xref: /dflybsd-src/sys/kern/subr_disk.c (revision 1fbe667443a3b9d7739b883913c42dd01b86b8ee)
1 /*
2  * Copyright (c) 2003,2004 The DragonFly Project.  All rights reserved.
3  *
4  * This code is derived from software contributed to The DragonFly Project
5  * by Matthew Dillon <dillon@backplane.com>
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  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in
15  *    the documentation and/or other materials provided with the
16  *    distribution.
17  * 3. Neither the name of The DragonFly Project nor the names of its
18  *    contributors may be used to endorse or promote products derived
19  *    from this software without specific, prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  * ----------------------------------------------------------------------------
35  * "THE BEER-WARE LICENSE" (Revision 42):
36  * <phk@FreeBSD.ORG> wrote this file.  As long as you retain this notice you
37  * can do whatever you want with this stuff. If we meet some day, and you think
38  * this stuff is worth it, you can buy me a beer in return.   Poul-Henning Kamp
39  * ----------------------------------------------------------------------------
40  *
41  * Copyright (c) 1982, 1986, 1988, 1993
42  *	The Regents of the University of California.  All rights reserved.
43  * (c) UNIX System Laboratories, Inc.
44  * All or some portions of this file are derived from material licensed
45  * to the University of California by American Telephone and Telegraph
46  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
47  * the permission of UNIX System Laboratories, Inc.
48  *
49  * Redistribution and use in source and binary forms, with or without
50  * modification, are permitted provided that the following conditions
51  * are met:
52  * 1. Redistributions of source code must retain the above copyright
53  *    notice, this list of conditions and the following disclaimer.
54  * 2. Redistributions in binary form must reproduce the above copyright
55  *    notice, this list of conditions and the following disclaimer in the
56  *    documentation and/or other materials provided with the distribution.
57  * 3. All advertising materials mentioning features or use of this software
58  *    must display the following acknowledgement:
59  *	This product includes software developed by the University of
60  *	California, Berkeley and its contributors.
61  * 4. Neither the name of the University nor the names of its contributors
62  *    may be used to endorse or promote products derived from this software
63  *    without specific prior written permission.
64  *
65  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
66  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
67  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
68  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
69  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
70  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
71  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
72  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
73  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
74  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
75  * SUCH DAMAGE.
76  *
77  *	@(#)ufs_disksubr.c	8.5 (Berkeley) 1/21/94
78  * $FreeBSD: src/sys/kern/subr_disk.c,v 1.20.2.6 2001/10/05 07:14:57 peter Exp $
79  * $FreeBSD: src/sys/ufs/ufs/ufs_disksubr.c,v 1.44.2.3 2001/03/05 05:42:19 obrien Exp $
80  * $DragonFly: src/sys/kern/subr_disk.c,v 1.40 2008/06/05 18:06:32 swildner Exp $
81  */
82 
83 #include <sys/param.h>
84 #include <sys/systm.h>
85 #include <sys/kernel.h>
86 #include <sys/proc.h>
87 #include <sys/sysctl.h>
88 #include <sys/buf.h>
89 #include <sys/conf.h>
90 #include <sys/disklabel.h>
91 #include <sys/disklabel32.h>
92 #include <sys/disklabel64.h>
93 #include <sys/diskslice.h>
94 #include <sys/diskmbr.h>
95 #include <sys/disk.h>
96 #include <sys/malloc.h>
97 #include <sys/sysctl.h>
98 #include <machine/md_var.h>
99 #include <sys/ctype.h>
100 #include <sys/syslog.h>
101 #include <sys/device.h>
102 #include <sys/msgport.h>
103 #include <sys/msgport2.h>
104 #include <sys/buf2.h>
105 #include <vfs/devfs/devfs.h>
106 #include <sys/thread.h>
107 #include <sys/thread2.h>
108 
109 #include <sys/queue.h>
110 #include <sys/lock.h>
111 
112 static MALLOC_DEFINE(M_DISK, "disk", "disk data");
113 
114 static void disk_msg_autofree_reply(lwkt_port_t, lwkt_msg_t);
115 static void disk_msg_core(void *);
116 static int disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe);
117 static void disk_probe(struct disk *dp, int reprobe);
118 static void _setdiskinfo(struct disk *disk, struct disk_info *info);
119 
120 static d_open_t diskopen;
121 static d_close_t diskclose;
122 static d_ioctl_t diskioctl;
123 static d_strategy_t diskstrategy;
124 static d_psize_t diskpsize;
125 static d_clone_t diskclone;
126 static d_dump_t diskdump;
127 
128 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
129 static struct lwkt_token disklist_token;
130 
131 static struct dev_ops disk_ops = {
132 	{ "disk", 0, D_DISK },
133 	.d_open = diskopen,
134 	.d_close = diskclose,
135 	.d_read = physread,
136 	.d_write = physwrite,
137 	.d_ioctl = diskioctl,
138 	.d_strategy = diskstrategy,
139 	.d_dump = diskdump,
140 	.d_psize = diskpsize,
141 	.d_clone = diskclone
142 };
143 
144 static struct objcache 	*disk_msg_cache;
145 
146 struct objcache_malloc_args disk_msg_malloc_args = {
147 	sizeof(struct disk_msg), M_DISK };
148 
149 static struct lwkt_port disk_dispose_port;
150 static struct lwkt_port disk_msg_port;
151 
152 
153 static int
154 disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe)
155 {
156 	struct disk_info *info = &dp->d_info;
157 	struct diskslice *sp = &dp->d_slice->dss_slices[slice];
158 	disklabel_ops_t ops;
159 	struct partinfo part;
160 	const char *msg;
161 	cdev_t ndev;
162 	unsigned long i;
163 
164 	//lp.opaque = NULL;
165 
166 	ops = &disklabel32_ops;
167 	msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
168 	if (msg && !strcmp(msg, "no disk label")) {
169 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: trying with disklabel64\n");
170 		ops = &disklabel64_ops;
171 		msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
172 	}
173 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: label: %s\n", (msg)?msg:"is NULL");
174 	if (msg == NULL) {
175 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: found %d partitions in the label\n", ops->op_getnumparts(sp->ds_label));
176 		if (slice != WHOLE_DISK_SLICE)
177 			ops->op_adjust_label_reserved(dp->d_slice, slice, sp);
178 		else
179 			sp->ds_reserved = 0;
180 
181 		sp->ds_ops = ops;
182 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: lp.opaque: %x\n", sp->ds_label.opaque);
183 		for (i = 0; i < ops->op_getnumparts(sp->ds_label); i++) {
184 			ops->op_loadpartinfo(sp->ds_label, i, &part);
185 			devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: partinfo says fstype=%d for part %d\n", part.fstype, i);
186 			if (part.fstype) {
187 				if (reprobe &&
188 					(ndev = devfs_find_device_by_name("%s%c",
189 					dev->si_name, 'a'+ (char)i))) {
190 					/* Device already exists and is still valid */
191 					devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: reprobing and device remained valid, mark it\n");
192 					ndev->si_flags |= SI_REPROBE_TEST;
193 				} else {
194 					ndev = make_dev(&disk_ops,
195 						dkmakeminor(dkunit(dp->d_cdev), slice, i),
196 						UID_ROOT, GID_OPERATOR, 0640,
197 						"%s%c", dev->si_name, 'a'+ (char)i);
198 					ndev->si_disk = dp;
199 					if (dp->d_info.d_serialno) {
200 						make_dev_alias(ndev, "serno/%s.s%d%c", dp->d_info.d_serialno, slice - 1, 'a' + (char)i);
201 					}
202 					ndev->si_flags |= SI_REPROBE_TEST;
203 				}
204 
205 				devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice:end: lp.opaque: %x\n", ndev->si_disk->d_slice->dss_slices[slice].ds_label.opaque);
206 			}
207 		}
208 	} else if (info->d_dsflags & DSO_COMPATLABEL) {
209 		msg = NULL;
210 		if (sp->ds_size >= 0x100000000ULL)
211 			ops = &disklabel64_ops;
212 		else
213 			ops = &disklabel32_ops;
214 		sp->ds_label = ops->op_clone_label(info, sp);
215 	} else {
216 		if (sp->ds_type == DOSPTYP_386BSD /* XXX */)
217 			log(LOG_WARNING, "%s: cannot find label (%s)\n",
218 			    dev->si_name, msg);
219 	}
220 
221 	if (msg == NULL) {
222 		sp->ds_wlabel = FALSE;
223 	}
224 
225 	return (msg ? EINVAL : 0);
226 }
227 
228 
229 static void
230 disk_probe(struct disk *dp, int reprobe)
231 {
232 	struct disk_info *info = &dp->d_info;
233 	cdev_t dev = dp->d_cdev;
234 	cdev_t ndev;
235 	int error, i;
236 
237 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe called for %s\n", dp->d_cdev->si_name);
238 	KKASSERT (info->d_media_blksize != 0);
239 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: info set!\n");
240 
241 	dp->d_slice = dsmakeslicestruct(BASE_SLICE, info);
242 
243 	error = mbrinit(dev, info, &(dp->d_slice));
244 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: &dp->d_slice is: %x, %x\n", &dp->d_slice, dp->d_slice);
245 	if (error != 0) {
246 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: mbrinit() failed with error: %d\n", error);
247 		return;
248 	} else {
249 		devfs_debug(DEVFS_DEBUG_DEBUG, "mbrinit succeeded, found %d slices\n", dp->d_slice->dss_nslices);
250 		if (dp->d_slice->dss_nslices == BASE_SLICE) {
251 			dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_size = info->d_media_blocks;
252 			dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_reserved = 0;
253 			if (reprobe &&
254 				(ndev = devfs_find_device_by_name("%ss%d",
255 				dev->si_name, COMPATIBILITY_SLICE))) {
256 				/* Device already exists and is still valid */
257 				devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: reprobing and device remained valid, mark it\n");
258 				ndev->si_flags |= SI_REPROBE_TEST;
259 			} else {
260 				ndev = make_dev(&disk_ops,
261 					dkmakewholeslice(dkunit(dev), COMPATIBILITY_SLICE),
262 					UID_ROOT, GID_OPERATOR, 0640,
263 					"%ss%d", dev->si_name, COMPATIBILITY_SLICE);
264 
265 				ndev->si_disk = dp;
266 				if (dp->d_info.d_serialno) {
267 					make_dev_alias(ndev, "serno/%s.s%d",
268 						       dp->d_info.d_serialno,
269 						       COMPATIBILITY_SLICE);
270 				}
271 				ndev->si_flags |= SI_REPROBE_TEST;
272 			}
273 
274 			dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_dev = ndev;
275 			devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: type of slice is :%x\n", dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_type );
276 
277 			dp->d_slice->dss_first_bsd_slice = COMPATIBILITY_SLICE;
278 			disk_probe_slice(dp, ndev, COMPATIBILITY_SLICE, reprobe);
279 
280 		}
281 		for (i = BASE_SLICE; i < dp->d_slice->dss_nslices; i++) {
282 			if (dp->d_slice->dss_slices[i].ds_size == 0)
283 				continue;
284 
285 			if (reprobe &&
286 				(ndev = devfs_find_device_by_name("%ss%d",
287 				dev->si_name, i-1))) {
288 				/* Device already exists and is still valid */
289 				devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: reprobing and device remained valid, mark it\n");
290 				ndev->si_flags |= SI_REPROBE_TEST;
291 			} else {
292 				ndev = make_dev(&disk_ops,
293 					dkmakewholeslice(dkunit(dev), i),
294 					UID_ROOT, GID_OPERATOR, 0640,
295 					"%ss%d", dev->si_name, i-1);
296 				if (dp->d_info.d_serialno) {
297 					make_dev_alias(ndev, "serno/%s.s%d",
298 						       dp->d_info.d_serialno,
299 						       i - 1);
300 				}
301 				ndev->si_disk = dp;
302 				ndev->si_flags |= SI_REPROBE_TEST;
303 			}
304 			dp->d_slice->dss_slices[i].ds_reserved = 0;
305 			dp->d_slice->dss_slices[i].ds_dev = ndev;
306 			devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe-> type of slice is :%x\n", dp->d_slice->dss_slices[i].ds_type );
307 			if (dp->d_slice->dss_slices[i].ds_type == DOSPTYP_386BSD) {
308 				if (!dp->d_slice->dss_first_bsd_slice)
309 					dp->d_slice->dss_first_bsd_slice = i;
310 				disk_probe_slice(dp, ndev, i, reprobe);
311 			}
312 		}
313 	}
314 }
315 
316 
317 static void
318 disk_msg_core(void *arg)
319 {
320     uint8_t  run = 1;
321 	struct disk	*dp;
322 	struct diskslice *sp;
323 	lwkt_tokref ilock;
324     disk_msg_t msg;
325 
326 	lwkt_initport_thread(&disk_msg_port, curthread);
327 	wakeup(curthread);
328 
329     while (run) {
330         msg = (disk_msg_t)lwkt_waitport(&disk_msg_port, 0);
331 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_msg_core, new msg: %x\n", (unsigned int)msg->hdr.u.ms_result);
332 
333         switch (msg->hdr.u.ms_result) {
334 
335         case DISK_DISK_PROBE:
336 			dp = (struct disk *)msg->load;
337 			disk_probe(dp, 0);
338 			break;
339 
340 		case DISK_DISK_DESTROY:
341 			dp = (struct disk *)msg->load;
342 			devfs_destroy_subnames(dp->d_cdev->si_name);
343 			devfs_destroy_dev(dp->d_cdev);
344 			lwkt_gettoken(&ilock, &disklist_token);
345 			LIST_REMOVE(dp, d_list);
346 			lwkt_reltoken(&ilock);
347 #if 0
348 			devfs_destroy_dev(dp->d_rawdev); /* XXX: needed? when? */
349 #endif
350 			if (dp->d_info.d_serialno) {
351 				kfree(dp->d_info.d_serialno, M_TEMP);
352 				dp->d_info.d_serialno = NULL;
353 			}
354 			break;
355 
356 		case DISK_UNPROBE:
357 			dp = (struct disk *)msg->load;
358 			devfs_destroy_subnames(dp->d_cdev->si_name);
359 			break;
360 
361 		case DISK_SLICE_REPROBE:
362 			dp = (struct disk *)msg->load;
363 			sp = (struct diskslice *)msg->load2;
364 			devfs_clr_subnames_flag(sp->ds_dev->si_name, SI_REPROBE_TEST);
365 			devfs_debug(DEVFS_DEBUG_DEBUG,
366 				    "DISK_SLICE_REPROBE: %s\n",
367 				    sp->ds_dev->si_name);
368 			disk_probe_slice(dp, sp->ds_dev, dkslice(sp->ds_dev), 1);
369 			devfs_destroy_subnames_without_flag(sp->ds_dev->si_name,
370 												SI_REPROBE_TEST);
371 			break;
372 
373 		case DISK_DISK_REPROBE:
374 			dp = (struct disk *)msg->load;
375 			devfs_clr_subnames_flag(dp->d_cdev->si_name, SI_REPROBE_TEST);
376 			devfs_debug(DEVFS_DEBUG_DEBUG,
377 				    "DISK_DISK_REPROBE: %s\n",
378 				    dp->d_cdev->si_name);
379 			disk_probe(dp, 1);
380 			devfs_destroy_subnames_without_flag(dp->d_cdev->si_name,
381 												SI_REPROBE_TEST);
382 			break;
383 
384 		case DISK_SYNC:
385 			break;
386 
387         default:
388             devfs_debug(DEVFS_DEBUG_WARNING, "disk_msg_core: unknown message received at core\n");
389         }
390 
391         lwkt_replymsg((lwkt_msg_t)msg, 0);
392     }
393 	lwkt_exit();
394 }
395 
396 
397 /**
398  * Acts as a message drain. Any message that is replied to here gets destroyed and
399  * the memory freed.
400  **/
401 static void
402 disk_msg_autofree_reply(lwkt_port_t port, lwkt_msg_t msg)
403 {
404     objcache_put(disk_msg_cache, msg);
405 }
406 
407 
408 void
409 disk_msg_send(uint32_t cmd, void *load, void *load2)
410 {
411     disk_msg_t disk_msg;
412 	lwkt_port_t port = &disk_msg_port;
413 
414     disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
415 
416     lwkt_initmsg(&disk_msg->hdr, &disk_dispose_port, 0);
417 
418 	disk_msg->hdr.u.ms_result = cmd;
419 	disk_msg->load = load;
420 	disk_msg->load2 = load2;
421 	KKASSERT(port);
422     lwkt_sendmsg(port, (lwkt_msg_t)disk_msg);
423 }
424 
425 void
426 disk_msg_send_sync(uint32_t cmd, void *load, void *load2)
427 {
428 	struct lwkt_port rep_port;
429 	disk_msg_t disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
430 	disk_msg_t	msg_incoming;
431 	lwkt_port_t port = &disk_msg_port;
432 
433 	lwkt_initport_thread(&rep_port, curthread);
434 	lwkt_initmsg(&disk_msg->hdr, &rep_port, 0);
435 
436 	disk_msg->hdr.u.ms_result = cmd;
437 	disk_msg->load = load;
438 	disk_msg->load2 = load2;
439 
440 	KKASSERT(port);
441     lwkt_sendmsg(port, (lwkt_msg_t)disk_msg);
442 	msg_incoming = lwkt_waitport(&rep_port, 0);
443 }
444 
445 /*
446  * Create a raw device for the dev_ops template (which is returned).  Also
447  * create a slice and unit managed disk and overload the user visible
448  * device space with it.
449  *
450  * NOTE: The returned raw device is NOT a slice and unit managed device.
451  * It is an actual raw device representing the raw disk as specified by
452  * the passed dev_ops.  The disk layer not only returns such a raw device,
453  * it also uses it internally when passing (modified) commands through.
454  */
455 cdev_t
456 disk_create(int unit, struct disk *dp, struct dev_ops *raw_ops)
457 {
458 	lwkt_tokref ilock;
459 	cdev_t rawdev;
460 
461 	rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
462 			    UID_ROOT, GID_OPERATOR, 0640,
463 			    "%s%d", raw_ops->head.name, unit);
464 
465 
466 	bzero(dp, sizeof(*dp));
467 
468 	dp->d_rawdev = rawdev;
469 	dp->d_raw_ops = raw_ops;
470 	dp->d_dev_ops = &disk_ops;
471 	dp->d_cdev = make_dev(&disk_ops,
472 			    dkmakewholedisk(unit),
473 			    UID_ROOT, GID_OPERATOR, 0640,
474 			    "%s%d", raw_ops->head.name, unit);
475 
476 	dp->d_cdev->si_disk = dp;
477 
478 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_create called for %s\n",
479 			dp->d_cdev->si_name);
480 	lwkt_gettoken(&ilock, &disklist_token);
481 	LIST_INSERT_HEAD(&disklist, dp, d_list);
482 	lwkt_reltoken(&ilock);
483 	return (dp->d_rawdev);
484 }
485 
486 
487 static void
488 _setdiskinfo(struct disk *disk, struct disk_info *info)
489 {
490 	char *oldserialno;
491 
492 	devfs_debug(DEVFS_DEBUG_DEBUG,
493 		    "_setdiskinfo called for disk -1-: %x\n", disk);
494 	oldserialno = disk->d_info.d_serialno;
495 	bcopy(info, &disk->d_info, sizeof(disk->d_info));
496 	info = &disk->d_info;
497 
498 	/*
499 	 * The serial number is duplicated so the caller can throw
500 	 * their copy away.
501 	 */
502 	if (info->d_serialno && info->d_serialno[0]) {
503 		info->d_serialno = kstrdup(info->d_serialno, M_TEMP);
504 		if (disk->d_cdev) {
505 			make_dev_alias(disk->d_cdev, "serno/%s",
506 					info->d_serialno);
507 		}
508 	} else {
509 		info->d_serialno = NULL;
510 	}
511 	if (oldserialno)
512 		kfree(oldserialno, M_TEMP);
513 
514 	/*
515 	 * The caller may set d_media_size or d_media_blocks and we
516 	 * calculate the other.
517 	 */
518 	KKASSERT(info->d_media_size == 0 || info->d_media_blksize == 0);
519 	if (info->d_media_size == 0 && info->d_media_blocks) {
520 		info->d_media_size = (u_int64_t)info->d_media_blocks *
521 				     info->d_media_blksize;
522 	} else if (info->d_media_size && info->d_media_blocks == 0 &&
523 		   info->d_media_blksize) {
524 		info->d_media_blocks = info->d_media_size /
525 				       info->d_media_blksize;
526 	}
527 
528 	/*
529 	 * The si_* fields for rawdev are not set until after the
530 	 * disk_create() call, so someone using the cooked version
531 	 * of the raw device (i.e. da0s0) will not get the right
532 	 * si_iosize_max unless we fix it up here.
533 	 */
534 	if (disk->d_cdev && disk->d_rawdev &&
535 	    disk->d_cdev->si_iosize_max == 0) {
536 		disk->d_cdev->si_iosize_max = disk->d_rawdev->si_iosize_max;
537 		disk->d_cdev->si_bsize_phys = disk->d_rawdev->si_bsize_phys;
538 		disk->d_cdev->si_bsize_best = disk->d_rawdev->si_bsize_best;
539 	}
540 }
541 
542 /*
543  * Disk drivers must call this routine when media parameters are available
544  * or have changed.
545  */
546 void
547 disk_setdiskinfo(struct disk *disk, struct disk_info *info)
548 {
549 	_setdiskinfo(disk, info);
550 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_setdiskinfo called for disk -2-: %x\n", disk);
551 	disk_msg_send(DISK_DISK_PROBE, disk, NULL);
552 }
553 
554 void
555 disk_setdiskinfo_sync(struct disk *disk, struct disk_info *info)
556 {
557 	_setdiskinfo(disk, info);
558 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_setdiskinfo_sync called for disk -2-: %x\n", disk);
559 	disk_msg_send_sync(DISK_DISK_PROBE, disk, NULL);
560 }
561 
562 /*
563  * This routine is called when an adapter detaches.  The higher level
564  * managed disk device is destroyed while the lower level raw device is
565  * released.
566  */
567 void
568 disk_destroy(struct disk *disk)
569 {
570 	disk_msg_send_sync(DISK_DISK_DESTROY, disk, NULL);
571 	return;
572 }
573 
574 int
575 disk_dumpcheck(cdev_t dev, u_int64_t *count, u_int64_t *blkno, u_int *secsize)
576 {
577 	struct partinfo pinfo;
578 	int error;
579 
580 	bzero(&pinfo, sizeof(pinfo));
581 	error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0, proc0.p_ucred);
582 	if (error)
583 		return (error);
584 	if (pinfo.media_blksize == 0)
585 		return (ENXIO);
586 	*count = (u_int64_t)Maxmem * PAGE_SIZE / pinfo.media_blksize;
587 	if (dumplo64 < pinfo.reserved_blocks ||
588 	    dumplo64 + *count > pinfo.media_blocks) {
589 		return (ENOSPC);
590 	}
591 	*blkno = dumplo64 + pinfo.media_offset / pinfo.media_blksize;
592 	*secsize = pinfo.media_blksize;
593 	return (0);
594 }
595 
596 void
597 disk_unprobe(struct disk *disk)
598 {
599 	if (disk == NULL)
600 		return;
601 
602 	disk_msg_send_sync(DISK_UNPROBE, disk, NULL);
603 }
604 
605 void
606 disk_invalidate (struct disk *disk)
607 {
608 	devfs_debug(DEVFS_DEBUG_INFO, "disk_invalidate for %s\n", disk->d_cdev->si_name);
609 	if (disk->d_slice)
610 		dsgone(&disk->d_slice);
611 }
612 
613 struct disk *
614 disk_enumerate(struct disk *disk)
615 {
616 	struct disk *dp;
617 	lwkt_tokref ilock;
618 
619 	lwkt_gettoken(&ilock, &disklist_token);
620 	if (!disk)
621 		dp = (LIST_FIRST(&disklist));
622 	else
623 		dp = (LIST_NEXT(disk, d_list));
624 	lwkt_reltoken(&ilock);
625 
626 	return dp;
627 }
628 
629 static
630 int
631 sysctl_disks(SYSCTL_HANDLER_ARGS)
632 {
633 	struct disk *disk;
634 	int error, first;
635 
636 	disk = NULL;
637 	first = 1;
638 
639 	while ((disk = disk_enumerate(disk))) {
640 		if (!first) {
641 			error = SYSCTL_OUT(req, " ", 1);
642 			if (error)
643 				return error;
644 		} else {
645 			first = 0;
646 		}
647 		error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
648 				   strlen(disk->d_rawdev->si_name));
649 		if (error)
650 			return error;
651 	}
652 	error = SYSCTL_OUT(req, "", 1);
653 	return error;
654 }
655 
656 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, NULL, 0,
657     sysctl_disks, "A", "names of available disks");
658 
659 /*
660  * Open a disk device or partition.
661  */
662 static
663 int
664 diskopen(struct dev_open_args *ap)
665 {
666 	cdev_t dev = ap->a_head.a_dev;
667 	struct disk *dp;
668 	int error;
669 
670 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskopen: name is %s\n", dev->si_name);
671 
672 	/*
673 	 * dp can't be NULL here XXX.
674 	 */
675 	dp = dev->si_disk;
676 	if (dp == NULL)
677 		return (ENXIO);
678 	error = 0;
679 
680 	/*
681 	 * Deal with open races
682 	 */
683 	while (dp->d_flags & DISKFLAG_LOCK) {
684 		dp->d_flags |= DISKFLAG_WANTED;
685 		error = tsleep(dp, PCATCH, "diskopen", hz);
686 		if (error)
687 			return (error);
688 	}
689 	dp->d_flags |= DISKFLAG_LOCK;
690 
691 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskopen: -2- name is %s\n", dev->si_name);
692 
693 	/*
694 	 * Open the underlying raw device.
695 	 */
696 	if (!dsisopen(dp->d_slice)) {
697 #if 0
698 		if (!pdev->si_iosize_max)
699 			pdev->si_iosize_max = dev->si_iosize_max;
700 #endif
701 		error = dev_dopen(dp->d_rawdev, ap->a_oflags,
702 				  ap->a_devtype, ap->a_cred);
703 	}
704 #if 0
705 	/*
706 	 * Inherit properties from the underlying device now that it is
707 	 * open.
708 	 */
709 	dev_dclone(dev);
710 #endif
711 
712 	if (error)
713 		goto out;
714 	error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags,
715 		       &dp->d_slice, &dp->d_info);
716 	if (!dsisopen(dp->d_slice)) {
717 		dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype);
718 	}
719 out:
720 	dp->d_flags &= ~DISKFLAG_LOCK;
721 	if (dp->d_flags & DISKFLAG_WANTED) {
722 		dp->d_flags &= ~DISKFLAG_WANTED;
723 		wakeup(dp);
724 	}
725 
726 	return(error);
727 }
728 
729 /*
730  * Close a disk device or partition
731  */
732 static
733 int
734 diskclose(struct dev_close_args *ap)
735 {
736 	cdev_t dev = ap->a_head.a_dev;
737 	struct disk *dp;
738 	int error;
739 
740 	error = 0;
741 	dp = dev->si_disk;
742 
743 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskclose: name %s\n", dev->si_name);
744 
745 	dsclose(dev, ap->a_devtype, dp->d_slice);
746 	if (!dsisopen(dp->d_slice)) {
747 		devfs_debug(DEVFS_DEBUG_DEBUG, "diskclose is closing underlying device\n");
748 		error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype);
749 	}
750 	return (error);
751 }
752 
753 /*
754  * First execute the ioctl on the disk device, and if it isn't supported
755  * try running it on the backing device.
756  */
757 static
758 int
759 diskioctl(struct dev_ioctl_args *ap)
760 {
761 	cdev_t dev = ap->a_head.a_dev;
762 	struct disk *dp;
763 	int error;
764 
765 	dp = dev->si_disk;
766 	if (dp == NULL)
767 		return (ENXIO);
768 
769 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl: cmd is: %x (name: %s)\n", ap->a_cmd, dev->si_name);
770 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl: &dp->d_slice is: %x, %x\n", &dp->d_slice, dp->d_slice);
771 
772 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl:1: says lp.opaque is: %x\n", dp->d_slice->dss_slices[0].ds_label.opaque);
773 
774 	error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag,
775 			&dp->d_slice, &dp->d_info);
776 
777 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl:2: says lp.opaque is: %x\n", dp->d_slice->dss_slices[0].ds_label.opaque);
778 
779 	if (error == ENOIOCTL) {
780 		devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl: going for dev_dioctl instead!\n");
781 		error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data,
782 				   ap->a_fflag, ap->a_cred);
783 	}
784 	return (error);
785 }
786 
787 /*
788  * Execute strategy routine
789  */
790 static
791 int
792 diskstrategy(struct dev_strategy_args *ap)
793 {
794 	cdev_t dev = ap->a_head.a_dev;
795 	struct bio *bio = ap->a_bio;
796 	struct bio *nbio;
797 	struct disk *dp;
798 
799 	dp = dev->si_disk;
800 
801 	if (dp == NULL) {
802 		bio->bio_buf->b_error = ENXIO;
803 		bio->bio_buf->b_flags |= B_ERROR;
804 		biodone(bio);
805 		return(0);
806 	}
807 	KKASSERT(dev->si_disk == dp);
808 
809 	/*
810 	 * The dscheck() function will also transform the slice relative
811 	 * block number i.e. bio->bio_offset into a block number that can be
812 	 * passed directly to the underlying raw device.  If dscheck()
813 	 * returns NULL it will have handled the bio for us (e.g. EOF
814 	 * or error due to being beyond the device size).
815 	 */
816 	if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL) {
817 		dev_dstrategy(dp->d_rawdev, nbio);
818 	} else {
819 		devfs_debug(DEVFS_DEBUG_DEBUG, "diskstrategy: dscheck NULL!!! biodone time!\n");
820 		biodone(bio);
821 	}
822 	return(0);
823 }
824 
825 /*
826  * Return the partition size in ?blocks?
827  */
828 static
829 int
830 diskpsize(struct dev_psize_args *ap)
831 {
832 	cdev_t dev = ap->a_head.a_dev;
833 	struct disk *dp;
834 
835 	dp = dev->si_disk;
836 	if (dp == NULL)
837 		return(ENODEV);
838 	ap->a_result = dssize(dev, &dp->d_slice);
839 	return(0);
840 }
841 
842 /*
843  * When new device entries are instantiated, make sure they inherit our
844  * si_disk structure and block and iosize limits from the raw device.
845  *
846  * This routine is always called synchronously in the context of the
847  * client.
848  *
849  * XXX The various io and block size constraints are not always initialized
850  * properly by devices.
851  */
852 static
853 int
854 diskclone(struct dev_clone_args *ap)
855 {
856 	cdev_t dev = ap->a_head.a_dev;
857 	struct disk *dp;
858 	dp = dev->si_disk;
859 
860 	KKASSERT(dp != NULL);
861 	dev->si_disk = dp;
862 	dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
863 	dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
864 	dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
865 	return(0);
866 }
867 
868 int
869 diskdump(struct dev_dump_args *ap)
870 {
871 	cdev_t dev = ap->a_head.a_dev;
872 	struct disk *dp = dev->si_disk;
873 	int error;
874 
875 	error = disk_dumpcheck(dev, &ap->a_count, &ap->a_blkno, &ap->a_secsize);
876 	if (error == 0) {
877 		ap->a_head.a_dev = dp->d_rawdev;
878 		error = dev_doperate(&ap->a_head);
879 	}
880 
881 	return(error);
882 }
883 
884 
885 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
886     0, sizeof(struct diskslices), "sizeof(struct diskslices)");
887 
888 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
889     0, sizeof(struct disk), "sizeof(struct disk)");
890 
891 
892 /*
893  * Seek sort for disks.
894  *
895  * The bio_queue keep two queues, sorted in ascending block order.  The first
896  * queue holds those requests which are positioned after the current block
897  * (in the first request); the second, which starts at queue->switch_point,
898  * holds requests which came in after their block number was passed.  Thus
899  * we implement a one way scan, retracting after reaching the end of the drive
900  * to the first request on the second queue, at which time it becomes the
901  * first queue.
902  *
903  * A one-way scan is natural because of the way UNIX read-ahead blocks are
904  * allocated.
905  */
906 void
907 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
908 {
909 	struct bio *bq;
910 	struct bio *bn;
911 	struct bio *be;
912 
913 	be = TAILQ_LAST(&bioq->queue, bio_queue);
914 	/*
915 	 * If the queue is empty or we are an
916 	 * ordered transaction, then it's easy.
917 	 */
918 	if ((bq = bioq_first(bioq)) == NULL ||
919 	    (bio->bio_buf->b_flags & B_ORDERED) != 0) {
920 		bioq_insert_tail(bioq, bio);
921 		return;
922 	} else if (bioq->insert_point != NULL) {
923 
924 		/*
925 		 * A certain portion of the list is
926 		 * "locked" to preserve ordering, so
927 		 * we can only insert after the insert
928 		 * point.
929 		 */
930 		bq = bioq->insert_point;
931 	} else {
932 
933 		/*
934 		 * If we lie before the last removed (currently active)
935 		 * request, and are not inserting ourselves into the
936 		 * "locked" portion of the list, then we must add ourselves
937 		 * to the second request list.
938 		 */
939 		if (bio->bio_offset < bioq->last_offset) {
940 			bq = bioq->switch_point;
941 			/*
942 			 * If we are starting a new secondary list,
943 			 * then it's easy.
944 			 */
945 			if (bq == NULL) {
946 				bioq->switch_point = bio;
947 				bioq_insert_tail(bioq, bio);
948 				return;
949 			}
950 			/*
951 			 * If we lie ahead of the current switch point,
952 			 * insert us before the switch point and move
953 			 * the switch point.
954 			 */
955 			if (bio->bio_offset < bq->bio_offset) {
956 				bioq->switch_point = bio;
957 				TAILQ_INSERT_BEFORE(bq, bio, bio_act);
958 				return;
959 			}
960 		} else {
961 			if (bioq->switch_point != NULL)
962 				be = TAILQ_PREV(bioq->switch_point,
963 						bio_queue, bio_act);
964 			/*
965 			 * If we lie between last_offset and bq,
966 			 * insert before bq.
967 			 */
968 			if (bio->bio_offset < bq->bio_offset) {
969 				TAILQ_INSERT_BEFORE(bq, bio, bio_act);
970 				return;
971 			}
972 		}
973 	}
974 
975 	/*
976 	 * Request is at/after our current position in the list.
977 	 * Optimize for sequential I/O by seeing if we go at the tail.
978 	 */
979 	if (bio->bio_offset > be->bio_offset) {
980 		TAILQ_INSERT_AFTER(&bioq->queue, be, bio, bio_act);
981 		return;
982 	}
983 
984 	/* Otherwise, insertion sort */
985 	while ((bn = TAILQ_NEXT(bq, bio_act)) != NULL) {
986 
987 		/*
988 		 * We want to go after the current request if it is the end
989 		 * of the first request list, or if the next request is a
990 		 * larger cylinder than our request.
991 		 */
992 		if (bn == bioq->switch_point
993 		 || bio->bio_offset < bn->bio_offset)
994 			break;
995 		bq = bn;
996 	}
997 	TAILQ_INSERT_AFTER(&bioq->queue, bq, bio, bio_act);
998 }
999 
1000 /*
1001  * Disk error is the preface to plaintive error messages
1002  * about failing disk transfers.  It prints messages of the form
1003 
1004 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
1005 
1006  * if the offset of the error in the transfer and a disk label
1007  * are both available.  blkdone should be -1 if the position of the error
1008  * is unknown; the disklabel pointer may be null from drivers that have not
1009  * been converted to use them.  The message is printed with kprintf
1010  * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
1011  * The message should be completed (with at least a newline) with kprintf
1012  * or log(-1, ...), respectively.  There is no trailing space.
1013  */
1014 void
1015 diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt)
1016 {
1017 	struct buf *bp = bio->bio_buf;
1018 	const char *term;
1019 
1020 	switch(bp->b_cmd) {
1021 	case BUF_CMD_READ:
1022 		term = "read";
1023 		break;
1024 	case BUF_CMD_WRITE:
1025 		term = "write";
1026 		break;
1027 	default:
1028 		term = "access";
1029 		break;
1030 	}
1031 	//sname = dsname(dev, unit, slice, part, partname);
1032 	kprintf("%s: %s %sing ", dev->si_name, what, term);
1033 	kprintf("offset %012llx for %d",
1034 		(long long)bio->bio_offset,
1035 		bp->b_bcount);
1036 
1037 	if (donecnt)
1038 		kprintf(" (%d bytes completed)", donecnt);
1039 }
1040 
1041 /*
1042  * Locate a disk device
1043  */
1044 cdev_t
1045 disk_locate(const char *devname)
1046 {
1047 	return devfs_find_device_by_name(devname);
1048 }
1049 
1050 
1051 void
1052 disk_config(void *arg)
1053 {
1054 	disk_msg_send_sync(DISK_SYNC, NULL, NULL);
1055 }
1056 
1057 
1058 static void
1059 disk_init(void)
1060 {
1061 	struct thread* td_core;
1062 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_init() called\n");
1063 
1064     disk_msg_cache = objcache_create("disk-msg-cache", 0, 0,
1065 			NULL, NULL, NULL,
1066 			objcache_malloc_alloc,
1067 			objcache_malloc_free,
1068 			&disk_msg_malloc_args );
1069 
1070 	lwkt_token_init(&disklist_token);
1071 
1072 	/* Initialize the reply-only port which acts as a message drain */
1073 	lwkt_initport_replyonly(&disk_dispose_port, disk_msg_autofree_reply);
1074 
1075 	lwkt_create(disk_msg_core, /*args*/NULL, &td_core, NULL,
1076 		    0, 0, "disk_msg_core");
1077 
1078 	tsleep(td_core, 0, "diskcore", 0);
1079 }
1080 
1081 
1082 static void
1083 disk_uninit(void)
1084 {
1085 	devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_uninit() called\n");
1086 
1087 	objcache_destroy(disk_msg_cache);
1088 
1089 }
1090 
1091 
1092 SYSINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_FIRST, disk_init, NULL);
1093 SYSUNINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_ANY, disk_uninit, NULL);
1094