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 1.25 2006/07/28 02:17:40 dillon 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/diskslice.h>
93 #include <sys/malloc.h>
94 #include <sys/sysctl.h>
95 #include <machine/md_var.h>
96 #include <sys/ctype.h>
97 #include <sys/syslog.h>
98 #include <sys/device.h>
99 #include <sys/msgport.h>
100 #include <sys/msgport2.h>
101 #include <sys/buf2.h>
103 static MALLOC_DEFINE(M_DISK, "disk", "disk data");
105 static d_open_t diskopen;
106 static d_close_t diskclose;
107 static d_ioctl_t diskioctl;
108 static d_strategy_t diskstrategy;
109 static d_psize_t diskpsize;
110 static d_clone_t diskclone;
111 static d_dump_t diskdump;
113 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
115 static struct dev_ops disk_ops = {
118 .d_close = diskclose,
120 .d_write = physwrite,
121 .d_ioctl = diskioctl,
122 .d_strategy = diskstrategy,
124 .d_psize = diskpsize,
129 * Create a raw device for the dev_ops template (which is returned). Also
130 * create a slice and unit managed disk and overload the user visible
131 * device space with it.
133 * NOTE: The returned raw device is NOT a slice and unit managed device.
134 * It is an actual raw device representing the raw disk as specified by
135 * the passed dev_ops. The disk layer not only returns such a raw device,
136 * it also uses it internally when passing (modified) commands through.
139 disk_create(int unit, struct disk *dp, int flags, struct dev_ops *raw_ops)
142 struct dev_ops *dev_ops;
145 * Create the raw backing device
147 compile_dev_ops(raw_ops);
148 rawdev = make_dev(raw_ops,
149 dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART),
150 UID_ROOT, GID_OPERATOR, 0640,
151 "%s%d", raw_ops->head.name, unit);
153 bzero(dp, sizeof(*dp));
156 * We install a custom cdevsw rather then the passed cdevsw,
157 * and save our disk structure in d_data so we can get at it easily
158 * without any complex cloning code.
160 dev_ops = dev_ops_add_override(rawdev, &disk_ops,
161 dkunitmask(), dkmakeunit(unit));
162 dev_ops->head.data = dp;
164 dp->d_rawdev = rawdev;
165 dp->d_raw_ops = raw_ops;
166 dp->d_dev_ops = dev_ops;
167 dp->d_cdev = make_dev(dev_ops,
168 dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART),
169 UID_ROOT, GID_OPERATOR, 0640,
170 "%s%d", dev_ops->head.name, unit);
172 dp->d_dsflags = flags;
173 LIST_INSERT_HEAD(&disklist, dp, d_list);
174 return (dp->d_rawdev);
178 * This routine is called when an adapter detaches. The higher level
179 * managed disk device is destroyed while the lower level raw device is
183 disk_destroy(struct disk *disk)
185 if (disk->d_dev_ops) {
186 dev_ops_remove(disk->d_dev_ops, dkunitmask(),
187 dkmakeunit(dkunit(disk->d_cdev)));
188 LIST_REMOVE(disk, d_list);
190 if (disk->d_raw_ops) {
191 destroy_all_devs(disk->d_raw_ops, dkunitmask(),
192 dkmakeunit(dkunit(disk->d_rawdev)));
194 bzero(disk, sizeof(*disk));
198 disk_dumpcheck(dev_t dev, u_int *count, u_int *blkno, u_int *secsize)
201 struct disklabel *dl;
209 dl = dsgetlabel(dev, dp->d_slice);
212 *count = Maxmem * (PAGE_SIZE / dl->d_secsize);
213 if (dumplo <= LABELSECTOR ||
214 (dumplo + *count > dl->d_partitions[dkpart(dev)].p_size))
216 boff = dl->d_partitions[dkpart(dev)].p_offset +
217 dp->d_slice->dss_slices[dkslice(dev)].ds_offset;
218 *blkno = boff + dumplo;
219 *secsize = dl->d_secsize;
225 disk_invalidate (struct disk *disk)
228 dsgone(&disk->d_slice);
232 disk_enumerate(struct disk *disk)
235 return (LIST_FIRST(&disklist));
237 return (LIST_NEXT(disk, d_list));
242 sysctl_disks(SYSCTL_HANDLER_ARGS)
250 while ((disk = disk_enumerate(disk))) {
252 error = SYSCTL_OUT(req, " ", 1);
258 error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
259 strlen(disk->d_rawdev->si_name));
263 error = SYSCTL_OUT(req, "", 1);
267 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, 0, NULL,
268 sysctl_disks, "A", "names of available disks");
271 * Open a disk device or partition.
275 diskopen(struct dev_open_args *ap)
277 dev_t dev = ap->a_head.a_dev;
282 * dp can't be NULL here XXX.
290 * Deal with open races
292 while (dp->d_flags & DISKFLAG_LOCK) {
293 dp->d_flags |= DISKFLAG_WANTED;
294 error = tsleep(dp, PCATCH, "diskopen", hz);
298 dp->d_flags |= DISKFLAG_LOCK;
301 * Open the underlying raw device.
303 if (!dsisopen(dp->d_slice)) {
305 if (!pdev->si_iosize_max)
306 pdev->si_iosize_max = dev->si_iosize_max;
308 error = dev_dopen(dp->d_rawdev, ap->a_oflags,
309 ap->a_devtype, ap->a_cred);
313 * Inherit properties from the underlying device now that it is
321 error = dsopen(dev, ap->a_devtype, dp->d_dsflags,
322 &dp->d_slice, &dp->d_label);
324 if (!dsisopen(dp->d_slice))
325 dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype);
327 dp->d_flags &= ~DISKFLAG_LOCK;
328 if (dp->d_flags & DISKFLAG_WANTED) {
329 dp->d_flags &= ~DISKFLAG_WANTED;
337 * Close a disk device or partition
341 diskclose(struct dev_close_args *ap)
343 dev_t dev = ap->a_head.a_dev;
350 dsclose(dev, ap->a_devtype, dp->d_slice);
351 if (!dsisopen(dp->d_slice))
352 error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype);
357 * First execute the ioctl on the disk device, and if it isn't supported
358 * try running it on the backing device.
362 diskioctl(struct dev_ioctl_args *ap)
364 dev_t dev = ap->a_head.a_dev;
371 error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag, &dp->d_slice);
372 if (error == ENOIOCTL) {
373 error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data,
374 ap->a_fflag, ap->a_cred);
380 * Execute strategy routine
384 diskstrategy(struct dev_strategy_args *ap)
386 dev_t dev = ap->a_head.a_dev;
387 struct bio *bio = ap->a_bio;
394 bio->bio_buf->b_error = ENXIO;
395 bio->bio_buf->b_flags |= B_ERROR;
399 KKASSERT(dev->si_disk == dp);
402 * The dscheck() function will also transform the slice relative
403 * block number i.e. bio->bio_offset into a block number that can be
404 * passed directly to the underlying raw device. If dscheck()
405 * returns NULL it will have handled the bio for us (e.g. EOF
406 * or error due to being beyond the device size).
408 if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL)
409 dev_dstrategy(dp->d_rawdev, nbio);
416 * Return the partition size in ?blocks?
420 diskpsize(struct dev_psize_args *ap)
422 dev_t dev = ap->a_head.a_dev;
428 ap->a_result = dssize(dev, &dp->d_slice);
433 * When new device entries are instantiated, make sure they inherit our
434 * si_disk structure and block and iosize limits from the raw device.
436 * This routine is always called synchronously in the context of the
439 * XXX The various io and block size constraints are not always initialized
440 * properly by devices.
444 diskclone(struct dev_clone_args *ap)
446 dev_t dev = ap->a_head.a_dev;
449 dp = dev->si_ops->head.data;
450 KKASSERT(dp != NULL);
452 dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
453 dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
454 dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
459 diskdump(struct dev_dump_args *ap)
461 dev_t dev = ap->a_head.a_dev;
462 struct disk *dp = dev->si_ops->head.data;
465 error = disk_dumpcheck(dev, &ap->a_count, &ap->a_blkno, &ap->a_secsize);
467 ap->a_head.a_dev = dp->d_rawdev;
468 error = dev_doperate(&ap->a_head);
475 SYSCTL_INT(_debug_sizeof, OID_AUTO, disklabel, CTLFLAG_RD,
476 0, sizeof(struct disklabel), "sizeof(struct disklabel)");
478 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
479 0, sizeof(struct diskslices), "sizeof(struct diskslices)");
481 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
482 0, sizeof(struct disk), "sizeof(struct disk)");
486 * Seek sort for disks.
488 * The bio_queue keep two queues, sorted in ascending block order. The first
489 * queue holds those requests which are positioned after the current block
490 * (in the first request); the second, which starts at queue->switch_point,
491 * holds requests which came in after their block number was passed. Thus
492 * we implement a one way scan, retracting after reaching the end of the drive
493 * to the first request on the second queue, at which time it becomes the
496 * A one-way scan is natural because of the way UNIX read-ahead blocks are
500 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
506 be = TAILQ_LAST(&bioq->queue, bio_queue);
508 * If the queue is empty or we are an
509 * ordered transaction, then it's easy.
511 if ((bq = bioq_first(bioq)) == NULL ||
512 (bio->bio_buf->b_flags & B_ORDERED) != 0) {
513 bioq_insert_tail(bioq, bio);
515 } else if (bioq->insert_point != NULL) {
518 * A certain portion of the list is
519 * "locked" to preserve ordering, so
520 * we can only insert after the insert
523 bq = bioq->insert_point;
527 * If we lie before the last removed (currently active)
528 * request, and are not inserting ourselves into the
529 * "locked" portion of the list, then we must add ourselves
530 * to the second request list.
532 if (bio->bio_offset < bioq->last_offset) {
533 bq = bioq->switch_point;
535 * If we are starting a new secondary list,
539 bioq->switch_point = bio;
540 bioq_insert_tail(bioq, bio);
544 * If we lie ahead of the current switch point,
545 * insert us before the switch point and move
548 if (bio->bio_offset < bq->bio_offset) {
549 bioq->switch_point = bio;
550 TAILQ_INSERT_BEFORE(bq, bio, bio_act);
554 if (bioq->switch_point != NULL)
555 be = TAILQ_PREV(bioq->switch_point,
558 * If we lie between last_offset and bq,
561 if (bio->bio_offset < bq->bio_offset) {
562 TAILQ_INSERT_BEFORE(bq, bio, bio_act);
569 * Request is at/after our current position in the list.
570 * Optimize for sequential I/O by seeing if we go at the tail.
572 if (bio->bio_offset > be->bio_offset) {
573 TAILQ_INSERT_AFTER(&bioq->queue, be, bio, bio_act);
577 /* Otherwise, insertion sort */
578 while ((bn = TAILQ_NEXT(bq, bio_act)) != NULL) {
581 * We want to go after the current request if it is the end
582 * of the first request list, or if the next request is a
583 * larger cylinder than our request.
585 if (bn == bioq->switch_point
586 || bio->bio_offset < bn->bio_offset)
590 TAILQ_INSERT_AFTER(&bioq->queue, bq, bio, bio_act);
595 * Attempt to read a disk label from a device using the indicated strategy
596 * routine. The label must be partly set up before this: secpercyl, secsize
597 * and anything required in the strategy routine (e.g., dummy bounds for the
598 * partition containing the label) must be filled in before calling us.
599 * Returns NULL on success and an error string on failure.
602 readdisklabel(dev_t dev, struct disklabel *lp)
605 struct disklabel *dlp;
608 bp = geteblk((int)lp->d_secsize);
609 bp->b_bio1.bio_offset = (off_t)LABELSECTOR * lp->d_secsize;
610 bp->b_bcount = lp->d_secsize;
611 bp->b_flags &= ~B_INVAL;
612 bp->b_cmd = BUF_CMD_READ;
613 dev_dstrategy(dev, &bp->b_bio1);
616 else for (dlp = (struct disklabel *)bp->b_data;
617 dlp <= (struct disklabel *)((char *)bp->b_data +
618 lp->d_secsize - sizeof(*dlp));
619 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
620 if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) {
622 msg = "no disk label";
623 } else if (dlp->d_npartitions > MAXPARTITIONS ||
625 msg = "disk label corrupted";
632 bp->b_flags |= B_INVAL | B_AGE;
638 * Check new disk label for sensibility before setting it.
641 setdisklabel(struct disklabel *olp, struct disklabel *nlp, u_long openmask)
644 struct partition *opp, *npp;
647 * Check it is actually a disklabel we are looking at.
649 if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC ||
653 * For each partition that we think is open,
655 while ((i = ffs((long)openmask)) != 0) {
658 * Check it is not changing....
660 openmask &= ~(1 << i);
661 if (nlp->d_npartitions <= i)
663 opp = &olp->d_partitions[i];
664 npp = &nlp->d_partitions[i];
665 if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size)
668 * Copy internally-set partition information
669 * if new label doesn't include it. XXX
670 * (If we are using it then we had better stay the same type)
671 * This is possibly dubious, as someone else noted (XXX)
673 if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) {
674 npp->p_fstype = opp->p_fstype;
675 npp->p_fsize = opp->p_fsize;
676 npp->p_frag = opp->p_frag;
677 npp->p_cpg = opp->p_cpg;
681 nlp->d_checksum = dkcksum(nlp);
687 * Write disk label back to device after modification.
690 writedisklabel(dev_t dev, struct disklabel *lp)
693 struct disklabel *dlp;
696 if (lp->d_partitions[RAW_PART].p_offset != 0)
697 return (EXDEV); /* not quite right */
698 bp = geteblk((int)lp->d_secsize);
699 bp->b_bio1.bio_offset = (off_t)LABELSECTOR * lp->d_secsize;
700 bp->b_bcount = lp->d_secsize;
703 * We read the label first to see if it's there,
704 * in which case we will put ours at the same offset into the block..
705 * (I think this is stupid [Julian])
706 * Note that you can't write a label out over a corrupted label!
707 * (also stupid.. how do you write the first one? by raw writes?)
709 bp->b_flags &= ~B_INVAL;
710 bp->b_cmd = BUF_CMD_READ;
711 dev_dstrategy(dkmodpart(dev, RAW_PART), &bp->b_bio1);
715 for (dlp = (struct disklabel *)bp->b_data;
716 dlp <= (struct disklabel *)
717 ((char *)bp->b_data + lp->d_secsize - sizeof(*dlp));
718 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
719 if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC &&
722 bp->b_cmd = BUF_CMD_WRITE;
723 dev_dstrategy(dkmodpart(dev, RAW_PART), &bp->b_bio1);
731 bzero(bp->b_data, lp->d_secsize);
732 dlp = (struct disklabel *)bp->b_data;
734 bp->b_flags &= ~B_INVAL;
735 bp->b_cmd = BUF_CMD_WRITE;
739 bp->b_flags |= B_INVAL | B_AGE;
745 * Disk error is the preface to plaintive error messages
746 * about failing disk transfers. It prints messages of the form
748 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
750 * if the offset of the error in the transfer and a disk label
751 * are both available. blkdone should be -1 if the position of the error
752 * is unknown; the disklabel pointer may be null from drivers that have not
753 * been converted to use them. The message is printed with printf
754 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
755 * The message should be completed (with at least a newline) with printf
756 * or addlog, respectively. There is no trailing space.
759 diskerr(struct bio *bio, dev_t dev, const char *what, int pri,
760 int donecnt, struct disklabel *lp)
762 struct buf *bp = bio->bio_buf;
763 int unit = dkunit(dev);
764 int slice = dkslice(dev);
765 int part = dkpart(dev);
769 sname = dsname(dev, unit, slice, part, partname);
770 printf("%s%s: %s %sing ", sname, partname, what,
771 (bp->b_cmd == BUF_CMD_READ) ? "read" : "writ");
772 printf("offset %012llx for %d", bio->bio_offset, bp->b_bcount);
774 printf(" (%d bytes completed)", donecnt);