xref: /dflybsd-src/sys/dev/disk/dm/device-mapper.c (revision a84e173e22bf0ea7d3d281a3cb30d0064312742f)
1 /*        $NetBSD: device-mapper.c,v 1.22 2010/03/26 15:46:04 jakllsch Exp $ */
2 
3 /*
4  * Copyright (c) 2010 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Adam Hamsik.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * I want to say thank you to all people who helped me with this project.
34  */
35 
36 #include <sys/types.h>
37 #include <sys/param.h>
38 
39 #include <sys/buf.h>
40 #include <sys/conf.h>
41 #include <sys/device.h>
42 #include <sys/devfs.h>
43 #include <sys/disk.h>
44 #include <sys/disklabel.h>
45 #include <sys/dtype.h>
46 #include <sys/ioccom.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 void dmattach(int);
63 static int dm_modcmd(module_t mod, int cmd, void *unused);
64 static int dmdestroy(void);
65 
66 static void dm_doinit(void);
67 
68 static int dm_cmd_to_fun(prop_dictionary_t);
69 static int disk_ioctl_switch(cdev_t, u_long, void *);
70 static int dm_ioctl_switch(u_long);
71 #if 0
72 static void dmminphys(struct buf *);
73 #endif
74 
75 struct devfs_bitmap dm_minor_bitmap;
76 
77 /* ***Variable-definitions*** */
78 struct dev_ops dm_ops = {
79 	{ "dm", 0, D_DISK | D_MPSAFE },
80 	.d_open		= dmopen,
81 	.d_close	= dmclose,
82 	.d_read		= physread,
83 	.d_write	= physwrite,
84 	.d_ioctl	= dmioctl,
85 	.d_strategy	= dmstrategy,
86 	.d_psize	= dmsize,
87 	.d_dump		= dmdump,
88 /* D_DISK */
89 };
90 
91 MALLOC_DEFINE(M_DM, "dm", "Device Mapper allocations");
92 
93 int dm_debug_level = 0;
94 
95 extern uint64_t dm_dev_counter;
96 
97 static cdev_t dmcdev;
98 
99 static moduledata_t dm_mod = {
100     "dm",
101     dm_modcmd,
102     NULL
103 };
104 DECLARE_MODULE(dm, dm_mod, SI_SUB_RAID, SI_ORDER_ANY);
105 MODULE_VERSION(dm, 1);
106 
107 /*
108  * This array is used to translate cmd to function pointer.
109  *
110  * Interface between libdevmapper and lvm2tools uses different
111  * names for one IOCTL call because libdevmapper do another thing
112  * then. When I run "info" or "mknodes" libdevmapper will send same
113  * ioctl to kernel but will do another things in userspace.
114  *
115  */
116 static struct cmd_function cmd_fn[] = {
117 		{ .cmd = "version", .fn = dm_get_version_ioctl},
118 		{ .cmd = "targets", .fn = dm_list_versions_ioctl},
119 		{ .cmd = "create",  .fn = dm_dev_create_ioctl},
120 		{ .cmd = "info",    .fn = dm_dev_status_ioctl},
121 		{ .cmd = "mknodes", .fn = dm_dev_status_ioctl},
122 		{ .cmd = "names",   .fn = dm_dev_list_ioctl},
123 		{ .cmd = "suspend", .fn = dm_dev_suspend_ioctl},
124 		{ .cmd = "remove",  .fn = dm_dev_remove_ioctl},
125 		{ .cmd = "rename",  .fn = dm_dev_rename_ioctl},
126 		{ .cmd = "resume",  .fn = dm_dev_resume_ioctl},
127 		{ .cmd = "clear",   .fn = dm_table_clear_ioctl},
128 		{ .cmd = "deps",    .fn = dm_table_deps_ioctl},
129 		{ .cmd = "reload",  .fn = dm_table_load_ioctl},
130 		{ .cmd = "status",  .fn = dm_table_status_ioctl},
131 		{ .cmd = "table",   .fn = dm_table_status_ioctl},
132 		{NULL, NULL}
133 };
134 
135 /* New module handle routine */
136 static int
137 dm_modcmd(module_t mod, int cmd, void *unused)
138 {
139 	int error, bmajor, cmajor;
140 
141 	error = 0;
142 	bmajor = -1;
143 	cmajor = -1;
144 
145 	switch (cmd) {
146 	case MOD_LOAD:
147 		devfs_clone_bitmap_init(&dm_minor_bitmap);
148 		dm_doinit();
149 		kprintf("Device Mapper version %d.%d.%d loaded\n",
150 		    DM_VERSION_MAJOR, DM_VERSION_MINOR, DM_VERSION_PATCHLEVEL);
151 		break;
152 
153 	case MOD_UNLOAD:
154 		/*
155 		 * Disable unloading of dm module if there are any devices
156 		 * defined in driver. This is probably too strong we need
157 		 * to disable auto-unload only if there is mounted dm device
158 		 * present.
159 		 */
160 		if (dm_dev_counter > 0)
161 			return EBUSY;
162 
163 		error = dmdestroy();
164 		if (error)
165 			break;
166 		kprintf("Device Mapper unloaded\n");
167 		break;
168 
169 	default:
170 		break;
171 	}
172 
173 	return error;
174 }
175 
176 /*
177  * dm_detach is called to completely destroy & remove a dm disk device.
178  */
179 int
180 dm_detach(dm_dev_t *dmv)
181 {
182 	int minor;
183 
184 	/* Remove device from list and wait for refcnt to drop to zero */
185 	dm_dev_rem(dmv, NULL, NULL, -1);
186 
187 	/* Destroy active table first.  */
188 	dm_table_destroy(&dmv->table_head, DM_TABLE_ACTIVE);
189 
190 	/* Destroy inactive table if exits, too. */
191 	dm_table_destroy(&dmv->table_head, DM_TABLE_INACTIVE);
192 
193 	dm_table_head_destroy(&dmv->table_head);
194 
195 	minor = dkunit(dmv->devt);
196 	disk_destroy(dmv->diskp);
197 	devstat_remove_entry(&dmv->stats);
198 
199 	release_dev(dmv->devt);
200 	devfs_clone_bitmap_put(&dm_minor_bitmap, minor);
201 
202 	/* Destroy device */
203 	(void)dm_dev_free(dmv);
204 
205 	/* Decrement device counter After removing device */
206 	--dm_dev_counter; /* XXX: was atomic 64 */
207 
208 	return 0;
209 }
210 
211 static void
212 dm_doinit(void)
213 {
214 	dm_target_init();
215 	dm_dev_init();
216 	dm_pdev_init();
217 	dmcdev = make_dev(&dm_ops, 0, UID_ROOT, GID_OPERATOR, 0640, "mapper/control");
218 }
219 
220 /* Destroy routine */
221 static int
222 dmdestroy(void)
223 {
224 	destroy_dev(dmcdev);
225 
226 	dm_dev_destroy();
227 	dm_pdev_destroy();
228 	dm_target_destroy();
229 
230 	return 0;
231 }
232 
233 static int
234 dmopen(struct dev_open_args *ap)
235 {
236 	cdev_t dev = ap->a_head.a_dev;
237 	dm_dev_t *dmv;
238 
239 	/* Shortcut for the control device */
240 	if (minor(dev) == 0)
241 		return 0;
242 
243 	if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
244 		return ENXIO;
245 
246 	dmv->is_open = 1;
247 	dm_dev_unbusy(dmv);
248 
249 	aprint_debug("dm open routine called %" PRIu32 "\n",
250 	    minor(ap->a_head.a_dev));
251 	return 0;
252 }
253 
254 static int
255 dmclose(struct dev_close_args *ap)
256 {
257 	cdev_t dev = ap->a_head.a_dev;
258 	dm_dev_t *dmv;
259 
260 	/* Shortcut for the control device */
261 	if (minor(dev) == 0)
262 		return 0;
263 
264 	if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
265 		return ENXIO;
266 
267 	dmv->is_open = 0;
268 	dm_dev_unbusy(dmv);
269 
270 	aprint_debug("dm close routine called %" PRIu32 "\n",
271 	    minor(ap->a_head.a_dev));
272 	return 0;
273 }
274 
275 
276 static int
277 dmioctl(struct dev_ioctl_args *ap)
278 {
279 	cdev_t dev = ap->a_head.a_dev;
280 	u_long cmd = ap->a_cmd;
281 	void *data = ap->a_data;
282 
283 	int r, err;
284 	prop_dictionary_t dm_dict_in;
285 
286 	err = r = 0;
287 
288 	aprint_debug("dmioctl called\n");
289 
290 	KKASSERT(data != NULL);
291 
292 	if (( r = disk_ioctl_switch(dev, cmd, data)) == ENOTTY) {
293 		struct plistref *pref = (struct plistref *) data;
294 
295 		/* Check if we were called with NETBSD_DM_IOCTL ioctl
296 		   otherwise quit. */
297 		if ((r = dm_ioctl_switch(cmd)) != 0)
298 			return r;
299 
300 		if((r = prop_dictionary_copyin_ioctl(pref, cmd, &dm_dict_in)) != 0)
301 			return r;
302 
303 		if ((r = dm_check_version(dm_dict_in)) != 0)
304 			goto cleanup_exit;
305 
306 		/* run ioctl routine */
307 		if ((err = dm_cmd_to_fun(dm_dict_in)) != 0)
308 			goto cleanup_exit;
309 
310 cleanup_exit:
311 		r = prop_dictionary_copyout_ioctl(pref, cmd, dm_dict_in);
312 		prop_object_release(dm_dict_in);
313 	}
314 
315 	/*
316 	 * Return the error of the actual command if one one has
317 	 * happened. Otherwise return 'r' which indicates errors
318 	 * that occurred during helper operations.
319 	 */
320 	return (err != 0)?err:r;
321 }
322 
323 /*
324  * Translate command sent from libdevmapper to func.
325  */
326 static int
327 dm_cmd_to_fun(prop_dictionary_t dm_dict){
328 	int i, r;
329 	prop_string_t command;
330 
331 	r = 0;
332 
333 	if ((command = prop_dictionary_get(dm_dict, DM_IOCTL_COMMAND)) == NULL)
334 		return EINVAL;
335 
336 	for(i = 0; cmd_fn[i].cmd != NULL; i++)
337 		if (prop_string_equals_cstring(command, cmd_fn[i].cmd))
338 			break;
339 
340 	if (cmd_fn[i].cmd == NULL)
341 		return EINVAL;
342 
343 	aprint_debug("ioctl %s called\n", cmd_fn[i].cmd);
344 	r = cmd_fn[i].fn(dm_dict);
345 
346 	return r;
347 }
348 
349 /* Call apropriate ioctl handler function. */
350 static int
351 dm_ioctl_switch(u_long cmd)
352 {
353 
354 	switch(cmd) {
355 
356 	case NETBSD_DM_IOCTL:
357 		aprint_debug("dm NetBSD_DM_IOCTL called\n");
358 		break;
359 	default:
360 		 aprint_debug("dm unknown ioctl called\n");
361 		 return ENOTTY;
362 		 break; /* NOT REACHED */
363 	}
364 
365 	 return 0;
366 }
367 
368  /*
369   * Check for disk specific ioctls.
370   */
371 
372 static int
373 disk_ioctl_switch(cdev_t dev, u_long cmd, void *data)
374 {
375 	dm_dev_t *dmv;
376 
377 	/* disk ioctls make sense only on block devices */
378 	if (minor(dev) == 0)
379 		return ENOTTY;
380 
381 	switch(cmd) {
382 	case DIOCGPART:
383 	{
384 		struct partinfo *dpart;
385 		u_int64_t size;
386 		dpart = (void *)data;
387 		bzero(dpart, sizeof(*dpart));
388 
389 		if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
390 			return ENODEV;
391 		if (dmv->diskp->d_info.d_media_blksize == 0) {
392 			dm_dev_unbusy(dmv);
393 			return ENOTSUP;
394 		} else {
395 			size = dm_table_size(&dmv->table_head);
396 			dpart->media_offset  = 0;
397 			dpart->media_size    = size * DEV_BSIZE;
398 			dpart->media_blocks  = size;
399 			dpart->media_blksize = DEV_BSIZE;
400 			dpart->fstype = FS_BSDFFS;
401 		}
402 		dm_dev_unbusy(dmv);
403 		break;
404 	}
405 
406 	default:
407 		aprint_debug("unknown disk_ioctl called\n");
408 		return ENOTTY;
409 		break; /* NOT REACHED */
410 	}
411 
412 	return 0;
413 }
414 
415 /*
416  * Do all IO operations on dm logical devices.
417  */
418 static int
419 dmstrategy(struct dev_strategy_args *ap)
420 {
421 	cdev_t dev = ap->a_head.a_dev;
422 	struct bio *bio = ap->a_bio;
423 	struct buf *bp = bio->bio_buf;
424 	int bypass;
425 
426 	dm_dev_t *dmv;
427 	dm_table_t  *tbl;
428 	dm_table_entry_t *table_en;
429 	struct buf *nestbuf;
430 
431 	uint32_t dev_type;
432 
433 	uint64_t buf_start, buf_len, issued_len;
434 	uint64_t table_start, table_end;
435 	uint64_t start, end;
436 
437 	buf_start = bio->bio_offset;
438 	buf_len = bp->b_bcount;
439 
440 	tbl = NULL;
441 
442 	table_end = 0;
443 	dev_type = 0;
444 	issued_len = 0;
445 
446 	if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL) {
447 		bp->b_error = EIO;
448 		bp->b_resid = bp->b_bcount;
449 		biodone(bio);
450 		return 0;
451 	}
452 
453 	switch(bp->b_cmd) {
454 	case BUF_CMD_READ:
455 	case BUF_CMD_WRITE:
456 	case BUF_CMD_FREEBLKS:
457 		bypass = 0;
458 		break;
459 	case BUF_CMD_FLUSH:
460 		bypass = 1;
461 		KKASSERT(buf_len == 0);
462 		break;
463 	default:
464 		dm_dev_unbusy(dmv);
465 		bp->b_error = EIO;
466 		bp->b_resid = bp->b_bcount;
467 		biodone(bio);
468 		return 0;
469 	}
470 
471 	if (bypass == 0 &&
472 	    bounds_check_with_mediasize(bio, DEV_BSIZE,
473 					dm_table_size(&dmv->table_head)) <= 0) {
474 		dm_dev_unbusy(dmv);
475 		bp->b_resid = bp->b_bcount;
476 		biodone(bio);
477 		return 0;
478 	}
479 
480 	/* Select active table */
481 	tbl = dm_table_get_entry(&dmv->table_head, DM_TABLE_ACTIVE);
482 
483 	nestiobuf_init(bio);
484 	devstat_start_transaction(&dmv->stats);
485 
486 	/*
487 	 * Find out what tables I want to select.
488 	 */
489 	SLIST_FOREACH(table_en, tbl, next) {
490 		/*
491 		 * I need need number of bytes not blocks.
492 		 */
493 		table_start = table_en->start * DEV_BSIZE;
494 		table_end = table_start + (table_en->length) * DEV_BSIZE;
495 
496 		/*
497 		 * Calculate the start and end
498 		 */
499 		start = MAX(table_start, buf_start);
500 		end = MIN(table_end, buf_start + buf_len);
501 
502 		aprint_debug("----------------------------------------\n");
503 		aprint_debug("table_start %010" PRIu64", table_end %010"
504 		    PRIu64 "\n", table_start, table_end);
505 		aprint_debug("buf_start %010" PRIu64", buf_len %010"
506 		    PRIu64"\n", buf_start, buf_len);
507 		aprint_debug("start-buf_start %010"PRIu64", end %010"
508 		    PRIu64"\n", start - buf_start, end);
509 		aprint_debug("start %010" PRIu64" , end %010"
510                     PRIu64"\n", start, end);
511 		aprint_debug("\n----------------------------------------\n");
512 
513 		if (bypass) {
514 			nestbuf = getpbuf(NULL);
515 			nestbuf->b_flags |= bio->bio_buf->b_flags & B_HASBOGUS;
516 
517 			nestiobuf_add(bio, nestbuf, 0, 0, &dmv->stats);
518 			nestbuf->b_bio1.bio_offset = 0;
519 			table_en->target->strategy(table_en, nestbuf);
520 		} else if (start < end) {
521 			nestbuf = getpbuf(NULL);
522 			nestbuf->b_flags |= bio->bio_buf->b_flags & B_HASBOGUS;
523 
524 			nestiobuf_add(bio, nestbuf,
525 				      start - buf_start, (end - start),
526 				      &dmv->stats);
527 			issued_len += end - start;
528 
529 			nestbuf->b_bio1.bio_offset = (start - table_start);
530 			table_en->target->strategy(table_en, nestbuf);
531 		}
532 	}
533 
534 	if (issued_len < buf_len)
535 		nestiobuf_error(bio, EINVAL);
536 	nestiobuf_start(bio);
537 	dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE);
538 	dm_dev_unbusy(dmv);
539 
540 	return 0;
541 }
542 
543 static int
544 dmdump(struct dev_dump_args *ap)
545 {
546 	cdev_t dev = ap->a_head.a_dev;
547 	dm_dev_t *dmv;
548 	dm_table_t  *tbl;
549 	dm_table_entry_t *table_en;
550 	uint32_t dev_type;
551 	uint64_t buf_start, buf_len, issued_len;
552 	uint64_t table_start, table_end;
553 	uint64_t start, end, data_offset;
554 	off_t offset;
555 	size_t length;
556 	int error = 0;
557 
558 	buf_start = ap->a_offset;
559 	buf_len = ap->a_length;
560 
561 	tbl = NULL;
562 
563 	table_end = 0;
564 	dev_type = 0;
565 	issued_len = 0;
566 
567 	if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL) {
568 		return EIO;
569 	}
570 
571 	/* Select active table */
572 	tbl = dm_table_get_entry(&dmv->table_head, DM_TABLE_ACTIVE);
573 
574 
575 	/*
576 	 * Find out what tables I want to select.
577 	 */
578 	SLIST_FOREACH(table_en, tbl, next) {
579 		/*
580 		 * I need need number of bytes not blocks.
581 		 */
582 		table_start = table_en->start * DEV_BSIZE;
583 		table_end = table_start + (table_en->length) * DEV_BSIZE;
584 
585 		/*
586 		 * Calculate the start and end
587 		 */
588 		start = MAX(table_start, buf_start);
589 		end = MIN(table_end, buf_start + buf_len);
590 
591 		if (ap->a_length == 0) {
592 			if (table_en->target->dump == NULL) {
593 				error = ENXIO;
594 				goto out;
595 			}
596 
597 			table_en->target->dump(table_en, NULL, 0, 0);
598 		} else if (start < end) {
599 			data_offset = start - buf_start;
600 			offset = start - table_start;
601 			length = end - start;
602 
603 			if (table_en->target->dump == NULL) {
604 				error = ENXIO;
605 				goto out;
606 			}
607 
608 			table_en->target->dump(table_en,
609 			    (char *)ap->a_virtual + data_offset,
610 			    length, offset);
611 
612 			issued_len += end - start;
613 		}
614 	}
615 
616 	if (issued_len < buf_len)
617 		error = EINVAL;
618 
619 out:
620 	dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE);
621 	dm_dev_unbusy(dmv);
622 
623 	return error;
624 }
625 
626 static int
627 dmsize(struct dev_psize_args *ap)
628 {
629 	cdev_t dev = ap->a_head.a_dev;
630 	dm_dev_t *dmv;
631 	uint64_t size;
632 
633 	size = 0;
634 
635 	if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
636 			return ENOENT;
637 
638 	size = dm_table_size(&dmv->table_head);
639 	dm_dev_unbusy(dmv);
640 
641 	ap->a_result = (int64_t)size;
642 
643 	return 0;
644 }
645 
646 #if 0
647 static void
648 dmminphys(struct buf *bp)
649 {
650 
651 	bp->b_bcount = MIN(bp->b_bcount, MAXPHYS);
652 }
653 #endif
654 
655 void
656 dmsetdiskinfo(struct disk *disk, dm_table_head_t *head)
657 {
658 	struct disk_info info;
659 	uint64_t dmp_size;
660 
661 	dmp_size = dm_table_size(head);
662 
663 	bzero(&info, sizeof(struct disk_info));
664 	info.d_media_blksize = DEV_BSIZE;
665 	info.d_media_blocks = dmp_size;
666 #if 0
667 	/* this is set by disk_setdiskinfo */
668 	info.d_media_size = dmp_size * DEV_BSIZE;
669 #endif
670 	info.d_dsflags = DSO_MBRQUIET | DSO_DEVICEMAPPER;
671 
672 	info.d_secpertrack = 32;
673 	info.d_nheads = 64;
674 	info.d_secpercyl = info.d_secpertrack * info.d_nheads;
675 	info.d_ncylinders = dmp_size / info.d_secpercyl;
676 
677 	disk_setdiskinfo(disk, &info);
678 }
679 
680 void
681 dm_builtin_init(void *arg)
682 {
683 	modeventhand_t evh = (modeventhand_t)arg;
684 
685 	KKASSERT(evh != NULL);
686 	evh(NULL, MOD_LOAD, NULL);
687 }
688 
689 void
690 dm_builtin_uninit(void *arg)
691 {
692 	modeventhand_t evh = (modeventhand_t)arg;
693 
694 	KKASSERT(evh != NULL);
695 	evh(NULL, MOD_UNLOAD, NULL);
696 }
697 
698 TUNABLE_INT("debug.dm_debug", &dm_debug_level);
699 SYSCTL_INT(_debug, OID_AUTO, dm_debug, CTLFLAG_RW, &dm_debug_level,
700 	       0, "Eanble device mapper debugging");
701 
702