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 1.34 2007/05/20 04:41:58 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, 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, dkmakewholedisk(unit),
149 UID_ROOT, GID_OPERATOR, 0640,
150 "%s%d", raw_ops->head.name, unit);
152 bzero(dp, sizeof(*dp));
155 * We install a custom cdevsw rather then the passed cdevsw,
156 * and save our disk structure in d_data so we can get at it easily
157 * without any complex cloning code.
159 dev_ops = dev_ops_add_override(rawdev, &disk_ops,
160 dkunitmask(), dkmakeunit(unit));
161 dev_ops->head.data = dp;
163 dp->d_rawdev = rawdev;
164 dp->d_raw_ops = raw_ops;
165 dp->d_dev_ops = dev_ops;
166 dp->d_cdev = make_dev(dev_ops,
167 dkmakewholedisk(unit),
168 UID_ROOT, GID_OPERATOR, 0640,
169 "%s%d", dev_ops->head.name, unit);
171 LIST_INSERT_HEAD(&disklist, dp, d_list);
172 return (dp->d_rawdev);
176 * Disk drivers must call this routine when media parameters are available
180 disk_setdiskinfo(struct disk *disk, struct disk_info *info)
182 bcopy(info, &disk->d_info, sizeof(disk->d_info));
183 info = &disk->d_info;
185 KKASSERT(info->d_media_size == 0 || info->d_media_blksize == 0);
186 if (info->d_media_size == 0 && info->d_media_blocks) {
187 info->d_media_size = (u_int64_t)info->d_media_blocks *
188 info->d_media_blksize;
189 } else if (info->d_media_size && info->d_media_blocks == 0 &&
190 info->d_media_blksize) {
191 info->d_media_blocks = info->d_media_size /
192 info->d_media_blksize;
197 * This routine is called when an adapter detaches. The higher level
198 * managed disk device is destroyed while the lower level raw device is
202 disk_destroy(struct disk *disk)
204 if (disk->d_dev_ops) {
205 dev_ops_remove(disk->d_dev_ops, dkunitmask(),
206 dkmakeunit(dkunit(disk->d_cdev)));
207 LIST_REMOVE(disk, d_list);
209 if (disk->d_raw_ops) {
210 destroy_all_devs(disk->d_raw_ops, dkunitmask(),
211 dkmakeunit(dkunit(disk->d_rawdev)));
213 bzero(disk, sizeof(*disk));
217 disk_dumpcheck(cdev_t dev, u_int64_t *count, u_int64_t *blkno, u_int *secsize)
219 struct partinfo pinfo;
222 bzero(&pinfo, sizeof(pinfo));
223 error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0, proc0.p_ucred);
226 if (pinfo.media_blksize == 0)
228 *count = (u_int64_t)Maxmem * PAGE_SIZE / pinfo.media_blksize;
229 if (dumplo64 < pinfo.skip_bsdlabel ||
230 dumplo64 + *count > pinfo.media_blocks) {
233 *blkno = dumplo64 + pinfo.media_offset / pinfo.media_blksize;
234 *secsize = pinfo.media_blksize;
239 disk_invalidate (struct disk *disk)
242 dsgone(&disk->d_slice);
246 disk_enumerate(struct disk *disk)
249 return (LIST_FIRST(&disklist));
251 return (LIST_NEXT(disk, d_list));
256 sysctl_disks(SYSCTL_HANDLER_ARGS)
264 while ((disk = disk_enumerate(disk))) {
266 error = SYSCTL_OUT(req, " ", 1);
272 error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
273 strlen(disk->d_rawdev->si_name));
277 error = SYSCTL_OUT(req, "", 1);
281 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, 0, NULL,
282 sysctl_disks, "A", "names of available disks");
285 * Open a disk device or partition.
289 diskopen(struct dev_open_args *ap)
291 cdev_t dev = ap->a_head.a_dev;
296 * dp can't be NULL here XXX.
304 * Deal with open races
306 while (dp->d_flags & DISKFLAG_LOCK) {
307 dp->d_flags |= DISKFLAG_WANTED;
308 error = tsleep(dp, PCATCH, "diskopen", hz);
312 dp->d_flags |= DISKFLAG_LOCK;
315 * Open the underlying raw device.
317 if (!dsisopen(dp->d_slice)) {
319 if (!pdev->si_iosize_max)
320 pdev->si_iosize_max = dev->si_iosize_max;
322 error = dev_dopen(dp->d_rawdev, ap->a_oflags,
323 ap->a_devtype, ap->a_cred);
327 * Inherit properties from the underlying device now that it is
335 error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags,
336 &dp->d_slice, &dp->d_info);
338 if (!dsisopen(dp->d_slice))
339 dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype);
341 dp->d_flags &= ~DISKFLAG_LOCK;
342 if (dp->d_flags & DISKFLAG_WANTED) {
343 dp->d_flags &= ~DISKFLAG_WANTED;
351 * Close a disk device or partition
355 diskclose(struct dev_close_args *ap)
357 cdev_t dev = ap->a_head.a_dev;
364 dsclose(dev, ap->a_devtype, dp->d_slice);
365 if (!dsisopen(dp->d_slice))
366 error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype);
371 * First execute the ioctl on the disk device, and if it isn't supported
372 * try running it on the backing device.
376 diskioctl(struct dev_ioctl_args *ap)
378 cdev_t dev = ap->a_head.a_dev;
385 error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag,
386 &dp->d_slice, &dp->d_info);
387 if (error == ENOIOCTL) {
388 error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data,
389 ap->a_fflag, ap->a_cred);
395 * Execute strategy routine
399 diskstrategy(struct dev_strategy_args *ap)
401 cdev_t dev = ap->a_head.a_dev;
402 struct bio *bio = ap->a_bio;
409 bio->bio_buf->b_error = ENXIO;
410 bio->bio_buf->b_flags |= B_ERROR;
414 KKASSERT(dev->si_disk == dp);
417 * The dscheck() function will also transform the slice relative
418 * block number i.e. bio->bio_offset into a block number that can be
419 * passed directly to the underlying raw device. If dscheck()
420 * returns NULL it will have handled the bio for us (e.g. EOF
421 * or error due to being beyond the device size).
423 if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL)
424 dev_dstrategy(dp->d_rawdev, nbio);
431 * Return the partition size in ?blocks?
435 diskpsize(struct dev_psize_args *ap)
437 cdev_t dev = ap->a_head.a_dev;
443 ap->a_result = dssize(dev, &dp->d_slice);
448 * When new device entries are instantiated, make sure they inherit our
449 * si_disk structure and block and iosize limits from the raw device.
451 * This routine is always called synchronously in the context of the
454 * XXX The various io and block size constraints are not always initialized
455 * properly by devices.
459 diskclone(struct dev_clone_args *ap)
461 cdev_t dev = ap->a_head.a_dev;
464 dp = dev->si_ops->head.data;
465 KKASSERT(dp != NULL);
467 dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
468 dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
469 dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
474 diskdump(struct dev_dump_args *ap)
476 cdev_t dev = ap->a_head.a_dev;
477 struct disk *dp = dev->si_ops->head.data;
480 error = disk_dumpcheck(dev, &ap->a_count, &ap->a_blkno, &ap->a_secsize);
482 ap->a_head.a_dev = dp->d_rawdev;
483 error = dev_doperate(&ap->a_head);
490 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
491 0, sizeof(struct diskslices), "sizeof(struct diskslices)");
493 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
494 0, sizeof(struct disk), "sizeof(struct disk)");
498 * Seek sort for disks.
500 * The bio_queue keep two queues, sorted in ascending block order. The first
501 * queue holds those requests which are positioned after the current block
502 * (in the first request); the second, which starts at queue->switch_point,
503 * holds requests which came in after their block number was passed. Thus
504 * we implement a one way scan, retracting after reaching the end of the drive
505 * to the first request on the second queue, at which time it becomes the
508 * A one-way scan is natural because of the way UNIX read-ahead blocks are
512 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
518 be = TAILQ_LAST(&bioq->queue, bio_queue);
520 * If the queue is empty or we are an
521 * ordered transaction, then it's easy.
523 if ((bq = bioq_first(bioq)) == NULL ||
524 (bio->bio_buf->b_flags & B_ORDERED) != 0) {
525 bioq_insert_tail(bioq, bio);
527 } else if (bioq->insert_point != NULL) {
530 * A certain portion of the list is
531 * "locked" to preserve ordering, so
532 * we can only insert after the insert
535 bq = bioq->insert_point;
539 * If we lie before the last removed (currently active)
540 * request, and are not inserting ourselves into the
541 * "locked" portion of the list, then we must add ourselves
542 * to the second request list.
544 if (bio->bio_offset < bioq->last_offset) {
545 bq = bioq->switch_point;
547 * If we are starting a new secondary list,
551 bioq->switch_point = bio;
552 bioq_insert_tail(bioq, bio);
556 * If we lie ahead of the current switch point,
557 * insert us before the switch point and move
560 if (bio->bio_offset < bq->bio_offset) {
561 bioq->switch_point = bio;
562 TAILQ_INSERT_BEFORE(bq, bio, bio_act);
566 if (bioq->switch_point != NULL)
567 be = TAILQ_PREV(bioq->switch_point,
570 * If we lie between last_offset and bq,
573 if (bio->bio_offset < bq->bio_offset) {
574 TAILQ_INSERT_BEFORE(bq, bio, bio_act);
581 * Request is at/after our current position in the list.
582 * Optimize for sequential I/O by seeing if we go at the tail.
584 if (bio->bio_offset > be->bio_offset) {
585 TAILQ_INSERT_AFTER(&bioq->queue, be, bio, bio_act);
589 /* Otherwise, insertion sort */
590 while ((bn = TAILQ_NEXT(bq, bio_act)) != NULL) {
593 * We want to go after the current request if it is the end
594 * of the first request list, or if the next request is a
595 * larger cylinder than our request.
597 if (bn == bioq->switch_point
598 || bio->bio_offset < bn->bio_offset)
602 TAILQ_INSERT_AFTER(&bioq->queue, bq, bio, bio_act);
607 * Attempt to read a disk label from a device using the indicated strategy
608 * routine. The label must be partly set up before this: secpercyl, secsize
609 * and anything required in the strategy routine (e.g., dummy bounds for the
610 * partition containing the label) must be filled in before calling us.
611 * Returns NULL on success and an error string on failure.
614 readdisklabel(cdev_t dev, struct disklabel *lp)
617 struct disklabel *dlp;
620 bp = geteblk((int)lp->d_secsize);
621 bp->b_bio1.bio_offset = (off_t)LABELSECTOR * lp->d_secsize;
622 bp->b_bcount = lp->d_secsize;
623 bp->b_flags &= ~B_INVAL;
624 bp->b_cmd = BUF_CMD_READ;
625 dev_dstrategy(dev, &bp->b_bio1);
628 else for (dlp = (struct disklabel *)bp->b_data;
629 dlp <= (struct disklabel *)((char *)bp->b_data +
630 lp->d_secsize - sizeof(*dlp));
631 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
632 if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) {
634 msg = "no disk label";
635 } else if (dlp->d_npartitions > MAXPARTITIONS ||
637 msg = "disk label corrupted";
644 bp->b_flags |= B_INVAL | B_AGE;
650 * Check new disk label for sensibility before setting it.
653 setdisklabel(struct disklabel *olp, struct disklabel *nlp, u_long openmask)
656 struct partition *opp, *npp;
659 * Check it is actually a disklabel we are looking at.
661 if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC ||
665 * For each partition that we think is open,
667 while ((i = ffs((long)openmask)) != 0) {
670 * Check it is not changing....
672 openmask &= ~(1 << i);
673 if (nlp->d_npartitions <= i)
675 opp = &olp->d_partitions[i];
676 npp = &nlp->d_partitions[i];
677 if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size)
680 * Copy internally-set partition information
681 * if new label doesn't include it. XXX
682 * (If we are using it then we had better stay the same type)
683 * This is possibly dubious, as someone else noted (XXX)
685 if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) {
686 npp->p_fstype = opp->p_fstype;
687 npp->p_fsize = opp->p_fsize;
688 npp->p_frag = opp->p_frag;
689 npp->p_cpg = opp->p_cpg;
693 nlp->d_checksum = dkcksum(nlp);
699 * Write disk label back to device after modification.
702 writedisklabel(cdev_t dev, struct disklabel *lp)
705 struct disklabel *dlp;
708 if (lp->d_partitions[RAW_PART].p_offset != 0)
709 return (EXDEV); /* not quite right */
710 bp = geteblk((int)lp->d_secsize);
711 bp->b_bio1.bio_offset = (off_t)LABELSECTOR * lp->d_secsize;
712 bp->b_bcount = lp->d_secsize;
715 * We read the label first to see if it's there,
716 * in which case we will put ours at the same offset into the block..
717 * (I think this is stupid [Julian])
718 * Note that you can't write a label out over a corrupted label!
719 * (also stupid.. how do you write the first one? by raw writes?)
721 bp->b_flags &= ~B_INVAL;
722 bp->b_cmd = BUF_CMD_READ;
723 dev_dstrategy(dkmodpart(dev, WHOLE_SLICE_PART), &bp->b_bio1);
727 for (dlp = (struct disklabel *)bp->b_data;
728 dlp <= (struct disklabel *)
729 ((char *)bp->b_data + lp->d_secsize - sizeof(*dlp));
730 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
731 if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC &&
734 bp->b_cmd = BUF_CMD_WRITE;
735 dev_dstrategy(dkmodpart(dev, WHOLE_SLICE_PART), &bp->b_bio1);
743 bzero(bp->b_data, lp->d_secsize);
744 dlp = (struct disklabel *)bp->b_data;
746 bp->b_flags &= ~B_INVAL;
747 bp->b_cmd = BUF_CMD_WRITE;
751 bp->b_flags |= B_INVAL | B_AGE;
757 * Disk error is the preface to plaintive error messages
758 * about failing disk transfers. It prints messages of the form
760 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
762 * if the offset of the error in the transfer and a disk label
763 * are both available. blkdone should be -1 if the position of the error
764 * is unknown; the disklabel pointer may be null from drivers that have not
765 * been converted to use them. The message is printed with kprintf
766 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
767 * The message should be completed (with at least a newline) with kprintf
768 * or log(-1, ...), respectively. There is no trailing space.
771 diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt)
773 struct buf *bp = bio->bio_buf;
774 int unit = dkunit(dev);
775 int slice = dkslice(dev);
776 int part = dkpart(dev);
792 sname = dsname(dev, unit, slice, part, partname);
793 kprintf("%s%s: %s %sing ", sname, partname, what, term);
794 kprintf("offset %012llx for %d", bio->bio_offset, bp->b_bcount);
796 kprintf(" (%d bytes completed)", donecnt);