xref: /freebsd-13-stable/sys/sys/buf.h (revision f4ebac8d805b94c99b08d96a262218cbe48cd3a6)
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  */
38 
39 #ifndef _SYS_BUF_H_
40 #define	_SYS_BUF_H_
41 
42 #include <sys/bufobj.h>
43 #include <sys/queue.h>
44 #include <sys/lock.h>
45 #include <sys/lockmgr.h>
46 #include <vm/uma.h>
47 
48 struct bio;
49 struct buf;
50 struct bufobj;
51 struct mount;
52 struct vnode;
53 struct uio;
54 
55 /*
56  * To avoid including <ufs/ffs/softdep.h>
57  */
58 LIST_HEAD(workhead, worklist);
59 /*
60  * These are currently used only by the soft dependency code, hence
61  * are stored once in a global variable. If other subsystems wanted
62  * to use these hooks, a pointer to a set of bio_ops could be added
63  * to each buffer.
64  */
65 extern struct bio_ops {
66 	void	(*io_start)(struct buf *);
67 	void	(*io_complete)(struct buf *);
68 	void	(*io_deallocate)(struct buf *);
69 	int	(*io_countdeps)(struct buf *, int);
70 } bioops;
71 
72 struct vm_object;
73 struct vm_page;
74 
75 typedef uint32_t b_xflags_t;
76 
77 /*
78  * The buffer header describes an I/O operation in the kernel.
79  *
80  * NOTES:
81  *	b_bufsize, b_bcount.  b_bufsize is the allocation size of the
82  *	buffer, either DEV_BSIZE or PAGE_SIZE aligned.  b_bcount is the
83  *	originally requested buffer size and can serve as a bounds check
84  *	against EOF.  For most, but not all uses, b_bcount == b_bufsize.
85  *
86  *	b_dirtyoff, b_dirtyend.  Buffers support piecemeal, unaligned
87  *	ranges of dirty data that need to be written to backing store.
88  *	The range is typically clipped at b_bcount ( not b_bufsize ).
89  *
90  *	b_resid.  Number of bytes remaining in I/O.  After an I/O operation
91  *	completes, b_resid is usually 0 indicating 100% success.
92  *
93  *	All fields are protected by the buffer lock except those marked:
94  *		V - Protected by owning bufobj lock
95  *		Q - Protected by the buf queue lock
96  *		D - Protected by an dependency implementation specific lock
97  */
98 struct buf {
99 	struct bufobj	*b_bufobj;
100 	long		b_bcount;
101 	void		*b_caller1;
102 	caddr_t		b_data;
103 	int		b_error;
104 	uint16_t	b_iocmd;	/* BIO_* bio_cmd from bio.h */
105 	uint16_t	b_ioflags;	/* BIO_* bio_flags from bio.h */
106 	off_t		b_iooffset;
107 	long		b_resid;
108 	void	(*b_iodone)(struct buf *);
109 	void	(*b_ckhashcalc)(struct buf *);
110 	uint64_t	b_ckhash;	/* B_CKHASH requested check-hash */
111 	daddr_t b_blkno;		/* Underlying physical block number. */
112 	off_t	b_offset;		/* Offset into file. */
113 	TAILQ_ENTRY(buf) b_bobufs;	/* (V) Buffer's associated vnode. */
114 	uint32_t	b_vflags;	/* (V) BV_* flags */
115 	uint8_t		b_qindex;	/* (Q) buffer queue index */
116 	uint8_t		b_domain;	/* (Q) buf domain this resides in */
117 	uint16_t	b_subqueue;	/* (Q) per-cpu q if any */
118 	uint32_t	b_flags;	/* B_* flags. */
119 	b_xflags_t b_xflags;		/* extra flags */
120 	struct lock b_lock;		/* Buffer lock */
121 	long	b_bufsize;		/* Allocated buffer size. */
122 	int	b_runningbufspace;	/* when I/O is running, pipelining */
123 	int	b_kvasize;		/* size of kva for buffer */
124 	int	b_dirtyoff;		/* Offset in buffer of dirty region. */
125 	int	b_dirtyend;		/* Offset of end of dirty region. */
126 	caddr_t	b_kvabase;		/* base kva for buffer */
127 	daddr_t b_lblkno;		/* Logical block number. */
128 	struct	vnode *b_vp;		/* Device vnode. */
129 	struct	ucred *b_rcred;		/* Read credentials reference. */
130 	struct	ucred *b_wcred;		/* Write credentials reference. */
131 	union {
132 		TAILQ_ENTRY(buf) b_freelist; /* (Q) */
133 		struct {
134 			void	(*b_pgiodone)(void *, vm_page_t *, int, int);
135 			int	b_pgbefore;
136 			int	b_pgafter;
137 		};
138 	};
139 	union	cluster_info {
140 		TAILQ_HEAD(cluster_list_head, buf) cluster_head;
141 		TAILQ_ENTRY(buf) cluster_entry;
142 	} b_cluster;
143 	int		b_npages;
144 	struct	workhead b_dep;		/* (D) List of filesystem dependencies. */
145 	void	*b_fsprivate1;
146 	void	*b_fsprivate2;
147 	void	*b_fsprivate3;
148 
149 #if defined(FULL_BUF_TRACKING)
150 #define BUF_TRACKING_SIZE	32
151 #define BUF_TRACKING_ENTRY(x)	((x) & (BUF_TRACKING_SIZE - 1))
152 	const char	*b_io_tracking[BUF_TRACKING_SIZE];
153 	uint32_t	b_io_tcnt;
154 #elif defined(BUF_TRACKING)
155 	const char	*b_io_tracking;
156 #endif
157 	struct	vm_page *b_pages[];
158 };
159 
160 #define b_object	b_bufobj->bo_object
161 
162 /*
163  * These flags are kept in b_flags.
164  *
165  * Notes:
166  *
167  *	B_ASYNC		VOP calls on bp's are usually async whether or not
168  *			B_ASYNC is set, but some subsystems, such as NFS, like
169  *			to know what is best for the caller so they can
170  *			optimize the I/O.
171  *
172  *	B_PAGING	Indicates that bp is being used by the paging system or
173  *			some paging system and that the bp is not linked into
174  *			the b_vp's clean/dirty linked lists or ref counts.
175  *			Buffer vp reassignments are illegal in this case.
176  *
177  *	B_CACHE		This may only be set if the buffer is entirely valid.
178  *			The situation where B_DELWRI is set and B_CACHE is
179  *			clear MUST be committed to disk by getblk() so
180  *			B_DELWRI can also be cleared.  See the comments for
181  *			getblk() in kern/vfs_bio.c.  If B_CACHE is clear,
182  *			the caller is expected to clear BIO_ERROR and B_INVAL,
183  *			set BIO_READ, and initiate an I/O.
184  *
185  *			The 'entire buffer' is defined to be the range from
186  *			0 through b_bcount.
187  *
188  *	B_MALLOC	Request that the buffer be allocated from the malloc
189  *			pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
190  *
191  *	B_CLUSTEROK	This flag is typically set for B_DELWRI buffers
192  *			by filesystems that allow clustering when the buffer
193  *			is fully dirty and indicates that it may be clustered
194  *			with other adjacent dirty buffers.  Note the clustering
195  *			may not be used with the stage 1 data write under NFS
196  *			but may be used for the commit rpc portion.
197  *
198  *	B_INVALONERR	This flag is set on dirty buffers.  It specifies that a
199  *			write error should forcibly invalidate the buffer
200  *			contents.  This flag should be used with caution, as it
201  *			discards data.  It is incompatible with B_ASYNC.
202  *
203  *	B_VMIO		Indicates that the buffer is tied into an VM object.
204  *			The buffer's data is always PAGE_SIZE aligned even
205  *			if b_bufsize and b_bcount are not.  ( b_bufsize is
206  *			always at least DEV_BSIZE aligned, though ).
207  *
208  *	B_DIRECT	Hint that we should attempt to completely free
209  *			the pages underlying the buffer.  B_DIRECT is
210  *			sticky until the buffer is released and typically
211  *			only has an effect when B_RELBUF is also set.
212  *
213  */
214 
215 #define	B_AGE		0x00000001	/* Move to age queue when I/O done. */
216 #define	B_NEEDCOMMIT	0x00000002	/* Append-write in progress. */
217 #define	B_ASYNC		0x00000004	/* Start I/O, do not wait. */
218 #define	B_DIRECT	0x00000008	/* direct I/O flag (pls free vmio) */
219 #define	B_DEFERRED	0x00000010	/* Skipped over for cleaning */
220 #define	B_CACHE		0x00000020	/* Bread found us in the cache. */
221 #define	B_VALIDSUSPWRT	0x00000040	/* Valid write during suspension. */
222 #define	B_DELWRI	0x00000080	/* Delay I/O until buffer reused. */
223 #define	B_CKHASH	0x00000100	/* checksum hash calculated on read */
224 #define	B_DONE		0x00000200	/* I/O completed. */
225 #define	B_EINTR		0x00000400	/* I/O was interrupted */
226 #define	B_NOREUSE	0x00000800	/* Contents not reused once released. */
227 #define	B_REUSE		0x00001000	/* Contents reused, second chance. */
228 #define	B_INVAL		0x00002000	/* Does not contain valid info. */
229 #define	B_BARRIER	0x00004000	/* Write this and all preceding first. */
230 #define	B_NOCACHE	0x00008000	/* Do not cache block after use. */
231 #define	B_MALLOC	0x00010000	/* malloced b_data */
232 #define	B_CLUSTEROK	0x00020000	/* Pagein op, so swap() can count it. */
233 #define	B_INVALONERR	0x00040000	/* Invalidate on write error. */
234 #define	B_IOSTARTED	0x00080000	/* buf_start() called */
235 #define	B_00100000	0x00100000	/* Available flag. */
236 #define	B_MAXPHYS	0x00200000	/* nitems(b_pages[]) = atop(MAXPHYS). */
237 #define	B_RELBUF	0x00400000	/* Release VMIO buffer. */
238 #define	B_FS_FLAG1	0x00800000	/* Available flag for FS use. */
239 #define	B_NOCOPY	0x01000000	/* Don't copy-on-write this buf. */
240 #define	B_INFREECNT	0x02000000	/* buf is counted in numfreebufs */
241 #define	B_PAGING	0x04000000	/* volatile paging I/O -- bypass VMIO */
242 #define B_MANAGED	0x08000000	/* Managed by FS. */
243 #define B_RAM		0x10000000	/* Read ahead mark (flag) */
244 #define B_VMIO		0x20000000	/* VMIO flag */
245 #define B_CLUSTER	0x40000000	/* pagein op, so swap() can count it */
246 #define B_REMFREE	0x80000000	/* Delayed bremfree */
247 
248 #define PRINT_BUF_FLAGS "\20\40remfree\37cluster\36vmio\35ram\34managed" \
249 	"\33paging\32infreecnt\31nocopy\30b23\27relbuf\26maxphys\25b20" \
250 	"\24iostarted\23invalonerr\22clusterok\21malloc\20nocache\17b14" \
251 	"\16inval\15reuse\14noreuse\13eintr\12done\11b8\10delwri" \
252 	"\7validsuspwrt\6cache\5deferred\4direct\3async\2needcommit\1age"
253 
254 /*
255  * These flags are kept in b_xflags.
256  *
257  * BX_FSPRIV reserves a set of eight flags that may be used by individual
258  * filesystems for their own purpose. Their specific definitions are
259  * found in the header files for each filesystem that uses them.
260  */
261 #define	BX_VNDIRTY	0x00000001	/* On vnode dirty list */
262 #define	BX_VNCLEAN	0x00000002	/* On vnode clean list */
263 #define	BX_CVTENXIO	0x00000004	/* Convert errors to ENXIO */
264 #define	BX_BKGRDWRITE	0x00000010	/* Do writes in background */
265 #define	BX_BKGRDMARKER	0x00000020	/* Mark buffer for splay tree */
266 #define	BX_ALTDATA	0x00000040	/* Holds extended data */
267 #define	BX_FSPRIV	0x00FF0000	/* Filesystem-specific flags mask */
268 
269 #define	PRINT_BUF_XFLAGS "\20\7altdata\6bkgrdmarker\5bkgrdwrite\3cvtenxio" \
270 	"\2clean\1dirty"
271 
272 #define	NOOFFSET	(-1LL)		/* No buffer offset calculated yet */
273 
274 /*
275  * These flags are kept in b_vflags.
276  */
277 #define	BV_SCANNED	0x00000001	/* VOP_FSYNC funcs mark written bufs */
278 #define	BV_BKGRDINPROG	0x00000002	/* Background write in progress */
279 #define	BV_BKGRDWAIT	0x00000004	/* Background write waiting */
280 #define	BV_BKGRDERR	0x00000008	/* Error from background write */
281 
282 #define	PRINT_BUF_VFLAGS "\20\4bkgrderr\3bkgrdwait\2bkgrdinprog\1scanned"
283 
284 #ifdef _KERNEL
285 
286 #ifndef NSWBUF_MIN
287 #define	NSWBUF_MIN	16
288 #endif
289 
290 /*
291  * Buffer locking
292  */
293 #include <sys/proc.h>			/* XXX for curthread */
294 #include <sys/mutex.h>
295 
296 /*
297  * Initialize a lock.
298  */
299 #define BUF_LOCKINIT(bp, wmesg)						\
300 	lockinit(&(bp)->b_lock, PRIBIO + 4, wmesg, 0, LK_NEW)
301 /*
302  *
303  * Get a lock sleeping non-interruptably until it becomes available.
304  */
305 #define	BUF_LOCK(bp, locktype, interlock)				\
306 	_lockmgr_args_rw(&(bp)->b_lock, (locktype), (interlock),	\
307 	    LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT,		\
308 	    LOCK_FILE, LOCK_LINE)
309 
310 /*
311  * Get a lock sleeping with specified interruptably and timeout.
312  */
313 #define	BUF_TIMELOCK(bp, locktype, interlock, wmesg, catch, timo)	\
314 	_lockmgr_args_rw(&(bp)->b_lock, (locktype) | LK_TIMELOCK,	\
315 	    (interlock), (wmesg), (PRIBIO + 4) | (catch), (timo),	\
316 	    LOCK_FILE, LOCK_LINE)
317 
318 /*
319  * Release a lock. Only the acquiring process may free the lock unless
320  * it has been handed off to biodone.
321  */
322 #define	BUF_UNLOCK(bp) do {						\
323 	KASSERT(((bp)->b_flags & B_REMFREE) == 0,			\
324 	    ("BUF_UNLOCK %p while B_REMFREE is still set.", (bp)));	\
325 									\
326 	BUF_UNLOCK_RAW((bp));						\
327 } while (0)
328 #define	BUF_UNLOCK_RAW(bp) do {						\
329 	(void)_lockmgr_args(&(bp)->b_lock, LK_RELEASE, NULL,		\
330 	    LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT,		\
331 	    LOCK_FILE, LOCK_LINE);					\
332 } while (0)
333 
334 /*
335  * Check if a buffer lock is recursed.
336  */
337 #define	BUF_LOCKRECURSED(bp)						\
338 	lockmgr_recursed(&(bp)->b_lock)
339 
340 /*
341  * Check if a buffer lock is currently held.
342  */
343 #define	BUF_ISLOCKED(bp)						\
344 	lockstatus(&(bp)->b_lock)
345 /*
346  * Free a buffer lock.
347  */
348 #define BUF_LOCKFREE(bp) 						\
349 	lockdestroy(&(bp)->b_lock)
350 
351 /*
352  * Print informations on a buffer lock.
353  */
354 #define BUF_LOCKPRINTINFO(bp) 						\
355 	lockmgr_printinfo(&(bp)->b_lock)
356 
357 /*
358  * Buffer lock assertions.
359  */
360 #if defined(INVARIANTS) && defined(INVARIANT_SUPPORT)
361 #define	BUF_ASSERT_LOCKED(bp)						\
362 	_lockmgr_assert(&(bp)->b_lock, KA_LOCKED, LOCK_FILE, LOCK_LINE)
363 #define	BUF_ASSERT_SLOCKED(bp)						\
364 	_lockmgr_assert(&(bp)->b_lock, KA_SLOCKED, LOCK_FILE, LOCK_LINE)
365 #define	BUF_ASSERT_XLOCKED(bp)						\
366 	_lockmgr_assert(&(bp)->b_lock, KA_XLOCKED, LOCK_FILE, LOCK_LINE)
367 #define	BUF_ASSERT_UNLOCKED(bp)						\
368 	_lockmgr_assert(&(bp)->b_lock, KA_UNLOCKED, LOCK_FILE, LOCK_LINE)
369 #else
370 #define	BUF_ASSERT_LOCKED(bp)
371 #define	BUF_ASSERT_SLOCKED(bp)
372 #define	BUF_ASSERT_XLOCKED(bp)
373 #define	BUF_ASSERT_UNLOCKED(bp)
374 #endif
375 
376 #ifdef _SYS_PROC_H_	/* Avoid #include <sys/proc.h> pollution */
377 /*
378  * When initiating asynchronous I/O, change ownership of the lock to the
379  * kernel. Once done, the lock may legally released by biodone. The
380  * original owning process can no longer acquire it recursively, but must
381  * wait until the I/O is completed and the lock has been freed by biodone.
382  */
383 #define	BUF_KERNPROC(bp)						\
384 	_lockmgr_disown(&(bp)->b_lock, LOCK_FILE, LOCK_LINE)
385 #endif
386 
387 #endif /* _KERNEL */
388 
389 struct buf_queue_head {
390 	TAILQ_HEAD(buf_queue, buf) queue;
391 	daddr_t last_pblkno;
392 	struct	buf *insert_point;
393 	struct	buf *switch_point;
394 };
395 
396 /*
397  * This structure describes a clustered I/O.
398  */
399 struct cluster_save {
400 	long	bs_bcount;		/* Saved b_bcount. */
401 	long	bs_bufsize;		/* Saved b_bufsize. */
402 	int	bs_nchildren;		/* Number of associated buffers. */
403 	struct buf **bs_children;	/* List of associated buffers. */
404 };
405 
406 #ifdef _KERNEL
407 
408 static __inline int
bwrite(struct buf * bp)409 bwrite(struct buf *bp)
410 {
411 
412 	KASSERT(bp->b_bufobj != NULL, ("bwrite: no bufobj bp=%p", bp));
413 	KASSERT(bp->b_bufobj->bo_ops != NULL, ("bwrite: no bo_ops bp=%p", bp));
414 	KASSERT(bp->b_bufobj->bo_ops->bop_write != NULL,
415 	    ("bwrite: no bop_write bp=%p", bp));
416 	return (BO_WRITE(bp->b_bufobj, bp));
417 }
418 
419 static __inline void
bstrategy(struct buf * bp)420 bstrategy(struct buf *bp)
421 {
422 
423 	KASSERT(bp->b_bufobj != NULL, ("bstrategy: no bufobj bp=%p", bp));
424 	KASSERT(bp->b_bufobj->bo_ops != NULL,
425 	    ("bstrategy: no bo_ops bp=%p", bp));
426 	KASSERT(bp->b_bufobj->bo_ops->bop_strategy != NULL,
427 	    ("bstrategy: no bop_strategy bp=%p", bp));
428 	BO_STRATEGY(bp->b_bufobj, bp);
429 }
430 
431 static __inline void
buf_start(struct buf * bp)432 buf_start(struct buf *bp)
433 {
434 	KASSERT((bp->b_flags & B_IOSTARTED) == 0,
435 	    ("recursed buf_start %p", bp));
436 	bp->b_flags |= B_IOSTARTED;
437 	if (bioops.io_start)
438 		(*bioops.io_start)(bp);
439 }
440 
441 static __inline void
buf_complete(struct buf * bp)442 buf_complete(struct buf *bp)
443 {
444 	if ((bp->b_flags & B_IOSTARTED) != 0) {
445 		bp->b_flags &= ~B_IOSTARTED;
446 		if (bioops.io_complete)
447 			(*bioops.io_complete)(bp);
448 	}
449 }
450 
451 static __inline void
buf_deallocate(struct buf * bp)452 buf_deallocate(struct buf *bp)
453 {
454 	if (bioops.io_deallocate)
455 		(*bioops.io_deallocate)(bp);
456 }
457 
458 static __inline int
buf_countdeps(struct buf * bp,int i)459 buf_countdeps(struct buf *bp, int i)
460 {
461 	if (bioops.io_countdeps)
462 		return ((*bioops.io_countdeps)(bp, i));
463 	else
464 		return (0);
465 }
466 
467 static __inline void
buf_track(struct buf * bp __unused,const char * location __unused)468 buf_track(struct buf *bp __unused, const char *location __unused)
469 {
470 
471 #if defined(FULL_BUF_TRACKING)
472 	bp->b_io_tracking[BUF_TRACKING_ENTRY(bp->b_io_tcnt++)] = location;
473 #elif defined(BUF_TRACKING)
474 	bp->b_io_tracking = location;
475 #endif
476 }
477 
478 #endif /* _KERNEL */
479 
480 /*
481  * Zero out the buffer's data area.
482  */
483 #define	clrbuf(bp) {							\
484 	bzero((bp)->b_data, (u_int)(bp)->b_bcount);			\
485 	(bp)->b_resid = 0;						\
486 }
487 
488 /*
489  * Flags for getblk's last parameter.
490  */
491 #define	GB_LOCK_NOWAIT	0x0001		/* Fail if we block on a buf lock. */
492 #define	GB_NOCREAT	0x0002		/* Don't create a buf if not found. */
493 #define	GB_NOWAIT_BD	0x0004		/* Do not wait for bufdaemon. */
494 #define	GB_UNMAPPED	0x0008		/* Do not mmap buffer pages. */
495 #define	GB_KVAALLOC	0x0010		/* But allocate KVA. */
496 #define	GB_CKHASH	0x0020		/* If reading, calc checksum hash */
497 #define	GB_NOSPARSE	0x0040		/* Do not instantiate holes */
498 #define	GB_CVTENXIO	0x0080		/* Convert errors to ENXIO */
499 #define	GB_NOWITNESS	0x0100		/* Do not record for WITNESS */
500 
501 #ifdef _KERNEL
502 extern int	nbuf;			/* The number of buffer headers */
503 extern u_long	maxswzone;		/* Max KVA for swap structures */
504 extern u_long	maxbcache;		/* Max KVA for buffer cache */
505 extern int	maxbcachebuf;		/* Max buffer cache block size */
506 extern long	runningbufspace;
507 extern long	hibufspace;
508 extern int	dirtybufthresh;
509 extern int	bdwriteskip;
510 extern int	dirtybufferflushes;
511 extern int	altbufferflushes;
512 extern int	nswbuf;			/* Number of swap I/O buffer headers. */
513 extern caddr_t __read_mostly unmapped_buf; /* Data address for unmapped
514 					      buffers. */
515 
516 static inline int
buf_mapped(struct buf * bp)517 buf_mapped(struct buf *bp)
518 {
519 
520 	return (bp->b_data != unmapped_buf);
521 }
522 
523 void	runningbufwakeup(struct buf *);
524 void	waitrunningbufspace(void);
525 caddr_t	kern_vfs_bio_buffer_alloc(caddr_t v, long physmem_est);
526 void	bufinit(void);
527 void	bufshutdown(int);
528 void	bdata2bio(struct buf *bp, struct bio *bip);
529 void	bwillwrite(void);
530 int	buf_dirty_count_severe(void);
531 void	bremfree(struct buf *);
532 void	bremfreef(struct buf *);	/* XXX Force bremfree, only for nfs. */
533 #define bread(vp, blkno, size, cred, bpp) \
534 	    breadn_flags(vp, blkno, blkno, size, NULL, NULL, 0, cred, 0, \
535 		NULL, bpp)
536 #define bread_gb(vp, blkno, size, cred, gbflags, bpp) \
537 	    breadn_flags(vp, blkno, blkno, size, NULL, NULL, 0, cred, \
538 		gbflags, NULL, bpp)
539 #define breadn(vp, blkno, size, rablkno, rabsize, cnt, cred, bpp) \
540 	    breadn_flags(vp, blkno, blkno, size, rablkno, rabsize, cnt, cred, \
541 		0, NULL, bpp)
542 int	breadn_flags(struct vnode *, daddr_t, daddr_t, int, daddr_t *, int *,
543 	    int, struct ucred *, int, void (*)(struct buf *), struct buf **);
544 void	bdwrite(struct buf *);
545 void	bawrite(struct buf *);
546 void	babarrierwrite(struct buf *);
547 int	bbarrierwrite(struct buf *);
548 void	bdirty(struct buf *);
549 void	bundirty(struct buf *);
550 void	bufstrategy(struct bufobj *, struct buf *);
551 void	brelse(struct buf *);
552 void	bqrelse(struct buf *);
553 int	vfs_bio_awrite(struct buf *);
554 void	vfs_busy_pages_acquire(struct buf *bp);
555 void	vfs_busy_pages_release(struct buf *bp);
556 struct buf *incore(struct bufobj *, daddr_t);
557 bool	inmem(struct vnode *, daddr_t);
558 struct buf *gbincore(struct bufobj *, daddr_t);
559 struct buf *gbincore_unlocked(struct bufobj *, daddr_t);
560 struct buf *getblk(struct vnode *, daddr_t, int, int, int, int);
561 int	getblkx(struct vnode *vp, daddr_t blkno, daddr_t dblkno, int size,
562 	    int slpflag, int slptimeo, int flags, struct buf **bpp);
563 struct buf *geteblk(int, int);
564 int	bufwait(struct buf *);
565 int	bufwrite(struct buf *);
566 void	bufdone(struct buf *);
567 void	bd_speedup(void);
568 
569 extern uma_zone_t pbuf_zone;
570 uma_zone_t pbuf_zsecond_create(const char *name, int max);
571 
572 int	cluster_read(struct vnode *, u_quad_t, daddr_t, long,
573 	    struct ucred *, long, int, int, struct buf **);
574 int	cluster_wbuild(struct vnode *, long, daddr_t, int, int);
575 void	cluster_write(struct vnode *, struct buf *, u_quad_t, int, int);
576 void	vfs_bio_brelse(struct buf *bp, int ioflags);
577 void	vfs_bio_bzero_buf(struct buf *bp, int base, int size);
578 void	vfs_bio_clrbuf(struct buf *);
579 void	vfs_bio_set_flags(struct buf *bp, int ioflags);
580 void	vfs_bio_set_valid(struct buf *, int base, int size);
581 void	vfs_busy_pages(struct buf *, int clear_modify);
582 void	vfs_unbusy_pages(struct buf *);
583 int	vmapbuf(struct buf *, void *, size_t, int);
584 void	vunmapbuf(struct buf *);
585 void	brelvp(struct buf *);
586 void	bgetvp(struct vnode *, struct buf *);
587 void	pbgetbo(struct bufobj *bo, struct buf *bp);
588 void	pbgetvp(struct vnode *, struct buf *);
589 void	pbrelbo(struct buf *);
590 void	pbrelvp(struct buf *);
591 int	allocbuf(struct buf *bp, int size);
592 void	reassignbuf(struct buf *);
593 void	bwait(struct buf *, u_char, const char *);
594 void	bdone(struct buf *);
595 
596 typedef daddr_t (vbg_get_lblkno_t)(struct vnode *, vm_ooffset_t);
597 typedef int (vbg_get_blksize_t)(struct vnode *, daddr_t, long *);
598 int	vfs_bio_getpages(struct vnode *vp, struct vm_page **ma, int count,
599 	    int *rbehind, int *rahead, vbg_get_lblkno_t get_lblkno,
600 	    vbg_get_blksize_t get_blksize);
601 
602 #endif /* _KERNEL */
603 
604 #endif /* !_SYS_BUF_H_ */
605