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 1.23 2006/04/30 17:22:17 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_strategy_t diskstrategy;
106 static d_open_t diskopen;
107 static d_close_t diskclose;
108 static d_ioctl_t diskioctl;
109 static d_psize_t diskpsize;
110 static d_clone_t diskclone;
111 static int disk_putport(lwkt_port_t port, lwkt_msg_t msg);
113 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
116 * Create a slice and unit managed disk.
118 * Our port layer will be responsible for assigning blkno and handling
119 * high level partition operations, then forwarding the requests to the
122 * The disk_create() function clones the provided rawsw for creating a
123 * managed disk device. In addition, the cdevsw intercept port is
124 * changed to disk_putport, which is used to transform requests for the
125 * managed disk device.
127 * The raw device (based on rawsw) is returned to the caller, NOT the
128 * slice and unit managed cdev. The caller typically sets various
129 * driver parameters and IO limits on the returned rawdev which we must
130 * inherit when our managed device is opened.
133 disk_create(int unit, struct disk *dp, int flags, struct cdevsw *rawsw)
136 struct cdevsw *devsw;
139 * Create the raw backing device
141 compile_devsw(rawsw);
142 rawdev = make_dev(rawsw, dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART),
143 UID_ROOT, GID_OPERATOR, 0640,
144 "%s%d", rawsw->d_name, unit);
147 * Initialize our intercept port
149 bzero(dp, sizeof(*dp));
150 lwkt_initport(&dp->d_port, NULL);
151 dp->d_port.mp_putport = disk_putport;
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 devsw = cdevsw_add_override(rawdev, dkunitmask(), dkmakeunit(unit));
160 devsw->d_port = &dp->d_port;
162 devsw->d_clone = diskclone;
164 dp->d_rawdev = rawdev;
165 dp->d_cdev = make_dev(devsw,
166 dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART),
167 UID_ROOT, GID_OPERATOR, 0640,
168 "%s%d", devsw->d_name, unit);
170 dp->d_dsflags = flags;
171 LIST_INSERT_HEAD(&disklist, dp, d_list);
172 return (dp->d_rawdev);
176 * This routine is called when an adapter detaches. The higher level
177 * managed disk device is destroyed while the lower level raw device is
181 disk_destroy(struct disk *disk)
184 cdevsw_remove(disk->d_devsw, dkunitmask(),
185 dkmakeunit(dkunit(disk->d_cdev)));
186 LIST_REMOVE(disk, d_list);
189 destroy_all_dev(disk->d_rawsw, dkunitmask(),
190 dkmakeunit(dkunit(disk->d_rawdev)));
192 bzero(disk, sizeof(*disk));
196 disk_dumpcheck(dev_t dev, u_int *count, u_int *blkno, u_int *secsize)
199 struct disklabel *dl;
207 dl = dsgetlabel(dev, dp->d_slice);
210 *count = Maxmem * (PAGE_SIZE / dl->d_secsize);
211 if (dumplo <= LABELSECTOR ||
212 (dumplo + *count > dl->d_partitions[dkpart(dev)].p_size))
214 boff = dl->d_partitions[dkpart(dev)].p_offset +
215 dp->d_slice->dss_slices[dkslice(dev)].ds_offset;
216 *blkno = boff + dumplo;
217 *secsize = dl->d_secsize;
223 disk_invalidate (struct disk *disk)
226 dsgone(&disk->d_slice);
230 disk_enumerate(struct disk *disk)
233 return (LIST_FIRST(&disklist));
235 return (LIST_NEXT(disk, d_list));
240 sysctl_disks(SYSCTL_HANDLER_ARGS)
248 while ((disk = disk_enumerate(disk))) {
250 error = SYSCTL_OUT(req, " ", 1);
256 error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
257 strlen(disk->d_rawdev->si_name));
261 error = SYSCTL_OUT(req, "", 1);
265 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, 0, NULL,
266 sysctl_disks, "A", "names of available disks");
269 * The port intercept functions
273 disk_putport(lwkt_port_t port, lwkt_msg_t lmsg)
275 struct disk *disk = (struct disk *)port;
276 cdevallmsg_t msg = (cdevallmsg_t)lmsg;
279 switch(msg->am_lmsg.ms_cmd.cm_op) {
282 msg->am_open.msg.dev,
284 msg->am_open.devtype,
289 msg->am_close.msg.dev,
291 msg->am_close.devtype,
296 msg->am_ioctl.msg.dev,
302 case CDEV_CMD_STRATEGY:
303 diskstrategy(msg->am_strategy.msg.dev, msg->am_strategy.bio);
307 msg->am_psize.result = diskpsize(msg->am_psize.msg.dev);
311 error = physio(msg->am_read.msg.dev,
312 msg->am_read.uio, msg->am_read.ioflag);
315 error = physio(msg->am_write.msg.dev,
316 msg->am_write.uio, msg->am_write.ioflag);
319 case CDEV_CMD_KQFILTER:
325 error = disk_dumpcheck(msg->am_dump.msg.dev,
328 &msg->am_dump.secsize);
330 msg->am_dump.msg.dev = disk->d_rawdev;
331 error = lwkt_forwardmsg(disk->d_rawdev->si_port,
332 &msg->am_dump.msg.msg);
333 printf("error2 %d\n", error);
344 * When new device entries are instantiated, make sure they inherit our
345 * si_disk structure and block and iosize limits from the raw device.
347 * This routine is always called synchronously in the context of the
350 * XXX The various io and block size constraints are not always initialized
351 * properly by devices.
359 dp = dev->si_devsw->d_data;
360 KKASSERT(dp != NULL);
362 dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
363 dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
364 dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
369 * Open a disk device or partition.
373 diskopen(dev_t dev, int oflags, int devtype, struct thread *td)
379 * dp can't be NULL here XXX.
387 * Deal with open races
389 while (dp->d_flags & DISKFLAG_LOCK) {
390 dp->d_flags |= DISKFLAG_WANTED;
391 error = tsleep(dp, PCATCH, "diskopen", hz);
395 dp->d_flags |= DISKFLAG_LOCK;
398 * Open the underlying raw device.
400 if (!dsisopen(dp->d_slice)) {
402 if (!pdev->si_iosize_max)
403 pdev->si_iosize_max = dev->si_iosize_max;
405 error = dev_dopen(dp->d_rawdev, oflags, devtype, td);
409 * Inherit properties from the underlying device now that it is
417 error = dsopen(dev, devtype, dp->d_dsflags, &dp->d_slice, &dp->d_label);
419 if (!dsisopen(dp->d_slice))
420 dev_dclose(dp->d_rawdev, oflags, devtype, td);
422 dp->d_flags &= ~DISKFLAG_LOCK;
423 if (dp->d_flags & DISKFLAG_WANTED) {
424 dp->d_flags &= ~DISKFLAG_WANTED;
432 * Close a disk device or partition
436 diskclose(dev_t dev, int fflag, int devtype, struct thread *td)
444 dsclose(dev, devtype, dp->d_slice);
445 if (!dsisopen(dp->d_slice))
446 error = dev_dclose(dp->d_rawdev, fflag, devtype, td);
451 * Execute strategy routine
455 diskstrategy(dev_t dev, struct bio *bio)
463 bio->bio_buf->b_error = ENXIO;
464 bio->bio_buf->b_flags |= B_ERROR;
468 KKASSERT(dev->si_disk == dp);
471 * The dscheck() function will also transform the slice relative
472 * block number i.e. bio->bio_offset into a block number that can be
473 * passed directly to the underlying raw device.
475 nbio = dscheck(dev, bio, dp->d_slice);
480 dev_dstrategy(dp->d_rawdev, nbio);
484 * First execute the ioctl on the disk device, and if it isn't supported
485 * try running it on the backing device.
489 diskioctl(dev_t dev, u_long cmd, caddr_t data, int fflag, struct thread *td)
498 error = dsioctl(dev, cmd, data, fflag, &dp->d_slice);
499 if (error == ENOIOCTL)
500 error = dev_dioctl(dp->d_rawdev, cmd, data, fflag, td);
516 return(dssize(dev, &dp->d_slice));
518 if (dp != dev->si_disk) {
519 dev->si_drv1 = pdev->si_drv1;
520 dev->si_drv2 = pdev->si_drv2;
521 /* XXX: don't set bp->b_dev->si_disk (?) */
526 SYSCTL_INT(_debug_sizeof, OID_AUTO, disklabel, CTLFLAG_RD,
527 0, sizeof(struct disklabel), "sizeof(struct disklabel)");
529 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
530 0, sizeof(struct diskslices), "sizeof(struct diskslices)");
532 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
533 0, sizeof(struct disk), "sizeof(struct disk)");
537 * Seek sort for disks.
539 * The bio_queue keep two queues, sorted in ascending block order. The first
540 * queue holds those requests which are positioned after the current block
541 * (in the first request); the second, which starts at queue->switch_point,
542 * holds requests which came in after their block number was passed. Thus
543 * we implement a one way scan, retracting after reaching the end of the drive
544 * to the first request on the second queue, at which time it becomes the
547 * A one-way scan is natural because of the way UNIX read-ahead blocks are
551 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
557 be = TAILQ_LAST(&bioq->queue, bio_queue);
559 * If the queue is empty or we are an
560 * ordered transaction, then it's easy.
562 if ((bq = bioq_first(bioq)) == NULL ||
563 (bio->bio_buf->b_flags & B_ORDERED) != 0) {
564 bioq_insert_tail(bioq, bio);
566 } else if (bioq->insert_point != NULL) {
569 * A certain portion of the list is
570 * "locked" to preserve ordering, so
571 * we can only insert after the insert
574 bq = bioq->insert_point;
578 * If we lie before the last removed (currently active)
579 * request, and are not inserting ourselves into the
580 * "locked" portion of the list, then we must add ourselves
581 * to the second request list.
583 if (bio->bio_offset < bioq->last_offset) {
584 bq = bioq->switch_point;
586 * If we are starting a new secondary list,
590 bioq->switch_point = bio;
591 bioq_insert_tail(bioq, bio);
595 * If we lie ahead of the current switch point,
596 * insert us before the switch point and move
599 if (bio->bio_offset < bq->bio_offset) {
600 bioq->switch_point = bio;
601 TAILQ_INSERT_BEFORE(bq, bio, bio_act);
605 if (bioq->switch_point != NULL)
606 be = TAILQ_PREV(bioq->switch_point,
609 * If we lie between last_offset and bq,
612 if (bio->bio_offset < bq->bio_offset) {
613 TAILQ_INSERT_BEFORE(bq, bio, bio_act);
620 * Request is at/after our current position in the list.
621 * Optimize for sequential I/O by seeing if we go at the tail.
623 if (bio->bio_offset > be->bio_offset) {
624 TAILQ_INSERT_AFTER(&bioq->queue, be, bio, bio_act);
628 /* Otherwise, insertion sort */
629 while ((bn = TAILQ_NEXT(bq, bio_act)) != NULL) {
632 * We want to go after the current request if it is the end
633 * of the first request list, or if the next request is a
634 * larger cylinder than our request.
636 if (bn == bioq->switch_point
637 || bio->bio_offset < bn->bio_offset)
641 TAILQ_INSERT_AFTER(&bioq->queue, bq, bio, bio_act);
646 * Attempt to read a disk label from a device using the indicated strategy
647 * routine. The label must be partly set up before this: secpercyl, secsize
648 * and anything required in the strategy routine (e.g., dummy bounds for the
649 * partition containing the label) must be filled in before calling us.
650 * Returns NULL on success and an error string on failure.
653 readdisklabel(dev_t dev, struct disklabel *lp)
656 struct disklabel *dlp;
659 bp = geteblk((int)lp->d_secsize);
660 bp->b_bio1.bio_offset = (off_t)LABELSECTOR * lp->d_secsize;
661 bp->b_bcount = lp->d_secsize;
662 bp->b_flags &= ~B_INVAL;
663 bp->b_cmd = BUF_CMD_READ;
664 dev_dstrategy(dev, &bp->b_bio1);
667 else for (dlp = (struct disklabel *)bp->b_data;
668 dlp <= (struct disklabel *)((char *)bp->b_data +
669 lp->d_secsize - sizeof(*dlp));
670 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
671 if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) {
673 msg = "no disk label";
674 } else if (dlp->d_npartitions > MAXPARTITIONS ||
676 msg = "disk label corrupted";
683 bp->b_flags |= B_INVAL | B_AGE;
689 * Check new disk label for sensibility before setting it.
692 setdisklabel(struct disklabel *olp, struct disklabel *nlp, u_long openmask)
695 struct partition *opp, *npp;
698 * Check it is actually a disklabel we are looking at.
700 if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC ||
704 * For each partition that we think is open,
706 while ((i = ffs((long)openmask)) != 0) {
709 * Check it is not changing....
711 openmask &= ~(1 << i);
712 if (nlp->d_npartitions <= i)
714 opp = &olp->d_partitions[i];
715 npp = &nlp->d_partitions[i];
716 if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size)
719 * Copy internally-set partition information
720 * if new label doesn't include it. XXX
721 * (If we are using it then we had better stay the same type)
722 * This is possibly dubious, as someone else noted (XXX)
724 if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) {
725 npp->p_fstype = opp->p_fstype;
726 npp->p_fsize = opp->p_fsize;
727 npp->p_frag = opp->p_frag;
728 npp->p_cpg = opp->p_cpg;
732 nlp->d_checksum = dkcksum(nlp);
738 * Write disk label back to device after modification.
741 writedisklabel(dev_t dev, struct disklabel *lp)
744 struct disklabel *dlp;
747 if (lp->d_partitions[RAW_PART].p_offset != 0)
748 return (EXDEV); /* not quite right */
749 bp = geteblk((int)lp->d_secsize);
750 bp->b_bio1.bio_offset = (off_t)LABELSECTOR * lp->d_secsize;
751 bp->b_bcount = lp->d_secsize;
754 * We read the label first to see if it's there,
755 * in which case we will put ours at the same offset into the block..
756 * (I think this is stupid [Julian])
757 * Note that you can't write a label out over a corrupted label!
758 * (also stupid.. how do you write the first one? by raw writes?)
760 bp->b_flags &= ~B_INVAL;
761 bp->b_cmd = BUF_CMD_READ;
762 dev_dstrategy(dkmodpart(dev, RAW_PART), &bp->b_bio1);
766 for (dlp = (struct disklabel *)bp->b_data;
767 dlp <= (struct disklabel *)
768 ((char *)bp->b_data + lp->d_secsize - sizeof(*dlp));
769 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
770 if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC &&
773 bp->b_cmd = BUF_CMD_WRITE;
774 dev_dstrategy(dkmodpart(dev, RAW_PART), &bp->b_bio1);
782 bzero(bp->b_data, lp->d_secsize);
783 dlp = (struct disklabel *)bp->b_data;
785 bp->b_flags &= ~B_INVAL;
786 bp->b_cmd = BUF_CMD_WRITE;
790 bp->b_flags |= B_INVAL | B_AGE;
796 * Disk error is the preface to plaintive error messages
797 * about failing disk transfers. It prints messages of the form
799 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
801 * if the offset of the error in the transfer and a disk label
802 * are both available. blkdone should be -1 if the position of the error
803 * is unknown; the disklabel pointer may be null from drivers that have not
804 * been converted to use them. The message is printed with printf
805 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
806 * The message should be completed (with at least a newline) with printf
807 * or addlog, respectively. There is no trailing space.
810 diskerr(struct bio *bio, dev_t dev, const char *what, int pri,
811 int donecnt, struct disklabel *lp)
813 struct buf *bp = bio->bio_buf;
814 int unit = dkunit(dev);
815 int slice = dkslice(dev);
816 int part = dkpart(dev);
820 sname = dsname(dev, unit, slice, part, partname);
821 printf("%s%s: %s %sing ", sname, partname, what,
822 (bp->b_cmd == BUF_CMD_READ) ? "read" : "writ");
823 printf("offset %012llx for %d", bio->bio_offset, bp->b_bcount);
825 printf(" (%d bytes completed)", donecnt);