2 * Copyright (c) 2003,2004 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * ----------------------------------------------------------------------------
35 * "THE BEER-WARE LICENSE" (Revision 42):
36 * <phk@FreeBSD.ORG> wrote this file. As long as you retain this notice you
37 * can do whatever you want with this stuff. If we meet some day, and you think
38 * this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp
39 * ----------------------------------------------------------------------------
41 * Copyright (c) 1982, 1986, 1988, 1993
42 * The Regents of the University of California. All rights reserved.
43 * (c) UNIX System Laboratories, Inc.
44 * All or some portions of this file are derived from material licensed
45 * to the University of California by American Telephone and Telegraph
46 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
47 * the permission of UNIX System Laboratories, Inc.
49 * Redistribution and use in source and binary forms, with or without
50 * modification, are permitted provided that the following conditions
52 * 1. Redistributions of source code must retain the above copyright
53 * notice, this list of conditions and the following disclaimer.
54 * 2. Redistributions in binary form must reproduce the above copyright
55 * notice, this list of conditions and the following disclaimer in the
56 * documentation and/or other materials provided with the distribution.
57 * 3. All advertising materials mentioning features or use of this software
58 * must display the following acknowledgement:
59 * This product includes software developed by the University of
60 * California, Berkeley and its contributors.
61 * 4. Neither the name of the University nor the names of its contributors
62 * may be used to endorse or promote products derived from this software
63 * without specific prior written permission.
65 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
66 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
67 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
68 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
69 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
70 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
71 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
72 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
73 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
74 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
77 * @(#)ufs_disksubr.c 8.5 (Berkeley) 1/21/94
78 * $FreeBSD: src/sys/kern/subr_disk.c,v 1.20.2.6 2001/10/05 07:14:57 peter Exp $
79 * $FreeBSD: src/sys/ufs/ufs/ufs_disksubr.c,v 1.44.2.3 2001/03/05 05:42:19 obrien Exp $
80 * $DragonFly: src/sys/kern/subr_disk.c,v 1.40 2008/06/05 18:06:32 swildner Exp $
83 #include <sys/param.h>
84 #include <sys/systm.h>
85 #include <sys/kernel.h>
87 #include <sys/sysctl.h>
90 #include <sys/diskslice.h>
92 #include <sys/malloc.h>
93 #include <sys/sysctl.h>
94 #include <machine/md_var.h>
95 #include <sys/ctype.h>
96 #include <sys/syslog.h>
97 #include <sys/device.h>
98 #include <sys/msgport.h>
99 #include <sys/msgport2.h>
100 #include <sys/buf2.h>
102 static MALLOC_DEFINE(M_DISK, "disk", "disk data");
104 static d_open_t diskopen;
105 static d_close_t diskclose;
106 static d_ioctl_t diskioctl;
107 static d_strategy_t diskstrategy;
108 static d_psize_t diskpsize;
109 static d_clone_t diskclone;
110 static d_dump_t diskdump;
112 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
114 static struct dev_ops disk_ops = {
115 { "disk", 0, D_DISK },
117 .d_close = diskclose,
119 .d_write = physwrite,
120 .d_ioctl = diskioctl,
121 .d_strategy = diskstrategy,
123 .d_psize = diskpsize,
128 * Create a raw device for the dev_ops template (which is returned). Also
129 * create a slice and unit managed disk and overload the user visible
130 * device space with it.
132 * NOTE: The returned raw device is NOT a slice and unit managed device.
133 * It is an actual raw device representing the raw disk as specified by
134 * the passed dev_ops. The disk layer not only returns such a raw device,
135 * it also uses it internally when passing (modified) commands through.
138 disk_create(int unit, struct disk *dp, struct dev_ops *raw_ops)
141 struct dev_ops *dev_ops;
144 * Create the raw backing device
146 compile_dev_ops(raw_ops);
147 rawdev = make_dev(raw_ops, dkmakewholedisk(unit),
148 UID_ROOT, GID_OPERATOR, 0640,
149 "%s%d", raw_ops->head.name, unit);
151 bzero(dp, sizeof(*dp));
154 * We install a custom cdevsw rather then the passed cdevsw,
155 * and save our disk structure in d_data so we can get at it easily
156 * without any complex cloning code.
158 dev_ops = dev_ops_add_override(rawdev, &disk_ops,
159 dkunitmask(), dkmakeunit(unit));
160 dev_ops->head.data = dp;
162 dp->d_rawdev = rawdev;
163 dp->d_raw_ops = raw_ops;
164 dp->d_dev_ops = dev_ops;
165 dp->d_cdev = make_dev(dev_ops,
166 dkmakewholedisk(unit),
167 UID_ROOT, GID_OPERATOR, 0640,
168 "%s%d", dev_ops->head.name, unit);
170 LIST_INSERT_HEAD(&disklist, dp, d_list);
171 return (dp->d_rawdev);
175 * Disk drivers must call this routine when media parameters are available
179 disk_setdiskinfo(struct disk *disk, struct disk_info *info)
181 bcopy(info, &disk->d_info, sizeof(disk->d_info));
182 info = &disk->d_info;
184 KKASSERT(info->d_media_size == 0 || info->d_media_blksize == 0);
185 if (info->d_media_size == 0 && info->d_media_blocks) {
186 info->d_media_size = (u_int64_t)info->d_media_blocks *
187 info->d_media_blksize;
188 } else if (info->d_media_size && info->d_media_blocks == 0 &&
189 info->d_media_blksize) {
190 info->d_media_blocks = info->d_media_size /
191 info->d_media_blksize;
196 * This routine is called when an adapter detaches. The higher level
197 * managed disk device is destroyed while the lower level raw device is
201 disk_destroy(struct disk *disk)
203 if (disk->d_dev_ops) {
204 dev_ops_remove(disk->d_dev_ops, dkunitmask(),
205 dkmakeunit(dkunit(disk->d_cdev)));
206 LIST_REMOVE(disk, d_list);
208 if (disk->d_raw_ops) {
209 destroy_all_devs(disk->d_raw_ops, dkunitmask(),
210 dkmakeunit(dkunit(disk->d_rawdev)));
212 bzero(disk, sizeof(*disk));
216 disk_dumpcheck(cdev_t dev, u_int64_t *count, u_int64_t *blkno, u_int *secsize)
218 struct partinfo pinfo;
221 bzero(&pinfo, sizeof(pinfo));
222 error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0, proc0.p_ucred);
225 if (pinfo.media_blksize == 0)
227 *count = (u_int64_t)Maxmem * PAGE_SIZE / pinfo.media_blksize;
228 if (dumplo64 < pinfo.reserved_blocks ||
229 dumplo64 + *count > pinfo.media_blocks) {
232 *blkno = dumplo64 + pinfo.media_offset / pinfo.media_blksize;
233 *secsize = pinfo.media_blksize;
238 disk_invalidate (struct disk *disk)
241 dsgone(&disk->d_slice);
245 disk_enumerate(struct disk *disk)
248 return (LIST_FIRST(&disklist));
250 return (LIST_NEXT(disk, d_list));
255 sysctl_disks(SYSCTL_HANDLER_ARGS)
263 while ((disk = disk_enumerate(disk))) {
265 error = SYSCTL_OUT(req, " ", 1);
271 error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
272 strlen(disk->d_rawdev->si_name));
276 error = SYSCTL_OUT(req, "", 1);
280 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, NULL, 0,
281 sysctl_disks, "A", "names of available disks");
284 * Open a disk device or partition.
288 diskopen(struct dev_open_args *ap)
290 cdev_t dev = ap->a_head.a_dev;
295 * dp can't be NULL here XXX.
303 * Deal with open races
305 while (dp->d_flags & DISKFLAG_LOCK) {
306 dp->d_flags |= DISKFLAG_WANTED;
307 error = tsleep(dp, PCATCH, "diskopen", hz);
311 dp->d_flags |= DISKFLAG_LOCK;
314 * Open the underlying raw device.
316 if (!dsisopen(dp->d_slice)) {
318 if (!pdev->si_iosize_max)
319 pdev->si_iosize_max = dev->si_iosize_max;
321 error = dev_dopen(dp->d_rawdev, ap->a_oflags,
322 ap->a_devtype, ap->a_cred);
326 * Inherit properties from the underlying device now that it is
334 error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags,
335 &dp->d_slice, &dp->d_info);
337 if (!dsisopen(dp->d_slice))
338 dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype);
340 dp->d_flags &= ~DISKFLAG_LOCK;
341 if (dp->d_flags & DISKFLAG_WANTED) {
342 dp->d_flags &= ~DISKFLAG_WANTED;
350 * Close a disk device or partition
354 diskclose(struct dev_close_args *ap)
356 cdev_t dev = ap->a_head.a_dev;
363 dsclose(dev, ap->a_devtype, dp->d_slice);
364 if (!dsisopen(dp->d_slice))
365 error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype);
370 * First execute the ioctl on the disk device, and if it isn't supported
371 * try running it on the backing device.
375 diskioctl(struct dev_ioctl_args *ap)
377 cdev_t dev = ap->a_head.a_dev;
384 error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag,
385 &dp->d_slice, &dp->d_info);
386 if (error == ENOIOCTL) {
387 error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data,
388 ap->a_fflag, ap->a_cred);
394 * Execute strategy routine
398 diskstrategy(struct dev_strategy_args *ap)
400 cdev_t dev = ap->a_head.a_dev;
401 struct bio *bio = ap->a_bio;
408 bio->bio_buf->b_error = ENXIO;
409 bio->bio_buf->b_flags |= B_ERROR;
413 KKASSERT(dev->si_disk == dp);
416 * The dscheck() function will also transform the slice relative
417 * block number i.e. bio->bio_offset into a block number that can be
418 * passed directly to the underlying raw device. If dscheck()
419 * returns NULL it will have handled the bio for us (e.g. EOF
420 * or error due to being beyond the device size).
422 if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL)
423 dev_dstrategy(dp->d_rawdev, nbio);
430 * Return the partition size in ?blocks?
434 diskpsize(struct dev_psize_args *ap)
436 cdev_t dev = ap->a_head.a_dev;
442 ap->a_result = dssize(dev, &dp->d_slice);
447 * When new device entries are instantiated, make sure they inherit our
448 * si_disk structure and block and iosize limits from the raw device.
450 * This routine is always called synchronously in the context of the
453 * XXX The various io and block size constraints are not always initialized
454 * properly by devices.
458 diskclone(struct dev_clone_args *ap)
460 cdev_t dev = ap->a_head.a_dev;
463 dp = dev->si_ops->head.data;
464 KKASSERT(dp != NULL);
466 dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
467 dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
468 dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
473 diskdump(struct dev_dump_args *ap)
475 cdev_t dev = ap->a_head.a_dev;
476 struct disk *dp = dev->si_ops->head.data;
479 error = disk_dumpcheck(dev, &ap->a_count, &ap->a_blkno, &ap->a_secsize);
481 ap->a_head.a_dev = dp->d_rawdev;
482 error = dev_doperate(&ap->a_head);
489 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
490 0, sizeof(struct diskslices), "sizeof(struct diskslices)");
492 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
493 0, sizeof(struct disk), "sizeof(struct disk)");
497 * Seek sort for disks.
499 * The bio_queue keep two queues, sorted in ascending block order. The first
500 * queue holds those requests which are positioned after the current block
501 * (in the first request); the second, which starts at queue->switch_point,
502 * holds requests which came in after their block number was passed. Thus
503 * we implement a one way scan, retracting after reaching the end of the drive
504 * to the first request on the second queue, at which time it becomes the
507 * A one-way scan is natural because of the way UNIX read-ahead blocks are
511 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
517 be = TAILQ_LAST(&bioq->queue, bio_queue);
519 * If the queue is empty or we are an
520 * ordered transaction, then it's easy.
522 if ((bq = bioq_first(bioq)) == NULL ||
523 (bio->bio_buf->b_flags & B_ORDERED) != 0) {
524 bioq_insert_tail(bioq, bio);
526 } else if (bioq->insert_point != NULL) {
529 * A certain portion of the list is
530 * "locked" to preserve ordering, so
531 * we can only insert after the insert
534 bq = bioq->insert_point;
538 * If we lie before the last removed (currently active)
539 * request, and are not inserting ourselves into the
540 * "locked" portion of the list, then we must add ourselves
541 * to the second request list.
543 if (bio->bio_offset < bioq->last_offset) {
544 bq = bioq->switch_point;
546 * If we are starting a new secondary list,
550 bioq->switch_point = bio;
551 bioq_insert_tail(bioq, bio);
555 * If we lie ahead of the current switch point,
556 * insert us before the switch point and move
559 if (bio->bio_offset < bq->bio_offset) {
560 bioq->switch_point = bio;
561 TAILQ_INSERT_BEFORE(bq, bio, bio_act);
565 if (bioq->switch_point != NULL)
566 be = TAILQ_PREV(bioq->switch_point,
569 * If we lie between last_offset and bq,
572 if (bio->bio_offset < bq->bio_offset) {
573 TAILQ_INSERT_BEFORE(bq, bio, bio_act);
580 * Request is at/after our current position in the list.
581 * Optimize for sequential I/O by seeing if we go at the tail.
583 if (bio->bio_offset > be->bio_offset) {
584 TAILQ_INSERT_AFTER(&bioq->queue, be, bio, bio_act);
588 /* Otherwise, insertion sort */
589 while ((bn = TAILQ_NEXT(bq, bio_act)) != NULL) {
592 * We want to go after the current request if it is the end
593 * of the first request list, or if the next request is a
594 * larger cylinder than our request.
596 if (bn == bioq->switch_point
597 || bio->bio_offset < bn->bio_offset)
601 TAILQ_INSERT_AFTER(&bioq->queue, bq, bio, bio_act);
605 * Disk error is the preface to plaintive error messages
606 * about failing disk transfers. It prints messages of the form
608 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
610 * if the offset of the error in the transfer and a disk label
611 * are both available. blkdone should be -1 if the position of the error
612 * is unknown; the disklabel pointer may be null from drivers that have not
613 * been converted to use them. The message is printed with kprintf
614 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
615 * The message should be completed (with at least a newline) with kprintf
616 * or log(-1, ...), respectively. There is no trailing space.
619 diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt)
621 struct buf *bp = bio->bio_buf;
622 int unit = dkunit(dev);
623 int slice = dkslice(dev);
624 int part = dkpart(dev);
640 sname = dsname(dev, unit, slice, part, partname);
641 kprintf("%s%s: %s %sing ", sname, partname, what, term);
642 kprintf("offset %012llx for %d", bio->bio_offset, bp->b_bcount);
644 kprintf(" (%d bytes completed)", donecnt);
648 * Locate a disk device
651 disk_locate(const char *devname)
664 for (i = 0; devname[i]; ++i) {
665 if (devname[i] >= '0' && devname[i] <= '9')
668 while (devname[i] >= '0' && devname[i] <= '9')
673 * Slice and partition. s1 starts at slice #2. s0 is slice #0.
674 * slice #1 is the WHOLE_DISK_SLICE.
676 if (devname[i] == 's') {
677 slice = strtol(devname + i + 1, &ptr, 10);
678 i = (const char *)ptr - devname;
682 slice = WHOLE_DISK_SLICE;
684 if (devname[i] >= 'a' && devname[i] <= 'z') {
685 part = devname[i] - 'a';
687 part = WHOLE_SLICE_PART;
693 LIST_FOREACH(dp, &disklist, d_list) {
695 if (strlen(dev->si_name) == prefix &&
696 strncmp(devname, dev->si_name, prefix) == 0
698 return(dkmodpart(dkmodslice(dev, slice), part));