xref: /dragonfly/sys/kern/subr_disk.c (revision 9bb2a92d)
1 /*
2  * ----------------------------------------------------------------------------
3  * "THE BEER-WARE LICENSE" (Revision 42):
4  * <phk@FreeBSD.ORG> wrote this file.  As long as you retain this notice you
5  * can do whatever you want with this stuff. If we meet some day, and you think
6  * this stuff is worth it, you can buy me a beer in return.   Poul-Henning Kamp
7  * ----------------------------------------------------------------------------
8  *
9  * Copyright (c) 1982, 1986, 1988, 1993
10  *	The Regents of the University of California.  All rights reserved.
11  * (c) UNIX System Laboratories, Inc.
12  * All or some portions of this file are derived from material licensed
13  * to the University of California by American Telephone and Telegraph
14  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
15  * the permission of UNIX System Laboratories, Inc.
16  *
17  * Redistribution and use in source and binary forms, with or without
18  * modification, are permitted provided that the following conditions
19  * are met:
20  * 1. Redistributions of source code must retain the above copyright
21  *    notice, this list of conditions and the following disclaimer.
22  * 2. Redistributions in binary form must reproduce the above copyright
23  *    notice, this list of conditions and the following disclaimer in the
24  *    documentation and/or other materials provided with the distribution.
25  * 3. All advertising materials mentioning features or use of this software
26  *    must display the following acknowledgement:
27  *	This product includes software developed by the University of
28  *	California, Berkeley and its contributors.
29  * 4. Neither the name of the University nor the names of its contributors
30  *    may be used to endorse or promote products derived from this software
31  *    without specific prior written permission.
32  *
33  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
34  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
35  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
36  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
37  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
38  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
39  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
40  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
41  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
42  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
43  * SUCH DAMAGE.
44  *
45  *	@(#)ufs_disksubr.c	8.5 (Berkeley) 1/21/94
46  * $FreeBSD: src/sys/kern/subr_disk.c,v 1.20.2.6 2001/10/05 07:14:57 peter Exp $
47  * $FreeBSD: src/sys/ufs/ufs/ufs_disksubr.c,v 1.44.2.3 2001/03/05 05:42:19 obrien Exp $
48  * $DragonFly: src/sys/kern/subr_disk.c,v 1.8 2004/02/18 06:59:15 dillon Exp $
49  */
50 
51 #include <sys/param.h>
52 #include <sys/systm.h>
53 #include <sys/kernel.h>
54 #include <sys/proc.h>
55 #include <sys/sysctl.h>
56 #include <sys/buf.h>
57 #include <sys/conf.h>
58 #include <sys/disklabel.h>
59 #include <sys/diskslice.h>
60 #include <sys/disk.h>
61 #include <sys/malloc.h>
62 #include <sys/sysctl.h>
63 #include <machine/md_var.h>
64 #include <sys/ctype.h>
65 #include <sys/syslog.h>
66 #include <sys/device.h>
67 #include <sys/msgport.h>
68 #include <sys/msgport2.h>
69 #include <sys/buf2.h>
70 
71 static MALLOC_DEFINE(M_DISK, "disk", "disk data");
72 
73 static d_strategy_t diskstrategy;
74 static d_open_t diskopen;
75 static d_close_t diskclose;
76 static d_ioctl_t diskioctl;
77 static d_psize_t diskpsize;
78 static int disk_putport(lwkt_port_t port, lwkt_msg_t msg);
79 
80 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
81 
82 static void
83 inherit_raw(dev_t pdev, dev_t dev)
84 {
85 	dev->si_disk = pdev->si_disk;
86 	dev->si_drv1 = pdev->si_drv1;
87 	dev->si_drv2 = pdev->si_drv2;
88 	dev->si_iosize_max = pdev->si_iosize_max;
89 	dev->si_bsize_phys = pdev->si_bsize_phys;
90 	dev->si_bsize_best = pdev->si_bsize_best;
91 }
92 
93 /*
94  * Create a slice and unit managed disk.  The underlying raw disk device
95  * is specified by cdevsw.  We create the device as a managed device by
96  * first creating it normally then overriding the message port with our
97  * own frontend (which will be responsible for assigning pblkno).
98  */
99 dev_t
100 disk_create(int unit, struct disk *dp, int flags, struct cdevsw *cdevsw)
101 {
102 	dev_t dev;
103 
104 	bzero(dp, sizeof(*dp));
105 	lwkt_initport(&dp->d_port, NULL);	/* intercept port */
106 	dp->d_port.mp_putport = disk_putport;
107 
108 	dev = makedev(cdevsw->d_maj, 0);	/* base device */
109 	dev->si_disk = dp;
110 						/* forwarding port */
111 	dp->d_fwdport = cdevsw_add_override(cdevsw, &dp->d_port);
112 
113 	if (bootverbose)
114 		printf("Creating DISK %s%d\n", cdevsw->d_name, unit);
115 
116   	/*
117 	 * The whole disk placemarker holds the disk structure.
118 	 */
119 	dev = make_dev(cdevsw, dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART),
120 	    UID_ROOT, GID_OPERATOR, 0640, "%s%d", cdevsw->d_name, unit);
121 	dev->si_disk = dp;
122 	dp->d_dev = dev;
123 	dp->d_dsflags = flags;
124 	LIST_INSERT_HEAD(&disklist, dp, d_list);
125 	return (dev);
126 }
127 
128 void
129 disk_destroy(struct disk *disk)
130 {
131 	dev_t dev = disk->d_dev;
132 
133 	LIST_REMOVE(disk, d_list);
134 	bzero(disk, sizeof(*disk));
135 	dev->si_disk = NULL;
136 	destroy_dev(dev);
137 	/* YYY remove cdevsw entries? */
138 	return;
139 }
140 
141 int
142 disk_dumpcheck(dev_t dev, u_int *count, u_int *blkno, u_int *secsize)
143 {
144 	struct disk *dp;
145 	struct disklabel *dl;
146 	u_int boff;
147 
148 	dp = dev->si_disk;
149 	if (!dp)
150 		return (ENXIO);
151 	if (!dp->d_slice)
152 		return (ENXIO);
153 	dl = dsgetlabel(dev, dp->d_slice);
154 	if (!dl)
155 		return (ENXIO);
156 	*count = Maxmem * (PAGE_SIZE / dl->d_secsize);
157 	if (dumplo <= LABELSECTOR ||
158 	    (dumplo + *count > dl->d_partitions[dkpart(dev)].p_size))
159 		return (EINVAL);
160 	boff = dl->d_partitions[dkpart(dev)].p_offset +
161 	    dp->d_slice->dss_slices[dkslice(dev)].ds_offset;
162 	*blkno = boff + dumplo;
163 	*secsize = dl->d_secsize;
164 	return (0);
165 
166 }
167 
168 void
169 disk_invalidate (struct disk *disk)
170 {
171 	if (disk->d_slice)
172 		dsgone(&disk->d_slice);
173 }
174 
175 struct disk *
176 disk_enumerate(struct disk *disk)
177 {
178 	if (!disk)
179 		return (LIST_FIRST(&disklist));
180 	else
181 		return (LIST_NEXT(disk, d_list));
182 }
183 
184 static int
185 sysctl_disks(SYSCTL_HANDLER_ARGS)
186 {
187 	struct disk *disk;
188 	int error, first;
189 
190 	disk = NULL;
191 	first = 1;
192 
193 	while ((disk = disk_enumerate(disk))) {
194 		if (!first) {
195 			error = SYSCTL_OUT(req, " ", 1);
196 			if (error)
197 				return error;
198 		} else {
199 			first = 0;
200 		}
201 		error = SYSCTL_OUT(req, disk->d_dev->si_name, strlen(disk->d_dev->si_name));
202 		if (error)
203 			return error;
204 	}
205 	error = SYSCTL_OUT(req, "", 1);
206 	return error;
207 }
208 
209 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, 0, NULL,
210     sysctl_disks, "A", "names of available disks");
211 
212 /*
213  * The port intercept functions
214  */
215 static
216 int
217 disk_putport(lwkt_port_t port, lwkt_msg_t lmsg)
218 {
219 	struct disk *disk = (struct disk *)port;
220 	cdevallmsg_t msg = (cdevallmsg_t)lmsg;
221 	int error;
222 
223 	switch(msg->am_lmsg.ms_cmd) {
224 	case CDEV_CMD_OPEN:
225 		error = diskopen(
226 			    msg->am_open.msg.dev,
227 			    msg->am_open.oflags,
228 			    msg->am_open.devtype,
229 			    msg->am_open.td);
230 		break;
231 	case CDEV_CMD_CLOSE:
232 		error = diskclose(
233 			    msg->am_close.msg.dev,
234 			    msg->am_close.fflag,
235 			    msg->am_close.devtype,
236 			    msg->am_close.td);
237 		break;
238 	case CDEV_CMD_IOCTL:
239 		error = diskioctl(
240 			    msg->am_ioctl.msg.dev,
241 			    msg->am_ioctl.cmd,
242 			    msg->am_ioctl.data,
243 			    msg->am_ioctl.fflag,
244 			    msg->am_ioctl.td);
245 		break;
246 	case CDEV_CMD_STRATEGY:
247 		diskstrategy(msg->am_strategy.bp);
248 		error = 0;
249 		break;
250 	case CDEV_CMD_PSIZE:
251 		msg->am_psize.result = diskpsize(msg->am_psize.msg.dev);
252 		error = 0;      /* XXX */
253 		break;
254 	default:
255 		error = lwkt_forwardmsg(disk->d_fwdport, &msg->am_lmsg);
256 		break;
257 	}
258 	return(error);
259 }
260 
261 static int
262 diskopen(dev_t dev, int oflags, int devtype, struct thread *td)
263 {
264 	dev_t pdev;
265 	struct disk *dp;
266 	int error;
267 
268 	error = 0;
269 	pdev = dkmodpart(dkmodslice(dev, WHOLE_DISK_SLICE), RAW_PART);
270 
271 	dp = pdev->si_disk;
272 	if (dp == NULL)
273 		return (ENXIO);
274 
275 	while (dp->d_flags & DISKFLAG_LOCK) {
276 		dp->d_flags |= DISKFLAG_WANTED;
277 		error = tsleep(dp, PCATCH, "diskopen", hz);
278 		if (error)
279 			return (error);
280 	}
281 	dp->d_flags |= DISKFLAG_LOCK;
282 
283 	if (!dsisopen(dp->d_slice)) {
284 		if (!pdev->si_iosize_max)
285 			pdev->si_iosize_max = dev->si_iosize_max;
286 		error = dev_port_dopen(dp->d_fwdport, pdev, oflags, devtype, td);
287 	}
288 
289 	/* Inherit properties from the whole/raw dev_t */
290 	inherit_raw(pdev, dev);
291 
292 	if (error)
293 		goto out;
294 
295 	error = dsopen(dev, devtype, dp->d_dsflags, &dp->d_slice, &dp->d_label);
296 
297 	if (!dsisopen(dp->d_slice))
298 		dev_port_dclose(dp->d_fwdport, pdev, oflags, devtype, td);
299 out:
300 	dp->d_flags &= ~DISKFLAG_LOCK;
301 	if (dp->d_flags & DISKFLAG_WANTED) {
302 		dp->d_flags &= ~DISKFLAG_WANTED;
303 		wakeup(dp);
304 	}
305 
306 	return(error);
307 }
308 
309 static int
310 diskclose(dev_t dev, int fflag, int devtype, struct thread *td)
311 {
312 	struct disk *dp;
313 	int error;
314 	dev_t pdev;
315 
316 	error = 0;
317 	pdev = dkmodpart(dkmodslice(dev, WHOLE_DISK_SLICE), RAW_PART);
318 	dp = pdev->si_disk;
319 	if (!dp)
320 		return (ENXIO);
321 	dsclose(dev, devtype, dp->d_slice);
322 	if (!dsisopen(dp->d_slice))
323 		error = dev_port_dclose(dp->d_fwdport, pdev, fflag, devtype, td);
324 	return (error);
325 }
326 
327 static void
328 diskstrategy(struct buf *bp)
329 {
330 	dev_t pdev;
331 	struct disk *dp;
332 
333 	pdev = dkmodpart(dkmodslice(bp->b_dev, WHOLE_DISK_SLICE), RAW_PART);
334 	dp = pdev->si_disk;
335 	if (dp != bp->b_dev->si_disk)
336 		inherit_raw(pdev, bp->b_dev);
337 
338 	if (!dp) {
339 		bp->b_error = ENXIO;
340 		bp->b_flags |= B_ERROR;
341 		biodone(bp);
342 		return;
343 	}
344 
345 	if (dscheck(bp, dp->d_slice) <= 0) {
346 		biodone(bp);
347 		return;
348 	}
349 	dev_port_dstrategy(dp->d_fwdport, dp->d_dev, bp);
350 }
351 
352 /*
353  * note: when forwarding the ioctl we use the original device rather then
354  * the whole disk slice.
355  */
356 static int
357 diskioctl(dev_t dev, u_long cmd, caddr_t data, int fflag, struct thread *td)
358 {
359 	struct disk *dp;
360 	int error;
361 	dev_t pdev;
362 
363 	pdev = dkmodpart(dkmodslice(dev, WHOLE_DISK_SLICE), RAW_PART);
364 	dp = pdev->si_disk;
365 	if (!dp)
366 		return (ENXIO);
367 	error = dsioctl(dev, cmd, data, fflag, &dp->d_slice);
368 	if (error == ENOIOCTL)
369 		error = dev_port_dioctl(dp->d_fwdport, dev, cmd, data, fflag, td);
370 	return (error);
371 }
372 
373 static int
374 diskpsize(dev_t dev)
375 {
376 	struct disk *dp;
377 	dev_t pdev;
378 
379 	pdev = dkmodpart(dkmodslice(dev, WHOLE_DISK_SLICE), RAW_PART);
380 	dp = pdev->si_disk;
381 	if (!dp)
382 		return (-1);
383 	if (dp != dev->si_disk) {
384 		dev->si_drv1 = pdev->si_drv1;
385 		dev->si_drv2 = pdev->si_drv2;
386 		/* XXX: don't set bp->b_dev->si_disk (?) */
387 	}
388 	return (dssize(dev, &dp->d_slice));
389 }
390 
391 SYSCTL_DECL(_debug_sizeof);
392 
393 SYSCTL_INT(_debug_sizeof, OID_AUTO, disklabel, CTLFLAG_RD,
394     0, sizeof(struct disklabel), "sizeof(struct disklabel)");
395 
396 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
397     0, sizeof(struct diskslices), "sizeof(struct diskslices)");
398 
399 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
400     0, sizeof(struct disk), "sizeof(struct disk)");
401 
402 
403 /*
404  * Seek sort for disks.
405  *
406  * The buf_queue keep two queues, sorted in ascending block order.  The first
407  * queue holds those requests which are positioned after the current block
408  * (in the first request); the second, which starts at queue->switch_point,
409  * holds requests which came in after their block number was passed.  Thus
410  * we implement a one way scan, retracting after reaching the end of the drive
411  * to the first request on the second queue, at which time it becomes the
412  * first queue.
413  *
414  * A one-way scan is natural because of the way UNIX read-ahead blocks are
415  * allocated.
416  */
417 void
418 bufqdisksort(bufq, bp)
419 	struct buf_queue_head *bufq;
420 	struct buf *bp;
421 {
422 	struct buf *bq;
423 	struct buf *bn;
424 	struct buf *be;
425 
426 	be = TAILQ_LAST(&bufq->queue, buf_queue);
427 	/*
428 	 * If the queue is empty or we are an
429 	 * ordered transaction, then it's easy.
430 	 */
431 	if ((bq = bufq_first(bufq)) == NULL
432 	 || (bp->b_flags & B_ORDERED) != 0) {
433 		bufq_insert_tail(bufq, bp);
434 		return;
435 	} else if (bufq->insert_point != NULL) {
436 
437 		/*
438 		 * A certain portion of the list is
439 		 * "locked" to preserve ordering, so
440 		 * we can only insert after the insert
441 		 * point.
442 		 */
443 		bq = bufq->insert_point;
444 	} else {
445 
446 		/*
447 		 * If we lie before the last removed (currently active)
448 		 * request, and are not inserting ourselves into the
449 		 * "locked" portion of the list, then we must add ourselves
450 		 * to the second request list.
451 		 */
452 		if (bp->b_pblkno < bufq->last_pblkno) {
453 
454 			bq = bufq->switch_point;
455 			/*
456 			 * If we are starting a new secondary list,
457 			 * then it's easy.
458 			 */
459 			if (bq == NULL) {
460 				bufq->switch_point = bp;
461 				bufq_insert_tail(bufq, bp);
462 				return;
463 			}
464 			/*
465 			 * If we lie ahead of the current switch point,
466 			 * insert us before the switch point and move
467 			 * the switch point.
468 			 */
469 			if (bp->b_pblkno < bq->b_pblkno) {
470 				bufq->switch_point = bp;
471 				TAILQ_INSERT_BEFORE(bq, bp, b_act);
472 				return;
473 			}
474 		} else {
475 			if (bufq->switch_point != NULL)
476 				be = TAILQ_PREV(bufq->switch_point,
477 						buf_queue, b_act);
478 			/*
479 			 * If we lie between last_pblkno and bq,
480 			 * insert before bq.
481 			 */
482 			if (bp->b_pblkno < bq->b_pblkno) {
483 				TAILQ_INSERT_BEFORE(bq, bp, b_act);
484 				return;
485 			}
486 		}
487 	}
488 
489 	/*
490 	 * Request is at/after our current position in the list.
491 	 * Optimize for sequential I/O by seeing if we go at the tail.
492 	 */
493 	if (bp->b_pblkno > be->b_pblkno) {
494 		TAILQ_INSERT_AFTER(&bufq->queue, be, bp, b_act);
495 		return;
496 	}
497 
498 	/* Otherwise, insertion sort */
499 	while ((bn = TAILQ_NEXT(bq, b_act)) != NULL) {
500 
501 		/*
502 		 * We want to go after the current request if it is the end
503 		 * of the first request list, or if the next request is a
504 		 * larger cylinder than our request.
505 		 */
506 		if (bn == bufq->switch_point
507 		 || bp->b_pblkno < bn->b_pblkno)
508 			break;
509 		bq = bn;
510 	}
511 	TAILQ_INSERT_AFTER(&bufq->queue, bq, bp, b_act);
512 }
513 
514 
515 /*
516  * Attempt to read a disk label from a device using the indicated strategy
517  * routine.  The label must be partly set up before this: secpercyl, secsize
518  * and anything required in the strategy routine (e.g., dummy bounds for the
519  * partition containing the label) must be filled in before calling us.
520  * Returns NULL on success and an error string on failure.
521  */
522 char *
523 readdisklabel(dev, lp)
524 	dev_t dev;
525 	struct disklabel *lp;
526 {
527 	struct buf *bp;
528 	struct disklabel *dlp;
529 	char *msg = NULL;
530 
531 	bp = geteblk((int)lp->d_secsize);
532 	bp->b_dev = dev;
533 	bp->b_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE);
534 	bp->b_bcount = lp->d_secsize;
535 	bp->b_flags &= ~B_INVAL;
536 	bp->b_flags |= B_READ;
537 	BUF_STRATEGY(bp, 1);
538 	if (biowait(bp))
539 		msg = "I/O error";
540 	else for (dlp = (struct disklabel *)bp->b_data;
541 	    dlp <= (struct disklabel *)((char *)bp->b_data +
542 	    lp->d_secsize - sizeof(*dlp));
543 	    dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
544 		if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) {
545 			if (msg == NULL)
546 				msg = "no disk label";
547 		} else if (dlp->d_npartitions > MAXPARTITIONS ||
548 			   dkcksum(dlp) != 0)
549 			msg = "disk label corrupted";
550 		else {
551 			*lp = *dlp;
552 			msg = NULL;
553 			break;
554 		}
555 	}
556 	bp->b_flags |= B_INVAL | B_AGE;
557 	brelse(bp);
558 	return (msg);
559 }
560 
561 /*
562  * Check new disk label for sensibility before setting it.
563  */
564 int
565 setdisklabel(olp, nlp, openmask)
566 	struct disklabel *olp, *nlp;
567 	u_long openmask;
568 {
569 	int i;
570 	struct partition *opp, *npp;
571 
572 	/*
573 	 * Check it is actually a disklabel we are looking at.
574 	 */
575 	if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC ||
576 	    dkcksum(nlp) != 0)
577 		return (EINVAL);
578 	/*
579 	 * For each partition that we think is open,
580 	 */
581 	while ((i = ffs((long)openmask)) != 0) {
582 		i--;
583 		/*
584 	 	 * Check it is not changing....
585 	 	 */
586 		openmask &= ~(1 << i);
587 		if (nlp->d_npartitions <= i)
588 			return (EBUSY);
589 		opp = &olp->d_partitions[i];
590 		npp = &nlp->d_partitions[i];
591 		if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size)
592 			return (EBUSY);
593 		/*
594 		 * Copy internally-set partition information
595 		 * if new label doesn't include it.		XXX
596 		 * (If we are using it then we had better stay the same type)
597 		 * This is possibly dubious, as someone else noted (XXX)
598 		 */
599 		if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) {
600 			npp->p_fstype = opp->p_fstype;
601 			npp->p_fsize = opp->p_fsize;
602 			npp->p_frag = opp->p_frag;
603 			npp->p_cpg = opp->p_cpg;
604 		}
605 	}
606  	nlp->d_checksum = 0;
607  	nlp->d_checksum = dkcksum(nlp);
608 	*olp = *nlp;
609 	return (0);
610 }
611 
612 /*
613  * Write disk label back to device after modification.
614  */
615 int
616 writedisklabel(dev, lp)
617 	dev_t dev;
618 	struct disklabel *lp;
619 {
620 	struct buf *bp;
621 	struct disklabel *dlp;
622 	int error = 0;
623 
624 	if (lp->d_partitions[RAW_PART].p_offset != 0)
625 		return (EXDEV);			/* not quite right */
626 	bp = geteblk((int)lp->d_secsize);
627 	bp->b_dev = dkmodpart(dev, RAW_PART);
628 	bp->b_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE);
629 	bp->b_bcount = lp->d_secsize;
630 #if 1
631 	/*
632 	 * We read the label first to see if it's there,
633 	 * in which case we will put ours at the same offset into the block..
634 	 * (I think this is stupid [Julian])
635 	 * Note that you can't write a label out over a corrupted label!
636 	 * (also stupid.. how do you write the first one? by raw writes?)
637 	 */
638 	bp->b_flags &= ~B_INVAL;
639 	bp->b_flags |= B_READ;
640 	BUF_STRATEGY(bp, 1);
641 	error = biowait(bp);
642 	if (error)
643 		goto done;
644 	for (dlp = (struct disklabel *)bp->b_data;
645 	    dlp <= (struct disklabel *)
646 	      ((char *)bp->b_data + lp->d_secsize - sizeof(*dlp));
647 	    dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
648 		if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC &&
649 		    dkcksum(dlp) == 0) {
650 			*dlp = *lp;
651 			bp->b_flags &= ~(B_DONE | B_READ);
652 			bp->b_flags |= B_WRITE;
653 #ifdef __alpha__
654 			alpha_fix_srm_checksum(bp);
655 #endif
656 			BUF_STRATEGY(bp, 1);
657 			error = biowait(bp);
658 			goto done;
659 		}
660 	}
661 	error = ESRCH;
662 done:
663 #else
664 	bzero(bp->b_data, lp->d_secsize);
665 	dlp = (struct disklabel *)bp->b_data;
666 	*dlp = *lp;
667 	bp->b_flags &= ~B_INVAL;
668 	bp->b_flags |= B_WRITE;
669 	BUF_STRATEGY(bp, 1);
670 	error = biowait(bp);
671 #endif
672 	bp->b_flags |= B_INVAL | B_AGE;
673 	brelse(bp);
674 	return (error);
675 }
676 
677 /*
678  * Disk error is the preface to plaintive error messages
679  * about failing disk transfers.  It prints messages of the form
680 
681 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
682 
683  * if the offset of the error in the transfer and a disk label
684  * are both available.  blkdone should be -1 if the position of the error
685  * is unknown; the disklabel pointer may be null from drivers that have not
686  * been converted to use them.  The message is printed with printf
687  * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
688  * The message should be completed (with at least a newline) with printf
689  * or addlog, respectively.  There is no trailing space.
690  */
691 void
692 diskerr(bp, what, pri, blkdone, lp)
693 	struct buf *bp;
694 	char *what;
695 	int pri, blkdone;
696 	struct disklabel *lp;
697 {
698 	int unit = dkunit(bp->b_dev);
699 	int slice = dkslice(bp->b_dev);
700 	int part = dkpart(bp->b_dev);
701 	char partname[2];
702 	char *sname;
703 	daddr_t sn;
704 
705 	sname = dsname(bp->b_dev, unit, slice, part, partname);
706 	printf("%s%s: %s %sing fsbn ", sname, partname, what,
707 	      bp->b_flags & B_READ ? "read" : "writ");
708 	sn = bp->b_blkno;
709 	if (bp->b_bcount <= DEV_BSIZE)
710 		printf("%ld", (long)sn);
711 	else {
712 		if (blkdone >= 0) {
713 			sn += blkdone;
714 			printf("%ld of ", (long)sn);
715 		}
716 		printf("%ld-%ld", (long)bp->b_blkno,
717 		    (long)(bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE));
718 	}
719 	if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) {
720 #ifdef tahoe
721 		sn *= DEV_BSIZE / lp->d_secsize;		/* XXX */
722 #endif
723 		sn += lp->d_partitions[part].p_offset;
724 		/*
725 		 * XXX should add slice offset and not print the slice,
726 		 * but we don't know the slice pointer.
727 		 * XXX should print bp->b_pblkno so that this will work
728 		 * independent of slices, labels and bad sector remapping,
729 		 * but some drivers don't set bp->b_pblkno.
730 		 */
731 		printf(" (%s bn %ld; cn %ld", sname, (long)sn,
732 		    (long)(sn / lp->d_secpercyl));
733 		sn %= (long)lp->d_secpercyl;
734 		printf(" tn %ld sn %ld)", (long)(sn / lp->d_nsectors),
735 		    (long)(sn % lp->d_nsectors));
736 	}
737 }
738