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>
6 * and Alex Hornung <ahornung@gmail.com>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
18 * 3. Neither the name of The DragonFly Project nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific, prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
25 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
26 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
27 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
28 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
29 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
30 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
31 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
32 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * ----------------------------------------------------------------------------
36 * "THE BEER-WARE LICENSE" (Revision 42):
37 * <phk@FreeBSD.ORG> wrote this file. As long as you retain this notice you
38 * can do whatever you want with this stuff. If we meet some day, and you think
39 * this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp
40 * ----------------------------------------------------------------------------
42 * Copyright (c) 1982, 1986, 1988, 1993
43 * The Regents of the University of California. All rights reserved.
44 * (c) UNIX System Laboratories, Inc.
45 * All or some portions of this file are derived from material licensed
46 * to the University of California by American Telephone and Telegraph
47 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
48 * the permission of UNIX System Laboratories, Inc.
50 * Redistribution and use in source and binary forms, with or without
51 * modification, are permitted provided that the following conditions
53 * 1. Redistributions of source code must retain the above copyright
54 * notice, this list of conditions and the following disclaimer.
55 * 2. Redistributions in binary form must reproduce the above copyright
56 * notice, this list of conditions and the following disclaimer in the
57 * documentation and/or other materials provided with the distribution.
58 * 3. All advertising materials mentioning features or use of this software
59 * must display the following acknowledgement:
60 * This product includes software developed by the University of
61 * California, Berkeley and its contributors.
62 * 4. Neither the name of the University nor the names of its contributors
63 * may be used to endorse or promote products derived from this software
64 * without specific prior written permission.
66 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
67 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
68 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
69 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
70 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
71 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
72 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
73 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
74 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
75 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
78 * @(#)ufs_disksubr.c 8.5 (Berkeley) 1/21/94
79 * $FreeBSD: src/sys/kern/subr_disk.c,v 1.20.2.6 2001/10/05 07:14:57 peter Exp $
80 * $FreeBSD: src/sys/ufs/ufs/ufs_disksubr.c,v 1.44.2.3 2001/03/05 05:42:19 obrien Exp $
81 * $DragonFly: src/sys/kern/subr_disk.c,v 1.40 2008/06/05 18:06:32 swildner Exp $
84 #include <sys/param.h>
85 #include <sys/systm.h>
86 #include <sys/kernel.h>
88 #include <sys/sysctl.h>
91 #include <sys/disklabel.h>
92 #include <sys/disklabel32.h>
93 #include <sys/disklabel64.h>
94 #include <sys/diskslice.h>
95 #include <sys/diskmbr.h>
97 #include <sys/kerneldump.h>
98 #include <sys/malloc.h>
99 #include <sys/sysctl.h>
100 #include <machine/md_var.h>
101 #include <sys/ctype.h>
102 #include <sys/syslog.h>
103 #include <sys/device.h>
104 #include <sys/msgport.h>
105 #include <sys/devfs.h>
106 #include <sys/thread.h>
107 #include <sys/dsched.h>
108 #include <sys/queue.h>
109 #include <sys/lock.h>
110 #include <sys/udev.h>
111 #include <sys/uuid.h>
113 #include <sys/buf2.h>
114 #include <sys/mplock2.h>
115 #include <sys/msgport2.h>
116 #include <sys/thread2.h>
118 static MALLOC_DEFINE(M_DISK, "disk", "disk data");
119 static int disk_debug_enable = 0;
121 static void disk_msg_autofree_reply(lwkt_port_t, lwkt_msg_t);
122 static void disk_msg_core(void *);
123 static int disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe);
124 static void disk_probe(struct disk *dp, int reprobe);
125 static void _setdiskinfo(struct disk *disk, struct disk_info *info);
126 static void bioqwritereorder(struct bio_queue_head *bioq);
127 static void disk_cleanserial(char *serno);
128 static int disk_debug(int, char *, ...) __printflike(2, 3);
130 static d_open_t diskopen;
131 static d_close_t diskclose;
132 static d_ioctl_t diskioctl;
133 static d_strategy_t diskstrategy;
134 static d_psize_t diskpsize;
135 static d_clone_t diskclone;
136 static d_dump_t diskdump;
138 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
139 static struct lwkt_token disklist_token;
141 static struct dev_ops disk_ops = {
142 { "disk", 0, D_DISK | D_MPSAFE },
144 .d_close = diskclose,
146 .d_write = physwrite,
147 .d_ioctl = diskioctl,
148 .d_strategy = diskstrategy,
150 .d_psize = diskpsize,
154 static struct objcache *disk_msg_cache;
156 struct objcache_malloc_args disk_msg_malloc_args = {
157 sizeof(struct disk_msg), M_DISK };
159 static struct lwkt_port disk_dispose_port;
160 static struct lwkt_port disk_msg_port;
163 disk_debug(int level, char *fmt, ...)
168 if (level <= disk_debug_enable)
176 disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe)
178 struct disk_info *info = &dp->d_info;
179 struct diskslice *sp = &dp->d_slice->dss_slices[slice];
181 struct partinfo part;
189 "disk_probe_slice (begin): %s (%s)\n",
190 dev->si_name, dp->d_cdev->si_name);
192 sno = slice ? slice - 1 : 0;
194 ops = &disklabel32_ops;
195 msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
196 if (msg && !strcmp(msg, "no disk label")) {
197 ops = &disklabel64_ops;
198 msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
201 if (slice != WHOLE_DISK_SLICE)
202 ops->op_adjust_label_reserved(dp->d_slice, slice, sp);
207 for (i = 0; i < ops->op_getnumparts(sp->ds_label); i++) {
208 ops->op_loadpartinfo(sp->ds_label, i, &part);
211 (ndev = devfs_find_device_by_name("%s%c",
212 dev->si_name, 'a' + i))
215 * Device already exists and
218 ndev->si_flags |= SI_REPROBE_TEST;
221 * Destroy old UUID alias
223 destroy_dev_alias(ndev, "part-by-uuid/*");
225 /* Create UUID alias */
226 if (!kuuid_is_nil(&part.storage_uuid)) {
227 snprintf_uuid(uuid_buf,
235 ndev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
236 dkmakeminor(dkunit(dp->d_cdev),
238 UID_ROOT, GID_OPERATOR, 0640,
239 "%s%c", dev->si_name, 'a'+ i);
241 udev_dict_set_cstr(ndev, "subsystem", "disk");
242 /* Inherit parent's disk type */
243 if (dp->d_disktype) {
244 udev_dict_set_cstr(ndev, "disk-type",
245 __DECONST(char *, dp->d_disktype));
248 /* Create serno alias */
249 if (dp->d_info.d_serialno) {
252 dp->d_info.d_serialno,
256 /* Create UUID alias */
257 if (!kuuid_is_nil(&part.storage_uuid)) {
258 snprintf_uuid(uuid_buf,
265 ndev->si_flags |= SI_REPROBE_TEST;
269 } else if (info->d_dsflags & DSO_COMPATLABEL) {
271 if (sp->ds_size >= 0x100000000ULL)
272 ops = &disklabel64_ops;
274 ops = &disklabel32_ops;
275 sp->ds_label = ops->op_clone_label(info, sp);
277 if (sp->ds_type == DOSPTYP_386BSD || /* XXX */
278 sp->ds_type == DOSPTYP_NETBSD ||
279 sp->ds_type == DOSPTYP_OPENBSD) {
280 log(LOG_WARNING, "%s: cannot find label (%s)\n",
286 sp->ds_wlabel = FALSE;
289 return (msg ? EINVAL : 0);
293 * This routine is only called for newly minted drives or to reprobe
294 * a drive with no open slices. disk_probe_slice() is called directly
295 * when reprobing partition changes within slices.
298 disk_probe(struct disk *dp, int reprobe)
300 struct disk_info *info = &dp->d_info;
301 cdev_t dev = dp->d_cdev;
304 struct diskslices *osp;
305 struct diskslice *sp;
308 KKASSERT (info->d_media_blksize != 0);
311 dp->d_slice = dsmakeslicestruct(BASE_SLICE, info);
312 disk_debug(1, "disk_probe (begin): %s\n", dp->d_cdev->si_name);
314 error = mbrinit(dev, info, &(dp->d_slice));
320 for (i = 0; i < dp->d_slice->dss_nslices; i++) {
322 * Ignore the whole-disk slice, it has already been created.
324 if (i == WHOLE_DISK_SLICE)
329 * Ignore the compatibility slice s0 if it's a device mapper
332 if ((i == COMPATIBILITY_SLICE) &&
333 (info->d_dsflags & DSO_DEVICEMAPPER))
337 sp = &dp->d_slice->dss_slices[i];
340 * Handle s0. s0 is a compatibility slice if there are no
341 * other slices and it has not otherwise been set up, else
344 if (i == COMPATIBILITY_SLICE) {
346 if (sp->ds_type == 0 &&
347 dp->d_slice->dss_nslices == BASE_SLICE) {
348 sp->ds_size = info->d_media_blocks;
357 * Ignore 0-length slices
359 if (sp->ds_size == 0)
363 (ndev = devfs_find_device_by_name("%ss%d",
364 dev->si_name, sno))) {
366 * Device already exists and is still valid
368 ndev->si_flags |= SI_REPROBE_TEST;
371 * Destroy old UUID alias
373 destroy_dev_alias(ndev, "slice-by-uuid/*");
375 /* Create UUID alias */
376 if (!kuuid_is_nil(&sp->ds_stor_uuid)) {
377 snprintf_uuid(uuid_buf, sizeof(uuid_buf),
379 make_dev_alias(ndev, "slice-by-uuid/%s",
384 * Else create new device
386 ndev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
387 dkmakewholeslice(dkunit(dev), i),
388 UID_ROOT, GID_OPERATOR, 0640,
389 (info->d_dsflags & DSO_DEVICEMAPPER)?
390 "%s.s%d" : "%ss%d", dev->si_name, sno);
391 udev_dict_set_cstr(ndev, "subsystem", "disk");
392 /* Inherit parent's disk type */
393 if (dp->d_disktype) {
394 udev_dict_set_cstr(ndev, "disk-type",
395 __DECONST(char *, dp->d_disktype));
398 /* Create serno alias */
399 if (dp->d_info.d_serialno) {
400 make_dev_alias(ndev, "serno/%s.s%d",
401 dp->d_info.d_serialno, sno);
404 /* Create UUID alias */
405 if (!kuuid_is_nil(&sp->ds_stor_uuid)) {
406 snprintf_uuid(uuid_buf, sizeof(uuid_buf),
408 make_dev_alias(ndev, "slice-by-uuid/%s",
413 ndev->si_flags |= SI_REPROBE_TEST;
418 * Probe appropriate slices for a disklabel
420 * XXX slice type 1 used by our gpt probe code.
421 * XXX slice type 0 used by mbr compat slice.
423 if (sp->ds_type == DOSPTYP_386BSD ||
424 sp->ds_type == DOSPTYP_NETBSD ||
425 sp->ds_type == DOSPTYP_OPENBSD ||
428 if (dp->d_slice->dss_first_bsd_slice == 0)
429 dp->d_slice->dss_first_bsd_slice = i;
430 disk_probe_slice(dp, ndev, i, reprobe);
434 disk_debug(1, "disk_probe (end): %s\n", dp->d_cdev->si_name);
439 disk_msg_core(void *arg)
442 struct diskslice *sp;
446 lwkt_gettoken(&disklist_token);
447 lwkt_initport_thread(&disk_msg_port, curthread);
448 wakeup(curthread); /* synchronous startup */
449 lwkt_reltoken(&disklist_token);
451 get_mplock(); /* not mpsafe yet? */
455 msg = (disk_msg_t)lwkt_waitport(&disk_msg_port, 0);
457 switch (msg->hdr.u.ms_result) {
458 case DISK_DISK_PROBE:
459 dp = (struct disk *)msg->load;
461 "DISK_DISK_PROBE: %s\n",
462 dp->d_cdev->si_name);
465 case DISK_DISK_DESTROY:
466 dp = (struct disk *)msg->load;
468 "DISK_DISK_DESTROY: %s\n",
469 dp->d_cdev->si_name);
470 devfs_destroy_subnames(dp->d_cdev->si_name);
471 devfs_destroy_dev(dp->d_cdev);
472 lwkt_gettoken(&disklist_token);
473 LIST_REMOVE(dp, d_list);
474 lwkt_reltoken(&disklist_token);
475 if (dp->d_info.d_serialno) {
476 kfree(dp->d_info.d_serialno, M_TEMP);
477 dp->d_info.d_serialno = NULL;
481 dp = (struct disk *)msg->load;
483 "DISK_DISK_UNPROBE: %s\n",
484 dp->d_cdev->si_name);
485 devfs_destroy_subnames(dp->d_cdev->si_name);
487 case DISK_SLICE_REPROBE:
488 dp = (struct disk *)msg->load;
489 sp = (struct diskslice *)msg->load2;
490 devfs_clr_subnames_flag(sp->ds_dev->si_name,
493 "DISK_SLICE_REPROBE: %s\n",
494 sp->ds_dev->si_name);
495 disk_probe_slice(dp, sp->ds_dev,
496 dkslice(sp->ds_dev), 1);
497 devfs_destroy_subnames_without_flag(
498 sp->ds_dev->si_name, SI_REPROBE_TEST);
500 case DISK_DISK_REPROBE:
501 dp = (struct disk *)msg->load;
502 devfs_clr_subnames_flag(dp->d_cdev->si_name, SI_REPROBE_TEST);
504 "DISK_DISK_REPROBE: %s\n",
505 dp->d_cdev->si_name);
507 devfs_destroy_subnames_without_flag(
508 dp->d_cdev->si_name, SI_REPROBE_TEST);
511 disk_debug(1, "DISK_SYNC\n");
514 devfs_debug(DEVFS_DEBUG_WARNING,
515 "disk_msg_core: unknown message "
516 "received at core\n");
519 lwkt_replymsg(&msg->hdr, 0);
526 * Acts as a message drain. Any message that is replied to here gets
527 * destroyed and the memory freed.
530 disk_msg_autofree_reply(lwkt_port_t port, lwkt_msg_t msg)
532 objcache_put(disk_msg_cache, msg);
537 disk_msg_send(uint32_t cmd, void *load, void *load2)
540 lwkt_port_t port = &disk_msg_port;
542 disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
544 lwkt_initmsg(&disk_msg->hdr, &disk_dispose_port, 0);
546 disk_msg->hdr.u.ms_result = cmd;
547 disk_msg->load = load;
548 disk_msg->load2 = load2;
550 lwkt_sendmsg(port, &disk_msg->hdr);
554 disk_msg_send_sync(uint32_t cmd, void *load, void *load2)
556 struct lwkt_port rep_port;
560 disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
561 port = &disk_msg_port;
563 /* XXX could probably use curthread's built-in msgport */
564 lwkt_initport_thread(&rep_port, curthread);
565 lwkt_initmsg(&disk_msg->hdr, &rep_port, 0);
567 disk_msg->hdr.u.ms_result = cmd;
568 disk_msg->load = load;
569 disk_msg->load2 = load2;
571 lwkt_sendmsg(port, &disk_msg->hdr);
572 lwkt_waitmsg(&disk_msg->hdr, 0);
573 objcache_put(disk_msg_cache, disk_msg);
577 * Create a raw device for the dev_ops template (which is returned). Also
578 * create a slice and unit managed disk and overload the user visible
579 * device space with it.
581 * NOTE: The returned raw device is NOT a slice and unit managed device.
582 * It is an actual raw device representing the raw disk as specified by
583 * the passed dev_ops. The disk layer not only returns such a raw device,
584 * it also uses it internally when passing (modified) commands through.
587 disk_create(int unit, struct disk *dp, struct dev_ops *raw_ops)
589 return disk_create_named(NULL, unit, dp, raw_ops);
593 disk_create_named(const char *name, int unit, struct disk *dp, struct dev_ops *raw_ops)
597 disk_debug(1, "disk_create (begin): %s%d\n", name, unit);
600 rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
601 UID_ROOT, GID_OPERATOR, 0640, "%s", name);
603 rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
604 UID_ROOT, GID_OPERATOR, 0640,
605 "%s%d", raw_ops->head.name, unit);
608 bzero(dp, sizeof(*dp));
610 dp->d_rawdev = rawdev;
611 dp->d_raw_ops = raw_ops;
612 dp->d_dev_ops = &disk_ops;
615 dp->d_cdev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
616 dkmakewholedisk(unit), UID_ROOT, GID_OPERATOR, 0640,
619 dp->d_cdev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
620 dkmakewholedisk(unit),
621 UID_ROOT, GID_OPERATOR, 0640,
622 "%s%d", raw_ops->head.name, unit);
625 udev_dict_set_cstr(dp->d_cdev, "subsystem", "disk");
626 dp->d_cdev->si_disk = dp;
629 dsched_disk_create_callback(dp, name, unit);
631 dsched_disk_create_callback(dp, raw_ops->head.name, unit);
633 lwkt_gettoken(&disklist_token);
634 LIST_INSERT_HEAD(&disklist, dp, d_list);
635 lwkt_reltoken(&disklist_token);
637 disk_debug(1, "disk_create (end): %s%d\n",
638 (name != NULL)?(name):(raw_ops->head.name), unit);
640 return (dp->d_rawdev);
644 disk_setdisktype(struct disk *disk, const char *type)
646 KKASSERT(disk != NULL);
648 disk->d_disktype = type;
649 return udev_dict_set_cstr(disk->d_cdev, "disk-type", __DECONST(char *, type));
653 _setdiskinfo(struct disk *disk, struct disk_info *info)
657 oldserialno = disk->d_info.d_serialno;
658 bcopy(info, &disk->d_info, sizeof(disk->d_info));
659 info = &disk->d_info;
662 "_setdiskinfo: %s\n",
663 disk->d_cdev->si_name);
666 * The serial number is duplicated so the caller can throw
669 if (info->d_serialno && info->d_serialno[0]) {
670 info->d_serialno = kstrdup(info->d_serialno, M_TEMP);
671 disk_cleanserial(info->d_serialno);
673 make_dev_alias(disk->d_cdev, "serno/%s",
677 info->d_serialno = NULL;
680 kfree(oldserialno, M_TEMP);
682 dsched_disk_update_callback(disk, info);
685 * The caller may set d_media_size or d_media_blocks and we
686 * calculate the other.
688 KKASSERT(info->d_media_size == 0 || info->d_media_blocks == 0);
689 if (info->d_media_size == 0 && info->d_media_blocks) {
690 info->d_media_size = (u_int64_t)info->d_media_blocks *
691 info->d_media_blksize;
692 } else if (info->d_media_size && info->d_media_blocks == 0 &&
693 info->d_media_blksize) {
694 info->d_media_blocks = info->d_media_size /
695 info->d_media_blksize;
699 * The si_* fields for rawdev are not set until after the
700 * disk_create() call, so someone using the cooked version
701 * of the raw device (i.e. da0s0) will not get the right
702 * si_iosize_max unless we fix it up here.
704 if (disk->d_cdev && disk->d_rawdev &&
705 disk->d_cdev->si_iosize_max == 0) {
706 disk->d_cdev->si_iosize_max = disk->d_rawdev->si_iosize_max;
707 disk->d_cdev->si_bsize_phys = disk->d_rawdev->si_bsize_phys;
708 disk->d_cdev->si_bsize_best = disk->d_rawdev->si_bsize_best;
711 /* Add the serial number to the udev_dictionary */
712 if (info->d_serialno)
713 udev_dict_set_cstr(disk->d_cdev, "serno", info->d_serialno);
717 * Disk drivers must call this routine when media parameters are available
721 disk_setdiskinfo(struct disk *disk, struct disk_info *info)
723 _setdiskinfo(disk, info);
724 disk_msg_send(DISK_DISK_PROBE, disk, NULL);
726 "disk_setdiskinfo: sent probe for %s\n",
727 disk->d_cdev->si_name);
731 disk_setdiskinfo_sync(struct disk *disk, struct disk_info *info)
733 _setdiskinfo(disk, info);
734 disk_msg_send_sync(DISK_DISK_PROBE, disk, NULL);
736 "disk_setdiskinfo_sync: sent probe for %s\n",
737 disk->d_cdev->si_name);
741 * This routine is called when an adapter detaches. The higher level
742 * managed disk device is destroyed while the lower level raw device is
746 disk_destroy(struct disk *disk)
748 dsched_disk_destroy_callback(disk);
749 disk_msg_send_sync(DISK_DISK_DESTROY, disk, NULL);
754 disk_dumpcheck(cdev_t dev, u_int64_t *size, u_int64_t *blkno, u_int32_t *secsize)
756 struct partinfo pinfo;
759 bzero(&pinfo, sizeof(pinfo));
760 error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0,
761 proc0.p_ucred, NULL);
765 if (pinfo.media_blksize == 0)
768 if (blkno) /* XXX: make sure this reserved stuff is right */
769 *blkno = pinfo.reserved_blocks +
770 pinfo.media_offset / pinfo.media_blksize;
772 *secsize = pinfo.media_blksize;
774 *size = (pinfo.media_blocks - pinfo.reserved_blocks);
780 disk_dumpconf(cdev_t dev, u_int onoff)
782 struct dumperinfo di;
783 u_int64_t size, blkno;
788 return set_dumper(NULL);
790 error = disk_dumpcheck(dev, &size, &blkno, &secsize);
795 bzero(&di, sizeof(struct dumperinfo));
796 di.dumper = diskdump;
798 di.blocksize = secsize;
799 di.mediaoffset = blkno * DEV_BSIZE;
800 di.mediasize = size * DEV_BSIZE;
802 return set_dumper(&di);
806 disk_unprobe(struct disk *disk)
811 disk_msg_send_sync(DISK_UNPROBE, disk, NULL);
815 disk_invalidate (struct disk *disk)
817 dsgone(&disk->d_slice);
821 disk_enumerate(struct disk *disk)
825 lwkt_gettoken(&disklist_token);
827 dp = (LIST_FIRST(&disklist));
829 dp = (LIST_NEXT(disk, d_list));
830 lwkt_reltoken(&disklist_token);
837 sysctl_disks(SYSCTL_HANDLER_ARGS)
845 while ((disk = disk_enumerate(disk))) {
847 error = SYSCTL_OUT(req, " ", 1);
853 error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
854 strlen(disk->d_rawdev->si_name));
858 error = SYSCTL_OUT(req, "", 1);
862 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, NULL, 0,
863 sysctl_disks, "A", "names of available disks");
866 * Open a disk device or partition.
870 diskopen(struct dev_open_args *ap)
872 cdev_t dev = ap->a_head.a_dev;
877 * dp can't be NULL here XXX.
879 * d_slice will be NULL if setdiskinfo() has not been called yet.
880 * setdiskinfo() is typically called whether the disk is present
881 * or not (e.g. CD), but the base disk device is created first
882 * and there may be a race.
885 if (dp == NULL || dp->d_slice == NULL)
890 * Deal with open races
893 while (dp->d_flags & DISKFLAG_LOCK) {
894 dp->d_flags |= DISKFLAG_WANTED;
895 error = tsleep(dp, PCATCH, "diskopen", hz);
901 dp->d_flags |= DISKFLAG_LOCK;
904 * Open the underlying raw device.
906 if (!dsisopen(dp->d_slice)) {
908 if (!pdev->si_iosize_max)
909 pdev->si_iosize_max = dev->si_iosize_max;
911 error = dev_dopen(dp->d_rawdev, ap->a_oflags,
912 ap->a_devtype, ap->a_cred);
916 * Inherit properties from the underlying device now that it is
924 error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags,
925 &dp->d_slice, &dp->d_info);
926 if (!dsisopen(dp->d_slice)) {
927 dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype);
930 dp->d_flags &= ~DISKFLAG_LOCK;
931 if (dp->d_flags & DISKFLAG_WANTED) {
932 dp->d_flags &= ~DISKFLAG_WANTED;
941 * Close a disk device or partition
945 diskclose(struct dev_close_args *ap)
947 cdev_t dev = ap->a_head.a_dev;
955 dsclose(dev, ap->a_devtype, dp->d_slice);
956 if (!dsisopen(dp->d_slice)) {
957 error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype);
964 * First execute the ioctl on the disk device, and if it isn't supported
965 * try running it on the backing device.
969 diskioctl(struct dev_ioctl_args *ap)
971 cdev_t dev = ap->a_head.a_dev;
980 devfs_debug(DEVFS_DEBUG_DEBUG,
981 "diskioctl: cmd is: %lx (name: %s)\n",
982 ap->a_cmd, dev->si_name);
983 devfs_debug(DEVFS_DEBUG_DEBUG,
984 "diskioctl: &dp->d_slice is: %p, %p\n",
985 &dp->d_slice, dp->d_slice);
987 if (ap->a_cmd == DIOCGKERNELDUMP) {
988 u = *(u_int *)ap->a_data;
989 return disk_dumpconf(dev, u);
992 if (&dp->d_slice == NULL || dp->d_slice == NULL ||
993 ((dp->d_info.d_dsflags & DSO_DEVICEMAPPER) &&
994 dkslice(dev) == WHOLE_DISK_SLICE)) {
998 error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag,
999 &dp->d_slice, &dp->d_info);
1003 if (error == ENOIOCTL) {
1004 error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data,
1005 ap->a_fflag, ap->a_cred, NULL);
1011 * Execute strategy routine
1015 diskstrategy(struct dev_strategy_args *ap)
1017 cdev_t dev = ap->a_head.a_dev;
1018 struct bio *bio = ap->a_bio;
1025 bio->bio_buf->b_error = ENXIO;
1026 bio->bio_buf->b_flags |= B_ERROR;
1030 KKASSERT(dev->si_disk == dp);
1033 * The dscheck() function will also transform the slice relative
1034 * block number i.e. bio->bio_offset into a block number that can be
1035 * passed directly to the underlying raw device. If dscheck()
1036 * returns NULL it will have handled the bio for us (e.g. EOF
1037 * or error due to being beyond the device size).
1039 if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL) {
1040 dsched_queue(dp, nbio);
1048 * Return the partition size in ?blocks?
1052 diskpsize(struct dev_psize_args *ap)
1054 cdev_t dev = ap->a_head.a_dev;
1061 ap->a_result = dssize(dev, &dp->d_slice);
1063 if ((ap->a_result == -1) &&
1064 (dp->d_info.d_dsflags & DSO_DEVICEMAPPER)) {
1065 ap->a_head.a_dev = dp->d_rawdev;
1066 return dev_doperate(&ap->a_head);
1072 * When new device entries are instantiated, make sure they inherit our
1073 * si_disk structure and block and iosize limits from the raw device.
1075 * This routine is always called synchronously in the context of the
1078 * XXX The various io and block size constraints are not always initialized
1079 * properly by devices.
1083 diskclone(struct dev_clone_args *ap)
1085 cdev_t dev = ap->a_head.a_dev;
1089 KKASSERT(dp != NULL);
1091 dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
1092 dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
1093 dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
1098 diskdump(struct dev_dump_args *ap)
1100 cdev_t dev = ap->a_head.a_dev;
1101 struct disk *dp = dev->si_disk;
1102 u_int64_t size, offset;
1105 error = disk_dumpcheck(dev, &size, &ap->a_blkno, &ap->a_secsize);
1106 /* XXX: this should probably go in disk_dumpcheck somehow */
1107 if (ap->a_length != 0) {
1109 offset = ap->a_blkno * DEV_BSIZE;
1110 if ((ap->a_offset < offset) ||
1111 (ap->a_offset + ap->a_length - offset > size)) {
1112 kprintf("Attempt to write outside dump device boundaries.\n");
1118 ap->a_head.a_dev = dp->d_rawdev;
1119 error = dev_doperate(&ap->a_head);
1126 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
1127 0, sizeof(struct diskslices), "sizeof(struct diskslices)");
1129 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
1130 0, sizeof(struct disk), "sizeof(struct disk)");
1133 * Reorder interval for burst write allowance and minor write
1136 * We always want to trickle some writes in to make use of the
1137 * disk's zone cache. Bursting occurs on a longer interval and only
1138 * runningbufspace is well over the hirunningspace limit.
1140 int bioq_reorder_burst_interval = 60; /* should be multiple of minor */
1141 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_interval,
1142 CTLFLAG_RW, &bioq_reorder_burst_interval, 0, "");
1143 int bioq_reorder_minor_interval = 5;
1144 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_interval,
1145 CTLFLAG_RW, &bioq_reorder_minor_interval, 0, "");
1147 int bioq_reorder_burst_bytes = 3000000;
1148 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_bytes,
1149 CTLFLAG_RW, &bioq_reorder_burst_bytes, 0, "");
1150 int bioq_reorder_minor_bytes = 262144;
1151 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_bytes,
1152 CTLFLAG_RW, &bioq_reorder_minor_bytes, 0, "");
1156 * Order I/Os. Generally speaking this code is designed to make better
1157 * use of drive zone caches. A drive zone cache can typically track linear
1158 * reads or writes for around 16 zones simultaniously.
1160 * Read prioritization issues: It is possible for hundreds of megabytes worth
1161 * of writes to be queued asynchronously. This creates a huge bottleneck
1162 * for reads which reduce read bandwidth to a trickle.
1164 * To solve this problem we generally reorder reads before writes.
1166 * However, a large number of random reads can also starve writes and
1167 * make poor use of the drive zone cache so we allow writes to trickle
1171 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
1174 * The BIO wants to be ordered. Adding to the tail also
1175 * causes transition to be set to NULL, forcing the ordering
1176 * of all prior I/O's.
1178 if (bio->bio_buf->b_flags & B_ORDERED) {
1179 bioq_insert_tail(bioq, bio);
1183 switch(bio->bio_buf->b_cmd) {
1185 if (bioq->transition) {
1187 * Insert before the first write. Bleedover writes
1188 * based on reorder intervals to prevent starvation.
1190 TAILQ_INSERT_BEFORE(bioq->transition, bio, bio_act);
1192 if (bioq->reorder % bioq_reorder_minor_interval == 0) {
1193 bioqwritereorder(bioq);
1194 if (bioq->reorder >=
1195 bioq_reorder_burst_interval) {
1201 * No writes queued (or ordering was forced),
1204 TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act);
1209 * Writes are always appended. If no writes were previously
1210 * queued or an ordered tail insertion occured the transition
1211 * field will be NULL.
1213 TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act);
1214 if (bioq->transition == NULL)
1215 bioq->transition = bio;
1219 * All other request types are forced to be ordered.
1221 bioq_insert_tail(bioq, bio);
1227 * Move the read-write transition point to prevent reads from
1228 * completely starving our writes. This brings a number of writes into
1229 * the fold every N reads.
1231 * We bring a few linear writes into the fold on a minor interval
1232 * and we bring a non-linear burst of writes into the fold on a major
1233 * interval. Bursting only occurs if runningbufspace is really high
1234 * (typically from syncs, fsyncs, or HAMMER flushes).
1238 bioqwritereorder(struct bio_queue_head *bioq)
1246 if (bioq->reorder < bioq_reorder_burst_interval ||
1247 !buf_runningbufspace_severe()) {
1248 left = (size_t)bioq_reorder_minor_bytes;
1251 left = (size_t)bioq_reorder_burst_bytes;
1255 next_offset = bioq->transition->bio_offset;
1256 while ((bio = bioq->transition) != NULL &&
1257 (check_off == 0 || next_offset == bio->bio_offset)
1259 n = bio->bio_buf->b_bcount;
1260 next_offset = bio->bio_offset + n;
1261 bioq->transition = TAILQ_NEXT(bio, bio_act);
1269 * Bounds checking against the media size, used for the raw partition.
1270 * secsize, mediasize and b_blkno must all be the same units.
1271 * Possibly this has to be DEV_BSIZE (512).
1274 bounds_check_with_mediasize(struct bio *bio, int secsize, uint64_t mediasize)
1276 struct buf *bp = bio->bio_buf;
1279 sz = howmany(bp->b_bcount, secsize);
1281 if (bio->bio_offset/DEV_BSIZE + sz > mediasize) {
1282 sz = mediasize - bio->bio_offset/DEV_BSIZE;
1284 /* If exactly at end of disk, return EOF. */
1285 bp->b_resid = bp->b_bcount;
1289 /* If past end of disk, return EINVAL. */
1290 bp->b_error = EINVAL;
1293 /* Otherwise, truncate request. */
1294 bp->b_bcount = sz * secsize;
1301 * Disk error is the preface to plaintive error messages
1302 * about failing disk transfers. It prints messages of the form
1304 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
1306 * if the offset of the error in the transfer and a disk label
1307 * are both available. blkdone should be -1 if the position of the error
1308 * is unknown; the disklabel pointer may be null from drivers that have not
1309 * been converted to use them. The message is printed with kprintf
1310 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
1311 * The message should be completed (with at least a newline) with kprintf
1312 * or log(-1, ...), respectively. There is no trailing space.
1315 diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt)
1317 struct buf *bp = bio->bio_buf;
1331 kprintf("%s: %s %sing ", dev->si_name, what, term);
1332 kprintf("offset %012llx for %d",
1333 (long long)bio->bio_offset,
1337 kprintf(" (%d bytes completed)", donecnt);
1341 * Locate a disk device
1344 disk_locate(const char *devname)
1346 return devfs_find_device_by_name(devname);
1350 disk_config(void *arg)
1352 disk_msg_send_sync(DISK_SYNC, NULL, NULL);
1358 struct thread* td_core;
1360 disk_msg_cache = objcache_create("disk-msg-cache", 0, 0,
1362 objcache_malloc_alloc,
1363 objcache_malloc_free,
1364 &disk_msg_malloc_args);
1366 lwkt_token_init(&disklist_token, 1, "disks");
1369 * Initialize the reply-only port which acts as a message drain
1371 lwkt_initport_replyonly(&disk_dispose_port, disk_msg_autofree_reply);
1373 lwkt_gettoken(&disklist_token);
1374 lwkt_create(disk_msg_core, /*args*/NULL, &td_core, NULL,
1375 0, 0, "disk_msg_core");
1376 tsleep(td_core, 0, "diskcore", 0);
1377 lwkt_reltoken(&disklist_token);
1383 objcache_destroy(disk_msg_cache);
1387 * Clean out illegal characters in serial numbers.
1390 disk_cleanserial(char *serno)
1394 while ((c = *serno) != 0) {
1395 if (c >= 'a' && c <= 'z')
1397 else if (c >= 'A' && c <= 'Z')
1399 else if (c >= '0' && c <= '9')
1401 else if (c == '-' || c == '@' || c == '+' || c == '.')
1409 TUNABLE_INT("kern.disk_debug", &disk_debug_enable);
1410 SYSCTL_INT(_kern, OID_AUTO, disk_debug, CTLFLAG_RW, &disk_debug_enable,
1411 0, "Enable subr_disk debugging");
1413 SYSINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_FIRST, disk_init, NULL);
1414 SYSUNINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_ANY, disk_uninit, NULL);