xref: /openbsd/sys/kern/kern_bufq.c (revision 2b46a8cb)
1 /*	$OpenBSD: kern_bufq.c,v 1.35 2022/12/05 23:18:37 deraadt Exp $	*/
2 /*
3  * Copyright (c) 2010 Thordur I. Bjornsson <thib@openbsd.org>
4  * Copyright (c) 2010 David Gwynne <dlg@openbsd.org>
5  *
6  * Permission to use, copy, modify, and distribute this software for any
7  * purpose with or without fee is hereby granted, provided that the above
8  * copyright notice and this permission notice appear in all copies.
9  *
10  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
13  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17  */
18 
19 #include <sys/param.h>
20 #include <sys/systm.h>
21 #include <sys/malloc.h>
22 #include <sys/mount.h>
23 #include <sys/mutex.h>
24 #include <sys/buf.h>
25 #include <sys/errno.h>
26 #include <sys/queue.h>
27 
28 SLIST_HEAD(, bufq)	bufqs = SLIST_HEAD_INITIALIZER(bufqs);
29 struct mutex		bufqs_mtx = MUTEX_INITIALIZER(IPL_NONE);
30 int			bufqs_stop;
31 
32 struct bufq_impl {
33 	void		*(*impl_create)(void);
34 	void		 (*impl_destroy)(void *);
35 
36 	void		 (*impl_queue)(void *, struct buf *);
37 	struct buf	*(*impl_dequeue)(void *);
38 	void		 (*impl_requeue)(void *, struct buf *);
39 	int		 (*impl_peek)(void *);
40 };
41 
42 void		*bufq_fifo_create(void);
43 void		 bufq_fifo_destroy(void *);
44 void		 bufq_fifo_queue(void *, struct buf *);
45 struct buf	*bufq_fifo_dequeue(void *);
46 void		 bufq_fifo_requeue(void *, struct buf *);
47 int		 bufq_fifo_peek(void *);
48 
49 void		*bufq_nscan_create(void);
50 void		 bufq_nscan_destroy(void *);
51 void		 bufq_nscan_queue(void *, struct buf *);
52 struct buf	*bufq_nscan_dequeue(void *);
53 void		 bufq_nscan_requeue(void *, struct buf *);
54 int		 bufq_nscan_peek(void *);
55 
56 const struct bufq_impl bufq_impls[BUFQ_HOWMANY] = {
57 	{
58 		bufq_fifo_create,
59 		bufq_fifo_destroy,
60 		bufq_fifo_queue,
61 		bufq_fifo_dequeue,
62 		bufq_fifo_requeue,
63 		bufq_fifo_peek
64 	},
65 	{
66 		bufq_nscan_create,
67 		bufq_nscan_destroy,
68 		bufq_nscan_queue,
69 		bufq_nscan_dequeue,
70 		bufq_nscan_requeue,
71 		bufq_nscan_peek
72 	}
73 };
74 
75 int
bufq_init(struct bufq * bq,int type)76 bufq_init(struct bufq *bq, int type)
77 {
78 	u_int hi = BUFQ_HI, low = BUFQ_LOW;
79 
80 	if (type >= BUFQ_HOWMANY)
81 		panic("bufq_init: type %i unknown", type);
82 
83 	/*
84 	 * Ensure that writes can't consume the entire amount of kva
85 	 * available the buffer cache if we only have a limited amount
86 	 * of kva available to us.
87 	 */
88 	if (hi >= (bcstats.kvaslots / 16)) {
89 		hi = bcstats.kvaslots / 16;
90 		if (hi < 2)
91 			hi = 2;
92 		low = hi / 2;
93 	}
94 
95 	mtx_init(&bq->bufq_mtx, IPL_BIO);
96 	bq->bufq_hi = hi;
97 	bq->bufq_low = low;
98 	bq->bufq_type = type;
99 	bq->bufq_impl = &bufq_impls[type];
100 	bq->bufq_data = bq->bufq_impl->impl_create();
101 	if (bq->bufq_data == NULL) {
102 		/*
103 		 * we should actually return failure so disks attaching after
104 		 * boot in low memory situations dont panic the system.
105 		 */
106 		panic("bufq init fail");
107 	}
108 
109 	mtx_enter(&bufqs_mtx);
110 	while (bufqs_stop) {
111 		msleep_nsec(&bufqs_stop, &bufqs_mtx, PRIBIO, "bqinit", INFSLP);
112 	}
113 	SLIST_INSERT_HEAD(&bufqs, bq, bufq_entries);
114 	mtx_leave(&bufqs_mtx);
115 
116 	return (0);
117 }
118 
119 int
bufq_switch(struct bufq * bq,int type)120 bufq_switch(struct bufq *bq, int type)
121 {
122 	void		*data;
123 	void		*odata;
124 	int		otype;
125 	struct buf	*bp;
126 	int		ret;
127 
128 	mtx_enter(&bq->bufq_mtx);
129 	ret = (bq->bufq_type == type);
130 	mtx_leave(&bq->bufq_mtx);
131 	if (ret)
132 		return (0);
133 
134 	data = bufq_impls[type].impl_create();
135 	if (data == NULL)
136 		return (ENOMEM);
137 
138 	mtx_enter(&bq->bufq_mtx);
139 	if (bq->bufq_type != type) { /* might have changed during create */
140 		odata = bq->bufq_data;
141 		otype = bq->bufq_type;
142 
143 		while ((bp = bufq_impls[otype].impl_dequeue(odata)) != NULL)
144 			bufq_impls[type].impl_queue(data, bp);
145 
146 		bq->bufq_data = data;
147 		bq->bufq_type = type;
148 		bq->bufq_impl = &bufq_impls[type];
149 	} else {
150 		otype = type;
151 		odata = data;
152 	}
153 	mtx_leave(&bq->bufq_mtx);
154 
155 	bufq_impls[otype].impl_destroy(odata);
156 
157 	return (0);
158 }
159 
160 void
bufq_destroy(struct bufq * bq)161 bufq_destroy(struct bufq *bq)
162 {
163 	bufq_drain(bq);
164 
165 	bq->bufq_impl->impl_destroy(bq->bufq_data);
166 	bq->bufq_data = NULL;
167 
168 	mtx_enter(&bufqs_mtx);
169 	while (bufqs_stop) {
170 		msleep_nsec(&bufqs_stop, &bufqs_mtx, PRIBIO, "bqdest", INFSLP);
171 	}
172 	SLIST_REMOVE(&bufqs, bq, bufq, bufq_entries);
173 	mtx_leave(&bufqs_mtx);
174 }
175 
176 
177 void
bufq_queue(struct bufq * bq,struct buf * bp)178 bufq_queue(struct bufq *bq, struct buf *bp)
179 {
180 	mtx_enter(&bq->bufq_mtx);
181 	while (bq->bufq_stop) {
182 		msleep_nsec(&bq->bufq_stop, &bq->bufq_mtx, PRIBIO, "bqqueue",
183 		    INFSLP);
184 	}
185 
186 	bp->b_bq = bq;
187 	bq->bufq_outstanding++;
188 	bq->bufq_impl->impl_queue(bq->bufq_data, bp);
189 	mtx_leave(&bq->bufq_mtx);
190 }
191 
192 struct buf *
bufq_dequeue(struct bufq * bq)193 bufq_dequeue(struct bufq *bq)
194 {
195 	struct buf	*bp;
196 
197 	mtx_enter(&bq->bufq_mtx);
198 	bp = bq->bufq_impl->impl_dequeue(bq->bufq_data);
199 	mtx_leave(&bq->bufq_mtx);
200 
201 	return (bp);
202 }
203 
204 void
bufq_requeue(struct bufq * bq,struct buf * bp)205 bufq_requeue(struct bufq *bq, struct buf *bp)
206 {
207 	mtx_enter(&bq->bufq_mtx);
208 	bq->bufq_impl->impl_requeue(bq->bufq_data, bp);
209 	mtx_leave(&bq->bufq_mtx);
210 }
211 
212 int
bufq_peek(struct bufq * bq)213 bufq_peek(struct bufq *bq)
214 {
215 	int		rv;
216 
217 	mtx_enter(&bq->bufq_mtx);
218 	rv = bq->bufq_impl->impl_peek(bq->bufq_data);
219 	mtx_leave(&bq->bufq_mtx);
220 
221 	return (rv);
222 }
223 
224 void
bufq_drain(struct bufq * bq)225 bufq_drain(struct bufq *bq)
226 {
227 	struct buf	*bp;
228 	int		 s;
229 
230 	while ((bp = bufq_dequeue(bq)) != NULL) {
231 		bp->b_error = ENXIO;
232 		bp->b_flags |= B_ERROR;
233 		s = splbio();
234 		biodone(bp);
235 		splx(s);
236 	}
237 }
238 
239 void
bufq_wait(struct bufq * bq)240 bufq_wait(struct bufq *bq)
241 {
242 	if (bq->bufq_hi) {
243 		assertwaitok();
244 		mtx_enter(&bq->bufq_mtx);
245 		while (bq->bufq_outstanding >= bq->bufq_hi) {
246 			bq->bufq_waiting++;
247 			msleep_nsec(&bq->bufq_waiting, &bq->bufq_mtx,
248 			    PRIBIO, "bqwait", INFSLP);
249 			bq->bufq_waiting--;
250 		}
251 		mtx_leave(&bq->bufq_mtx);
252 	}
253 }
254 
255 void
bufq_done(struct bufq * bq,struct buf * bp)256 bufq_done(struct bufq *bq, struct buf *bp)
257 {
258 	mtx_enter(&bq->bufq_mtx);
259 	KASSERT(bq->bufq_outstanding > 0);
260 	bq->bufq_outstanding--;
261 	if (bq->bufq_stop && bq->bufq_outstanding == 0)
262 		wakeup(&bq->bufq_outstanding);
263 	if (bq->bufq_waiting && bq->bufq_outstanding < bq->bufq_low)
264 		wakeup(&bq->bufq_waiting);
265 	mtx_leave(&bq->bufq_mtx);
266 	bp->b_bq = NULL;
267 }
268 
269 void
bufq_quiesce(void)270 bufq_quiesce(void)
271 {
272 	struct bufq		*bq;
273 
274 	mtx_enter(&bufqs_mtx);
275 	bufqs_stop = 1;
276 	mtx_leave(&bufqs_mtx);
277 	/*
278 	 * We can safely walk the list since it can't be modified as
279 	 * long as bufqs_stop is non-zero.
280 	 */
281 	SLIST_FOREACH(bq, &bufqs, bufq_entries) {
282 		mtx_enter(&bq->bufq_mtx);
283 		bq->bufq_stop = 1;
284 		while (bq->bufq_outstanding) {
285 			msleep_nsec(&bq->bufq_outstanding, &bq->bufq_mtx,
286 			    PRIBIO, "bqquies", INFSLP);
287 		}
288 		mtx_leave(&bq->bufq_mtx);
289 	}
290 }
291 
292 void
bufq_restart(void)293 bufq_restart(void)
294 {
295 	struct bufq		*bq;
296 
297 	mtx_enter(&bufqs_mtx);
298 	SLIST_FOREACH(bq, &bufqs, bufq_entries) {
299 		mtx_enter(&bq->bufq_mtx);
300 		bq->bufq_stop = 0;
301 		wakeup(&bq->bufq_stop);
302 		mtx_leave(&bq->bufq_mtx);
303 	}
304 	bufqs_stop = 0;
305 	wakeup(&bufqs_stop);
306 	mtx_leave(&bufqs_mtx);
307 }
308 
309 
310 /*
311  * fifo implementation
312  */
313 
314 void *
bufq_fifo_create(void)315 bufq_fifo_create(void)
316 {
317 	struct bufq_fifo_head	*head;
318 
319 	head = malloc(sizeof(*head), M_DEVBUF, M_NOWAIT | M_ZERO);
320 	if (head == NULL)
321 		return (NULL);
322 
323 	SIMPLEQ_INIT(head);
324 
325 	return (head);
326 }
327 
328 void
bufq_fifo_destroy(void * data)329 bufq_fifo_destroy(void *data)
330 {
331 	struct bufq_fifo_head	*head = data;
332 
333 	free(head, M_DEVBUF, sizeof(*head));
334 }
335 
336 void
bufq_fifo_queue(void * data,struct buf * bp)337 bufq_fifo_queue(void *data, struct buf *bp)
338 {
339 	struct bufq_fifo_head	*head = data;
340 
341 	SIMPLEQ_INSERT_TAIL(head, bp, b_bufq.bufq_data_fifo.bqf_entries);
342 }
343 
344 struct buf *
bufq_fifo_dequeue(void * data)345 bufq_fifo_dequeue(void *data)
346 {
347 	struct bufq_fifo_head	*head = data;
348 	struct buf		*bp;
349 
350 	bp = SIMPLEQ_FIRST(head);
351 	if (bp != NULL)
352 		SIMPLEQ_REMOVE_HEAD(head, b_bufq.bufq_data_fifo.bqf_entries);
353 
354 	return (bp);
355 }
356 
357 void
bufq_fifo_requeue(void * data,struct buf * bp)358 bufq_fifo_requeue(void *data, struct buf *bp)
359 {
360 	struct bufq_fifo_head	*head = data;
361 
362 	SIMPLEQ_INSERT_HEAD(head, bp, b_bufq.bufq_data_fifo.bqf_entries);
363 }
364 
365 int
bufq_fifo_peek(void * data)366 bufq_fifo_peek(void *data)
367 {
368 	struct bufq_fifo_head	*head = data;
369 
370 	return (SIMPLEQ_FIRST(head) != NULL);
371 }
372 
373 /*
374  * nscan implementation
375  */
376 
377 #define BUF_INORDER(ba, bb) ((ba)->b_blkno < (bb)->b_blkno)
378 
379 #define dsentries b_bufq.bufq_data_nscan.bqf_entries
380 
381 struct bufq_nscan_data {
382 	struct bufq_nscan_head sorted;
383 	struct bufq_nscan_head fifo;
384 	int leftoverroom; /* Remaining number of buffer inserts allowed  */
385 };
386 
387 void bufq_nscan_resort(struct bufq_nscan_data *data);
388 void bufq_simple_nscan(struct bufq_nscan_head *, struct buf *);
389 
390 void
bufq_simple_nscan(struct bufq_nscan_head * head,struct buf * bp)391 bufq_simple_nscan(struct bufq_nscan_head *head, struct buf *bp)
392 {
393 	struct buf *cur, *prev;
394 
395 	prev = NULL;
396 	/*
397 	 * We look for the first slot where we would fit, then insert
398 	 * after the element we just passed.
399 	 */
400 	SIMPLEQ_FOREACH(cur, head, dsentries) {
401 		if (BUF_INORDER(bp, cur))
402 			break;
403 		prev = cur;
404 	}
405 	if (prev)
406 		SIMPLEQ_INSERT_AFTER(head, prev, bp, dsentries);
407 	else
408 		SIMPLEQ_INSERT_HEAD(head, bp, dsentries);
409 
410 }
411 
412 /*
413  * Take N elements from the fifo queue and sort them
414  */
415 void
bufq_nscan_resort(struct bufq_nscan_data * data)416 bufq_nscan_resort(struct bufq_nscan_data *data)
417 {
418 	struct bufq_nscan_head *fifo = &data->fifo;
419 	struct bufq_nscan_head *sorted = &data->sorted;
420 	int count, segmentsize = BUFQ_NSCAN_N;
421 	struct buf *bp;
422 
423 	for (count = 0; count < segmentsize; count++) {
424 		bp = SIMPLEQ_FIRST(fifo);
425 		if (!bp)
426 			break;
427 		SIMPLEQ_REMOVE_HEAD(fifo, dsentries);
428 		bufq_simple_nscan(sorted, bp);
429 	}
430 	data->leftoverroom = segmentsize - count;
431 }
432 
433 void *
bufq_nscan_create(void)434 bufq_nscan_create(void)
435 {
436 	struct bufq_nscan_data *data;
437 
438 	data = malloc(sizeof(*data), M_DEVBUF, M_NOWAIT | M_ZERO);
439 	if (!data)
440 		return NULL;
441 	SIMPLEQ_INIT(&data->sorted);
442 	SIMPLEQ_INIT(&data->fifo);
443 
444 	return data;
445 }
446 
447 void
bufq_nscan_destroy(void * vdata)448 bufq_nscan_destroy(void *vdata)
449 {
450 	struct bufq_nscan_data *data = vdata;
451 
452 	free(data, M_DEVBUF, sizeof(*data));
453 }
454 
455 void
bufq_nscan_queue(void * vdata,struct buf * bp)456 bufq_nscan_queue(void *vdata, struct buf *bp)
457 {
458 	struct bufq_nscan_data *data = vdata;
459 
460 	/*
461 	 * If the previous sorted segment was small, we will continue
462 	 * packing in bufs as long as they're in order.
463 	 */
464 	if (data->leftoverroom) {
465 		struct buf *next = SIMPLEQ_FIRST(&data->sorted);
466 		if (next && BUF_INORDER(next, bp)) {
467 			bufq_simple_nscan(&data->sorted, bp);
468 			data->leftoverroom--;
469 			return;
470 		}
471 	}
472 
473 	SIMPLEQ_INSERT_TAIL(&data->fifo, bp, dsentries);
474 
475 }
476 
477 struct buf *
bufq_nscan_dequeue(void * vdata)478 bufq_nscan_dequeue(void *vdata)
479 {
480 	struct bufq_nscan_data *data = vdata;
481 	struct bufq_nscan_head *sorted = &data->sorted;
482 	struct buf	*bp;
483 
484 	if (SIMPLEQ_FIRST(sorted) == NULL)
485 		bufq_nscan_resort(data);
486 
487 	bp = SIMPLEQ_FIRST(sorted);
488 	if (bp != NULL)
489 		SIMPLEQ_REMOVE_HEAD(sorted, dsentries);
490 
491 	return (bp);
492 }
493 
494 void
bufq_nscan_requeue(void * vdata,struct buf * bp)495 bufq_nscan_requeue(void *vdata, struct buf *bp)
496 {
497 	struct bufq_nscan_data *data = vdata;
498 
499 	SIMPLEQ_INSERT_HEAD(&data->fifo, bp, dsentries);
500 }
501 
502 int
bufq_nscan_peek(void * vdata)503 bufq_nscan_peek(void *vdata)
504 {
505 	struct bufq_nscan_data *data = vdata;
506 
507 	return (SIMPLEQ_FIRST(&data->sorted) != NULL) ||
508 	    (SIMPLEQ_FIRST(&data->fifo) != NULL);
509 }
510