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 1.15 2005/04/30 23:04:21 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/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 pblkno and handling
119 * high level partition operations, then forwarding the requests to the
122 * The raw device (based on rawsw) is returned to the caller, NOT the
123 * slice and unit managed cdev. The caller typically sets various
124 * driver parameters and IO limits on the returned rawdev which we must
125 * inherit when our managed device is opened.
128 disk_create(int unit, struct disk *dp, int flags, struct cdevsw *rawsw)
131 struct cdevsw *devsw;
134 * Create the raw backing device
136 compile_devsw(rawsw);
137 rawdev = make_dev(rawsw, dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART),
138 UID_ROOT, GID_OPERATOR, 0640,
139 "%s%d", rawsw->d_name, unit);
142 * Initialize our intercept port
144 bzero(dp, sizeof(*dp));
145 lwkt_initport(&dp->d_port, NULL);
146 dp->d_port.mp_putport = disk_putport;
150 * We install a custom cdevsw rather then the passed cdevsw,
151 * and save our disk structure in d_data so we can get at it easily
152 * without any complex cloning code.
154 devsw = cdevsw_add_override(rawdev, dkunitmask(), dkmakeunit(unit));
155 devsw->d_port = &dp->d_port;
157 devsw->d_clone = diskclone;
159 dp->d_rawdev = rawdev;
160 dp->d_cdev = make_dev(devsw,
161 dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART),
162 UID_ROOT, GID_OPERATOR, 0640,
163 "%s%d", devsw->d_name, unit);
165 dp->d_dsflags = flags;
166 LIST_INSERT_HEAD(&disklist, dp, d_list);
167 return (dp->d_rawdev);
171 * This routine is called when an adapter detaches. The higher level
172 * managed disk device is destroyed while the lower level raw device is
176 disk_destroy(struct disk *disk)
179 cdevsw_remove(disk->d_devsw, dkunitmask(),
180 dkmakeunit(dkunit(disk->d_cdev)));
181 LIST_REMOVE(disk, d_list);
184 destroy_all_dev(disk->d_rawsw, dkunitmask(),
185 dkmakeunit(dkunit(disk->d_rawdev)));
187 bzero(disk, sizeof(*disk));
191 disk_dumpcheck(dev_t dev, u_int *count, u_int *blkno, u_int *secsize)
194 struct disklabel *dl;
202 dl = dsgetlabel(dev, dp->d_slice);
205 *count = Maxmem * (PAGE_SIZE / dl->d_secsize);
206 if (dumplo <= LABELSECTOR ||
207 (dumplo + *count > dl->d_partitions[dkpart(dev)].p_size))
209 boff = dl->d_partitions[dkpart(dev)].p_offset +
210 dp->d_slice->dss_slices[dkslice(dev)].ds_offset;
211 *blkno = boff + dumplo;
212 *secsize = dl->d_secsize;
218 disk_invalidate (struct disk *disk)
221 dsgone(&disk->d_slice);
225 disk_enumerate(struct disk *disk)
228 return (LIST_FIRST(&disklist));
230 return (LIST_NEXT(disk, d_list));
235 sysctl_disks(SYSCTL_HANDLER_ARGS)
243 while ((disk = disk_enumerate(disk))) {
245 error = SYSCTL_OUT(req, " ", 1);
251 error = SYSCTL_OUT(req, disk->d_rawdev->si_name, strlen(disk->d_rawdev->si_name));
255 error = SYSCTL_OUT(req, "", 1);
259 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, 0, NULL,
260 sysctl_disks, "A", "names of available disks");
263 * The port intercept functions
267 disk_putport(lwkt_port_t port, lwkt_msg_t lmsg)
269 struct disk *disk = (struct disk *)port;
270 cdevallmsg_t msg = (cdevallmsg_t)lmsg;
273 switch(msg->am_lmsg.ms_cmd.cm_op) {
276 msg->am_open.msg.dev,
278 msg->am_open.devtype,
283 msg->am_close.msg.dev,
285 msg->am_close.devtype,
290 msg->am_ioctl.msg.dev,
296 case CDEV_CMD_STRATEGY:
297 diskstrategy(msg->am_strategy.bp);
301 msg->am_psize.result = diskpsize(msg->am_psize.msg.dev);
305 error = physio(msg->am_read.msg.dev,
306 msg->am_read.uio, msg->am_read.ioflag);
309 error = physio(msg->am_write.msg.dev,
310 msg->am_write.uio, msg->am_write.ioflag);
313 case CDEV_CMD_KQFILTER:
319 error = disk_dumpcheck(msg->am_dump.msg.dev,
322 &msg->am_dump.secsize);
324 msg->am_dump.msg.dev = disk->d_rawdev;
325 error = lwkt_forwardmsg(disk->d_rawdev->si_port,
326 &msg->am_dump.msg.msg);
327 printf("error2 %d\n", error);
338 * When new device entries are instantiated, make sure they inherit our
339 * si_disk structure and block and iosize limits from the raw device.
341 * This routine is always called synchronously in the context of the
344 * XXX The various io and block size constraints are not always initialized
345 * properly by devices.
353 dp = dev->si_devsw->d_data;
354 KKASSERT(dp != NULL);
356 dev->si_iosize_max = dp->d_rawdev->si_iosize_max;
357 dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys;
358 dev->si_bsize_best = dp->d_rawdev->si_bsize_best;
363 * Open a disk device or partition.
367 diskopen(dev_t dev, int oflags, int devtype, struct thread *td)
373 * dp can't be NULL here XXX.
381 * Deal with open races
383 while (dp->d_flags & DISKFLAG_LOCK) {
384 dp->d_flags |= DISKFLAG_WANTED;
385 error = tsleep(dp, PCATCH, "diskopen", hz);
389 dp->d_flags |= DISKFLAG_LOCK;
392 * Open the underlying raw device.
394 if (!dsisopen(dp->d_slice)) {
396 if (!pdev->si_iosize_max)
397 pdev->si_iosize_max = dev->si_iosize_max;
399 error = dev_dopen(dp->d_rawdev, oflags, devtype, td);
403 * Inherit properties from the underlying device now that it is
411 error = dsopen(dev, devtype, dp->d_dsflags, &dp->d_slice, &dp->d_label);
413 if (!dsisopen(dp->d_slice))
414 dev_dclose(dp->d_rawdev, oflags, devtype, td);
416 dp->d_flags &= ~DISKFLAG_LOCK;
417 if (dp->d_flags & DISKFLAG_WANTED) {
418 dp->d_flags &= ~DISKFLAG_WANTED;
426 * Close a disk device or partition
430 diskclose(dev_t dev, int fflag, int devtype, struct thread *td)
438 dsclose(dev, devtype, dp->d_slice);
439 if (!dsisopen(dp->d_slice))
440 error = dev_dclose(dp->d_rawdev, fflag, devtype, td);
445 * Execute strategy routine
449 diskstrategy(struct buf *bp)
453 dp = bp->b_dev->si_disk;
457 bp->b_flags |= B_ERROR;
461 KKASSERT(bp->b_dev->si_disk == dp);
463 if (dscheck(bp, dp->d_slice) <= 0) {
467 bp->b_dev = dp->d_rawdev;
468 dev_dstrategy(dp->d_rawdev, bp);
472 * First execute the ioctl on the disk device, and if it isn't supported
473 * try running it on the backing device.
477 diskioctl(dev_t dev, u_long cmd, caddr_t data, int fflag, struct thread *td)
486 error = dsioctl(dev, cmd, data, fflag, &dp->d_slice);
487 if (error == ENOIOCTL)
488 error = dev_dioctl(dp->d_rawdev, cmd, data, fflag, td);
504 return(dssize(dev, &dp->d_slice));
506 if (dp != dev->si_disk) {
507 dev->si_drv1 = pdev->si_drv1;
508 dev->si_drv2 = pdev->si_drv2;
509 /* XXX: don't set bp->b_dev->si_disk (?) */
514 SYSCTL_DECL(_debug_sizeof);
516 SYSCTL_INT(_debug_sizeof, OID_AUTO, disklabel, CTLFLAG_RD,
517 0, sizeof(struct disklabel), "sizeof(struct disklabel)");
519 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
520 0, sizeof(struct diskslices), "sizeof(struct diskslices)");
522 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
523 0, sizeof(struct disk), "sizeof(struct disk)");
527 * Seek sort for disks.
529 * The buf_queue keep two queues, sorted in ascending block order. The first
530 * queue holds those requests which are positioned after the current block
531 * (in the first request); the second, which starts at queue->switch_point,
532 * holds requests which came in after their block number was passed. Thus
533 * we implement a one way scan, retracting after reaching the end of the drive
534 * to the first request on the second queue, at which time it becomes the
537 * A one-way scan is natural because of the way UNIX read-ahead blocks are
541 bufqdisksort(struct buf_queue_head *bufq, struct buf *bp)
547 be = TAILQ_LAST(&bufq->queue, buf_queue);
549 * If the queue is empty or we are an
550 * ordered transaction, then it's easy.
552 if ((bq = bufq_first(bufq)) == NULL ||
553 (bp->b_flags & B_ORDERED) != 0) {
554 bufq_insert_tail(bufq, bp);
556 } else if (bufq->insert_point != NULL) {
559 * A certain portion of the list is
560 * "locked" to preserve ordering, so
561 * we can only insert after the insert
564 bq = bufq->insert_point;
568 * If we lie before the last removed (currently active)
569 * request, and are not inserting ourselves into the
570 * "locked" portion of the list, then we must add ourselves
571 * to the second request list.
573 if (bp->b_pblkno < bufq->last_pblkno) {
575 bq = bufq->switch_point;
577 * If we are starting a new secondary list,
581 bufq->switch_point = bp;
582 bufq_insert_tail(bufq, bp);
586 * If we lie ahead of the current switch point,
587 * insert us before the switch point and move
590 if (bp->b_pblkno < bq->b_pblkno) {
591 bufq->switch_point = bp;
592 TAILQ_INSERT_BEFORE(bq, bp, b_act);
596 if (bufq->switch_point != NULL)
597 be = TAILQ_PREV(bufq->switch_point,
600 * If we lie between last_pblkno and bq,
603 if (bp->b_pblkno < bq->b_pblkno) {
604 TAILQ_INSERT_BEFORE(bq, bp, b_act);
611 * Request is at/after our current position in the list.
612 * Optimize for sequential I/O by seeing if we go at the tail.
614 if (bp->b_pblkno > be->b_pblkno) {
615 TAILQ_INSERT_AFTER(&bufq->queue, be, bp, b_act);
619 /* Otherwise, insertion sort */
620 while ((bn = TAILQ_NEXT(bq, b_act)) != NULL) {
623 * We want to go after the current request if it is the end
624 * of the first request list, or if the next request is a
625 * larger cylinder than our request.
627 if (bn == bufq->switch_point
628 || bp->b_pblkno < bn->b_pblkno)
632 TAILQ_INSERT_AFTER(&bufq->queue, bq, bp, b_act);
637 * Attempt to read a disk label from a device using the indicated strategy
638 * routine. The label must be partly set up before this: secpercyl, secsize
639 * and anything required in the strategy routine (e.g., dummy bounds for the
640 * partition containing the label) must be filled in before calling us.
641 * Returns NULL on success and an error string on failure.
644 readdisklabel(dev_t dev, struct disklabel *lp)
647 struct disklabel *dlp;
650 bp = geteblk((int)lp->d_secsize);
652 bp->b_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE);
653 bp->b_bcount = lp->d_secsize;
654 bp->b_flags &= ~B_INVAL;
655 bp->b_flags |= B_READ;
659 else for (dlp = (struct disklabel *)bp->b_data;
660 dlp <= (struct disklabel *)((char *)bp->b_data +
661 lp->d_secsize - sizeof(*dlp));
662 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
663 if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) {
665 msg = "no disk label";
666 } else if (dlp->d_npartitions > MAXPARTITIONS ||
668 msg = "disk label corrupted";
675 bp->b_flags |= B_INVAL | B_AGE;
681 * Check new disk label for sensibility before setting it.
684 setdisklabel(struct disklabel *olp, struct disklabel *nlp, u_long openmask)
687 struct partition *opp, *npp;
690 * Check it is actually a disklabel we are looking at.
692 if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC ||
696 * For each partition that we think is open,
698 while ((i = ffs((long)openmask)) != 0) {
701 * Check it is not changing....
703 openmask &= ~(1 << i);
704 if (nlp->d_npartitions <= i)
706 opp = &olp->d_partitions[i];
707 npp = &nlp->d_partitions[i];
708 if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size)
711 * Copy internally-set partition information
712 * if new label doesn't include it. XXX
713 * (If we are using it then we had better stay the same type)
714 * This is possibly dubious, as someone else noted (XXX)
716 if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) {
717 npp->p_fstype = opp->p_fstype;
718 npp->p_fsize = opp->p_fsize;
719 npp->p_frag = opp->p_frag;
720 npp->p_cpg = opp->p_cpg;
724 nlp->d_checksum = dkcksum(nlp);
730 * Write disk label back to device after modification.
733 writedisklabel(dev_t dev, struct disklabel *lp)
736 struct disklabel *dlp;
739 if (lp->d_partitions[RAW_PART].p_offset != 0)
740 return (EXDEV); /* not quite right */
741 bp = geteblk((int)lp->d_secsize);
742 bp->b_dev = dkmodpart(dev, RAW_PART);
743 bp->b_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE);
744 bp->b_bcount = lp->d_secsize;
747 * We read the label first to see if it's there,
748 * in which case we will put ours at the same offset into the block..
749 * (I think this is stupid [Julian])
750 * Note that you can't write a label out over a corrupted label!
751 * (also stupid.. how do you write the first one? by raw writes?)
753 bp->b_flags &= ~B_INVAL;
754 bp->b_flags |= B_READ;
759 for (dlp = (struct disklabel *)bp->b_data;
760 dlp <= (struct disklabel *)
761 ((char *)bp->b_data + lp->d_secsize - sizeof(*dlp));
762 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
763 if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC &&
766 bp->b_flags &= ~(B_DONE | B_READ);
767 bp->b_flags |= B_WRITE;
768 bp->b_dev = dkmodpart(dev, RAW_PART);
777 bzero(bp->b_data, lp->d_secsize);
778 dlp = (struct disklabel *)bp->b_data;
780 bp->b_flags &= ~B_INVAL;
781 bp->b_flags |= B_WRITE;
785 bp->b_flags |= B_INVAL | B_AGE;
791 * Disk error is the preface to plaintive error messages
792 * about failing disk transfers. It prints messages of the form
794 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
796 * if the offset of the error in the transfer and a disk label
797 * are both available. blkdone should be -1 if the position of the error
798 * is unknown; the disklabel pointer may be null from drivers that have not
799 * been converted to use them. The message is printed with printf
800 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
801 * The message should be completed (with at least a newline) with printf
802 * or addlog, respectively. There is no trailing space.
805 diskerr(struct buf *bp, dev_t dev, char *what, int pri,
806 int blkdone, struct disklabel *lp)
808 int unit = dkunit(dev);
809 int slice = dkslice(dev);
810 int part = dkpart(dev);
815 sname = dsname(dev, unit, slice, part, partname);
816 printf("%s%s: %s %sing fsbn ", sname, partname, what,
817 bp->b_flags & B_READ ? "read" : "writ");
819 if (bp->b_bcount <= DEV_BSIZE) {
820 printf("%ld", (long)sn);
824 printf("%ld of ", (long)sn);
826 printf("%ld-%ld", (long)bp->b_blkno,
827 (long)(bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE));
829 if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) {
831 sn *= DEV_BSIZE / lp->d_secsize; /* XXX */
833 sn += lp->d_partitions[part].p_offset;
835 * XXX should add slice offset and not print the slice,
836 * but we don't know the slice pointer.
837 * XXX should print bp->b_pblkno so that this will work
838 * independent of slices, labels and bad sector remapping,
839 * but some drivers don't set bp->b_pblkno.
841 printf(" (%s bn %ld; cn %ld", sname, (long)sn,
842 (long)(sn / lp->d_secpercyl));
843 sn %= (long)lp->d_secpercyl;
844 printf(" tn %ld sn %ld)", (long)(sn / lp->d_nsectors),
845 (long)(sn % lp->d_nsectors));