xref: /dflybsd-src/sys/kern/subr_disk.c (revision 3e82b46c18bc48fdb3c1d60729c7661b3a0bf6bf)
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);
117 static void disk_probe(struct disk *dp);
118 
119 static d_open_t diskopen;
120 static d_close_t diskclose;
121 static d_ioctl_t diskioctl;
122 static d_strategy_t diskstrategy;
123 static d_psize_t diskpsize;
124 static d_clone_t diskclone;
125 static d_dump_t diskdump;
126 
127 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
128 
129 static struct dev_ops disk_ops = {
130 	{ "disk", 0, D_DISK },
131 	.d_open = diskopen,
132 	.d_close = diskclose,
133 	.d_read = physread,
134 	.d_write = physwrite,
135 	.d_ioctl = diskioctl,
136 	.d_strategy = diskstrategy,
137 	.d_dump = diskdump,
138 	.d_psize = diskpsize,
139 	.d_clone = diskclone
140 };
141 
142 static struct objcache 	*disk_msg_cache;
143 
144 struct objcache_malloc_args disk_msg_malloc_args = {
145 	sizeof(struct disk_msg), M_DISK };
146 
147 static struct lwkt_port disk_dispose_port;
148 static struct lwkt_port disk_msg_port;
149 
150 
151 static int
152 disk_probe_slice(struct disk *dp, cdev_t dev, int slice)
153 {
154 	struct disk_info *info = &dp->d_info;
155 	struct diskslice *sp = &dp->d_slice->dss_slices[slice];
156 	disklabel_ops_t ops;
157 	struct partinfo part;
158 	const char *msg;
159 	cdev_t ndev;
160 	unsigned long i;
161 
162 	//lp.opaque = NULL;
163 
164 	ops = &disklabel32_ops;
165 	msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
166 	if (msg && !strcmp(msg, "no disk label")) {
167 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: trying with disklabel64\n");
168 		ops = &disklabel64_ops;
169 		msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
170 	}
171 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: label: %s\n", (msg)?msg:"is NULL");
172 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: found %d partitions in the label\n", ops->op_getnumparts(sp->ds_label));
173 	if (msg == NULL) {
174 		if (slice != WHOLE_DISK_SLICE)
175 			ops->op_adjust_label_reserved(dp->d_slice, slice, sp);
176 		else
177 			sp->ds_reserved = 0;
178 
179 		sp->ds_ops = ops;
180 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: lp.opaque: %x\n", sp->ds_label.opaque);
181 		for (i = 0; i < ops->op_getnumparts(sp->ds_label); i++) {
182 			ops->op_loadpartinfo(sp->ds_label, i, &part);
183 			devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: partinfo says fstype=%d for part %d\n", part.fstype, i);
184 			if (part.fstype) {
185 				ndev = make_only_devfs_dev(&disk_ops,
186 					dkmakeminor(dkunit(dp->d_cdev), slice, i),
187 					UID_ROOT, GID_OPERATOR, 0640,
188 					"%s%c", dev->si_name, 'a'+ (char)i);
189 #if 0
190 				make_dev_alias(ndev, "disk-by-id/diskTEST-sliceTEST-part%d", i);
191 #endif
192 				ndev->si_disk = dp;
193 				devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice:end: lp.opaque: %x\n", ndev->si_disk->d_slice->dss_slices[slice].ds_label.opaque);
194 			}
195 		}
196 	} else if (info->d_dsflags & DSO_COMPATLABEL) {
197 		msg = NULL;
198 		if (sp->ds_size >= 0x100000000ULL)
199 			ops = &disklabel64_ops;
200 		else
201 			ops = &disklabel32_ops;
202 		sp->ds_label = ops->op_clone_label(info, sp);
203 	} else {
204 		if (sp->ds_type == DOSPTYP_386BSD /* XXX */)
205 			log(LOG_WARNING, "%s: cannot find label (%s)\n",
206 			    dev->si_name, msg);
207 	}
208 
209 	if (msg == NULL) {
210 		sp->ds_wlabel = FALSE;
211 	}
212 
213 	return (msg ? EINVAL : 0);
214 }
215 
216 
217 static void
218 disk_probe(struct disk *dp)
219 {
220 	struct disk_info *info = &dp->d_info;
221 	cdev_t dev = dp->d_cdev;
222 	cdev_t ndev;
223 	int error, i;
224 
225 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe called for %s\n", dp->d_cdev->si_name);
226 	KKASSERT (info->d_media_blksize != 0);
227 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: info set!\n");
228 
229 	dp->d_slice = dsmakeslicestruct(BASE_SLICE, info);
230 
231 	error = mbrinit(dev, info, &(dp->d_slice));
232 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: &dp->d_slice is: %x, %x\n", &dp->d_slice, dp->d_slice);
233 	if (error != 0) {
234 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: mbrinit() failed with error: %d\n", error);
235 		return;
236 	} else {
237 		devfs_debug(DEVFS_DEBUG_DEBUG, "mbrinit succeeded, found %d slices\n", dp->d_slice->dss_nslices);
238 		if (dp->d_slice->dss_nslices == BASE_SLICE) {
239 			dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_size = info->d_media_blocks;
240 			dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_reserved = 0;
241 			ndev = make_only_devfs_dev(&disk_ops,
242 			    dkmakewholeslice(dkunit(dev), COMPATIBILITY_SLICE),
243 			    UID_ROOT, GID_OPERATOR, 0640,
244 			    "%ss%d", dev->si_name, COMPATIBILITY_SLICE);
245 
246 			ndev->si_disk = dp;
247 			dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_dev = ndev;
248 			devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: type of slice is :%x\n", dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_type );
249 			//if (dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_type == DOSPTYP_386BSD) {
250 			dp->d_slice->dss_first_bsd_slice = COMPATIBILITY_SLICE;
251 			disk_probe_slice(dp, ndev, COMPATIBILITY_SLICE);
252 			//}
253 		}
254 		for (i = BASE_SLICE; i < dp->d_slice->dss_nslices; i++) {
255 			ndev = make_only_devfs_dev(&disk_ops,
256 			    dkmakewholeslice(dkunit(dev), i),
257 			    UID_ROOT, GID_OPERATOR, 0640,
258 			    "%ss%d", dev->si_name, i-1);
259 			make_dev_alias(ndev, "disk-by-id/diskTEST-slice%d", i-1);
260 
261 			ndev->si_disk = dp;
262 			dp->d_slice->dss_slices[i].ds_dev = ndev;
263 			devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe-> type of slice is :%x\n", dp->d_slice->dss_slices[i].ds_type );
264 			if (dp->d_slice->dss_slices[i].ds_type == DOSPTYP_386BSD) {
265 				if (!dp->d_slice->dss_first_bsd_slice)
266 					dp->d_slice->dss_first_bsd_slice = i;
267 				disk_probe_slice(dp, ndev, i);
268 			}
269 		}
270 	}
271 }
272 
273 
274 static void
275 disk_msg_core(void *arg)
276 {
277     uint8_t  run = 1;
278 	struct disk	*dp;
279 	struct diskslice *sp;
280     disk_msg_t msg;
281 
282 
283 	lwkt_initport_thread(&disk_msg_port, curthread);
284 	wakeup(curthread);
285 
286     while (run) {
287         msg = (disk_msg_t)lwkt_waitport(&disk_msg_port, 0);
288 		devfs_debug(DEVFS_DEBUG_DEBUG, "disk_msg_core, new msg: %x\n", (unsigned int)msg->hdr.u.ms_result);
289 
290         switch (msg->hdr.u.ms_result) {
291 
292         case DISK_DISK_PROBE:
293 			dp = (struct disk *)msg->load;
294 			disk_probe(dp);
295 			break;
296 
297 		case DISK_DISK_DESTROY:
298 			dp = (struct disk *)msg->load;
299 			devfs_destroy_subnames(dp->d_cdev->si_name);
300 			devfs_destroy_dev(dp->d_cdev);
301 			//devfs_destroy_dev(dp->d_rawdev); //XXX: needed? when?
302 			break;
303 
304 		case DISK_SLICE_REPROBE:
305 			dp = (struct disk *)msg->load;
306 			sp = (struct diskslice *)msg->load2;
307 			devfs_destroy_subnames(sp->ds_dev->si_name);
308 			disk_probe_slice(dp, sp->ds_dev, dkslice(sp->ds_dev));
309 			break;
310 
311 		case DISK_DISK_REPROBE:
312 			dp = (struct disk *)msg->load;
313 			devfs_destroy_subnames(dp->d_cdev->si_name);
314 			disk_probe(dp);
315 			break;
316 
317 		case DISK_SYNC:
318 			break;
319 
320         default:
321             devfs_debug(DEVFS_DEBUG_WARNING, "disk_msg_core: unknown message received at core\n");
322         }
323 
324         lwkt_replymsg((lwkt_msg_t)msg, 0);
325     }
326 	lwkt_exit();
327 }
328 
329 
330 /**
331  * Acts as a message drain. Any message that is replied to here gets destroyed and
332  * the memory freed.
333  **/
334 static void
335 disk_msg_autofree_reply(lwkt_port_t port, lwkt_msg_t msg)
336 {
337     objcache_put(disk_msg_cache, msg);
338 }
339 
340 
341 void
342 disk_msg_send(uint32_t cmd, void *load, void *load2)
343 {
344     disk_msg_t disk_msg;
345 	lwkt_port_t port = &disk_msg_port;
346 
347     disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
348 
349     lwkt_initmsg(&disk_msg->hdr, &disk_dispose_port, 0);
350 
351 	disk_msg->hdr.u.ms_result = cmd;
352 	disk_msg->load = load;
353 	disk_msg->load2 = load2;
354 	KKASSERT(port);
355     lwkt_sendmsg(port, (lwkt_msg_t)disk_msg);
356 }
357 
358 /*
359  * Create a raw device for the dev_ops template (which is returned).  Also
360  * create a slice and unit managed disk and overload the user visible
361  * device space with it.
362  *
363  * NOTE: The returned raw device is NOT a slice and unit managed device.
364  * It is an actual raw device representing the raw disk as specified by
365  * the passed dev_ops.  The disk layer not only returns such a raw device,
366  * it also uses it internally when passing (modified) commands through.
367  */
368 cdev_t
369 disk_create(int unit, struct disk *dp, struct dev_ops *raw_ops)
370 {
371 	cdev_t rawdev;
372 
373 	rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
374 			    UID_ROOT, GID_OPERATOR, 0640,
375 			    "%s%d", raw_ops->head.name, unit);
376 
377 
378 	bzero(dp, sizeof(*dp));
379 
380 	dp->d_rawdev = rawdev;
381 	dp->d_raw_ops = raw_ops;
382 	dp->d_dev_ops = &disk_ops;
383 	dp->d_cdev = make_only_devfs_dev(&disk_ops,
384 			    dkmakewholedisk(unit),
385 			    UID_ROOT, GID_OPERATOR, 0640,
386 			    "%s%d", raw_ops->head.name, unit);
387 
388 	dp->d_cdev->si_disk = dp;
389 
390 	disk_ops.head.data = dp;
391 
392 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_create called for %s\n", dp->d_cdev->si_name);
393 	LIST_INSERT_HEAD(&disklist, dp, d_list);
394 	return (dp->d_rawdev);
395 }
396 
397 /*
398  * Disk drivers must call this routine when media parameters are available
399  * or have changed.
400  */
401 void
402 disk_setdiskinfo(struct disk *disk, struct disk_info *info)
403 {
404 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_setdiskinfo called for disk -1-: %x\n", disk);
405 	bcopy(info, &disk->d_info, sizeof(disk->d_info));
406 	info = &disk->d_info;
407 
408 	KKASSERT(info->d_media_size == 0 || info->d_media_blksize == 0);
409 	if (info->d_media_size == 0 && info->d_media_blocks) {
410 		info->d_media_size = (u_int64_t)info->d_media_blocks *
411 				     info->d_media_blksize;
412 	} else if (info->d_media_size && info->d_media_blocks == 0 &&
413 		   info->d_media_blksize) {
414 		info->d_media_blocks = info->d_media_size /
415 				       info->d_media_blksize;
416 	}
417 
418 	/*
419 	 * The si_* fields for rawdev are not set until after the
420 	 * disk_create() call, so someone using the cooked version
421 	 * of the raw device (i.e. da0s0) will not get the right
422 	 * si_iosize_max unless we fix it up here.
423 	 */
424 	if (disk->d_cdev && disk->d_rawdev &&
425 	    disk->d_cdev->si_iosize_max == 0) {
426 		disk->d_cdev->si_iosize_max = disk->d_rawdev->si_iosize_max;
427 		disk->d_cdev->si_bsize_phys = disk->d_rawdev->si_bsize_phys;
428 		disk->d_cdev->si_bsize_best = disk->d_rawdev->si_bsize_best;
429 	}
430 
431 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_setdiskinfo called for disk -2-: %x\n", disk);
432 	disk_msg_send(DISK_DISK_PROBE, disk, NULL);
433 }
434 
435 /*
436  * This routine is called when an adapter detaches.  The higher level
437  * managed disk device is destroyed while the lower level raw device is
438  * released.
439  */
440 void
441 disk_destroy(struct disk *disk)
442 {
443 	disk_msg_send(DISK_DISK_DESTROY, disk, NULL);
444 	return;
445 }
446 
447 int
448 disk_dumpcheck(cdev_t dev, u_int64_t *count, u_int64_t *blkno, u_int *secsize)
449 {
450 	struct partinfo pinfo;
451 	int error;
452 
453 	bzero(&pinfo, sizeof(pinfo));
454 	error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0, proc0.p_ucred);
455 	if (error)
456 		return (error);
457 	if (pinfo.media_blksize == 0)
458 		return (ENXIO);
459 	*count = (u_int64_t)Maxmem * PAGE_SIZE / pinfo.media_blksize;
460 	if (dumplo64 < pinfo.reserved_blocks ||
461 	    dumplo64 + *count > pinfo.media_blocks) {
462 		return (ENOSPC);
463 	}
464 	*blkno = dumplo64 + pinfo.media_offset / pinfo.media_blksize;
465 	*secsize = pinfo.media_blksize;
466 	return (0);
467 }
468 
469 void
470 disk_invalidate (struct disk *disk)
471 {
472 	devfs_debug(DEVFS_DEBUG_INFO, "disk_invalidate for %s\n", disk->d_cdev->si_name);
473 	if (disk->d_slice)
474 		dsgone(&disk->d_slice);
475 }
476 
477 struct disk *
478 disk_enumerate(struct disk *disk)
479 {
480 	if (!disk)
481 		return (LIST_FIRST(&disklist));
482 	else
483 		return (LIST_NEXT(disk, d_list));
484 }
485 
486 static
487 int
488 sysctl_disks(SYSCTL_HANDLER_ARGS)
489 {
490 	struct disk *disk;
491 	int error, first;
492 
493 	disk = NULL;
494 	first = 1;
495 
496 	while ((disk = disk_enumerate(disk))) {
497 		if (!first) {
498 			error = SYSCTL_OUT(req, " ", 1);
499 			if (error)
500 				return error;
501 		} else {
502 			first = 0;
503 		}
504 		error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
505 				   strlen(disk->d_rawdev->si_name));
506 		if (error)
507 			return error;
508 	}
509 	error = SYSCTL_OUT(req, "", 1);
510 	return error;
511 }
512 
513 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, NULL, 0,
514     sysctl_disks, "A", "names of available disks");
515 
516 /*
517  * Open a disk device or partition.
518  */
519 static
520 int
521 diskopen(struct dev_open_args *ap)
522 {
523 	cdev_t dev = ap->a_head.a_dev;
524 	struct disk *dp;
525 	int error;
526 
527 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskopen: name is %s\n", dev->si_name);
528 
529 	/*
530 	 * dp can't be NULL here XXX.
531 	 */
532 	dp = dev->si_disk;
533 	if (dp == NULL)
534 		return (ENXIO);
535 	error = 0;
536 
537 	/*
538 	 * Deal with open races
539 	 */
540 	while (dp->d_flags & DISKFLAG_LOCK) {
541 		dp->d_flags |= DISKFLAG_WANTED;
542 		error = tsleep(dp, PCATCH, "diskopen", hz);
543 		if (error)
544 			return (error);
545 	}
546 	dp->d_flags |= DISKFLAG_LOCK;
547 
548 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskopen: -2- name is %s\n", dev->si_name);
549 
550 	/*
551 	 * Open the underlying raw device.
552 	 */
553 	if (!dsisopen(dp->d_slice)) {
554 #if 0
555 		if (!pdev->si_iosize_max)
556 			pdev->si_iosize_max = dev->si_iosize_max;
557 #endif
558 		error = dev_dopen(dp->d_rawdev, ap->a_oflags,
559 				  ap->a_devtype, ap->a_cred);
560 	}
561 #if 0
562 	/*
563 	 * Inherit properties from the underlying device now that it is
564 	 * open.
565 	 */
566 	dev_dclone(dev);
567 #endif
568 
569 	if (error)
570 		goto out;
571 	error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags,
572 		       &dp->d_slice, &dp->d_info);
573 	if (!dsisopen(dp->d_slice)) {
574 		dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype);
575 	}
576 out:
577 	dp->d_flags &= ~DISKFLAG_LOCK;
578 	if (dp->d_flags & DISKFLAG_WANTED) {
579 		dp->d_flags &= ~DISKFLAG_WANTED;
580 		wakeup(dp);
581 	}
582 
583 	return(error);
584 }
585 
586 /*
587  * Close a disk device or partition
588  */
589 static
590 int
591 diskclose(struct dev_close_args *ap)
592 {
593 	cdev_t dev = ap->a_head.a_dev;
594 	struct disk *dp;
595 	int error;
596 
597 	error = 0;
598 	dp = dev->si_disk;
599 
600 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskclose: name %s\n", dev->si_name);
601 
602 	dsclose(dev, ap->a_devtype, dp->d_slice);
603 	if (!dsisopen(dp->d_slice)) {
604 		devfs_debug(DEVFS_DEBUG_DEBUG, "diskclose is closing underlying device\n");
605 		error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype);
606 	}
607 	return (error);
608 }
609 
610 /*
611  * First execute the ioctl on the disk device, and if it isn't supported
612  * try running it on the backing device.
613  */
614 static
615 int
616 diskioctl(struct dev_ioctl_args *ap)
617 {
618 	cdev_t dev = ap->a_head.a_dev;
619 	struct disk *dp;
620 	int error;
621 
622 	dp = dev->si_disk;
623 	if (dp == NULL)
624 		return (ENXIO);
625 
626 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl: cmd is: %x (name: %s)\n", ap->a_cmd, dev->si_name);
627 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl: &dp->d_slice is: %x, %x\n", &dp->d_slice, dp->d_slice);
628 
629 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl:1: says lp.opaque is: %x\n", dp->d_slice->dss_slices[0].ds_label.opaque);
630 
631 	error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag,
632 			&dp->d_slice, &dp->d_info);
633 
634 	devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl:2: says lp.opaque is: %x\n", dp->d_slice->dss_slices[0].ds_label.opaque);
635 
636 	if (error == ENOIOCTL) {
637 		devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl: going for dev_dioctl instead!\n");
638 		error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data,
639 				   ap->a_fflag, ap->a_cred);
640 	}
641 	return (error);
642 }
643 
644 /*
645  * Execute strategy routine
646  */
647 static
648 int
649 diskstrategy(struct dev_strategy_args *ap)
650 {
651 	cdev_t dev = ap->a_head.a_dev;
652 	struct bio *bio = ap->a_bio;
653 	struct bio *nbio;
654 	struct disk *dp;
655 
656 	dp = dev->si_disk;
657 
658 	if (dp == NULL) {
659 		bio->bio_buf->b_error = ENXIO;
660 		bio->bio_buf->b_flags |= B_ERROR;
661 		biodone(bio);
662 		return(0);
663 	}
664 	KKASSERT(dev->si_disk == dp);
665 
666 	/*
667 	 * The dscheck() function will also transform the slice relative
668 	 * block number i.e. bio->bio_offset into a block number that can be
669 	 * passed directly to the underlying raw device.  If dscheck()
670 	 * returns NULL it will have handled the bio for us (e.g. EOF
671 	 * or error due to being beyond the device size).
672 	 */
673 	if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL) {
674 		dev_dstrategy(dp->d_rawdev, nbio);
675 	} else {
676 		devfs_debug(DEVFS_DEBUG_DEBUG, "diskstrategy: dscheck NULL!!! biodone time!\n");
677 		biodone(bio);
678 	}
679 	return(0);
680 }
681 
682 /*
683  * Return the partition size in ?blocks?
684  */
685 static
686 int
687 diskpsize(struct dev_psize_args *ap)
688 {
689 	cdev_t dev = ap->a_head.a_dev;
690 	struct disk *dp;
691 
692 	dp = dev->si_disk;
693 	if (dp == NULL)
694 		return(ENODEV);
695 	ap->a_result = dssize(dev, &dp->d_slice);
696 	return(0);
697 }
698 
699 /*
700  * When new device entries are instantiated, make sure they inherit our
701  * si_disk structure and block and iosize limits from the raw device.
702  *
703  * This routine is always called synchronously in the context of the
704  * client.
705  *
706  * XXX The various io and block size constraints are not always initialized
707  * properly by devices.
708  */
709 static
710 int
711 diskclone(struct dev_clone_args *ap)
712 {
713 	cdev_t dev = ap->a_head.a_dev;
714 	struct disk *dp;
715 //XXX: need changes for devfs
716 	dp = dev->si_ops->head.data;
717 	KKASSERT(dp != NULL);
718 	dev->si_disk = dp;
719 	dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
720 	dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
721 	dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
722 	return(0);
723 }
724 
725 int
726 diskdump(struct dev_dump_args *ap)
727 {
728 	cdev_t dev = ap->a_head.a_dev;
729 	struct disk *dp = dev->si_ops->head.data;
730 	int error;
731 
732 	error = disk_dumpcheck(dev, &ap->a_count, &ap->a_blkno, &ap->a_secsize);
733 	if (error == 0) {
734 		ap->a_head.a_dev = dp->d_rawdev;
735 		error = dev_doperate(&ap->a_head);
736 	}
737 
738 	return(error);
739 }
740 
741 
742 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
743     0, sizeof(struct diskslices), "sizeof(struct diskslices)");
744 
745 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
746     0, sizeof(struct disk), "sizeof(struct disk)");
747 
748 
749 /*
750  * Seek sort for disks.
751  *
752  * The bio_queue keep two queues, sorted in ascending block order.  The first
753  * queue holds those requests which are positioned after the current block
754  * (in the first request); the second, which starts at queue->switch_point,
755  * holds requests which came in after their block number was passed.  Thus
756  * we implement a one way scan, retracting after reaching the end of the drive
757  * to the first request on the second queue, at which time it becomes the
758  * first queue.
759  *
760  * A one-way scan is natural because of the way UNIX read-ahead blocks are
761  * allocated.
762  */
763 void
764 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
765 {
766 	struct bio *bq;
767 	struct bio *bn;
768 	struct bio *be;
769 
770 	be = TAILQ_LAST(&bioq->queue, bio_queue);
771 	/*
772 	 * If the queue is empty or we are an
773 	 * ordered transaction, then it's easy.
774 	 */
775 	if ((bq = bioq_first(bioq)) == NULL ||
776 	    (bio->bio_buf->b_flags & B_ORDERED) != 0) {
777 		bioq_insert_tail(bioq, bio);
778 		return;
779 	} else if (bioq->insert_point != NULL) {
780 
781 		/*
782 		 * A certain portion of the list is
783 		 * "locked" to preserve ordering, so
784 		 * we can only insert after the insert
785 		 * point.
786 		 */
787 		bq = bioq->insert_point;
788 	} else {
789 
790 		/*
791 		 * If we lie before the last removed (currently active)
792 		 * request, and are not inserting ourselves into the
793 		 * "locked" portion of the list, then we must add ourselves
794 		 * to the second request list.
795 		 */
796 		if (bio->bio_offset < bioq->last_offset) {
797 			bq = bioq->switch_point;
798 			/*
799 			 * If we are starting a new secondary list,
800 			 * then it's easy.
801 			 */
802 			if (bq == NULL) {
803 				bioq->switch_point = bio;
804 				bioq_insert_tail(bioq, bio);
805 				return;
806 			}
807 			/*
808 			 * If we lie ahead of the current switch point,
809 			 * insert us before the switch point and move
810 			 * the switch point.
811 			 */
812 			if (bio->bio_offset < bq->bio_offset) {
813 				bioq->switch_point = bio;
814 				TAILQ_INSERT_BEFORE(bq, bio, bio_act);
815 				return;
816 			}
817 		} else {
818 			if (bioq->switch_point != NULL)
819 				be = TAILQ_PREV(bioq->switch_point,
820 						bio_queue, bio_act);
821 			/*
822 			 * If we lie between last_offset and bq,
823 			 * insert before bq.
824 			 */
825 			if (bio->bio_offset < bq->bio_offset) {
826 				TAILQ_INSERT_BEFORE(bq, bio, bio_act);
827 				return;
828 			}
829 		}
830 	}
831 
832 	/*
833 	 * Request is at/after our current position in the list.
834 	 * Optimize for sequential I/O by seeing if we go at the tail.
835 	 */
836 	if (bio->bio_offset > be->bio_offset) {
837 		TAILQ_INSERT_AFTER(&bioq->queue, be, bio, bio_act);
838 		return;
839 	}
840 
841 	/* Otherwise, insertion sort */
842 	while ((bn = TAILQ_NEXT(bq, bio_act)) != NULL) {
843 
844 		/*
845 		 * We want to go after the current request if it is the end
846 		 * of the first request list, or if the next request is a
847 		 * larger cylinder than our request.
848 		 */
849 		if (bn == bioq->switch_point
850 		 || bio->bio_offset < bn->bio_offset)
851 			break;
852 		bq = bn;
853 	}
854 	TAILQ_INSERT_AFTER(&bioq->queue, bq, bio, bio_act);
855 }
856 
857 /*
858  * Disk error is the preface to plaintive error messages
859  * about failing disk transfers.  It prints messages of the form
860 
861 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
862 
863  * if the offset of the error in the transfer and a disk label
864  * are both available.  blkdone should be -1 if the position of the error
865  * is unknown; the disklabel pointer may be null from drivers that have not
866  * been converted to use them.  The message is printed with kprintf
867  * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
868  * The message should be completed (with at least a newline) with kprintf
869  * or log(-1, ...), respectively.  There is no trailing space.
870  */
871 void
872 diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt)
873 {
874 	struct buf *bp = bio->bio_buf;
875 	const char *term;
876 
877 	switch(bp->b_cmd) {
878 	case BUF_CMD_READ:
879 		term = "read";
880 		break;
881 	case BUF_CMD_WRITE:
882 		term = "write";
883 		break;
884 	default:
885 		term = "access";
886 		break;
887 	}
888 	//sname = dsname(dev, unit, slice, part, partname);
889 	kprintf("%s: %s %sing ", dev->si_name, what, term);
890 	kprintf("offset %012llx for %d",
891 		(long long)bio->bio_offset,
892 		bp->b_bcount);
893 
894 	if (donecnt)
895 		kprintf(" (%d bytes completed)", donecnt);
896 }
897 
898 /*
899  * Locate a disk device
900  */
901 cdev_t
902 disk_locate(const char *devname)
903 {
904 	return devfs_find_device_by_name(devname);
905 }
906 
907 
908 void
909 disk_config(void *arg)
910 {
911 	struct lwkt_port rep_port;
912 	disk_msg_t disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
913 	disk_msg_t	msg_incoming;
914 	lwkt_port_t port = &disk_msg_port;
915 
916 	lwkt_initport_thread(&rep_port, curthread);
917 	lwkt_initmsg(&disk_msg->hdr, &rep_port, 0);
918 	kprintf("disk_config: sync'ing up\n");
919 	disk_msg->hdr.u.ms_result = DISK_SYNC;
920 
921 	lwkt_sendmsg(port, (lwkt_msg_t)disk_msg);
922 	msg_incoming = lwkt_waitport(&rep_port, 0);
923 }
924 
925 
926 static void
927 disk_init(void)
928 {
929 	struct thread* td_core;
930 	devfs_debug(DEVFS_DEBUG_DEBUG, "disk_init() called\n");
931 
932     disk_msg_cache = objcache_create("disk-msg-cache", 0, 0,
933 			NULL, NULL, NULL,
934 			objcache_malloc_alloc,
935 			objcache_malloc_free,
936 			&disk_msg_malloc_args );
937 
938 	/* Initialize the reply-only port which acts as a message drain */
939 	lwkt_initport_replyonly(&disk_dispose_port, disk_msg_autofree_reply);
940 
941 	lwkt_create(disk_msg_core, /*args*/NULL, &td_core, NULL,
942 		    0, 0, "disk_msg_core");
943 
944 	tsleep(td_core, 0, "diskcore", 0);
945 }
946 
947 
948 static void
949 disk_uninit(void)
950 {
951 	devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_uninit() called\n");
952 
953 	objcache_destroy(disk_msg_cache);
954 
955 }
956 
957 
958 SYSINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_FIRST, disk_init, NULL);
959 SYSUNINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_ANY, disk_uninit, NULL);
960