2 * Copyright (c) 2003,2004 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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
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.
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
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 * ----------------------------------------------------------------------------
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.
49 * Redistribution and use in source and binary forms, with or without
50 * modification, are permitted provided that the following conditions
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.
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
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 $
83 #include <sys/param.h>
84 #include <sys/systm.h>
85 #include <sys/kernel.h>
87 #include <sys/sysctl.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>
96 #include <sys/malloc.h>
97 #include <sys/sysctl.h>
98 #include <machine/md_var.h>
99 #include <sys/ctype.h>
100 #include <sys/syslog.h>
101 #include <sys/device.h>
102 #include <sys/msgport.h>
103 #include <sys/msgport2.h>
104 #include <sys/buf2.h>
105 #include <vfs/devfs/devfs.h>
106 #include <sys/thread.h>
107 #include <sys/thread2.h>
109 #include <sys/queue.h>
110 #include <sys/lock.h>
112 static MALLOC_DEFINE(M_DISK, "disk", "disk data");
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);
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;
128 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
129 static struct lwkt_token disklist_token;
131 static struct dev_ops disk_ops = {
132 { "disk", 0, D_DISK },
134 .d_close = diskclose,
136 .d_write = physwrite,
137 .d_ioctl = diskioctl,
138 .d_strategy = diskstrategy,
140 .d_psize = diskpsize,
144 static struct objcache *disk_msg_cache;
146 struct objcache_malloc_args disk_msg_malloc_args = {
147 sizeof(struct disk_msg), M_DISK };
149 static struct lwkt_port disk_dispose_port;
150 static struct lwkt_port disk_msg_port;
154 disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe)
156 struct disk_info *info = &dp->d_info;
157 struct diskslice *sp = &dp->d_slice->dss_slices[slice];
159 struct partinfo part;
166 ops = &disklabel32_ops;
167 msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
168 if (msg && !strcmp(msg, "no disk label")) {
169 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: trying with disklabel64\n");
170 ops = &disklabel64_ops;
171 msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
173 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: label: %s\n", (msg)?msg:"is NULL");
175 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: found %d partitions in the label\n", ops->op_getnumparts(sp->ds_label));
176 if (slice != WHOLE_DISK_SLICE)
177 ops->op_adjust_label_reserved(dp->d_slice, slice, sp);
182 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: lp.opaque: %x\n", sp->ds_label.opaque);
183 for (i = 0; i < ops->op_getnumparts(sp->ds_label); i++) {
184 ops->op_loadpartinfo(sp->ds_label, i, &part);
185 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: partinfo says fstype=%d for part %d\n", part.fstype, i);
188 (ndev = devfs_find_device_by_name("%s%c",
189 dev->si_name, 'a'+ (char)i))) {
190 /* Device already exists and is still valid */
191 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice: reprobing and device remained valid, mark it\n");
192 ndev->si_flags |= SI_REPROBE_TEST;
194 ndev = make_dev(&disk_ops,
195 dkmakeminor(dkunit(dp->d_cdev), slice, i),
196 UID_ROOT, GID_OPERATOR, 0640,
197 "%s%c", dev->si_name, 'a'+ (char)i);
199 if (dp->d_info.d_serialno) {
200 make_dev_alias(ndev, "serno/%s.s%d%c", dp->d_info.d_serialno, slice - 1, 'a' + (char)i);
202 ndev->si_flags |= SI_REPROBE_TEST;
205 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe_slice:end: lp.opaque: %x\n", ndev->si_disk->d_slice->dss_slices[slice].ds_label.opaque);
208 } else if (info->d_dsflags & DSO_COMPATLABEL) {
210 if (sp->ds_size >= 0x100000000ULL)
211 ops = &disklabel64_ops;
213 ops = &disklabel32_ops;
214 sp->ds_label = ops->op_clone_label(info, sp);
216 if (sp->ds_type == DOSPTYP_386BSD /* XXX */)
217 log(LOG_WARNING, "%s: cannot find label (%s)\n",
222 sp->ds_wlabel = FALSE;
225 return (msg ? EINVAL : 0);
230 disk_probe(struct disk *dp, int reprobe)
232 struct disk_info *info = &dp->d_info;
233 cdev_t dev = dp->d_cdev;
237 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe called for %s\n", dp->d_cdev->si_name);
238 KKASSERT (info->d_media_blksize != 0);
239 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: info set!\n");
241 dp->d_slice = dsmakeslicestruct(BASE_SLICE, info);
243 error = mbrinit(dev, info, &(dp->d_slice));
244 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: &dp->d_slice is: %x, %x\n", &dp->d_slice, dp->d_slice);
246 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: mbrinit() failed with error: %d\n", error);
249 devfs_debug(DEVFS_DEBUG_DEBUG, "mbrinit succeeded, found %d slices\n", dp->d_slice->dss_nslices);
250 if (dp->d_slice->dss_nslices == BASE_SLICE) {
251 dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_size = info->d_media_blocks;
252 dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_reserved = 0;
254 (ndev = devfs_find_device_by_name("%ss%d",
255 dev->si_name, COMPATIBILITY_SLICE))) {
256 /* Device already exists and is still valid */
257 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: reprobing and device remained valid, mark it\n");
258 ndev->si_flags |= SI_REPROBE_TEST;
260 ndev = make_dev(&disk_ops,
261 dkmakewholeslice(dkunit(dev), COMPATIBILITY_SLICE),
262 UID_ROOT, GID_OPERATOR, 0640,
263 "%ss%d", dev->si_name, COMPATIBILITY_SLICE);
266 if (dp->d_info.d_serialno) {
267 make_dev_alias(ndev, "serno/%s.s%d",
268 dp->d_info.d_serialno,
269 COMPATIBILITY_SLICE);
271 ndev->si_flags |= SI_REPROBE_TEST;
274 dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_dev = ndev;
275 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: type of slice is :%x\n", dp->d_slice->dss_slices[COMPATIBILITY_SLICE].ds_type );
277 dp->d_slice->dss_first_bsd_slice = COMPATIBILITY_SLICE;
278 disk_probe_slice(dp, ndev, COMPATIBILITY_SLICE, reprobe);
281 for (i = BASE_SLICE; i < dp->d_slice->dss_nslices; i++) {
283 (ndev = devfs_find_device_by_name("%ss%d",
284 dev->si_name, i-1))) {
285 /* Device already exists and is still valid */
286 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe: reprobing and device remained valid, mark it\n");
287 ndev->si_flags |= SI_REPROBE_TEST;
289 ndev = make_dev(&disk_ops,
290 dkmakewholeslice(dkunit(dev), i),
291 UID_ROOT, GID_OPERATOR, 0640,
292 "%ss%d", dev->si_name, i-1);
293 if (dp->d_info.d_serialno) {
294 make_dev_alias(ndev, "serno/%s.s%d",
295 dp->d_info.d_serialno,
299 ndev->si_flags |= SI_REPROBE_TEST;
301 dp->d_slice->dss_slices[i].ds_reserved = 0;
302 dp->d_slice->dss_slices[i].ds_dev = ndev;
303 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_probe-> type of slice is :%x\n", dp->d_slice->dss_slices[i].ds_type );
304 if (dp->d_slice->dss_slices[i].ds_type == DOSPTYP_386BSD) {
305 if (!dp->d_slice->dss_first_bsd_slice)
306 dp->d_slice->dss_first_bsd_slice = i;
307 disk_probe_slice(dp, ndev, i, reprobe);
315 disk_msg_core(void *arg)
319 struct diskslice *sp;
323 lwkt_initport_thread(&disk_msg_port, curthread);
327 msg = (disk_msg_t)lwkt_waitport(&disk_msg_port, 0);
328 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_msg_core, new msg: %x\n", (unsigned int)msg->hdr.u.ms_result);
330 switch (msg->hdr.u.ms_result) {
332 case DISK_DISK_PROBE:
333 dp = (struct disk *)msg->load;
337 case DISK_DISK_DESTROY:
338 dp = (struct disk *)msg->load;
339 devfs_destroy_subnames(dp->d_cdev->si_name);
340 devfs_destroy_dev(dp->d_cdev);
341 lwkt_gettoken(&ilock, &disklist_token);
342 LIST_REMOVE(dp, d_list);
343 lwkt_reltoken(&ilock);
345 devfs_destroy_dev(dp->d_rawdev); /* XXX: needed? when? */
347 if (dp->d_info.d_serialno) {
348 kfree(dp->d_info.d_serialno, M_TEMP);
349 dp->d_info.d_serialno = NULL;
354 dp = (struct disk *)msg->load;
355 devfs_destroy_subnames(dp->d_cdev->si_name);
358 case DISK_SLICE_REPROBE:
359 dp = (struct disk *)msg->load;
360 sp = (struct diskslice *)msg->load2;
361 devfs_clr_subnames_flag(sp->ds_dev->si_name, SI_REPROBE_TEST);
362 devfs_debug(DEVFS_DEBUG_DEBUG,
363 "DISK_SLICE_REPROBE: %s\n",
364 sp->ds_dev->si_name);
365 disk_probe_slice(dp, sp->ds_dev, dkslice(sp->ds_dev), 1);
366 devfs_destroy_subnames_without_flag(sp->ds_dev->si_name,
370 case DISK_DISK_REPROBE:
371 dp = (struct disk *)msg->load;
372 devfs_clr_subnames_flag(dp->d_cdev->si_name, SI_REPROBE_TEST);
373 devfs_debug(DEVFS_DEBUG_DEBUG,
374 "DISK_DISK_REPROBE: %s\n",
375 dp->d_cdev->si_name);
377 devfs_destroy_subnames_without_flag(dp->d_cdev->si_name,
385 devfs_debug(DEVFS_DEBUG_WARNING, "disk_msg_core: unknown message received at core\n");
388 lwkt_replymsg((lwkt_msg_t)msg, 0);
395 * Acts as a message drain. Any message that is replied to here gets destroyed and
399 disk_msg_autofree_reply(lwkt_port_t port, lwkt_msg_t msg)
401 objcache_put(disk_msg_cache, msg);
406 disk_msg_send(uint32_t cmd, void *load, void *load2)
409 lwkt_port_t port = &disk_msg_port;
411 disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
413 lwkt_initmsg(&disk_msg->hdr, &disk_dispose_port, 0);
415 disk_msg->hdr.u.ms_result = cmd;
416 disk_msg->load = load;
417 disk_msg->load2 = load2;
419 lwkt_sendmsg(port, (lwkt_msg_t)disk_msg);
423 disk_msg_send_sync(uint32_t cmd, void *load, void *load2)
425 struct lwkt_port rep_port;
426 disk_msg_t disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
427 disk_msg_t msg_incoming;
428 lwkt_port_t port = &disk_msg_port;
430 lwkt_initport_thread(&rep_port, curthread);
431 lwkt_initmsg(&disk_msg->hdr, &rep_port, 0);
433 disk_msg->hdr.u.ms_result = cmd;
434 disk_msg->load = load;
435 disk_msg->load2 = load2;
438 lwkt_sendmsg(port, (lwkt_msg_t)disk_msg);
439 msg_incoming = lwkt_waitport(&rep_port, 0);
443 * Create a raw device for the dev_ops template (which is returned). Also
444 * create a slice and unit managed disk and overload the user visible
445 * device space with it.
447 * NOTE: The returned raw device is NOT a slice and unit managed device.
448 * It is an actual raw device representing the raw disk as specified by
449 * the passed dev_ops. The disk layer not only returns such a raw device,
450 * it also uses it internally when passing (modified) commands through.
453 disk_create(int unit, struct disk *dp, struct dev_ops *raw_ops)
458 rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
459 UID_ROOT, GID_OPERATOR, 0640,
460 "%s%d", raw_ops->head.name, unit);
463 bzero(dp, sizeof(*dp));
465 dp->d_rawdev = rawdev;
466 dp->d_raw_ops = raw_ops;
467 dp->d_dev_ops = &disk_ops;
468 dp->d_cdev = make_dev(&disk_ops,
469 dkmakewholedisk(unit),
470 UID_ROOT, GID_OPERATOR, 0640,
471 "%s%d", raw_ops->head.name, unit);
473 dp->d_cdev->si_disk = dp;
475 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_create called for %s\n",
476 dp->d_cdev->si_name);
477 lwkt_gettoken(&ilock, &disklist_token);
478 LIST_INSERT_HEAD(&disklist, dp, d_list);
479 lwkt_reltoken(&ilock);
480 return (dp->d_rawdev);
485 _setdiskinfo(struct disk *disk, struct disk_info *info)
489 devfs_debug(DEVFS_DEBUG_DEBUG,
490 "_setdiskinfo called for disk -1-: %x\n", disk);
491 oldserialno = disk->d_info.d_serialno;
492 bcopy(info, &disk->d_info, sizeof(disk->d_info));
493 info = &disk->d_info;
496 * The serial number is duplicated so the caller can throw
499 if (info->d_serialno && info->d_serialno[0]) {
500 info->d_serialno = kstrdup(info->d_serialno, M_TEMP);
502 make_dev_alias(disk->d_cdev, "serno/%s",
506 info->d_serialno = NULL;
509 kfree(oldserialno, M_TEMP);
512 * The caller may set d_media_size or d_media_blocks and we
513 * calculate the other.
515 KKASSERT(info->d_media_size == 0 || info->d_media_blksize == 0);
516 if (info->d_media_size == 0 && info->d_media_blocks) {
517 info->d_media_size = (u_int64_t)info->d_media_blocks *
518 info->d_media_blksize;
519 } else if (info->d_media_size && info->d_media_blocks == 0 &&
520 info->d_media_blksize) {
521 info->d_media_blocks = info->d_media_size /
522 info->d_media_blksize;
526 * The si_* fields for rawdev are not set until after the
527 * disk_create() call, so someone using the cooked version
528 * of the raw device (i.e. da0s0) will not get the right
529 * si_iosize_max unless we fix it up here.
531 if (disk->d_cdev && disk->d_rawdev &&
532 disk->d_cdev->si_iosize_max == 0) {
533 disk->d_cdev->si_iosize_max = disk->d_rawdev->si_iosize_max;
534 disk->d_cdev->si_bsize_phys = disk->d_rawdev->si_bsize_phys;
535 disk->d_cdev->si_bsize_best = disk->d_rawdev->si_bsize_best;
540 * Disk drivers must call this routine when media parameters are available
544 disk_setdiskinfo(struct disk *disk, struct disk_info *info)
546 _setdiskinfo(disk, info);
547 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_setdiskinfo called for disk -2-: %x\n", disk);
548 disk_msg_send(DISK_DISK_PROBE, disk, NULL);
552 disk_setdiskinfo_sync(struct disk *disk, struct disk_info *info)
554 _setdiskinfo(disk, info);
555 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_setdiskinfo_sync called for disk -2-: %x\n", disk);
556 disk_msg_send_sync(DISK_DISK_PROBE, disk, NULL);
560 * This routine is called when an adapter detaches. The higher level
561 * managed disk device is destroyed while the lower level raw device is
565 disk_destroy(struct disk *disk)
567 disk_msg_send_sync(DISK_DISK_DESTROY, disk, NULL);
572 disk_dumpcheck(cdev_t dev, u_int64_t *count, u_int64_t *blkno, u_int *secsize)
574 struct partinfo pinfo;
577 bzero(&pinfo, sizeof(pinfo));
578 error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0, proc0.p_ucred);
581 if (pinfo.media_blksize == 0)
583 *count = (u_int64_t)Maxmem * PAGE_SIZE / pinfo.media_blksize;
584 if (dumplo64 < pinfo.reserved_blocks ||
585 dumplo64 + *count > pinfo.media_blocks) {
588 *blkno = dumplo64 + pinfo.media_offset / pinfo.media_blksize;
589 *secsize = pinfo.media_blksize;
594 disk_unprobe(struct disk *disk)
599 disk_msg_send_sync(DISK_UNPROBE, disk, NULL);
603 disk_invalidate (struct disk *disk)
605 devfs_debug(DEVFS_DEBUG_INFO, "disk_invalidate for %s\n", disk->d_cdev->si_name);
607 dsgone(&disk->d_slice);
611 disk_enumerate(struct disk *disk)
616 lwkt_gettoken(&ilock, &disklist_token);
618 dp = (LIST_FIRST(&disklist));
620 dp = (LIST_NEXT(disk, d_list));
621 lwkt_reltoken(&ilock);
628 sysctl_disks(SYSCTL_HANDLER_ARGS)
636 while ((disk = disk_enumerate(disk))) {
638 error = SYSCTL_OUT(req, " ", 1);
644 error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
645 strlen(disk->d_rawdev->si_name));
649 error = SYSCTL_OUT(req, "", 1);
653 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, NULL, 0,
654 sysctl_disks, "A", "names of available disks");
657 * Open a disk device or partition.
661 diskopen(struct dev_open_args *ap)
663 cdev_t dev = ap->a_head.a_dev;
667 devfs_debug(DEVFS_DEBUG_DEBUG, "diskopen: name is %s\n", dev->si_name);
670 * dp can't be NULL here XXX.
678 * Deal with open races
680 while (dp->d_flags & DISKFLAG_LOCK) {
681 dp->d_flags |= DISKFLAG_WANTED;
682 error = tsleep(dp, PCATCH, "diskopen", hz);
686 dp->d_flags |= DISKFLAG_LOCK;
688 devfs_debug(DEVFS_DEBUG_DEBUG, "diskopen: -2- name is %s\n", dev->si_name);
691 * Open the underlying raw device.
693 if (!dsisopen(dp->d_slice)) {
695 if (!pdev->si_iosize_max)
696 pdev->si_iosize_max = dev->si_iosize_max;
698 error = dev_dopen(dp->d_rawdev, ap->a_oflags,
699 ap->a_devtype, ap->a_cred);
703 * Inherit properties from the underlying device now that it is
711 error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags,
712 &dp->d_slice, &dp->d_info);
713 if (!dsisopen(dp->d_slice)) {
714 dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype);
717 dp->d_flags &= ~DISKFLAG_LOCK;
718 if (dp->d_flags & DISKFLAG_WANTED) {
719 dp->d_flags &= ~DISKFLAG_WANTED;
727 * Close a disk device or partition
731 diskclose(struct dev_close_args *ap)
733 cdev_t dev = ap->a_head.a_dev;
740 devfs_debug(DEVFS_DEBUG_DEBUG, "diskclose: name %s\n", dev->si_name);
742 dsclose(dev, ap->a_devtype, dp->d_slice);
743 if (!dsisopen(dp->d_slice)) {
744 devfs_debug(DEVFS_DEBUG_DEBUG, "diskclose is closing underlying device\n");
745 error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype);
751 * First execute the ioctl on the disk device, and if it isn't supported
752 * try running it on the backing device.
756 diskioctl(struct dev_ioctl_args *ap)
758 cdev_t dev = ap->a_head.a_dev;
766 devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl: cmd is: %x (name: %s)\n", ap->a_cmd, dev->si_name);
767 devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl: &dp->d_slice is: %x, %x\n", &dp->d_slice, dp->d_slice);
769 devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl:1: says lp.opaque is: %x\n", dp->d_slice->dss_slices[0].ds_label.opaque);
771 error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag,
772 &dp->d_slice, &dp->d_info);
774 devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl:2: says lp.opaque is: %x\n", dp->d_slice->dss_slices[0].ds_label.opaque);
776 if (error == ENOIOCTL) {
777 devfs_debug(DEVFS_DEBUG_DEBUG, "diskioctl: going for dev_dioctl instead!\n");
778 error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data,
779 ap->a_fflag, ap->a_cred);
785 * Execute strategy routine
789 diskstrategy(struct dev_strategy_args *ap)
791 cdev_t dev = ap->a_head.a_dev;
792 struct bio *bio = ap->a_bio;
799 bio->bio_buf->b_error = ENXIO;
800 bio->bio_buf->b_flags |= B_ERROR;
804 KKASSERT(dev->si_disk == dp);
807 * The dscheck() function will also transform the slice relative
808 * block number i.e. bio->bio_offset into a block number that can be
809 * passed directly to the underlying raw device. If dscheck()
810 * returns NULL it will have handled the bio for us (e.g. EOF
811 * or error due to being beyond the device size).
813 if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL) {
814 dev_dstrategy(dp->d_rawdev, nbio);
816 devfs_debug(DEVFS_DEBUG_DEBUG, "diskstrategy: dscheck NULL!!! biodone time!\n");
823 * Return the partition size in ?blocks?
827 diskpsize(struct dev_psize_args *ap)
829 cdev_t dev = ap->a_head.a_dev;
835 ap->a_result = dssize(dev, &dp->d_slice);
840 * When new device entries are instantiated, make sure they inherit our
841 * si_disk structure and block and iosize limits from the raw device.
843 * This routine is always called synchronously in the context of the
846 * XXX The various io and block size constraints are not always initialized
847 * properly by devices.
851 diskclone(struct dev_clone_args *ap)
853 cdev_t dev = ap->a_head.a_dev;
857 KKASSERT(dp != NULL);
859 dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
860 dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
861 dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
866 diskdump(struct dev_dump_args *ap)
868 cdev_t dev = ap->a_head.a_dev;
869 struct disk *dp = dev->si_disk;
872 error = disk_dumpcheck(dev, &ap->a_count, &ap->a_blkno, &ap->a_secsize);
874 ap->a_head.a_dev = dp->d_rawdev;
875 error = dev_doperate(&ap->a_head);
882 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
883 0, sizeof(struct diskslices), "sizeof(struct diskslices)");
885 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
886 0, sizeof(struct disk), "sizeof(struct disk)");
890 * Seek sort for disks.
892 * The bio_queue keep two queues, sorted in ascending block order. The first
893 * queue holds those requests which are positioned after the current block
894 * (in the first request); the second, which starts at queue->switch_point,
895 * holds requests which came in after their block number was passed. Thus
896 * we implement a one way scan, retracting after reaching the end of the drive
897 * to the first request on the second queue, at which time it becomes the
900 * A one-way scan is natural because of the way UNIX read-ahead blocks are
904 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
910 be = TAILQ_LAST(&bioq->queue, bio_queue);
912 * If the queue is empty or we are an
913 * ordered transaction, then it's easy.
915 if ((bq = bioq_first(bioq)) == NULL ||
916 (bio->bio_buf->b_flags & B_ORDERED) != 0) {
917 bioq_insert_tail(bioq, bio);
919 } else if (bioq->insert_point != NULL) {
922 * A certain portion of the list is
923 * "locked" to preserve ordering, so
924 * we can only insert after the insert
927 bq = bioq->insert_point;
931 * If we lie before the last removed (currently active)
932 * request, and are not inserting ourselves into the
933 * "locked" portion of the list, then we must add ourselves
934 * to the second request list.
936 if (bio->bio_offset < bioq->last_offset) {
937 bq = bioq->switch_point;
939 * If we are starting a new secondary list,
943 bioq->switch_point = bio;
944 bioq_insert_tail(bioq, bio);
948 * If we lie ahead of the current switch point,
949 * insert us before the switch point and move
952 if (bio->bio_offset < bq->bio_offset) {
953 bioq->switch_point = bio;
954 TAILQ_INSERT_BEFORE(bq, bio, bio_act);
958 if (bioq->switch_point != NULL)
959 be = TAILQ_PREV(bioq->switch_point,
962 * If we lie between last_offset and bq,
965 if (bio->bio_offset < bq->bio_offset) {
966 TAILQ_INSERT_BEFORE(bq, bio, bio_act);
973 * Request is at/after our current position in the list.
974 * Optimize for sequential I/O by seeing if we go at the tail.
976 if (bio->bio_offset > be->bio_offset) {
977 TAILQ_INSERT_AFTER(&bioq->queue, be, bio, bio_act);
981 /* Otherwise, insertion sort */
982 while ((bn = TAILQ_NEXT(bq, bio_act)) != NULL) {
985 * We want to go after the current request if it is the end
986 * of the first request list, or if the next request is a
987 * larger cylinder than our request.
989 if (bn == bioq->switch_point
990 || bio->bio_offset < bn->bio_offset)
994 TAILQ_INSERT_AFTER(&bioq->queue, bq, bio, bio_act);
998 * Disk error is the preface to plaintive error messages
999 * about failing disk transfers. It prints messages of the form
1001 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
1003 * if the offset of the error in the transfer and a disk label
1004 * are both available. blkdone should be -1 if the position of the error
1005 * is unknown; the disklabel pointer may be null from drivers that have not
1006 * been converted to use them. The message is printed with kprintf
1007 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
1008 * The message should be completed (with at least a newline) with kprintf
1009 * or log(-1, ...), respectively. There is no trailing space.
1012 diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt)
1014 struct buf *bp = bio->bio_buf;
1028 //sname = dsname(dev, unit, slice, part, partname);
1029 kprintf("%s: %s %sing ", dev->si_name, what, term);
1030 kprintf("offset %012llx for %d",
1031 (long long)bio->bio_offset,
1035 kprintf(" (%d bytes completed)", donecnt);
1039 * Locate a disk device
1042 disk_locate(const char *devname)
1044 return devfs_find_device_by_name(devname);
1049 disk_config(void *arg)
1051 disk_msg_send_sync(DISK_SYNC, NULL, NULL);
1058 struct thread* td_core;
1059 devfs_debug(DEVFS_DEBUG_DEBUG, "disk_init() called\n");
1061 disk_msg_cache = objcache_create("disk-msg-cache", 0, 0,
1063 objcache_malloc_alloc,
1064 objcache_malloc_free,
1065 &disk_msg_malloc_args );
1067 lwkt_token_init(&disklist_token);
1069 /* Initialize the reply-only port which acts as a message drain */
1070 lwkt_initport_replyonly(&disk_dispose_port, disk_msg_autofree_reply);
1072 lwkt_create(disk_msg_core, /*args*/NULL, &td_core, NULL,
1073 0, 0, "disk_msg_core");
1075 tsleep(td_core, 0, "diskcore", 0);
1082 devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_uninit() called\n");
1084 objcache_destroy(disk_msg_cache);
1089 SYSINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_FIRST, disk_init, NULL);
1090 SYSUNINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_ANY, disk_uninit, NULL);