xref: /dflybsd-src/sys/dev/disk/dm/device-mapper.c (revision a3c6b8ec99ce8a3fe6f1ca8ace1d35c9b45e2bd4)
1 /*        $NetBSD: device-mapper.c,v 1.22 2010/03/26 15:46:04 jakllsch Exp $ */
2 
3 /*
4  * Copyright (c) 2010-2011 Alex Hornung <alex@alexhornung.com>
5  * Copyright (c) 2010 The NetBSD Foundation, Inc.
6  * All rights reserved.
7  *
8  * This code is derived from software contributed to The NetBSD Foundation
9  * by Adam Hamsik.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
22  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
23  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
24  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30  * POSSIBILITY OF SUCH DAMAGE.
31  */
32 
33 /*
34  * I want to say thank you to all people who helped me with this project.
35  */
36 
37 #include <sys/types.h>
38 #include <sys/param.h>
39 #include <sys/ctype.h>
40 
41 #include <sys/buf.h>
42 #include <sys/conf.h>
43 #include <sys/device.h>
44 #include <sys/disk.h>
45 #include <sys/disklabel.h>
46 #include <sys/dtype.h>
47 #include <sys/malloc.h>
48 #include <sys/module.h>
49 #include <sys/sysctl.h>
50 #include <dev/disk/dm/dm.h>
51 
52 #include "netbsd-dm.h"
53 
54 static	d_ioctl_t	dmioctl;
55 static	d_open_t	dmopen;
56 static	d_close_t	dmclose;
57 static	d_psize_t	dmsize;
58 static	d_strategy_t	dmstrategy;
59 static	d_dump_t	dmdump;
60 
61 /* attach and detach routines */
62 static int dm_modcmd(module_t mod, int cmd, void *unused);
63 static int dmdestroy(void);
64 
65 static void dm_doinit(void);
66 
67 static int dm_cmd_to_fun(prop_dictionary_t);
68 static int disk_ioctl_switch(cdev_t, u_long, void *);
69 static int dm_ioctl_switch(u_long);
70 #if 0
71 static void dmminphys(struct buf *);
72 #endif
73 
74 /* ***Variable-definitions*** */
75 struct dev_ops dm_ops = {
76 	{ "dm", 0, D_DISK | D_MPSAFE },
77 	.d_open		= dmopen,
78 	.d_close	= dmclose,
79 	.d_read		= physread,
80 	.d_write	= physwrite,
81 	.d_ioctl	= dmioctl,
82 	.d_strategy	= dmstrategy,
83 	.d_psize	= dmsize,
84 	.d_dump		= dmdump,
85 /* D_DISK */
86 };
87 
88 MALLOC_DEFINE(M_DM, "dm", "Device Mapper allocations");
89 
90 int dm_debug_level = 0;
91 
92 extern uint64_t dm_dev_counter;
93 
94 static cdev_t dmcdev;
95 
96 static moduledata_t dm_mod = {
97     "dm",
98     dm_modcmd,
99     NULL
100 };
101 DECLARE_MODULE(dm, dm_mod, SI_SUB_RAID, SI_ORDER_ANY);
102 MODULE_VERSION(dm, 1);
103 
104 /*
105  * This array is used to translate cmd to function pointer.
106  *
107  * Interface between libdevmapper and lvm2tools uses different
108  * names for one IOCTL call because libdevmapper do another thing
109  * then. When I run "info" or "mknodes" libdevmapper will send same
110  * ioctl to kernel but will do another things in userspace.
111  *
112  */
113 static struct cmd_function cmd_fn[] = {
114 		{ .cmd = "version", .fn = dm_get_version_ioctl},
115 		{ .cmd = "targets", .fn = dm_list_versions_ioctl},
116 		{ .cmd = "create",  .fn = dm_dev_create_ioctl},
117 		{ .cmd = "info",    .fn = dm_dev_status_ioctl},
118 		{ .cmd = "mknodes", .fn = dm_dev_status_ioctl},
119 		{ .cmd = "names",   .fn = dm_dev_list_ioctl},
120 		{ .cmd = "suspend", .fn = dm_dev_suspend_ioctl},
121 		{ .cmd = "remove",  .fn = dm_dev_remove_ioctl},
122 		{ .cmd = "remove_all", .fn = dm_dev_remove_all_ioctl},
123 		{ .cmd = "rename",  .fn = dm_dev_rename_ioctl},
124 		{ .cmd = "resume",  .fn = dm_dev_resume_ioctl},
125 		{ .cmd = "clear",   .fn = dm_table_clear_ioctl},
126 		{ .cmd = "deps",    .fn = dm_table_deps_ioctl},
127 		{ .cmd = "reload",  .fn = dm_table_load_ioctl},
128 		{ .cmd = "status",  .fn = dm_table_status_ioctl},
129 		{ .cmd = "table",   .fn = dm_table_status_ioctl},
130 		{ .cmd = "message", .fn = dm_message_ioctl},
131 		{NULL, NULL}
132 };
133 
134 /* New module handle routine */
135 static int
136 dm_modcmd(module_t mod, int cmd, void *unused)
137 {
138 	int error;
139 
140 	error = 0;
141 
142 	switch (cmd) {
143 	case MOD_LOAD:
144 		dm_doinit();
145 		kprintf("Device Mapper version %d.%d.%d loaded\n",
146 		    DM_VERSION_MAJOR, DM_VERSION_MINOR, DM_VERSION_PATCHLEVEL);
147 		break;
148 
149 	case MOD_UNLOAD:
150 		/*
151 		 * Disable unloading of dm module if there are any devices
152 		 * defined in driver. This is probably too strong we need
153 		 * to disable auto-unload only if there is mounted dm device
154 		 * present.
155 		 */
156 		if (dm_dev_counter > 0)
157 			return EBUSY;
158 
159 		error = dmdestroy();
160 		if (error)
161 			break;
162 		kprintf("Device Mapper unloaded\n");
163 		break;
164 
165 	default:
166 		break;
167 	}
168 
169 	return error;
170 }
171 
172 static void
173 dm_doinit(void)
174 {
175 	dm_target_init();
176 	dm_dev_init();
177 	dm_pdev_init();
178 	dmcdev = make_dev(&dm_ops, 0, UID_ROOT, GID_OPERATOR, 0640, "mapper/control");
179 }
180 
181 /* Destroy routine */
182 static int
183 dmdestroy(void)
184 {
185 	destroy_dev(dmcdev);
186 
187 	dm_dev_uninit();
188 	dm_pdev_uninit();
189 	dm_target_uninit();
190 
191 	return 0;
192 }
193 
194 static int
195 dmopen(struct dev_open_args *ap)
196 {
197 	cdev_t dev = ap->a_head.a_dev;
198 	dm_dev_t *dmv;
199 
200 	/* Shortcut for the control device */
201 	if (minor(dev) == 0)
202 		return 0;
203 
204 	if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
205 		return ENXIO;
206 
207 	dmv->is_open = 1;
208 	dm_dev_unbusy(dmv);
209 
210 	aprint_debug("dm open routine called %" PRIu32 "\n",
211 	    minor(ap->a_head.a_dev));
212 	return 0;
213 }
214 
215 static int
216 dmclose(struct dev_close_args *ap)
217 {
218 	cdev_t dev = ap->a_head.a_dev;
219 	dm_dev_t *dmv;
220 
221 	/* Shortcut for the control device */
222 	if (minor(dev) == 0)
223 		return 0;
224 
225 	if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
226 		return ENXIO;
227 
228 	dmv->is_open = 0;
229 	dm_dev_unbusy(dmv);
230 
231 	aprint_debug("dm close routine called %" PRIu32 "\n",
232 	    minor(ap->a_head.a_dev));
233 	return 0;
234 }
235 
236 
237 static int
238 dmioctl(struct dev_ioctl_args *ap)
239 {
240 	cdev_t dev = ap->a_head.a_dev;
241 	u_long cmd = ap->a_cmd;
242 	void *data = ap->a_data;
243 
244 	int r, err;
245 	prop_dictionary_t dm_dict_in;
246 
247 	err = r = 0;
248 
249 	aprint_debug("dmioctl called\n");
250 
251 	KKASSERT(data != NULL);
252 
253 	if ((r = disk_ioctl_switch(dev, cmd, data)) == ENOTTY) {
254 		struct plistref *pref = (struct plistref *) data;
255 
256 		/* Check if we were called with NETBSD_DM_IOCTL ioctl
257 		   otherwise quit. */
258 		if ((r = dm_ioctl_switch(cmd)) != 0)
259 			return r;
260 
261 		if ((r = prop_dictionary_copyin_ioctl(pref, cmd, &dm_dict_in)) != 0)
262 			return r;
263 
264 		if ((r = dm_check_version(dm_dict_in)) != 0)
265 			goto cleanup_exit;
266 
267 		/* run ioctl routine */
268 		if ((err = dm_cmd_to_fun(dm_dict_in)) != 0)
269 			goto cleanup_exit;
270 
271 cleanup_exit:
272 		r = prop_dictionary_copyout_ioctl(pref, cmd, dm_dict_in);
273 		prop_object_release(dm_dict_in);
274 	}
275 
276 	/*
277 	 * Return the error of the actual command if one one has
278 	 * happened. Otherwise return 'r' which indicates errors
279 	 * that occurred during helper operations.
280 	 */
281 	return (err != 0)?err:r;
282 }
283 
284 /*
285  * Translate command sent from libdevmapper to func.
286  */
287 static int
288 dm_cmd_to_fun(prop_dictionary_t dm_dict)
289 {
290 	int i, r;
291 	prop_string_t command;
292 
293 	r = 0;
294 
295 	if ((command = prop_dictionary_get(dm_dict, DM_IOCTL_COMMAND)) == NULL)
296 		return EINVAL;
297 
298 	for (i = 0; cmd_fn[i].cmd != NULL; i++)
299 		if (prop_string_equals_cstring(command, cmd_fn[i].cmd))
300 			break;
301 
302 	if (cmd_fn[i].cmd == NULL)
303 		return EINVAL;
304 
305 	aprint_debug("ioctl %s called\n", cmd_fn[i].cmd);
306 	r = cmd_fn[i].fn(dm_dict);
307 
308 	return r;
309 }
310 
311 /* Call apropriate ioctl handler function. */
312 static int
313 dm_ioctl_switch(u_long cmd)
314 {
315 
316 	switch(cmd) {
317 
318 	case NETBSD_DM_IOCTL:
319 		aprint_debug("dm NetBSD_DM_IOCTL called\n");
320 		break;
321 	default:
322 		aprint_debug("dm unknown ioctl called\n");
323 		return ENOTTY;
324 		break; /* NOT REACHED */
325 	}
326 
327 	return 0;
328 }
329 
330  /*
331   * Check for disk specific ioctls.
332   */
333 
334 static int
335 disk_ioctl_switch(cdev_t dev, u_long cmd, void *data)
336 {
337 	dm_dev_t *dmv;
338 
339 	/* disk ioctls make sense only on block devices */
340 	if (minor(dev) == 0)
341 		return ENOTTY;
342 
343 	switch(cmd) {
344 	case DIOCGPART:
345 	{
346 		struct partinfo *dpart;
347 		u_int64_t size;
348 		dpart = data;
349 		bzero(dpart, sizeof(*dpart));
350 
351 		if ((dmv = dev->si_drv1) == NULL)
352 			return ENODEV;
353 		if (dmv->diskp->d_info.d_media_blksize == 0) {
354 			return ENOTSUP;
355 		} else {
356 			size = dm_table_size(&dmv->table_head);
357 			dpart->media_offset  = 0;
358 			dpart->media_size    = size * DEV_BSIZE;
359 			dpart->media_blocks  = size;
360 			dpart->media_blksize = DEV_BSIZE;
361 			dpart->fstype = FS_BSDFFS;
362 		}
363 		break;
364 	}
365 
366 	default:
367 		aprint_debug("unknown disk_ioctl called\n");
368 		return ENOTTY;
369 		break; /* NOT REACHED */
370 	}
371 
372 	return 0;
373 }
374 
375 /*
376  * Do all IO operations on dm logical devices.
377  */
378 static int
379 dmstrategy(struct dev_strategy_args *ap)
380 {
381 	cdev_t dev = ap->a_head.a_dev;
382 	struct bio *bio = ap->a_bio;
383 	struct buf *bp = bio->bio_buf;
384 	int bypass;
385 
386 	dm_dev_t *dmv;
387 	dm_table_t  *tbl;
388 	dm_table_entry_t *table_en;
389 	struct buf *nestbuf;
390 
391 	uint64_t buf_start, buf_len, issued_len;
392 	uint64_t table_start, table_end;
393 	uint64_t start, end;
394 
395 	buf_start = bio->bio_offset;
396 	buf_len = bp->b_bcount;
397 
398 	tbl = NULL;
399 
400 	table_end = 0;
401 	issued_len = 0;
402 
403 	dmv = dev->si_drv1;
404 
405 	switch(bp->b_cmd) {
406 	case BUF_CMD_READ:
407 	case BUF_CMD_WRITE:
408 	case BUF_CMD_FREEBLKS:
409 		bypass = 0;
410 		break;
411 	case BUF_CMD_FLUSH:
412 		bypass = 1;
413 		KKASSERT(buf_len == 0);
414 		break;
415 	default:
416 		bp->b_error = EIO;
417 		bp->b_resid = bp->b_bcount;
418 		biodone(bio);
419 		return 0;
420 	}
421 
422 	if (bypass == 0 &&
423 	    bounds_check_with_mediasize(bio, DEV_BSIZE,
424 					dm_table_size(&dmv->table_head)) <= 0) {
425 		bp->b_resid = bp->b_bcount;
426 		biodone(bio);
427 		return 0;
428 	}
429 
430 	/* Select active table */
431 	tbl = dm_table_get_entry(&dmv->table_head, DM_TABLE_ACTIVE);
432 
433 	nestiobuf_init(bio);
434 	devstat_start_transaction(&dmv->stats);
435 
436 	/*
437 	 * Find out what tables I want to select.
438 	 */
439 	SLIST_FOREACH(table_en, tbl, next) {
440 		/*
441 		 * I need need number of bytes not blocks.
442 		 */
443 		table_start = table_en->start * DEV_BSIZE;
444 		table_end = table_start + (table_en->length) * DEV_BSIZE;
445 
446 		/*
447 		 * Calculate the start and end
448 		 */
449 		start = MAX(table_start, buf_start);
450 		end = MIN(table_end, buf_start + buf_len);
451 
452 		if (dm_debug_level) {
453 			aprint_normal("----------------------------------------\n");
454 			aprint_normal("table_start %010" PRIu64", table_end %010"
455 			    PRIu64 "\n", table_start, table_end);
456 			aprint_normal("buf_start %010" PRIu64", buf_len %010"
457 			    PRIu64"\n", buf_start, buf_len);
458 			aprint_normal("start-buf_start %010"PRIu64", end %010"
459 			    PRIu64"\n", start - buf_start, end);
460 			aprint_normal("start %010" PRIu64" , end %010"
461 			    PRIu64"\n", start, end);
462 			aprint_normal("\n----------------------------------------\n");
463 		}
464 
465 		if (bypass) {
466 			nestbuf = getpbuf(NULL);
467 			nestbuf->b_flags |= bio->bio_buf->b_flags & B_HASBOGUS;
468 
469 			nestiobuf_add(bio, nestbuf, 0, 0, &dmv->stats);
470 			nestbuf->b_bio1.bio_offset = 0;
471 			table_en->target->strategy(table_en, nestbuf);
472 		} else if (start < end) {
473 			nestbuf = getpbuf(NULL);
474 			nestbuf->b_flags |= bio->bio_buf->b_flags & B_HASBOGUS;
475 
476 			nestiobuf_add(bio, nestbuf,
477 				      start - buf_start, (end - start),
478 				      &dmv->stats);
479 			issued_len += end - start;
480 
481 			nestbuf->b_bio1.bio_offset = (start - table_start);
482 			table_en->target->strategy(table_en, nestbuf);
483 		}
484 	}
485 
486 	if (issued_len < buf_len)
487 		nestiobuf_error(bio, EINVAL);
488 	nestiobuf_start(bio);
489 	dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE);
490 
491 	return 0;
492 }
493 
494 static int
495 dmdump(struct dev_dump_args *ap)
496 {
497 	cdev_t dev = ap->a_head.a_dev;
498 	dm_dev_t *dmv;
499 	dm_table_t  *tbl;
500 	dm_table_entry_t *table_en;
501 	uint64_t buf_start, buf_len, issued_len;
502 	uint64_t table_start, table_end;
503 	uint64_t start, end, data_offset;
504 	off_t offset;
505 	size_t length;
506 	int error = 0;
507 
508 	buf_start = ap->a_offset;
509 	buf_len = ap->a_length;
510 
511 	tbl = NULL;
512 
513 	table_end = 0;
514 	issued_len = 0;
515 
516 	dmv = dev->si_drv1;
517 
518 	/* Select active table */
519 	tbl = dm_table_get_entry(&dmv->table_head, DM_TABLE_ACTIVE);
520 
521 
522 	/*
523 	 * Find out what tables I want to select.
524 	 */
525 	SLIST_FOREACH(table_en, tbl, next) {
526 		/*
527 		 * I need need number of bytes not blocks.
528 		 */
529 		table_start = table_en->start * DEV_BSIZE;
530 		table_end = table_start + (table_en->length) * DEV_BSIZE;
531 
532 		/*
533 		 * Calculate the start and end
534 		 */
535 		start = MAX(table_start, buf_start);
536 		end = MIN(table_end, buf_start + buf_len);
537 
538 		if (ap->a_length == 0) {
539 			if (table_en->target->dump == NULL) {
540 				error = ENXIO;
541 				goto out;
542 			}
543 
544 			table_en->target->dump(table_en, NULL, 0, 0);
545 		} else if (start < end) {
546 			data_offset = start - buf_start;
547 			offset = start - table_start;
548 			length = end - start;
549 
550 			if (table_en->target->dump == NULL) {
551 				error = ENXIO;
552 				goto out;
553 			}
554 
555 			table_en->target->dump(table_en,
556 			    (char *)ap->a_virtual + data_offset,
557 			    length, offset);
558 
559 			issued_len += end - start;
560 		}
561 	}
562 
563 	if (issued_len < buf_len)
564 		error = EINVAL;
565 
566 out:
567 	dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE);
568 
569 	return error;
570 }
571 
572 static int
573 dmsize(struct dev_psize_args *ap)
574 {
575 	cdev_t dev = ap->a_head.a_dev;
576 	dm_dev_t *dmv;
577 	uint64_t size;
578 
579 	size = 0;
580 
581 	if ((dmv = dev->si_drv1) == NULL)
582 		return ENXIO;
583 
584 	size = dm_table_size(&dmv->table_head);
585 	ap->a_result = (int64_t)size;
586 
587 	return 0;
588 }
589 
590 #if 0
591 static void
592 dmminphys(struct buf *bp)
593 {
594 
595 	bp->b_bcount = MIN(bp->b_bcount, MAXPHYS);
596 }
597 #endif
598 
599 void
600 dmsetdiskinfo(struct disk *disk, dm_table_head_t *head)
601 {
602 	struct disk_info info;
603 	uint64_t dmp_size;
604 
605 	dmp_size = dm_table_size(head);
606 
607 	bzero(&info, sizeof(struct disk_info));
608 	info.d_media_blksize = DEV_BSIZE;
609 	info.d_media_blocks = dmp_size;
610 #if 0
611 	/* this is set by disk_setdiskinfo */
612 	info.d_media_size = dmp_size * DEV_BSIZE;
613 #endif
614 	info.d_dsflags = DSO_MBRQUIET | DSO_DEVICEMAPPER | DSO_RAWPSIZE;
615 
616 	info.d_secpertrack = 32;
617 	info.d_nheads = 64;
618 	info.d_secpercyl = info.d_secpertrack * info.d_nheads;
619 	info.d_ncylinders = dmp_size / info.d_secpercyl;
620 
621 	disk_setdiskinfo(disk, &info);
622 }
623 
624 /*
625  * Transform char s to uint64_t offset number.
626  */
627 uint64_t
628 atoi64(const char *s)
629 {
630 	uint64_t n;
631 	n = 0;
632 
633 	while (*s != '\0') {
634 		if (!isdigit(*s))
635 			break;
636 
637 		n = (10 * n) + (*s - '0');
638 		s++;
639 	}
640 
641 	return n;
642 }
643 
644 void
645 dm_builtin_init(void *arg)
646 {
647 	modeventhand_t evh = (modeventhand_t)arg;
648 
649 	KKASSERT(evh != NULL);
650 	evh(NULL, MOD_LOAD, NULL);
651 }
652 
653 void
654 dm_builtin_uninit(void *arg)
655 {
656 	modeventhand_t evh = (modeventhand_t)arg;
657 
658 	KKASSERT(evh != NULL);
659 	evh(NULL, MOD_UNLOAD, NULL);
660 }
661 
662 TUNABLE_INT("debug.dm_debug", &dm_debug_level);
663 SYSCTL_INT(_debug, OID_AUTO, dm_debug, CTLFLAG_RW, &dm_debug_level,
664 	       0, "Enable device mapper debugging");
665 
666