xref: /dflybsd-src/sys/kern/subr_disk.c (revision 3b9337bbddcb34a1fbcf503d3655c77bdc98c6f7)
1 /*
2  * Copyright (c) 2003,2004,2009 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  * and Alex Hornung <ahornung@gmail.com>
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  *
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in
16  *    the documentation and/or other materials provided with the
17  *    distribution.
18  * 3. Neither the name of The DragonFly Project nor the names of its
19  *    contributors may be used to endorse or promote products derived
20  *    from this software without specific, prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
25  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
26  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
27  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
28  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
29  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
30  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
31  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
32  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  * ----------------------------------------------------------------------------
36  * "THE BEER-WARE LICENSE" (Revision 42):
37  * <phk@FreeBSD.ORG> wrote this file.  As long as you retain this notice you
38  * can do whatever you want with this stuff. If we meet some day, and you think
39  * this stuff is worth it, you can buy me a beer in return.   Poul-Henning Kamp
40  * ----------------------------------------------------------------------------
41  *
42  * Copyright (c) 1982, 1986, 1988, 1993
43  *	The Regents of the University of California.  All rights reserved.
44  * (c) UNIX System Laboratories, Inc.
45  * All or some portions of this file are derived from material licensed
46  * to the University of California by American Telephone and Telegraph
47  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
48  * the permission of UNIX System Laboratories, Inc.
49  *
50  * Redistribution and use in source and binary forms, with or without
51  * modification, are permitted provided that the following conditions
52  * are met:
53  * 1. Redistributions of source code must retain the above copyright
54  *    notice, this list of conditions and the following disclaimer.
55  * 2. Redistributions in binary form must reproduce the above copyright
56  *    notice, this list of conditions and the following disclaimer in the
57  *    documentation and/or other materials provided with the distribution.
58  * 3. All advertising materials mentioning features or use of this software
59  *    must display the following acknowledgement:
60  *	This product includes software developed by the University of
61  *	California, Berkeley and its contributors.
62  * 4. Neither the name of the University nor the names of its contributors
63  *    may be used to endorse or promote products derived from this software
64  *    without specific prior written permission.
65  *
66  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
67  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
68  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
69  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
70  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
71  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
72  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
73  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
74  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
75  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
76  * SUCH DAMAGE.
77  *
78  *	@(#)ufs_disksubr.c	8.5 (Berkeley) 1/21/94
79  * $FreeBSD: src/sys/kern/subr_disk.c,v 1.20.2.6 2001/10/05 07:14:57 peter Exp $
80  * $FreeBSD: src/sys/ufs/ufs/ufs_disksubr.c,v 1.44.2.3 2001/03/05 05:42:19 obrien Exp $
81  * $DragonFly: src/sys/kern/subr_disk.c,v 1.40 2008/06/05 18:06:32 swildner Exp $
82  */
83 
84 #include <sys/param.h>
85 #include <sys/systm.h>
86 #include <sys/kernel.h>
87 #include <sys/proc.h>
88 #include <sys/sysctl.h>
89 #include <sys/buf.h>
90 #include <sys/conf.h>
91 #include <sys/disklabel.h>
92 #include <sys/disklabel32.h>
93 #include <sys/disklabel64.h>
94 #include <sys/diskslice.h>
95 #include <sys/diskmbr.h>
96 #include <sys/disk.h>
97 #include <sys/kerneldump.h>
98 #include <sys/malloc.h>
99 #include <sys/sysctl.h>
100 #include <machine/md_var.h>
101 #include <sys/ctype.h>
102 #include <sys/syslog.h>
103 #include <sys/device.h>
104 #include <sys/msgport.h>
105 #include <sys/devfs.h>
106 #include <sys/thread.h>
107 #include <sys/dsched.h>
108 #include <sys/queue.h>
109 #include <sys/lock.h>
110 #include <sys/udev.h>
111 #include <sys/uuid.h>
112 
113 #include <sys/buf2.h>
114 #include <sys/mplock2.h>
115 #include <sys/msgport2.h>
116 #include <sys/thread2.h>
117 
118 static MALLOC_DEFINE(M_DISK, "disk", "disk data");
119 static int disk_debug_enable = 0;
120 
121 static void disk_msg_autofree_reply(lwkt_port_t, lwkt_msg_t);
122 static void disk_msg_core(void *);
123 static int disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe);
124 static void disk_probe(struct disk *dp, int reprobe);
125 static void _setdiskinfo(struct disk *disk, struct disk_info *info);
126 static void bioqwritereorder(struct bio_queue_head *bioq);
127 static void disk_cleanserial(char *serno);
128 static int disk_debug(int, char *, ...) __printflike(2, 3);
129 static cdev_t _disk_create_named(const char *name, int unit, struct disk *dp,
130     struct dev_ops *raw_ops, int clone);
131 
132 static d_open_t diskopen;
133 static d_close_t diskclose;
134 static d_ioctl_t diskioctl;
135 static d_strategy_t diskstrategy;
136 static d_psize_t diskpsize;
137 static d_dump_t diskdump;
138 
139 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
140 static struct lwkt_token disklist_token;
141 
142 static struct dev_ops disk_ops = {
143 	{ "disk", 0, D_DISK | D_MPSAFE | D_TRACKCLOSE },
144 	.d_open = diskopen,
145 	.d_close = diskclose,
146 	.d_read = physread,
147 	.d_write = physwrite,
148 	.d_ioctl = diskioctl,
149 	.d_strategy = diskstrategy,
150 	.d_dump = diskdump,
151 	.d_psize = diskpsize,
152 };
153 
154 static struct objcache 	*disk_msg_cache;
155 
156 struct objcache_malloc_args disk_msg_malloc_args = {
157 	sizeof(struct disk_msg), M_DISK };
158 
159 static struct lwkt_port disk_dispose_port;
160 static struct lwkt_port disk_msg_port;
161 
162 static int
163 disk_debug(int level, char *fmt, ...)
164 {
165 	__va_list ap;
166 
167 	__va_start(ap, fmt);
168 	if (level <= disk_debug_enable)
169 		kvprintf(fmt, ap);
170 	__va_end(ap);
171 
172 	return 0;
173 }
174 
175 static int
176 disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe)
177 {
178 	struct disk_info *info = &dp->d_info;
179 	struct diskslice *sp = &dp->d_slice->dss_slices[slice];
180 	disklabel_ops_t ops;
181 	struct partinfo part;
182 	const char *msg;
183 	char uuid_buf[128];
184 	cdev_t ndev;
185 	int sno;
186 	u_int i;
187 
188 	disk_debug(2,
189 		    "disk_probe_slice (begin): %s (%s)\n",
190 			dev->si_name, dp->d_cdev->si_name);
191 
192 	sno = slice ? slice - 1 : 0;
193 
194 	ops = &disklabel32_ops;
195 	msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
196 	if (msg && !strcmp(msg, "no disk label")) {
197 		ops = &disklabel64_ops;
198 		msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
199 	}
200 	if (msg == NULL) {
201 		if (slice != WHOLE_DISK_SLICE)
202 			ops->op_adjust_label_reserved(dp->d_slice, slice, sp);
203 		else
204 			sp->ds_reserved = 0;
205 
206 		sp->ds_ops = ops;
207 		for (i = 0; i < ops->op_getnumparts(sp->ds_label); i++) {
208 			ops->op_loadpartinfo(sp->ds_label, i, &part);
209 			if (part.fstype) {
210 				if (reprobe &&
211 				    (ndev = devfs_find_device_by_name("%s%c",
212 						dev->si_name, 'a' + i))
213 				) {
214 					/*
215 					 * Device already exists and
216 					 * is still valid.
217 					 */
218 					ndev->si_flags |= SI_REPROBE_TEST;
219 
220 					/*
221 					 * Destroy old UUID alias
222 					 */
223 					destroy_dev_alias(ndev, "part-by-uuid/*");
224 
225 					/* Create UUID alias */
226 					if (!kuuid_is_nil(&part.storage_uuid)) {
227 						snprintf_uuid(uuid_buf,
228 						    sizeof(uuid_buf),
229 						    &part.storage_uuid);
230 						make_dev_alias(ndev,
231 						    "part-by-uuid/%s",
232 						    uuid_buf);
233 					}
234 				} else {
235 					ndev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
236 						dkmakeminor(dkunit(dp->d_cdev),
237 							    slice, i),
238 						UID_ROOT, GID_OPERATOR, 0640,
239 						"%s%c", dev->si_name, 'a'+ i);
240 					ndev->si_disk = dp;
241 					udev_dict_set_cstr(ndev, "subsystem", "disk");
242 					/* Inherit parent's disk type */
243 					if (dp->d_disktype) {
244 						udev_dict_set_cstr(ndev, "disk-type",
245 						    __DECONST(char *, dp->d_disktype));
246 					}
247 
248 					/* Create serno alias */
249 					if (dp->d_info.d_serialno) {
250 						make_dev_alias(ndev,
251 						    "serno/%s.s%d%c",
252 						    dp->d_info.d_serialno,
253 						    sno, 'a' + i);
254 					}
255 
256 					/* Create UUID alias */
257 					if (!kuuid_is_nil(&part.storage_uuid)) {
258 						snprintf_uuid(uuid_buf,
259 						    sizeof(uuid_buf),
260 						    &part.storage_uuid);
261 						make_dev_alias(ndev,
262 						    "part-by-uuid/%s",
263 						    uuid_buf);
264 					}
265 					ndev->si_flags |= SI_REPROBE_TEST;
266 				}
267 			}
268 		}
269 	} else if (info->d_dsflags & DSO_COMPATLABEL) {
270 		msg = NULL;
271 		if (sp->ds_size >= 0x100000000ULL)
272 			ops = &disklabel64_ops;
273 		else
274 			ops = &disklabel32_ops;
275 		sp->ds_label = ops->op_clone_label(info, sp);
276 	} else {
277 		if (sp->ds_type == DOSPTYP_386BSD || /* XXX */
278 		    sp->ds_type == DOSPTYP_NETBSD ||
279 		    sp->ds_type == DOSPTYP_OPENBSD) {
280 			log(LOG_WARNING, "%s: cannot find label (%s)\n",
281 			    dev->si_name, msg);
282 		}
283 	}
284 
285 	if (msg == NULL) {
286 		sp->ds_wlabel = FALSE;
287 	}
288 
289 	return (msg ? EINVAL : 0);
290 }
291 
292 /*
293  * This routine is only called for newly minted drives or to reprobe
294  * a drive with no open slices.  disk_probe_slice() is called directly
295  * when reprobing partition changes within slices.
296  */
297 static void
298 disk_probe(struct disk *dp, int reprobe)
299 {
300 	struct disk_info *info = &dp->d_info;
301 	cdev_t dev = dp->d_cdev;
302 	cdev_t ndev;
303 	int error, i, sno;
304 	struct diskslices *osp;
305 	struct diskslice *sp;
306 	char uuid_buf[128];
307 
308 	KKASSERT (info->d_media_blksize != 0);
309 
310 	osp = dp->d_slice;
311 	dp->d_slice = dsmakeslicestruct(BASE_SLICE, info);
312 	disk_debug(1, "disk_probe (begin): %s\n", dp->d_cdev->si_name);
313 
314 	error = mbrinit(dev, info, &(dp->d_slice));
315 	if (error) {
316 		dsgone(&osp);
317 		return;
318 	}
319 
320 	for (i = 0; i < dp->d_slice->dss_nslices; i++) {
321 		/*
322 		 * Ignore the whole-disk slice, it has already been created.
323 		 */
324 		if (i == WHOLE_DISK_SLICE)
325 			continue;
326 
327 #if 1
328 		/*
329 		 * Ignore the compatibility slice s0 if it's a device mapper
330 		 * volume.
331 		 */
332 		if ((i == COMPATIBILITY_SLICE) &&
333 		    (info->d_dsflags & DSO_DEVICEMAPPER))
334 			continue;
335 #endif
336 
337 		sp = &dp->d_slice->dss_slices[i];
338 
339 		/*
340 		 * Handle s0.  s0 is a compatibility slice if there are no
341 		 * other slices and it has not otherwise been set up, else
342 		 * we ignore it.
343 		 */
344 		if (i == COMPATIBILITY_SLICE) {
345 			sno = 0;
346 			if (sp->ds_type == 0 &&
347 			    dp->d_slice->dss_nslices == BASE_SLICE) {
348 				sp->ds_size = info->d_media_blocks;
349 				sp->ds_reserved = 0;
350 			}
351 		} else {
352 			sno = i - 1;
353 			sp->ds_reserved = 0;
354 		}
355 
356 		/*
357 		 * Ignore 0-length slices
358 		 */
359 		if (sp->ds_size == 0)
360 			continue;
361 
362 		if (reprobe &&
363 		    (ndev = devfs_find_device_by_name("%ss%d",
364 						      dev->si_name, sno))) {
365 			/*
366 			 * Device already exists and is still valid
367 			 */
368 			ndev->si_flags |= SI_REPROBE_TEST;
369 
370 			/*
371 			 * Destroy old UUID alias
372 			 */
373 			destroy_dev_alias(ndev, "slice-by-uuid/*");
374 
375 			/* Create UUID alias */
376 			if (!kuuid_is_nil(&sp->ds_stor_uuid)) {
377 				snprintf_uuid(uuid_buf, sizeof(uuid_buf),
378 				    &sp->ds_stor_uuid);
379 				make_dev_alias(ndev, "slice-by-uuid/%s",
380 				    uuid_buf);
381 			}
382 		} else {
383 			/*
384 			 * Else create new device
385 			 */
386 			ndev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
387 					dkmakewholeslice(dkunit(dev), i),
388 					UID_ROOT, GID_OPERATOR, 0640,
389 					(info->d_dsflags & DSO_DEVICEMAPPER)?
390 					"%s.s%d" : "%ss%d", dev->si_name, sno);
391 			udev_dict_set_cstr(ndev, "subsystem", "disk");
392 			/* Inherit parent's disk type */
393 			if (dp->d_disktype) {
394 				udev_dict_set_cstr(ndev, "disk-type",
395 				    __DECONST(char *, dp->d_disktype));
396 			}
397 
398 			/* Create serno alias */
399 			if (dp->d_info.d_serialno) {
400 				make_dev_alias(ndev, "serno/%s.s%d",
401 					       dp->d_info.d_serialno, sno);
402 			}
403 
404 			/* Create UUID alias */
405 			if (!kuuid_is_nil(&sp->ds_stor_uuid)) {
406 				snprintf_uuid(uuid_buf, sizeof(uuid_buf),
407 				    &sp->ds_stor_uuid);
408 				make_dev_alias(ndev, "slice-by-uuid/%s",
409 				    uuid_buf);
410 			}
411 
412 			ndev->si_disk = dp;
413 			ndev->si_flags |= SI_REPROBE_TEST;
414 		}
415 		sp->ds_dev = ndev;
416 
417 		/*
418 		 * Probe appropriate slices for a disklabel
419 		 *
420 		 * XXX slice type 1 used by our gpt probe code.
421 		 * XXX slice type 0 used by mbr compat slice.
422 		 */
423 		if (sp->ds_type == DOSPTYP_386BSD ||
424 		    sp->ds_type == DOSPTYP_NETBSD ||
425 		    sp->ds_type == DOSPTYP_OPENBSD ||
426 		    sp->ds_type == 0 ||
427 		    sp->ds_type == 1) {
428 			if (dp->d_slice->dss_first_bsd_slice == 0)
429 				dp->d_slice->dss_first_bsd_slice = i;
430 			disk_probe_slice(dp, ndev, i, reprobe);
431 		}
432 	}
433 	dsgone(&osp);
434 	disk_debug(1, "disk_probe (end): %s\n", dp->d_cdev->si_name);
435 }
436 
437 
438 static void
439 disk_msg_core(void *arg)
440 {
441 	struct disk	*dp;
442 	struct diskslice *sp;
443 	disk_msg_t msg;
444 	int run;
445 
446 	lwkt_gettoken(&disklist_token);
447 	lwkt_initport_thread(&disk_msg_port, curthread);
448 	wakeup(curthread);	/* synchronous startup */
449 	lwkt_reltoken(&disklist_token);
450 
451 	get_mplock();	/* not mpsafe yet? */
452 	run = 1;
453 
454 	while (run) {
455 		msg = (disk_msg_t)lwkt_waitport(&disk_msg_port, 0);
456 
457 		switch (msg->hdr.u.ms_result) {
458 		case DISK_DISK_PROBE:
459 			dp = (struct disk *)msg->load;
460 			disk_debug(1,
461 				    "DISK_DISK_PROBE: %s\n",
462 					dp->d_cdev->si_name);
463 			disk_probe(dp, 0);
464 			break;
465 		case DISK_DISK_DESTROY:
466 			dp = (struct disk *)msg->load;
467 			disk_debug(1,
468 				    "DISK_DISK_DESTROY: %s\n",
469 					dp->d_cdev->si_name);
470 			devfs_destroy_subnames(dp->d_cdev->si_name);
471 			destroy_dev(dp->d_cdev);
472 			destroy_only_dev(dp->d_rawdev);
473 			lwkt_gettoken(&disklist_token);
474 			LIST_REMOVE(dp, d_list);
475 			lwkt_reltoken(&disklist_token);
476 			if (dp->d_info.d_serialno) {
477 				kfree(dp->d_info.d_serialno, M_TEMP);
478 				dp->d_info.d_serialno = NULL;
479 			}
480 			break;
481 		case DISK_UNPROBE:
482 			dp = (struct disk *)msg->load;
483 			disk_debug(1,
484 				    "DISK_DISK_UNPROBE: %s\n",
485 					dp->d_cdev->si_name);
486 			devfs_destroy_subnames(dp->d_cdev->si_name);
487 			break;
488 		case DISK_SLICE_REPROBE:
489 			dp = (struct disk *)msg->load;
490 			sp = (struct diskslice *)msg->load2;
491 			devfs_clr_subnames_flag(sp->ds_dev->si_name,
492 						SI_REPROBE_TEST);
493 			disk_debug(1,
494 				    "DISK_SLICE_REPROBE: %s\n",
495 				    sp->ds_dev->si_name);
496 			disk_probe_slice(dp, sp->ds_dev,
497 					 dkslice(sp->ds_dev), 1);
498 			devfs_destroy_subnames_without_flag(
499 					sp->ds_dev->si_name, SI_REPROBE_TEST);
500 			break;
501 		case DISK_DISK_REPROBE:
502 			dp = (struct disk *)msg->load;
503 			devfs_clr_subnames_flag(dp->d_cdev->si_name, SI_REPROBE_TEST);
504 			disk_debug(1,
505 				    "DISK_DISK_REPROBE: %s\n",
506 				    dp->d_cdev->si_name);
507 			disk_probe(dp, 1);
508 			devfs_destroy_subnames_without_flag(
509 					dp->d_cdev->si_name, SI_REPROBE_TEST);
510 			break;
511 		case DISK_SYNC:
512 			disk_debug(1, "DISK_SYNC\n");
513 			break;
514 		default:
515 			devfs_debug(DEVFS_DEBUG_WARNING,
516 				    "disk_msg_core: unknown message "
517 				    "received at core\n");
518 			break;
519 		}
520 		lwkt_replymsg(&msg->hdr, 0);
521 	}
522 	lwkt_exit();
523 }
524 
525 
526 /*
527  * Acts as a message drain. Any message that is replied to here gets
528  * destroyed and the memory freed.
529  */
530 static void
531 disk_msg_autofree_reply(lwkt_port_t port, lwkt_msg_t msg)
532 {
533 	objcache_put(disk_msg_cache, msg);
534 }
535 
536 
537 void
538 disk_msg_send(uint32_t cmd, void *load, void *load2)
539 {
540 	disk_msg_t disk_msg;
541 	lwkt_port_t port = &disk_msg_port;
542 
543 	disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
544 
545 	lwkt_initmsg(&disk_msg->hdr, &disk_dispose_port, 0);
546 
547 	disk_msg->hdr.u.ms_result = cmd;
548 	disk_msg->load = load;
549 	disk_msg->load2 = load2;
550 	KKASSERT(port);
551 	lwkt_sendmsg(port, &disk_msg->hdr);
552 }
553 
554 void
555 disk_msg_send_sync(uint32_t cmd, void *load, void *load2)
556 {
557 	struct lwkt_port rep_port;
558 	disk_msg_t disk_msg;
559 	lwkt_port_t port;
560 
561 	disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
562 	port = &disk_msg_port;
563 
564 	/* XXX could probably use curthread's built-in msgport */
565 	lwkt_initport_thread(&rep_port, curthread);
566 	lwkt_initmsg(&disk_msg->hdr, &rep_port, 0);
567 
568 	disk_msg->hdr.u.ms_result = cmd;
569 	disk_msg->load = load;
570 	disk_msg->load2 = load2;
571 
572 	lwkt_sendmsg(port, &disk_msg->hdr);
573 	lwkt_waitmsg(&disk_msg->hdr, 0);
574 	objcache_put(disk_msg_cache, disk_msg);
575 }
576 
577 /*
578  * Create a raw device for the dev_ops template (which is returned).  Also
579  * create a slice and unit managed disk and overload the user visible
580  * device space with it.
581  *
582  * NOTE: The returned raw device is NOT a slice and unit managed device.
583  * It is an actual raw device representing the raw disk as specified by
584  * the passed dev_ops.  The disk layer not only returns such a raw device,
585  * it also uses it internally when passing (modified) commands through.
586  */
587 cdev_t
588 disk_create(int unit, struct disk *dp, struct dev_ops *raw_ops)
589 {
590 	return _disk_create_named(NULL, unit, dp, raw_ops, 0);
591 }
592 
593 cdev_t
594 disk_create_clone(int unit, struct disk *dp, struct dev_ops *raw_ops)
595 {
596 	return _disk_create_named(NULL, unit, dp, raw_ops, 1);
597 }
598 
599 cdev_t
600 disk_create_named(const char *name, int unit, struct disk *dp, struct dev_ops *raw_ops)
601 {
602 	return _disk_create_named(name, unit, dp, raw_ops, 0);
603 }
604 
605 cdev_t
606 disk_create_named_clone(const char *name, int unit, struct disk *dp, struct dev_ops *raw_ops)
607 {
608 	return _disk_create_named(name, unit, dp, raw_ops, 1);
609 }
610 
611 static cdev_t
612 _disk_create_named(const char *name, int unit, struct disk *dp, struct dev_ops *raw_ops, int clone)
613 {
614 	cdev_t rawdev;
615 
616 	disk_debug(1, "disk_create (begin): %s%d\n", name, unit);
617 
618 	if (name) {
619 		rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
620 		    UID_ROOT, GID_OPERATOR, 0640, "%s", name);
621 	} else {
622 		rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
623 		    UID_ROOT, GID_OPERATOR, 0640,
624 		    "%s%d", raw_ops->head.name, unit);
625 	}
626 
627 	bzero(dp, sizeof(*dp));
628 
629 	dp->d_rawdev = rawdev;
630 	dp->d_raw_ops = raw_ops;
631 	dp->d_dev_ops = &disk_ops;
632 
633 	if (name) {
634 		if (clone) {
635 			dp->d_cdev = make_only_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
636 			    dkmakewholedisk(unit), UID_ROOT, GID_OPERATOR, 0640,
637 			    "%s", name);
638 		} else {
639 			dp->d_cdev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
640 			    dkmakewholedisk(unit), UID_ROOT, GID_OPERATOR, 0640,
641 			    "%s", name);
642 		}
643 	} else {
644 		if (clone) {
645 			dp->d_cdev = make_only_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
646 			    dkmakewholedisk(unit),
647 			    UID_ROOT, GID_OPERATOR, 0640,
648 			    "%s%d", raw_ops->head.name, unit);
649 		} else {
650 			dp->d_cdev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
651 			    dkmakewholedisk(unit),
652 			    UID_ROOT, GID_OPERATOR, 0640,
653 			    "%s%d", raw_ops->head.name, unit);
654 		}
655 	}
656 
657 	udev_dict_set_cstr(dp->d_cdev, "subsystem", "disk");
658 	dp->d_cdev->si_disk = dp;
659 
660 	if (name)
661 		dsched_disk_create_callback(dp, name, unit);
662 	else
663 		dsched_disk_create_callback(dp, raw_ops->head.name, unit);
664 
665 	lwkt_gettoken(&disklist_token);
666 	LIST_INSERT_HEAD(&disklist, dp, d_list);
667 	lwkt_reltoken(&disklist_token);
668 
669 	disk_debug(1, "disk_create (end): %s%d\n",
670 	    (name != NULL)?(name):(raw_ops->head.name), unit);
671 
672 	return (dp->d_rawdev);
673 }
674 
675 int
676 disk_setdisktype(struct disk *disk, const char *type)
677 {
678 	KKASSERT(disk != NULL);
679 
680 	disk->d_disktype = type;
681 	return udev_dict_set_cstr(disk->d_cdev, "disk-type", __DECONST(char *, type));
682 }
683 
684 int
685 disk_getopencount(struct disk *disk)
686 {
687 	return disk->d_opencount;
688 }
689 
690 static void
691 _setdiskinfo(struct disk *disk, struct disk_info *info)
692 {
693 	char *oldserialno;
694 
695 	oldserialno = disk->d_info.d_serialno;
696 	bcopy(info, &disk->d_info, sizeof(disk->d_info));
697 	info = &disk->d_info;
698 
699 	disk_debug(1,
700 		    "_setdiskinfo: %s\n",
701 			disk->d_cdev->si_name);
702 
703 	/*
704 	 * The serial number is duplicated so the caller can throw
705 	 * their copy away.
706 	 */
707 	if (info->d_serialno && info->d_serialno[0]) {
708 		info->d_serialno = kstrdup(info->d_serialno, M_TEMP);
709 		disk_cleanserial(info->d_serialno);
710 		if (disk->d_cdev) {
711 			make_dev_alias(disk->d_cdev, "serno/%s",
712 					info->d_serialno);
713 		}
714 	} else {
715 		info->d_serialno = NULL;
716 	}
717 	if (oldserialno)
718 		kfree(oldserialno, M_TEMP);
719 
720 	dsched_disk_update_callback(disk, info);
721 
722 	/*
723 	 * The caller may set d_media_size or d_media_blocks and we
724 	 * calculate the other.
725 	 */
726 	KKASSERT(info->d_media_size == 0 || info->d_media_blocks == 0);
727 	if (info->d_media_size == 0 && info->d_media_blocks) {
728 		info->d_media_size = (u_int64_t)info->d_media_blocks *
729 				     info->d_media_blksize;
730 	} else if (info->d_media_size && info->d_media_blocks == 0 &&
731 		   info->d_media_blksize) {
732 		info->d_media_blocks = info->d_media_size /
733 				       info->d_media_blksize;
734 	}
735 
736 	/*
737 	 * The si_* fields for rawdev are not set until after the
738 	 * disk_create() call, so someone using the cooked version
739 	 * of the raw device (i.e. da0s0) will not get the right
740 	 * si_iosize_max unless we fix it up here.
741 	 */
742 	if (disk->d_cdev && disk->d_rawdev &&
743 	    disk->d_cdev->si_iosize_max == 0) {
744 		disk->d_cdev->si_iosize_max = disk->d_rawdev->si_iosize_max;
745 		disk->d_cdev->si_bsize_phys = disk->d_rawdev->si_bsize_phys;
746 		disk->d_cdev->si_bsize_best = disk->d_rawdev->si_bsize_best;
747 	}
748 
749 	/* Add the serial number to the udev_dictionary */
750 	if (info->d_serialno)
751 		udev_dict_set_cstr(disk->d_cdev, "serno", info->d_serialno);
752 }
753 
754 /*
755  * Disk drivers must call this routine when media parameters are available
756  * or have changed.
757  */
758 void
759 disk_setdiskinfo(struct disk *disk, struct disk_info *info)
760 {
761 	_setdiskinfo(disk, info);
762 	disk_msg_send(DISK_DISK_PROBE, disk, NULL);
763 	disk_debug(1,
764 		    "disk_setdiskinfo: sent probe for %s\n",
765 			disk->d_cdev->si_name);
766 }
767 
768 void
769 disk_setdiskinfo_sync(struct disk *disk, struct disk_info *info)
770 {
771 	_setdiskinfo(disk, info);
772 	disk_msg_send_sync(DISK_DISK_PROBE, disk, NULL);
773 	disk_debug(1,
774 		    "disk_setdiskinfo_sync: sent probe for %s\n",
775 			disk->d_cdev->si_name);
776 }
777 
778 /*
779  * This routine is called when an adapter detaches.  The higher level
780  * managed disk device is destroyed while the lower level raw device is
781  * released.
782  */
783 void
784 disk_destroy(struct disk *disk)
785 {
786 	dsched_disk_destroy_callback(disk);
787 	disk_msg_send_sync(DISK_DISK_DESTROY, disk, NULL);
788 	return;
789 }
790 
791 int
792 disk_dumpcheck(cdev_t dev, u_int64_t *size, u_int64_t *blkno, u_int32_t *secsize)
793 {
794 	struct partinfo pinfo;
795 	int error;
796 
797 	bzero(&pinfo, sizeof(pinfo));
798 	error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0,
799 			   proc0.p_ucred, NULL);
800 	if (error)
801 		return (error);
802 
803 	if (pinfo.media_blksize == 0)
804 		return (ENXIO);
805 
806 	if (blkno) /* XXX: make sure this reserved stuff is right */
807 		*blkno = pinfo.reserved_blocks +
808 			pinfo.media_offset / pinfo.media_blksize;
809 	if (secsize)
810 		*secsize = pinfo.media_blksize;
811 	if (size)
812 		*size = (pinfo.media_blocks - pinfo.reserved_blocks);
813 
814 	return (0);
815 }
816 
817 int
818 disk_dumpconf(cdev_t dev, u_int onoff)
819 {
820 	struct dumperinfo di;
821 	u_int64_t	size, blkno;
822 	u_int32_t	secsize;
823 	int error;
824 
825 	if (!onoff)
826 		return set_dumper(NULL);
827 
828 	error = disk_dumpcheck(dev, &size, &blkno, &secsize);
829 
830 	if (error)
831 		return ENXIO;
832 
833 	bzero(&di, sizeof(struct dumperinfo));
834 	di.dumper = diskdump;
835 	di.priv = dev;
836 	di.blocksize = secsize;
837 	di.mediaoffset = blkno * DEV_BSIZE;
838 	di.mediasize = size * DEV_BSIZE;
839 
840 	return set_dumper(&di);
841 }
842 
843 void
844 disk_unprobe(struct disk *disk)
845 {
846 	if (disk == NULL)
847 		return;
848 
849 	disk_msg_send_sync(DISK_UNPROBE, disk, NULL);
850 }
851 
852 void
853 disk_invalidate (struct disk *disk)
854 {
855 	dsgone(&disk->d_slice);
856 }
857 
858 struct disk *
859 disk_enumerate(struct disk *disk)
860 {
861 	struct disk *dp;
862 
863 	lwkt_gettoken(&disklist_token);
864 	if (!disk)
865 		dp = (LIST_FIRST(&disklist));
866 	else
867 		dp = (LIST_NEXT(disk, d_list));
868 	lwkt_reltoken(&disklist_token);
869 
870 	return dp;
871 }
872 
873 static
874 int
875 sysctl_disks(SYSCTL_HANDLER_ARGS)
876 {
877 	struct disk *disk;
878 	int error, first;
879 
880 	disk = NULL;
881 	first = 1;
882 
883 	while ((disk = disk_enumerate(disk))) {
884 		if (!first) {
885 			error = SYSCTL_OUT(req, " ", 1);
886 			if (error)
887 				return error;
888 		} else {
889 			first = 0;
890 		}
891 		error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
892 				   strlen(disk->d_rawdev->si_name));
893 		if (error)
894 			return error;
895 	}
896 	error = SYSCTL_OUT(req, "", 1);
897 	return error;
898 }
899 
900 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, NULL, 0,
901     sysctl_disks, "A", "names of available disks");
902 
903 /*
904  * Open a disk device or partition.
905  */
906 static
907 int
908 diskopen(struct dev_open_args *ap)
909 {
910 	cdev_t dev = ap->a_head.a_dev;
911 	struct disk *dp;
912 	int error;
913 
914 	/*
915 	 * dp can't be NULL here XXX.
916 	 *
917 	 * d_slice will be NULL if setdiskinfo() has not been called yet.
918 	 * setdiskinfo() is typically called whether the disk is present
919 	 * or not (e.g. CD), but the base disk device is created first
920 	 * and there may be a race.
921 	 */
922 	dp = dev->si_disk;
923 	if (dp == NULL || dp->d_slice == NULL)
924 		return (ENXIO);
925 	error = 0;
926 
927 	/*
928 	 * Deal with open races
929 	 */
930 	get_mplock();
931 	while (dp->d_flags & DISKFLAG_LOCK) {
932 		dp->d_flags |= DISKFLAG_WANTED;
933 		error = tsleep(dp, PCATCH, "diskopen", hz);
934 		if (error) {
935 			rel_mplock();
936 			return (error);
937 		}
938 	}
939 	dp->d_flags |= DISKFLAG_LOCK;
940 
941 	/*
942 	 * Open the underlying raw device.
943 	 */
944 	if (!dsisopen(dp->d_slice)) {
945 #if 0
946 		if (!pdev->si_iosize_max)
947 			pdev->si_iosize_max = dev->si_iosize_max;
948 #endif
949 		error = dev_dopen(dp->d_rawdev, ap->a_oflags,
950 				  ap->a_devtype, ap->a_cred);
951 	}
952 
953 	if (error)
954 		goto out;
955 	error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags,
956 		       &dp->d_slice, &dp->d_info);
957 	if (!dsisopen(dp->d_slice)) {
958 		dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype);
959 	}
960 out:
961 	dp->d_flags &= ~DISKFLAG_LOCK;
962 	if (dp->d_flags & DISKFLAG_WANTED) {
963 		dp->d_flags &= ~DISKFLAG_WANTED;
964 		wakeup(dp);
965 	}
966 	rel_mplock();
967 
968 	KKASSERT(dp->d_opencount >= 0);
969 	/* If the open was successful, bump open count */
970 	if (error == 0)
971 		atomic_add_int(&dp->d_opencount, 1);
972 
973 	return(error);
974 }
975 
976 /*
977  * Close a disk device or partition
978  */
979 static
980 int
981 diskclose(struct dev_close_args *ap)
982 {
983 	cdev_t dev = ap->a_head.a_dev;
984 	struct disk *dp;
985 	int error;
986 
987 	error = 0;
988 	dp = dev->si_disk;
989 
990 	KKASSERT(dp->d_opencount >= 1);
991 	/* If this is not the last close, just ignore it */
992 	if ((atomic_fetchadd_int(&dp->d_opencount, -1)) > 1)
993 		return 0;
994 
995 	get_mplock();
996 	dsclose(dev, ap->a_devtype, dp->d_slice);
997 	if (!dsisopen(dp->d_slice)) {
998 		error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype);
999 	}
1000 	rel_mplock();
1001 	return (error);
1002 }
1003 
1004 /*
1005  * First execute the ioctl on the disk device, and if it isn't supported
1006  * try running it on the backing device.
1007  */
1008 static
1009 int
1010 diskioctl(struct dev_ioctl_args *ap)
1011 {
1012 	cdev_t dev = ap->a_head.a_dev;
1013 	struct disk *dp;
1014 	int error;
1015 	u_int u;
1016 
1017 	dp = dev->si_disk;
1018 	if (dp == NULL)
1019 		return (ENXIO);
1020 
1021 	devfs_debug(DEVFS_DEBUG_DEBUG,
1022 		    "diskioctl: cmd is: %lx (name: %s)\n",
1023 		    ap->a_cmd, dev->si_name);
1024 	devfs_debug(DEVFS_DEBUG_DEBUG,
1025 		    "diskioctl: &dp->d_slice is: %p, %p\n",
1026 		    &dp->d_slice, dp->d_slice);
1027 
1028 	if (ap->a_cmd == DIOCGKERNELDUMP) {
1029 		u = *(u_int *)ap->a_data;
1030 		return disk_dumpconf(dev, u);
1031 	}
1032 
1033 	if (&dp->d_slice == NULL || dp->d_slice == NULL ||
1034 	    ((dp->d_info.d_dsflags & DSO_DEVICEMAPPER) &&
1035 	     dkslice(dev) == WHOLE_DISK_SLICE)) {
1036 		error = ENOIOCTL;
1037 	} else {
1038 		get_mplock();
1039 		error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag,
1040 				&dp->d_slice, &dp->d_info);
1041 		rel_mplock();
1042 	}
1043 
1044 	if (error == ENOIOCTL) {
1045 		error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data,
1046 				   ap->a_fflag, ap->a_cred, NULL);
1047 	}
1048 	return (error);
1049 }
1050 
1051 /*
1052  * Execute strategy routine
1053  */
1054 static
1055 int
1056 diskstrategy(struct dev_strategy_args *ap)
1057 {
1058 	cdev_t dev = ap->a_head.a_dev;
1059 	struct bio *bio = ap->a_bio;
1060 	struct bio *nbio;
1061 	struct disk *dp;
1062 
1063 	dp = dev->si_disk;
1064 
1065 	if (dp == NULL) {
1066 		bio->bio_buf->b_error = ENXIO;
1067 		bio->bio_buf->b_flags |= B_ERROR;
1068 		biodone(bio);
1069 		return(0);
1070 	}
1071 	KKASSERT(dev->si_disk == dp);
1072 
1073 	/*
1074 	 * The dscheck() function will also transform the slice relative
1075 	 * block number i.e. bio->bio_offset into a block number that can be
1076 	 * passed directly to the underlying raw device.  If dscheck()
1077 	 * returns NULL it will have handled the bio for us (e.g. EOF
1078 	 * or error due to being beyond the device size).
1079 	 */
1080 	if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL) {
1081 		dsched_queue(dp, nbio);
1082 	} else {
1083 		biodone(bio);
1084 	}
1085 	return(0);
1086 }
1087 
1088 /*
1089  * Return the partition size in ?blocks?
1090  */
1091 static
1092 int
1093 diskpsize(struct dev_psize_args *ap)
1094 {
1095 	cdev_t dev = ap->a_head.a_dev;
1096 	struct disk *dp;
1097 
1098 	dp = dev->si_disk;
1099 	if (dp == NULL)
1100 		return(ENODEV);
1101 
1102 	ap->a_result = dssize(dev, &dp->d_slice);
1103 
1104 	if ((ap->a_result == -1) &&
1105 	   (dp->d_info.d_dsflags & DSO_DEVICEMAPPER)) {
1106 		ap->a_head.a_dev = dp->d_rawdev;
1107 		return dev_doperate(&ap->a_head);
1108 	}
1109 	return(0);
1110 }
1111 
1112 int
1113 diskdump(struct dev_dump_args *ap)
1114 {
1115 	cdev_t dev = ap->a_head.a_dev;
1116 	struct disk *dp = dev->si_disk;
1117 	u_int64_t size, offset;
1118 	int error;
1119 
1120 	error = disk_dumpcheck(dev, &size, &ap->a_blkno, &ap->a_secsize);
1121 	/* XXX: this should probably go in disk_dumpcheck somehow */
1122 	if (ap->a_length != 0) {
1123 		size *= DEV_BSIZE;
1124 		offset = ap->a_blkno * DEV_BSIZE;
1125 		if ((ap->a_offset < offset) ||
1126 		    (ap->a_offset + ap->a_length - offset > size)) {
1127 			kprintf("Attempt to write outside dump device boundaries.\n");
1128 			error = ENOSPC;
1129 		}
1130 	}
1131 
1132 	if (error == 0) {
1133 		ap->a_head.a_dev = dp->d_rawdev;
1134 		error = dev_doperate(&ap->a_head);
1135 	}
1136 
1137 	return(error);
1138 }
1139 
1140 
1141 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
1142     0, sizeof(struct diskslices), "sizeof(struct diskslices)");
1143 
1144 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
1145     0, sizeof(struct disk), "sizeof(struct disk)");
1146 
1147 /*
1148  * Reorder interval for burst write allowance and minor write
1149  * allowance.
1150  *
1151  * We always want to trickle some writes in to make use of the
1152  * disk's zone cache.  Bursting occurs on a longer interval and only
1153  * runningbufspace is well over the hirunningspace limit.
1154  */
1155 int bioq_reorder_burst_interval = 60;	/* should be multiple of minor */
1156 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_interval,
1157 	   CTLFLAG_RW, &bioq_reorder_burst_interval, 0, "");
1158 int bioq_reorder_minor_interval = 5;
1159 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_interval,
1160 	   CTLFLAG_RW, &bioq_reorder_minor_interval, 0, "");
1161 
1162 int bioq_reorder_burst_bytes = 3000000;
1163 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_bytes,
1164 	   CTLFLAG_RW, &bioq_reorder_burst_bytes, 0, "");
1165 int bioq_reorder_minor_bytes = 262144;
1166 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_bytes,
1167 	   CTLFLAG_RW, &bioq_reorder_minor_bytes, 0, "");
1168 
1169 
1170 /*
1171  * Order I/Os.  Generally speaking this code is designed to make better
1172  * use of drive zone caches.  A drive zone cache can typically track linear
1173  * reads or writes for around 16 zones simultaniously.
1174  *
1175  * Read prioritization issues:  It is possible for hundreds of megabytes worth
1176  * of writes to be queued asynchronously.  This creates a huge bottleneck
1177  * for reads which reduce read bandwidth to a trickle.
1178  *
1179  * To solve this problem we generally reorder reads before writes.
1180  *
1181  * However, a large number of random reads can also starve writes and
1182  * make poor use of the drive zone cache so we allow writes to trickle
1183  * in every N reads.
1184  */
1185 void
1186 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
1187 {
1188 	/*
1189 	 * The BIO wants to be ordered.  Adding to the tail also
1190 	 * causes transition to be set to NULL, forcing the ordering
1191 	 * of all prior I/O's.
1192 	 */
1193 	if (bio->bio_buf->b_flags & B_ORDERED) {
1194 		bioq_insert_tail(bioq, bio);
1195 		return;
1196 	}
1197 
1198 	switch(bio->bio_buf->b_cmd) {
1199 	case BUF_CMD_READ:
1200 		if (bioq->transition) {
1201 			/*
1202 			 * Insert before the first write.  Bleedover writes
1203 			 * based on reorder intervals to prevent starvation.
1204 			 */
1205 			TAILQ_INSERT_BEFORE(bioq->transition, bio, bio_act);
1206 			++bioq->reorder;
1207 			if (bioq->reorder % bioq_reorder_minor_interval == 0) {
1208 				bioqwritereorder(bioq);
1209 				if (bioq->reorder >=
1210 				    bioq_reorder_burst_interval) {
1211 					bioq->reorder = 0;
1212 				}
1213 			}
1214 		} else {
1215 			/*
1216 			 * No writes queued (or ordering was forced),
1217 			 * insert at tail.
1218 			 */
1219 			TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act);
1220 		}
1221 		break;
1222 	case BUF_CMD_WRITE:
1223 		/*
1224 		 * Writes are always appended.  If no writes were previously
1225 		 * queued or an ordered tail insertion occured the transition
1226 		 * field will be NULL.
1227 		 */
1228 		TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act);
1229 		if (bioq->transition == NULL)
1230 			bioq->transition = bio;
1231 		break;
1232 	default:
1233 		/*
1234 		 * All other request types are forced to be ordered.
1235 		 */
1236 		bioq_insert_tail(bioq, bio);
1237 		break;
1238 	}
1239 }
1240 
1241 /*
1242  * Move the read-write transition point to prevent reads from
1243  * completely starving our writes.  This brings a number of writes into
1244  * the fold every N reads.
1245  *
1246  * We bring a few linear writes into the fold on a minor interval
1247  * and we bring a non-linear burst of writes into the fold on a major
1248  * interval.  Bursting only occurs if runningbufspace is really high
1249  * (typically from syncs, fsyncs, or HAMMER flushes).
1250  */
1251 static
1252 void
1253 bioqwritereorder(struct bio_queue_head *bioq)
1254 {
1255 	struct bio *bio;
1256 	off_t next_offset;
1257 	size_t left;
1258 	size_t n;
1259 	int check_off;
1260 
1261 	if (bioq->reorder < bioq_reorder_burst_interval ||
1262 	    !buf_runningbufspace_severe()) {
1263 		left = (size_t)bioq_reorder_minor_bytes;
1264 		check_off = 1;
1265 	} else {
1266 		left = (size_t)bioq_reorder_burst_bytes;
1267 		check_off = 0;
1268 	}
1269 
1270 	next_offset = bioq->transition->bio_offset;
1271 	while ((bio = bioq->transition) != NULL &&
1272 	       (check_off == 0 || next_offset == bio->bio_offset)
1273 	) {
1274 		n = bio->bio_buf->b_bcount;
1275 		next_offset = bio->bio_offset + n;
1276 		bioq->transition = TAILQ_NEXT(bio, bio_act);
1277 		if (left < n)
1278 			break;
1279 		left -= n;
1280 	}
1281 }
1282 
1283 /*
1284  * Bounds checking against the media size, used for the raw partition.
1285  * secsize, mediasize and b_blkno must all be the same units.
1286  * Possibly this has to be DEV_BSIZE (512).
1287  */
1288 int
1289 bounds_check_with_mediasize(struct bio *bio, int secsize, uint64_t mediasize)
1290 {
1291 	struct buf *bp = bio->bio_buf;
1292 	int64_t sz;
1293 
1294 	sz = howmany(bp->b_bcount, secsize);
1295 
1296 	if (bio->bio_offset/DEV_BSIZE + sz > mediasize) {
1297 		sz = mediasize - bio->bio_offset/DEV_BSIZE;
1298 		if (sz == 0) {
1299 			/* If exactly at end of disk, return EOF. */
1300 			bp->b_resid = bp->b_bcount;
1301 			return 0;
1302 		}
1303 		if (sz < 0) {
1304 			/* If past end of disk, return EINVAL. */
1305 			bp->b_error = EINVAL;
1306 			return 0;
1307 		}
1308 		/* Otherwise, truncate request. */
1309 		bp->b_bcount = sz * secsize;
1310 	}
1311 
1312 	return 1;
1313 }
1314 
1315 /*
1316  * Disk error is the preface to plaintive error messages
1317  * about failing disk transfers.  It prints messages of the form
1318 
1319 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
1320 
1321  * if the offset of the error in the transfer and a disk label
1322  * are both available.  blkdone should be -1 if the position of the error
1323  * is unknown; the disklabel pointer may be null from drivers that have not
1324  * been converted to use them.  The message is printed with kprintf
1325  * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
1326  * The message should be completed (with at least a newline) with kprintf
1327  * or log(-1, ...), respectively.  There is no trailing space.
1328  */
1329 void
1330 diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt)
1331 {
1332 	struct buf *bp = bio->bio_buf;
1333 	const char *term;
1334 
1335 	switch(bp->b_cmd) {
1336 	case BUF_CMD_READ:
1337 		term = "read";
1338 		break;
1339 	case BUF_CMD_WRITE:
1340 		term = "write";
1341 		break;
1342 	default:
1343 		term = "access";
1344 		break;
1345 	}
1346 	kprintf("%s: %s %sing ", dev->si_name, what, term);
1347 	kprintf("offset %012llx for %d",
1348 		(long long)bio->bio_offset,
1349 		bp->b_bcount);
1350 
1351 	if (donecnt)
1352 		kprintf(" (%d bytes completed)", donecnt);
1353 }
1354 
1355 /*
1356  * Locate a disk device
1357  */
1358 cdev_t
1359 disk_locate(const char *devname)
1360 {
1361 	return devfs_find_device_by_name(devname);
1362 }
1363 
1364 void
1365 disk_config(void *arg)
1366 {
1367 	disk_msg_send_sync(DISK_SYNC, NULL, NULL);
1368 }
1369 
1370 static void
1371 disk_init(void)
1372 {
1373 	struct thread* td_core;
1374 
1375 	disk_msg_cache = objcache_create("disk-msg-cache", 0, 0,
1376 					 NULL, NULL, NULL,
1377 					 objcache_malloc_alloc,
1378 					 objcache_malloc_free,
1379 					 &disk_msg_malloc_args);
1380 
1381 	lwkt_token_init(&disklist_token, 1, "disks");
1382 
1383 	/*
1384 	 * Initialize the reply-only port which acts as a message drain
1385 	 */
1386 	lwkt_initport_replyonly(&disk_dispose_port, disk_msg_autofree_reply);
1387 
1388 	lwkt_gettoken(&disklist_token);
1389 	lwkt_create(disk_msg_core, /*args*/NULL, &td_core, NULL,
1390 		    0, 0, "disk_msg_core");
1391 	tsleep(td_core, 0, "diskcore", 0);
1392 	lwkt_reltoken(&disklist_token);
1393 }
1394 
1395 static void
1396 disk_uninit(void)
1397 {
1398 	objcache_destroy(disk_msg_cache);
1399 }
1400 
1401 /*
1402  * Clean out illegal characters in serial numbers.
1403  */
1404 static void
1405 disk_cleanserial(char *serno)
1406 {
1407 	char c;
1408 
1409 	while ((c = *serno) != 0) {
1410 		if (c >= 'a' && c <= 'z')
1411 			;
1412 		else if (c >= 'A' && c <= 'Z')
1413 			;
1414 		else if (c >= '0' && c <= '9')
1415 			;
1416 		else if (c == '-' || c == '@' || c == '+' || c == '.')
1417 			;
1418 		else
1419 			c = '_';
1420 		*serno++= c;
1421 	}
1422 }
1423 
1424 TUNABLE_INT("kern.disk_debug", &disk_debug_enable);
1425 SYSCTL_INT(_kern, OID_AUTO, disk_debug, CTLFLAG_RW, &disk_debug_enable,
1426 		0, "Enable subr_disk debugging");
1427 
1428 SYSINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_FIRST, disk_init, NULL);
1429 SYSUNINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_ANY, disk_uninit, NULL);
1430