf_writeout_node(struct udf_node *udf_node, int waitfor)
4698 {
4699 	union dscrptr *dscr;
4700 	struct long_ad *loc;
4701 	int error;
4702 
4703 	DPRINTF(NODE, ("udf_writeout_node called\n"));
4704 
4705 	KASSERT(udf_node->outstanding_bufs == 0);
4706 	KASSERT(udf_node->outstanding_nodedscr == 0);
4707 
4708 	KASSERT(LIST_EMPTY(&udf_node->vnode->v_dirtyblkhd));
4709 
4710 	if (udf_node->i_flags & IN_DELETED) {
4711 		DPRINTF(NODE, ("\tnode deleted; not writing out\n"));
4712 		return 0;
4713 	}
4714 
4715 	/* we're going to write out the descriptor so clear the flags */
4716 	udf_node->i_flags &= ~(IN_MODIFIED | IN_ACCESSED);
4717 
4718 	if (udf_node->fe) {
4719 		dscr = (union dscrptr *) udf_node->fe;
4720 	} else {
4721 		KASSERT(udf_node->efe);
4722 		dscr = (union dscrptr *) udf_node->efe;
4723 	}
4724 	KASSERT(dscr);
4725 
4726 	loc = &udf_node->write_loc;
4727 	error = udf_write_logvol_dscr(udf_node, dscr, loc, waitfor);
4728 	return error;
4729 }
4730 
4731 /* --------------------------------------------------------------------- */
4732 
4733 int
4734 udf_dispose_node(struct udf_node *udf_node)
4735 {
4736 	struct vnode *vp;
4737 
4738 	DPRINTF(NODE, ("udf_dispose_node called on node %p\n", udf_node));
4739 	if (!udf_node) {
4740 		DPRINTF(NODE, ("UDF: Dispose node on node NULL, ignoring\n"));
4741 		return 0;
4742 	}
4743 
4744 	vp  = udf_node->vnode;
4745 #ifdef DIAGNOSTIC
4746 	if (vp->v_numoutput)
4747 		panic("disposing UDF node with pending I/O's, udf_node = %p, "
4748 				"v_numoutput = %d", udf_node, vp->v_numoutput);
4749 #endif
4750 
4751 	/* wait until out of sync (just in case we happen to stumble over one */
4752 	KASSERT(!mutex_owned(&mntvnode_lock));
4753 	mutex_enter(&mntvnode_lock);
4754 	while (udf_node->i_flags & IN_SYNCED) {
4755 		cv_timedwait(&udf_node->ump->dirtynodes_cv, &mntvnode_lock,
4756 			hz/16);
4757 	}
4758 	mutex_exit(&mntvnode_lock);
4759 
4760 	/* TODO extended attributes and streamdir */
4761 
4762 	/* remove from our hash lookup table */
4763 	udf_deregister_node(udf_node);
4764 
4765 	/* destroy our lock */
4766 	mutex_destroy(&udf_node->node_mutex);
4767 	cv_destroy(&udf_node->node_lock);
4768 
4769 	/* dissociate our udf_node from the vnode */
4770 	genfs_node_destroy(udf_node->vnode);
4771 	vp->v_data = NULL;
4772 
4773 	/* free associated memory and the node itself */
4774 	if (udf_node->fe)
4775 		udf_free_logvol_dscr(udf_node->ump, &udf_node->loc, udf_node->fe);
4776 	if (udf_node->efe)
4777 		udf_free_logvol_dscr(udf_node->ump, &udf_node->loc, udf_node->efe);
4778 
4779 	udf_node->fe  = (void *) 0xdeadaaaa;
4780 	udf_node->efe = (void *) 0xdeadbbbb;
4781 	udf_node->ump = (void *) 0xdeadbeef;
4782 	pool_put(&udf_node_pool, udf_node);
4783 
4784 	return 0;
4785 }
4786 
4787 
4788 
4789 /*
4790  * create a new node using the specified vnodeops, vap and cnp but with the
4791  * udf_file_type. This allows special files to be created. Use with care.
4792  */
4793 
4794 static int
4795 udf_create_node_raw(struct vnode *dvp, struct vnode **vpp, int udf_file_type,
4796 	int (**vnodeops)(void *), struct vattr *vap, struct componentname *cnp)
4797 {
4798 	union dscrptr *dscr;
4799 	struct udf_node *dir_node = VTOI(dvp);;
4800 	struct udf_node *udf_node;
4801 	struct udf_mount *ump = dir_node->ump;
4802 	struct vnode *nvp;
4803 	struct long_ad node_icb_loc;
4804 	uint64_t parent_unique_id;
4805 	uint64_t lmapping, pmapping;
4806 	uint32_t lb_size, lb_num;
4807 	uint16_t vpart_num;
4808 	int fid_size, error;
4809 
4810 	lb_size = udf_rw32(ump->logical_vol->lb_size);
4811 	*vpp = NULL;
4812 
4813 	/* allocate vnode */
4814 	error = getnewvnode(VT_UDF, ump->vfs_mountp, vnodeops, &nvp);
4815         if (error)
4816 		return error;
4817 
4818 	/* lock node */
4819 	error = vn_lock(nvp, LK_EXCLUSIVE | LK_RETRY);
4820 	if (error) {
4821 		nvp->v_data = NULL;
4822 		ungetnewvnode(nvp);
4823 		return error;
4824 	}
4825 
4826 	/* get disc allocation for one logical block */
4827 	error = udf_pre_allocate_space(ump, UDF_C_NODE, 1,
4828 			&vpart_num, &lmapping, &pmapping);
4829 	lb_num = lmapping;
4830 	if (error) {
4831 		vlockmgr(nvp->v_vnlock, LK_RELEASE);
4832 		ungetnewvnode(nvp);
4833 		return error;
4834 	}
4835 
4836 	/* initialise pointer to location */
4837 	memset(&node_icb_loc, 0, sizeof(struct long_ad));
4838 	node_icb_loc.len = lb_size;
4839 	node_icb_loc.loc.lb_num   = udf_rw32(lb_num);
4840 	node_icb_loc.loc.part_num = udf_rw16(vpart_num);
4841 
4842 	/* build udf_node (do initialise!) */
4843 	udf_node = pool_get(&udf_node_pool, PR_WAITOK);
4844 	memset(udf_node, 0, sizeof(struct udf_node));
4845 
4846 	/* initialise crosslinks, note location of fe/efe for hashing */
4847 	/* bugalert: synchronise with udf_get_node() */
4848 	udf_node->ump       = ump;
4849 	udf_node->vnode     = nvp;
4850 	nvp->v_data         = udf_node;
4851 	udf_node->loc       = node_icb_loc;
4852 	udf_node->write_loc = node_icb_loc;
4853 	udf_node->lockf     = 0;
4854 	mutex_init(&udf_node->node_mutex, MUTEX_DEFAULT, IPL_NONE);
4855 	cv_init(&udf_node->node_lock, "udf_nlk");
4856 	udf_node->outstanding_bufs = 0;
4857 	udf_node->outstanding_nodedscr = 0;
4858 
4859 	/* initialise genfs */
4860 	genfs_node_init(nvp, &udf_genfsops);
4861 
4862 	/* insert into the hash lookup */
4863 	udf_register_node(udf_node);
4864 
4865 	/* get parent's unique ID for refering '..' if its a directory */
4866 	if (dir_node->fe) {
4867 		parent_unique_id = udf_rw64(dir_node->fe->unique_id);
4868 	} else {
4869 		parent_unique_id = udf_rw64(dir_node->efe->unique_id);
4870 	}
4871 
4872 	/* get descriptor */
4873 	udf_create_logvol_dscr(ump, udf_node, &node_icb_loc, &dscr);
4874 
4875 	/* choose a fe or an efe for it */
4876 	if (ump->logical_vol->tag.descriptor_ver == 2) {
4877 		udf_node->fe = &dscr->fe;
4878 		fid_size = udf_create_new_fe(ump, udf_node->fe,
4879 			udf_file_type, &udf_node->loc,
4880 			&dir_node->loc, parent_unique_id);
4881 		/* TODO add extended attribute for creation time */
4882 	} else {
4883 		udf_node->efe = &dscr->efe;
4884 		fid_size = udf_create_new_efe(ump, udf_node->efe,
4885 			udf_file_type, &udf_node->loc,
4886 			&dir_node->loc, parent_unique_id);
4887 	}
4888 	KASSERT(dscr->tag.tag_loc == udf_node->loc.loc.lb_num);
4889 
4890 	/* update vnode's size and type */
4891 	nvp->v_type = vap->va_type;
4892 	uvm_vnp_setsize(nvp, fid_size);
4893 
4894 	/* set access mode */
4895 	udf_setaccessmode(udf_node, vap->va_mode);
4896 
4897 	/* set ownership */
4898 	udf_setownership(udf_node, vap->va_uid, vap->va_gid);
4899 
4900 	error = udf_dir_attach(ump, dir_node, udf_node, vap, cnp);
4901 	if (error) {
4902 		/* free disc allocation for node */
4903 		udf_free_allocated_space(ump, lb_num, vpart_num, 1);
4904 
4905 		/* recycle udf_node */
4906 		udf_dispose_node(udf_node);
4907 		vput(nvp);
4908 
4909 		*vpp = NULL;
4910 		return error;
4911 	}
4912 
4913 	/* adjust file count */
4914 	udf_adjust_filecount(udf_node, 1);
4915 
4916 	/* return result */
4917 	*vpp = nvp;
4918 
4919 	return 0;
4920 }
4921 
4922 
4923 int
4924 udf_create_node(struct vnode *dvp, struct vnode **vpp, struct vattr *vap,
4925 	struct componentname *cnp)
4926 {
4927 	int (**vnodeops)(void *);
4928 	int udf_file_type;
4929 
4930 	DPRINTF(NODE, ("udf_create_node called\n"));
4931 
4932 	/* what type are we creating ? */
4933 	vnodeops = udf_vnodeop_p;
4934 	/* start with a default */
4935 	udf_file_type = UDF_ICB_FILETYPE_RANDOMACCESS;
4936 
4937 	*vpp = NULL;
4938 
4939 	switch (vap->va_type) {
4940 	case VREG :
4941 		udf_file_type = UDF_ICB_FILETYPE_RANDOMACCESS;
4942 		break;
4943 	case VDIR :
4944 		udf_file_type = UDF_ICB_FILETYPE_DIRECTORY;
4945 		break;
4946 	case VLNK :
4947 		udf_file_type = UDF_ICB_FILETYPE_SYMLINK;
4948 		break;
4949 	case VBLK :
4950 		udf_file_type = UDF_ICB_FILETYPE_BLOCKDEVICE;
4951 		/* specfs */
4952 		return ENOTSUP;
4953 		break;
4954 	case VCHR :
4955 		udf_file_type = UDF_ICB_FILETYPE_CHARDEVICE;
4956 		/* specfs */
4957 		return ENOTSUP;
4958 		break;
4959 	case VFIFO :
4960 		udf_file_type = UDF_ICB_FILETYPE_FIFO;
4961 		/* specfs */
4962 		return ENOTSUP;
4963 		break;
4964 	case VSOCK :
4965 		udf_file_type = UDF_ICB_FILETYPE_SOCKET;
4966 		/* specfs */
4967 		return ENOTSUP;
4968 		break;
4969 	case VNON :
4970 	case VBAD :
4971 	default :
4972 		/* nothing; can we even create these? */
4973 		return EINVAL;
4974 	}
4975 
4976 	return udf_create_node_raw(dvp, vpp, udf_file_type, vnodeops, vap, cnp);
4977 }
4978 
4979 /* --------------------------------------------------------------------- */
4980 
4981 static void
4982 udf_free_descriptor_space(struct udf_node *udf_node, struct long_ad *loc, void *mem)
4983 {
4984 	struct udf_mount *ump = udf_node->ump;
4985 	uint32_t lb_size, lb_num, len, num_lb;
4986 	uint16_t vpart_num;
4987 
4988 	/* is there really one? */
4989 	if (mem == NULL)
4990 		return;
4991 
4992 	/* got a descriptor here */
4993 	len       = udf_rw32(loc->len);
4994 	lb_num    = udf_rw32(loc->loc.lb_num);
4995 	vpart_num = udf_rw16(loc->loc.part_num);
4996 
4997 	lb_size = udf_rw32(ump->logical_vol->lb_size);
4998 	num_lb = (len + lb_size -1) / lb_size;
4999 
5000 	udf_free_allocated_space(ump, lb_num, vpart_num, num_lb);
5001 }
5002 
5003 void
5004 udf_delete_node(struct udf_node *udf_node)
5005 {
5006 	void *dscr;
5007 	struct udf_mount *ump;
5008 	struct long_ad *loc;
5009 	int extnr, lvint, dummy;
5010 
5011 	ump = udf_node->ump;
5012 
5013 	/* paranoia check on integrity; should be open!; we could panic */
5014 	lvint = udf_rw32(udf_node->ump->logvol_integrity->integrity_type);
5015 	if (lvint == UDF_INTEGRITY_CLOSED)
5016 		printf("\tIntegrity was CLOSED!\n");
5017 
5018 	/* whatever the node type, change its size to zero */
5019 	(void) udf_resize_node(udf_node, 0, &dummy);
5020 
5021 	/* force it to be `clean'; no use writing it out */
5022 	udf_node->i_flags &= ~(IN_MODIFIED | IN_ACCESSED | IN_ACCESS |
5023 		IN_CHANGE | IN_UPDATE | IN_MODIFY);
5024 
5025 	/* adjust file count */
5026 	udf_adjust_filecount(udf_node, -1);
5027 
5028 	/*
5029 	 * Free its allocated descriptors; memory will be released when
5030 	 * vop_reclaim() is called.
5031 	 */
5032 	loc = &udf_node->loc;
5033 
5034 	dscr = udf_node->fe;
5035 	udf_free_descriptor_space(udf_node, loc, dscr);
5036 	dscr = udf_node->efe;
5037 	udf_free_descriptor_space(udf_node, loc, dscr);
5038 
5039 	for (extnr = 0; extnr < UDF_MAX_ALLOC_EXTENTS; extnr++) {
5040 		dscr =  udf_node->ext[extnr];
5041 		loc  = &udf_node->ext_loc[extnr];
5042 		udf_free_descriptor_space(udf_node, loc, dscr);
5043 	}
5044 }
5045 
5046 /* --------------------------------------------------------------------- */
5047 
5048 /* set new filesize; node but be LOCKED on entry and is locked on exit */
5049 int
5050 udf_resize_node(struct udf_node *udf_node, uint64_t new_size, int *extended)
5051 {
5052 	struct file_entry    *fe  = udf_node->fe;
5053 	struct extfile_entry *efe = udf_node->efe;
5054 	uint64_t file_size;
5055 	int error;
5056 
5057 	if (fe) {
5058 		file_size  = udf_rw64(fe->inf_len);
5059 	} else {
5060 		assert(udf_node->efe);
5061 		file_size  = udf_rw64(efe->inf_len);
5062 	}
5063 
5064 	DPRINTF(ATTR, ("\tchanging file length from %"PRIu64" to %"PRIu64"\n",
5065 			file_size, new_size));
5066 
5067 	/* if not changing, we're done */
5068 	if (file_size == new_size)
5069 		return 0;
5070 
5071 	*extended = (new_size > file_size);
5072 	if (*extended) {
5073 		error = udf_grow_node(udf_node, new_size);
5074 	} else {
5075 		error = udf_shrink_node(udf_node, new_size);
5076 	}
5077 
5078 	return error;
5079 }
5080 
5081 
5082 /* --------------------------------------------------------------------- */
5083 
5084 void
5085 udf_itimes(struct udf_node *udf_node, struct timespec *acc,
5086 	struct timespec *mod, struct timespec *changed)
5087 {
5088 	struct timespec now;
5089 	struct file_entry    *fe;
5090 	struct extfile_entry *efe;
5091 	struct timestamp *atime, *mtime, *attrtime;
5092 
5093 	/* protect against rogue values */
5094 	if (!udf_node)
5095 		return;
5096 
5097 	fe  = udf_node->fe;
5098 	efe = udf_node->efe;
5099 
5100 	if (!(udf_node->i_flags & (IN_ACCESS|IN_CHANGE|IN_UPDATE|IN_MODIFY)))
5101 		return;
5102 
5103 	/* get descriptor information */
5104 	if (fe) {
5105 		atime    = &fe->atime;
5106 		mtime    = &fe->mtime;
5107 		attrtime = &fe->attrtime;
5108 	} else {
5109 		assert(udf_node->efe);
5110 		atime    = &efe->atime;
5111 		mtime    = &efe->mtime;
5112 		attrtime = &efe->attrtime;
5113 	}
5114 
5115 	vfs_timestamp(&now);
5116 
5117 	/* set access time */
5118 	if (udf_node->i_flags & IN_ACCESS) {
5119 		if (acc == NULL)
5120 			acc = &now;
5121 		udf_timespec_to_timestamp(acc, atime);
5122 	}
5123 
5124 	/* set modification time */
5125 	if (udf_node->i_flags & (IN_UPDATE | IN_MODIFY)) {
5126 		if (mod == NULL)
5127 			mod = &now;
5128 		udf_timespec_to_timestamp(mod, mtime);
5129 	}
5130 
5131 	/* set change time */
5132 	if (udf_node->i_flags & (IN_CHANGE | IN_MODIFY)) {
5133 		if (changed == NULL)
5134 			changed = &now;
5135 		udf_timespec_to_timestamp(changed, attrtime);
5136 	}
5137 
5138 	/* notify updates to the node itself */
5139 	if (udf_node->i_flags & (IN_ACCESS | IN_MODIFY))
5140 		udf_node->i_flags |= IN_ACCESSED;
5141 	if (udf_node->i_flags & (IN_UPDATE | IN_CHANGE))
5142 		udf_node->i_flags |= IN_MODIFIED;
5143 
5144 	/* clear modification flags */
5145 	udf_node->i_flags &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY);
5146 }
5147 
5148 /* --------------------------------------------------------------------- */
5149 
5150 int
5151 udf_update(struct vnode *vp, struct timespec *acc,
5152 	struct timespec *mod, int updflags)
5153 {
5154 	struct udf_node  *udf_node = VTOI(vp);
5155 	struct udf_mount *ump = udf_node->ump;
5156 	struct regid     *impl_id;
5157 	int mnt_async = (vp->v_mount->mnt_flag & MNT_ASYNC);
5158 	int waitfor, flags;
5159 
5160 #ifdef DEBUG
5161 	char bits[128];
5162 	DPRINTF(CALL, ("udf_update(node, %p, %p, %d)\n", acc, mod, updflags));
5163 	bitmask_snprintf(udf_node->i_flags, IN_FLAGBITS, bits, sizeof(bits));
5164 	DPRINTF(CALL, ("\tnode flags %s\n", bits));
5165 	DPRINTF(CALL, ("\t\tmnt_async = %d\n", mnt_async));
5166 #endif
5167 
5168 	/* set our times */
5169 	udf_itimes(udf_node, acc, mod, NULL);
5170 
5171 	/* set our implementation id */
5172 	if (udf_node->fe) {
5173 		impl_id = &udf_node->fe->imp_id;
5174 	} else {
5175 		impl_id = &udf_node->efe->imp_id;
5176 	}
5177 	udf_set_regid(impl_id, IMPL_NAME);
5178 	udf_add_impl_regid(ump, impl_id);
5179 
5180 	/* if called when mounted readonly, never write back */
5181 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
5182 		return 0;
5183 
5184 	/* check if the node is dirty 'enough'*/
5185 	if (updflags & UPDATE_CLOSE) {
5186 		flags = udf_node->i_flags & (IN_MODIFIED | IN_ACCESSED);
5187 	} else {
5188 		flags = udf_node->i_flags & IN_MODIFIED;
5189 	}
5190 	if (flags == 0)
5191 		return 0;
5192 
5193 	/* determine if we need to write sync or async */
5194 	waitfor = 0;
5195 	if ((flags & IN_MODIFIED) && (mnt_async == 0)) {
5196 		/* sync mounted */
5197 		waitfor = updflags & UPDATE_WAIT;
5198 		if (updflags & UPDATE_DIROP)
5199 			waitfor |= UPDATE_WAIT;
5200 	}
5201 	if (waitfor)
5202 		return VOP_FSYNC(vp, FSCRED, FSYNC_WAIT, 0,0);
5203 
5204 	return 0;
5205 }
5206 
5207 
5208 /* --------------------------------------------------------------------- */
5209 
5210 /*
5211  * Read one fid and process it into a dirent and advance to the next (*fid)
5212  * has to be allocated a logical block in size, (*dirent) struct dirent length
5213  */
5214 
5215 int
5216 udf_read_fid_stream(struct vnode *vp, uint64_t *offset,
5217 		    struct fileid_desc *fid, struct dirent *dirent)
5218 {
5219 	struct udf_node  *dir_node = VTOI(vp);
5220 	struct udf_mount *ump = dir_node->ump;
5221 	struct file_entry    *fe  = dir_node->fe;
5222 	struct extfile_entry *efe = dir_node->efe;
5223 	uint32_t      fid_size, lb_size;
5224 	uint64_t      file_size;
5225 	char         *fid_name;
5226 	int           enough, error;
5227 
5228 	assert(fid);
5229 	assert(dirent);
5230 	assert(dir_node);
5231 	assert(offset);
5232 	assert(*offset != 1);
5233 
5234 	DPRINTF(FIDS, ("read_fid_stream called at offset %"PRIu64"\n", *offset));
5235 	/* check if we're past the end of the directory */
5236 	if (fe) {
5237 		file_size = udf_rw64(fe->inf_len);
5238 	} else {
5239 		assert(dir_node->efe);
5240 		file_size = udf_rw64(efe->inf_len);
5241 	}
5242 	if (*offset >= file_size)
5243 		return EINVAL;
5244 
5245 	/* get maximum length of FID descriptor */
5246 	lb_size = udf_rw32(ump->logical_vol->lb_size);
5247 
5248 	/* initialise return values */
5249 	fid_size = 0;
5250 	memset(dirent, 0, sizeof(struct dirent));
5251 	memset(fid, 0, lb_size);
5252 
5253 	enough  = (file_size - (*offset) >= UDF_FID_SIZE);
5254 	if (!enough) {
5255 		/* short dir ... */
5256 		return EIO;
5257 	}
5258 
5259 	error = vn_rdwr(UIO_READ, vp,
5260 			fid, MIN(file_size - (*offset), lb_size), *offset,
5261 			UIO_SYSSPACE, IO_ALTSEMANTICS | IO_NODELOCKED, FSCRED,
5262 			NULL, NULL);
5263 	if (error)
5264 		return error;
5265 
5266 	DPRINTF(FIDS, ("\tfid piece read in fine\n"));
5267 	/*
5268 	 * Check if we got a whole descriptor.
5269 	 * TODO Try to `resync' directory stream when something is very wrong.
5270 	 */
5271 
5272 	/* check if our FID header is OK */
5273 	error = udf_check_tag(fid);
5274 	if (error) {
5275 		goto brokendir;
5276 	}
5277 	DPRINTF(FIDS, ("\ttag check ok\n"));
5278 
5279 	if (udf_rw16(fid->tag.id) != TAGID_FID) {
5280 		error = EIO;
5281 		goto brokendir;
5282 	}
5283 	DPRINTF(FIDS, ("\ttag checked ok: got TAGID_FID\n"));
5284 
5285 	/* check for length */
5286 	fid_size = udf_fidsize(fid);
5287 	enough = (file_size - (*offset) >= fid_size);
5288 	if (!enough) {
5289 		error = EIO;
5290 		goto brokendir;
5291 	}
5292 	DPRINTF(FIDS, ("\tthe complete fid is read in\n"));
5293 
5294 	/* check FID contents */
5295 	error = udf_check_tag_payload((union dscrptr *) fid, lb_size);
5296 brokendir:
5297 	if (error) {
5298 		/* note that is sometimes a bit quick to report */
5299 		printf("BROKEN DIRECTORY ENTRY\n");
5300 		/* RESYNC? */
5301 		/* TODO: use udf_resync_fid_stream */
5302 		return EIO;
5303 	}
5304 	DPRINTF(FIDS, ("\tpayload checked ok\n"));
5305 
5306 	/* we got a whole and valid descriptor! */
5307 	DPRINTF(FIDS, ("\tinterpret FID\n"));
5308 
5309 	/* create resulting dirent structure */
5310 	fid_name = (char *) fid->data + udf_rw16(fid->l_iu);
5311 	udf_to_unix_name(dirent->d_name, MAXNAMLEN,
5312 		fid_name, fid->l_fi, &ump->logical_vol->desc_charset);
5313 
5314 	/* '..' has no name, so provide one */
5315 	if (fid->file_char & UDF_FILE_CHAR_PAR)
5316 		strcpy(dirent->d_name, "..");
5317 
5318 	dirent->d_fileno = udf_calchash(&fid->icb);	/* inode hash XXX */
5319 	dirent->d_namlen = strlen(dirent->d_name);
5320 	dirent->d_reclen = _DIRENT_SIZE(dirent);
5321 
5322 	/*
5323 	 * Note that its not worth trying to go for the filetypes now... its
5324 	 * too expensive too
5325 	 */
5326 	dirent->d_type = DT_UNKNOWN;
5327 
5328 	/* initial guess for filetype we can make */
5329 	if (fid->file_char & UDF_FILE_CHAR_DIR)
5330 		dirent->d_type = DT_DIR;
5331 
5332 	/* advance */
5333 	*offset += fid_size;
5334 
5335 	return error;
5336 }
5337 
5338 
5339 /* --------------------------------------------------------------------- */
5340 
5341 static void
5342 udf_sync_pass(struct udf_mount *ump, kauth_cred_t cred, int waitfor,
5343 	int pass, int *ndirty)
5344 {
5345 	struct udf_node *udf_node, *n_udf_node;
5346 	struct vnode *vp;
5347 	int vdirty, error;
5348 	int on_type, on_flags, on_vnode;
5349 
5350 derailed:
5351 	KASSERT(mutex_owned(&mntvnode_lock));
5352 
5353 	DPRINTF(SYNC, ("sync_pass %d\n", pass));
5354 	udf_node = LIST_FIRST(&ump->sorted_udf_nodes);
5355 	for (;udf_node; udf_node = n_udf_node) {
5356 		DPRINTF(SYNC, ("."));
5357 
5358 		udf_node->i_flags &= ~IN_SYNCED;
5359 		vp = udf_node->vnode;
5360 
5361 		mutex_enter(&vp->v_interlock);
5362 		n_udf_node = LIST_NEXT(udf_node, sortchain);
5363 		if (n_udf_node)
5364 			n_udf_node->i_flags |= IN_SYNCED;
5365 
5366 		/* system nodes are not synced this way */
5367 		if (vp->v_vflag & VV_SYSTEM) {
5368 			mutex_exit(&vp->v_interlock);
5369 			continue;
5370 		}
5371 
5372 		/* check if its dirty enough to even try */
5373 		on_type  = (waitfor == MNT_LAZY || vp->v_type == VNON);
5374 		on_flags = ((udf_node->i_flags &
5375 			(IN_ACCESSED | IN_UPDATE | IN_MODIFIED)) == 0);
5376 		on_vnode = LIST_EMPTY(&vp->v_dirtyblkhd)
5377 			&& UVM_OBJ_IS_CLEAN(&vp->v_uobj);
5378 		if (on_type || (on_flags || on_vnode)) { /* XXX */
5379 			/* not dirty (enough?) */
5380 			mutex_exit(&vp->v_interlock);
5381 			continue;
5382 		}
5383 
5384 		mutex_exit(&mntvnode_lock);
5385 		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK);
5386 		if (error) {
5387 			mutex_enter(&mntvnode_lock);
5388 			if (error == ENOENT)
5389 				goto derailed;
5390 			*ndirty += 1;
5391 			continue;
5392 		}
5393 
5394 		switch (pass) {
5395 		case 1:
5396 			VOP_FSYNC(vp, cred, 0 | FSYNC_DATAONLY,0,0);
5397 			break;
5398 		case 2:
5399 			vdirty = vp->v_numoutput;
5400 			if (vp->v_tag == VT_UDF)
5401 				vdirty += udf_node->outstanding_bufs +
5402 					udf_node->outstanding_nodedscr;
5403 			if (vdirty == 0)
5404 				VOP_FSYNC(vp, cred, 0,0,0);
5405 			*ndirty += vdirty;
5406 			break;
5407 		case 3:
5408 			vdirty = vp->v_numoutput;
5409 			if (vp->v_tag == VT_UDF)
5410 				vdirty += udf_node->outstanding_bufs +
5411 					udf_node->outstanding_nodedscr;
5412 			*ndirty += vdirty;
5413 			break;
5414 		}
5415 
5416 		vput(vp);
5417 		mutex_enter(&mntvnode_lock);
5418 	}
5419 	DPRINTF(SYNC, ("END sync_pass %d\n", pass));
5420 }
5421 
5422 
5423 void
5424 udf_do_sync(struct udf_mount *ump, kauth_cred_t cred, int waitfor)
5425 {
5426 	int dummy, ndirty;
5427 
5428 	mutex_enter(&mntvnode_lock);
5429 recount:
5430 	dummy = 0;
5431 	DPRINTF(CALL, ("issue VOP_FSYNC(DATA only) on all nodes\n"));
5432 	DPRINTF(SYNC, ("issue VOP_FSYNC(DATA only) on all nodes\n"));
5433 	udf_sync_pass(ump, cred, waitfor, 1, &dummy);
5434 
5435 	DPRINTF(CALL, ("issue VOP_FSYNC(COMPLETE) on all finished nodes\n"));
5436 	DPRINTF(SYNC, ("issue VOP_FSYNC(COMPLETE) on all finished nodes\n"));
5437 	udf_sync_pass(ump, cred, waitfor, 2, &dummy);
5438 
5439 	if (waitfor == MNT_WAIT) {
5440 		ndirty = ump->devvp->v_numoutput;
5441 		DPRINTF(NODE, ("counting pending blocks: on devvp %d\n",
5442 			ndirty));
5443 		udf_sync_pass(ump, cred, waitfor, 3, &ndirty);
5444 		DPRINTF(NODE, ("counted num dirty pending blocks %d\n",
5445 			ndirty));
5446 
5447 		if (ndirty) {
5448 			/* 1/4 second wait */
5449 			cv_timedwait(&ump->dirtynodes_cv, &mntvnode_lock,
5450 				hz/4);
5451 			goto recount;
5452 		}
5453 	}
5454 
5455 	mutex_exit(&mntvnode_lock);
5456 }
5457 
5458 /* --------------------------------------------------------------------- */
5459 
5460 /*
5461  * Read and write file extent in/from the buffer.
5462  *
5463  * The splitup of the extent into seperate request-buffers is to minimise
5464  * copying around as much as possible.
5465  *
5466  * block based file reading and writing
5467  */
5468 
5469 static int
5470 udf_read_internal(struct udf_node *node, uint8_t *blob)
5471 {
5472 	struct udf_mount *ump;
5473 	struct file_entry     *fe = node->fe;
5474 	struct extfile_entry *efe = node->efe;
5475 	uint64_t inflen;
5476 	uint32_t sector_size;
5477 	uint8_t  *pos;
5478 	int icbflags, addr_type;
5479 
5480 	/* get extent and do some paranoia checks */
5481 	ump = node->ump;
5482 	sector_size = ump->discinfo.sector_size;
5483 
5484 	if (fe) {
5485 		inflen   = udf_rw64(fe->inf_len);
5486 		pos      = &fe->data[0] + udf_rw32(fe->l_ea);
5487 		icbflags = udf_rw16(fe->icbtag.flags);
5488 	} else {
5489 		assert(node->efe);
5490 		inflen   = udf_rw64(efe->inf_len);
5491 		pos      = &efe->data[0] + udf_rw32(efe->l_ea);
5492 		icbflags = udf_rw16(efe->icbtag.flags);
5493 	}
5494 	addr_type = icbflags & UDF_ICB_TAG_FLAGS_ALLOC_MASK;
5495 
5496 	assert(addr_type == UDF_ICB_INTERN_ALLOC);
5497 	assert(inflen < sector_size);
5498 
5499 	/* copy out info */
5500 	memset(blob, 0, sector_size);
5501 	memcpy(blob, pos, inflen);
5502 
5503 	return 0;
5504 }
5505 
5506 
5507 static int
5508 udf_write_internal(struct udf_node *node, uint8_t *blob)
5509 {
5510 	struct udf_mount *ump;
5511 	struct file_entry     *fe = node->fe;
5512 	struct extfile_entry *efe = node->efe;
5513 	uint64_t inflen;
5514 	uint32_t sector_size;
5515 	uint8_t  *pos;
5516 	int icbflags, addr_type;
5517 
5518 	/* get extent and do some paranoia checks */
5519 	ump = node->ump;
5520 	sector_size = ump->discinfo.sector_size;
5521 
5522 	if (fe) {
5523 		inflen   = udf_rw64(fe->inf_len);
5524 		pos      = &fe->data[0] + udf_rw32(fe->l_ea);
5525 		icbflags = udf_rw16(fe->icbtag.flags);
5526 	} else {
5527 		assert(node->efe);
5528 		inflen   = udf_rw64(efe->inf_len);
5529 		pos      = &efe->data[0] + udf_rw32(efe->l_ea);
5530 		icbflags = udf_rw16(efe->icbtag.flags);
5531 	}
5532 	addr_type = icbflags & UDF_ICB_TAG_FLAGS_ALLOC_MASK;
5533 
5534 	assert(addr_type == UDF_ICB_INTERN_ALLOC);
5535 	assert(inflen < sector_size);
5536 
5537 	/* copy in blob */
5538 	/* memset(pos, 0, inflen); */
5539 	memcpy(pos, blob, inflen);
5540 
5541 	return 0;
5542 }
5543 
5544 
5545 void
5546 udf_read_filebuf(struct udf_node *udf_node, struct buf *buf)
5547 {
5548 	struct buf *nestbuf;
5549 	struct udf_mount *ump = udf_node->ump;
5550 	uint64_t   *mapping;
5551 	uint64_t    run_start;
5552 	uint32_t    sector_size;
5553 	uint32_t    buf_offset, sector, rbuflen, rblk;
5554 	uint32_t    from, lblkno;
5555 	uint32_t    sectors;
5556 	uint8_t    *buf_pos;
5557 	int error, run_length, isdir, what;
5558 
5559 	sector_size = udf_node->ump->discinfo.sector_size;
5560 
5561 	from    = buf->b_blkno;
5562 	sectors = buf->b_bcount / sector_size;
5563 
5564 	isdir   = (udf_node->vnode->v_type == VDIR);
5565 	what    = isdir ? UDF_C_FIDS : UDF_C_USERDATA;
5566 
5567 	/* assure we have enough translation slots */
5568 	KASSERT(buf->b_bcount / sector_size <= UDF_MAX_MAPPINGS);
5569 	KASSERT(MAXPHYS / sector_size <= UDF_MAX_MAPPINGS);
5570 
5571 	if (sectors > UDF_MAX_MAPPINGS) {
5572 		printf("udf_read_filebuf: implementation limit on bufsize\n");
5573 		buf->b_error  = EIO;
5574 		biodone(buf);
5575 		return;
5576 	}
5577 
5578 	mapping = malloc(sizeof(*mapping) * UDF_MAX_MAPPINGS, M_TEMP, M_WAITOK);
5579 
5580 	error = 0;
5581 	DPRINTF(READ, ("\ttranslate %d-%d\n", from, sectors));
5582 	error = udf_translate_file_extent(udf_node, from, sectors, mapping);
5583 	if (error) {
5584 		buf->b_error  = error;
5585 		biodone(buf);
5586 		goto out;
5587 	}
5588 	DPRINTF(READ, ("\ttranslate extent went OK\n"));
5589 
5590 	/* pre-check if its an internal */
5591 	if (*mapping == UDF_TRANS_INTERN) {
5592 		error = udf_read_internal(udf_node, (uint8_t *) buf->b_data);
5593 		if (error)
5594 			buf->b_error  = error;
5595 		biodone(buf);
5596 		goto out;
5597 	}
5598 	DPRINTF(READ, ("\tnot intern\n"));
5599 
5600 #ifdef DEBUG
5601 	if (udf_verbose & UDF_DEBUG_TRANSLATE) {
5602 		printf("Returned translation table:\n");
5603 		for (sector = 0; sector < sectors; sector++) {
5604 			printf("%d : %"PRIu64"\n", sector, mapping[sector]);
5605 		}
5606 	}
5607 #endif
5608 
5609 	/* request read-in of data from disc sheduler */
5610 	buf->b_resid = buf->b_bcount;
5611 	for (sector = 0; sector < sectors; sector++) {
5612 		buf_offset = sector * sector_size;
5613 		buf_pos    = (uint8_t *) buf->b_data + buf_offset;
5614 		DPRINTF(READ, ("\tprocessing rel sector %d\n", sector));
5615 
5616 		/* check if its zero or unmapped to stop reading */
5617 		switch (mapping[sector]) {
5618 		case UDF_TRANS_UNMAPPED:
5619 		case UDF_TRANS_ZERO:
5620 			/* copy zero sector TODO runlength like below */
5621 			memset(buf_pos, 0, sector_size);
5622 			DPRINTF(READ, ("\treturning zero sector\n"));
5623 			nestiobuf_done(buf, sector_size, 0);
5624 			break;
5625 		default :
5626 			DPRINTF(READ, ("\tread sector "
5627 			    "%"PRIu64"\n", mapping[sector]));
5628 
5629 			lblkno = from + sector;
5630 			run_start  = mapping[sector];
5631 			run_length = 1;
5632 			while (sector < sectors-1) {
5633 				if (mapping[sector+1] != mapping[sector]+1)
5634 					break;
5635 				run_length++;
5636 				sector++;
5637 			}
5638 
5639 			/*
5640 			 * nest an iobuf and mark it for async reading. Since
5641 			 * we're using nested buffers, they can't be cached by
5642 			 * design.
5643 			 */
5644 			rbuflen = run_length * sector_size;
5645 			rblk    = run_start  * (sector_size/DEV_BSIZE);
5646 
5647 			nestbuf = getiobuf(NULL, true);
5648 			nestiobuf_setup(buf, nestbuf, buf_offset, rbuflen);
5649 			/* nestbuf is B_ASYNC */
5650 
5651 			/* identify this nestbuf */
5652 			nestbuf->b_lblkno   = lblkno;
5653 			assert(nestbuf->b_vp == udf_node->vnode);
5654 
5655 			/* CD shedules on raw blkno */
5656 			nestbuf->b_blkno      = rblk;
5657 			nestbuf->b_proc       = NULL;
5658 			nestbuf->b_rawblkno   = rblk;
5659 			nestbuf->b_udf_c_type = what;
5660 
5661 			udf_discstrat_queuebuf(ump, nestbuf);
5662 		}
5663 	}
5664 out:
5665 	/* if we're synchronously reading, wait for the completion */
5666 	if ((buf->b_flags & B_ASYNC) == 0)
5667 		biowait(buf);
5668 
5669 	DPRINTF(READ, ("\tend of read_filebuf\n"));
5670 	free(mapping, M_TEMP);
5671 	return;
5672 }
5673 
5674 
5675 void
5676 udf_write_filebuf(struct udf_node *udf_node, struct buf *buf)
5677 {
5678 	struct buf *nestbuf;
5679 	struct udf_mount *ump = udf_node->ump;
5680 	uint64_t   *mapping;
5681 	uint64_t    run_start;
5682 	uint32_t    lb_size;
5683 	uint32_t    buf_offset, lb_num, rbuflen, rblk;
5684 	uint32_t    from, lblkno;
5685 	uint32_t    num_lb;
5686 	uint8_t    *buf_pos;
5687 	int error, run_length, isdir, what, s;
5688 
5689 	lb_size = udf_rw32(udf_node->ump->logical_vol->lb_size);
5690 
5691 	from   = buf->b_blkno;
5692 	num_lb = buf->b_bcount / lb_size;
5693 
5694 	isdir  = (udf_node->vnode->v_type == VDIR);
5695 	what   = isdir ? UDF_C_FIDS : UDF_C_USERDATA;
5696 
5697 	/* assure we have enough translation slots */
5698 	KASSERT(buf->b_bcount / lb_size <= UDF_MAX_MAPPINGS);
5699 	KASSERT(MAXPHYS / lb_size <= UDF_MAX_MAPPINGS);
5700 
5701 	if (num_lb > UDF_MAX_MAPPINGS) {
5702 		printf("udf_write_filebuf: implementation limit on bufsize\n");
5703 		buf->b_error  = EIO;
5704 		biodone(buf);
5705 		return;
5706 	}
5707 
5708 	mapping = malloc(sizeof(*mapping) * UDF_MAX_MAPPINGS, M_TEMP, M_WAITOK);
5709 
5710 	error = 0;
5711 	DPRINTF(WRITE, ("\ttranslate %d-%d\n", from, num_lb));
5712 	error = udf_translate_file_extent(udf_node, from, num_lb, mapping);
5713 	if (error) {
5714 		buf->b_error  = error;
5715 		biodone(buf);
5716 		goto out;
5717 	}
5718 	DPRINTF(WRITE, ("\ttranslate extent went OK\n"));
5719 
5720 	/* if its internally mapped, we can write it in the descriptor itself */
5721 	if (*mapping == UDF_TRANS_INTERN) {
5722 		/* TODO paranoia check if we ARE going to have enough space */
5723 		error = udf_write_internal(udf_node, (uint8_t *) buf->b_data);
5724 		if (error)
5725 			buf->b_error  = error;
5726 		biodone(buf);
5727 		goto out;
5728 	}
5729 	DPRINTF(WRITE, ("\tnot intern\n"));
5730 
5731 	/* request write out of data to disc sheduler */
5732 	buf->b_resid = buf->b_bcount;
5733 	for (lb_num = 0; lb_num < num_lb; lb_num++) {
5734 		buf_offset = lb_num * lb_size;
5735 		buf_pos    = (uint8_t *) buf->b_data + buf_offset;
5736 		DPRINTF(WRITE, ("\tprocessing rel lb_num %d\n", lb_num));
5737 
5738 		/*
5739 		 * Mappings are not that important here. Just before we write
5740 		 * the lb_num we late-allocate them when needed and update the
5741 		 * mapping in the udf_node.
5742 		 */
5743 
5744 		/* XXX why not ignore the mapping altogether ? */
5745 		/* TODO estimate here how much will be late-allocated */
5746 		DPRINTF(WRITE, ("\twrite lb_num "
5747 		    "%"PRIu64, mapping[lb_num]));
5748 
5749 		lblkno = from + lb_num;
5750 		run_start  = mapping[lb_num];
5751 		run_length = 1;
5752 		while (lb_num < num_lb-1) {
5753 			if (mapping[lb_num+1] != mapping[lb_num]+1)
5754 				if (mapping[lb_num+1] != mapping[lb_num])
5755 					break;
5756 			run_length++;
5757 			lb_num++;
5758 		}
5759 		DPRINTF(WRITE, ("+ %d\n", run_length));
5760 
5761 		/* nest an iobuf on the master buffer for the extent */
5762 		rbuflen = run_length * lb_size;
5763 		rblk = run_start * (lb_size/DEV_BSIZE);
5764 
5765 #if 0
5766 		/* if its zero or unmapped, our blknr gets -1 for unmapped */
5767 		switch (mapping[lb_num]) {
5768 		case UDF_TRANS_UNMAPPED:
5769 		case UDF_TRANS_ZERO:
5770 			rblk = -1;
5771 			break;
5772 		default:
5773 			rblk = run_start * (lb_size/DEV_BSIZE);
5774 			break;
5775 		}
5776 #endif
5777 
5778 		nestbuf = getiobuf(NULL, true);
5779 		nestiobuf_setup(buf, nestbuf, buf_offset, rbuflen);
5780 		/* nestbuf is B_ASYNC */
5781 
5782 		/* identify this nestbuf */
5783 		nestbuf->b_lblkno   = lblkno;
5784 		KASSERT(nestbuf->b_vp == udf_node->vnode);
5785 
5786 		/* CD shedules on raw blkno */
5787 		nestbuf->b_blkno      = rblk;
5788 		nestbuf->b_proc       = NULL;
5789 		nestbuf->b_rawblkno   = rblk;
5790 		nestbuf->b_udf_c_type = what;
5791 
5792 		/* increment our outstanding bufs counter */
5793 		s = splbio();
5794 			udf_node->outstanding_bufs++;
5795 		splx(s);
5796 
5797 		udf_discstrat_queuebuf(ump, nestbuf);
5798 	}
5799 out:
5800 	/* if we're synchronously writing, wait for the completion */
5801 	if ((buf->b_flags & B_ASYNC) == 0)
5802 		biowait(buf);
5803 
5804 	DPRINTF(WRITE, ("\tend of write_filebuf\n"));
5805 	free(mapping, M_TEMP);
5806 	return;
5807 }
5808 
5809 /* --------------------------------------------------------------------- */
5810 
5811 
5812