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>
110 #include <sys/uuid.h>
112 #include <sys/buf2.h>
113 #include <sys/mplock2.h>
114 #include <sys/msgport2.h>
115 #include <sys/thread2.h>
117 static MALLOC_DEFINE(M_DISK, "disk", "disk data");
118 static int disk_debug_enable = 0;
120 static void disk_msg_autofree_reply(lwkt_port_t, lwkt_msg_t);
121 static void disk_msg_core(void *);
122 static int disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe);
123 static void disk_probe(struct disk *dp, int reprobe);
124 static void _setdiskinfo(struct disk *disk, struct disk_info *info);
125 static void bioqwritereorder(struct bio_queue_head *bioq);
126 static void disk_cleanserial(char *serno);
127 static int disk_debug(int, char *, ...) __printflike(2, 3);
129 static d_open_t diskopen;
130 static d_close_t diskclose;
131 static d_ioctl_t diskioctl;
132 static d_strategy_t diskstrategy;
133 static d_psize_t diskpsize;
134 static d_clone_t diskclone;
135 static d_dump_t diskdump;
137 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
138 static struct lwkt_token disklist_token;
140 static struct dev_ops disk_ops = {
141 { "disk", 0, D_DISK | D_MPSAFE },
143 .d_close = diskclose,
145 .d_write = physwrite,
146 .d_ioctl = diskioctl,
147 .d_strategy = diskstrategy,
149 .d_psize = diskpsize,
153 static struct objcache *disk_msg_cache;
155 struct objcache_malloc_args disk_msg_malloc_args = {
156 sizeof(struct disk_msg), M_DISK };
158 static struct lwkt_port disk_dispose_port;
159 static struct lwkt_port disk_msg_port;
162 disk_debug(int level, char *fmt, ...)
167 if (level <= disk_debug_enable)
175 disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe)
177 struct disk_info *info = &dp->d_info;
178 struct diskslice *sp = &dp->d_slice->dss_slices[slice];
180 struct partinfo part;
188 "disk_probe_slice (begin): %s (%s)\n",
189 dev->si_name, dp->d_cdev->si_name);
191 sno = slice ? slice - 1 : 0;
193 ops = &disklabel32_ops;
194 msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
195 if (msg && !strcmp(msg, "no disk label")) {
196 ops = &disklabel64_ops;
197 msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
200 if (slice != WHOLE_DISK_SLICE)
201 ops->op_adjust_label_reserved(dp->d_slice, slice, sp);
206 for (i = 0; i < ops->op_getnumparts(sp->ds_label); i++) {
207 ops->op_loadpartinfo(sp->ds_label, i, &part);
210 (ndev = devfs_find_device_by_name("%s%c",
211 dev->si_name, 'a' + i))
214 * Device already exists and
217 ndev->si_flags |= SI_REPROBE_TEST;
220 * Destroy old UUID alias
222 destroy_dev_alias(ndev, "part-by-uuid/*");
224 /* Create UUID alias */
225 if (!kuuid_is_nil(&part.storage_uuid)) {
226 snprintf_uuid(uuid_buf,
234 ndev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
235 dkmakeminor(dkunit(dp->d_cdev),
237 UID_ROOT, GID_OPERATOR, 0640,
238 "%s%c", dev->si_name, 'a'+ i);
240 udev_dict_set_cstr(ndev, "subsystem", "disk");
241 /* Inherit parent's disk type */
242 if (dp->d_disktype) {
243 udev_dict_set_cstr(ndev, "disk-type",
244 __DECONST(char *, dp->d_disktype));
247 /* Create serno alias */
248 if (dp->d_info.d_serialno) {
251 dp->d_info.d_serialno,
255 /* Create UUID alias */
256 if (!kuuid_is_nil(&part.storage_uuid)) {
257 snprintf_uuid(uuid_buf,
264 ndev->si_flags |= SI_REPROBE_TEST;
268 } else if (info->d_dsflags & DSO_COMPATLABEL) {
270 if (sp->ds_size >= 0x100000000ULL)
271 ops = &disklabel64_ops;
273 ops = &disklabel32_ops;
274 sp->ds_label = ops->op_clone_label(info, sp);
276 if (sp->ds_type == DOSPTYP_386BSD || /* XXX */
277 sp->ds_type == DOSPTYP_NETBSD ||
278 sp->ds_type == DOSPTYP_OPENBSD) {
279 log(LOG_WARNING, "%s: cannot find label (%s)\n",
285 sp->ds_wlabel = FALSE;
288 return (msg ? EINVAL : 0);
292 * This routine is only called for newly minted drives or to reprobe
293 * a drive with no open slices. disk_probe_slice() is called directly
294 * when reprobing partition changes within slices.
297 disk_probe(struct disk *dp, int reprobe)
299 struct disk_info *info = &dp->d_info;
300 cdev_t dev = dp->d_cdev;
303 struct diskslices *osp;
304 struct diskslice *sp;
307 KKASSERT (info->d_media_blksize != 0);
310 dp->d_slice = dsmakeslicestruct(BASE_SLICE, info);
311 disk_debug(1, "disk_probe (begin): %s\n", dp->d_cdev->si_name);
313 error = mbrinit(dev, info, &(dp->d_slice));
319 for (i = 0; i < dp->d_slice->dss_nslices; i++) {
321 * Ignore the whole-disk slice, it has already been created.
323 if (i == WHOLE_DISK_SLICE)
328 * Ignore the compatibility slice s0 if it's a device mapper
331 if ((i == COMPATIBILITY_SLICE) &&
332 (info->d_dsflags & DSO_DEVICEMAPPER))
336 sp = &dp->d_slice->dss_slices[i];
339 * Handle s0. s0 is a compatibility slice if there are no
340 * other slices and it has not otherwise been set up, else
343 if (i == COMPATIBILITY_SLICE) {
345 if (sp->ds_type == 0 &&
346 dp->d_slice->dss_nslices == BASE_SLICE) {
347 sp->ds_size = info->d_media_blocks;
356 * Ignore 0-length slices
358 if (sp->ds_size == 0)
362 (ndev = devfs_find_device_by_name("%ss%d",
363 dev->si_name, sno))) {
365 * Device already exists and is still valid
367 ndev->si_flags |= SI_REPROBE_TEST;
370 * Destroy old UUID alias
372 destroy_dev_alias(ndev, "slice-by-uuid/*");
374 /* Create UUID alias */
375 if (!kuuid_is_nil(&sp->ds_stor_uuid)) {
376 snprintf_uuid(uuid_buf, sizeof(uuid_buf),
378 make_dev_alias(ndev, "slice-by-uuid/%s",
383 * Else create new device
385 ndev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
386 dkmakewholeslice(dkunit(dev), i),
387 UID_ROOT, GID_OPERATOR, 0640,
388 (info->d_dsflags & DSO_DEVICEMAPPER)?
389 "%s.s%d" : "%ss%d", dev->si_name, sno);
390 udev_dict_set_cstr(ndev, "subsystem", "disk");
391 /* Inherit parent's disk type */
392 if (dp->d_disktype) {
393 udev_dict_set_cstr(ndev, "disk-type",
394 __DECONST(char *, dp->d_disktype));
397 /* Create serno alias */
398 if (dp->d_info.d_serialno) {
399 make_dev_alias(ndev, "serno/%s.s%d",
400 dp->d_info.d_serialno, sno);
403 /* Create UUID alias */
404 if (!kuuid_is_nil(&sp->ds_stor_uuid)) {
405 snprintf_uuid(uuid_buf, sizeof(uuid_buf),
407 make_dev_alias(ndev, "slice-by-uuid/%s",
412 ndev->si_flags |= SI_REPROBE_TEST;
417 * Probe appropriate slices for a disklabel
419 * XXX slice type 1 used by our gpt probe code.
420 * XXX slice type 0 used by mbr compat slice.
422 if (sp->ds_type == DOSPTYP_386BSD ||
423 sp->ds_type == DOSPTYP_NETBSD ||
424 sp->ds_type == DOSPTYP_OPENBSD ||
427 if (dp->d_slice->dss_first_bsd_slice == 0)
428 dp->d_slice->dss_first_bsd_slice = i;
429 disk_probe_slice(dp, ndev, i, reprobe);
433 disk_debug(1, "disk_probe (end): %s\n", dp->d_cdev->si_name);
438 disk_msg_core(void *arg)
441 struct diskslice *sp;
445 lwkt_gettoken(&disklist_token);
446 lwkt_initport_thread(&disk_msg_port, curthread);
447 wakeup(curthread); /* synchronous startup */
448 lwkt_reltoken(&disklist_token);
450 get_mplock(); /* not mpsafe yet? */
454 msg = (disk_msg_t)lwkt_waitport(&disk_msg_port, 0);
456 switch (msg->hdr.u.ms_result) {
457 case DISK_DISK_PROBE:
458 dp = (struct disk *)msg->load;
460 "DISK_DISK_PROBE: %s\n",
461 dp->d_cdev->si_name);
464 case DISK_DISK_DESTROY:
465 dp = (struct disk *)msg->load;
467 "DISK_DISK_DESTROY: %s\n",
468 dp->d_cdev->si_name);
469 devfs_destroy_subnames(dp->d_cdev->si_name);
470 devfs_destroy_dev(dp->d_cdev);
471 lwkt_gettoken(&disklist_token);
472 LIST_REMOVE(dp, d_list);
473 lwkt_reltoken(&disklist_token);
474 if (dp->d_info.d_serialno) {
475 kfree(dp->d_info.d_serialno, M_TEMP);
476 dp->d_info.d_serialno = NULL;
480 dp = (struct disk *)msg->load;
482 "DISK_DISK_UNPROBE: %s\n",
483 dp->d_cdev->si_name);
484 devfs_destroy_subnames(dp->d_cdev->si_name);
486 case DISK_SLICE_REPROBE:
487 dp = (struct disk *)msg->load;
488 sp = (struct diskslice *)msg->load2;
489 devfs_clr_subnames_flag(sp->ds_dev->si_name,
492 "DISK_SLICE_REPROBE: %s\n",
493 sp->ds_dev->si_name);
494 disk_probe_slice(dp, sp->ds_dev,
495 dkslice(sp->ds_dev), 1);
496 devfs_destroy_subnames_without_flag(
497 sp->ds_dev->si_name, SI_REPROBE_TEST);
499 case DISK_DISK_REPROBE:
500 dp = (struct disk *)msg->load;
501 devfs_clr_subnames_flag(dp->d_cdev->si_name, SI_REPROBE_TEST);
503 "DISK_DISK_REPROBE: %s\n",
504 dp->d_cdev->si_name);
506 devfs_destroy_subnames_without_flag(
507 dp->d_cdev->si_name, SI_REPROBE_TEST);
510 disk_debug(1, "DISK_SYNC\n");
513 devfs_debug(DEVFS_DEBUG_WARNING,
514 "disk_msg_core: unknown message "
515 "received at core\n");
518 lwkt_replymsg(&msg->hdr, 0);
525 * Acts as a message drain. Any message that is replied to here gets
526 * destroyed and the memory freed.
529 disk_msg_autofree_reply(lwkt_port_t port, lwkt_msg_t msg)
531 objcache_put(disk_msg_cache, msg);
536 disk_msg_send(uint32_t cmd, void *load, void *load2)
539 lwkt_port_t port = &disk_msg_port;
541 disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
543 lwkt_initmsg(&disk_msg->hdr, &disk_dispose_port, 0);
545 disk_msg->hdr.u.ms_result = cmd;
546 disk_msg->load = load;
547 disk_msg->load2 = load2;
549 lwkt_sendmsg(port, &disk_msg->hdr);
553 disk_msg_send_sync(uint32_t cmd, void *load, void *load2)
555 struct lwkt_port rep_port;
559 disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
560 port = &disk_msg_port;
562 /* XXX could probably use curthread's built-in msgport */
563 lwkt_initport_thread(&rep_port, curthread);
564 lwkt_initmsg(&disk_msg->hdr, &rep_port, 0);
566 disk_msg->hdr.u.ms_result = cmd;
567 disk_msg->load = load;
568 disk_msg->load2 = load2;
570 lwkt_sendmsg(port, &disk_msg->hdr);
571 lwkt_waitmsg(&disk_msg->hdr, 0);
572 objcache_put(disk_msg_cache, disk_msg);
576 * Create a raw device for the dev_ops template (which is returned). Also
577 * create a slice and unit managed disk and overload the user visible
578 * device space with it.
580 * NOTE: The returned raw device is NOT a slice and unit managed device.
581 * It is an actual raw device representing the raw disk as specified by
582 * the passed dev_ops. The disk layer not only returns such a raw device,
583 * it also uses it internally when passing (modified) commands through.
586 disk_create(int unit, struct disk *dp, struct dev_ops *raw_ops)
588 return disk_create_named(NULL, unit, dp, raw_ops);
592 disk_create_named(const char *name, int unit, struct disk *dp, struct dev_ops *raw_ops)
596 disk_debug(1, "disk_create (begin): %s%d\n", name, unit);
599 rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
600 UID_ROOT, GID_OPERATOR, 0640, "%s", name);
602 rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
603 UID_ROOT, GID_OPERATOR, 0640,
604 "%s%d", raw_ops->head.name, unit);
607 bzero(dp, sizeof(*dp));
609 dp->d_rawdev = rawdev;
610 dp->d_raw_ops = raw_ops;
611 dp->d_dev_ops = &disk_ops;
614 dp->d_cdev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
615 dkmakewholedisk(unit), UID_ROOT, GID_OPERATOR, 0640,
618 dp->d_cdev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
619 dkmakewholedisk(unit),
620 UID_ROOT, GID_OPERATOR, 0640,
621 "%s%d", raw_ops->head.name, unit);
624 udev_dict_set_cstr(dp->d_cdev, "subsystem", "disk");
625 dp->d_cdev->si_disk = dp;
628 dsched_disk_create_callback(dp, name, unit);
630 dsched_disk_create_callback(dp, raw_ops->head.name, unit);
632 lwkt_gettoken(&disklist_token);
633 LIST_INSERT_HEAD(&disklist, dp, d_list);
634 lwkt_reltoken(&disklist_token);
636 disk_debug(1, "disk_create (end): %s%d\n",
637 (name != NULL)?(name):(raw_ops->head.name), unit);
639 return (dp->d_rawdev);
643 disk_setdisktype(struct disk *disk, const char *type)
645 KKASSERT(disk != NULL);
647 disk->d_disktype = type;
648 return udev_dict_set_cstr(disk->d_cdev, "disk-type", __DECONST(char *, type));
652 _setdiskinfo(struct disk *disk, struct disk_info *info)
656 oldserialno = disk->d_info.d_serialno;
657 bcopy(info, &disk->d_info, sizeof(disk->d_info));
658 info = &disk->d_info;
661 "_setdiskinfo: %s\n",
662 disk->d_cdev->si_name);
665 * The serial number is duplicated so the caller can throw
668 if (info->d_serialno && info->d_serialno[0]) {
669 info->d_serialno = kstrdup(info->d_serialno, M_TEMP);
670 disk_cleanserial(info->d_serialno);
672 make_dev_alias(disk->d_cdev, "serno/%s",
676 info->d_serialno = NULL;
679 kfree(oldserialno, M_TEMP);
681 dsched_disk_update_callback(disk, info);
684 * The caller may set d_media_size or d_media_blocks and we
685 * calculate the other.
687 KKASSERT(info->d_media_size == 0 || info->d_media_blocks == 0);
688 if (info->d_media_size == 0 && info->d_media_blocks) {
689 info->d_media_size = (u_int64_t)info->d_media_blocks *
690 info->d_media_blksize;
691 } else if (info->d_media_size && info->d_media_blocks == 0 &&
692 info->d_media_blksize) {
693 info->d_media_blocks = info->d_media_size /
694 info->d_media_blksize;
698 * The si_* fields for rawdev are not set until after the
699 * disk_create() call, so someone using the cooked version
700 * of the raw device (i.e. da0s0) will not get the right
701 * si_iosize_max unless we fix it up here.
703 if (disk->d_cdev && disk->d_rawdev &&
704 disk->d_cdev->si_iosize_max == 0) {
705 disk->d_cdev->si_iosize_max = disk->d_rawdev->si_iosize_max;
706 disk->d_cdev->si_bsize_phys = disk->d_rawdev->si_bsize_phys;
707 disk->d_cdev->si_bsize_best = disk->d_rawdev->si_bsize_best;
710 /* Add the serial number to the udev_dictionary */
711 if (info->d_serialno)
712 udev_dict_set_cstr(disk->d_cdev, "serno", info->d_serialno);
716 * Disk drivers must call this routine when media parameters are available
720 disk_setdiskinfo(struct disk *disk, struct disk_info *info)
722 _setdiskinfo(disk, info);
723 disk_msg_send(DISK_DISK_PROBE, disk, NULL);
725 "disk_setdiskinfo: sent probe for %s\n",
726 disk->d_cdev->si_name);
730 disk_setdiskinfo_sync(struct disk *disk, struct disk_info *info)
732 _setdiskinfo(disk, info);
733 disk_msg_send_sync(DISK_DISK_PROBE, disk, NULL);
735 "disk_setdiskinfo_sync: sent probe for %s\n",
736 disk->d_cdev->si_name);
740 * This routine is called when an adapter detaches. The higher level
741 * managed disk device is destroyed while the lower level raw device is
745 disk_destroy(struct disk *disk)
747 dsched_disk_destroy_callback(disk);
748 disk_msg_send_sync(DISK_DISK_DESTROY, disk, NULL);
753 disk_dumpcheck(cdev_t dev, u_int64_t *size, u_int64_t *blkno, u_int32_t *secsize)
755 struct partinfo pinfo;
758 bzero(&pinfo, sizeof(pinfo));
759 error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0,
760 proc0.p_ucred, NULL);
764 if (pinfo.media_blksize == 0)
767 if (blkno) /* XXX: make sure this reserved stuff is right */
768 *blkno = pinfo.reserved_blocks +
769 pinfo.media_offset / pinfo.media_blksize;
771 *secsize = pinfo.media_blksize;
773 *size = (pinfo.media_blocks - pinfo.reserved_blocks);
779 disk_dumpconf(cdev_t dev, u_int onoff)
781 struct dumperinfo di;
782 u_int64_t size, blkno;
787 return set_dumper(NULL);
789 error = disk_dumpcheck(dev, &size, &blkno, &secsize);
794 bzero(&di, sizeof(struct dumperinfo));
795 di.dumper = diskdump;
797 di.blocksize = secsize;
798 di.mediaoffset = blkno * DEV_BSIZE;
799 di.mediasize = size * DEV_BSIZE;
801 return set_dumper(&di);
805 disk_unprobe(struct disk *disk)
810 disk_msg_send_sync(DISK_UNPROBE, disk, NULL);
814 disk_invalidate (struct disk *disk)
816 dsgone(&disk->d_slice);
820 disk_enumerate(struct disk *disk)
824 lwkt_gettoken(&disklist_token);
826 dp = (LIST_FIRST(&disklist));
828 dp = (LIST_NEXT(disk, d_list));
829 lwkt_reltoken(&disklist_token);
836 sysctl_disks(SYSCTL_HANDLER_ARGS)
844 while ((disk = disk_enumerate(disk))) {
846 error = SYSCTL_OUT(req, " ", 1);
852 error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
853 strlen(disk->d_rawdev->si_name));
857 error = SYSCTL_OUT(req, "", 1);
861 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, NULL, 0,
862 sysctl_disks, "A", "names of available disks");
865 * Open a disk device or partition.
869 diskopen(struct dev_open_args *ap)
871 cdev_t dev = ap->a_head.a_dev;
876 * dp can't be NULL here XXX.
878 * d_slice will be NULL if setdiskinfo() has not been called yet.
879 * setdiskinfo() is typically called whether the disk is present
880 * or not (e.g. CD), but the base disk device is created first
881 * and there may be a race.
884 if (dp == NULL || dp->d_slice == NULL)
889 * Deal with open races
892 while (dp->d_flags & DISKFLAG_LOCK) {
893 dp->d_flags |= DISKFLAG_WANTED;
894 error = tsleep(dp, PCATCH, "diskopen", hz);
900 dp->d_flags |= DISKFLAG_LOCK;
903 * Open the underlying raw device.
905 if (!dsisopen(dp->d_slice)) {
907 if (!pdev->si_iosize_max)
908 pdev->si_iosize_max = dev->si_iosize_max;
910 error = dev_dopen(dp->d_rawdev, ap->a_oflags,
911 ap->a_devtype, ap->a_cred);
915 * Inherit properties from the underlying device now that it is
923 error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags,
924 &dp->d_slice, &dp->d_info);
925 if (!dsisopen(dp->d_slice)) {
926 dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype);
929 dp->d_flags &= ~DISKFLAG_LOCK;
930 if (dp->d_flags & DISKFLAG_WANTED) {
931 dp->d_flags &= ~DISKFLAG_WANTED;
940 * Close a disk device or partition
944 diskclose(struct dev_close_args *ap)
946 cdev_t dev = ap->a_head.a_dev;
954 dsclose(dev, ap->a_devtype, dp->d_slice);
955 if (!dsisopen(dp->d_slice)) {
956 error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype);
963 * First execute the ioctl on the disk device, and if it isn't supported
964 * try running it on the backing device.
968 diskioctl(struct dev_ioctl_args *ap)
970 cdev_t dev = ap->a_head.a_dev;
979 devfs_debug(DEVFS_DEBUG_DEBUG,
980 "diskioctl: cmd is: %lx (name: %s)\n",
981 ap->a_cmd, dev->si_name);
982 devfs_debug(DEVFS_DEBUG_DEBUG,
983 "diskioctl: &dp->d_slice is: %p, %p\n",
984 &dp->d_slice, dp->d_slice);
986 if (ap->a_cmd == DIOCGKERNELDUMP) {
987 u = *(u_int *)ap->a_data;
988 return disk_dumpconf(dev, u);
991 if (&dp->d_slice == NULL || dp->d_slice == NULL) {
995 error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag,
996 &dp->d_slice, &dp->d_info);
1000 if (error == ENOIOCTL) {
1001 error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data,
1002 ap->a_fflag, ap->a_cred, NULL);
1008 * Execute strategy routine
1012 diskstrategy(struct dev_strategy_args *ap)
1014 cdev_t dev = ap->a_head.a_dev;
1015 struct bio *bio = ap->a_bio;
1022 bio->bio_buf->b_error = ENXIO;
1023 bio->bio_buf->b_flags |= B_ERROR;
1027 KKASSERT(dev->si_disk == dp);
1030 * The dscheck() function will also transform the slice relative
1031 * block number i.e. bio->bio_offset into a block number that can be
1032 * passed directly to the underlying raw device. If dscheck()
1033 * returns NULL it will have handled the bio for us (e.g. EOF
1034 * or error due to being beyond the device size).
1036 if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL) {
1037 dsched_queue(dp, nbio);
1045 * Return the partition size in ?blocks?
1049 diskpsize(struct dev_psize_args *ap)
1051 cdev_t dev = ap->a_head.a_dev;
1057 ap->a_result = dssize(dev, &dp->d_slice);
1062 * When new device entries are instantiated, make sure they inherit our
1063 * si_disk structure and block and iosize limits from the raw device.
1065 * This routine is always called synchronously in the context of the
1068 * XXX The various io and block size constraints are not always initialized
1069 * properly by devices.
1073 diskclone(struct dev_clone_args *ap)
1075 cdev_t dev = ap->a_head.a_dev;
1079 KKASSERT(dp != NULL);
1081 dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
1082 dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
1083 dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
1088 diskdump(struct dev_dump_args *ap)
1090 cdev_t dev = ap->a_head.a_dev;
1091 struct disk *dp = dev->si_disk;
1092 u_int64_t size, offset;
1095 error = disk_dumpcheck(dev, &size, &ap->a_blkno, &ap->a_secsize);
1096 /* XXX: this should probably go in disk_dumpcheck somehow */
1097 if (ap->a_length != 0) {
1099 offset = ap->a_blkno * DEV_BSIZE;
1100 if ((ap->a_offset < offset) ||
1101 (ap->a_offset + ap->a_length - offset > size)) {
1102 kprintf("Attempt to write outside dump device boundaries.\n");
1108 ap->a_head.a_dev = dp->d_rawdev;
1109 error = dev_doperate(&ap->a_head);
1116 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
1117 0, sizeof(struct diskslices), "sizeof(struct diskslices)");
1119 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
1120 0, sizeof(struct disk), "sizeof(struct disk)");
1123 * Reorder interval for burst write allowance and minor write
1126 * We always want to trickle some writes in to make use of the
1127 * disk's zone cache. Bursting occurs on a longer interval and only
1128 * runningbufspace is well over the hirunningspace limit.
1130 int bioq_reorder_burst_interval = 60; /* should be multiple of minor */
1131 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_interval,
1132 CTLFLAG_RW, &bioq_reorder_burst_interval, 0, "");
1133 int bioq_reorder_minor_interval = 5;
1134 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_interval,
1135 CTLFLAG_RW, &bioq_reorder_minor_interval, 0, "");
1137 int bioq_reorder_burst_bytes = 3000000;
1138 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_bytes,
1139 CTLFLAG_RW, &bioq_reorder_burst_bytes, 0, "");
1140 int bioq_reorder_minor_bytes = 262144;
1141 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_bytes,
1142 CTLFLAG_RW, &bioq_reorder_minor_bytes, 0, "");
1146 * Order I/Os. Generally speaking this code is designed to make better
1147 * use of drive zone caches. A drive zone cache can typically track linear
1148 * reads or writes for around 16 zones simultaniously.
1150 * Read prioritization issues: It is possible for hundreds of megabytes worth
1151 * of writes to be queued asynchronously. This creates a huge bottleneck
1152 * for reads which reduce read bandwidth to a trickle.
1154 * To solve this problem we generally reorder reads before writes.
1156 * However, a large number of random reads can also starve writes and
1157 * make poor use of the drive zone cache so we allow writes to trickle
1161 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
1164 * The BIO wants to be ordered. Adding to the tail also
1165 * causes transition to be set to NULL, forcing the ordering
1166 * of all prior I/O's.
1168 if (bio->bio_buf->b_flags & B_ORDERED) {
1169 bioq_insert_tail(bioq, bio);
1173 switch(bio->bio_buf->b_cmd) {
1175 if (bioq->transition) {
1177 * Insert before the first write. Bleedover writes
1178 * based on reorder intervals to prevent starvation.
1180 TAILQ_INSERT_BEFORE(bioq->transition, bio, bio_act);
1182 if (bioq->reorder % bioq_reorder_minor_interval == 0) {
1183 bioqwritereorder(bioq);
1184 if (bioq->reorder >=
1185 bioq_reorder_burst_interval) {
1191 * No writes queued (or ordering was forced),
1194 TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act);
1199 * Writes are always appended. If no writes were previously
1200 * queued or an ordered tail insertion occured the transition
1201 * field will be NULL.
1203 TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act);
1204 if (bioq->transition == NULL)
1205 bioq->transition = bio;
1209 * All other request types are forced to be ordered.
1211 bioq_insert_tail(bioq, bio);
1217 * Move the read-write transition point to prevent reads from
1218 * completely starving our writes. This brings a number of writes into
1219 * the fold every N reads.
1221 * We bring a few linear writes into the fold on a minor interval
1222 * and we bring a non-linear burst of writes into the fold on a major
1223 * interval. Bursting only occurs if runningbufspace is really high
1224 * (typically from syncs, fsyncs, or HAMMER flushes).
1228 bioqwritereorder(struct bio_queue_head *bioq)
1236 if (bioq->reorder < bioq_reorder_burst_interval ||
1237 !buf_runningbufspace_severe()) {
1238 left = (size_t)bioq_reorder_minor_bytes;
1241 left = (size_t)bioq_reorder_burst_bytes;
1245 next_offset = bioq->transition->bio_offset;
1246 while ((bio = bioq->transition) != NULL &&
1247 (check_off == 0 || next_offset == bio->bio_offset)
1249 n = bio->bio_buf->b_bcount;
1250 next_offset = bio->bio_offset + n;
1251 bioq->transition = TAILQ_NEXT(bio, bio_act);
1259 * Bounds checking against the media size, used for the raw partition.
1260 * secsize, mediasize and b_blkno must all be the same units.
1261 * Possibly this has to be DEV_BSIZE (512).
1264 bounds_check_with_mediasize(struct bio *bio, int secsize, uint64_t mediasize)
1266 struct buf *bp = bio->bio_buf;
1269 sz = howmany(bp->b_bcount, secsize);
1271 if (bio->bio_offset/DEV_BSIZE + sz > mediasize) {
1272 sz = mediasize - bio->bio_offset/DEV_BSIZE;
1274 /* If exactly at end of disk, return EOF. */
1275 bp->b_resid = bp->b_bcount;
1279 /* If past end of disk, return EINVAL. */
1280 bp->b_error = EINVAL;
1283 /* Otherwise, truncate request. */
1284 bp->b_bcount = sz * secsize;
1291 * Disk error is the preface to plaintive error messages
1292 * about failing disk transfers. It prints messages of the form
1294 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
1296 * if the offset of the error in the transfer and a disk label
1297 * are both available. blkdone should be -1 if the position of the error
1298 * is unknown; the disklabel pointer may be null from drivers that have not
1299 * been converted to use them. The message is printed with kprintf
1300 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
1301 * The message should be completed (with at least a newline) with kprintf
1302 * or log(-1, ...), respectively. There is no trailing space.
1305 diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt)
1307 struct buf *bp = bio->bio_buf;
1321 kprintf("%s: %s %sing ", dev->si_name, what, term);
1322 kprintf("offset %012llx for %d",
1323 (long long)bio->bio_offset,
1327 kprintf(" (%d bytes completed)", donecnt);
1331 * Locate a disk device
1334 disk_locate(const char *devname)
1336 return devfs_find_device_by_name(devname);
1340 disk_config(void *arg)
1342 disk_msg_send_sync(DISK_SYNC, NULL, NULL);
1348 struct thread* td_core;
1350 disk_msg_cache = objcache_create("disk-msg-cache", 0, 0,
1352 objcache_malloc_alloc,
1353 objcache_malloc_free,
1354 &disk_msg_malloc_args);
1356 lwkt_token_init(&disklist_token, 1, "disks");
1359 * Initialize the reply-only port which acts as a message drain
1361 lwkt_initport_replyonly(&disk_dispose_port, disk_msg_autofree_reply);
1363 lwkt_gettoken(&disklist_token);
1364 lwkt_create(disk_msg_core, /*args*/NULL, &td_core, NULL,
1365 0, 0, "disk_msg_core");
1366 tsleep(td_core, 0, "diskcore", 0);
1367 lwkt_reltoken(&disklist_token);
1373 objcache_destroy(disk_msg_cache);
1377 * Clean out illegal characters in serial numbers.
1380 disk_cleanserial(char *serno)
1384 while ((c = *serno) != 0) {
1385 if (c >= 'a' && c <= 'z')
1387 else if (c >= 'A' && c <= 'Z')
1389 else if (c >= '0' && c <= '9')
1391 else if (c == '-' || c == '@' || c == '+' || c == '.')
1399 TUNABLE_INT("kern.disk_debug", &disk_debug_enable);
1400 SYSCTL_INT(_kern, OID_AUTO, disk_debug, CTLFLAG_RW, &disk_debug_enable,
1401 0, "Enable subr_disk debugging");
1403 SYSINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_FIRST, disk_init, NULL);
1404 SYSUNINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_ANY, disk_uninit, NULL);