2 * Copyright (c) 2003,2004,2009 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/kerneldump.h>
97 #include <sys/malloc.h>
98 #include <sys/sysctl.h>
99 #include <machine/md_var.h>
100 #include <sys/ctype.h>
101 #include <sys/syslog.h>
102 #include <sys/device.h>
103 #include <sys/msgport.h>
104 #include <sys/devfs.h>
105 #include <sys/thread.h>
106 #include <sys/dsched.h>
107 #include <sys/queue.h>
108 #include <sys/lock.h>
109 #include <sys/udev.h>
111 #include <sys/buf2.h>
112 #include <sys/mplock2.h>
113 #include <sys/msgport2.h>
114 #include <sys/thread2.h>
116 static MALLOC_DEFINE(M_DISK, "disk", "disk data");
117 static int disk_debug_enable = 0;
119 static void disk_msg_autofree_reply(lwkt_port_t, lwkt_msg_t);
120 static void disk_msg_core(void *);
121 static int disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe);
122 static void disk_probe(struct disk *dp, int reprobe);
123 static void _setdiskinfo(struct disk *disk, struct disk_info *info);
124 static void bioqwritereorder(struct bio_queue_head *bioq);
125 static void disk_cleanserial(char *serno);
126 static int disk_debug(int, char *, ...) __printflike(2, 3);
128 static d_open_t diskopen;
129 static d_close_t diskclose;
130 static d_ioctl_t diskioctl;
131 static d_strategy_t diskstrategy;
132 static d_psize_t diskpsize;
133 static d_clone_t diskclone;
134 static d_dump_t diskdump;
136 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
137 static struct lwkt_token disklist_token;
139 static struct dev_ops disk_ops = {
140 { "disk", 0, D_DISK | D_MPSAFE },
142 .d_close = diskclose,
144 .d_write = physwrite,
145 .d_ioctl = diskioctl,
146 .d_strategy = diskstrategy,
148 .d_psize = diskpsize,
152 static struct objcache *disk_msg_cache;
154 struct objcache_malloc_args disk_msg_malloc_args = {
155 sizeof(struct disk_msg), M_DISK };
157 static struct lwkt_port disk_dispose_port;
158 static struct lwkt_port disk_msg_port;
161 disk_debug(int level, char *fmt, ...)
166 if (level <= disk_debug_enable)
174 disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe)
176 struct disk_info *info = &dp->d_info;
177 struct diskslice *sp = &dp->d_slice->dss_slices[slice];
179 struct partinfo part;
186 "disk_probe_slice (begin): %s (%s)\n",
187 dev->si_name, dp->d_cdev->si_name);
189 sno = slice ? slice - 1 : 0;
191 ops = &disklabel32_ops;
192 msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
193 if (msg && !strcmp(msg, "no disk label")) {
194 ops = &disklabel64_ops;
195 msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
198 if (slice != WHOLE_DISK_SLICE)
199 ops->op_adjust_label_reserved(dp->d_slice, slice, sp);
204 for (i = 0; i < ops->op_getnumparts(sp->ds_label); i++) {
205 ops->op_loadpartinfo(sp->ds_label, i, &part);
208 (ndev = devfs_find_device_by_name("%s%c",
209 dev->si_name, 'a' + i))
212 * Device already exists and
215 ndev->si_flags |= SI_REPROBE_TEST;
217 ndev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
218 dkmakeminor(dkunit(dp->d_cdev),
220 UID_ROOT, GID_OPERATOR, 0640,
221 "%s%c", dev->si_name, 'a'+ i);
223 udev_dict_set_cstr(ndev, "subsystem", "disk");
224 /* Inherit parent's disk type */
225 if (dp->d_disktype) {
226 udev_dict_set_cstr(ndev, "disk-type",
227 __DECONST(char *, dp->d_disktype));
229 if (dp->d_info.d_serialno) {
232 dp->d_info.d_serialno,
235 ndev->si_flags |= SI_REPROBE_TEST;
239 } else if (info->d_dsflags & DSO_COMPATLABEL) {
241 if (sp->ds_size >= 0x100000000ULL)
242 ops = &disklabel64_ops;
244 ops = &disklabel32_ops;
245 sp->ds_label = ops->op_clone_label(info, sp);
247 if (sp->ds_type == DOSPTYP_386BSD || /* XXX */
248 sp->ds_type == DOSPTYP_NETBSD ||
249 sp->ds_type == DOSPTYP_OPENBSD) {
250 log(LOG_WARNING, "%s: cannot find label (%s)\n",
256 sp->ds_wlabel = FALSE;
259 return (msg ? EINVAL : 0);
263 * This routine is only called for newly minted drives or to reprobe
264 * a drive with no open slices. disk_probe_slice() is called directly
265 * when reprobing partition changes within slices.
268 disk_probe(struct disk *dp, int reprobe)
270 struct disk_info *info = &dp->d_info;
271 cdev_t dev = dp->d_cdev;
274 struct diskslices *osp;
275 struct diskslice *sp;
277 KKASSERT (info->d_media_blksize != 0);
280 dp->d_slice = dsmakeslicestruct(BASE_SLICE, info);
281 disk_debug(1, "disk_probe (begin): %s\n", dp->d_cdev->si_name);
283 error = mbrinit(dev, info, &(dp->d_slice));
289 for (i = 0; i < dp->d_slice->dss_nslices; i++) {
291 * Ignore the whole-disk slice, it has already been created.
293 if (i == WHOLE_DISK_SLICE)
295 sp = &dp->d_slice->dss_slices[i];
298 * Handle s0. s0 is a compatibility slice if there are no
299 * other slices and it has not otherwise been set up, else
302 if (i == COMPATIBILITY_SLICE) {
304 if (sp->ds_type == 0 &&
305 dp->d_slice->dss_nslices == BASE_SLICE) {
306 sp->ds_size = info->d_media_blocks;
315 * Ignore 0-length slices
317 if (sp->ds_size == 0)
321 (ndev = devfs_find_device_by_name("%ss%d",
322 dev->si_name, sno))) {
324 * Device already exists and is still valid
326 ndev->si_flags |= SI_REPROBE_TEST;
329 * Else create new device
331 ndev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
332 dkmakewholeslice(dkunit(dev), i),
333 UID_ROOT, GID_OPERATOR, 0640,
334 "%ss%d", dev->si_name, sno);
335 udev_dict_set_cstr(ndev, "subsystem", "disk");
336 /* Inherit parent's disk type */
337 if (dp->d_disktype) {
338 udev_dict_set_cstr(ndev, "disk-type",
339 __DECONST(char *, dp->d_disktype));
341 if (dp->d_info.d_serialno) {
342 make_dev_alias(ndev, "serno/%s.s%d",
343 dp->d_info.d_serialno, sno);
346 ndev->si_flags |= SI_REPROBE_TEST;
351 * Probe appropriate slices for a disklabel
353 * XXX slice type 1 used by our gpt probe code.
354 * XXX slice type 0 used by mbr compat slice.
356 if (sp->ds_type == DOSPTYP_386BSD ||
357 sp->ds_type == DOSPTYP_NETBSD ||
358 sp->ds_type == DOSPTYP_OPENBSD ||
361 if (dp->d_slice->dss_first_bsd_slice == 0)
362 dp->d_slice->dss_first_bsd_slice = i;
363 disk_probe_slice(dp, ndev, i, reprobe);
367 disk_debug(1, "disk_probe (end): %s\n", dp->d_cdev->si_name);
372 disk_msg_core(void *arg)
375 struct diskslice *sp;
379 lwkt_gettoken(&disklist_token);
380 lwkt_initport_thread(&disk_msg_port, curthread);
381 wakeup(curthread); /* synchronous startup */
382 lwkt_reltoken(&disklist_token);
384 get_mplock(); /* not mpsafe yet? */
388 msg = (disk_msg_t)lwkt_waitport(&disk_msg_port, 0);
390 switch (msg->hdr.u.ms_result) {
391 case DISK_DISK_PROBE:
392 dp = (struct disk *)msg->load;
394 "DISK_DISK_PROBE: %s\n",
395 dp->d_cdev->si_name);
398 case DISK_DISK_DESTROY:
399 dp = (struct disk *)msg->load;
401 "DISK_DISK_DESTROY: %s\n",
402 dp->d_cdev->si_name);
403 devfs_destroy_subnames(dp->d_cdev->si_name);
404 devfs_destroy_dev(dp->d_cdev);
405 lwkt_gettoken(&disklist_token);
406 LIST_REMOVE(dp, d_list);
407 lwkt_reltoken(&disklist_token);
408 if (dp->d_info.d_serialno) {
409 kfree(dp->d_info.d_serialno, M_TEMP);
410 dp->d_info.d_serialno = NULL;
414 dp = (struct disk *)msg->load;
416 "DISK_DISK_UNPROBE: %s\n",
417 dp->d_cdev->si_name);
418 devfs_destroy_subnames(dp->d_cdev->si_name);
420 case DISK_SLICE_REPROBE:
421 dp = (struct disk *)msg->load;
422 sp = (struct diskslice *)msg->load2;
423 devfs_clr_subnames_flag(sp->ds_dev->si_name,
426 "DISK_SLICE_REPROBE: %s\n",
427 sp->ds_dev->si_name);
428 disk_probe_slice(dp, sp->ds_dev,
429 dkslice(sp->ds_dev), 1);
430 devfs_destroy_subnames_without_flag(
431 sp->ds_dev->si_name, SI_REPROBE_TEST);
433 case DISK_DISK_REPROBE:
434 dp = (struct disk *)msg->load;
435 devfs_clr_subnames_flag(dp->d_cdev->si_name, SI_REPROBE_TEST);
437 "DISK_DISK_REPROBE: %s\n",
438 dp->d_cdev->si_name);
440 devfs_destroy_subnames_without_flag(
441 dp->d_cdev->si_name, SI_REPROBE_TEST);
444 disk_debug(1, "DISK_SYNC\n");
447 devfs_debug(DEVFS_DEBUG_WARNING,
448 "disk_msg_core: unknown message "
449 "received at core\n");
452 lwkt_replymsg(&msg->hdr, 0);
459 * Acts as a message drain. Any message that is replied to here gets
460 * destroyed and the memory freed.
463 disk_msg_autofree_reply(lwkt_port_t port, lwkt_msg_t msg)
465 objcache_put(disk_msg_cache, msg);
470 disk_msg_send(uint32_t cmd, void *load, void *load2)
473 lwkt_port_t port = &disk_msg_port;
475 disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
477 lwkt_initmsg(&disk_msg->hdr, &disk_dispose_port, 0);
479 disk_msg->hdr.u.ms_result = cmd;
480 disk_msg->load = load;
481 disk_msg->load2 = load2;
483 lwkt_sendmsg(port, &disk_msg->hdr);
487 disk_msg_send_sync(uint32_t cmd, void *load, void *load2)
489 struct lwkt_port rep_port;
493 disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
494 port = &disk_msg_port;
496 /* XXX could probably use curthread's built-in msgport */
497 lwkt_initport_thread(&rep_port, curthread);
498 lwkt_initmsg(&disk_msg->hdr, &rep_port, 0);
500 disk_msg->hdr.u.ms_result = cmd;
501 disk_msg->load = load;
502 disk_msg->load2 = load2;
504 lwkt_sendmsg(port, &disk_msg->hdr);
505 lwkt_waitmsg(&disk_msg->hdr, 0);
506 objcache_put(disk_msg_cache, disk_msg);
510 * Create a raw device for the dev_ops template (which is returned). Also
511 * create a slice and unit managed disk and overload the user visible
512 * device space with it.
514 * NOTE: The returned raw device is NOT a slice and unit managed device.
515 * It is an actual raw device representing the raw disk as specified by
516 * the passed dev_ops. The disk layer not only returns such a raw device,
517 * it also uses it internally when passing (modified) commands through.
520 disk_create(int unit, struct disk *dp, struct dev_ops *raw_ops)
522 return disk_create_named(NULL, unit, dp, raw_ops);
526 disk_create_named(const char *name, int unit, struct disk *dp, struct dev_ops *raw_ops)
530 disk_debug(1, "disk_create (begin): %s%d\n", name, unit);
533 rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
534 UID_ROOT, GID_OPERATOR, 0640, "%s", name);
536 rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
537 UID_ROOT, GID_OPERATOR, 0640,
538 "%s%d", raw_ops->head.name, unit);
541 bzero(dp, sizeof(*dp));
543 dp->d_rawdev = rawdev;
544 dp->d_raw_ops = raw_ops;
545 dp->d_dev_ops = &disk_ops;
548 dp->d_cdev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
549 dkmakewholedisk(unit), UID_ROOT, GID_OPERATOR, 0640,
552 dp->d_cdev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
553 dkmakewholedisk(unit),
554 UID_ROOT, GID_OPERATOR, 0640,
555 "%s%d", raw_ops->head.name, unit);
558 udev_dict_set_cstr(dp->d_cdev, "subsystem", "disk");
559 dp->d_cdev->si_disk = dp;
562 dsched_disk_create_callback(dp, name, unit);
564 dsched_disk_create_callback(dp, raw_ops->head.name, unit);
566 lwkt_gettoken(&disklist_token);
567 LIST_INSERT_HEAD(&disklist, dp, d_list);
568 lwkt_reltoken(&disklist_token);
570 disk_debug(1, "disk_create (end): %s%d\n",
571 (name != NULL)?(name):(raw_ops->head.name), unit);
573 return (dp->d_rawdev);
577 disk_setdisktype(struct disk *disk, const char *type)
579 KKASSERT(disk != NULL);
581 disk->d_disktype = type;
582 return udev_dict_set_cstr(disk->d_cdev, "disk-type", __DECONST(char *, type));
586 _setdiskinfo(struct disk *disk, struct disk_info *info)
590 oldserialno = disk->d_info.d_serialno;
591 bcopy(info, &disk->d_info, sizeof(disk->d_info));
592 info = &disk->d_info;
595 "_setdiskinfo: %s\n",
596 disk->d_cdev->si_name);
599 * The serial number is duplicated so the caller can throw
602 if (info->d_serialno && info->d_serialno[0]) {
603 info->d_serialno = kstrdup(info->d_serialno, M_TEMP);
604 disk_cleanserial(info->d_serialno);
606 make_dev_alias(disk->d_cdev, "serno/%s",
610 info->d_serialno = NULL;
613 kfree(oldserialno, M_TEMP);
615 dsched_disk_update_callback(disk, info);
618 * The caller may set d_media_size or d_media_blocks and we
619 * calculate the other.
621 KKASSERT(info->d_media_size == 0 || info->d_media_blksize == 0);
622 if (info->d_media_size == 0 && info->d_media_blocks) {
623 info->d_media_size = (u_int64_t)info->d_media_blocks *
624 info->d_media_blksize;
625 } else if (info->d_media_size && info->d_media_blocks == 0 &&
626 info->d_media_blksize) {
627 info->d_media_blocks = info->d_media_size /
628 info->d_media_blksize;
632 * The si_* fields for rawdev are not set until after the
633 * disk_create() call, so someone using the cooked version
634 * of the raw device (i.e. da0s0) will not get the right
635 * si_iosize_max unless we fix it up here.
637 if (disk->d_cdev && disk->d_rawdev &&
638 disk->d_cdev->si_iosize_max == 0) {
639 disk->d_cdev->si_iosize_max = disk->d_rawdev->si_iosize_max;
640 disk->d_cdev->si_bsize_phys = disk->d_rawdev->si_bsize_phys;
641 disk->d_cdev->si_bsize_best = disk->d_rawdev->si_bsize_best;
644 /* Add the serial number to the udev_dictionary */
645 if (info->d_serialno)
646 udev_dict_set_cstr(disk->d_cdev, "serno", info->d_serialno);
650 * Disk drivers must call this routine when media parameters are available
654 disk_setdiskinfo(struct disk *disk, struct disk_info *info)
656 _setdiskinfo(disk, info);
657 disk_msg_send(DISK_DISK_PROBE, disk, NULL);
659 "disk_setdiskinfo: sent probe for %s\n",
660 disk->d_cdev->si_name);
664 disk_setdiskinfo_sync(struct disk *disk, struct disk_info *info)
666 _setdiskinfo(disk, info);
667 disk_msg_send_sync(DISK_DISK_PROBE, disk, NULL);
669 "disk_setdiskinfo_sync: sent probe for %s\n",
670 disk->d_cdev->si_name);
674 * This routine is called when an adapter detaches. The higher level
675 * managed disk device is destroyed while the lower level raw device is
679 disk_destroy(struct disk *disk)
681 dsched_disk_destroy_callback(disk);
682 disk_msg_send_sync(DISK_DISK_DESTROY, disk, NULL);
687 disk_dumpcheck(cdev_t dev, u_int64_t *size, u_int64_t *blkno, u_int32_t *secsize)
689 struct partinfo pinfo;
692 bzero(&pinfo, sizeof(pinfo));
693 error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0,
694 proc0.p_ucred, NULL);
698 if (pinfo.media_blksize == 0)
701 if (blkno) /* XXX: make sure this reserved stuff is right */
702 *blkno = pinfo.reserved_blocks +
703 pinfo.media_offset / pinfo.media_blksize;
705 *secsize = pinfo.media_blksize;
707 *size = (pinfo.media_blocks - pinfo.reserved_blocks);
713 disk_dumpconf(cdev_t dev, u_int onoff)
715 struct dumperinfo di;
716 u_int64_t size, blkno;
721 return set_dumper(NULL);
723 error = disk_dumpcheck(dev, &size, &blkno, &secsize);
728 bzero(&di, sizeof(struct dumperinfo));
729 di.dumper = diskdump;
731 di.blocksize = secsize;
732 di.mediaoffset = blkno * DEV_BSIZE;
733 di.mediasize = size * DEV_BSIZE;
735 return set_dumper(&di);
739 disk_unprobe(struct disk *disk)
744 disk_msg_send_sync(DISK_UNPROBE, disk, NULL);
748 disk_invalidate (struct disk *disk)
750 dsgone(&disk->d_slice);
754 disk_enumerate(struct disk *disk)
758 lwkt_gettoken(&disklist_token);
760 dp = (LIST_FIRST(&disklist));
762 dp = (LIST_NEXT(disk, d_list));
763 lwkt_reltoken(&disklist_token);
770 sysctl_disks(SYSCTL_HANDLER_ARGS)
778 while ((disk = disk_enumerate(disk))) {
780 error = SYSCTL_OUT(req, " ", 1);
786 error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
787 strlen(disk->d_rawdev->si_name));
791 error = SYSCTL_OUT(req, "", 1);
795 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, NULL, 0,
796 sysctl_disks, "A", "names of available disks");
799 * Open a disk device or partition.
803 diskopen(struct dev_open_args *ap)
805 cdev_t dev = ap->a_head.a_dev;
810 * dp can't be NULL here XXX.
812 * d_slice will be NULL if setdiskinfo() has not been called yet.
813 * setdiskinfo() is typically called whether the disk is present
814 * or not (e.g. CD), but the base disk device is created first
815 * and there may be a race.
818 if (dp == NULL || dp->d_slice == NULL)
823 * Deal with open races
826 while (dp->d_flags & DISKFLAG_LOCK) {
827 dp->d_flags |= DISKFLAG_WANTED;
828 error = tsleep(dp, PCATCH, "diskopen", hz);
834 dp->d_flags |= DISKFLAG_LOCK;
837 * Open the underlying raw device.
839 if (!dsisopen(dp->d_slice)) {
841 if (!pdev->si_iosize_max)
842 pdev->si_iosize_max = dev->si_iosize_max;
844 error = dev_dopen(dp->d_rawdev, ap->a_oflags,
845 ap->a_devtype, ap->a_cred);
849 * Inherit properties from the underlying device now that it is
857 error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags,
858 &dp->d_slice, &dp->d_info);
859 if (!dsisopen(dp->d_slice)) {
860 dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype);
863 dp->d_flags &= ~DISKFLAG_LOCK;
864 if (dp->d_flags & DISKFLAG_WANTED) {
865 dp->d_flags &= ~DISKFLAG_WANTED;
874 * Close a disk device or partition
878 diskclose(struct dev_close_args *ap)
880 cdev_t dev = ap->a_head.a_dev;
888 dsclose(dev, ap->a_devtype, dp->d_slice);
889 if (!dsisopen(dp->d_slice)) {
890 error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype);
897 * First execute the ioctl on the disk device, and if it isn't supported
898 * try running it on the backing device.
902 diskioctl(struct dev_ioctl_args *ap)
904 cdev_t dev = ap->a_head.a_dev;
913 devfs_debug(DEVFS_DEBUG_DEBUG,
914 "diskioctl: cmd is: %lx (name: %s)\n",
915 ap->a_cmd, dev->si_name);
916 devfs_debug(DEVFS_DEBUG_DEBUG,
917 "diskioctl: &dp->d_slice is: %p, %p\n",
918 &dp->d_slice, dp->d_slice);
920 if (ap->a_cmd == DIOCGKERNELDUMP) {
921 u = *(u_int *)ap->a_data;
922 return disk_dumpconf(dev, u);
925 if (&dp->d_slice == NULL || dp->d_slice == NULL) {
929 error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag,
930 &dp->d_slice, &dp->d_info);
934 if (error == ENOIOCTL) {
935 error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data,
936 ap->a_fflag, ap->a_cred, NULL);
942 * Execute strategy routine
946 diskstrategy(struct dev_strategy_args *ap)
948 cdev_t dev = ap->a_head.a_dev;
949 struct bio *bio = ap->a_bio;
956 bio->bio_buf->b_error = ENXIO;
957 bio->bio_buf->b_flags |= B_ERROR;
961 KKASSERT(dev->si_disk == dp);
964 * The dscheck() function will also transform the slice relative
965 * block number i.e. bio->bio_offset into a block number that can be
966 * passed directly to the underlying raw device. If dscheck()
967 * returns NULL it will have handled the bio for us (e.g. EOF
968 * or error due to being beyond the device size).
970 if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL) {
971 dsched_queue(dp, nbio);
979 * Return the partition size in ?blocks?
983 diskpsize(struct dev_psize_args *ap)
985 cdev_t dev = ap->a_head.a_dev;
991 ap->a_result = dssize(dev, &dp->d_slice);
996 * When new device entries are instantiated, make sure they inherit our
997 * si_disk structure and block and iosize limits from the raw device.
999 * This routine is always called synchronously in the context of the
1002 * XXX The various io and block size constraints are not always initialized
1003 * properly by devices.
1007 diskclone(struct dev_clone_args *ap)
1009 cdev_t dev = ap->a_head.a_dev;
1013 KKASSERT(dp != NULL);
1015 dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
1016 dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
1017 dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
1022 diskdump(struct dev_dump_args *ap)
1024 cdev_t dev = ap->a_head.a_dev;
1025 struct disk *dp = dev->si_disk;
1026 u_int64_t size, offset;
1029 error = disk_dumpcheck(dev, &size, &ap->a_blkno, &ap->a_secsize);
1030 /* XXX: this should probably go in disk_dumpcheck somehow */
1031 if (ap->a_length != 0) {
1033 offset = ap->a_blkno * DEV_BSIZE;
1034 if ((ap->a_offset < offset) ||
1035 (ap->a_offset + ap->a_length - offset > size)) {
1036 kprintf("Attempt to write outside dump device boundaries.\n");
1042 ap->a_head.a_dev = dp->d_rawdev;
1043 error = dev_doperate(&ap->a_head);
1050 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
1051 0, sizeof(struct diskslices), "sizeof(struct diskslices)");
1053 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
1054 0, sizeof(struct disk), "sizeof(struct disk)");
1057 * Reorder interval for burst write allowance and minor write
1060 * We always want to trickle some writes in to make use of the
1061 * disk's zone cache. Bursting occurs on a longer interval and only
1062 * runningbufspace is well over the hirunningspace limit.
1064 int bioq_reorder_burst_interval = 60; /* should be multiple of minor */
1065 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_interval,
1066 CTLFLAG_RW, &bioq_reorder_burst_interval, 0, "");
1067 int bioq_reorder_minor_interval = 5;
1068 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_interval,
1069 CTLFLAG_RW, &bioq_reorder_minor_interval, 0, "");
1071 int bioq_reorder_burst_bytes = 3000000;
1072 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_bytes,
1073 CTLFLAG_RW, &bioq_reorder_burst_bytes, 0, "");
1074 int bioq_reorder_minor_bytes = 262144;
1075 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_bytes,
1076 CTLFLAG_RW, &bioq_reorder_minor_bytes, 0, "");
1080 * Order I/Os. Generally speaking this code is designed to make better
1081 * use of drive zone caches. A drive zone cache can typically track linear
1082 * reads or writes for around 16 zones simultaniously.
1084 * Read prioritization issues: It is possible for hundreds of megabytes worth
1085 * of writes to be queued asynchronously. This creates a huge bottleneck
1086 * for reads which reduce read bandwidth to a trickle.
1088 * To solve this problem we generally reorder reads before writes.
1090 * However, a large number of random reads can also starve writes and
1091 * make poor use of the drive zone cache so we allow writes to trickle
1095 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
1098 * The BIO wants to be ordered. Adding to the tail also
1099 * causes transition to be set to NULL, forcing the ordering
1100 * of all prior I/O's.
1102 if (bio->bio_buf->b_flags & B_ORDERED) {
1103 bioq_insert_tail(bioq, bio);
1107 switch(bio->bio_buf->b_cmd) {
1109 if (bioq->transition) {
1111 * Insert before the first write. Bleedover writes
1112 * based on reorder intervals to prevent starvation.
1114 TAILQ_INSERT_BEFORE(bioq->transition, bio, bio_act);
1116 if (bioq->reorder % bioq_reorder_minor_interval == 0) {
1117 bioqwritereorder(bioq);
1118 if (bioq->reorder >=
1119 bioq_reorder_burst_interval) {
1125 * No writes queued (or ordering was forced),
1128 TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act);
1133 * Writes are always appended. If no writes were previously
1134 * queued or an ordered tail insertion occured the transition
1135 * field will be NULL.
1137 TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act);
1138 if (bioq->transition == NULL)
1139 bioq->transition = bio;
1143 * All other request types are forced to be ordered.
1145 bioq_insert_tail(bioq, bio);
1151 * Move the read-write transition point to prevent reads from
1152 * completely starving our writes. This brings a number of writes into
1153 * the fold every N reads.
1155 * We bring a few linear writes into the fold on a minor interval
1156 * and we bring a non-linear burst of writes into the fold on a major
1157 * interval. Bursting only occurs if runningbufspace is really high
1158 * (typically from syncs, fsyncs, or HAMMER flushes).
1162 bioqwritereorder(struct bio_queue_head *bioq)
1170 if (bioq->reorder < bioq_reorder_burst_interval ||
1171 !buf_runningbufspace_severe()) {
1172 left = (size_t)bioq_reorder_minor_bytes;
1175 left = (size_t)bioq_reorder_burst_bytes;
1179 next_offset = bioq->transition->bio_offset;
1180 while ((bio = bioq->transition) != NULL &&
1181 (check_off == 0 || next_offset == bio->bio_offset)
1183 n = bio->bio_buf->b_bcount;
1184 next_offset = bio->bio_offset + n;
1185 bioq->transition = TAILQ_NEXT(bio, bio_act);
1193 * Bounds checking against the media size, used for the raw partition.
1194 * secsize, mediasize and b_blkno must all be the same units.
1195 * Possibly this has to be DEV_BSIZE (512).
1198 bounds_check_with_mediasize(struct bio *bio, int secsize, uint64_t mediasize)
1200 struct buf *bp = bio->bio_buf;
1203 sz = howmany(bp->b_bcount, secsize);
1205 if (bio->bio_offset/DEV_BSIZE + sz > mediasize) {
1206 sz = mediasize - bio->bio_offset/DEV_BSIZE;
1208 /* If exactly at end of disk, return EOF. */
1209 bp->b_resid = bp->b_bcount;
1213 /* If past end of disk, return EINVAL. */
1214 bp->b_error = EINVAL;
1217 /* Otherwise, truncate request. */
1218 bp->b_bcount = sz * secsize;
1225 * Disk error is the preface to plaintive error messages
1226 * about failing disk transfers. It prints messages of the form
1228 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
1230 * if the offset of the error in the transfer and a disk label
1231 * are both available. blkdone should be -1 if the position of the error
1232 * is unknown; the disklabel pointer may be null from drivers that have not
1233 * been converted to use them. The message is printed with kprintf
1234 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
1235 * The message should be completed (with at least a newline) with kprintf
1236 * or log(-1, ...), respectively. There is no trailing space.
1239 diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt)
1241 struct buf *bp = bio->bio_buf;
1255 kprintf("%s: %s %sing ", dev->si_name, what, term);
1256 kprintf("offset %012llx for %d",
1257 (long long)bio->bio_offset,
1261 kprintf(" (%d bytes completed)", donecnt);
1265 * Locate a disk device
1268 disk_locate(const char *devname)
1270 return devfs_find_device_by_name(devname);
1274 disk_config(void *arg)
1276 disk_msg_send_sync(DISK_SYNC, NULL, NULL);
1282 struct thread* td_core;
1284 disk_msg_cache = objcache_create("disk-msg-cache", 0, 0,
1286 objcache_malloc_alloc,
1287 objcache_malloc_free,
1288 &disk_msg_malloc_args);
1290 lwkt_token_init(&disklist_token, 1, "disks");
1293 * Initialize the reply-only port which acts as a message drain
1295 lwkt_initport_replyonly(&disk_dispose_port, disk_msg_autofree_reply);
1297 lwkt_gettoken(&disklist_token);
1298 lwkt_create(disk_msg_core, /*args*/NULL, &td_core, NULL,
1299 0, 0, "disk_msg_core");
1300 tsleep(td_core, 0, "diskcore", 0);
1301 lwkt_reltoken(&disklist_token);
1307 objcache_destroy(disk_msg_cache);
1311 * Clean out illegal characters in serial numbers.
1314 disk_cleanserial(char *serno)
1318 while ((c = *serno) != 0) {
1319 if (c >= 'a' && c <= 'z')
1321 else if (c >= 'A' && c <= 'Z')
1323 else if (c >= '0' && c <= '9')
1325 else if (c == '-' || c == '@' || c == '+' || c == '.')
1333 TUNABLE_INT("kern.disk_debug", &disk_debug_enable);
1334 SYSCTL_INT(_kern, OID_AUTO, disk_debug, CTLFLAG_RW, &disk_debug_enable,
1335 0, "Enable subr_disk debugging");
1337 SYSINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_FIRST, disk_init, NULL);
1338 SYSUNINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_ANY, disk_uninit, NULL);