xref: /dflybsd-src/sys/kern/subr_disk.c (revision 6507240b2fcfebaacc0f92f997dad76922e1d8c0)
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 
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 
130 static struct dev_ops disk_ops = {
131 	{ "disk", 0, D_DISK },
132 	.d_open = diskopen,
133 	.d_close = diskclose,
134 	.d_read = physread,
135 	.d_write = physwrite,
136 	.d_ioctl = diskioctl,
137 	.d_strategy = diskstrategy,
138 	.d_dump = diskdump,
139 	.d_psize = diskpsize,
140 	.d_clone = diskclone
141 };
142 
143 static struct objcache 	*disk_msg_cache;
144 
145 struct objcache_malloc_args disk_msg_malloc_args = {
146 	sizeof(struct disk_msg), M_DISK };
147 
148 static struct lwkt_port disk_dispose_port;
149 static struct lwkt_port disk_msg_port;
150 
151 
152 static int
153 disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe)
154 {
155 	struct disk_info *info = &dp->d_info;
156 	struct diskslice *sp = &dp->d_slice->dss_slices[slice];
157 	disklabel_ops_t ops;
158 	struct partinfo part;
159 	const char *msg;
160 	cdev_t ndev;
161 	unsigned long i;
162 
163 	//lp.opaque = NULL;
164 
165 	ops = &disklabel32_ops;
166 	msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
167 	if (msg && !strcmp(msg, "no disk label")) {
168 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: trying with disklabel64\n");
169 		ops = &disklabel64_ops;
170 		msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
171 	}
172 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: label: %s\n", (msg)?msg:"is NULL");
173 	if (msg == NULL) {
174 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: found %d partitions in the label\n", ops->op_getnumparts(sp->ds_label));
175 		if (slice != WHOLE_DISK_SLICE)
176 			ops->op_adjust_label_reserved(dp->d_slice, slice, sp);
177 		else
178 			sp->ds_reserved = 0;
179 
180 		sp->ds_ops = ops;
181 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: lp.opaque: %x\n", sp->ds_label.opaque);
182 		for (i = 0; i < ops->op_getnumparts(sp->ds_label); i++) {
183 			ops->op_loadpartinfo(sp->ds_label, i, &part);
184 			devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: partinfo says fstype=%d for part %d\n", part.fstype, i);
185 			if (part.fstype) {
186 				if (reprobe &&
187 					(ndev = devfs_find_device_by_name("%s%c",
188 					dev->si_name, 'a'+ (char)i))) {
189 					/* Device already exists and is still valid */
190 					devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: reprobing and device remained valid, mark it\n");
191 					ndev->si_flags |= SI_REPROBE_TEST;
192 				} else {
193 					ndev = make_dev(&disk_ops,
194 						dkmakeminor(dkunit(dp->d_cdev), slice, i),
195 						UID_ROOT, GID_OPERATOR, 0640,
196 						"%s%c", dev->si_name, 'a'+ (char)i);
197 					ndev->si_disk = dp;
198 					if (dp->d_info.d_serialno) {
199 						make_dev_alias(ndev, "serno/%s.s%d%c", dp->d_info.d_serialno, slice - 1, 'a' + (char)i);
200 					}
201 					ndev->si_flags |= SI_REPROBE_TEST;
202 				}
203 
204 				devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice:end: lp.opaque: %x\n", ndev->si_disk->d_slice->dss_slices[slice].ds_label.opaque);
205 			}
206 		}
207 	} else if (info->d_dsflags & DSO_COMPATLABEL) {
208 		msg = NULL;
209 		if (sp->ds_size >= 0x100000000ULL)
210 			ops = &disklabel64_ops;
211 		else
212 			ops = &disklabel32_ops;
213 		sp->ds_label = ops->op_clone_label(info, sp);
214 	} else {
215 		if (sp->ds_type == DOSPTYP_386BSD /* XXX */)
216 			log(LOG_WARNING, "%s: cannot find label (%s)\n",
217 			    dev->si_name, msg);
218 	}
219 
220 	if (msg == NULL) {
221 		sp->ds_wlabel = FALSE;
222 	}
223 
224 	return (msg ? EINVAL : 0);
225 }
226 
227 
228 static void
229 disk_probe(struct disk *dp, int reprobe)
230 {
231 	struct disk_info *info = &dp->d_info;
232 	cdev_t dev = dp->d_cdev;
233 	cdev_t ndev;
234 	int error, i;
235 
236 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe called for %s\n", dp->d_cdev->si_name);
237 	KKASSERT (info->d_media_blksize != 0);
238 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: info set!\n");
239 
240 	dp->d_slice = dsmakeslicestruct(BASE_SLICE, info);
241 
242 	error = mbrinit(dev, info, &(dp->d_slice));
243 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: &dp->d_slice is: %x, %x\n", &dp->d_slice, dp->d_slice);
244 	if (error != 0) {
245 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: mbrinit() failed with error: %d\n", error);
246 		return;
247 	} else {
248 		devfs_debug(DEVFS_DEBUG_DEBUG, "mbrinit succeeded, found %d slices\n", dp->d_slice->dss_nslices);
249 		if (dp->d_slice->dss_nslices == BASE_SLICE) {
250 			dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_size = info->d_media_blocks;
251 			dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_reserved = 0;
252 			if (reprobe &&
253 				(ndev = devfs_find_device_by_name("%ss%d",
254 				dev->si_name, COMPATIBILITY_SLICE))) {
255 				/* Device already exists and is still valid */
256 				devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: reprobing and device remained valid, mark it\n");
257 				ndev->si_flags |= SI_REPROBE_TEST;
258 			} else {
259 				ndev = make_dev(&disk_ops,
260 					dkmakewholeslice(dkunit(dev), COMPATIBILITY_SLICE),
261 					UID_ROOT, GID_OPERATOR, 0640,
262 					"%ss%d", dev->si_name, COMPATIBILITY_SLICE);
263 
264 				ndev->si_disk = dp;
265 				if (dp->d_info.d_serialno) {
266 					make_dev_alias(ndev, "serno/%s.s%d",
267 						       dp->d_info.d_serialno,
268 						       COMPATIBILITY_SLICE);
269 				}
270 				ndev->si_flags |= SI_REPROBE_TEST;
271 			}
272 
273 			dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_dev = ndev;
274 			devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: type of slice is :%x\n", dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_type );
275 
276 			dp->d_slice->dss_first_bsd_slice = COMPATIBILITY_SLICE;
277 			disk_probe_slice(dp, ndev, COMPATIBILITY_SLICE, reprobe);
278 
279 		}
280 		for (i = BASE_SLICE; i < dp->d_slice->dss_nslices; i++) {
281 			if (reprobe &&
282 				(ndev = devfs_find_device_by_name("%ss%d",
283 				dev->si_name, i-1))) {
284 				/* Device already exists and is still valid */
285 				devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: reprobing and device remained valid, mark it\n");
286 				ndev->si_flags |= SI_REPROBE_TEST;
287 			} else {
288 				ndev = make_dev(&disk_ops,
289 					dkmakewholeslice(dkunit(dev), i),
290 					UID_ROOT, GID_OPERATOR, 0640,
291 					"%ss%d", dev->si_name, i-1);
292 				if (dp->d_info.d_serialno) {
293 					make_dev_alias(ndev, "serno/%s.s%d",
294 						       dp->d_info.d_serialno,
295 						       i - 1);
296 				}
297 				ndev->si_disk = dp;
298 				ndev->si_flags |= SI_REPROBE_TEST;
299 			}
300 			dp->d_slice->dss_slices[i].ds_reserved = 0;
301 			dp->d_slice->dss_slices[i].ds_dev = ndev;
302 			devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe-> type of slice is :%x\n", dp->d_slice->dss_slices[i].ds_type );
303 			if (dp->d_slice->dss_slices[i].ds_type == DOSPTYP_386BSD) {
304 				if (!dp->d_slice->dss_first_bsd_slice)
305 					dp->d_slice->dss_first_bsd_slice = i;
306 				disk_probe_slice(dp, ndev, i, reprobe);
307 			}
308 		}
309 	}
310 }
311 
312 
313 static void
314 disk_msg_core(void *arg)
315 {
316     uint8_t  run = 1;
317 	struct disk	*dp;
318 	struct diskslice *sp;
319     disk_msg_t msg;
320 
321 
322 	lwkt_initport_thread(&disk_msg_port, curthread);
323 	wakeup(curthread);
324 
325     while (run) {
326         msg = (disk_msg_t)lwkt_waitport(&disk_msg_port, 0);
327 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_msg_core, new msg: %x\n", (unsigned int)msg->hdr.u.ms_result);
328 
329         switch (msg->hdr.u.ms_result) {
330 
331         case DISK_DISK_PROBE:
332 			dp = (struct disk *)msg->load;
333 			disk_probe(dp, 0);
334 			break;
335 
336 		case DISK_DISK_DESTROY:
337 			dp = (struct disk *)msg->load;
338 			devfs_destroy_subnames(dp->d_cdev->si_name);
339 			devfs_destroy_dev(dp->d_cdev);
340 			LIST_REMOVE(dp, d_list);
341 			//devfs_destroy_dev(dp->d_rawdev); //XXX: needed? when?
342 			if (dp->d_info.d_serialno) {
343 				kfree(dp->d_info.d_serialno, M_TEMP);
344 				dp->d_info.d_serialno = NULL;
345 			}
346 			break;
347 
348 		case DISK_UNPROBE:
349 			dp = (struct disk *)msg->load;
350 			devfs_destroy_subnames(dp->d_cdev->si_name);
351 			break;
352 
353 		case DISK_SLICE_REPROBE:
354 			dp = (struct disk *)msg->load;
355 			sp = (struct diskslice *)msg->load2;
356 			devfs_clr_subnames_flag(sp->ds_dev->si_name, SI_REPROBE_TEST);
357 			devfs_debug(DEVFS_DEBUG_DEBUG,
358 				    "DISK_SLICE_REPROBE: %s\n",
359 				    sp->ds_dev->si_name);
360 			disk_probe_slice(dp, sp->ds_dev, dkslice(sp->ds_dev), 1);
361 			devfs_destroy_subnames_without_flag(sp->ds_dev->si_name,
362 												SI_REPROBE_TEST);
363 			break;
364 
365 		case DISK_DISK_REPROBE:
366 			dp = (struct disk *)msg->load;
367 			devfs_clr_subnames_flag(dp->d_cdev->si_name, SI_REPROBE_TEST);
368 			devfs_debug(DEVFS_DEBUG_DEBUG,
369 				    "DISK_DISK_REPROBE: %s\n",
370 				    dp->d_cdev->si_name);
371 			disk_probe(dp, 1);
372 			devfs_destroy_subnames_without_flag(dp->d_cdev->si_name,
373 												SI_REPROBE_TEST);
374 			break;
375 
376 		case DISK_SYNC:
377 			break;
378 
379         default:
380             devfs_debug(DEVFS_DEBUG_WARNING, "disk_msg_core: unknown message received at core\n");
381         }
382 
383         lwkt_replymsg((lwkt_msg_t)msg, 0);
384     }
385 	lwkt_exit();
386 }
387 
388 
389 /**
390  * Acts as a message drain. Any message that is replied to here gets destroyed and
391  * the memory freed.
392  **/
393 static void
394 disk_msg_autofree_reply(lwkt_port_t port, lwkt_msg_t msg)
395 {
396     objcache_put(disk_msg_cache, msg);
397 }
398 
399 
400 void
401 disk_msg_send(uint32_t cmd, void *load, void *load2)
402 {
403     disk_msg_t disk_msg;
404 	lwkt_port_t port = &disk_msg_port;
405 
406     disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
407 
408     lwkt_initmsg(&disk_msg->hdr, &disk_dispose_port, 0);
409 
410 	disk_msg->hdr.u.ms_result = cmd;
411 	disk_msg->load = load;
412 	disk_msg->load2 = load2;
413 	KKASSERT(port);
414     lwkt_sendmsg(port, (lwkt_msg_t)disk_msg);
415 }
416 
417 void
418 disk_msg_send_sync(uint32_t cmd, void *load, void *load2)
419 {
420 	struct lwkt_port rep_port;
421 	disk_msg_t disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
422 	disk_msg_t	msg_incoming;
423 	lwkt_port_t port = &disk_msg_port;
424 
425 	lwkt_initport_thread(&rep_port, curthread);
426 	lwkt_initmsg(&disk_msg->hdr, &rep_port, 0);
427 
428 	disk_msg->hdr.u.ms_result = cmd;
429 	disk_msg->load = load;
430 	disk_msg->load2 = load2;
431 
432 	KKASSERT(port);
433     lwkt_sendmsg(port, (lwkt_msg_t)disk_msg);
434 	msg_incoming = lwkt_waitport(&rep_port, 0);
435 }
436 
437 /*
438  * Create a raw device for the dev_ops template (which is returned).  Also
439  * create a slice and unit managed disk and overload the user visible
440  * device space with it.
441  *
442  * NOTE: The returned raw device is NOT a slice and unit managed device.
443  * It is an actual raw device representing the raw disk as specified by
444  * the passed dev_ops.  The disk layer not only returns such a raw device,
445  * it also uses it internally when passing (modified) commands through.
446  */
447 cdev_t
448 disk_create(int unit, struct disk *dp, struct dev_ops *raw_ops)
449 {
450 	cdev_t rawdev;
451 
452 	rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
453 			    UID_ROOT, GID_OPERATOR, 0640,
454 			    "%s%d", raw_ops->head.name, unit);
455 
456 
457 	bzero(dp, sizeof(*dp));
458 
459 	dp->d_rawdev = rawdev;
460 	dp->d_raw_ops = raw_ops;
461 	dp->d_dev_ops = &disk_ops;
462 	dp->d_cdev = make_dev(&disk_ops,
463 			    dkmakewholedisk(unit),
464 			    UID_ROOT, GID_OPERATOR, 0640,
465 			    "%s%d", raw_ops->head.name, unit);
466 
467 	dp->d_cdev->si_disk = dp;
468 
469 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_create called for %s\n", dp->d_cdev->si_name);
470 	LIST_INSERT_HEAD(&disklist, dp, d_list);
471 	return (dp->d_rawdev);
472 }
473 
474 
475 static void
476 _setdiskinfo(struct disk *disk, struct disk_info *info)
477 {
478 	char *oldserialno;
479 
480 	devfs_debug(DEVFS_DEBUG_DEBUG,
481 		    "_setdiskinfo called for disk -1-: %x\n", disk);
482 	oldserialno = disk->d_info.d_serialno;
483 	bcopy(info, &disk->d_info, sizeof(disk->d_info));
484 	info = &disk->d_info;
485 
486 	/*
487 	 * The serial number is duplicated so the caller can throw
488 	 * their copy away.
489 	 */
490 	if (info->d_serialno && info->d_serialno[0]) {
491 		info->d_serialno = kstrdup(info->d_serialno, M_TEMP);
492 		if (disk->d_cdev) {
493 			make_dev_alias(disk->d_cdev, "serno/%s",
494 					info->d_serialno);
495 		}
496 	} else {
497 		info->d_serialno = NULL;
498 	}
499 	if (oldserialno)
500 		kfree(oldserialno, M_TEMP);
501 
502 	/*
503 	 * The caller may set d_media_size or d_media_blocks and we
504 	 * calculate the other.
505 	 */
506 	KKASSERT(info->d_media_size == 0 || info->d_media_blksize == 0);
507 	if (info->d_media_size == 0 && info->d_media_blocks) {
508 		info->d_media_size = (u_int64_t)info->d_media_blocks *
509 				     info->d_media_blksize;
510 	} else if (info->d_media_size && info->d_media_blocks == 0 &&
511 		   info->d_media_blksize) {
512 		info->d_media_blocks = info->d_media_size /
513 				       info->d_media_blksize;
514 	}
515 
516 	/*
517 	 * The si_* fields for rawdev are not set until after the
518 	 * disk_create() call, so someone using the cooked version
519 	 * of the raw device (i.e. da0s0) will not get the right
520 	 * si_iosize_max unless we fix it up here.
521 	 */
522 	if (disk->d_cdev && disk->d_rawdev &&
523 	    disk->d_cdev->si_iosize_max == 0) {
524 		disk->d_cdev->si_iosize_max = disk->d_rawdev->si_iosize_max;
525 		disk->d_cdev->si_bsize_phys = disk->d_rawdev->si_bsize_phys;
526 		disk->d_cdev->si_bsize_best = disk->d_rawdev->si_bsize_best;
527 	}
528 }
529 
530 /*
531  * Disk drivers must call this routine when media parameters are available
532  * or have changed.
533  */
534 void
535 disk_setdiskinfo(struct disk *disk, struct disk_info *info)
536 {
537 	_setdiskinfo(disk, info);
538 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_setdiskinfo called for disk -2-: %x\n", disk);
539 	disk_msg_send(DISK_DISK_PROBE, disk, NULL);
540 }
541 
542 void
543 disk_setdiskinfo_sync(struct disk *disk, struct disk_info *info)
544 {
545 	_setdiskinfo(disk, info);
546 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_setdiskinfo_sync called for disk -2-: %x\n", disk);
547 	disk_msg_send_sync(DISK_DISK_PROBE, disk, NULL);
548 }
549 
550 /*
551  * This routine is called when an adapter detaches.  The higher level
552  * managed disk device is destroyed while the lower level raw device is
553  * released.
554  */
555 void
556 disk_destroy(struct disk *disk)
557 {
558 	disk_msg_send_sync(DISK_DISK_DESTROY, disk, NULL);
559 	return;
560 }
561 
562 int
563 disk_dumpcheck(cdev_t dev, u_int64_t *count, u_int64_t *blkno, u_int *secsize)
564 {
565 	struct partinfo pinfo;
566 	int error;
567 
568 	bzero(&pinfo, sizeof(pinfo));
569 	error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0, proc0.p_ucred);
570 	if (error)
571 		return (error);
572 	if (pinfo.media_blksize == 0)
573 		return (ENXIO);
574 	*count = (u_int64_t)Maxmem * PAGE_SIZE / pinfo.media_blksize;
575 	if (dumplo64 < pinfo.reserved_blocks ||
576 	    dumplo64 + *count > pinfo.media_blocks) {
577 		return (ENOSPC);
578 	}
579 	*blkno = dumplo64 + pinfo.media_offset / pinfo.media_blksize;
580 	*secsize = pinfo.media_blksize;
581 	return (0);
582 }
583 
584 void
585 disk_unprobe(struct disk *disk)
586 {
587 	if (disk == NULL)
588 		return;
589 
590 	disk_msg_send_sync(DISK_UNPROBE, disk, NULL);
591 }
592 
593 void
594 disk_invalidate (struct disk *disk)
595 {
596 	devfs_debug(DEVFS_DEBUG_INFO, "disk_invalidate for %s\n", disk->d_cdev->si_name);
597 	if (disk->d_slice)
598 		dsgone(&disk->d_slice);
599 }
600 
601 struct disk *
602 disk_enumerate(struct disk *disk)
603 {
604 	if (!disk)
605 		return (LIST_FIRST(&disklist));
606 	else
607 		return (LIST_NEXT(disk, d_list));
608 }
609 
610 static
611 int
612 sysctl_disks(SYSCTL_HANDLER_ARGS)
613 {
614 	struct disk *disk;
615 	int error, first;
616 
617 	disk = NULL;
618 	first = 1;
619 
620 	while ((disk = disk_enumerate(disk))) {
621 		if (!first) {
622 			error = SYSCTL_OUT(req, " ", 1);
623 			if (error)
624 				return error;
625 		} else {
626 			first = 0;
627 		}
628 		error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
629 				   strlen(disk->d_rawdev->si_name));
630 		if (error)
631 			return error;
632 	}
633 	error = SYSCTL_OUT(req, "", 1);
634 	return error;
635 }
636 
637 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, NULL, 0,
638     sysctl_disks, "A", "names of available disks");
639 
640 /*
641  * Open a disk device or partition.
642  */
643 static
644 int
645 diskopen(struct dev_open_args *ap)
646 {
647 	cdev_t dev = ap->a_head.a_dev;
648 	struct disk *dp;
649 	int error;
650 
651 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskopen: name is %s\n", dev->si_name);
652 
653 	/*
654 	 * dp can't be NULL here XXX.
655 	 */
656 	dp = dev->si_disk;
657 	if (dp == NULL)
658 		return (ENXIO);
659 	error = 0;
660 
661 	/*
662 	 * Deal with open races
663 	 */
664 	while (dp->d_flags & DISKFLAG_LOCK) {
665 		dp->d_flags |= DISKFLAG_WANTED;
666 		error = tsleep(dp, PCATCH, "diskopen", hz);
667 		if (error)
668 			return (error);
669 	}
670 	dp->d_flags |= DISKFLAG_LOCK;
671 
672 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskopen: -2- name is %s\n", dev->si_name);
673 
674 	/*
675 	 * Open the underlying raw device.
676 	 */
677 	if (!dsisopen(dp->d_slice)) {
678 #if 0
679 		if (!pdev->si_iosize_max)
680 			pdev->si_iosize_max = dev->si_iosize_max;
681 #endif
682 		error = dev_dopen(dp->d_rawdev, ap->a_oflags,
683 				  ap->a_devtype, ap->a_cred);
684 	}
685 #if 0
686 	/*
687 	 * Inherit properties from the underlying device now that it is
688 	 * open.
689 	 */
690 	dev_dclone(dev);
691 #endif
692 
693 	if (error)
694 		goto out;
695 	error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags,
696 		       &dp->d_slice, &dp->d_info);
697 	if (!dsisopen(dp->d_slice)) {
698 		dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype);
699 	}
700 out:
701 	dp->d_flags &= ~DISKFLAG_LOCK;
702 	if (dp->d_flags & DISKFLAG_WANTED) {
703 		dp->d_flags &= ~DISKFLAG_WANTED;
704 		wakeup(dp);
705 	}
706 
707 	return(error);
708 }
709 
710 /*
711  * Close a disk device or partition
712  */
713 static
714 int
715 diskclose(struct dev_close_args *ap)
716 {
717 	cdev_t dev = ap->a_head.a_dev;
718 	struct disk *dp;
719 	int error;
720 
721 	error = 0;
722 	dp = dev->si_disk;
723 
724 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskclose: name %s\n", dev->si_name);
725 
726 	dsclose(dev, ap->a_devtype, dp->d_slice);
727 	if (!dsisopen(dp->d_slice)) {
728 		devfs_debug(DEVFS_DEBUG_DEBUG, "diskclose is closing underlying device\n");
729 		error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype);
730 	}
731 	return (error);
732 }
733 
734 /*
735  * First execute the ioctl on the disk device, and if it isn't supported
736  * try running it on the backing device.
737  */
738 static
739 int
740 diskioctl(struct dev_ioctl_args *ap)
741 {
742 	cdev_t dev = ap->a_head.a_dev;
743 	struct disk *dp;
744 	int error;
745 
746 	dp = dev->si_disk;
747 	if (dp == NULL)
748 		return (ENXIO);
749 
750 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl: cmd is: %x (name: %s)\n", ap->a_cmd, dev->si_name);
751 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl: &dp->d_slice is: %x, %x\n", &dp->d_slice, dp->d_slice);
752 
753 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl:1: says lp.opaque is: %x\n", dp->d_slice->dss_slices[0].ds_label.opaque);
754 
755 	error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag,
756 			&dp->d_slice, &dp->d_info);
757 
758 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl:2: says lp.opaque is: %x\n", dp->d_slice->dss_slices[0].ds_label.opaque);
759 
760 	if (error == ENOIOCTL) {
761 		devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl: going for dev_dioctl instead!\n");
762 		error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data,
763 				   ap->a_fflag, ap->a_cred);
764 	}
765 	return (error);
766 }
767 
768 /*
769  * Execute strategy routine
770  */
771 static
772 int
773 diskstrategy(struct dev_strategy_args *ap)
774 {
775 	cdev_t dev = ap->a_head.a_dev;
776 	struct bio *bio = ap->a_bio;
777 	struct bio *nbio;
778 	struct disk *dp;
779 
780 	dp = dev->si_disk;
781 
782 	if (dp == NULL) {
783 		bio->bio_buf->b_error = ENXIO;
784 		bio->bio_buf->b_flags |= B_ERROR;
785 		biodone(bio);
786 		return(0);
787 	}
788 	KKASSERT(dev->si_disk == dp);
789 
790 	/*
791 	 * The dscheck() function will also transform the slice relative
792 	 * block number i.e. bio->bio_offset into a block number that can be
793 	 * passed directly to the underlying raw device.  If dscheck()
794 	 * returns NULL it will have handled the bio for us (e.g. EOF
795 	 * or error due to being beyond the device size).
796 	 */
797 	if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL) {
798 		dev_dstrategy(dp->d_rawdev, nbio);
799 	} else {
800 		devfs_debug(DEVFS_DEBUG_DEBUG, "diskstrategy: dscheck NULL!!! biodone time!\n");
801 		biodone(bio);
802 	}
803 	return(0);
804 }
805 
806 /*
807  * Return the partition size in ?blocks?
808  */
809 static
810 int
811 diskpsize(struct dev_psize_args *ap)
812 {
813 	cdev_t dev = ap->a_head.a_dev;
814 	struct disk *dp;
815 
816 	dp = dev->si_disk;
817 	if (dp == NULL)
818 		return(ENODEV);
819 	ap->a_result = dssize(dev, &dp->d_slice);
820 	return(0);
821 }
822 
823 /*
824  * When new device entries are instantiated, make sure they inherit our
825  * si_disk structure and block and iosize limits from the raw device.
826  *
827  * This routine is always called synchronously in the context of the
828  * client.
829  *
830  * XXX The various io and block size constraints are not always initialized
831  * properly by devices.
832  */
833 static
834 int
835 diskclone(struct dev_clone_args *ap)
836 {
837 	cdev_t dev = ap->a_head.a_dev;
838 	struct disk *dp;
839 	dp = dev->si_disk;
840 
841 	KKASSERT(dp != NULL);
842 	dev->si_disk = dp;
843 	dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
844 	dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
845 	dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
846 	return(0);
847 }
848 
849 int
850 diskdump(struct dev_dump_args *ap)
851 {
852 	cdev_t dev = ap->a_head.a_dev;
853 	struct disk *dp = dev->si_disk;
854 	int error;
855 
856 	error = disk_dumpcheck(dev, &ap->a_count, &ap->a_blkno, &ap->a_secsize);
857 	if (error == 0) {
858 		ap->a_head.a_dev = dp->d_rawdev;
859 		error = dev_doperate(&ap->a_head);
860 	}
861 
862 	return(error);
863 }
864 
865 
866 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
867     0, sizeof(struct diskslices), "sizeof(struct diskslices)");
868 
869 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
870     0, sizeof(struct disk), "sizeof(struct disk)");
871 
872 
873 /*
874  * Seek sort for disks.
875  *
876  * The bio_queue keep two queues, sorted in ascending block order.  The first
877  * queue holds those requests which are positioned after the current block
878  * (in the first request); the second, which starts at queue->switch_point,
879  * holds requests which came in after their block number was passed.  Thus
880  * we implement a one way scan, retracting after reaching the end of the drive
881  * to the first request on the second queue, at which time it becomes the
882  * first queue.
883  *
884  * A one-way scan is natural because of the way UNIX read-ahead blocks are
885  * allocated.
886  */
887 void
888 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
889 {
890 	struct bio *bq;
891 	struct bio *bn;
892 	struct bio *be;
893 
894 	be = TAILQ_LAST(&bioq->queue, bio_queue);
895 	/*
896 	 * If the queue is empty or we are an
897 	 * ordered transaction, then it's easy.
898 	 */
899 	if ((bq = bioq_first(bioq)) == NULL ||
900 	    (bio->bio_buf->b_flags & B_ORDERED) != 0) {
901 		bioq_insert_tail(bioq, bio);
902 		return;
903 	} else if (bioq->insert_point != NULL) {
904 
905 		/*
906 		 * A certain portion of the list is
907 		 * "locked" to preserve ordering, so
908 		 * we can only insert after the insert
909 		 * point.
910 		 */
911 		bq = bioq->insert_point;
912 	} else {
913 
914 		/*
915 		 * If we lie before the last removed (currently active)
916 		 * request, and are not inserting ourselves into the
917 		 * "locked" portion of the list, then we must add ourselves
918 		 * to the second request list.
919 		 */
920 		if (bio->bio_offset < bioq->last_offset) {
921 			bq = bioq->switch_point;
922 			/*
923 			 * If we are starting a new secondary list,
924 			 * then it's easy.
925 			 */
926 			if (bq == NULL) {
927 				bioq->switch_point = bio;
928 				bioq_insert_tail(bioq, bio);
929 				return;
930 			}
931 			/*
932 			 * If we lie ahead of the current switch point,
933 			 * insert us before the switch point and move
934 			 * the switch point.
935 			 */
936 			if (bio->bio_offset < bq->bio_offset) {
937 				bioq->switch_point = bio;
938 				TAILQ_INSERT_BEFORE(bq, bio, bio_act);
939 				return;
940 			}
941 		} else {
942 			if (bioq->switch_point != NULL)
943 				be = TAILQ_PREV(bioq->switch_point,
944 						bio_queue, bio_act);
945 			/*
946 			 * If we lie between last_offset and bq,
947 			 * insert before bq.
948 			 */
949 			if (bio->bio_offset < bq->bio_offset) {
950 				TAILQ_INSERT_BEFORE(bq, bio, bio_act);
951 				return;
952 			}
953 		}
954 	}
955 
956 	/*
957 	 * Request is at/after our current position in the list.
958 	 * Optimize for sequential I/O by seeing if we go at the tail.
959 	 */
960 	if (bio->bio_offset > be->bio_offset) {
961 		TAILQ_INSERT_AFTER(&bioq->queue, be, bio, bio_act);
962 		return;
963 	}
964 
965 	/* Otherwise, insertion sort */
966 	while ((bn = TAILQ_NEXT(bq, bio_act)) != NULL) {
967 
968 		/*
969 		 * We want to go after the current request if it is the end
970 		 * of the first request list, or if the next request is a
971 		 * larger cylinder than our request.
972 		 */
973 		if (bn == bioq->switch_point
974 		 || bio->bio_offset < bn->bio_offset)
975 			break;
976 		bq = bn;
977 	}
978 	TAILQ_INSERT_AFTER(&bioq->queue, bq, bio, bio_act);
979 }
980 
981 /*
982  * Disk error is the preface to plaintive error messages
983  * about failing disk transfers.  It prints messages of the form
984 
985 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
986 
987  * if the offset of the error in the transfer and a disk label
988  * are both available.  blkdone should be -1 if the position of the error
989  * is unknown; the disklabel pointer may be null from drivers that have not
990  * been converted to use them.  The message is printed with kprintf
991  * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
992  * The message should be completed (with at least a newline) with kprintf
993  * or log(-1, ...), respectively.  There is no trailing space.
994  */
995 void
996 diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt)
997 {
998 	struct buf *bp = bio->bio_buf;
999 	const char *term;
1000 
1001 	switch(bp->b_cmd) {
1002 	case BUF_CMD_READ:
1003 		term = "read";
1004 		break;
1005 	case BUF_CMD_WRITE:
1006 		term = "write";
1007 		break;
1008 	default:
1009 		term = "access";
1010 		break;
1011 	}
1012 	//sname = dsname(dev, unit, slice, part, partname);
1013 	kprintf("%s: %s %sing ", dev->si_name, what, term);
1014 	kprintf("offset %012llx for %d",
1015 		(long long)bio->bio_offset,
1016 		bp->b_bcount);
1017 
1018 	if (donecnt)
1019 		kprintf(" (%d bytes completed)", donecnt);
1020 }
1021 
1022 /*
1023  * Locate a disk device
1024  */
1025 cdev_t
1026 disk_locate(const char *devname)
1027 {
1028 	return devfs_find_device_by_name(devname);
1029 }
1030 
1031 
1032 void
1033 disk_config(void *arg)
1034 {
1035 	disk_msg_send_sync(DISK_SYNC, NULL, NULL);
1036 }
1037 
1038 
1039 static void
1040 disk_init(void)
1041 {
1042 	struct thread* td_core;
1043 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_init() called\n");
1044 
1045     disk_msg_cache = objcache_create("disk-msg-cache", 0, 0,
1046 			NULL, NULL, NULL,
1047 			objcache_malloc_alloc,
1048 			objcache_malloc_free,
1049 			&disk_msg_malloc_args );
1050 
1051 	/* Initialize the reply-only port which acts as a message drain */
1052 	lwkt_initport_replyonly(&disk_dispose_port, disk_msg_autofree_reply);
1053 
1054 	lwkt_create(disk_msg_core, /*args*/NULL, &td_core, NULL,
1055 		    0, 0, "disk_msg_core");
1056 
1057 	tsleep(td_core, 0, "diskcore", 0);
1058 }
1059 
1060 
1061 static void
1062 disk_uninit(void)
1063 {
1064 	devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_uninit() called\n");
1065 
1066 	objcache_destroy(disk_msg_cache);
1067 
1068 }
1069 
1070 
1071 SYSINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_FIRST, disk_init, NULL);
1072 SYSUNINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_ANY, disk_uninit, NULL);
1073