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/msgport2.h>
105 #include <sys/buf2.h>
106 #include <sys/devfs.h>
107 #include <sys/thread.h>
108 #include <sys/thread2.h>
110 #include <sys/queue.h>
111 #include <sys/lock.h>
113 static MALLOC_DEFINE(M_DISK
, "disk", "disk data");
114 static int disk_debug_enable
= 0;
116 static void disk_msg_autofree_reply(lwkt_port_t
, lwkt_msg_t
);
117 static void disk_msg_core(void *);
118 static int disk_probe_slice(struct disk
*dp
, cdev_t dev
, int slice
, int reprobe
);
119 static void disk_probe(struct disk
*dp
, int reprobe
);
120 static void _setdiskinfo(struct disk
*disk
, struct disk_info
*info
);
121 static void bioqwritereorder(struct bio_queue_head
*bioq
);
122 static void disk_cleanserial(char *serno
);
124 static d_open_t diskopen
;
125 static d_close_t diskclose
;
126 static d_ioctl_t diskioctl
;
127 static d_strategy_t diskstrategy
;
128 static d_psize_t diskpsize
;
129 static d_clone_t diskclone
;
130 static d_dump_t diskdump
;
132 static LIST_HEAD(, disk
) disklist
= LIST_HEAD_INITIALIZER(&disklist
);
133 static struct lwkt_token disklist_token
;
135 static struct dev_ops disk_ops
= {
136 { "disk", 0, D_DISK
},
138 .d_close
= diskclose
,
140 .d_write
= physwrite
,
141 .d_ioctl
= diskioctl
,
142 .d_strategy
= diskstrategy
,
144 .d_psize
= diskpsize
,
148 static struct objcache
*disk_msg_cache
;
150 struct objcache_malloc_args disk_msg_malloc_args
= {
151 sizeof(struct disk_msg
), M_DISK
};
153 static struct lwkt_port disk_dispose_port
;
154 static struct lwkt_port disk_msg_port
;
157 disk_debug(int level
, char *fmt
, ...)
162 if (level
<= disk_debug_enable
)
170 disk_probe_slice(struct disk
*dp
, cdev_t dev
, int slice
, int reprobe
)
172 struct disk_info
*info
= &dp
->d_info
;
173 struct diskslice
*sp
= &dp
->d_slice
->dss_slices
[slice
];
175 struct partinfo part
;
182 "disk_probe_slice (begin): %s (%s)\n",
183 dev
->si_name
, dp
->d_cdev
->si_name
);
185 sno
= slice
? slice
- 1 : 0;
187 ops
= &disklabel32_ops
;
188 msg
= ops
->op_readdisklabel(dev
, sp
, &sp
->ds_label
, info
);
189 if (msg
&& !strcmp(msg
, "no disk label")) {
190 ops
= &disklabel64_ops
;
191 msg
= ops
->op_readdisklabel(dev
, sp
, &sp
->ds_label
, info
);
194 if (slice
!= WHOLE_DISK_SLICE
)
195 ops
->op_adjust_label_reserved(dp
->d_slice
, slice
, sp
);
200 for (i
= 0; i
< ops
->op_getnumparts(sp
->ds_label
); i
++) {
201 ops
->op_loadpartinfo(sp
->ds_label
, i
, &part
);
204 (ndev
= devfs_find_device_by_name("%s%c",
205 dev
->si_name
, 'a' + i
))
208 * Device already exists and
211 ndev
->si_flags
|= SI_REPROBE_TEST
;
213 ndev
= make_dev_covering(&disk_ops
, dp
->d_rawdev
->si_ops
,
214 dkmakeminor(dkunit(dp
->d_cdev
),
216 UID_ROOT
, GID_OPERATOR
, 0640,
217 "%s%c", dev
->si_name
, 'a'+ i
);
219 if (dp
->d_info
.d_serialno
) {
222 dp
->d_info
.d_serialno
,
225 ndev
->si_flags
|= SI_REPROBE_TEST
;
229 } else if (info
->d_dsflags
& DSO_COMPATLABEL
) {
231 if (sp
->ds_size
>= 0x100000000ULL
)
232 ops
= &disklabel64_ops
;
234 ops
= &disklabel32_ops
;
235 sp
->ds_label
= ops
->op_clone_label(info
, sp
);
237 if (sp
->ds_type
== DOSPTYP_386BSD
|| /* XXX */
238 sp
->ds_type
== DOSPTYP_NETBSD
||
239 sp
->ds_type
== DOSPTYP_OPENBSD
) {
240 log(LOG_WARNING
, "%s: cannot find label (%s)\n",
246 sp
->ds_wlabel
= FALSE
;
249 return (msg
? EINVAL
: 0);
253 * This routine is only called for newly minted drives or to reprobe
254 * a drive with no open slices. disk_probe_slice() is called directly
255 * when reprobing partition changes within slices.
258 disk_probe(struct disk
*dp
, int reprobe
)
260 struct disk_info
*info
= &dp
->d_info
;
261 cdev_t dev
= dp
->d_cdev
;
264 struct diskslices
*osp
;
265 struct diskslice
*sp
;
267 KKASSERT (info
->d_media_blksize
!= 0);
270 dp
->d_slice
= dsmakeslicestruct(BASE_SLICE
, info
);
271 disk_debug(1, "disk_probe (begin): %s\n", dp
->d_cdev
->si_name
);
273 error
= mbrinit(dev
, info
, &(dp
->d_slice
));
279 for (i
= 0; i
< dp
->d_slice
->dss_nslices
; i
++) {
281 * Ignore the whole-disk slice, it has already been created.
283 if (i
== WHOLE_DISK_SLICE
)
285 sp
= &dp
->d_slice
->dss_slices
[i
];
288 * Handle s0. s0 is a compatibility slice if there are no
289 * other slices and it has not otherwise been set up, else
292 if (i
== COMPATIBILITY_SLICE
) {
294 if (sp
->ds_type
== 0 &&
295 dp
->d_slice
->dss_nslices
== BASE_SLICE
) {
296 sp
->ds_size
= info
->d_media_blocks
;
305 * Ignore 0-length slices
307 if (sp
->ds_size
== 0)
311 (ndev
= devfs_find_device_by_name("%ss%d",
312 dev
->si_name
, sno
))) {
314 * Device already exists and is still valid
316 ndev
->si_flags
|= SI_REPROBE_TEST
;
319 * Else create new device
321 ndev
= make_dev_covering(&disk_ops
, dp
->d_rawdev
->si_ops
,
322 dkmakewholeslice(dkunit(dev
), i
),
323 UID_ROOT
, GID_OPERATOR
, 0640,
324 "%ss%d", dev
->si_name
, sno
);
325 if (dp
->d_info
.d_serialno
) {
326 make_dev_alias(ndev
, "serno/%s.s%d",
327 dp
->d_info
.d_serialno
, sno
);
330 ndev
->si_flags
|= SI_REPROBE_TEST
;
335 * Probe appropriate slices for a disklabel
337 * XXX slice type 1 used by our gpt probe code.
338 * XXX slice type 0 used by mbr compat slice.
340 if (sp
->ds_type
== DOSPTYP_386BSD
||
341 sp
->ds_type
== DOSPTYP_NETBSD
||
342 sp
->ds_type
== DOSPTYP_OPENBSD
||
345 if (dp
->d_slice
->dss_first_bsd_slice
== 0)
346 dp
->d_slice
->dss_first_bsd_slice
= i
;
347 disk_probe_slice(dp
, ndev
, i
, reprobe
);
351 disk_debug(1, "disk_probe (end): %s\n", dp
->d_cdev
->si_name
);
356 disk_msg_core(void *arg
)
359 struct diskslice
*sp
;
364 lwkt_initport_thread(&disk_msg_port
, curthread
);
369 msg
= (disk_msg_t
)lwkt_waitport(&disk_msg_port
, 0);
371 switch (msg
->hdr
.u
.ms_result
) {
372 case DISK_DISK_PROBE
:
373 dp
= (struct disk
*)msg
->load
;
375 "DISK_DISK_PROBE: %s\n",
376 dp
->d_cdev
->si_name
);
379 case DISK_DISK_DESTROY
:
380 dp
= (struct disk
*)msg
->load
;
382 "DISK_DISK_DESTROY: %s\n",
383 dp
->d_cdev
->si_name
);
384 devfs_destroy_subnames(dp
->d_cdev
->si_name
);
385 devfs_destroy_dev(dp
->d_cdev
);
386 lwkt_gettoken(&ilock
, &disklist_token
);
387 LIST_REMOVE(dp
, d_list
);
388 lwkt_reltoken(&ilock
);
389 if (dp
->d_info
.d_serialno
) {
390 kfree(dp
->d_info
.d_serialno
, M_TEMP
);
391 dp
->d_info
.d_serialno
= NULL
;
395 dp
= (struct disk
*)msg
->load
;
397 "DISK_DISK_UNPROBE: %s\n",
398 dp
->d_cdev
->si_name
);
399 devfs_destroy_subnames(dp
->d_cdev
->si_name
);
401 case DISK_SLICE_REPROBE
:
402 dp
= (struct disk
*)msg
->load
;
403 sp
= (struct diskslice
*)msg
->load2
;
404 devfs_clr_subnames_flag(sp
->ds_dev
->si_name
,
407 "DISK_SLICE_REPROBE: %s\n",
408 sp
->ds_dev
->si_name
);
409 disk_probe_slice(dp
, sp
->ds_dev
,
410 dkslice(sp
->ds_dev
), 1);
411 devfs_destroy_subnames_without_flag(
412 sp
->ds_dev
->si_name
, SI_REPROBE_TEST
);
414 case DISK_DISK_REPROBE
:
415 dp
= (struct disk
*)msg
->load
;
416 devfs_clr_subnames_flag(dp
->d_cdev
->si_name
, SI_REPROBE_TEST
);
418 "DISK_DISK_REPROBE: %s\n",
419 dp
->d_cdev
->si_name
);
421 devfs_destroy_subnames_without_flag(
422 dp
->d_cdev
->si_name
, SI_REPROBE_TEST
);
425 disk_debug(1, "DISK_SYNC\n");
428 devfs_debug(DEVFS_DEBUG_WARNING
,
429 "disk_msg_core: unknown message "
430 "received at core\n");
433 lwkt_replymsg(&msg
->hdr
, 0);
440 * Acts as a message drain. Any message that is replied to here gets
441 * destroyed and the memory freed.
444 disk_msg_autofree_reply(lwkt_port_t port
, lwkt_msg_t msg
)
446 objcache_put(disk_msg_cache
, msg
);
451 disk_msg_send(uint32_t cmd
, void *load
, void *load2
)
454 lwkt_port_t port
= &disk_msg_port
;
456 disk_msg
= objcache_get(disk_msg_cache
, M_WAITOK
);
458 lwkt_initmsg(&disk_msg
->hdr
, &disk_dispose_port
, 0);
460 disk_msg
->hdr
.u
.ms_result
= cmd
;
461 disk_msg
->load
= load
;
462 disk_msg
->load2
= load2
;
464 lwkt_sendmsg(port
, &disk_msg
->hdr
);
468 disk_msg_send_sync(uint32_t cmd
, void *load
, void *load2
)
470 struct lwkt_port rep_port
;
474 disk_msg
= objcache_get(disk_msg_cache
, M_WAITOK
);
475 port
= &disk_msg_port
;
477 /* XXX could probably use curthread's built-in msgport */
478 lwkt_initport_thread(&rep_port
, curthread
);
479 lwkt_initmsg(&disk_msg
->hdr
, &rep_port
, 0);
481 disk_msg
->hdr
.u
.ms_result
= cmd
;
482 disk_msg
->load
= load
;
483 disk_msg
->load2
= load2
;
485 lwkt_sendmsg(port
, &disk_msg
->hdr
);
486 lwkt_waitmsg(&disk_msg
->hdr
, 0);
487 objcache_put(disk_msg_cache
, disk_msg
);
491 * Create a raw device for the dev_ops template (which is returned). Also
492 * create a slice and unit managed disk and overload the user visible
493 * device space with it.
495 * NOTE: The returned raw device is NOT a slice and unit managed device.
496 * It is an actual raw device representing the raw disk as specified by
497 * the passed dev_ops. The disk layer not only returns such a raw device,
498 * it also uses it internally when passing (modified) commands through.
501 disk_create(int unit
, struct disk
*dp
, struct dev_ops
*raw_ops
)
507 "disk_create (begin): %s%d\n",
508 raw_ops
->head
.name
, unit
);
510 rawdev
= make_only_dev(raw_ops
, dkmakewholedisk(unit
),
511 UID_ROOT
, GID_OPERATOR
, 0640,
512 "%s%d", raw_ops
->head
.name
, unit
);
514 bzero(dp
, sizeof(*dp
));
516 dp
->d_rawdev
= rawdev
;
517 dp
->d_raw_ops
= raw_ops
;
518 dp
->d_dev_ops
= &disk_ops
;
519 dp
->d_cdev
= make_dev_covering(&disk_ops
, dp
->d_rawdev
->si_ops
,
520 dkmakewholedisk(unit
),
521 UID_ROOT
, GID_OPERATOR
, 0640,
522 "%s%d", raw_ops
->head
.name
, unit
);
524 dp
->d_cdev
->si_disk
= dp
;
526 lwkt_gettoken(&ilock
, &disklist_token
);
527 LIST_INSERT_HEAD(&disklist
, dp
, d_list
);
528 lwkt_reltoken(&ilock
);
531 "disk_create (end): %s%d\n",
532 raw_ops
->head
.name
, unit
);
534 return (dp
->d_rawdev
);
539 _setdiskinfo(struct disk
*disk
, struct disk_info
*info
)
543 oldserialno
= disk
->d_info
.d_serialno
;
544 bcopy(info
, &disk
->d_info
, sizeof(disk
->d_info
));
545 info
= &disk
->d_info
;
548 "_setdiskinfo: %s\n",
549 disk
->d_cdev
->si_name
);
552 * The serial number is duplicated so the caller can throw
555 if (info
->d_serialno
&& info
->d_serialno
[0]) {
556 info
->d_serialno
= kstrdup(info
->d_serialno
, M_TEMP
);
557 disk_cleanserial(info
->d_serialno
);
559 make_dev_alias(disk
->d_cdev
, "serno/%s",
563 info
->d_serialno
= NULL
;
566 kfree(oldserialno
, M_TEMP
);
569 * The caller may set d_media_size or d_media_blocks and we
570 * calculate the other.
572 KKASSERT(info
->d_media_size
== 0 || info
->d_media_blksize
== 0);
573 if (info
->d_media_size
== 0 && info
->d_media_blocks
) {
574 info
->d_media_size
= (u_int64_t
)info
->d_media_blocks
*
575 info
->d_media_blksize
;
576 } else if (info
->d_media_size
&& info
->d_media_blocks
== 0 &&
577 info
->d_media_blksize
) {
578 info
->d_media_blocks
= info
->d_media_size
/
579 info
->d_media_blksize
;
583 * The si_* fields for rawdev are not set until after the
584 * disk_create() call, so someone using the cooked version
585 * of the raw device (i.e. da0s0) will not get the right
586 * si_iosize_max unless we fix it up here.
588 if (disk
->d_cdev
&& disk
->d_rawdev
&&
589 disk
->d_cdev
->si_iosize_max
== 0) {
590 disk
->d_cdev
->si_iosize_max
= disk
->d_rawdev
->si_iosize_max
;
591 disk
->d_cdev
->si_bsize_phys
= disk
->d_rawdev
->si_bsize_phys
;
592 disk
->d_cdev
->si_bsize_best
= disk
->d_rawdev
->si_bsize_best
;
597 * Disk drivers must call this routine when media parameters are available
601 disk_setdiskinfo(struct disk
*disk
, struct disk_info
*info
)
603 _setdiskinfo(disk
, info
);
604 disk_msg_send(DISK_DISK_PROBE
, disk
, NULL
);
606 "disk_setdiskinfo: sent probe for %s\n",
607 disk
->d_cdev
->si_name
);
611 disk_setdiskinfo_sync(struct disk
*disk
, struct disk_info
*info
)
613 _setdiskinfo(disk
, info
);
614 disk_msg_send_sync(DISK_DISK_PROBE
, disk
, NULL
);
616 "disk_setdiskinfo_sync: sent probe for %s\n",
617 disk
->d_cdev
->si_name
);
621 * This routine is called when an adapter detaches. The higher level
622 * managed disk device is destroyed while the lower level raw device is
626 disk_destroy(struct disk
*disk
)
628 disk_msg_send_sync(DISK_DISK_DESTROY
, disk
, NULL
);
633 disk_dumpcheck(cdev_t dev
, u_int64_t
*size
, u_int64_t
*blkno
, u_int32_t
*secsize
)
635 struct partinfo pinfo
;
638 bzero(&pinfo
, sizeof(pinfo
));
639 error
= dev_dioctl(dev
, DIOCGPART
, (void *)&pinfo
, 0,
640 proc0
.p_ucred
, NULL
);
644 if (pinfo
.media_blksize
== 0)
647 if (blkno
) /* XXX: make sure this reserved stuff is right */
648 *blkno
= pinfo
.reserved_blocks
+
649 pinfo
.media_offset
/ pinfo
.media_blksize
;
651 *secsize
= pinfo
.media_blksize
;
653 *size
= (pinfo
.media_blocks
- pinfo
.reserved_blocks
);
659 disk_dumpconf(cdev_t dev
, u_int onoff
)
661 struct dumperinfo di
;
662 u_int64_t size
, blkno
;
667 return set_dumper(NULL
);
669 error
= disk_dumpcheck(dev
, &size
, &blkno
, &secsize
);
674 bzero(&di
, sizeof(struct dumperinfo
));
675 di
.dumper
= diskdump
;
677 di
.blocksize
= secsize
;
678 di
.mediaoffset
= blkno
* DEV_BSIZE
;
679 di
.mediasize
= size
* DEV_BSIZE
;
681 return set_dumper(&di
);
685 disk_unprobe(struct disk
*disk
)
690 disk_msg_send_sync(DISK_UNPROBE
, disk
, NULL
);
694 disk_invalidate (struct disk
*disk
)
696 dsgone(&disk
->d_slice
);
700 disk_enumerate(struct disk
*disk
)
705 lwkt_gettoken(&ilock
, &disklist_token
);
707 dp
= (LIST_FIRST(&disklist
));
709 dp
= (LIST_NEXT(disk
, d_list
));
710 lwkt_reltoken(&ilock
);
717 sysctl_disks(SYSCTL_HANDLER_ARGS
)
725 while ((disk
= disk_enumerate(disk
))) {
727 error
= SYSCTL_OUT(req
, " ", 1);
733 error
= SYSCTL_OUT(req
, disk
->d_rawdev
->si_name
,
734 strlen(disk
->d_rawdev
->si_name
));
738 error
= SYSCTL_OUT(req
, "", 1);
742 SYSCTL_PROC(_kern
, OID_AUTO
, disks
, CTLTYPE_STRING
| CTLFLAG_RD
, NULL
, 0,
743 sysctl_disks
, "A", "names of available disks");
746 * Open a disk device or partition.
750 diskopen(struct dev_open_args
*ap
)
752 cdev_t dev
= ap
->a_head
.a_dev
;
757 * dp can't be NULL here XXX.
759 * d_slice will be NULL if setdiskinfo() has not been called yet.
760 * setdiskinfo() is typically called whether the disk is present
761 * or not (e.g. CD), but the base disk device is created first
762 * and there may be a race.
765 if (dp
== NULL
|| dp
->d_slice
== NULL
)
770 * Deal with open races
772 while (dp
->d_flags
& DISKFLAG_LOCK
) {
773 dp
->d_flags
|= DISKFLAG_WANTED
;
774 error
= tsleep(dp
, PCATCH
, "diskopen", hz
);
778 dp
->d_flags
|= DISKFLAG_LOCK
;
781 * Open the underlying raw device.
783 if (!dsisopen(dp
->d_slice
)) {
785 if (!pdev
->si_iosize_max
)
786 pdev
->si_iosize_max
= dev
->si_iosize_max
;
788 error
= dev_dopen(dp
->d_rawdev
, ap
->a_oflags
,
789 ap
->a_devtype
, ap
->a_cred
);
793 * Inherit properties from the underlying device now that it is
801 error
= dsopen(dev
, ap
->a_devtype
, dp
->d_info
.d_dsflags
,
802 &dp
->d_slice
, &dp
->d_info
);
803 if (!dsisopen(dp
->d_slice
)) {
804 dev_dclose(dp
->d_rawdev
, ap
->a_oflags
, ap
->a_devtype
);
807 dp
->d_flags
&= ~DISKFLAG_LOCK
;
808 if (dp
->d_flags
& DISKFLAG_WANTED
) {
809 dp
->d_flags
&= ~DISKFLAG_WANTED
;
817 * Close a disk device or partition
821 diskclose(struct dev_close_args
*ap
)
823 cdev_t dev
= ap
->a_head
.a_dev
;
830 dsclose(dev
, ap
->a_devtype
, dp
->d_slice
);
831 if (!dsisopen(dp
->d_slice
)) {
832 error
= dev_dclose(dp
->d_rawdev
, ap
->a_fflag
, ap
->a_devtype
);
838 * First execute the ioctl on the disk device, and if it isn't supported
839 * try running it on the backing device.
843 diskioctl(struct dev_ioctl_args
*ap
)
845 cdev_t dev
= ap
->a_head
.a_dev
;
854 devfs_debug(DEVFS_DEBUG_DEBUG
,
855 "diskioctl: cmd is: %x (name: %s)\n",
856 ap
->a_cmd
, dev
->si_name
);
857 devfs_debug(DEVFS_DEBUG_DEBUG
,
858 "diskioctl: &dp->d_slice is: %x, %x\n",
859 &dp
->d_slice
, dp
->d_slice
);
861 if (ap
->a_cmd
== DIOCGKERNELDUMP
) {
862 u
= *(u_int
*)ap
->a_data
;
863 return disk_dumpconf(dev
, u
);
866 error
= dsioctl(dev
, ap
->a_cmd
, ap
->a_data
, ap
->a_fflag
,
867 &dp
->d_slice
, &dp
->d_info
);
869 if (error
== ENOIOCTL
) {
870 error
= dev_dioctl(dp
->d_rawdev
, ap
->a_cmd
, ap
->a_data
,
871 ap
->a_fflag
, ap
->a_cred
, NULL
);
877 * Execute strategy routine
881 diskstrategy(struct dev_strategy_args
*ap
)
883 cdev_t dev
= ap
->a_head
.a_dev
;
884 struct bio
*bio
= ap
->a_bio
;
891 bio
->bio_buf
->b_error
= ENXIO
;
892 bio
->bio_buf
->b_flags
|= B_ERROR
;
896 KKASSERT(dev
->si_disk
== dp
);
899 * The dscheck() function will also transform the slice relative
900 * block number i.e. bio->bio_offset into a block number that can be
901 * passed directly to the underlying raw device. If dscheck()
902 * returns NULL it will have handled the bio for us (e.g. EOF
903 * or error due to being beyond the device size).
905 if ((nbio
= dscheck(dev
, bio
, dp
->d_slice
)) != NULL
) {
906 dev_dstrategy(dp
->d_rawdev
, nbio
);
914 * Return the partition size in ?blocks?
918 diskpsize(struct dev_psize_args
*ap
)
920 cdev_t dev
= ap
->a_head
.a_dev
;
926 ap
->a_result
= dssize(dev
, &dp
->d_slice
);
931 * When new device entries are instantiated, make sure they inherit our
932 * si_disk structure and block and iosize limits from the raw device.
934 * This routine is always called synchronously in the context of the
937 * XXX The various io and block size constraints are not always initialized
938 * properly by devices.
942 diskclone(struct dev_clone_args
*ap
)
944 cdev_t dev
= ap
->a_head
.a_dev
;
948 KKASSERT(dp
!= NULL
);
950 dev
->si_iosize_max
= dp
->d_rawdev
->si_iosize_max
;
951 dev
->si_bsize_phys
= dp
->d_rawdev
->si_bsize_phys
;
952 dev
->si_bsize_best
= dp
->d_rawdev
->si_bsize_best
;
957 diskdump(struct dev_dump_args
*ap
)
959 cdev_t dev
= ap
->a_head
.a_dev
;
960 struct disk
*dp
= dev
->si_disk
;
961 u_int64_t size
, offset
;
964 error
= disk_dumpcheck(dev
, &size
, &ap
->a_blkno
, &ap
->a_secsize
);
965 /* XXX: this should probably go in disk_dumpcheck somehow */
966 if (ap
->a_length
!= 0) {
968 offset
= ap
->a_blkno
* DEV_BSIZE
;
969 if ((ap
->a_offset
< offset
) ||
970 (ap
->a_offset
+ ap
->a_length
- offset
> size
)) {
971 kprintf("Attempt to write outside dump device boundaries.\n");
977 ap
->a_head
.a_dev
= dp
->d_rawdev
;
978 error
= dev_doperate(&ap
->a_head
);
985 SYSCTL_INT(_debug_sizeof
, OID_AUTO
, diskslices
, CTLFLAG_RD
,
986 0, sizeof(struct diskslices
), "sizeof(struct diskslices)");
988 SYSCTL_INT(_debug_sizeof
, OID_AUTO
, disk
, CTLFLAG_RD
,
989 0, sizeof(struct disk
), "sizeof(struct disk)");
992 * Reorder interval for burst write allowance and minor write
995 * We always want to trickle some writes in to make use of the
996 * disk's zone cache. Bursting occurs on a longer interval and only
997 * runningbufspace is well over the hirunningspace limit.
999 int bioq_reorder_burst_interval
= 60; /* should be multiple of minor */
1000 SYSCTL_INT(_kern
, OID_AUTO
, bioq_reorder_burst_interval
,
1001 CTLFLAG_RW
, &bioq_reorder_burst_interval
, 0, "");
1002 int bioq_reorder_minor_interval
= 5;
1003 SYSCTL_INT(_kern
, OID_AUTO
, bioq_reorder_minor_interval
,
1004 CTLFLAG_RW
, &bioq_reorder_minor_interval
, 0, "");
1006 int bioq_reorder_burst_bytes
= 3000000;
1007 SYSCTL_INT(_kern
, OID_AUTO
, bioq_reorder_burst_bytes
,
1008 CTLFLAG_RW
, &bioq_reorder_burst_bytes
, 0, "");
1009 int bioq_reorder_minor_bytes
= 262144;
1010 SYSCTL_INT(_kern
, OID_AUTO
, bioq_reorder_minor_bytes
,
1011 CTLFLAG_RW
, &bioq_reorder_minor_bytes
, 0, "");
1015 * Order I/Os. Generally speaking this code is designed to make better
1016 * use of drive zone caches. A drive zone cache can typically track linear
1017 * reads or writes for around 16 zones simultaniously.
1019 * Read prioritization issues: It is possible for hundreds of megabytes worth
1020 * of writes to be queued asynchronously. This creates a huge bottleneck
1021 * for reads which reduce read bandwidth to a trickle.
1023 * To solve this problem we generally reorder reads before writes.
1025 * However, a large number of random reads can also starve writes and
1026 * make poor use of the drive zone cache so we allow writes to trickle
1030 bioqdisksort(struct bio_queue_head
*bioq
, struct bio
*bio
)
1033 * The BIO wants to be ordered. Adding to the tail also
1034 * causes transition to be set to NULL, forcing the ordering
1035 * of all prior I/O's.
1037 if (bio
->bio_buf
->b_flags
& B_ORDERED
) {
1038 bioq_insert_tail(bioq
, bio
);
1042 switch(bio
->bio_buf
->b_cmd
) {
1044 if (bioq
->transition
) {
1046 * Insert before the first write. Bleedover writes
1047 * based on reorder intervals to prevent starvation.
1049 TAILQ_INSERT_BEFORE(bioq
->transition
, bio
, bio_act
);
1051 if (bioq
->reorder
% bioq_reorder_minor_interval
== 0) {
1052 bioqwritereorder(bioq
);
1053 if (bioq
->reorder
>=
1054 bioq_reorder_burst_interval
) {
1060 * No writes queued (or ordering was forced),
1063 TAILQ_INSERT_TAIL(&bioq
->queue
, bio
, bio_act
);
1068 * Writes are always appended. If no writes were previously
1069 * queued or an ordered tail insertion occured the transition
1070 * field will be NULL.
1072 TAILQ_INSERT_TAIL(&bioq
->queue
, bio
, bio_act
);
1073 if (bioq
->transition
== NULL
)
1074 bioq
->transition
= bio
;
1078 * All other request types are forced to be ordered.
1080 bioq_insert_tail(bioq
, bio
);
1086 * Move the read-write transition point to prevent reads from
1087 * completely starving our writes. This brings a number of writes into
1088 * the fold every N reads.
1090 * We bring a few linear writes into the fold on a minor interval
1091 * and we bring a non-linear burst of writes into the fold on a major
1092 * interval. Bursting only occurs if runningbufspace is really high
1093 * (typically from syncs, fsyncs, or HAMMER flushes).
1097 bioqwritereorder(struct bio_queue_head
*bioq
)
1105 if (bioq
->reorder
< bioq_reorder_burst_interval
||
1106 !buf_runningbufspace_severe()) {
1107 left
= (size_t)bioq_reorder_minor_bytes
;
1110 left
= (size_t)bioq_reorder_burst_bytes
;
1114 next_offset
= bioq
->transition
->bio_offset
;
1115 while ((bio
= bioq
->transition
) != NULL
&&
1116 (check_off
== 0 || next_offset
== bio
->bio_offset
)
1118 n
= bio
->bio_buf
->b_bcount
;
1119 next_offset
= bio
->bio_offset
+ n
;
1120 bioq
->transition
= TAILQ_NEXT(bio
, bio_act
);
1128 * Disk error is the preface to plaintive error messages
1129 * about failing disk transfers. It prints messages of the form
1131 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
1133 * if the offset of the error in the transfer and a disk label
1134 * are both available. blkdone should be -1 if the position of the error
1135 * is unknown; the disklabel pointer may be null from drivers that have not
1136 * been converted to use them. The message is printed with kprintf
1137 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
1138 * The message should be completed (with at least a newline) with kprintf
1139 * or log(-1, ...), respectively. There is no trailing space.
1142 diskerr(struct bio
*bio
, cdev_t dev
, const char *what
, int pri
, int donecnt
)
1144 struct buf
*bp
= bio
->bio_buf
;
1158 kprintf("%s: %s %sing ", dev
->si_name
, what
, term
);
1159 kprintf("offset %012llx for %d",
1160 (long long)bio
->bio_offset
,
1164 kprintf(" (%d bytes completed)", donecnt
);
1168 * Locate a disk device
1171 disk_locate(const char *devname
)
1173 return devfs_find_device_by_name(devname
);
1177 disk_config(void *arg
)
1179 disk_msg_send_sync(DISK_SYNC
, NULL
, NULL
);
1185 struct thread
* td_core
;
1187 disk_msg_cache
= objcache_create("disk-msg-cache", 0, 0,
1189 objcache_malloc_alloc
,
1190 objcache_malloc_free
,
1191 &disk_msg_malloc_args
);
1193 lwkt_token_init(&disklist_token
);
1196 * Initialize the reply-only port which acts as a message drain
1198 lwkt_initport_replyonly(&disk_dispose_port
, disk_msg_autofree_reply
);
1200 lwkt_create(disk_msg_core
, /*args*/NULL
, &td_core
, NULL
,
1201 0, 0, "disk_msg_core");
1203 tsleep(td_core
, 0, "diskcore", 0);
1209 objcache_destroy(disk_msg_cache
);
1213 * Clean out illegal characters in serial numbers.
1216 disk_cleanserial(char *serno
)
1220 while ((c
= *serno
) != 0) {
1221 if (c
>= 'a' && c
<= 'z')
1223 else if (c
>= 'A' && c
<= 'Z')
1225 else if (c
>= '0' && c
<= '9')
1227 else if (c
== '-' || c
== '@' || c
== '+' || c
== '.')
1235 TUNABLE_INT("kern.disk_debug", &disk_debug_enable
);
1236 SYSCTL_INT(_kern
, OID_AUTO
, disk_debug
, CTLFLAG_RW
, &disk_debug_enable
,
1237 0, "Enable subr_disk debugging");
1239 SYSINIT(disk_register
, SI_SUB_PRE_DRIVERS
, SI_ORDER_FIRST
, disk_init
, NULL
);
1240 SYSUNINIT(disk_register
, SI_SUB_PRE_DRIVERS
, SI_ORDER_ANY
, disk_uninit
, NULL
);