xref: /freebsd/sys/sys/buf.h (revision c7046f76)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)buf.h	8.9 (Berkeley) 3/30/95
37  * $FreeBSD$
38  */
39 
40 #ifndef _SYS_BUF_H_
41 #define	_SYS_BUF_H_
42 
43 #include <sys/bufobj.h>
44 #include <sys/queue.h>
45 #include <sys/lock.h>
46 #include <sys/lockmgr.h>
47 #include <vm/uma.h>
48 
49 struct bio;
50 struct buf;
51 struct bufobj;
52 struct mount;
53 struct vnode;
54 struct uio;
55 
56 /*
57  * To avoid including <ufs/ffs/softdep.h>
58  */
59 LIST_HEAD(workhead, worklist);
60 /*
61  * These are currently used only by the soft dependency code, hence
62  * are stored once in a global variable. If other subsystems wanted
63  * to use these hooks, a pointer to a set of bio_ops could be added
64  * to each buffer.
65  */
66 extern struct bio_ops {
67 	void	(*io_start)(struct buf *);
68 	void	(*io_complete)(struct buf *);
69 	void	(*io_deallocate)(struct buf *);
70 	int	(*io_countdeps)(struct buf *, int);
71 } bioops;
72 
73 struct vm_object;
74 struct vm_page;
75 
76 typedef uint32_t b_xflags_t;
77 
78 /*
79  * The buffer header describes an I/O operation in the kernel.
80  *
81  * NOTES:
82  *	b_bufsize, b_bcount.  b_bufsize is the allocation size of the
83  *	buffer, either DEV_BSIZE or PAGE_SIZE aligned.  b_bcount is the
84  *	originally requested buffer size and can serve as a bounds check
85  *	against EOF.  For most, but not all uses, b_bcount == b_bufsize.
86  *
87  *	b_dirtyoff, b_dirtyend.  Buffers support piecemeal, unaligned
88  *	ranges of dirty data that need to be written to backing store.
89  *	The range is typically clipped at b_bcount ( not b_bufsize ).
90  *
91  *	b_resid.  Number of bytes remaining in I/O.  After an I/O operation
92  *	completes, b_resid is usually 0 indicating 100% success.
93  *
94  *	All fields are protected by the buffer lock except those marked:
95  *		V - Protected by owning bufobj lock
96  *		Q - Protected by the buf queue lock
97  *		D - Protected by an dependency implementation specific lock
98  */
99 struct buf {
100 	struct bufobj	*b_bufobj;
101 	long		b_bcount;
102 	void		*b_caller1;
103 	caddr_t		b_data;
104 	int		b_error;
105 	uint16_t	b_iocmd;	/* BIO_* bio_cmd from bio.h */
106 	uint16_t	b_ioflags;	/* BIO_* bio_flags from bio.h */
107 	off_t		b_iooffset;
108 	long		b_resid;
109 	void	(*b_iodone)(struct buf *);
110 	void	(*b_ckhashcalc)(struct buf *);
111 	uint64_t	b_ckhash;	/* B_CKHASH requested check-hash */
112 	daddr_t b_blkno;		/* Underlying physical block number. */
113 	off_t	b_offset;		/* Offset into file. */
114 	TAILQ_ENTRY(buf) b_bobufs;	/* (V) Buffer's associated vnode. */
115 	uint32_t	b_vflags;	/* (V) BV_* flags */
116 	uint8_t		b_qindex;	/* (Q) buffer queue index */
117 	uint8_t		b_domain;	/* (Q) buf domain this resides in */
118 	uint16_t	b_subqueue;	/* (Q) per-cpu q if any */
119 	uint32_t	b_flags;	/* B_* flags. */
120 	b_xflags_t b_xflags;		/* extra flags */
121 	struct lock b_lock;		/* Buffer lock */
122 	long	b_bufsize;		/* Allocated buffer size. */
123 	int	b_runningbufspace;	/* when I/O is running, pipelining */
124 	int	b_kvasize;		/* size of kva for buffer */
125 	int	b_dirtyoff;		/* Offset in buffer of dirty region. */
126 	int	b_dirtyend;		/* Offset of end of dirty region. */
127 	caddr_t	b_kvabase;		/* base kva for buffer */
128 	daddr_t b_lblkno;		/* Logical block number. */
129 	struct	vnode *b_vp;		/* Device vnode. */
130 	struct	ucred *b_rcred;		/* Read credentials reference. */
131 	struct	ucred *b_wcred;		/* Write credentials reference. */
132 	union {
133 		TAILQ_ENTRY(buf) b_freelist; /* (Q) */
134 		struct {
135 			void	(*b_pgiodone)(void *, struct vm_page **,
136 				    int, int);
137 			int	b_pgbefore;
138 			int	b_pgafter;
139 		};
140 	};
141 	union	cluster_info {
142 		TAILQ_HEAD(cluster_list_head, buf) cluster_head;
143 		TAILQ_ENTRY(buf) cluster_entry;
144 	} b_cluster;
145 	int		b_npages;
146 	struct	workhead b_dep;		/* (D) List of filesystem dependencies. */
147 	void	*b_fsprivate1;
148 	void	*b_fsprivate2;
149 	void	*b_fsprivate3;
150 
151 #if defined(FULL_BUF_TRACKING)
152 #define BUF_TRACKING_SIZE	32
153 #define BUF_TRACKING_ENTRY(x)	((x) & (BUF_TRACKING_SIZE - 1))
154 	const char	*b_io_tracking[BUF_TRACKING_SIZE];
155 	uint32_t	b_io_tcnt;
156 #elif defined(BUF_TRACKING)
157 	const char	*b_io_tracking;
158 #endif
159 	struct	vm_page *b_pages[];
160 };
161 
162 #define b_object	b_bufobj->bo_object
163 
164 /*
165  * These flags are kept in b_flags.
166  *
167  * Notes:
168  *
169  *	B_ASYNC		VOP calls on bp's are usually async whether or not
170  *			B_ASYNC is set, but some subsystems, such as NFS, like
171  *			to know what is best for the caller so they can
172  *			optimize the I/O.
173  *
174  *	B_PAGING	Indicates that bp is being used by the paging system or
175  *			some paging system and that the bp is not linked into
176  *			the b_vp's clean/dirty linked lists or ref counts.
177  *			Buffer vp reassignments are illegal in this case.
178  *
179  *	B_CACHE		This may only be set if the buffer is entirely valid.
180  *			The situation where B_DELWRI is set and B_CACHE is
181  *			clear MUST be committed to disk by getblk() so
182  *			B_DELWRI can also be cleared.  See the comments for
183  *			getblk() in kern/vfs_bio.c.  If B_CACHE is clear,
184  *			the caller is expected to clear BIO_ERROR and B_INVAL,
185  *			set BIO_READ, and initiate an I/O.
186  *
187  *			The 'entire buffer' is defined to be the range from
188  *			0 through b_bcount.
189  *
190  *	B_MALLOC	Request that the buffer be allocated from the malloc
191  *			pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
192  *
193  *	B_CLUSTEROK	This flag is typically set for B_DELWRI buffers
194  *			by filesystems that allow clustering when the buffer
195  *			is fully dirty and indicates that it may be clustered
196  *			with other adjacent dirty buffers.  Note the clustering
197  *			may not be used with the stage 1 data write under NFS
198  *			but may be used for the commit rpc portion.
199  *
200  *	B_INVALONERR	This flag is set on dirty buffers.  It specifies that a
201  *			write error should forcibly invalidate the buffer
202  *			contents.  This flag should be used with caution, as it
203  *			discards data.  It is incompatible with B_ASYNC.
204  *
205  *	B_VMIO		Indicates that the buffer is tied into an VM object.
206  *			The buffer's data is always PAGE_SIZE aligned even
207  *			if b_bufsize and b_bcount are not.  ( b_bufsize is
208  *			always at least DEV_BSIZE aligned, though ).
209  *
210  *	B_DIRECT	Hint that we should attempt to completely free
211  *			the pages underlying the buffer.  B_DIRECT is
212  *			sticky until the buffer is released and typically
213  *			only has an effect when B_RELBUF is also set.
214  *
215  */
216 
217 #define	B_AGE		0x00000001	/* Move to age queue when I/O done. */
218 #define	B_NEEDCOMMIT	0x00000002	/* Append-write in progress. */
219 #define	B_ASYNC		0x00000004	/* Start I/O, do not wait. */
220 #define	B_DIRECT	0x00000008	/* direct I/O flag (pls free vmio) */
221 #define	B_DEFERRED	0x00000010	/* Skipped over for cleaning */
222 #define	B_CACHE		0x00000020	/* Bread found us in the cache. */
223 #define	B_VALIDSUSPWRT	0x00000040	/* Valid write during suspension. */
224 #define	B_DELWRI	0x00000080	/* Delay I/O until buffer reused. */
225 #define	B_CKHASH	0x00000100	/* checksum hash calculated on read */
226 #define	B_DONE		0x00000200	/* I/O completed. */
227 #define	B_EINTR		0x00000400	/* I/O was interrupted */
228 #define	B_NOREUSE	0x00000800	/* Contents not reused once released. */
229 #define	B_REUSE		0x00001000	/* Contents reused, second chance. */
230 #define	B_INVAL		0x00002000	/* Does not contain valid info. */
231 #define	B_BARRIER	0x00004000	/* Write this and all preceding first. */
232 #define	B_NOCACHE	0x00008000	/* Do not cache block after use. */
233 #define	B_MALLOC	0x00010000	/* malloced b_data */
234 #define	B_CLUSTEROK	0x00020000	/* Pagein op, so swap() can count it. */
235 #define	B_INVALONERR	0x00040000	/* Invalidate on write error. */
236 #define	B_IOSTARTED	0x00080000	/* buf_start() called */
237 #define	B_00100000	0x00100000	/* Available flag. */
238 #define	B_MAXPHYS	0x00200000	/* nitems(b_pages[]) = atop(MAXPHYS). */
239 #define	B_RELBUF	0x00400000	/* Release VMIO buffer. */
240 #define	B_FS_FLAG1	0x00800000	/* Available flag for FS use. */
241 #define	B_NOCOPY	0x01000000	/* Don't copy-on-write this buf. */
242 #define	B_INFREECNT	0x02000000	/* buf is counted in numfreebufs */
243 #define	B_PAGING	0x04000000	/* volatile paging I/O -- bypass VMIO */
244 #define B_MANAGED	0x08000000	/* Managed by FS. */
245 #define B_RAM		0x10000000	/* Read ahead mark (flag) */
246 #define B_VMIO		0x20000000	/* VMIO flag */
247 #define B_CLUSTER	0x40000000	/* pagein op, so swap() can count it */
248 #define B_REMFREE	0x80000000	/* Delayed bremfree */
249 
250 #define PRINT_BUF_FLAGS "\20\40remfree\37cluster\36vmio\35ram\34managed" \
251 	"\33paging\32infreecnt\31nocopy\30b23\27relbuf\26maxphys\25b20" \
252 	"\24iostarted\23invalonerr\22clusterok\21malloc\20nocache\17b14" \
253 	"\16inval\15reuse\14noreuse\13eintr\12done\11b8\10delwri" \
254 	"\7validsuspwrt\6cache\5deferred\4direct\3async\2needcommit\1age"
255 
256 /*
257  * These flags are kept in b_xflags.
258  *
259  * BX_FSPRIV reserves a set of eight flags that may be used by individual
260  * filesystems for their own purpose. Their specific definitions are
261  * found in the header files for each filesystem that uses them.
262  */
263 #define	BX_VNDIRTY	0x00000001	/* On vnode dirty list */
264 #define	BX_VNCLEAN	0x00000002	/* On vnode clean list */
265 #define	BX_CVTENXIO	0x00000004	/* Convert errors to ENXIO */
266 #define	BX_BKGRDWRITE	0x00000010	/* Do writes in background */
267 #define	BX_BKGRDMARKER	0x00000020	/* Mark buffer for splay tree */
268 #define	BX_ALTDATA	0x00000040	/* Holds extended data */
269 #define	BX_FSPRIV	0x00FF0000	/* Filesystem-specific flags mask */
270 
271 #define	PRINT_BUF_XFLAGS "\20\7altdata\6bkgrdmarker\5bkgrdwrite\3cvtenxio" \
272 	"\2clean\1dirty"
273 
274 #define	NOOFFSET	(-1LL)		/* No buffer offset calculated yet */
275 
276 /*
277  * These flags are kept in b_vflags.
278  */
279 #define	BV_SCANNED	0x00000001	/* VOP_FSYNC funcs mark written bufs */
280 #define	BV_BKGRDINPROG	0x00000002	/* Background write in progress */
281 #define	BV_BKGRDWAIT	0x00000004	/* Background write waiting */
282 #define	BV_BKGRDERR	0x00000008	/* Error from background write */
283 
284 #define	PRINT_BUF_VFLAGS "\20\4bkgrderr\3bkgrdwait\2bkgrdinprog\1scanned"
285 
286 #ifdef _KERNEL
287 
288 #ifndef NSWBUF_MIN
289 #define	NSWBUF_MIN	16
290 #endif
291 
292 /*
293  * Buffer locking
294  */
295 #include <sys/proc.h>			/* XXX for curthread */
296 #include <sys/mutex.h>
297 
298 /*
299  * Initialize a lock.
300  */
301 #define BUF_LOCKINIT(bp, wmesg)						\
302 	lockinit(&(bp)->b_lock, PRIBIO + 4, wmesg, 0, LK_NEW)
303 /*
304  *
305  * Get a lock sleeping non-interruptably until it becomes available.
306  */
307 #define	BUF_LOCK(bp, locktype, interlock)				\
308 	_lockmgr_args_rw(&(bp)->b_lock, (locktype), (interlock),	\
309 	    LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT,		\
310 	    LOCK_FILE, LOCK_LINE)
311 
312 /*
313  * Get a lock sleeping with specified interruptably and timeout.
314  */
315 #define	BUF_TIMELOCK(bp, locktype, interlock, wmesg, catch, timo)	\
316 	_lockmgr_args_rw(&(bp)->b_lock, (locktype) | LK_TIMELOCK,	\
317 	    (interlock), (wmesg), (PRIBIO + 4) | (catch), (timo),	\
318 	    LOCK_FILE, LOCK_LINE)
319 
320 /*
321  * Release a lock. Only the acquiring process may free the lock unless
322  * it has been handed off to biodone.
323  */
324 #define	BUF_UNLOCK(bp) do {						\
325 	KASSERT(((bp)->b_flags & B_REMFREE) == 0,			\
326 	    ("BUF_UNLOCK %p while B_REMFREE is still set.", (bp)));	\
327 									\
328 	BUF_UNLOCK_RAW((bp));						\
329 } while (0)
330 #define	BUF_UNLOCK_RAW(bp) do {						\
331 	(void)_lockmgr_args(&(bp)->b_lock, LK_RELEASE, NULL,		\
332 	    LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT,		\
333 	    LOCK_FILE, LOCK_LINE);					\
334 } while (0)
335 
336 /*
337  * Check if a buffer lock is recursed.
338  */
339 #define	BUF_LOCKRECURSED(bp)						\
340 	lockmgr_recursed(&(bp)->b_lock)
341 
342 /*
343  * Check if a buffer lock is currently held.
344  */
345 #define	BUF_ISLOCKED(bp)						\
346 	lockstatus(&(bp)->b_lock)
347 /*
348  * Free a buffer lock.
349  */
350 #define BUF_LOCKFREE(bp) 						\
351 	lockdestroy(&(bp)->b_lock)
352 
353 /*
354  * Print informations on a buffer lock.
355  */
356 #define BUF_LOCKPRINTINFO(bp) 						\
357 	lockmgr_printinfo(&(bp)->b_lock)
358 
359 /*
360  * Buffer lock assertions.
361  */
362 #if defined(INVARIANTS) && defined(INVARIANT_SUPPORT)
363 #define	BUF_ASSERT_LOCKED(bp)						\
364 	_lockmgr_assert(&(bp)->b_lock, KA_LOCKED, LOCK_FILE, LOCK_LINE)
365 #define	BUF_ASSERT_SLOCKED(bp)						\
366 	_lockmgr_assert(&(bp)->b_lock, KA_SLOCKED, LOCK_FILE, LOCK_LINE)
367 #define	BUF_ASSERT_XLOCKED(bp)						\
368 	_lockmgr_assert(&(bp)->b_lock, KA_XLOCKED, LOCK_FILE, LOCK_LINE)
369 #define	BUF_ASSERT_UNLOCKED(bp)						\
370 	_lockmgr_assert(&(bp)->b_lock, KA_UNLOCKED, LOCK_FILE, LOCK_LINE)
371 #else
372 #define	BUF_ASSERT_LOCKED(bp)
373 #define	BUF_ASSERT_SLOCKED(bp)
374 #define	BUF_ASSERT_XLOCKED(bp)
375 #define	BUF_ASSERT_UNLOCKED(bp)
376 #endif
377 
378 #ifdef _SYS_PROC_H_	/* Avoid #include <sys/proc.h> pollution */
379 /*
380  * When initiating asynchronous I/O, change ownership of the lock to the
381  * kernel. Once done, the lock may legally released by biodone. The
382  * original owning process can no longer acquire it recursively, but must
383  * wait until the I/O is completed and the lock has been freed by biodone.
384  */
385 #define	BUF_KERNPROC(bp)						\
386 	_lockmgr_disown(&(bp)->b_lock, LOCK_FILE, LOCK_LINE)
387 #endif
388 
389 #endif /* _KERNEL */
390 
391 struct buf_queue_head {
392 	TAILQ_HEAD(buf_queue, buf) queue;
393 	daddr_t last_pblkno;
394 	struct	buf *insert_point;
395 	struct	buf *switch_point;
396 };
397 
398 /*
399  * This structure describes a clustered I/O.
400  */
401 struct cluster_save {
402 	long	bs_bcount;		/* Saved b_bcount. */
403 	long	bs_bufsize;		/* Saved b_bufsize. */
404 	int	bs_nchildren;		/* Number of associated buffers. */
405 	struct buf **bs_children;	/* List of associated buffers. */
406 };
407 
408 /*
409  * Vnode clustering tracker
410  */
411 struct vn_clusterw {
412 	daddr_t	v_cstart;			/* v start block of cluster */
413 	daddr_t	v_lasta;			/* v last allocation  */
414 	daddr_t	v_lastw;			/* v last write  */
415 	int	v_clen;				/* v length of cur. cluster */
416 };
417 
418 #ifdef _KERNEL
419 
420 static __inline int
421 bwrite(struct buf *bp)
422 {
423 
424 	KASSERT(bp->b_bufobj != NULL, ("bwrite: no bufobj bp=%p", bp));
425 	KASSERT(bp->b_bufobj->bo_ops != NULL, ("bwrite: no bo_ops bp=%p", bp));
426 	KASSERT(bp->b_bufobj->bo_ops->bop_write != NULL,
427 	    ("bwrite: no bop_write bp=%p", bp));
428 	return (BO_WRITE(bp->b_bufobj, bp));
429 }
430 
431 static __inline void
432 bstrategy(struct buf *bp)
433 {
434 
435 	KASSERT(bp->b_bufobj != NULL, ("bstrategy: no bufobj bp=%p", bp));
436 	KASSERT(bp->b_bufobj->bo_ops != NULL,
437 	    ("bstrategy: no bo_ops bp=%p", bp));
438 	KASSERT(bp->b_bufobj->bo_ops->bop_strategy != NULL,
439 	    ("bstrategy: no bop_strategy bp=%p", bp));
440 	BO_STRATEGY(bp->b_bufobj, bp);
441 }
442 
443 static __inline void
444 buf_start(struct buf *bp)
445 {
446 	KASSERT((bp->b_flags & B_IOSTARTED) == 0,
447 	    ("recursed buf_start %p", bp));
448 	bp->b_flags |= B_IOSTARTED;
449 	if (bioops.io_start)
450 		(*bioops.io_start)(bp);
451 }
452 
453 static __inline void
454 buf_complete(struct buf *bp)
455 {
456 	if ((bp->b_flags & B_IOSTARTED) != 0) {
457 		bp->b_flags &= ~B_IOSTARTED;
458 		if (bioops.io_complete)
459 			(*bioops.io_complete)(bp);
460 	}
461 }
462 
463 static __inline void
464 buf_deallocate(struct buf *bp)
465 {
466 	if (bioops.io_deallocate)
467 		(*bioops.io_deallocate)(bp);
468 }
469 
470 static __inline int
471 buf_countdeps(struct buf *bp, int i)
472 {
473 	if (bioops.io_countdeps)
474 		return ((*bioops.io_countdeps)(bp, i));
475 	else
476 		return (0);
477 }
478 
479 static __inline void
480 buf_track(struct buf *bp __unused, const char *location __unused)
481 {
482 
483 #if defined(FULL_BUF_TRACKING)
484 	bp->b_io_tracking[BUF_TRACKING_ENTRY(bp->b_io_tcnt++)] = location;
485 #elif defined(BUF_TRACKING)
486 	bp->b_io_tracking = location;
487 #endif
488 }
489 
490 #endif /* _KERNEL */
491 
492 /*
493  * Zero out the buffer's data area.
494  */
495 #define	clrbuf(bp) {							\
496 	bzero((bp)->b_data, (u_int)(bp)->b_bcount);			\
497 	(bp)->b_resid = 0;						\
498 }
499 
500 /*
501  * Flags for getblk's last parameter.
502  */
503 #define	GB_LOCK_NOWAIT	0x0001		/* Fail if we block on a buf lock. */
504 #define	GB_NOCREAT	0x0002		/* Don't create a buf if not found. */
505 #define	GB_NOWAIT_BD	0x0004		/* Do not wait for bufdaemon. */
506 #define	GB_UNMAPPED	0x0008		/* Do not mmap buffer pages. */
507 #define	GB_KVAALLOC	0x0010		/* But allocate KVA. */
508 #define	GB_CKHASH	0x0020		/* If reading, calc checksum hash */
509 #define	GB_NOSPARSE	0x0040		/* Do not instantiate holes */
510 #define	GB_CVTENXIO	0x0080		/* Convert errors to ENXIO */
511 #define	GB_NOWITNESS	0x0100		/* Do not record for WITNESS */
512 
513 #ifdef _KERNEL
514 extern int	nbuf;			/* The number of buffer headers */
515 extern u_long	maxswzone;		/* Max KVA for swap structures */
516 extern u_long	maxbcache;		/* Max KVA for buffer cache */
517 extern int	maxbcachebuf;		/* Max buffer cache block size */
518 extern long	runningbufspace;
519 extern long	hibufspace;
520 extern int	dirtybufthresh;
521 extern int	bdwriteskip;
522 extern int	dirtybufferflushes;
523 extern int	altbufferflushes;
524 extern int	nswbuf;			/* Number of swap I/O buffer headers. */
525 extern caddr_t __read_mostly unmapped_buf; /* Data address for unmapped
526 					      buffers. */
527 
528 static inline int
529 buf_mapped(struct buf *bp)
530 {
531 
532 	return (bp->b_data != unmapped_buf);
533 }
534 
535 void	runningbufwakeup(struct buf *);
536 void	waitrunningbufspace(void);
537 caddr_t	kern_vfs_bio_buffer_alloc(caddr_t v, long physmem_est);
538 void	bufinit(void);
539 void	bufshutdown(int);
540 void	bdata2bio(struct buf *bp, struct bio *bip);
541 void	bwillwrite(void);
542 int	buf_dirty_count_severe(void);
543 void	bremfree(struct buf *);
544 void	bremfreef(struct buf *);	/* XXX Force bremfree, only for nfs. */
545 #define bread(vp, blkno, size, cred, bpp) \
546 	    breadn_flags(vp, blkno, blkno, size, NULL, NULL, 0, cred, 0, \
547 		NULL, bpp)
548 #define bread_gb(vp, blkno, size, cred, gbflags, bpp) \
549 	    breadn_flags(vp, blkno, blkno, size, NULL, NULL, 0, cred, \
550 		gbflags, NULL, bpp)
551 #define breadn(vp, blkno, size, rablkno, rabsize, cnt, cred, bpp) \
552 	    breadn_flags(vp, blkno, blkno, size, rablkno, rabsize, cnt, cred, \
553 		0, NULL, bpp)
554 int	breadn_flags(struct vnode *, daddr_t, daddr_t, int, daddr_t *, int *,
555 	    int, struct ucred *, int, void (*)(struct buf *), struct buf **);
556 void	bdwrite(struct buf *);
557 void	bawrite(struct buf *);
558 void	babarrierwrite(struct buf *);
559 int	bbarrierwrite(struct buf *);
560 void	bdirty(struct buf *);
561 void	bundirty(struct buf *);
562 void	bufstrategy(struct bufobj *, struct buf *);
563 void	brelse(struct buf *);
564 void	bqrelse(struct buf *);
565 int	vfs_bio_awrite(struct buf *);
566 void	vfs_busy_pages_acquire(struct buf *bp);
567 void	vfs_busy_pages_release(struct buf *bp);
568 struct buf *incore(struct bufobj *, daddr_t);
569 bool	inmem(struct vnode *, daddr_t);
570 struct buf *gbincore(struct bufobj *, daddr_t);
571 struct buf *gbincore_unlocked(struct bufobj *, daddr_t);
572 struct buf *getblk(struct vnode *, daddr_t, int, int, int, int);
573 int	getblkx(struct vnode *vp, daddr_t blkno, daddr_t dblkno, int size,
574 	    int slpflag, int slptimeo, int flags, struct buf **bpp);
575 struct buf *geteblk(int, int);
576 int	bufwait(struct buf *);
577 int	bufwrite(struct buf *);
578 void	bufdone(struct buf *);
579 void	bd_speedup(void);
580 
581 extern uma_zone_t pbuf_zone;
582 uma_zone_t pbuf_zsecond_create(const char *name, int max);
583 
584 struct vn_clusterw;
585 
586 void	cluster_init_vn(struct vn_clusterw *vnc);
587 int	cluster_read(struct vnode *, u_quad_t, daddr_t, long,
588 	    struct ucred *, long, int, int, struct buf **);
589 int	cluster_wbuild(struct vnode *, long, daddr_t, int, int);
590 void	cluster_write(struct vnode *, struct vn_clusterw *, struct buf *,
591 	    u_quad_t, int, int);
592 void	vfs_bio_brelse(struct buf *bp, int ioflags);
593 void	vfs_bio_bzero_buf(struct buf *bp, int base, int size);
594 void	vfs_bio_clrbuf(struct buf *);
595 void	vfs_bio_set_flags(struct buf *bp, int ioflags);
596 void	vfs_bio_set_valid(struct buf *, int base, int size);
597 void	vfs_busy_pages(struct buf *, int clear_modify);
598 void	vfs_unbusy_pages(struct buf *);
599 int	vmapbuf(struct buf *, void *, size_t, int);
600 void	vunmapbuf(struct buf *);
601 void	brelvp(struct buf *);
602 void	bgetvp(struct vnode *, struct buf *);
603 void	pbgetbo(struct bufobj *bo, struct buf *bp);
604 void	pbgetvp(struct vnode *, struct buf *);
605 void	pbrelbo(struct buf *);
606 void	pbrelvp(struct buf *);
607 int	allocbuf(struct buf *bp, int size);
608 void	reassignbuf(struct buf *);
609 void	bwait(struct buf *, u_char, const char *);
610 void	bdone(struct buf *);
611 
612 typedef daddr_t (vbg_get_lblkno_t)(struct vnode *, vm_ooffset_t);
613 typedef int (vbg_get_blksize_t)(struct vnode *, daddr_t, long *);
614 int	vfs_bio_getpages(struct vnode *vp, struct vm_page **ma, int count,
615 	    int *rbehind, int *rahead, vbg_get_lblkno_t get_lblkno,
616 	    vbg_get_blksize_t get_blksize);
617 
618 #endif /* _KERNEL */
619 
620 #endif /* !_SYS_BUF_H_ */
621