xref: /netbsd-src/sys/kern/subr_disk.c (revision 08c81a9c2dc8c7300e893321eb65c0925d60871c)
1 /*	$NetBSD: subr_disk.c,v 1.42 2002/08/30 15:43:40 hannken Exp $	*/
2 
3 /*-
4  * Copyright (c) 1996, 1997, 1999, 2000 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
9  * NASA Ames Research Center.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
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.
48  *
49  * Redistribution and use in source and binary forms, with or without
50  * modification, are permitted provided that the following conditions
51  * are met:
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.
64  *
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
75  * SUCH DAMAGE.
76  *
77  *	@(#)ufs_disksubr.c	8.5 (Berkeley) 1/21/94
78  */
79 
80 #include <sys/cdefs.h>
81 __KERNEL_RCSID(0, "$NetBSD: subr_disk.c,v 1.42 2002/08/30 15:43:40 hannken Exp $");
82 
83 #include <sys/param.h>
84 #include <sys/kernel.h>
85 #include <sys/malloc.h>
86 #include <sys/buf.h>
87 #include <sys/syslog.h>
88 #include <sys/disklabel.h>
89 #include <sys/disk.h>
90 #include <sys/sysctl.h>
91 
92 /*
93  * A global list of all disks attached to the system.  May grow or
94  * shrink over time.
95  */
96 struct	disklist_head disklist;	/* TAILQ_HEAD */
97 int	disk_count;		/* number of drives in global disklist */
98 struct simplelock disklist_slock = SIMPLELOCK_INITIALIZER;
99 
100 /*
101  * Compute checksum for disk label.
102  */
103 u_int
104 dkcksum(struct disklabel *lp)
105 {
106 	u_short *start, *end;
107 	u_short sum = 0;
108 
109 	start = (u_short *)lp;
110 	end = (u_short *)&lp->d_partitions[lp->d_npartitions];
111 	while (start < end)
112 		sum ^= *start++;
113 	return (sum);
114 }
115 
116 /*
117  * Disk error is the preface to plaintive error messages
118  * about failing disk transfers.  It prints messages of the form
119 
120 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
121 
122  * if the offset of the error in the transfer and a disk label
123  * are both available.  blkdone should be -1 if the position of the error
124  * is unknown; the disklabel pointer may be null from drivers that have not
125  * been converted to use them.  The message is printed with printf
126  * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
127  * The message should be completed (with at least a newline) with printf
128  * or addlog, respectively.  There is no trailing space.
129  */
130 void
131 diskerr(const struct buf *bp, const char *dname, const char *what, int pri,
132     int blkdone, const struct disklabel *lp)
133 {
134 	int unit = DISKUNIT(bp->b_dev), part = DISKPART(bp->b_dev);
135 	void (*pr)(const char *, ...);
136 	char partname = 'a' + part;
137 	int sn;
138 
139 	if (pri != LOG_PRINTF) {
140 		static const char fmt[] = "";
141 		log(pri, fmt);
142 		pr = addlog;
143 	} else
144 		pr = printf;
145 	(*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what,
146 	    bp->b_flags & B_READ ? "read" : "writ");
147 	sn = bp->b_blkno;
148 	if (bp->b_bcount <= DEV_BSIZE)
149 		(*pr)("%d", sn);
150 	else {
151 		if (blkdone >= 0) {
152 			sn += blkdone;
153 			(*pr)("%d of ", sn);
154 		}
155 		(*pr)("%d-%d", bp->b_blkno,
156 		    bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE);
157 	}
158 	if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) {
159 		sn += lp->d_partitions[part].p_offset;
160 		(*pr)(" (%s%d bn %d; cn %d", dname, unit, sn,
161 		    sn / lp->d_secpercyl);
162 		sn %= lp->d_secpercyl;
163 		(*pr)(" tn %d sn %d)", sn / lp->d_nsectors,
164 		    sn % lp->d_nsectors);
165 	}
166 }
167 
168 /*
169  * Initialize the disklist.  Called by main() before autoconfiguration.
170  */
171 void
172 disk_init(void)
173 {
174 
175 	TAILQ_INIT(&disklist);
176 	disk_count = 0;
177 }
178 
179 /*
180  * Searches the disklist for the disk corresponding to the
181  * name provided.
182  */
183 struct disk *
184 disk_find(char *name)
185 {
186 	struct disk *diskp;
187 
188 	if ((name == NULL) || (disk_count <= 0))
189 		return (NULL);
190 
191 	simple_lock(&disklist_slock);
192 	for (diskp = TAILQ_FIRST(&disklist); diskp != NULL;
193 	    diskp = TAILQ_NEXT(diskp, dk_link))
194 		if (strcmp(diskp->dk_name, name) == 0) {
195 			simple_unlock(&disklist_slock);
196 			return (diskp);
197 		}
198 	simple_unlock(&disklist_slock);
199 
200 	return (NULL);
201 }
202 
203 /*
204  * Attach a disk.
205  */
206 void
207 disk_attach(struct disk *diskp)
208 {
209 	int s;
210 
211 	/*
212 	 * Allocate and initialize the disklabel structures.  Note that
213 	 * it's not safe to sleep here, since we're probably going to be
214 	 * called during autoconfiguration.
215 	 */
216 	diskp->dk_label = malloc(sizeof(struct disklabel), M_DEVBUF, M_NOWAIT);
217 	diskp->dk_cpulabel = malloc(sizeof(struct cpu_disklabel), M_DEVBUF,
218 	    M_NOWAIT);
219 	if ((diskp->dk_label == NULL) || (diskp->dk_cpulabel == NULL))
220 		panic("disk_attach: can't allocate storage for disklabel");
221 
222 	memset(diskp->dk_label, 0, sizeof(struct disklabel));
223 	memset(diskp->dk_cpulabel, 0, sizeof(struct cpu_disklabel));
224 
225 	/*
226 	 * Set the attached timestamp.
227 	 */
228 	s = splclock();
229 	diskp->dk_attachtime = mono_time;
230 	splx(s);
231 
232 	/*
233 	 * Link into the disklist.
234 	 */
235 	simple_lock(&disklist_slock);
236 	TAILQ_INSERT_TAIL(&disklist, diskp, dk_link);
237 	simple_unlock(&disklist_slock);
238 	++disk_count;
239 }
240 
241 /*
242  * Detach a disk.
243  */
244 void
245 disk_detach(struct disk *diskp)
246 {
247 
248 	/*
249 	 * Remove from the disklist.
250 	 */
251 	if (--disk_count < 0)
252 		panic("disk_detach: disk_count < 0");
253 	simple_lock(&disklist_slock);
254 	TAILQ_REMOVE(&disklist, diskp, dk_link);
255 	simple_unlock(&disklist_slock);
256 
257 	/*
258 	 * Free the space used by the disklabel structures.
259 	 */
260 	free(diskp->dk_label, M_DEVBUF);
261 	free(diskp->dk_cpulabel, M_DEVBUF);
262 }
263 
264 /*
265  * Increment a disk's busy counter.  If the counter is going from
266  * 0 to 1, set the timestamp.
267  */
268 void
269 disk_busy(struct disk *diskp)
270 {
271 	int s;
272 
273 	/*
274 	 * XXX We'd like to use something as accurate as microtime(),
275 	 * but that doesn't depend on the system TOD clock.
276 	 */
277 	if (diskp->dk_busy++ == 0) {
278 		s = splclock();
279 		diskp->dk_timestamp = mono_time;
280 		splx(s);
281 	}
282 }
283 
284 /*
285  * Decrement a disk's busy counter, increment the byte count, total busy
286  * time, and reset the timestamp.
287  */
288 void
289 disk_unbusy(struct disk *diskp, long bcount)
290 {
291 	int s;
292 	struct timeval dv_time, diff_time;
293 
294 	if (diskp->dk_busy-- == 0) {
295 		printf("%s: dk_busy < 0\n", diskp->dk_name);
296 		panic("disk_unbusy");
297 	}
298 
299 	s = splclock();
300 	dv_time = mono_time;
301 	splx(s);
302 
303 	timersub(&dv_time, &diskp->dk_timestamp, &diff_time);
304 	timeradd(&diskp->dk_time, &diff_time, &diskp->dk_time);
305 
306 	diskp->dk_timestamp = dv_time;
307 	if (bcount > 0) {
308 		diskp->dk_bytes += bcount;
309 		diskp->dk_xfer++;
310 	}
311 }
312 
313 /*
314  * Reset the metrics counters on the given disk.  Note that we cannot
315  * reset the busy counter, as it may case a panic in disk_unbusy().
316  * We also must avoid playing with the timestamp information, as it
317  * may skew any pending transfer results.
318  */
319 void
320 disk_resetstat(struct disk *diskp)
321 {
322 	int s = splbio(), t;
323 
324 	diskp->dk_xfer = 0;
325 	diskp->dk_bytes = 0;
326 
327 	t = splclock();
328 	diskp->dk_attachtime = mono_time;
329 	splx(t);
330 
331 	timerclear(&diskp->dk_time);
332 
333 	splx(s);
334 }
335 
336 int
337 sysctl_disknames(void *vwhere, size_t *sizep)
338 {
339 	char buf[DK_DISKNAMELEN + 1];
340 	char *where = vwhere;
341 	struct disk *diskp;
342 	size_t needed, left, slen;
343 	int error, first;
344 
345 	first = 1;
346 	error = 0;
347 	needed = 0;
348 	left = *sizep;
349 
350 	simple_lock(&disklist_slock);
351 	for (diskp = TAILQ_FIRST(&disklist); diskp != NULL;
352 	    diskp = TAILQ_NEXT(diskp, dk_link)) {
353 		if (where == NULL)
354 			needed += strlen(diskp->dk_name) + 1;
355 		else {
356 			memset(buf, 0, sizeof(buf));
357 			if (first) {
358 				strncpy(buf, diskp->dk_name, sizeof(buf));
359 				first = 0;
360 			} else {
361 				buf[0] = ' ';
362 				strncpy(buf + 1, diskp->dk_name,
363 				    sizeof(buf) - 1);
364 			}
365 			buf[DK_DISKNAMELEN] = '\0';
366 			slen = strlen(buf);
367 			if (left < slen + 1)
368 				break;
369 			/* +1 to copy out the trailing NUL byte */
370 			error = copyout(buf, where, slen + 1);
371 			if (error)
372 				break;
373 			where += slen;
374 			needed += slen;
375 			left -= slen;
376 		}
377 	}
378 	simple_unlock(&disklist_slock);
379 	*sizep = needed;
380 	return (error);
381 }
382 
383 int
384 sysctl_diskstats(int *name, u_int namelen, void *vwhere, size_t *sizep)
385 {
386 	struct disk_sysctl sdisk;
387 	struct disk *diskp;
388 	char *where = vwhere;
389 	size_t tocopy, left;
390 	int error;
391 
392 	if (where == NULL) {
393 		*sizep = disk_count * sizeof(struct disk_sysctl);
394 		return (0);
395 	}
396 
397 	if (namelen == 0)
398 		tocopy = sizeof(sdisk);
399 	else
400 		tocopy = name[0];
401 
402 	error = 0;
403 	left = *sizep;
404 	memset(&sdisk, 0, sizeof(sdisk));
405 	*sizep = 0;
406 
407 	simple_lock(&disklist_slock);
408 	TAILQ_FOREACH(diskp, &disklist, dk_link) {
409 		if (left < sizeof(struct disk_sysctl))
410 			break;
411 		strncpy(sdisk.dk_name, diskp->dk_name, sizeof(sdisk.dk_name));
412 		sdisk.dk_xfer = diskp->dk_xfer;
413 		sdisk.dk_seek = diskp->dk_seek;
414 		sdisk.dk_bytes = diskp->dk_bytes;
415 		sdisk.dk_attachtime_sec = diskp->dk_attachtime.tv_sec;
416 		sdisk.dk_attachtime_usec = diskp->dk_attachtime.tv_usec;
417 		sdisk.dk_timestamp_sec = diskp->dk_timestamp.tv_sec;
418 		sdisk.dk_timestamp_usec = diskp->dk_timestamp.tv_usec;
419 		sdisk.dk_time_sec = diskp->dk_time.tv_sec;
420 		sdisk.dk_time_usec = diskp->dk_time.tv_usec;
421 		sdisk.dk_busy = diskp->dk_busy;
422 
423 		error = copyout(&sdisk, where, min(tocopy, sizeof(sdisk)));
424 		if (error)
425 			break;
426 		where += tocopy;
427 		*sizep += tocopy;
428 		left -= tocopy;
429 	}
430 	simple_unlock(&disklist_slock);
431 	return (error);
432 }
433 
434 
435 struct bufq_fcfs {
436 	TAILQ_HEAD(, buf) bq_head;	/* actual list of buffers */
437 };
438 
439 struct bufq_disksort {
440 	TAILQ_HEAD(, buf) bq_head;	/* actual list of buffers */
441 };
442 
443 #define PRIO_READ_BURST		48
444 #define PRIO_WRITE_REQ		16
445 
446 struct bufq_prio {
447 	TAILQ_HEAD(, buf) bq_read, bq_write; /* actual list of buffers */
448 	struct buf *bq_write_next;	/* next request in bq_write */
449 	struct buf *bq_next;		/* current request */
450 	int bq_read_burst;		/* # of consecutive reads */
451 };
452 
453 
454 /*
455  * Check if two buf's are in ascending order.
456  */
457 static __inline int
458 buf_inorder(struct buf *bp, struct buf *bq, int sortby)
459 {
460 	int r;
461 
462 	if (bp == NULL || bq == NULL)
463 		return(bq == NULL);
464 
465 	if (sortby == BUFQ_SORT_CYLINDER)
466 		r = bp->b_cylinder - bq->b_cylinder;
467 	else
468 		r = 0;
469 
470 	if (r == 0)
471 		r = bp->b_rawblkno - bq->b_rawblkno;
472 
473 	return(r <= 0);
474 }
475 
476 
477 /*
478  * First-come first-served sort for disks.
479  *
480  * Requests are appended to the queue without any reordering.
481  */
482 static void
483 bufq_fcfs_put(struct bufq_state *bufq, struct buf *bp)
484 {
485 	struct bufq_fcfs *fcfs = bufq->bq_private;
486 
487 	TAILQ_INSERT_TAIL(&fcfs->bq_head, bp, b_actq);
488 }
489 
490 static struct buf *
491 bufq_fcfs_get(struct bufq_state *bufq, int remove)
492 {
493 	struct bufq_fcfs *fcfs = bufq->bq_private;
494 	struct buf *bp;
495 
496 	bp = TAILQ_FIRST(&fcfs->bq_head);
497 
498 	if (bp != NULL && remove)
499 		TAILQ_REMOVE(&fcfs->bq_head, bp, b_actq);
500 
501 	return(bp);
502 }
503 
504 
505 /*
506  * Seek sort for disks.
507  *
508  * There are actually two queues, sorted in ascendening order.  The first
509  * queue holds those requests which are positioned after the current block;
510  * the second holds requests which came in after their position was passed.
511  * Thus we implement a one-way scan, retracting after reaching the end of
512  * the drive to the first request on the second queue, at which time it
513  * becomes the first queue.
514  *
515  * A one-way scan is natural because of the way UNIX read-ahead blocks are
516  * allocated.
517  */
518 static void
519 bufq_disksort_put(struct bufq_state *bufq, struct buf *bp)
520 {
521 	struct bufq_disksort *disksort = bufq->bq_private;
522 	struct buf *bq, *nbq;
523 	int sortby;
524 
525 	sortby = bufq->bq_flags & BUFQ_SORT_MASK;
526 
527 	bq = TAILQ_FIRST(&disksort->bq_head);
528 
529 	/*
530 	 * If the queue is empty it's easy; we just go on the end.
531 	 */
532 	if (bq == NULL) {
533 		TAILQ_INSERT_TAIL(&disksort->bq_head, bp, b_actq);
534 		return;
535 	}
536 
537 	/*
538 	 * If we lie before the currently active request, then we
539 	 * must locate the second request list and add ourselves to it.
540 	 */
541 	if (buf_inorder(bp, bq, sortby)) {
542 		while ((nbq = TAILQ_NEXT(bq, b_actq)) != NULL) {
543 			/*
544 			 * Check for an ``inversion'' in the normally ascending
545 			 * block numbers, indicating the start of the second
546 			 * request list.
547 			 */
548 			if (buf_inorder(nbq, bq, sortby)) {
549 				/*
550 				 * Search the second request list for the first
551 				 * request at a larger block number.  We go
552 				 * after that; if there is no such request, we
553 				 * go at the end.
554 				 */
555 				do {
556 					if (buf_inorder(bp, nbq, sortby))
557 						goto insert;
558 					bq = nbq;
559 				} while ((nbq = TAILQ_NEXT(bq, b_actq)) != NULL);
560 				goto insert;		/* after last */
561 			}
562 			bq = nbq;
563 		}
564 		/*
565 		 * No inversions... we will go after the last, and
566 		 * be the first request in the second request list.
567 		 */
568 		goto insert;
569 	}
570 	/*
571 	 * Request is at/after the current request...
572 	 * sort in the first request list.
573 	 */
574 	while ((nbq = TAILQ_NEXT(bq, b_actq)) != NULL) {
575 		/*
576 		 * We want to go after the current request if there is an
577 		 * inversion after it (i.e. it is the end of the first
578 		 * request list), or if the next request is a larger cylinder
579 		 * than our request.
580 		 */
581 		if (buf_inorder(nbq, bq, sortby) ||
582 		    buf_inorder(bp, nbq, sortby))
583 			goto insert;
584 		bq = nbq;
585 	}
586 	/*
587 	 * Neither a second list nor a larger request... we go at the end of
588 	 * the first list, which is the same as the end of the whole schebang.
589 	 */
590 insert:	TAILQ_INSERT_AFTER(&disksort->bq_head, bq, bp, b_actq);
591 }
592 
593 static struct buf *
594 bufq_disksort_get(struct bufq_state *bufq, int remove)
595 {
596 	struct bufq_disksort *disksort = bufq->bq_private;
597 	struct buf *bp;
598 
599 	bp = TAILQ_FIRST(&disksort->bq_head);
600 
601 	if (bp != NULL && remove)
602 		TAILQ_REMOVE(&disksort->bq_head, bp, b_actq);
603 
604 	return(bp);
605 }
606 
607 
608 /*
609  * Seek sort for disks.
610  *
611  * There are two queues.  The first queue holds read requests; the second
612  * holds write requests.  The read queue is first-come first-served; the
613  * write queue is sorted in ascendening block order.
614  * The read queue is processed first.  After PRIO_READ_BURST consecutive
615  * read requests with non-empty write queue PRIO_WRITE_REQ requests from
616  * the write queue will be processed.
617  */
618 static void
619 bufq_prio_put(struct bufq_state *bufq, struct buf *bp)
620 {
621 	struct bufq_prio *prio = bufq->bq_private;
622 	struct buf *bq;
623 	int sortby;
624 
625 	sortby = bufq->bq_flags & BUFQ_SORT_MASK;
626 
627 	/*
628 	 * If it's a read request append it to the list.
629 	 */
630 	if ((bp->b_flags & B_READ) == B_READ) {
631 		TAILQ_INSERT_TAIL(&prio->bq_read, bp, b_actq);
632 		return;
633 	}
634 
635 	bq = TAILQ_FIRST(&prio->bq_write);
636 
637 	/*
638 	 * If the write list is empty, simply append it to the list.
639 	 */
640 	if (bq == NULL) {
641 		TAILQ_INSERT_TAIL(&prio->bq_write, bp, b_actq);
642 		prio->bq_write_next = bp;
643 		return;
644 	}
645 
646 	/*
647 	 * If we lie after the next request, insert after this request.
648 	 */
649 	if (buf_inorder(prio->bq_write_next, bp, sortby))
650 		bq = prio->bq_write_next;
651 
652 	/*
653 	 * Search for the first request at a larger block number.
654 	 * We go before this request if it exists.
655 	 */
656 	while (bq != NULL && buf_inorder(bq, bp, sortby))
657 		bq = TAILQ_NEXT(bq, b_actq);
658 
659 	if (bq != NULL)
660 		TAILQ_INSERT_BEFORE(bq, bp, b_actq);
661 	else
662 		TAILQ_INSERT_TAIL(&prio->bq_write, bp, b_actq);
663 }
664 
665 static struct buf *
666 bufq_prio_get(struct bufq_state *bufq, int remove)
667 {
668 	struct bufq_prio *prio = bufq->bq_private;
669 	struct buf *bp;
670 
671 	/*
672 	 * If no current request, get next from the lists.
673 	 */
674 	if (prio->bq_next == NULL) {
675 		/*
676 		 * If at least one list is empty, select the other.
677 		 */
678 
679 		if (TAILQ_FIRST(&prio->bq_read) == NULL) {
680 			prio->bq_next = prio->bq_write_next;
681 			prio->bq_read_burst = 0;
682 		} else if (prio->bq_write_next == NULL) {
683 			prio->bq_next = TAILQ_FIRST(&prio->bq_read);
684 			prio->bq_read_burst = 0;
685 		} else {
686 			/*
687 			 * Both list have requests.  Select the read list up
688 			 * to PRIO_READ_BURST times, then select the write
689 			 * list PRIO_WRITE_REQ times.
690 			 */
691 
692 			if (prio->bq_read_burst++ < PRIO_READ_BURST)
693 				prio->bq_next = TAILQ_FIRST(&prio->bq_read);
694 			else if (prio->bq_read_burst <
695 				     PRIO_READ_BURST + PRIO_WRITE_REQ)
696 				prio->bq_next = prio->bq_write_next;
697 			else {
698 				prio->bq_next = TAILQ_FIRST(&prio->bq_read);
699 				prio->bq_read_burst = 0;
700 			}
701 		}
702 	}
703 
704 	bp = prio->bq_next;
705 
706 	if (prio->bq_next != NULL && remove) {
707 		if ((prio->bq_next->b_flags & B_READ) == B_READ)
708 			TAILQ_REMOVE(&prio->bq_read, prio->bq_next, b_actq);
709 		else {
710 			TAILQ_REMOVE(&prio->bq_write, prio->bq_next, b_actq);
711 			/*
712 			 * Advance the write pointer.
713 			 */
714 			prio->bq_write_next =
715 			    TAILQ_NEXT(prio->bq_write_next, b_actq);
716 			if (prio->bq_write_next == NULL)
717 				prio->bq_write_next =
718 				    TAILQ_FIRST(&prio->bq_write);
719 		}
720 
721 		prio->bq_next = NULL;
722 	}
723 
724 	return(bp);
725 }
726 
727 /*
728  * Create a device buffer queue.
729  */
730 void
731 bufq_alloc(struct bufq_state *bufq, int flags)
732 {
733 	struct bufq_fcfs *fcfs;
734 	struct bufq_disksort *disksort;
735 	struct bufq_prio *prio;
736 
737 	bufq->bq_flags = flags;
738 
739 	switch (flags & BUFQ_SORT_MASK) {
740 	case BUFQ_SORT_RAWBLOCK:
741 	case BUFQ_SORT_CYLINDER:
742 		break;
743 	case 0:
744 		if ((flags & BUFQ_METHOD_MASK) == BUFQ_FCFS)
745 			break;
746 		/* FALLTHROUGH */
747 	default:
748 		panic("bufq_alloc: sort out of range");
749 	}
750 
751 	switch (flags & BUFQ_METHOD_MASK) {
752 	case BUFQ_FCFS:
753 		bufq->bq_get = bufq_fcfs_get;
754 		bufq->bq_put = bufq_fcfs_put;
755 		MALLOC(bufq->bq_private, struct bufq_fcfs *,
756 		    sizeof(struct bufq_fcfs), M_DEVBUF, M_ZERO);
757 		fcfs = (struct bufq_fcfs *)bufq->bq_private;
758 		TAILQ_INIT(&fcfs->bq_head);
759 		break;
760 	case BUFQ_DISKSORT:
761 		bufq->bq_get = bufq_disksort_get;
762 		bufq->bq_put = bufq_disksort_put;
763 		MALLOC(bufq->bq_private, struct bufq_disksort *,
764 		    sizeof(struct bufq_disksort), M_DEVBUF, M_ZERO);
765 		disksort = (struct bufq_disksort *)bufq->bq_private;
766 		TAILQ_INIT(&disksort->bq_head);
767 		break;
768 	case BUFQ_READ_PRIO:
769 		bufq->bq_get = bufq_prio_get;
770 		bufq->bq_put = bufq_prio_put;
771 		MALLOC(bufq->bq_private, struct bufq_prio *,
772 		    sizeof(struct bufq_prio), M_DEVBUF, M_ZERO);
773 		prio = (struct bufq_prio *)bufq->bq_private;
774 		TAILQ_INIT(&prio->bq_read);
775 		TAILQ_INIT(&prio->bq_write);
776 		break;
777 	default:
778 		panic("bufq_alloc: method out of range");
779 	}
780 }
781 
782 /*
783  * Destroy a device buffer queue.
784  */
785 void
786 bufq_free(struct bufq_state *bufq)
787 {
788 	KASSERT(bufq->bq_private != NULL);
789 	KASSERT(BUFQ_PEEK(bufq) == NULL);
790 
791 	FREE(bufq->bq_private, M_DEVBUF);
792 	bufq->bq_get = NULL;
793 	bufq->bq_put = NULL;
794 }
795