1/* $NetBSD: vfs_bio.c,v 1.278 2018/11/24 17:52:39 maxv Exp $ */
2
3/*-
4 * Copyright (c) 2007, 2008, 2009 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Andrew Doran, and by Wasabi Systems, Inc.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29 * POSSIBILITY OF SUCH DAMAGE.
30 */
31
32/*-
33 * Copyright (c) 1982, 1986, 1989, 1993
34 * The Regents of the University of California. All rights reserved.
35 * (c) UNIX System Laboratories, Inc.
36 * All or some portions of this file are derived from material licensed
37 * to the University of California by American Telephone and Telegraph
38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
39 * the permission of UNIX System Laboratories, Inc.
40 *
41 * Redistribution and use in source and binary forms, with or without
42 * modification, are permitted provided that the following conditions
43 * are met:
44 * 1. Redistributions of source code must retain the above copyright
45 * notice, this list of conditions and the following disclaimer.
46 * 2. Redistributions in binary form must reproduce the above copyright
47 * notice, this list of conditions and the following disclaimer in the
48 * documentation and/or other materials provided with the distribution.
49 * 3. Neither the name of the University nor the names of its contributors
50 * may be used to endorse or promote products derived from this software
51 * without specific prior written permission.
52 *
53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63 * SUCH DAMAGE.
64 *
65 * @(#)vfs_bio.c 8.6 (Berkeley) 1/11/94
66 */
67
68/*-
69 * Copyright (c) 1994 Christopher G. Demetriou
70 *
71 * Redistribution and use in source and binary forms, with or without
72 * modification, are permitted provided that the following conditions
73 * are met:
74 * 1. Redistributions of source code must retain the above copyright
75 * notice, this list of conditions and the following disclaimer.
76 * 2. Redistributions in binary form must reproduce the above copyright
77 * notice, this list of conditions and the following disclaimer in the
78 * documentation and/or other materials provided with the distribution.
79 * 3. All advertising materials mentioning features or use of this software
80 * must display the following acknowledgement:
81 * This product includes software developed by the University of
82 * California, Berkeley and its contributors.
83 * 4. Neither the name of the University nor the names of its contributors
84 * may be used to endorse or promote products derived from this software
85 * without specific prior written permission.
86 *
87 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
88 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
89 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
90 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
91 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
92 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
93 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
94 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
95 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
96 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
97 * SUCH DAMAGE.
98 *
99 * @(#)vfs_bio.c 8.6 (Berkeley) 1/11/94
100 */
101
102/*
103 * The buffer cache subsystem.
104 *
105 * Some references:
106 * Bach: The Design of the UNIX Operating System (Prentice Hall, 1986)
107 * Leffler, et al.: The Design and Implementation of the 4.3BSD
108 * UNIX Operating System (Addison Welley, 1989)
109 *
110 * Locking
111 *
112 * There are three locks:
113 * - bufcache_lock: protects global buffer cache state.
114 * - BC_BUSY: a long term per-buffer lock.
115 * - buf_t::b_objlock: lock on completion (biowait vs biodone).
116 *
117 * For buffers associated with vnodes (a most common case) b_objlock points
118 * to the vnode_t::v_interlock. Otherwise, it points to generic buffer_lock.
119 *
120 * Lock order:
121 * bufcache_lock ->
122 * buf_t::b_objlock
123 */
124
125#include <sys/cdefs.h>
126__KERNEL_RCSID(0, "$NetBSD: vfs_bio.c,v 1.278 2018/11/24 17:52:39 maxv Exp $");
127
128#ifdef _KERNEL_OPT
129#include "opt_bufcache.h"
130#include "opt_dtrace.h"
131#include "opt_biohist.h"
132#endif
133
134#include <sys/param.h>
135#include <sys/systm.h>
136#include <sys/kernel.h>
137#include <sys/proc.h>
138#include <sys/buf.h>
139#include <sys/vnode.h>
140#include <sys/mount.h>
141#include <sys/resourcevar.h>
142#include <sys/sysctl.h>
143#include <sys/conf.h>
144#include <sys/kauth.h>
145#include <sys/fstrans.h>
146#include <sys/intr.h>
147#include <sys/cpu.h>
148#include <sys/wapbl.h>
149#include <sys/bitops.h>
150#include <sys/cprng.h>
151#include <sys/sdt.h>
152
153#include <uvm/uvm.h> /* extern struct uvm uvm */
154
155#include <miscfs/specfs/specdev.h>
156
157#ifndef BUFPAGES
158# define BUFPAGES 0
159#endif
160
161#ifdef BUFCACHE
162# if (BUFCACHE < 5) || (BUFCACHE > 95)
163# error BUFCACHE is not between 5 and 95
164# endif
165#else
166# define BUFCACHE 15
167#endif
168
169u_int nbuf; /* desired number of buffer headers */
170u_int bufpages = BUFPAGES; /* optional hardwired count */
171u_int bufcache = BUFCACHE; /* max % of RAM to use for buffer cache */
172
173/*
174 * Definitions for the buffer free lists.
175 */
176#define BQUEUES 3 /* number of free buffer queues */
177
178#define BQ_LOCKED 0 /* super-blocks &c */
179#define BQ_LRU 1 /* lru, useful buffers */
180#define BQ_AGE 2 /* rubbish */
181
182struct bqueue {
183 TAILQ_HEAD(, buf) bq_queue;
184 uint64_t bq_bytes;
185 buf_t *bq_marker;
186};
187static struct bqueue bufqueues[BQUEUES];
188
189/* Function prototypes */
190static void buf_setwm(void);
191static int buf_trim(void);
192static void *bufpool_page_alloc(struct pool *, int);
193static void bufpool_page_free(struct pool *, void *);
194static buf_t *bio_doread(struct vnode *, daddr_t, int, int);
195static buf_t *getnewbuf(int, int, int);
196static int buf_lotsfree(void);
197static int buf_canrelease(void);
198static u_long buf_mempoolidx(u_long);
199static u_long buf_roundsize(u_long);
200static void *buf_alloc(size_t);
201static void buf_mrelease(void *, size_t);
202static void binsheadfree(buf_t *, struct bqueue *);
203static void binstailfree(buf_t *, struct bqueue *);
204#ifdef DEBUG
205static int checkfreelist(buf_t *, struct bqueue *, int);
206#endif
207static void biointr(void *);
208static void biodone2(buf_t *);
209static void bref(buf_t *);
210static void brele(buf_t *);
211static void sysctl_kern_buf_setup(void);
212static void sysctl_vm_buf_setup(void);
213
214/* Initialization for biohist */
215
216#include <sys/biohist.h>
217
218BIOHIST_DEFINE(biohist);
219
220void
221biohist_init(void)
222{
223
224 BIOHIST_INIT(biohist, BIOHIST_SIZE);
225}
226
227/*
228 * Definitions for the buffer hash lists.
229 */
230#define BUFHASH(dvp, lbn) \
231 (&bufhashtbl[(((long)(dvp) >> 8) + (int)(lbn)) & bufhash])
232LIST_HEAD(bufhashhdr, buf) *bufhashtbl, invalhash;
233u_long bufhash;
234
235static kcondvar_t needbuffer_cv;
236
237/*
238 * Buffer queue lock.
239 */
240kmutex_t bufcache_lock;
241kmutex_t buffer_lock;
242
243/* Software ISR for completed transfers. */
244static void *biodone_sih;
245
246/* Buffer pool for I/O buffers. */
247static pool_cache_t buf_cache;
248static pool_cache_t bufio_cache;
249
250#define MEMPOOL_INDEX_OFFSET (ilog2(DEV_BSIZE)) /* smallest pool is 512 bytes */
251#define NMEMPOOLS (ilog2(MAXBSIZE) - MEMPOOL_INDEX_OFFSET + 1)
252__CTASSERT((1 << (NMEMPOOLS + MEMPOOL_INDEX_OFFSET - 1)) == MAXBSIZE);
253
254/* Buffer memory pools */
255static struct pool bmempools[NMEMPOOLS];
256
257static struct vm_map *buf_map;
258
259/*
260 * Buffer memory pool allocator.
261 */
262static void *
263bufpool_page_alloc(struct pool *pp, int flags)
264{
265
266 return (void *)uvm_km_alloc(buf_map,
267 MAXBSIZE, MAXBSIZE,
268 ((flags & PR_WAITOK) ? 0 : UVM_KMF_NOWAIT|UVM_KMF_TRYLOCK)
269 | UVM_KMF_WIRED);
270}
271
272static void
273bufpool_page_free(struct pool *pp, void *v)
274{
275
276 uvm_km_free(buf_map, (vaddr_t)v, MAXBSIZE, UVM_KMF_WIRED);
277}
278
279static struct pool_allocator bufmempool_allocator = {
280 .pa_alloc = bufpool_page_alloc,
281 .pa_free = bufpool_page_free,
282 .pa_pagesz = MAXBSIZE,
283};
284
285/* Buffer memory management variables */
286u_long bufmem_valimit;
287u_long bufmem_hiwater;
288u_long bufmem_lowater;
289u_long bufmem;
290
291/*
292 * MD code can call this to set a hard limit on the amount
293 * of virtual memory used by the buffer cache.
294 */
295int
296buf_setvalimit(vsize_t sz)
297{
298
299 /* We need to accommodate at least NMEMPOOLS of MAXBSIZE each */
300 if (sz < NMEMPOOLS * MAXBSIZE)
301 return EINVAL;
302
303 bufmem_valimit = sz;
304 return 0;
305}
306
307static void
308buf_setwm(void)
309{
310
311 bufmem_hiwater = buf_memcalc();
312 /* lowater is approx. 2% of memory (with bufcache = 15) */
313#define BUFMEM_WMSHIFT 3
314#define BUFMEM_HIWMMIN (64 * 1024 << BUFMEM_WMSHIFT)
315 if (bufmem_hiwater < BUFMEM_HIWMMIN)
316 /* Ensure a reasonable minimum value */
317 bufmem_hiwater = BUFMEM_HIWMMIN;
318 bufmem_lowater = bufmem_hiwater >> BUFMEM_WMSHIFT;
319}
320
321#ifdef DEBUG
322int debug_verify_freelist = 0;
323static int
324checkfreelist(buf_t *bp, struct bqueue *dp, int ison)
325{
326 buf_t *b;
327
328 if (!debug_verify_freelist)
329 return 1;
330
331 TAILQ_FOREACH(b, &dp->bq_queue, b_freelist) {
332 if (b == bp)
333 return ison ? 1 : 0;
334 }
335
336 return ison ? 0 : 1;
337}
338#endif
339
340/*
341 * Insq/Remq for the buffer hash lists.
342 * Call with buffer queue locked.
343 */
344static void
345binsheadfree(buf_t *bp, struct bqueue *dp)
346{
347
348 KASSERT(mutex_owned(&bufcache_lock));
349 KASSERT(bp->b_freelistindex == -1);
350 TAILQ_INSERT_HEAD(&dp->bq_queue, bp, b_freelist);
351 dp->bq_bytes += bp->b_bufsize;
352 bp->b_freelistindex = dp - bufqueues;
353}
354
355static void
356binstailfree(buf_t *bp, struct bqueue *dp)
357{
358
359 KASSERT(mutex_owned(&bufcache_lock));
360 KASSERTMSG(bp->b_freelistindex == -1, "double free of buffer? "
361 "bp=%p, b_freelistindex=%d\n", bp, bp->b_freelistindex);
362 TAILQ_INSERT_TAIL(&dp->bq_queue, bp, b_freelist);
363 dp->bq_bytes += bp->b_bufsize;
364 bp->b_freelistindex = dp - bufqueues;
365}
366
367void
368bremfree(buf_t *bp)
369{
370 struct bqueue *dp;
371 int bqidx = bp->b_freelistindex;
372
373 KASSERT(mutex_owned(&bufcache_lock));
374
375 KASSERT(bqidx != -1);
376 dp = &bufqueues[bqidx];
377 KDASSERT(checkfreelist(bp, dp, 1));
378 KASSERT(dp->bq_bytes >= bp->b_bufsize);
379 TAILQ_REMOVE(&dp->bq_queue, bp, b_freelist);
380 dp->bq_bytes -= bp->b_bufsize;
381
382 /* For the sysctl helper. */
383 if (bp == dp->bq_marker)
384 dp->bq_marker = NULL;
385
386#if defined(DIAGNOSTIC)
387 bp->b_freelistindex = -1;
388#endif /* defined(DIAGNOSTIC) */
389}
390
391/*
392 * Add a reference to an buffer structure that came from buf_cache.
393 */
394static inline void
395bref(buf_t *bp)
396{
397
398 KASSERT(mutex_owned(&bufcache_lock));
399 KASSERT(bp->b_refcnt > 0);
400
401 bp->b_refcnt++;
402}
403
404/*
405 * Free an unused buffer structure that came from buf_cache.
406 */
407static inline void
408brele(buf_t *bp)
409{
410
411 KASSERT(mutex_owned(&bufcache_lock));
412 KASSERT(bp->b_refcnt > 0);
413
414 if (bp->b_refcnt-- == 1) {
415 buf_destroy(bp);
416#ifdef DEBUG
417 memset((char *)bp, 0, sizeof(*bp));
418#endif
419 pool_cache_put(buf_cache, bp);
420 }
421}
422
423/*
424 * note that for some ports this is used by pmap bootstrap code to
425 * determine kva size.
426 */
427u_long
428buf_memcalc(void)
429{
430 u_long n;
431 vsize_t mapsz = 0;
432
433 /*
434 * Determine the upper bound of memory to use for buffers.
435 *
436 * - If bufpages is specified, use that as the number
437 * pages.
438 *
439 * - Otherwise, use bufcache as the percentage of
440 * physical memory.
441 */
442 if (bufpages != 0) {
443 n = bufpages;
444 } else {
445 if (bufcache < 5) {
446 printf("forcing bufcache %d -> 5", bufcache);
447 bufcache = 5;
448 }
449 if (bufcache > 95) {
450 printf("forcing bufcache %d -> 95", bufcache);
451 bufcache = 95;
452 }
453 if (buf_map != NULL)
454 mapsz = vm_map_max(buf_map) - vm_map_min(buf_map);
455 n = calc_cache_size(mapsz, bufcache,
456 (buf_map != kernel_map) ? 100 : BUFCACHE_VA_MAXPCT)
457 / PAGE_SIZE;
458 }
459
460 n <<= PAGE_SHIFT;
461 if (bufmem_valimit != 0 && n > bufmem_valimit)
462 n = bufmem_valimit;
463
464 return (n);
465}
466
467/*
468 * Initialize buffers and hash links for buffers.
469 */
470void
471bufinit(void)
472{
473 struct bqueue *dp;
474 int use_std;
475 u_int i;
476
477 biodone_vfs = biodone;
478
479 mutex_init(&bufcache_lock, MUTEX_DEFAULT, IPL_NONE);
480 mutex_init(&buffer_lock, MUTEX_DEFAULT, IPL_NONE);
481 cv_init(&needbuffer_cv, "needbuf");
482
483 if (bufmem_valimit != 0) {
484 vaddr_t minaddr = 0, maxaddr;
485 buf_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
486 bufmem_valimit, 0, false, 0);
487 if (buf_map == NULL)
488 panic("bufinit: cannot allocate submap");
489 } else
490 buf_map = kernel_map;
491
492 /*
493 * Initialize buffer cache memory parameters.
494 */
495 bufmem = 0;
496 buf_setwm();
497
498 /* On "small" machines use small pool page sizes where possible */
499 use_std = (physmem < atop(16*1024*1024));
500
501 /*
502 * Also use them on systems that can map the pool pages using
503 * a direct-mapped segment.
504 */
505#ifdef PMAP_MAP_POOLPAGE
506 use_std = 1;
507#endif
508
509 buf_cache = pool_cache_init(sizeof(buf_t), 0, 0, 0,
510 "bufpl", NULL, IPL_SOFTBIO, NULL, NULL, NULL);
511 bufio_cache = pool_cache_init(sizeof(buf_t), 0, 0, 0,
512 "biopl", NULL, IPL_BIO, NULL, NULL, NULL);
513
514 for (i = 0; i < NMEMPOOLS; i++) {
515 struct pool_allocator *pa;
516 struct pool *pp = &bmempools[i];
517 u_int size = 1 << (i + MEMPOOL_INDEX_OFFSET);
518 char *name = kmem_alloc(8, KM_SLEEP); /* XXX: never freed */
519 if (__predict_false(size >= 1048576))
520 (void)snprintf(name, 8, "buf%um", size / 1048576);
521 else if (__predict_true(size >= 1024))
522 (void)snprintf(name, 8, "buf%uk", size / 1024);
523 else
524 (void)snprintf(name, 8, "buf%ub", size);
525 pa = (size <= PAGE_SIZE && use_std)
526 ? &pool_allocator_nointr
527 : &bufmempool_allocator;
528 pool_init(pp, size, 0, 0, 0, name, pa, IPL_NONE);
529 pool_setlowat(pp, 1);
530 pool_sethiwat(pp, 1);
531 }
532
533 /* Initialize the buffer queues */
534 for (dp = bufqueues; dp < &bufqueues[BQUEUES]; dp++) {
535 TAILQ_INIT(&dp->bq_queue);
536 dp->bq_bytes = 0;
537 }
538
539 /*
540 * Estimate hash table size based on the amount of memory we
541 * intend to use for the buffer cache. The average buffer
542 * size is dependent on our clients (i.e. filesystems).
543 *
544 * For now, use an empirical 3K per buffer.
545 */
546 nbuf = (bufmem_hiwater / 1024) / 3;
547 bufhashtbl = hashinit(nbuf, HASH_LIST, true, &bufhash);
548
549 sysctl_kern_buf_setup();
550 sysctl_vm_buf_setup();
551}
552
553void
554bufinit2(void)
555{
556
557 biodone_sih = softint_establish(SOFTINT_BIO | SOFTINT_MPSAFE, biointr,
558 NULL);
559 if (biodone_sih == NULL)
560 panic("bufinit2: can't establish soft interrupt");
561}
562
563static int
564buf_lotsfree(void)
565{
566 u_long guess;
567
568 /* Always allocate if less than the low water mark. */
569 if (bufmem < bufmem_lowater)
570 return 1;
571
572 /* Never allocate if greater than the high water mark. */
573 if (bufmem > bufmem_hiwater)
574 return 0;
575
576 /* If there's anything on the AGE list, it should be eaten. */
577 if (TAILQ_FIRST(&bufqueues[BQ_AGE].bq_queue) != NULL)
578 return 0;
579
580 /*
581 * The probabily of getting a new allocation is inversely
582 * proportional to the current size of the cache above
583 * the low water mark. Divide the total first to avoid overflows
584 * in the product.
585 */
586 guess = cprng_fast32() % 16;
587
588 if ((bufmem_hiwater - bufmem_lowater) / 16 * guess >=
589 (bufmem - bufmem_lowater))
590 return 1;
591
592 /* Otherwise don't allocate. */
593 return 0;
594}
595
596/*
597 * Return estimate of bytes we think need to be
598 * released to help resolve low memory conditions.
599 *
600 * => called with bufcache_lock held.
601 */
602static int
603buf_canrelease(void)
604{
605 int pagedemand, ninvalid = 0;
606
607 KASSERT(mutex_owned(&bufcache_lock));
608
609 if (bufmem < bufmem_lowater)
610 return 0;
611
612 if (bufmem > bufmem_hiwater)
613 return bufmem - bufmem_hiwater;
614
615 ninvalid += bufqueues[BQ_AGE].bq_bytes;
616
617 pagedemand = uvmexp.freetarg - uvmexp.free;
618 if (pagedemand < 0)
619 return ninvalid;
620 return MAX(ninvalid, MIN(2 * MAXBSIZE,
621 MIN((bufmem - bufmem_lowater) / 16, pagedemand * PAGE_SIZE)));
622}
623
624/*
625 * Buffer memory allocation helper functions
626 */
627static u_long
628buf_mempoolidx(u_long size)
629{
630 u_int n = 0;
631
632 size -= 1;
633 size >>= MEMPOOL_INDEX_OFFSET;
634 while (size) {
635 size >>= 1;
636 n += 1;
637 }
638 if (n >= NMEMPOOLS)
639 panic("buf mem pool index %d", n);
640 return n;
641}
642
643static u_long
644buf_roundsize(u_long size)
645{
646 /* Round up to nearest power of 2 */
647 return (1 << (buf_mempoolidx(size) + MEMPOOL_INDEX_OFFSET));
648}
649
650static void *
651buf_alloc(size_t size)
652{
653 u_int n = buf_mempoolidx(size);
654 void *addr;
655
656 while (1) {
657 addr = pool_get(&bmempools[n], PR_NOWAIT);
658 if (addr != NULL)
659 break;
660
661 /* No memory, see if we can free some. If so, try again */
662 mutex_enter(&bufcache_lock);
663 if (buf_drain(1) > 0) {
664 mutex_exit(&bufcache_lock);
665 continue;
666 }
667
668 if (curlwp == uvm.pagedaemon_lwp) {
669 mutex_exit(&bufcache_lock);
670 return NULL;
671 }
672
673 /* Wait for buffers to arrive on the LRU queue */
674 cv_timedwait(&needbuffer_cv, &bufcache_lock, hz / 4);
675 mutex_exit(&bufcache_lock);
676 }
677
678 return addr;
679}
680
681static void
682buf_mrelease(void *addr, size_t size)
683{
684
685 pool_put(&bmempools[buf_mempoolidx(size)], addr);
686}
687
688/*
689 * bread()/breadn() helper.
690 */
691static buf_t *
692bio_doread(struct vnode *vp, daddr_t blkno, int size, int async)
693{
694 buf_t *bp;
695 struct mount *mp;
696
697 bp = getblk(vp, blkno, size, 0, 0);
698
699 /*
700 * getblk() may return NULL if we are the pagedaemon.
701 */
702 if (bp == NULL) {
703 KASSERT(curlwp == uvm.pagedaemon_lwp);
704 return NULL;
705 }
706
707 /*
708 * If buffer does not have data valid, start a read.
709 * Note that if buffer is BC_INVAL, getblk() won't return it.
710 * Therefore, it's valid if its I/O has completed or been delayed.
711 */
712 if (!ISSET(bp->b_oflags, (BO_DONE | BO_DELWRI))) {
713 /* Start I/O for the buffer. */
714 SET(bp->b_flags, B_READ | async);
715 if (async)
716 BIO_SETPRIO(bp, BPRIO_TIMELIMITED);
717 else
718 BIO_SETPRIO(bp, BPRIO_TIMECRITICAL);
719 VOP_STRATEGY(vp, bp);
720
721 /* Pay for the read. */
722 curlwp->l_ru.ru_inblock++;
723 } else if (async)
724 brelse(bp, 0);
725
726 if (vp->v_type == VBLK)
727 mp = spec_node_getmountedfs(vp);
728 else
729 mp = vp->v_mount;
730
731 /*
732 * Collect statistics on synchronous and asynchronous reads.
733 * Reads from block devices are charged to their associated
734 * filesystem (if any).
735 */
736 if (mp != NULL) {
737 if (async == 0)
738 mp->mnt_stat.f_syncreads++;
739 else
740 mp->mnt_stat.f_asyncreads++;
741 }
742
743 return (bp);
744}
745
746/*
747 * Read a disk block.
748 * This algorithm described in Bach (p.54).
749 */
750int
751bread(struct vnode *vp, daddr_t blkno, int size, int flags, buf_t **bpp)
752{
753 buf_t *bp;
754 int error;
755
756 BIOHIST_FUNC(__func__); BIOHIST_CALLED(biohist);
757
758 /* Get buffer for block. */
759 bp = *bpp = bio_doread(vp, blkno, size, 0);
760 if (bp == NULL)
761 return ENOMEM;
762
763 /* Wait for the read to complete, and return result. */
764 error = biowait(bp);
765 if (error == 0 && (flags & B_MODIFY) != 0)
766 error = fscow_run(bp, true);
767 if (error) {
768 brelse(bp, 0);
769 *bpp = NULL;
770 }
771
772 return error;
773}
774
775/*
776 * Read-ahead multiple disk blocks. The first is sync, the rest async.
777 * Trivial modification to the breada algorithm presented in Bach (p.55).
778 */
779int
780breadn(struct vnode *vp, daddr_t blkno, int size, daddr_t *rablks,
781 int *rasizes, int nrablks, int flags, buf_t **bpp)
782{
783 buf_t *bp;
784 int error, i;
785
786 BIOHIST_FUNC(__func__); BIOHIST_CALLED(biohist);
787
788 bp = *bpp = bio_doread(vp, blkno, size, 0);
789 if (bp == NULL)
790 return ENOMEM;
791
792 /*
793 * For each of the read-ahead blocks, start a read, if necessary.
794 */
795 mutex_enter(&bufcache_lock);
796 for (i = 0; i < nrablks; i++) {
797 /* If it's in the cache, just go on to next one. */
798 if (incore(vp, rablks[i]))
799 continue;
800
801 /* Get a buffer for the read-ahead block */
802 mutex_exit(&bufcache_lock);
803 (void) bio_doread(vp, rablks[i], rasizes[i], B_ASYNC);
804 mutex_enter(&bufcache_lock);
805 }
806 mutex_exit(&bufcache_lock);
807
808 /* Otherwise, we had to start a read for it; wait until it's valid. */
809 error = biowait(bp);
810 if (error == 0 && (flags & B_MODIFY) != 0)
811 error = fscow_run(bp, true);
812 if (error) {
813 brelse(bp, 0);
814 *bpp = NULL;
815 }
816
817 return error;
818}
819
820/*
821 * Block write. Described in Bach (p.56)
822 */
823int
824bwrite(buf_t *bp)
825{
826 int rv, sync, wasdelayed;
827 struct vnode *vp;
828 struct mount *mp;
829
830 BIOHIST_FUNC(__func__); BIOHIST_CALLARGS(biohist, "bp=%#jx",
831 (uintptr_t)bp, 0, 0, 0);
832
833 KASSERT(ISSET(bp->b_cflags, BC_BUSY));
834 KASSERT(!cv_has_waiters(&bp->b_done));
835
836 vp = bp->b_vp;
837
838 /*
839 * dholland 20160728 AFAICT vp==NULL must be impossible as it
840 * will crash upon reaching VOP_STRATEGY below... see further
841 * analysis on tech-kern.
842 */
843 KASSERTMSG(vp != NULL, "bwrite given buffer with null vnode");
844
845 if (vp != NULL) {
846 KASSERT(bp->b_objlock == vp->v_interlock);
847 if (vp->v_type == VBLK)
848 mp = spec_node_getmountedfs(vp);
849 else
850 mp = vp->v_mount;
851 } else {
852 mp = NULL;
853 }
854
855 if (mp && mp->mnt_wapbl) {
856 if (bp->b_iodone != mp->mnt_wapbl_op->wo_wapbl_biodone) {
857 bdwrite(bp);
858 return 0;
859 }
860 }
861
862 /*
863 * Remember buffer type, to switch on it later. If the write was
864 * synchronous, but the file system was mounted with MNT_ASYNC,
865 * convert it to a delayed write.
866 * XXX note that this relies on delayed tape writes being converted
867 * to async, not sync writes (which is safe, but ugly).
868 */
869 sync = !ISSET(bp->b_flags, B_ASYNC);
870 if (sync && mp != NULL && ISSET(mp->mnt_flag, MNT_ASYNC)) {
871 bdwrite(bp);
872 return (0);
873 }
874
875 /*
876 * Collect statistics on synchronous and asynchronous writes.
877 * Writes to block devices are charged to their associated
878 * filesystem (if any).
879 */
880 if (mp != NULL) {
881 if (sync)
882 mp->mnt_stat.f_syncwrites++;
883 else
884 mp->mnt_stat.f_asyncwrites++;
885 }
886
887 /*
888 * Pay for the I/O operation and make sure the buf is on the correct
889 * vnode queue.
890 */
891 bp->b_error = 0;
892 wasdelayed = ISSET(bp->b_oflags, BO_DELWRI);
893 CLR(bp->b_flags, B_READ);
894 if (wasdelayed) {
895 mutex_enter(&bufcache_lock);
896 mutex_enter(bp->b_objlock);
897 CLR(bp->b_oflags, BO_DONE | BO_DELWRI);
898 reassignbuf(bp, bp->b_vp);
899 mutex_exit(&bufcache_lock);
900 } else {
901 curlwp->l_ru.ru_oublock++;
902 mutex_enter(bp->b_objlock);
903 CLR(bp->b_oflags, BO_DONE | BO_DELWRI);
904 }
905 if (vp != NULL)
906 vp->v_numoutput++;
907 mutex_exit(bp->b_objlock);
908
909 /* Initiate disk write. */
910 if (sync)
911 BIO_SETPRIO(bp, BPRIO_TIMECRITICAL);
912 else
913 BIO_SETPRIO(bp, BPRIO_TIMELIMITED);
914
915 VOP_STRATEGY(vp, bp);
916
917 if (sync) {
918 /* If I/O was synchronous, wait for it to complete. */
919 rv = biowait(bp);
920
921 /* Release the buffer. */
922 brelse(bp, 0);
923
924 return (rv);
925 } else {
926 return (0);
927 }
928}
929
930int
931vn_bwrite(void *v)
932{
933 struct vop_bwrite_args *ap = v;
934
935 return (bwrite(ap->a_bp));
936}
937
938/*
939 * Delayed write.
940 *
941 * The buffer is marked dirty, but is not queued for I/O.
942 * This routine should be used when the buffer is expected
943 * to be modified again soon, typically a small write that
944 * partially fills a buffer.
945 *
946 * NB: magnetic tapes cannot be delayed; they must be
947 * written in the order that the writes are requested.
948 *
949 * Described in Leffler, et al. (pp. 208-213).
950 */
951void
952bdwrite(buf_t *bp)
953{
954
955 BIOHIST_FUNC(__func__); BIOHIST_CALLARGS(biohist, "bp=%#jx",
956 (uintptr_t)bp, 0, 0, 0);
957
958 KASSERT(bp->b_vp == NULL || bp->b_vp->v_tag != VT_UFS ||
959 bp->b_vp->v_type == VBLK || ISSET(bp->b_flags, B_COWDONE));
960 KASSERT(ISSET(bp->b_cflags, BC_BUSY));
961 KASSERT(!cv_has_waiters(&bp->b_done));
962
963 /* If this is a tape block, write the block now. */
964 if (bdev_type(bp->b_dev) == D_TAPE) {
965 bawrite(bp);
966 return;
967 }
968
969 if (wapbl_vphaswapbl(bp->b_vp)) {
970 struct mount *mp = wapbl_vptomp(bp->b_vp);
971
972 if (bp->b_iodone != mp->mnt_wapbl_op->wo_wapbl_biodone) {
973 WAPBL_ADD_BUF(mp, bp);
974 }
975 }
976
977 /*
978 * If the block hasn't been seen before:
979 * (1) Mark it as having been seen,
980 * (2) Charge for the write,
981 * (3) Make sure it's on its vnode's correct block list.
982 */
983 KASSERT(bp->b_vp == NULL || bp->b_objlock == bp->b_vp->v_interlock);
984
985 if (!ISSET(bp->b_oflags, BO_DELWRI)) {
986 mutex_enter(&bufcache_lock);
987 mutex_enter(bp->b_objlock);
988 SET(bp->b_oflags, BO_DELWRI);
989 curlwp->l_ru.ru_oublock++;
990 reassignbuf(bp, bp->b_vp);
991 mutex_exit(&bufcache_lock);
992 } else {
993 mutex_enter(bp->b_objlock);
994 }
995 /* Otherwise, the "write" is done, so mark and release the buffer. */
996 CLR(bp->b_oflags, BO_DONE);
997 mutex_exit(bp->b_objlock);
998
999 brelse(bp, 0);
1000}
1001
1002/*
1003 * Asynchronous block write; just an asynchronous bwrite().
1004 */
1005void
1006bawrite(buf_t *bp)
1007{
1008
1009 KASSERT(ISSET(bp->b_cflags, BC_BUSY));
1010 KASSERT(bp->b_vp != NULL);
1011
1012 SET(bp->b_flags, B_ASYNC);
1013 VOP_BWRITE(bp->b_vp, bp);
1014}
1015
1016/*
1017 * Release a buffer on to the free lists.
1018 * Described in Bach (p. 46).
1019 */
1020void
1021brelsel(buf_t *bp, int set)
1022{
1023 struct bqueue *bufq;
1024 struct vnode *vp;
1025
1026 KASSERT(bp != NULL);
1027 KASSERT(mutex_owned(&bufcache_lock));
1028 KASSERT(!cv_has_waiters(&bp->b_done));
1029 KASSERT(bp->b_refcnt > 0);
1030
1031 SET(bp->b_cflags, set);
1032
1033 KASSERT(ISSET(bp->b_cflags, BC_BUSY));
1034 KASSERT(bp->b_iodone == NULL);
1035
1036 /* Wake up any processes waiting for any buffer to become free. */
1037 cv_signal(&needbuffer_cv);
1038
1039 /* Wake up any proceeses waiting for _this_ buffer to become free */
1040 if (ISSET(bp->b_cflags, BC_WANTED))
1041 CLR(bp->b_cflags, BC_WANTED|BC_AGE);
1042
1043 /* If it's clean clear the copy-on-write flag. */
1044 if (ISSET(bp->b_flags, B_COWDONE)) {
1045 mutex_enter(bp->b_objlock);
1046 if (!ISSET(bp->b_oflags, BO_DELWRI))
1047 CLR(bp->b_flags, B_COWDONE);
1048 mutex_exit(bp->b_objlock);
1049 }
1050
1051 /*
1052 * Determine which queue the buffer should be on, then put it there.
1053 */
1054
1055 /* If it's locked, don't report an error; try again later. */
1056 if (ISSET(bp->b_flags, B_LOCKED))
1057 bp->b_error = 0;
1058
1059 /* If it's not cacheable, or an error, mark it invalid. */
1060 if (ISSET(bp->b_cflags, BC_NOCACHE) || bp->b_error != 0)
1061 SET(bp->b_cflags, BC_INVAL);
1062
1063 if (ISSET(bp->b_cflags, BC_VFLUSH)) {
1064 /*
1065 * This is a delayed write buffer that was just flushed to
1066 * disk. It is still on the LRU queue. If it's become
1067 * invalid, then we need to move it to a different queue;
1068 * otherwise leave it in its current position.
1069 */
1070 CLR(bp->b_cflags, BC_VFLUSH);
1071 if (!ISSET(bp->b_cflags, BC_INVAL|BC_AGE) &&
1072 !ISSET(bp->b_flags, B_LOCKED) && bp->b_error == 0) {
1073 KDASSERT(checkfreelist(bp, &bufqueues[BQ_LRU], 1));
1074 goto already_queued;
1075 } else {
1076 bremfree(bp);
1077 }
1078 }
1079
1080 KDASSERT(checkfreelist(bp, &bufqueues[BQ_AGE], 0));
1081 KDASSERT(checkfreelist(bp, &bufqueues[BQ_LRU], 0));
1082 KDASSERT(checkfreelist(bp, &bufqueues[BQ_LOCKED], 0));
1083
1084 if ((bp->b_bufsize <= 0) || ISSET(bp->b_cflags, BC_INVAL)) {
1085 /*
1086 * If it's invalid or empty, dissociate it from its vnode
1087 * and put on the head of the appropriate queue.
1088 */
1089 if (ISSET(bp->b_flags, B_LOCKED)) {
1090 if (wapbl_vphaswapbl(vp = bp->b_vp)) {
1091 struct mount *mp = wapbl_vptomp(vp);
1092
1093 KASSERT(bp->b_iodone
1094 != mp->mnt_wapbl_op->wo_wapbl_biodone);
1095 WAPBL_REMOVE_BUF(mp, bp);
1096 }
1097 }
1098
1099 mutex_enter(bp->b_objlock);
1100 CLR(bp->b_oflags, BO_DONE|BO_DELWRI);
1101 if ((vp = bp->b_vp) != NULL) {
1102 KASSERT(bp->b_objlock == vp->v_interlock);
1103 reassignbuf(bp, bp->b_vp);
1104 brelvp(bp);
1105 mutex_exit(vp->v_interlock);
1106 } else {
1107 KASSERT(bp->b_objlock == &buffer_lock);
1108 mutex_exit(bp->b_objlock);
1109 }
1110
1111 if (bp->b_bufsize <= 0)
1112 /* no data */
1113 goto already_queued;
1114 else
1115 /* invalid data */
1116 bufq = &bufqueues[BQ_AGE];
1117 binsheadfree(bp, bufq);
1118 } else {
1119 /*
1120 * It has valid data. Put it on the end of the appropriate
1121 * queue, so that it'll stick around for as long as possible.
1122 * If buf is AGE, but has dependencies, must put it on last
1123 * bufqueue to be scanned, ie LRU. This protects against the
1124 * livelock where BQ_AGE only has buffers with dependencies,
1125 * and we thus never get to the dependent buffers in BQ_LRU.
1126 */
1127 if (ISSET(bp->b_flags, B_LOCKED)) {
1128 /* locked in core */
1129 bufq = &bufqueues[BQ_LOCKED];
1130 } else if (!ISSET(bp->b_cflags, BC_AGE)) {
1131 /* valid data */
1132 bufq = &bufqueues[BQ_LRU];
1133 } else {
1134 /* stale but valid data */
1135 bufq = &bufqueues[BQ_AGE];
1136 }
1137 binstailfree(bp, bufq);
1138 }
1139already_queued:
1140 /* Unlock the buffer. */
1141 CLR(bp->b_cflags, BC_AGE|BC_BUSY|BC_NOCACHE);
1142 CLR(bp->b_flags, B_ASYNC);
1143 cv_broadcast(&bp->b_busy);
1144
1145 if (bp->b_bufsize <= 0)
1146 brele(bp);
1147}
1148
1149void
1150brelse(buf_t *bp, int set)
1151{
1152
1153 mutex_enter(&bufcache_lock);
1154 brelsel(bp, set);
1155 mutex_exit(&bufcache_lock);
1156}
1157
1158/*
1159 * Determine if a block is in the cache.
1160 * Just look on what would be its hash chain. If it's there, return
1161 * a pointer to it, unless it's marked invalid. If it's marked invalid,
1162 * we normally don't return the buffer, unless the caller explicitly
1163 * wants us to.
1164 */
1165buf_t *
1166incore(struct vnode *vp, daddr_t blkno)
1167{
1168 buf_t *bp;
1169
1170 KASSERT(mutex_owned(&bufcache_lock));
1171
1172 /* Search hash chain */
1173 LIST_FOREACH(bp, BUFHASH(vp, blkno), b_hash) {
1174 if (bp->b_lblkno == blkno && bp->b_vp == vp &&
1175 !ISSET(bp->b_cflags, BC_INVAL)) {
1176 KASSERT(bp->b_objlock == vp->v_interlock);
1177 return (bp);
1178 }
1179 }
1180
1181 return (NULL);
1182}
1183
1184/*
1185 * Get a block of requested size that is associated with
1186 * a given vnode and block offset. If it is found in the
1187 * block cache, mark it as having been found, make it busy
1188 * and return it. Otherwise, return an empty block of the
1189 * correct size. It is up to the caller to insure that the
1190 * cached blocks be of the correct size.
1191 */
1192buf_t *
1193getblk(struct vnode *vp, daddr_t blkno, int size, int slpflag, int slptimeo)
1194{
1195 int err, preserve;
1196 buf_t *bp;
1197
1198 mutex_enter(&bufcache_lock);
1199 loop:
1200 bp = incore(vp, blkno);
1201 if (bp != NULL) {
1202 err = bbusy(bp, ((slpflag & PCATCH) != 0), slptimeo, NULL);
1203 if (err != 0) {
1204 if (err == EPASSTHROUGH)
1205 goto loop;
1206 mutex_exit(&bufcache_lock);
1207 return (NULL);
1208 }
1209 KASSERT(!cv_has_waiters(&bp->b_done));
1210#ifdef DIAGNOSTIC
1211 if (ISSET(bp->b_oflags, BO_DONE|BO_DELWRI) &&
1212 bp->b_bcount < size && vp->v_type != VBLK)
1213 panic("getblk: block size invariant failed");
1214#endif
1215 bremfree(bp);
1216 preserve = 1;
1217 } else {
1218 if ((bp = getnewbuf(slpflag, slptimeo, 0)) == NULL)
1219 goto loop;
1220
1221 if (incore(vp, blkno) != NULL) {
1222 /* The block has come into memory in the meantime. */
1223 brelsel(bp, 0);
1224 goto loop;
1225 }
1226
1227 LIST_INSERT_HEAD(BUFHASH(vp, blkno), bp, b_hash);
1228 bp->b_blkno = bp->b_lblkno = bp->b_rawblkno = blkno;
1229 mutex_enter(vp->v_interlock);
1230 bgetvp(vp, bp);
1231 mutex_exit(vp->v_interlock);
1232 preserve = 0;
1233 }
1234 mutex_exit(&bufcache_lock);
1235
1236 /*
1237 * LFS can't track total size of B_LOCKED buffer (locked_queue_bytes)
1238 * if we re-size buffers here.
1239 */
1240 if (ISSET(bp->b_flags, B_LOCKED)) {
1241 KASSERT(bp->b_bufsize >= size);
1242 } else {
1243 if (allocbuf(bp, size, preserve)) {
1244 mutex_enter(&bufcache_lock);
1245 LIST_REMOVE(bp, b_hash);
1246 brelsel(bp, BC_INVAL);
1247 mutex_exit(&bufcache_lock);
1248 return NULL;
1249 }
1250 }
1251 BIO_SETPRIO(bp, BPRIO_DEFAULT);
1252 return (bp);
1253}
1254
1255/*
1256 * Get an empty, disassociated buffer of given size.
1257 */
1258buf_t *
1259geteblk(int size)
1260{
1261 buf_t *bp;
1262 int error __diagused;
1263
1264 mutex_enter(&bufcache_lock);
1265 while ((bp = getnewbuf(0, 0, 0)) == NULL)
1266 ;
1267
1268 SET(bp->b_cflags, BC_INVAL);
1269 LIST_INSERT_HEAD(&invalhash, bp, b_hash);
1270 mutex_exit(&bufcache_lock);
1271 BIO_SETPRIO(bp, BPRIO_DEFAULT);
1272 error = allocbuf(bp, size, 0);
1273 KASSERT(error == 0);
1274 return (bp);
1275}
1276
1277/*
1278 * Expand or contract the actual memory allocated to a buffer.
1279 *
1280 * If the buffer shrinks, data is lost, so it's up to the
1281 * caller to have written it out *first*; this routine will not
1282 * start a write. If the buffer grows, it's the callers
1283 * responsibility to fill out the buffer's additional contents.
1284 */
1285int
1286allocbuf(buf_t *bp, int size, int preserve)
1287{
1288 void *addr;
1289 vsize_t oldsize, desired_size;
1290 int oldcount;
1291 int delta;
1292
1293 desired_size = buf_roundsize(size);
1294 if (desired_size > MAXBSIZE)
1295 printf("allocbuf: buffer larger than MAXBSIZE requested");
1296
1297 oldcount = bp->b_bcount;
1298
1299 bp->b_bcount = size;
1300
1301 oldsize = bp->b_bufsize;
1302 if (oldsize == desired_size) {
1303 /*
1304 * Do not short cut the WAPBL resize, as the buffer length
1305 * could still have changed and this would corrupt the
1306 * tracking of the transaction length.
1307 */
1308 goto out;
1309 }
1310
1311 /*
1312 * If we want a buffer of a different size, re-allocate the
1313 * buffer's memory; copy old content only if needed.
1314 */
1315 addr = buf_alloc(desired_size);
1316 if (addr == NULL)
1317 return ENOMEM;
1318 if (preserve)
1319 memcpy(addr, bp->b_data, MIN(oldsize,desired_size));
1320 if (bp->b_data != NULL)
1321 buf_mrelease(bp->b_data, oldsize);
1322 bp->b_data = addr;
1323 bp->b_bufsize = desired_size;
1324
1325 /*
1326 * Update overall buffer memory counter (protected by bufcache_lock)
1327 */
1328 delta = (long)desired_size - (long)oldsize;
1329
1330 mutex_enter(&bufcache_lock);
1331 if ((bufmem += delta) > bufmem_hiwater) {
1332 /*
1333 * Need to trim overall memory usage.
1334 */
1335 while (buf_canrelease()) {
1336 if (curcpu()->ci_schedstate.spc_flags &
1337 SPCF_SHOULDYIELD) {
1338 mutex_exit(&bufcache_lock);
1339 preempt();
1340 mutex_enter(&bufcache_lock);
1341 }
1342 if (buf_trim() == 0)
1343 break;
1344 }
1345 }
1346 mutex_exit(&bufcache_lock);
1347
1348 out:
1349 if (wapbl_vphaswapbl(bp->b_vp))
1350 WAPBL_RESIZE_BUF(wapbl_vptomp(bp->b_vp), bp, oldsize, oldcount);
1351
1352 return 0;
1353}
1354
1355/*
1356 * Find a buffer which is available for use.
1357 * Select something from a free list.
1358 * Preference is to AGE list, then LRU list.
1359 *
1360 * Called with the buffer queues locked.
1361 * Return buffer locked.
1362 */
1363static buf_t *
1364getnewbuf(int slpflag, int slptimeo, int from_bufq)
1365{
1366 buf_t *bp;
1367 struct vnode *vp;
1368 struct mount *transmp = NULL;
1369
1370 start:
1371 KASSERT(mutex_owned(&bufcache_lock));
1372
1373 /*
1374 * Get a new buffer from the pool.
1375 */
1376 if (!from_bufq && buf_lotsfree()) {
1377 mutex_exit(&bufcache_lock);
1378 bp = pool_cache_get(buf_cache, PR_NOWAIT);
1379 if (bp != NULL) {
1380 memset((char *)bp, 0, sizeof(*bp));
1381 buf_init(bp);
1382 SET(bp->b_cflags, BC_BUSY); /* mark buffer busy */
1383 mutex_enter(&bufcache_lock);
1384#if defined(DIAGNOSTIC)
1385 bp->b_freelistindex = -1;
1386#endif /* defined(DIAGNOSTIC) */
1387 return (bp);
1388 }
1389 mutex_enter(&bufcache_lock);
1390 }
1391
1392 KASSERT(mutex_owned(&bufcache_lock));
1393 if ((bp = TAILQ_FIRST(&bufqueues[BQ_AGE].bq_queue)) != NULL) {
1394 KASSERT(!ISSET(bp->b_oflags, BO_DELWRI));
1395 } else {
1396 TAILQ_FOREACH(bp, &bufqueues[BQ_LRU].bq_queue, b_freelist) {
1397 if (ISSET(bp->b_cflags, BC_VFLUSH) ||
1398 !ISSET(bp->b_oflags, BO_DELWRI))
1399 break;
1400 if (fstrans_start_nowait(bp->b_vp->v_mount) == 0) {
1401 KASSERT(transmp == NULL);
1402 transmp = bp->b_vp->v_mount;
1403 break;
1404 }
1405 }
1406 }
1407 if (bp != NULL) {
1408 KASSERT(!ISSET(bp->b_cflags, BC_BUSY) || ISSET(bp->b_cflags, BC_VFLUSH));
1409 bremfree(bp);
1410
1411 /* Buffer is no longer on free lists. */
1412 SET(bp->b_cflags, BC_BUSY);
1413 } else {
1414 /*
1415 * XXX: !from_bufq should be removed.
1416 */
1417 if (!from_bufq || curlwp != uvm.pagedaemon_lwp) {
1418 /* wait for a free buffer of any kind */
1419 if ((slpflag & PCATCH) != 0)
1420 (void)cv_timedwait_sig(&needbuffer_cv,
1421 &bufcache_lock, slptimeo);
1422 else
1423 (void)cv_timedwait(&needbuffer_cv,
1424 &bufcache_lock, slptimeo);
1425 }
1426 return (NULL);
1427 }
1428
1429#ifdef DIAGNOSTIC
1430 if (bp->b_bufsize <= 0)
1431 panic("buffer %p: on queue but empty", bp);
1432#endif
1433
1434 if (ISSET(bp->b_cflags, BC_VFLUSH)) {
1435 /*
1436 * This is a delayed write buffer being flushed to disk. Make
1437 * sure it gets aged out of the queue when it's finished, and
1438 * leave it off the LRU queue.
1439 */
1440 CLR(bp->b_cflags, BC_VFLUSH);
1441 SET(bp->b_cflags, BC_AGE);
1442 goto start;
1443 }
1444
1445 KASSERT(ISSET(bp->b_cflags, BC_BUSY));
1446 KASSERT(bp->b_refcnt > 0);
1447 KASSERT(!cv_has_waiters(&bp->b_done));
1448
1449 /*
1450 * If buffer was a delayed write, start it and return NULL
1451 * (since we might sleep while starting the write).
1452 */
1453 if (ISSET(bp->b_oflags, BO_DELWRI)) {
1454 /*
1455 * This buffer has gone through the LRU, so make sure it gets
1456 * reused ASAP.
1457 */
1458 SET(bp->b_cflags, BC_AGE);
1459 mutex_exit(&bufcache_lock);
1460 bawrite(bp);
1461 KASSERT(transmp != NULL);
1462 fstrans_done(transmp);
1463 mutex_enter(&bufcache_lock);
1464 return (NULL);
1465 }
1466
1467 KASSERT(transmp == NULL);
1468
1469 vp = bp->b_vp;
1470
1471 /* clear out various other fields */
1472 bp->b_cflags = BC_BUSY;
1473 bp->b_oflags = 0;
1474 bp->b_flags = 0;
1475 bp->b_dev = NODEV;
1476 bp->b_blkno = 0;
1477 bp->b_lblkno = 0;
1478 bp->b_rawblkno = 0;
1479 bp->b_iodone = 0;
1480 bp->b_error = 0;
1481 bp->b_resid = 0;
1482 bp->b_bcount = 0;
1483
1484 LIST_REMOVE(bp, b_hash);
1485
1486 /* Disassociate us from our vnode, if we had one... */
1487 if (vp != NULL) {
1488 mutex_enter(vp->v_interlock);
1489 brelvp(bp);
1490 mutex_exit(vp->v_interlock);
1491 }
1492
1493 return (bp);
1494}
1495
1496/*
1497 * Attempt to free an aged buffer off the queues.
1498 * Called with queue lock held.
1499 * Returns the amount of buffer memory freed.
1500 */
1501static int
1502buf_trim(void)
1503{
1504 buf_t *bp;
1505 long size;
1506
1507 KASSERT(mutex_owned(&bufcache_lock));
1508
1509 /* Instruct getnewbuf() to get buffers off the queues */
1510 if ((bp = getnewbuf(PCATCH, 1, 1)) == NULL)
1511 return 0;
1512
1513 KASSERT((bp->b_cflags & BC_WANTED) == 0);
1514 size = bp->b_bufsize;
1515 bufmem -= size;
1516 if (size > 0) {
1517 buf_mrelease(bp->b_data, size);
1518 bp->b_bcount = bp->b_bufsize = 0;
1519 }
1520 /* brelse() will return the buffer to the global buffer pool */
1521 brelsel(bp, 0);
1522 return size;
1523}
1524
1525int
1526buf_drain(int n)
1527{
1528 int size = 0, sz;
1529
1530 KASSERT(mutex_owned(&bufcache_lock));
1531
1532 while (size < n && bufmem > bufmem_lowater) {
1533 sz = buf_trim();
1534 if (sz <= 0)
1535 break;
1536 size += sz;
1537 }
1538
1539 return size;
1540}
1541
1542SDT_PROVIDER_DEFINE(io);
1543
1544SDT_PROBE_DEFINE1(io, kernel, , wait__start, "struct buf *"/*bp*/);
1545SDT_PROBE_DEFINE1(io, kernel, , wait__done, "struct buf *"/*bp*/);
1546
1547/*
1548 * Wait for operations on the buffer to complete.
1549 * When they do, extract and return the I/O's error value.
1550 */
1551int
1552biowait(buf_t *bp)
1553{
1554
1555 BIOHIST_FUNC(__func__);
1556
1557 KASSERT(ISSET(bp->b_cflags, BC_BUSY));
1558 KASSERT(bp->b_refcnt > 0);
1559
1560 SDT_PROBE1(io, kernel, , wait__start, bp);
1561
1562 mutex_enter(bp->b_objlock);
1563
1564 BIOHIST_CALLARGS(biohist, "bp=%#jx, oflags=0x%jx, ret_addr=%#jx",
1565 (uintptr_t)bp, bp->b_oflags,
1566 (uintptr_t)__builtin_return_address(0), 0);
1567
1568 while (!ISSET(bp->b_oflags, BO_DONE | BO_DELWRI)) {
1569 BIOHIST_LOG(biohist, "waiting bp=%#jx", (uintptr_t)bp, 0, 0, 0);
1570 cv_wait(&bp->b_done, bp->b_objlock);
1571 }
1572 mutex_exit(bp->b_objlock);
1573
1574 SDT_PROBE1(io, kernel, , wait__done, bp);
1575
1576 BIOHIST_LOG(biohist, "return %jd", bp->b_error, 0, 0, 0);
1577
1578 return bp->b_error;
1579}
1580
1581/*
1582 * Mark I/O complete on a buffer.
1583 *
1584 * If a callback has been requested, e.g. the pageout
1585 * daemon, do so. Otherwise, awaken waiting processes.
1586 *
1587 * [ Leffler, et al., says on p.247:
1588 * "This routine wakes up the blocked process, frees the buffer
1589 * for an asynchronous write, or, for a request by the pagedaemon
1590 * process, invokes a procedure specified in the buffer structure" ]
1591 *
1592 * In real life, the pagedaemon (or other system processes) wants
1593 * to do async stuff too, and doesn't want the buffer brelse()'d.
1594 * (for swap pager, that puts swap buffers on the free lists (!!!),
1595 * for the vn device, that puts allocated buffers on the free lists!)
1596 */
1597void
1598biodone(buf_t *bp)
1599{
1600 int s;
1601
1602 BIOHIST_FUNC(__func__);
1603
1604 KASSERT(!ISSET(bp->b_oflags, BO_DONE));
1605
1606 if (cpu_intr_p()) {
1607 /* From interrupt mode: defer to a soft interrupt. */
1608 s = splvm();
1609 TAILQ_INSERT_TAIL(&curcpu()->ci_data.cpu_biodone, bp, b_actq);
1610
1611 BIOHIST_CALLARGS(biohist, "bp=%#jx, softint scheduled",
1612 (uintptr_t)bp, 0, 0, 0);
1613 softint_schedule(biodone_sih);
1614 splx(s);
1615 } else {
1616 /* Process now - the buffer may be freed soon. */
1617 biodone2(bp);
1618 }
1619}
1620
1621SDT_PROBE_DEFINE1(io, kernel, , done, "struct buf *"/*bp*/);
1622
1623static void
1624biodone2(buf_t *bp)
1625{
1626 void (*callout)(buf_t *);
1627
1628 SDT_PROBE1(io, kernel, ,done, bp);
1629
1630 BIOHIST_FUNC(__func__);
1631 BIOHIST_CALLARGS(biohist, "bp=%#jx", (uintptr_t)bp, 0, 0, 0);
1632
1633 mutex_enter(bp->b_objlock);
1634 /* Note that the transfer is done. */
1635 if (ISSET(bp->b_oflags, BO_DONE))
1636 panic("biodone2 already");
1637 CLR(bp->b_flags, B_COWDONE);
1638 SET(bp->b_oflags, BO_DONE);
1639 BIO_SETPRIO(bp, BPRIO_DEFAULT);
1640
1641 /* Wake up waiting writers. */
1642 if (!ISSET(bp->b_flags, B_READ))
1643 vwakeup(bp);
1644
1645 if ((callout = bp->b_iodone) != NULL) {
1646 BIOHIST_LOG(biohist, "callout %#jx", (uintptr_t)callout,
1647 0, 0, 0);
1648
1649 /* Note callout done, then call out. */
1650 KASSERT(!cv_has_waiters(&bp->b_done));
1651 KERNEL_LOCK(1, NULL); /* XXXSMP */
1652 bp->b_iodone = NULL;
1653 mutex_exit(bp->b_objlock);
1654 (*callout)(bp);
1655 KERNEL_UNLOCK_ONE(NULL); /* XXXSMP */
1656 } else if (ISSET(bp->b_flags, B_ASYNC)) {
1657 /* If async, release. */
1658 BIOHIST_LOG(biohist, "async", 0, 0, 0, 0);
1659 KASSERT(!cv_has_waiters(&bp->b_done));
1660 mutex_exit(bp->b_objlock);
1661 brelse(bp, 0);
1662 } else {
1663 /* Otherwise just wake up waiters in biowait(). */
1664 BIOHIST_LOG(biohist, "wake-up", 0, 0, 0, 0);
1665 cv_broadcast(&bp->b_done);
1666 mutex_exit(bp->b_objlock);
1667 }
1668}
1669
1670static void
1671biointr(void *cookie)
1672{
1673 struct cpu_info *ci;
1674 buf_t *bp;
1675 int s;
1676
1677 BIOHIST_FUNC(__func__); BIOHIST_CALLED(biohist);
1678
1679 ci = curcpu();
1680
1681 s = splvm();
1682 while (!TAILQ_EMPTY(&ci->ci_data.cpu_biodone)) {
1683 KASSERT(curcpu() == ci);
1684
1685 bp = TAILQ_FIRST(&ci->ci_data.cpu_biodone);
1686 TAILQ_REMOVE(&ci->ci_data.cpu_biodone, bp, b_actq);
1687 splx(s);
1688
1689 BIOHIST_LOG(biohist, "bp=%#jx", (uintptr_t)bp, 0, 0, 0);
1690 biodone2(bp);
1691
1692 s = splvm();
1693 }
1694 splx(s);
1695}
1696
1697/*
1698 * Wait for all buffers to complete I/O
1699 * Return the number of "stuck" buffers.
1700 */
1701int
1702buf_syncwait(void)
1703{
1704 buf_t *bp;
1705 int iter, nbusy, nbusy_prev = 0, ihash;
1706
1707 BIOHIST_FUNC(__func__); BIOHIST_CALLED(biohist);
1708
1709 for (iter = 0; iter < 20;) {
1710 mutex_enter(&bufcache_lock);
1711 nbusy = 0;
1712 for (ihash = 0; ihash < bufhash+1; ihash++) {
1713 LIST_FOREACH(bp, &bufhashtbl[ihash], b_hash) {
1714 if ((bp->b_cflags & (BC_BUSY|BC_INVAL)) == BC_BUSY)
1715 nbusy += ((bp->b_flags & B_READ) == 0);
1716 }
1717 }
1718 mutex_exit(&bufcache_lock);
1719
1720 if (nbusy == 0)
1721 break;
1722 if (nbusy_prev == 0)
1723 nbusy_prev = nbusy;
1724 printf("%d ", nbusy);
1725 kpause("bflush", false, MAX(1, hz / 25 * iter), NULL);
1726 if (nbusy >= nbusy_prev) /* we didn't flush anything */
1727 iter++;
1728 else
1729 nbusy_prev = nbusy;
1730 }
1731
1732 if (nbusy) {
1733#if defined(DEBUG) || defined(DEBUG_HALT_BUSY)
1734 printf("giving up\nPrinting vnodes for busy buffers\n");
1735 for (ihash = 0; ihash < bufhash+1; ihash++) {
1736 LIST_FOREACH(bp, &bufhashtbl[ihash], b_hash) {
1737 if ((bp->b_cflags & (BC_BUSY|BC_INVAL)) == BC_BUSY &&
1738 (bp->b_flags & B_READ) == 0)
1739 vprint(NULL, bp->b_vp);
1740 }
1741 }
1742#endif
1743 }
1744
1745 return nbusy;
1746}
1747
1748static void
1749sysctl_fillbuf(const buf_t *i, struct buf_sysctl *o)
1750{
1751 const bool allowaddr = get_expose_address(curproc);
1752
1753 memset(o, 0, sizeof(*o));
1754
1755 o->b_flags = i->b_flags | i->b_cflags | i->b_oflags;
1756 o->b_error = i->b_error;
1757 o->b_prio = i->b_prio;
1758 o->b_dev = i->b_dev;
1759 o->b_bufsize = i->b_bufsize;
1760 o->b_bcount = i->b_bcount;
1761 o->b_resid = i->b_resid;
1762 COND_SET_VALUE(o->b_addr, PTRTOUINT64(i->b_data), allowaddr);
1763 o->b_blkno = i->b_blkno;
1764 o->b_rawblkno = i->b_rawblkno;
1765 COND_SET_VALUE(o->b_iodone, PTRTOUINT64(i->b_iodone), allowaddr);
1766 COND_SET_VALUE(o->b_proc, PTRTOUINT64(i->b_proc), allowaddr);
1767 COND_SET_VALUE(o->b_vp, PTRTOUINT64(i->b_vp), allowaddr);
1768 COND_SET_VALUE(o->b_saveaddr, PTRTOUINT64(i->b_saveaddr), allowaddr);
1769 o->b_lblkno = i->b_lblkno;
1770}
1771
1772#define KERN_BUFSLOP 20
1773static int
1774sysctl_dobuf(SYSCTLFN_ARGS)
1775{
1776 buf_t *bp;
1777 struct buf_sysctl bs;
1778 struct bqueue *bq;
1779 char *dp;
1780 u_int i, op, arg;
1781 size_t len, needed, elem_size, out_size;
1782 int error, elem_count, retries;
1783
1784 if (namelen == 1 && name[0] == CTL_QUERY)
1785 return (sysctl_query(SYSCTLFN_CALL(rnode)));
1786
1787 if (namelen != 4)
1788 return (EINVAL);
1789
1790 retries = 100;
1791 retry:
1792 dp = oldp;
1793 len = (oldp != NULL) ? *oldlenp : 0;
1794 op = name[0];
1795 arg = name[1];
1796 elem_size = name[2];
1797 elem_count = name[3];
1798 out_size = MIN(sizeof(bs), elem_size);
1799
1800 /*
1801 * at the moment, these are just "placeholders" to make the
1802 * API for retrieving kern.buf data more extensible in the
1803 * future.
1804 *
1805 * XXX kern.buf currently has "netbsd32" issues. hopefully
1806 * these will be resolved at a later point.
1807 */
1808 if (op != KERN_BUF_ALL || arg != KERN_BUF_ALL ||
1809 elem_size < 1 || elem_count < 0)
1810 return (EINVAL);
1811
1812 error = 0;
1813 needed = 0;
1814 sysctl_unlock();
1815 mutex_enter(&bufcache_lock);
1816 for (i = 0; i < BQUEUES; i++) {
1817 bq = &bufqueues[i];
1818 TAILQ_FOREACH(bp, &bq->bq_queue, b_freelist) {
1819 bq->bq_marker = bp;
1820 if (len >= elem_size && elem_count > 0) {
1821 sysctl_fillbuf(bp, &bs);
1822 mutex_exit(&bufcache_lock);
1823 error = copyout(&bs, dp, out_size);
1824 mutex_enter(&bufcache_lock);
1825 if (error)
1826 break;
1827 if (bq->bq_marker != bp) {
1828 /*
1829 * This sysctl node is only for
1830 * statistics. Retry; if the
1831 * queue keeps changing, then
1832 * bail out.
1833 */
1834 if (retries-- == 0) {
1835 error = EAGAIN;
1836 break;
1837 }
1838 mutex_exit(&bufcache_lock);
1839 sysctl_relock();
1840 goto retry;
1841 }
1842 dp += elem_size;
1843 len -= elem_size;
1844 }
1845 needed += elem_size;
1846 if (elem_count > 0 && elem_count != INT_MAX)
1847 elem_count--;
1848 }
1849 if (error != 0)
1850 break;
1851 }
1852 mutex_exit(&bufcache_lock);
1853 sysctl_relock();
1854
1855 *oldlenp = needed;
1856 if (oldp == NULL)
1857 *oldlenp += KERN_BUFSLOP * sizeof(buf_t);
1858
1859 return (error);
1860}
1861
1862static int
1863sysctl_bufvm_update(SYSCTLFN_ARGS)
1864{
1865 int error, rv;
1866 struct sysctlnode node;
1867 unsigned int temp_bufcache;
1868 unsigned long temp_water;
1869
1870 /* Take a copy of the supplied node and its data */
1871 node = *rnode;
1872 if (node.sysctl_data == &bufcache) {
1873 node.sysctl_data = &temp_bufcache;
1874 temp_bufcache = *(unsigned int *)rnode->sysctl_data;
1875 } else {
1876 node.sysctl_data = &temp_water;
1877 temp_water = *(unsigned long *)rnode->sysctl_data;
1878 }
1879
1880 /* Update the copy */
1881 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1882 if (error || newp == NULL)
1883 return (error);
1884
1885 if (rnode->sysctl_data == &bufcache) {
1886 if (temp_bufcache > 100)
1887 return (EINVAL);
1888 bufcache = temp_bufcache;
1889 buf_setwm();
1890 } else if (rnode->sysctl_data == &bufmem_lowater) {
1891 if (bufmem_hiwater - temp_water < 16)
1892 return (EINVAL);
1893 bufmem_lowater = temp_water;
1894 } else if (rnode->sysctl_data == &bufmem_hiwater) {
1895 if (temp_water - bufmem_lowater < 16)
1896 return (EINVAL);
1897 bufmem_hiwater = temp_water;
1898 } else
1899 return (EINVAL);
1900
1901 /* Drain until below new high water mark */
1902 sysctl_unlock();
1903 mutex_enter(&bufcache_lock);
1904 while (bufmem > bufmem_hiwater) {
1905 rv = buf_drain((bufmem - bufmem_hiwater) / (2 * 1024));
1906 if (rv <= 0)
1907 break;
1908 }
1909 mutex_exit(&bufcache_lock);
1910 sysctl_relock();
1911
1912 return 0;
1913}
1914
1915static struct sysctllog *vfsbio_sysctllog;
1916
1917static void
1918sysctl_kern_buf_setup(void)
1919{
1920
1921 sysctl_createv(&vfsbio_sysctllog, 0, NULL, NULL,
1922 CTLFLAG_PERMANENT,
1923 CTLTYPE_NODE, "buf",
1924 SYSCTL_DESCR("Kernel buffer cache information"),
1925 sysctl_dobuf, 0, NULL, 0,
1926 CTL_KERN, KERN_BUF, CTL_EOL);
1927}
1928
1929static void
1930sysctl_vm_buf_setup(void)
1931{
1932
1933 sysctl_createv(&vfsbio_sysctllog, 0, NULL, NULL,
1934 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1935 CTLTYPE_INT, "bufcache",
1936 SYSCTL_DESCR("Percentage of physical memory to use for "
1937 "buffer cache"),
1938 sysctl_bufvm_update, 0, &bufcache, 0,
1939 CTL_VM, CTL_CREATE, CTL_EOL);
1940 sysctl_createv(&vfsbio_sysctllog, 0, NULL, NULL,
1941 CTLFLAG_PERMANENT|CTLFLAG_READONLY,
1942 CTLTYPE_LONG, "bufmem",
1943 SYSCTL_DESCR("Amount of kernel memory used by buffer "
1944 "cache"),
1945 NULL, 0, &bufmem, 0,
1946 CTL_VM, CTL_CREATE, CTL_EOL);
1947 sysctl_createv(&vfsbio_sysctllog, 0, NULL, NULL,
1948 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1949 CTLTYPE_LONG, "bufmem_lowater",
1950 SYSCTL_DESCR("Minimum amount of kernel memory to "
1951 "reserve for buffer cache"),
1952 sysctl_bufvm_update, 0, &bufmem_lowater, 0,
1953 CTL_VM, CTL_CREATE, CTL_EOL);
1954 sysctl_createv(&vfsbio_sysctllog, 0, NULL, NULL,
1955 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1956 CTLTYPE_LONG, "bufmem_hiwater",
1957 SYSCTL_DESCR("Maximum amount of kernel memory to use "
1958 "for buffer cache"),
1959 sysctl_bufvm_update, 0, &bufmem_hiwater, 0,
1960 CTL_VM, CTL_CREATE, CTL_EOL);
1961}
1962
1963#ifdef DEBUG
1964/*
1965 * Print out statistics on the current allocation of the buffer pool.
1966 * Can be enabled to print out on every ``sync'' by setting "syncprt"
1967 * in vfs_syscalls.c using sysctl.
1968 */
1969void
1970vfs_bufstats(void)
1971{
1972 int i, j, count;
1973 buf_t *bp;
1974 struct bqueue *dp;
1975 int counts[MAXBSIZE / MIN_PAGE_SIZE + 1];
1976 static const char *bname[BQUEUES] = { "LOCKED", "LRU", "AGE" };
1977
1978 for (dp = bufqueues, i = 0; dp < &bufqueues[BQUEUES]; dp++, i++) {
1979 count = 0;
1980 memset(counts, 0, sizeof(counts));
1981 TAILQ_FOREACH(bp, &dp->bq_queue, b_freelist) {
1982 counts[bp->b_bufsize / PAGE_SIZE]++;
1983 count++;
1984 }
1985 printf("%s: total-%d", bname[i], count);
1986 for (j = 0; j <= MAXBSIZE / PAGE_SIZE; j++)
1987 if (counts[j] != 0)
1988 printf(", %d-%d", j * PAGE_SIZE, counts[j]);
1989 printf("\n");
1990 }
1991}
1992#endif /* DEBUG */
1993
1994/* ------------------------------ */
1995
1996buf_t *
1997getiobuf(struct vnode *vp, bool waitok)
1998{
1999 buf_t *bp;
2000
2001 bp = pool_cache_get(bufio_cache, (waitok ? PR_WAITOK : PR_NOWAIT));
2002 if (bp == NULL)
2003 return bp;
2004
2005 buf_init(bp);
2006
2007 if ((bp->b_vp = vp) != NULL) {
2008 bp->b_objlock = vp->v_interlock;
2009 } else {
2010 KASSERT(bp->b_objlock == &buffer_lock);
2011 }
2012
2013 return bp;
2014}
2015
2016void
2017putiobuf(buf_t *bp)
2018{
2019
2020 buf_destroy(bp);
2021 pool_cache_put(bufio_cache, bp);
2022}
2023
2024/*
2025 * nestiobuf_iodone: b_iodone callback for nested buffers.
2026 */
2027
2028void
2029nestiobuf_iodone(buf_t *bp)
2030{
2031 buf_t *mbp = bp->b_private;
2032 int error;
2033 int donebytes;
2034
2035 KASSERT(bp->b_bcount <= bp->b_bufsize);
2036 KASSERT(mbp != bp);
2037
2038 error = bp->b_error;
2039 if (bp->b_error == 0 &&
2040 (bp->b_bcount < bp->b_bufsize || bp->b_resid > 0)) {
2041 /*
2042 * Not all got transfered, raise an error. We have no way to
2043 * propagate these conditions to mbp.
2044 */
2045 error = EIO;
2046 }
2047
2048 donebytes = bp->b_bufsize;
2049
2050 putiobuf(bp);
2051 nestiobuf_done(mbp, donebytes, error);
2052}
2053
2054/*
2055 * nestiobuf_setup: setup a "nested" buffer.
2056 *
2057 * => 'mbp' is a "master" buffer which is being divided into sub pieces.
2058 * => 'bp' should be a buffer allocated by getiobuf.
2059 * => 'offset' is a byte offset in the master buffer.
2060 * => 'size' is a size in bytes of this nested buffer.
2061 */
2062
2063void
2064nestiobuf_setup(buf_t *mbp, buf_t *bp, int offset, size_t size)
2065{
2066 const int b_pass = mbp->b_flags & (B_READ|B_MEDIA_FLAGS);
2067 struct vnode *vp = mbp->b_vp;
2068
2069 KASSERT(mbp->b_bcount >= offset + size);
2070 bp->b_vp = vp;
2071 bp->b_dev = mbp->b_dev;
2072 bp->b_objlock = mbp->b_objlock;
2073 bp->b_cflags = BC_BUSY;
2074 bp->b_flags = B_ASYNC | b_pass;
2075 bp->b_iodone = nestiobuf_iodone;
2076 bp->b_data = (char *)mbp->b_data + offset;
2077 bp->b_resid = bp->b_bcount = size;
2078 bp->b_bufsize = bp->b_bcount;
2079 bp->b_private = mbp;
2080 BIO_COPYPRIO(bp, mbp);
2081 if (BUF_ISWRITE(bp) && vp != NULL) {
2082 mutex_enter(vp->v_interlock);
2083 vp->v_numoutput++;
2084 mutex_exit(vp->v_interlock);
2085 }
2086}
2087
2088/*
2089 * nestiobuf_done: propagate completion to the master buffer.
2090 *
2091 * => 'donebytes' specifies how many bytes in the 'mbp' is completed.
2092 * => 'error' is an errno(2) that 'donebytes' has been completed with.
2093 */
2094
2095void
2096nestiobuf_done(buf_t *mbp, int donebytes, int error)
2097{
2098
2099 if (donebytes == 0) {
2100 return;
2101 }
2102 mutex_enter(mbp->b_objlock);
2103 KASSERT(mbp->b_resid >= donebytes);
2104 mbp->b_resid -= donebytes;
2105 if (error)
2106 mbp->b_error = error;
2107 if (mbp->b_resid == 0) {
2108 if (mbp->b_error)
2109 mbp->b_resid = mbp->b_bcount;
2110 mutex_exit(mbp->b_objlock);
2111 biodone(mbp);
2112 } else
2113 mutex_exit(mbp->b_objlock);
2114}
2115
2116void
2117buf_init(buf_t *bp)
2118{
2119
2120 cv_init(&bp->b_busy, "biolock");
2121 cv_init(&bp->b_done, "biowait");
2122 bp->b_dev = NODEV;
2123 bp->b_error = 0;
2124 bp->b_flags = 0;
2125 bp->b_cflags = 0;
2126 bp->b_oflags = 0;
2127 bp->b_objlock = &buffer_lock;
2128 bp->b_iodone = NULL;
2129 bp->b_refcnt = 1;
2130 bp->b_dev = NODEV;
2131 bp->b_vnbufs.le_next = NOLIST;
2132 BIO_SETPRIO(bp, BPRIO_DEFAULT);
2133}
2134
2135void
2136buf_destroy(buf_t *bp)
2137{
2138
2139 cv_destroy(&bp->b_done);
2140 cv_destroy(&bp->b_busy);
2141}
2142
2143int
2144bbusy(buf_t *bp, bool intr, int timo, kmutex_t *interlock)
2145{
2146 int error;
2147
2148 KASSERT(mutex_owned(&bufcache_lock));
2149
2150 if ((bp->b_cflags & BC_BUSY) != 0) {
2151 if (curlwp == uvm.pagedaemon_lwp)
2152 return EDEADLK;
2153 bp->b_cflags |= BC_WANTED;
2154 bref(bp);
2155 if (interlock != NULL)
2156 mutex_exit(interlock);
2157 if (intr) {
2158 error = cv_timedwait_sig(&bp->b_busy, &bufcache_lock,
2159 timo);
2160 } else {
2161 error = cv_timedwait(&bp->b_busy, &bufcache_lock,
2162 timo);
2163 }
2164 brele(bp);
2165 if (interlock != NULL)
2166 mutex_enter(interlock);
2167 if (error != 0)
2168 return error;
2169 return EPASSTHROUGH;
2170 }
2171 bp->b_cflags |= BC_BUSY;
2172
2173 return 0;
2174}
2175
2176/*
2177 * Nothing outside this file should really need to know about nbuf,
2178 * but a few things still want to read it, so give them a way to do that.
2179 */
2180int
2181buf_nbuf(void)
2182{
2183
2184 return nbuf;
2185}
2186