xref: /titanic_50/usr/src/uts/common/fs/udfs/udf_subr.c (revision c77a61a72b5ecdc507d6cf104142edd371a16c84)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License, Version 1.0 only
6  * (the "License").  You may not use this file except in compliance
7  * with the License.
8  *
9  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10  * or http://www.opensolaris.org/os/licensing.
11  * See the License for the specific language governing permissions
12  * and limitations under the License.
13  *
14  * When distributing Covered Code, include this CDDL HEADER in each
15  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16  * If applicable, add the following below this CDDL HEADER, with the
17  * fields enclosed by brackets "[]" replaced with your own identifying
18  * information: Portions Copyright [yyyy] [name of copyright owner]
19  *
20  * CDDL HEADER END
21  */
22 /*
23  * Copyright 2004 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 #pragma ident	"%Z%%M%	%I%	%E% SMI"
28 
29 #include <sys/types.h>
30 #include <sys/t_lock.h>
31 #include <sys/param.h>
32 #include <sys/time.h>
33 #include <sys/systm.h>
34 #include <sys/sysmacros.h>
35 #include <sys/resource.h>
36 #include <sys/signal.h>
37 #include <sys/cred.h>
38 #include <sys/user.h>
39 #include <sys/buf.h>
40 #include <sys/vfs.h>
41 #include <sys/stat.h>
42 #include <sys/vnode.h>
43 #include <sys/mode.h>
44 #include <sys/proc.h>
45 #include <sys/disp.h>
46 #include <sys/file.h>
47 #include <sys/fcntl.h>
48 #include <sys/flock.h>
49 #include <sys/kmem.h>
50 #include <sys/uio.h>
51 #include <sys/dnlc.h>
52 #include <sys/conf.h>
53 #include <sys/errno.h>
54 #include <sys/mman.h>
55 #include <sys/fbuf.h>
56 #include <sys/pathname.h>
57 #include <sys/debug.h>
58 #include <sys/vmsystm.h>
59 #include <sys/cmn_err.h>
60 #include <sys/dirent.h>
61 #include <sys/errno.h>
62 #include <sys/modctl.h>
63 #include <sys/statvfs.h>
64 #include <sys/mount.h>
65 #include <sys/sunddi.h>
66 #include <sys/bootconf.h>
67 #include <sys/policy.h>
68 
69 #include <vm/hat.h>
70 #include <vm/page.h>
71 #include <vm/pvn.h>
72 #include <vm/as.h>
73 #include <vm/seg.h>
74 #include <vm/seg_map.h>
75 #include <vm/seg_kmem.h>
76 #include <vm/seg_vn.h>
77 #include <vm/rm.h>
78 #include <vm/page.h>
79 #include <sys/swap.h>
80 
81 
82 #include <fs/fs_subr.h>
83 
84 
85 #include <sys/fs/udf_volume.h>
86 #include <sys/fs/udf_inode.h>
87 
88 int32_t ud_trace;
89 
90 /*
91  * HASH chains and mutex
92  */
93 extern union ihead ud_ihead[UD_HASH_SZ];
94 extern kmutex_t ud_icache_lock;
95 
96 
97 extern kmutex_t ud_sync_busy;
98 /*
99  * udf_vfs list manipulation routines
100  */
101 extern kmutex_t udf_vfs_mutex;
102 extern struct udf_vfs *udf_vfs_instances;
103 
104 /*
105  * Used to verify that a given entry on the udf_instances list (see below)
106  * still refers to a mounted file system.
107  *
108  * XXX: This is a crock that substitutes for proper locking to coordinate
109  *      updates to and uses of the entries in udf_instances.
110  */
111 struct check_node {
112 	struct vfs	*vfsp;
113 	struct udf_vfs	*udf_vfs;
114 	dev_t		vfs_dev;
115 };
116 
117 vfs_t *ud_still_mounted(struct check_node *);
118 void ud_checkclean(struct vfs *,
119 		struct udf_vfs *, dev_t, time_t);
120 int32_t ud_icheck(struct udf_vfs *);
121 void ud_flushi(int32_t);
122 
123 /*
124  * Link udf_vfsp in at the head of the list of udf_vfs_instances.
125  */
126 void
127 ud_vfs_add(struct udf_vfs *udf_vfsp)
128 {
129 	mutex_enter(&udf_vfs_mutex);
130 	udf_vfsp->udf_next = udf_vfs_instances;
131 	udf_vfs_instances = udf_vfsp;
132 	mutex_exit(&udf_vfs_mutex);
133 }
134 
135 /*
136  * Remove udf_vfsp from the list of udf_vfs_instances.
137  *
138  * Does no error checking; udf_vfsp is assumed to actually be on the list.
139  */
140 void
141 ud_vfs_remove(struct udf_vfs *udf_vfsp)
142 {
143 	struct udf_vfs **delpt = &udf_vfs_instances;
144 
145 	mutex_enter(&udf_vfs_mutex);
146 	for (; *delpt != NULL; delpt = &((*delpt)->udf_next)) {
147 		if (*delpt == udf_vfsp) {
148 			*delpt = udf_vfsp->udf_next;
149 			udf_vfsp->udf_next = NULL;
150 			break;
151 		}
152 	}
153 	mutex_exit(&udf_vfs_mutex);
154 }
155 
156 /*
157  * Search for the prn in the array
158  * of partitions and translate
159  * to the disk block number
160  */
161 daddr_t
162 ud_xlate_to_daddr(struct udf_vfs *udf_vfsp,
163 	uint16_t prn, uint32_t blkno, int32_t nblks, uint32_t *count)
164 {
165 	int32_t i;
166 	struct ud_map *map;
167 	struct ud_part *ud_parts;
168 	uint32_t lblkno, retblkno = 0, *addr;
169 	uint32_t begin_req, end_req;
170 	uint32_t begin_bad, end_bad;
171 
172 	ud_printf("ud_xlate_to_daddr\n");
173 
174 	/* Is prn valid */
175 	if (prn < udf_vfsp->udf_nmaps) {
176 		map = &(udf_vfsp->udf_maps[prn]);
177 
178 		if (map->udm_flags == UDM_MAP_VPM) {
179 			/*
180 			 * Map is Virtual Parition Map
181 			 * first check for the appropriate
182 			 * table and then return the converted
183 			 * block number
184 			 */
185 			for (i = 0; i < map->udm_nent; i++) {
186 				if (blkno < map->udm_count[i]) {
187 					addr = map->udm_addr[i];
188 					lblkno = SWAP_32(addr[blkno]);
189 					*count = 1;
190 					break;
191 				} else {
192 					blkno -= map->udm_count[i];
193 				}
194 			}
195 		} else if (map->udm_flags == UDM_MAP_SPM) {
196 			struct stbl *stbl;
197 			struct stbl_entry *te;
198 			int32_t entry_count;
199 
200 			/*
201 			 * Map type is Sparable Parition Map
202 			 * if the block is in the map
203 			 * return the translated block
204 			 * other wise use the regular
205 			 * partition stuff
206 			 */
207 			begin_req = blkno;
208 			end_req = begin_req + nblks;
209 
210 			stbl = (struct stbl *)map->udm_spaddr[0];
211 			te = (struct stbl_entry *)&stbl->stbl_entry;
212 			entry_count = SWAP_16(stbl->stbl_len);
213 
214 			for (i = 0; i < entry_count; i++, te++) {
215 				begin_bad = SWAP_32(te->sent_ol);
216 				end_bad = begin_bad + map->udm_plen;
217 
218 				/*
219 				 * Either unmapped or reserved
220 				 * or defective. need not consider
221 				 */
222 				if (begin_bad >= (uint32_t)0xFFFFFFF0) {
223 					continue;
224 				}
225 				if ((end_req < begin_bad) ||
226 					(begin_req >= end_bad)) {
227 					continue;
228 				}
229 
230 				if (begin_req < begin_bad) {
231 					ASSERT(end_req >= begin_bad);
232 					end_req = begin_bad;
233 				} else {
234 					retblkno = SWAP_32(te->sent_ml) +
235 						begin_req - begin_bad;
236 					if (end_req < end_bad) {
237 						*count = end_req - begin_req;
238 					} else {
239 						*count = end_bad - begin_req;
240 					}
241 					goto end;
242 				}
243 			}
244 
245 			lblkno = blkno;
246 			*count = end_req - begin_req;
247 		} else {
248 			/*
249 			 * regular partition
250 			 */
251 			lblkno = blkno;
252 			*count = nblks;
253 		}
254 		ud_parts = udf_vfsp->udf_parts;
255 		for (i = 0; i < udf_vfsp->udf_npart; i++) {
256 			if (map->udm_pn == ud_parts->udp_number) {
257 				/*
258 				 * Check if the block is inside
259 				 * the partition or not
260 				 */
261 				if (lblkno >= ud_parts->udp_length) {
262 					retblkno = 0;
263 				} else {
264 					retblkno = ud_parts->udp_start + lblkno;
265 				}
266 				goto end;
267 			}
268 			ud_parts ++;
269 		}
270 	}
271 
272 end:
273 	return (retblkno);
274 }
275 
276 #ifdef	UNDEF
277 uint32_t
278 ud_xlate_to_addr(struct udf_vfs *udf_vfsp,
279 	uint16_t prn, daddr_t blkno, int32_t lad)
280 {
281 	int32_t i;
282 	struct ud_part *ud_parts;
283 
284 	ud_printf("ud_xlate_to_addr\n");
285 
286 	if (lad == 0) {
287 		return (blkno);
288 	}
289 	ud_parts = udf_vfsp->udf_parts;
290 	for (i = 0; i < udf_vfsp->udf_npart; i++) {
291 		if (prn == ud_parts->udp_number) {
292 			return (blkno - ud_parts->udp_start);
293 		}
294 	}
295 	return (0);
296 }
297 #endif
298 
299 /*
300  * Directories do not have holes
301  */
302 int32_t
303 ud_ip_off2bno(struct ud_inode *ip, uint32_t offset, uint32_t *bno)
304 {
305 	int32_t i, error;
306 	struct icb_ext *iext;
307 
308 	ASSERT(ip->i_type == VDIR);
309 
310 	if (ip->i_desc_type == ICB_FLAG_ONE_AD) {
311 		*bno = ip->i_icb_block;
312 		return (0);
313 	}
314 
315 	if ((error = ud_read_icb_till_off(ip, (u_offset_t)offset)) != 0) {
316 		return (error);
317 	}
318 
319 	for (i = 0; i < ip->i_ext_used; i++) {
320 		iext = &ip->i_ext[i];
321 		if ((iext->ib_offset <= offset) &&
322 			(offset < (iext->ib_offset + iext->ib_count))) {
323 			*bno = iext->ib_block +
324 				((offset - iext->ib_offset) >>
325 				ip->i_udf->udf_l2b_shift);
326 			break;
327 		}
328 	}
329 	return (0);
330 }
331 
332 static uint32_t cum_sec[] = {
333 	0x0, 0x28de80, 0x4dc880, 0x76a700, 0x9e3400, 0xc71280,
334 	0xee9f80, 0x1177e00, 0x1405c80, 0x167e980, 0x190c800, 0x1b85500
335 };
336 static uint32_t cum_sec_leap[] = {
337 	0x0, 0x28de80, 0x4f1a00, 0x77f880, 0x9f8580, 0xc86400,
338 	0xeff100, 0x118cf80, 0x141ae00, 0x1693b00, 0x1921980, 0x1b9a680
339 };
340 
341 #define	SECS_PER_MIN	60
342 #define	SECS_PER_HOUR	3600
343 #define	DAYS_PER_YEAR	365
344 
345 #define	SEC_PER_DAY	0x15180
346 #define	SEC_PER_YEAR	0x1e13380
347 
348 
349 /* This holds good till yr 2100 */
350 void
351 ud_dtime2utime(struct timespec32 *utime,
352 	struct tstamp const *dtime)
353 {
354 	int16_t year, tzone;
355 	int32_t	sec;
356 	uint32_t *cp;
357 
358 	ud_printf("ud_dtime2utime\n");
359 
360 	year = SWAP_16(dtime->ts_year);
361 	cp = (year % 4) ? cum_sec : cum_sec_leap;
362 
363 	utime->tv_sec = cp[dtime->ts_month - 1];
364 	utime->tv_sec += (dtime->ts_day - 1) * SEC_PER_DAY;
365 	utime->tv_sec += ((dtime->ts_hour * 60) +
366 				dtime->ts_min) * 60 +
367 				dtime->ts_sec;
368 
369 	tzone = SWAP_16(dtime->ts_tzone);
370 	if ((tzone & TMODE) == 0x1000) {
371 		/* Local time */
372 		if ((tzone & TINVALID) != TINVALID) {
373 			if (tzone & TSIGN) {
374 				/*
375 				 * Sign extend the tzone
376 				 */
377 				sec = tzone | 0xFFFFF000;
378 			} else {
379 				sec = tzone & TOFFSET;
380 			}
381 			sec *= 60;
382 			utime->tv_sec -= sec;
383 		}
384 	}
385 
386 	utime->tv_nsec = ((((dtime->ts_csec * 100) +
387 					dtime->ts_husec) * 100) +
388 					dtime->ts_usec) * 1000;
389 	if (year >= 1970) {
390 		utime->tv_sec += (year - 1970) * SEC_PER_YEAR;
391 		utime->tv_sec += ((year - 1969) / 4) * SEC_PER_DAY;
392 	} else {
393 		utime->tv_sec = ((1970 - year) * SEC_PER_YEAR +
394 					((1972 - year) / 4) * SEC_PER_DAY -
395 					utime->tv_sec) * -1;
396 		if (utime->tv_nsec) {
397 			utime->tv_sec++;
398 			utime->tv_nsec = 1000 * 1000 * 1000 - utime->tv_nsec;
399 		}
400 	}
401 }
402 
403 void
404 ud_utime2dtime(struct timespec32 const *utime,
405 	struct tstamp *dtime)
406 {
407 	time32_t sec = utime->tv_sec;
408 	int32_t usec = utime->tv_nsec / 1000;
409 	uint32_t lyrs, nyrs, dummy;
410 	uint32_t *cp;
411 	int32_t before = 0;
412 
413 	ud_printf("ud_utime2dtime\n");
414 
415 	if (sec < 0) {
416 		before = 1;
417 		sec = sec * -1;
418 		if (usec) {
419 			sec = sec + 1;
420 			usec = 1000 * 1000 - usec;
421 		}
422 	}
423 
424 	dtime->ts_csec = usec / 10000;
425 	usec %= 10000;
426 	dtime->ts_husec = usec / 100;
427 	dtime->ts_usec = usec % 100;
428 
429 	nyrs = sec / SEC_PER_YEAR;
430 	if (before == 0) {
431 		lyrs = (nyrs + 1) / 4;
432 	} else {
433 		lyrs = (nyrs + 2) / 4;
434 	}
435 	if (nyrs != ((sec - (lyrs * SEC_PER_DAY)) / SEC_PER_YEAR)) {
436 		nyrs--;
437 		if (before == 0) {
438 			lyrs = (nyrs + 1) / 4;
439 		} else {
440 			lyrs = (nyrs + 2) / 4;
441 		}
442 	}
443 	sec -= nyrs * SEC_PER_YEAR + lyrs * SEC_PER_DAY;
444 
445 	if (before == 1) {
446 		nyrs = 1970 - nyrs;
447 		if (sec != 0) {
448 			nyrs --;
449 			if ((nyrs % 4) == 0) {
450 				sec = SEC_PER_YEAR + SEC_PER_DAY - sec;
451 			} else {
452 				sec = SEC_PER_YEAR - sec;
453 			}
454 		}
455 	} else {
456 		nyrs += 1970;
457 	}
458 	cp = (nyrs % 4) ? cum_sec : cum_sec_leap;
459 	dummy = sec / (SEC_PER_DAY * 29);
460 	if (dummy > 11) {
461 		dummy = 11;
462 	}
463 	if (sec < cp[dummy]) {
464 		dummy--;
465 	}
466 	dtime->ts_year = SWAP_16(nyrs);
467 	dtime->ts_month = dummy;
468 	sec -= cp[dtime->ts_month];
469 	dtime->ts_month++;
470 	dtime->ts_day = sec / SEC_PER_DAY;
471 	sec -= dtime->ts_day * SEC_PER_DAY;
472 	dtime->ts_day++;
473 	dtime->ts_hour = sec / SECS_PER_HOUR;
474 	sec -= dtime->ts_hour * SECS_PER_HOUR;
475 	dtime->ts_min = sec / SECS_PER_MIN;
476 	sec -= dtime->ts_min * SECS_PER_MIN;
477 	dtime->ts_sec = (uint8_t)sec;
478 
479 	/* GMT offset is 0 */
480 	dtime->ts_tzone = SWAP_16(0x1000);
481 }
482 
483 
484 int32_t
485 ud_syncip(struct ud_inode *ip, int32_t flags, int32_t waitfor)
486 {
487 	int32_t error;
488 	struct vnode *vp = ITOV(ip);
489 
490 	ud_printf("ud_syncip\n");
491 
492 	if (ip->i_udf == NULL) {
493 		return (0);
494 	}
495 
496 	if (!vn_has_cached_data(vp) || (vp->v_type == VCHR)) {
497 		error = 0;
498 	} else {
499 		rw_exit(&ip->i_contents);
500 		error = VOP_PUTPAGE(vp, (offset_t)0,
501 				(uint32_t)0, flags, CRED());
502 		rw_enter(&ip->i_contents, RW_WRITER);
503 	}
504 
505 	if (ip->i_flag & (IUPD |IACC | ICHG | IMOD)) {
506 		ud_iupdat(ip, waitfor);
507 	}
508 
509 	return (error);
510 }
511 
512 
513 /* ARGSUSED */
514 int32_t
515 ud_fbwrite(struct fbuf *fbp, struct ud_inode *ip)
516 {
517 	ud_printf("ud_fbwrite\n");
518 
519 	ASSERT(fbp != NULL);
520 
521 	return (fbwrite(fbp));
522 }
523 
524 
525 void
526 ud_sbwrite(struct udf_vfs *udf_vfsp)
527 {
528 	struct log_vol_int_desc *lvid;
529 	struct ud_part *ud_part;
530 	struct lvid_iu *iu;
531 	uint32_t *temp;
532 	int32_t i, c;
533 
534 	ud_printf("ud_sbwrite\n");
535 	ASSERT(udf_vfsp);
536 	ASSERT(MUTEX_HELD(&udf_vfsp->udf_lock));
537 
538 	/*
539 	 * updatable information in the superblock
540 	 * integrity type, udf_maxuniq, udf_nfiles, udf_ndirs
541 	 * udp_nfree in lvid
542 	 */
543 	lvid = (struct log_vol_int_desc *)udf_vfsp->udf_lvid;
544 	if (udf_vfsp->udf_clean == UDF_DIRTY) {
545 		lvid->lvid_int_type = SWAP_32(LOG_VOL_OPEN_INT);
546 	} else {
547 		lvid->lvid_int_type = SWAP_32(LOG_VOL_CLOSE_INT);
548 	}
549 	lvid->lvid_uniqid = SWAP_64(udf_vfsp->udf_maxuniq);
550 	temp = lvid->lvid_fst;
551 	c = SWAP_32(lvid->lvid_npart);
552 	ud_part = udf_vfsp->udf_parts;
553 	for (i = 0; i < c; i++) {
554 		temp[i] = SWAP_32(ud_part->udp_nfree);
555 		ud_part++;
556 	}
557 	iu = (struct lvid_iu *)(temp + c * 2);
558 	iu->lvidiu_nfiles = SWAP_32(udf_vfsp->udf_nfiles);
559 	iu->lvidiu_ndirs = SWAP_32(udf_vfsp->udf_ndirs);
560 
561 	ud_update_regid(&iu->lvidiu_regid);
562 
563 	ud_make_tag(udf_vfsp, &lvid->lvid_tag,
564 		UD_LOG_VOL_INT, udf_vfsp->udf_iseq_loc,
565 			sizeof (struct log_vol_int_desc) - 8 +
566 				8 * udf_vfsp->udf_npart +
567 				SWAP_32(lvid->lvid_liu));
568 
569 	/*
570 	 * Don't release the buffer after writing to the disk
571 	 */
572 	bwrite2(udf_vfsp->udf_iseq);
573 }
574 
575 
576 int32_t
577 ud_sync_indir(struct ud_inode *ip)
578 {
579 	int32_t elen;
580 
581 	ud_printf("ud_sync_indir\n");
582 
583 	if (ip->i_desc_type == ICB_FLAG_ONE_AD) {
584 		return (0);
585 	} else if (ip->i_desc_type == ICB_FLAG_SHORT_AD) {
586 		elen = sizeof (struct short_ad);
587 	} else if (ip->i_desc_type == ICB_FLAG_LONG_AD) {
588 		elen = sizeof (struct long_ad);
589 	} else {
590 		return (EINVAL);
591 	}
592 
593 	if (ip->i_astrat == STRAT_TYPE4) {
594 		int32_t ndentry;
595 
596 		ndentry = ip->i_max_emb / elen;
597 		if (ip->i_ext_used < ndentry) {
598 			return (0);
599 		}
600 		ASSERT(ip->i_con);
601 	} else {
602 		cmn_err(CE_WARN, "unsupported strategy type\n");
603 		return (EINVAL);
604 	}
605 
606 	return (0);
607 }
608 
609 void
610 ud_update(int32_t flag)
611 {
612 	struct vfs *vfsp;
613 	struct udf_vfs *udfsp, *udfsnext, *update_list = NULL;
614 	int32_t check_cnt = 0;
615 	size_t check_size;
616 	struct check_node *check_list, *ptr;
617 	time_t start_time;
618 
619 	ud_printf("ud_update\n");
620 
621 	mutex_enter(&ud_sync_busy);
622 	/*
623 	 * Examine all udf_vfs structures and add those that we can lock to the
624 	 * update list.  This is so that we don't hold the list lock for a
625 	 * long time.  If vfs_lock fails for a file system instance, then skip
626 	 * it because somebody is doing a unmount on it.
627 	 */
628 	mutex_enter(&udf_vfs_mutex);
629 	for (udfsp = udf_vfs_instances;
630 			udfsp != NULL; udfsp = udfsp->udf_next) {
631 		vfsp = udfsp->udf_vfs;
632 		if (vfs_lock(vfsp) != 0) {
633 			continue;
634 		}
635 		udfsp->udf_wnext = update_list;
636 		update_list = udfsp;
637 		check_cnt++;
638 	}
639 	mutex_exit(&udf_vfs_mutex);
640 
641 	if (update_list == NULL) {
642 		mutex_exit(&ud_sync_busy);
643 		return;
644 	}
645 
646 	check_size = sizeof (struct check_node) * check_cnt;
647 	check_list = ptr = kmem_alloc(check_size, KM_NOSLEEP);
648 
649 	/*
650 	 * Write back modified superblocks.
651 	 * Consistency check that the superblock of
652 	 * each file system is still in the buffer cache.
653 	 *
654 	 * Note that the update_list traversal is done without the protection
655 	 * of an overall list lock, so it's necessary to rely on the fact that
656 	 * each entry of the list is vfs_locked when moving from one entry to
657 	 * the next.  This works because a concurrent attempt to add an entry
658 	 * to another thread's update_list won't find it, since it'll already
659 	 * be locked.
660 	 */
661 	check_cnt = 0;
662 	for (udfsp = update_list; udfsp != NULL; udfsp = udfsnext) {
663 		/*
664 		 * Need to grab the next ptr before we unlock this one so
665 		 * another thread doesn't grab it and change it before we move
666 		 * on to the next vfs.  (Once we unlock it, it's ok if another
667 		 * thread finds it to add it to its own update_list; we don't
668 		 * attempt to refer to it through our list any more.)
669 		 */
670 		udfsnext = udfsp->udf_wnext;
671 		vfsp = udfsp->udf_vfs;
672 
673 		if (!vfsp->vfs_data) {
674 			vfs_unlock(vfsp);
675 			continue;
676 		}
677 		mutex_enter(&udfsp->udf_lock);
678 
679 		/*
680 		 * Build up the STABLE check list, so we can unlock the vfs
681 		 * until we do the actual checking.
682 		 */
683 		if (check_list != NULL) {
684 			if ((udfsp->udf_flags & UDF_FL_RDONLY) == 0) {
685 				ptr->vfsp = vfsp;
686 				ptr->udf_vfs = udfsp;
687 				ptr->vfs_dev = vfsp->vfs_dev;
688 				ptr++;
689 				check_cnt++;
690 			}
691 		}
692 
693 		/*
694 		 * superblock is not modified
695 		 */
696 		if (udfsp->udf_mod == 0) {
697 			mutex_exit(&udfsp->udf_lock);
698 			vfs_unlock(vfsp);
699 			continue;
700 		}
701 		if ((udfsp->udf_flags & UDF_FL_RDONLY) == 0) {
702 			mutex_exit(&udfsp->udf_lock);
703 			mutex_exit(&ud_sync_busy);
704 			cmn_err(CE_WARN, "update ro udfs mod\n");
705 			return;
706 		}
707 		udfsp->udf_mod = 0;
708 		mutex_exit(&udfsp->udf_lock);
709 
710 		ud_update_superblock(vfsp);
711 		vfs_unlock(vfsp);
712 	}
713 
714 	ud_flushi(flag);
715 	/*
716 	 * Force stale buffer cache information to be flushed,
717 	 * for all devices.  This should cause any remaining control
718 	 * information (e.g., inode info) to be flushed back.
719 	 */
720 	bflush((dev_t)NODEV);
721 
722 	if (check_list == NULL) {
723 		mutex_exit(&ud_sync_busy);
724 		return;
725 	}
726 
727 	/*
728 	 * For each udf filesystem in the STABLE check_list, update
729 	 * the clean flag if warranted.
730 	 */
731 	start_time = gethrestime_sec();
732 	for (ptr = check_list; check_cnt > 0; check_cnt--, ptr++) {
733 		/*
734 		 * ud_still_mounted() returns with vfsp and the vfs_reflock
735 		 * held if ptr refers to a vfs that is still mounted.
736 		 */
737 		if ((vfsp = ud_still_mounted(ptr)) == NULL) {
738 			continue;
739 		}
740 		ud_checkclean(vfsp, ptr->udf_vfs, ptr->vfs_dev, start_time);
741 		vfs_unlock(vfsp);
742 	}
743 	mutex_exit(&ud_sync_busy);
744 	kmem_free(check_list, check_size);
745 }
746 
747 
748 /*
749  * Returns vfsp and hold the lock if the vfs is still being mounted.
750  * Otherwise, returns 0.
751  *
752  * For our purposes, "still mounted" means that the file system still appears
753  * on the list of UFS file system instances.
754  */
755 vfs_t *
756 ud_still_mounted(struct check_node *checkp)
757 {
758 	struct vfs *vfsp;
759 	struct udf_vfs *udf_vfsp;
760 
761 	ud_printf("ud_still_mounted\n");
762 
763 	mutex_enter(&udf_vfs_mutex);
764 	for (udf_vfsp = udf_vfs_instances;
765 		udf_vfsp != NULL; udf_vfsp = udf_vfsp->udf_next) {
766 		if (udf_vfsp != checkp->udf_vfs) {
767 			continue;
768 		}
769 		/*
770 		 * Tentative match:  verify it and try to lock.  (It's not at
771 		 * all clear how the verification could fail, given that we've
772 		 * gotten this far.  We would have had to reallocate the
773 		 * ufsvfs struct at hand for a new incarnation; is that really
774 		 * possible in the interval from constructing the check_node
775 		 * to here?)
776 		 */
777 		vfsp = udf_vfsp->udf_vfs;
778 		if (vfsp != checkp->vfsp) {
779 			continue;
780 		}
781 		if (vfsp->vfs_dev != checkp->vfs_dev) {
782 			continue;
783 		}
784 		if (vfs_lock(vfsp) != 0) {
785 			continue;
786 		}
787 		mutex_exit(&udf_vfs_mutex);
788 		return (vfsp);
789 	}
790 	mutex_exit(&udf_vfs_mutex);
791 	return (NULL);
792 }
793 
794 /* ARGSUSED */
795 void
796 ud_checkclean(struct vfs *vfsp,
797 	struct udf_vfs *udf_vfsp, dev_t dev, time_t timev)
798 {
799 	ud_printf("ud_checkclean\n");
800 	udf_vfsp = (struct udf_vfs *)vfsp->vfs_data;
801 	/*
802 	 * ignore if buffers or inodes are busy
803 	 */
804 	if ((bcheck(dev, udf_vfsp->udf_iseq)) ||
805 		(ud_icheck(udf_vfsp))) {
806 		return;
807 	}
808 	mutex_enter(&udf_vfsp->udf_lock);
809 	ud_sbwrite(udf_vfsp);
810 	mutex_exit(&udf_vfsp->udf_lock);
811 }
812 
813 int32_t
814 ud_icheck(struct udf_vfs *udf_vfsp)
815 {
816 	int32_t index, error = 0;
817 	union ihead *ih;
818 	struct ud_inode *ip;
819 
820 	mutex_enter(&ud_icache_lock);
821 	for (index = 0; index < UD_HASH_SZ; index++) {
822 		ih = &ud_ihead[index];
823 		for (ip = ih->ih_chain[0];
824 			ip != (struct ud_inode *)ih; ip = ip->i_forw) {
825 			if ((ip->i_udf == udf_vfsp) &&
826 				((ip->i_flag & (IMOD|IUPD|ICHG)) ||
827 				(RW_ISWRITER(&ip->i_rwlock)) ||
828 				((ip->i_nlink <= 0) && (ip->i_flag & IREF)))) {
829 					error = 1;
830 					goto end;
831 			}
832 		}
833 	}
834 end:
835 	mutex_exit(&ud_icache_lock);
836 	return (error);
837 }
838 
839 void
840 ud_flushi(int32_t flag)
841 {
842 	struct ud_inode *ip, *lip;
843 	struct vnode *vp;
844 	int cheap = flag & SYNC_ATTR;
845 	int32_t index;
846 	union  ihead *ih;
847 
848 	/*
849 	 * Write back each (modified) inode,
850 	 * but don't sync back pages if vnode is
851 	 * part of the virtual swap device.
852 	 */
853 	mutex_enter(&ud_icache_lock);
854 	for (index = 0; index < UD_HASH_SZ; index++) {
855 		ih = &ud_ihead[index];
856 		lip = NULL;
857 
858 		for (ip = ih->ih_chain[0], lip = NULL;
859 				ip && ip != (struct ud_inode *)ih;
860 				ip = ip->i_forw) {
861 			int flag = ip->i_flag;
862 
863 			vp = ITOV(ip);
864 			/*
865 			 * Skip locked & inactive inodes.
866 			 * Skip vnodes w/ no cached data and no inode changes.
867 			 * Skip read-only vnodes
868 			 */
869 			if ((flag & IREF) == 0 ||
870 				(!vn_has_cached_data(vp) &&
871 				((flag & (IMOD|IACC|IUPD|ICHG)) == 0)) ||
872 				(vp->v_vfsp == NULL) || vn_is_readonly(vp)) {
873 				continue;
874 			}
875 
876 			if (!rw_tryenter(&ip->i_contents, RW_WRITER)) {
877 				continue;
878 			}
879 
880 			VN_HOLD(vp);
881 
882 			if (lip != NULL) {
883 				ITIMES(lip);
884 				VN_RELE(ITOV(lip));
885 			}
886 			lip = ip;
887 
888 			/*
889 			 * If this is an inode sync for file system hardening
890 			 * or this is a full sync but file is a swap file,
891 			 * don't sync pages but make sure the inode is up
892 			 * to date.  In other cases, push everything out.
893 			 */
894 			if (cheap || IS_SWAPVP(vp)) {
895 				ud_iupdat(ip, 0);
896 			} else {
897 				(void) ud_syncip(ip, B_ASYNC, I_SYNC);
898 			}
899 			rw_exit(&ip->i_contents);
900 		}
901 		if (lip != NULL) {
902 			ITIMES(lip);
903 			VN_RELE(ITOV(lip));
904 		}
905 	}
906 	mutex_exit(&ud_icache_lock);
907 }
908 
909 
910 void
911 ud_update_regid(struct regid *reg)
912 {
913 	ud_printf("ud_update_regid\n");
914 
915 	bzero(reg->reg_id, 23);
916 	(void) strncpy(reg->reg_id, SUN_IMPL_ID, SUN_IMPL_ID_LEN);
917 	reg->reg_ids[0] = SUN_OS_CLASS;
918 	reg->reg_ids[1] = SUN_OS_ID;
919 }
920 
921 /* ARGSUSED4 */
922 void
923 ud_make_tag(struct udf_vfs *udf_vfsp,
924 	struct tag *tag, uint16_t tag_id, uint32_t blkno, uint16_t crc_len)
925 {
926 	int32_t i;
927 	uint16_t crc;
928 	uint8_t *addr, cksum = 0;
929 
930 	ud_printf("ud_make_tag\n");
931 
932 	ASSERT(crc_len > 0x10);
933 	addr = (uint8_t *)tag;
934 	crc_len -= sizeof (struct tag);
935 	crc = ud_crc(addr + 0x10, crc_len);
936 
937 	tag->tag_id = SWAP_16(tag_id);
938 	tag->tag_desc_ver = SWAP_16(2);
939 	tag->tag_cksum = 0;
940 	tag->tag_res = 0;
941 	tag->tag_sno = SWAP_16(udf_vfsp->udf_tsno);
942 	tag->tag_crc = SWAP_16(crc);
943 
944 	tag->tag_crc_len = SWAP_16(crc_len);
945 	tag->tag_loc = SWAP_32(blkno);
946 
947 	addr = (uint8_t *)tag;
948 	for (i = 0; i <= 15; i++) {
949 		cksum += addr[i];
950 	}
951 	tag->tag_cksum = cksum;
952 }
953 
954 int32_t
955 ud_make_dev_spec_ear(struct dev_spec_ear *ds,
956 	major_t major, minor_t minor)
957 {
958 	int32_t attr_len;
959 
960 	ud_printf("ud_make_dev_spec_ear\n");
961 
962 	bzero(ds, sizeof (struct dev_spec_ear));
963 
964 	attr_len = sizeof (struct dev_spec_ear);
965 	ds->ds_atype = SWAP_32(12);
966 	ds->ds_astype = 1;
967 	ds->ds_attr_len = SWAP_32(attr_len);
968 	ds->ds_iu_len = 0;
969 	ds->ds_major_id = SWAP_32(major);
970 	ds->ds_minor_id = SWAP_32(minor);
971 
972 	return (attr_len);
973 }
974 
975 
976 int32_t
977 ud_get_next_fid(struct ud_inode *ip, struct fbuf **fbp, uint32_t offset,
978 	struct file_id **fid, uint8_t **name, uint8_t *buf)
979 {
980 	struct vnode *vp = ITOV(ip);
981 	caddr_t beg, end;
982 	int32_t error, lbsize, lbmask, sz, iulen, idlen, copied = 0;
983 	struct udf_vfs *udf_vfsp;
984 	uint8_t *obuf;
985 	int32_t count;
986 	uint32_t tbno;
987 	uint16_t crc_len;
988 	uint32_t len;
989 
990 	ud_printf("ud_get_next_fid\n");
991 
992 	obuf = buf;
993 	udf_vfsp = ip->i_udf;
994 	lbsize = udf_vfsp->udf_lbsize;
995 	lbmask = udf_vfsp->udf_lbmask;
996 
997 	if ((error = ud_ip_off2bno(ip, offset, &tbno)) != 0) {
998 		return (error);
999 	}
1000 	/* First time read */
1001 	if (*fbp == NULL) {
1002 		if ((error = fbread(vp, (offset_t)(offset & ~lbmask),
1003 		    lbsize, S_READ, fbp)) != 0) {
1004 			return (error);
1005 		}
1006 	}
1007 
1008 	end = (*fbp)->fb_addr + (*fbp)->fb_count;
1009 	beg = (*fbp)->fb_addr + (offset & lbmask);
1010 
1011 
1012 	if ((offset % lbsize) ||
1013 		(offset == 0)) {
1014 		sz = end - beg;
1015 	} else {
1016 		sz = 0;
1017 	}
1018 
1019 
1020 	if (F_LEN <= sz) {
1021 		*fid = (struct file_id *)beg;
1022 		beg += F_LEN;
1023 	} else {
1024 		copied = 1;
1025 		bcopy(beg, buf, sz);
1026 		fbrelse(*fbp, S_OTHER);
1027 		*fbp = NULL;
1028 
1029 		/* Skip to next block */
1030 		if (offset & lbmask) {
1031 			offset = (offset & ~lbmask) + lbsize;
1032 		}
1033 		if ((error = fbread(vp, (offset_t)offset,
1034 		    lbsize, S_READ, fbp)) != 0) {
1035 			return (error);
1036 		}
1037 		end = (*fbp)->fb_addr + (*fbp)->fb_count;
1038 		beg = (*fbp)->fb_addr;
1039 
1040 		bcopy(beg, buf + sz, F_LEN - sz);
1041 		beg = beg + F_LEN - sz;
1042 		*fid = (struct file_id *)buf;
1043 
1044 		buf += F_LEN;
1045 	}
1046 
1047 
1048 	/*
1049 	 * Check if this a valid file_identifier
1050 	 */
1051 	if (ud_verify_tag_and_desc(&(*fid)->fid_tag, UD_FILE_ID_DESC,
1052 	    tbno, 0, lbsize) != 0) {
1053 		/*
1054 		 * Either end of directory or corrupted
1055 		 */
1056 		return (EINVAL);
1057 	}
1058 
1059 	crc_len = SWAP_16((*fid)->fid_tag.tag_crc_len);
1060 	if (crc_len > udf_vfsp->udf_lbsize) {
1061 		/*
1062 		 * Entries cannot be larger than
1063 		 * blocksize
1064 		 */
1065 		return (EINVAL);
1066 	}
1067 
1068 	if (crc_len < (F_LEN - sizeof (struct tag))) {
1069 		iulen = SWAP_16((*fid)->fid_iulen);
1070 		idlen = FID_LEN(*fid) - F_LEN;
1071 		goto use_id_iu_len;
1072 	}
1073 
1074 	/*
1075 	 * By now beg points to the start fo the file name
1076 	 */
1077 
1078 	sz = end - beg;
1079 	len = crc_len + sizeof (struct tag) - (F_LEN);
1080 	if (len <= sz) {
1081 		if (copied == 1) {
1082 			bcopy(beg, buf, len);
1083 			buf += len;
1084 		}
1085 		beg += len;
1086 	} else {
1087 		copied = 1;
1088 		/*
1089 		 * We are releasing the
1090 		 * old buffer so copy fid to buf
1091 		 */
1092 		if (obuf == buf) {
1093 			count = F_LEN + sz;
1094 			bcopy(*fid, buf, count);
1095 			*fid = (struct file_id *)buf;
1096 			buf += count;
1097 		} else {
1098 			bcopy(beg, buf, sz);
1099 			*fid = (struct file_id *)buf;
1100 			buf += sz;
1101 		}
1102 		fbrelse(*fbp, S_OTHER);
1103 		*fbp = NULL;
1104 
1105 		/* Skip to next block */
1106 		if (offset & lbmask) {
1107 			offset = (offset & ~lbmask) + lbsize;
1108 		}
1109 		if ((error = fbread(vp, (offset_t)offset,
1110 		    lbsize, S_READ, fbp)) != 0) {
1111 			return (error);
1112 		}
1113 		end = (*fbp)->fb_addr + (*fbp)->fb_count;
1114 		beg = (*fbp)->fb_addr;
1115 		count = len - sz;
1116 		bcopy(beg, buf, count);
1117 		beg += count;
1118 	}
1119 
1120 	/*
1121 	 * First we verify that the tag id and the FID_LEN are valid.
1122 	 * Next we verify the crc of the descriptor.
1123 	 */
1124 	if (ud_verify_tag_and_desc(&(*fid)->fid_tag, UD_FILE_ID_DESC,
1125 	    tbno, 0, lbsize) != 0) {
1126 		/* directory is corrupted */
1127 		return (EINVAL);
1128 	}
1129 	if (ud_verify_tag_and_desc(&(*fid)->fid_tag, UD_FILE_ID_DESC,
1130 	    tbno, 1, FID_LEN(*fid)) != 0) {
1131 		/* directory is corrupted */
1132 		return (EINVAL);
1133 	}
1134 
1135 	idlen = FID_LEN(*fid);
1136 
1137 	idlen -= F_LEN;
1138 	iulen = SWAP_16((*fid)->fid_iulen);
1139 	if (crc_len < (F_LEN - sizeof (struct tag) + idlen)) {
1140 use_id_iu_len:
1141 		len = (F_LEN - sizeof (struct tag) + idlen) - crc_len;
1142 		sz = end - beg;
1143 		if (len <= sz) {
1144 			if (copied == 1) {
1145 				bcopy(beg, buf, len);
1146 			}
1147 		} else {
1148 			if (obuf == buf) {
1149 				count = crc_len + sizeof (struct tag);
1150 				bcopy(*fid, buf, count);
1151 				*fid = (struct file_id *)buf;
1152 				buf += count;
1153 			} else {
1154 				bcopy(beg, buf, sz);
1155 				*fid = (struct file_id *)buf;
1156 				buf += sz;
1157 			}
1158 			fbrelse(*fbp, S_OTHER);
1159 			*fbp = NULL;
1160 
1161 			/* Skip to next block */
1162 			if (offset & lbmask) {
1163 				offset = (offset & ~lbmask) + lbsize;
1164 			}
1165 			if ((error = fbread(vp, (offset_t)offset,
1166 			    lbsize, S_READ, fbp)) != 0) {
1167 				return (error);
1168 			}
1169 			end = (*fbp)->fb_addr + (*fbp)->fb_count;
1170 			beg = (*fbp)->fb_addr;
1171 			count = len - sz;
1172 			bcopy(beg, buf, count);
1173 			beg += count;
1174 		}
1175 	}
1176 
1177 	*name = ((uint8_t *)*fid) + F_LEN + iulen;
1178 
1179 	return (0);
1180 }
1181 
1182 
1183 int32_t
1184 ud_verify_tag_and_desc(struct tag *tag, uint16_t id, uint32_t blockno,
1185 			int32_t verify_desc, int32_t desc_len)
1186 {
1187 	int32_t i;
1188 	uint8_t *addr, cksum = 0;
1189 	uint16_t crc;
1190 	file_entry_t	*fe;
1191 	struct ext_attr_hdr *eah;
1192 	struct file_id	*fid;
1193 	int32_t fidlen, ea_off;
1194 
1195 	if (tag->tag_id != SWAP_16(id)) {
1196 		return (1);
1197 	}
1198 	addr = (uint8_t *)tag;
1199 	eah = (struct ext_attr_hdr *)tag;
1200 	for (i = 0; i < 4; i++) {
1201 		cksum += addr[i];
1202 	}
1203 	for (i = 5; i <= 15; i++) {
1204 		cksum += addr[i];
1205 	}
1206 	if (cksum != tag->tag_cksum) {
1207 		cmn_err(CE_NOTE,
1208 		"Checksum Does not Verify TAG %x CALC %x blockno 0x%x\n",
1209 		    tag->tag_cksum, cksum, blockno);
1210 		return (1);
1211 	}
1212 	/*
1213 	 * Validate the meta data for UD_FILE_ID_DESC.
1214 	 * The FID_LEN should not exceed the desc_len.
1215 	 * This validation is done before the entire descriptor is read.
1216 	 * A call to this routine is made initially with verify_desc set as 0
1217 	 * but a non zero value in desc_len.
1218 	 */
1219 	if (id == UD_FILE_ID_DESC) {
1220 		fid = (struct file_id *)tag;
1221 		fidlen = FID_LEN(fid);
1222 		if (fidlen > desc_len) {
1223 			cmn_err(CE_NOTE,
1224 	"Invalid FID_LEN(0x%x). Greater than expected(0x%x) blockno 0x%x\n",
1225 			    fidlen, desc_len, blockno);
1226 				return (1);
1227 		}
1228 	}
1229 	if (verify_desc == 0)
1230 		return (0);
1231 	/*
1232 	 * We are done verifying the tag. We proceed with verifying the
1233 	 * the descriptor. desc_len indicates the size of the structure
1234 	 * pointed to by argument tag. It includes the size of struct tag.
1235 	 * We first check the tag_crc_len since we use this to compute the
1236 	 * crc of the descriptor.
1237 	 * Verifying the crc is normally sufficient to ensure the integrity
1238 	 * of the meta data in the descriptor. However given the paranoia
1239 	 * about the panic caused by illegal meta data values we do an
1240 	 * additional check of the meta data for decriptor UD_FILE_ENTRY.
1241 	 * (The original panic was caused because this routine was not called
1242 	 * to verify the integrity of the tag and descriptor.)
1243 	 */
1244 	if (SWAP_16(tag->tag_crc_len) > (desc_len - sizeof (struct tag))) {
1245 		cmn_err(CE_NOTE,
1246 	"tag_crc_len(0x%x) is greater than expected len(0x%x) blockno 0x%x\n",
1247 		    SWAP_16(tag->tag_crc_len),
1248 		    desc_len, blockno);
1249 		return (1);
1250 	}
1251 	if (tag->tag_crc_len) {
1252 		crc = ud_crc(addr + 0x10, SWAP_16(tag->tag_crc_len));
1253 		if (crc != SWAP_16(tag->tag_crc)) {
1254 			cmn_err(CE_NOTE, "CRC mismatch TAG_ID 0x%x TAG_CRC 0x%x"
1255 			" Computed crc 0x%x tag_loc %x blockno 0x%x\n",
1256 			    id, SWAP_16(tag->tag_crc), crc,
1257 			    SWAP_32(tag->tag_loc), blockno);
1258 			return (1);
1259 		}
1260 	}
1261 	switch (id) {
1262 		case UD_FILE_ENTRY:
1263 			fe = (file_entry_t *)tag;
1264 			if ((offsetof(struct file_entry, fe_spec) +
1265 			    SWAP_32(fe->fe_len_ear) +
1266 			    SWAP_32(fe->fe_len_adesc)) > desc_len) {
1267 				cmn_err(CE_NOTE,
1268 	"fe_len_ear(0x%x) fe_len_adesc(0x%x) fields are not OK. blockno 0x%x\n",
1269 				    SWAP_32(fe->fe_len_ear),
1270 				    SWAP_32(fe->fe_len_adesc),
1271 				    blockno);
1272 				return (1);
1273 			}
1274 			break;
1275 		case UD_EXT_ATTR_HDR:
1276 			eah = (struct ext_attr_hdr *)tag;
1277 			if (SWAP_32(eah->eah_aal) > desc_len) {
1278 				cmn_err(CE_NOTE,
1279 			"eah_all(0x%x) exceeds desc. len(0x%x) blockno 0x%x\n",
1280 				    SWAP_32(eah->eah_aal), desc_len, blockno);
1281 			    return (1);
1282 			}
1283 			ea_off = GET_32(&eah->eah_ial);
1284 			if (ea_off >= desc_len) {
1285 				cmn_err(CE_NOTE,
1286 		"ea_off(0x%x) is not less than ea_len(0x%x) blockno 0x%x\n",
1287 				    ea_off, desc_len, blockno);
1288 				return (1);
1289 			}
1290 			break;
1291 		default:
1292 			break;
1293 	}
1294 	if (SWAP_32(blockno) != tag->tag_loc) {
1295 		cmn_err(CE_NOTE,
1296 			"Tag Location mismatch blockno %x tag_blockno %x\n",
1297 			blockno, SWAP_32(tag->tag_loc));
1298 		return (1);
1299 	}
1300 	return (0);
1301 }
1302 
1303 /* **************** udf specific subroutines *********************** */
1304 
1305 uint16_t ud_crc_table[256] = {
1306 	0x0000, 0x1021, 0x2042, 0x3063, 0x4084, 0x50A5, 0x60C6, 0x70E7,
1307 	0x8108, 0x9129, 0xA14A, 0xB16B, 0xC18C, 0xD1AD, 0xE1CE, 0xF1EF,
1308 	0x1231, 0x0210, 0x3273, 0x2252, 0x52B5, 0x4294, 0x72F7, 0x62D6,
1309 	0x9339, 0x8318, 0xB37B, 0xA35A, 0xD3BD, 0xC39C, 0xF3FF, 0xE3DE,
1310 	0x2462, 0x3443, 0x0420, 0x1401, 0x64E6, 0x74C7, 0x44A4, 0x5485,
1311 	0xA56A, 0xB54B, 0x8528, 0x9509, 0xE5EE, 0xF5CF, 0xC5AC, 0xD58D,
1312 	0x3653, 0x2672, 0x1611, 0x0630, 0x76D7, 0x66F6, 0x5695, 0x46B4,
1313 	0xB75B, 0xA77A, 0x9719, 0x8738, 0xF7DF, 0xE7FE, 0xD79D, 0xC7BC,
1314 	0x48C4, 0x58E5, 0x6886, 0x78A7, 0x0840, 0x1861, 0x2802, 0x3823,
1315 	0xC9CC, 0xD9ED, 0xE98E, 0xF9AF, 0x8948, 0x9969, 0xA90A, 0xB92B,
1316 	0x5AF5, 0x4AD4, 0x7AB7, 0x6A96, 0x1A71, 0x0A50, 0x3A33, 0x2A12,
1317 	0xDBFD, 0xCBDC, 0xFBBF, 0xEB9E, 0x9B79, 0x8B58, 0xBB3B, 0xAB1A,
1318 	0x6CA6, 0x7C87, 0x4CE4, 0x5CC5, 0x2C22, 0x3C03, 0x0C60, 0x1C41,
1319 	0xEDAE, 0xFD8F, 0xCDEC, 0xDDCD, 0xAD2A, 0xBD0B, 0x8D68, 0x9D49,
1320 	0x7E97, 0x6EB6, 0x5ED5, 0x4EF4, 0x3E13, 0x2E32, 0x1E51, 0x0E70,
1321 	0xFF9F, 0xEFBE, 0xDFDD, 0xCFFC, 0xBF1B, 0xAF3A, 0x9F59, 0x8F78,
1322 	0x9188, 0x81A9, 0xB1CA, 0xA1EB, 0xD10C, 0xC12D, 0xF14E, 0xE16F,
1323 	0x1080, 0x00A1, 0x30C2, 0x20E3, 0x5004, 0x4025, 0x7046, 0x6067,
1324 	0x83B9, 0x9398, 0xA3FB, 0xB3DA, 0xC33D, 0xD31C, 0xE37F, 0xF35E,
1325 	0x02B1, 0x1290, 0x22F3, 0x32D2, 0x4235, 0x5214, 0x6277, 0x7256,
1326 	0xB5EA, 0xA5CB, 0x95A8, 0x8589, 0xF56E, 0xE54F, 0xD52C, 0xC50D,
1327 	0x34E2, 0x24C3, 0x14A0, 0x0481, 0x7466, 0x6447, 0x5424, 0x4405,
1328 	0xA7DB, 0xB7FA, 0x8799, 0x97B8, 0xE75F, 0xF77E, 0xC71D, 0xD73C,
1329 	0x26D3, 0x36F2, 0x0691, 0x16B0, 0x6657, 0x7676, 0x4615, 0x5634,
1330 	0xD94C, 0xC96D, 0xF90E, 0xE92F, 0x99C8, 0x89E9, 0xB98A, 0xA9AB,
1331 	0x5844, 0x4865, 0x7806, 0x6827, 0x18C0, 0x08E1, 0x3882, 0x28A3,
1332 	0xCB7D, 0xDB5C, 0xEB3F, 0xFB1E, 0x8BF9, 0x9BD8, 0xABBB, 0xBB9A,
1333 	0x4A75, 0x5A54, 0x6A37, 0x7A16, 0x0AF1, 0x1AD0, 0x2AB3, 0x3A92,
1334 	0xFD2E, 0xED0F, 0xDD6C, 0xCD4D, 0xBDAA, 0xAD8B, 0x9DE8, 0x8DC9,
1335 	0x7C26, 0x6C07, 0x5C64, 0x4C45, 0x3CA2, 0x2C83, 0x1CE0, 0x0CC1,
1336 	0xEF1F, 0xFF3E, 0xCF5D, 0xDF7C, 0xAF9B, 0xBFBA, 0x8FD9, 0x9FF8,
1337 	0x6E17, 0x7E36, 0x4E55, 0x5E74, 0x2E93, 0x3EB2, 0x0ED1, 0x1EF0
1338 };
1339 
1340 uint16_t
1341 ud_crc(uint8_t *addr, int32_t len)
1342 {
1343 	uint16_t crc = 0;
1344 
1345 	while (len-- > 0) {
1346 		crc = ud_crc_table[(crc >> 8 ^ *addr++) & 0xff] ^ (crc<<8);
1347 	}
1348 
1349 	return (crc);
1350 }
1351 
1352 typedef unsigned short unicode_t;
1353 
1354 #define	POUND		0x0023
1355 #define	DOT		0x002E
1356 #define	SLASH		0x002F
1357 #define	UNDERBAR	0x005F
1358 
1359 
1360 static uint16_t htoc[16] = {'0', '1', '2', '3',
1361 	'4', '5', '6', '7', '8', '9', 'a', 'b', 'c', 'd', 'e', 'f'};
1362 
1363 
1364 /*
1365  * An unrecorded block will return all
1366  * 0's on a WORM media. to simulate
1367  * a unrecorded block on a rw media
1368  * we fill it with all zero's
1369  *	return 0 : If unrecorded
1370  *	return 1 : If recorded.
1371  */
1372 uint32_t
1373 ud_check_te_unrec(struct udf_vfs *udf_vfsp, caddr_t addr, uint32_t blkno)
1374 {
1375 	int32_t i, lbsize;
1376 	struct term_entry *te;
1377 
1378 	ASSERT(udf_vfsp);
1379 	ASSERT(addr);
1380 
1381 	te = (struct term_entry *)addr;
1382 	if (ud_verify_tag_and_desc(&te->te_tag, UD_TERMINAL_ENT,
1383 	    blkno, 1, udf_vfsp->udf_lbsize) != 0) {
1384 		lbsize = udf_vfsp->udf_lbsize;
1385 		for (i = 0; i < lbsize; i++) {
1386 			if (addr[i] != 0) {
1387 				return (1);
1388 			}
1389 		}
1390 	}
1391 	return (0);
1392 }
1393 
1394 
1395 /*
1396  * The algorithms ud_utf82utf16 and ud_utf162utf8
1397  * donot handle surrogates. This is unicode 1.1 as I
1398  * understand. When writing udf2.0 this code has
1399  * to be changed to process surrogates also
1400  * (Dont ask me what is a surrogate character)
1401  */
1402 
1403 /*
1404  * This will take a utf8 string convert the first character
1405  * to utf16 and return the number of bytes consumed in this
1406  * process. A 0 will be returned if the character is invalid
1407  */
1408 uint8_t bytes_from_utf8[] = {
1409 0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,
1410 0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,
1411 0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,
1412 0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,
1413 0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,
1414 0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,
1415 0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,
1416 0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,
1417 0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,
1418 0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,
1419 0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,
1420 0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,  0, 0, 0, 0,
1421 1, 1, 1, 1,  1, 1, 1, 1,  1, 1, 1, 1,  1, 1, 1, 1,
1422 1, 1, 1, 1,  1, 1, 1, 1,  1, 1, 1, 1,  1, 1, 1, 1,
1423 2, 2, 2, 2,  2, 2, 2, 2,  2, 2, 2, 2,  2, 2, 2, 2,
1424 3, 3, 3, 3,  3, 3, 3, 3,  4, 4, 4, 4,  5, 5, 5, 5
1425 };
1426 int32_t
1427 ud_utf82utf16(uint8_t *s_8, uint16_t *c_16, int32_t count)
1428 {
1429 	int32_t extra_bytes;
1430 	uint32_t c_32;
1431 	ASSERT(s_8);
1432 	ASSERT(c_16);
1433 
1434 	/*
1435 	 * First convert to a 32-bit
1436 	 * character
1437 	 */
1438 	c_32 = 0;
1439 	extra_bytes = bytes_from_utf8[*s_8];
1440 	if (extra_bytes > count) {
1441 		return (0);
1442 	}
1443 
1444 	/*
1445 	 * verify if the string is a valid
1446 	 * utf8 string
1447 	 */
1448 	if (extra_bytes == 0) {
1449 		/*
1450 		 * Apply one byte rule
1451 		 */
1452 		if (*s_8 & 0x80) {
1453 			return (0);
1454 		}
1455 		c_32 = *s_8 & 0x7F;
1456 	} else if (extra_bytes == 1) {
1457 		if (((*s_8 & 0xE0) != 0xC0) ||
1458 			((*(s_8 + 1) & 0xC0) != 0x80)) {
1459 			return (0);
1460 		}
1461 		c_32 = *s_8 & 0x1F;
1462 	} else if (extra_bytes == 2) {
1463 		if (((*s_8 & 0xF0) != 0xE0) ||
1464 			((*(s_8 + 1) & 0xC0) != 0x80) ||
1465 			((*(s_8 + 2) & 0xC0) != 0x80)) {
1466 			return (0);
1467 		}
1468 		c_32 = *s_8 & 0x0F;
1469 	} else if (extra_bytes == 3) {
1470 		if (((*s_8 & 0xF8) != 0xF0) ||
1471 			((*(s_8 + 1) & 0xC0) != 0x80) ||
1472 			((*(s_8 + 2) & 0xC0) != 0x80) ||
1473 			((*(s_8 + 3) & 0xC0) != 0x80)) {
1474 			return (0);
1475 		}
1476 		c_32 = *s_8 & 0x07;
1477 	} else if (extra_bytes == 4) {
1478 		if (((*s_8 & 0xFC) != 0xF8) ||
1479 			((*(s_8 + 1) & 0xC0) != 0x80) ||
1480 			((*(s_8 + 2) & 0xC0) != 0x80) ||
1481 			((*(s_8 + 3) & 0xC0) != 0x80) ||
1482 			((*(s_8 + 4) & 0xC0) != 0x80)) {
1483 			return (0);
1484 		}
1485 		c_32 = *s_8 & 0x03;
1486 	} else if (extra_bytes == 5) {
1487 		if (((*s_8 & 0xFE) != 0xFC) ||
1488 			((*(s_8 + 1) & 0xC0) != 0x80) ||
1489 			((*(s_8 + 2) & 0xC0) != 0x80) ||
1490 			((*(s_8 + 3) & 0xC0) != 0x80) ||
1491 			((*(s_8 + 4) & 0xC0) != 0x80) ||
1492 			((*(s_8 + 5) & 0xC0) != 0x80)) {
1493 			return (0);
1494 		}
1495 		c_32 = *s_8 & 0x01;
1496 	} else {
1497 		return (0);
1498 	}
1499 	s_8++;
1500 
1501 	/*
1502 	 * Convert to 32-bit character
1503 	 */
1504 	switch (extra_bytes) {
1505 		case 5 :
1506 			c_32 <<= 6;
1507 			c_32 += (*s_8++ & 0x3F);
1508 			/* FALLTHROUGH */
1509 		case 4 :
1510 			c_32 <<= 6;
1511 			c_32 += (*s_8++ & 0x3F);
1512 			/* FALLTHROUGH */
1513 		case 3 :
1514 			c_32 <<= 6;
1515 			c_32 += (*s_8++ & 0x3F);
1516 			/* FALLTHROUGH */
1517 		case 2 :
1518 			c_32 <<= 6;
1519 			c_32 += (*s_8++ & 0x3F);
1520 			/* FALLTHROUGH */
1521 		case 1 :
1522 			c_32 <<= 6;
1523 			c_32 += (*s_8++ & 0x3F);
1524 			/* FALLTHROUGH */
1525 		case 0 :
1526 			break;
1527 	}
1528 
1529 	/*
1530 	 * now convert the 32-bit
1531 	 * character into a 16-bit character
1532 	 */
1533 	*c_16 = c_32;
1534 	return (extra_bytes + 1);
1535 }
1536 
1537 /*
1538  * Convert to a form that can be put on the media
1539  * out_len has the size of out_str when we are called.
1540  * This routine will set out_len to actual bytes written to out_str.
1541  * We make sure that we will not attempt to write beyond the out_str_len.
1542  */
1543 int32_t
1544 ud_compress(int32_t in_len, int32_t *out_len,
1545 		uint8_t *in_str, uint8_t *out_str)
1546 {
1547 	int32_t error, in_index, out_index, index, c_tx_sz, out_str_len;
1548 	uint16_t w2_char, *w2_str;
1549 	uint8_t comp_id;
1550 
1551 	out_str_len = *out_len;
1552 	if (in_len > (out_str_len - 2)) {
1553 		return (ENAMETOOLONG);
1554 	}
1555 
1556 	*out_len = 0;
1557 	w2_str = (uint16_t *)kmem_zalloc(512, KM_SLEEP);
1558 
1559 	error = in_index = out_index = c_tx_sz = 0;
1560 	comp_id = 8;
1561 	for (in_index = 0; in_index < in_len; in_index += c_tx_sz) {
1562 		if ((c_tx_sz = ud_utf82utf16(&in_str[in_index],
1563 			&w2_char, in_len - in_index)) == 0) {
1564 			error = EINVAL;
1565 			goto end;
1566 		}
1567 		/*
1568 		 * utf-8 characters can be
1569 		 * of 1 - 6 bytes in length
1570 		 */
1571 		ASSERT(c_tx_sz > 0);
1572 		ASSERT(c_tx_sz < 7);
1573 		if ((comp_id == 8) && (w2_char & 0xff00)) {
1574 			comp_id = 0x10;
1575 		}
1576 		w2_str[out_index++] = w2_char;
1577 	}
1578 	if (((comp_id == 0x10) && (out_index > ((out_str_len - 2)/2))) ||
1579 		((comp_id == 0x8) && (out_index > (out_str_len - 2)))) {
1580 		error = ENAMETOOLONG;
1581 		goto end;
1582 	}
1583 
1584 	in_index = out_index;
1585 	out_index = 0;
1586 	out_str[out_index++] = comp_id;
1587 	for (index = 0; index < in_index; index++) {
1588 		if (comp_id == 0x10) {
1589 			out_str[out_index++] = (w2_str[index] & 0xFF00) >> 8;
1590 		}
1591 		out_str[out_index++] = w2_str[index] & 0xFF;
1592 	}
1593 	ASSERT(out_index <= (out_str_len - 1));
1594 	*out_len = out_index;
1595 end:
1596 	if (w2_str != NULL) {
1597 		kmem_free((caddr_t)w2_str, 512);
1598 	}
1599 	return (error);
1600 }
1601 
1602 /*
1603  * Take a utf16 character and convert
1604  * it into a utf8 character.
1605  * A 0 will be returned if the conversion fails
1606  */
1607 uint8_t first_byte_mark[7] = { 0x00, 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC};
1608 int32_t
1609 ud_utf162utf8(uint16_t c_16, uint8_t *s_8)
1610 {
1611 	int32_t nc;
1612 	uint32_t c_32;
1613 	uint32_t byte_mask = 0xBF;
1614 	uint32_t byte_mark = 0x80;
1615 
1616 	ASSERT(s_8);
1617 
1618 	/*
1619 	 * Convert the 16-bit character to
1620 	 * a 32-bit character
1621 	 */
1622 	c_32 = c_16;
1623 
1624 	/*
1625 	 * By here the 16-bit character is converted
1626 	 * to a 32-bit wide character
1627 	 */
1628 	if (c_32 < 0x80) {
1629 		nc = 1;
1630 	} else if (c_32 < 0x800) {
1631 		nc = 2;
1632 	} else if (c_32 < 0x10000) {
1633 		nc = 3;
1634 	} else if (c_32 < 0x200000) {
1635 		nc = 4;
1636 	} else if (c_32 < 0x4000000) {
1637 		nc = 5;
1638 	} else if (c_32 < (uint32_t)0x80000000) {
1639 		nc = 6;
1640 	} else {
1641 		nc = 0;
1642 	}
1643 	s_8 += nc;
1644 	switch (nc) {
1645 		case 6 :
1646 			*(--s_8) = (c_32 | byte_mark)  & byte_mask;
1647 			c_32 >>= 6;
1648 			/* FALLTHROUGH */
1649 		case 5 :
1650 			*(--s_8) = (c_32 | byte_mark)  & byte_mask;
1651 			c_32 >>= 6;
1652 			/* FALLTHROUGH */
1653 		case 4 :
1654 			*(--s_8) = (c_32 | byte_mark)  & byte_mask;
1655 			c_32 >>= 6;
1656 			/* FALLTHROUGH */
1657 		case 3 :
1658 			*(--s_8) = (c_32 | byte_mark)  & byte_mask;
1659 			c_32 >>= 6;
1660 			/* FALLTHROUGH */
1661 		case 2 :
1662 			*(--s_8) = (c_32 | byte_mark)  & byte_mask;
1663 			c_32 >>= 6;
1664 			/* FALLTHROUGH */
1665 		case 1 :
1666 			*(--s_8) = c_32 | first_byte_mark[nc];
1667 	}
1668 	return (nc);
1669 }
1670 
1671 /*
1672  * Convert to a form that can be transfered to the user
1673  * Assumption's
1674  * in_length < 256, out_str is atleast 255 bytes long
1675  * The converted byte stream length is returned in out_len
1676  */
1677 #define	MAX_ALLOWABLE_STRING 250
1678 
1679 int32_t
1680 ud_uncompress(int32_t in_len, int32_t *out_len,
1681 		uint8_t *in_str, uint8_t *out_str)
1682 {
1683 	uint8_t comp_id, utf8[6];
1684 	uint16_t w2_char, crc;
1685 	int32_t error, index, c_tx_sz, len_till_now;
1686 	int32_t make_crc, lic, dot_loc, crc_start_loc = 0, k = 0;
1687 
1688 	if (in_len == 0) {
1689 		*out_len = 0;
1690 		out_str[0] = '\0';
1691 		return (0);
1692 	}
1693 
1694 	error = len_till_now = make_crc = 0;
1695 	dot_loc = lic = -2;
1696 	*out_len = 0;
1697 	crc = 0;
1698 	comp_id = in_str[0];
1699 
1700 	/*
1701 	 * File names "." and ".." are invalid under unix.
1702 	 * Transform them into something
1703 	 */
1704 	if (comp_id == 8) {
1705 		if ((in_str[1] == DOT) &&
1706 			((in_len == 2) || ((in_len == 3) &&
1707 				(in_str[2] == DOT)))) {
1708 			out_str[k++] = UNDERBAR;
1709 			len_till_now = 1;
1710 			goto make_append_crc;
1711 		}
1712 	} else if (comp_id == 0x10) {
1713 		if (((in_str[1] << 8 | in_str[2]) == DOT) &&
1714 			((in_len == 3) || ((in_len == 5) &&
1715 				((in_str[3] << 8 | in_str[4]) == DOT)))) {
1716 			out_str[k++] = UNDERBAR;
1717 			len_till_now = 1;
1718 			goto make_append_crc;
1719 		}
1720 	} else {
1721 		*out_len = 0;
1722 		return (EINVAL);
1723 	}
1724 
1725 	for (index = 1; index < in_len; ) {
1726 
1727 		/*
1728 		 * Uncompress each character
1729 		 */
1730 		if (comp_id == 0x10) {
1731 			w2_char = in_str[index++] << 8;
1732 			w2_char |= in_str[index++];
1733 		} else {
1734 			w2_char = in_str[index++];
1735 		}
1736 
1737 		if (make_crc != 0) {
1738 			crc += w2_char;
1739 		}
1740 
1741 		if (w2_char == DOT) {
1742 			dot_loc = len_till_now;
1743 		}
1744 
1745 		/*
1746 		 * Get rid of invalid characters
1747 		 */
1748 		if ((w2_char == SLASH) ||
1749 			(w2_char == NULL)) {
1750 			make_crc = 1;
1751 			if (((comp_id == 8) &&
1752 				(lic != (index - 1))) ||
1753 				(comp_id == 0x10) &&
1754 				(lic != (index - 2))) {
1755 				w2_char = UNDERBAR;
1756 				lic = index;
1757 			} else {
1758 				lic = index;
1759 				continue;
1760 			}
1761 		}
1762 
1763 		/*
1764 		 * Conver a 16bit character to a
1765 		 * utf8 byte stream
1766 		 */
1767 		if ((c_tx_sz = ud_utf162utf8(w2_char, utf8)) == 0) {
1768 			error = EINVAL;
1769 			goto end;
1770 		}
1771 		ASSERT(c_tx_sz > 0);
1772 		ASSERT(c_tx_sz < 7);
1773 
1774 		/*
1775 		 * The output string is larger than
1776 		 * the maximum allowed string length
1777 		 */
1778 		if ((crc_start_loc == 0) &&
1779 			((len_till_now + c_tx_sz) > MAX_ALLOWABLE_STRING)) {
1780 			crc_start_loc = len_till_now;
1781 		}
1782 
1783 		if ((len_till_now + c_tx_sz) < MAXNAMELEN) {
1784 			(void) strncpy((caddr_t)&out_str[len_till_now],
1785 				(caddr_t)utf8, c_tx_sz);
1786 			len_till_now += c_tx_sz;
1787 		} else {
1788 			break;
1789 		}
1790 	}
1791 
1792 	/*
1793 	 * If we need to append CRC do it now
1794 	 */
1795 
1796 	if (make_crc) {
1797 
1798 		if (len_till_now > MAX_ALLOWABLE_STRING) {
1799 			len_till_now = crc_start_loc;
1800 		}
1801 
1802 		if (dot_loc > 0) {
1803 			/*
1804 			 * Make space for crc before the DOT
1805 			 * move the rest of the file name to the end
1806 			 */
1807 			for (k = len_till_now - 1; k >= dot_loc; k--) {
1808 				out_str[k + 5] = out_str[k];
1809 			}
1810 			k = dot_loc;
1811 		} else {
1812 			k = len_till_now;
1813 		}
1814 make_append_crc:
1815 		crc = ud_crc(in_str, in_len);
1816 		out_str[k++] = POUND;
1817 		out_str[k++] = htoc[(uint16_t)(crc & 0xf000) >> 12];
1818 		out_str[k++] = htoc[(uint16_t)(crc & 0xf00) >> 8];
1819 		out_str[k++] = htoc[(uint16_t)(crc & 0xf0) >> 4];
1820 		out_str[k++] = htoc[crc & 0xf];
1821 		len_till_now += 5;
1822 	}
1823 	*out_len = len_till_now;
1824 end:
1825 	return (error);
1826 }
1827 
1828 
1829 struct buf *
1830 ud_bread(dev_t dev, daddr_t blkno, long bsize)
1831 {
1832 	struct buf *bp;
1833 
1834 begin:
1835 	bp = bread(dev, blkno, bsize);
1836 
1837 	if (((bp->b_flags & B_ERROR) == 0) &&
1838 		(bp->b_bcount != bsize)) {
1839 		/*
1840 		 * Buffer cache returned a
1841 		 * wrong number of bytes
1842 		 * flush the old buffer and
1843 		 * reread it again
1844 		 */
1845 		if (bp->b_flags & B_DELWRI) {
1846 			bwrite(bp);
1847 		} else {
1848 			bp->b_flags |= (B_AGE | B_STALE);
1849 			brelse(bp);
1850 		}
1851 		goto begin;
1852 	}
1853 
1854 	return (bp);
1855 }
1856 
1857 /*
1858  * Decide whether it is okay to remove within a sticky directory.
1859  * Two conditions need to be met:  write access to the directory
1860  * is needed.  In sticky directories, write access is not sufficient;
1861  * you can remove entries from a directory only if you own the directory,
1862  * if you are privileged, if you own the entry or if they entry is
1863  * a plain file and you have write access to that file.
1864  * Function returns 0 if remove access is granted.
1865  */
1866 int
1867 ud_sticky_remove_access(struct ud_inode *dir, struct ud_inode *entry,
1868 	struct cred *cr)
1869 {
1870 	uid_t uid;
1871 
1872 	if ((dir->i_char & ISVTX) &&
1873 	    (uid = crgetuid(cr)) != dir->i_uid &&
1874 	    uid != entry->i_uid &&
1875 	    (entry->i_type != VREG ||
1876 	    ud_iaccess(entry, IWRITE, cr) != 0))
1877 		return (secpolicy_vnode_remove(cr));
1878 
1879 	return (0);
1880 }
1881