xref: /freebsd/sys/sys/buf.h (revision 4efe531c9d50a803a28d001fab9cc3011eb1f587)
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 
37 #ifndef _SYS_BUF_H_
38 #define	_SYS_BUF_H_
39 
40 #include <sys/bufobj.h>
41 #include <sys/queue.h>
42 #include <sys/lock.h>
43 #include <sys/lockmgr.h>
44 #include <vm/uma.h>
45 
46 struct bio;
47 struct buf;
48 struct bufobj;
49 struct mount;
50 struct vnode;
51 struct uio;
52 
53 /*
54  * To avoid including <ufs/ffs/softdep.h>
55  */
56 LIST_HEAD(workhead, worklist);
57 /*
58  * These are currently used only by the soft dependency code, hence
59  * are stored once in a global variable. If other subsystems wanted
60  * to use these hooks, a pointer to a set of bio_ops could be added
61  * to each buffer.
62  */
63 extern struct bio_ops {
64 	void	(*io_start)(struct buf *);
65 	void	(*io_complete)(struct buf *);
66 	void	(*io_deallocate)(struct buf *);
67 	int	(*io_countdeps)(struct buf *, int);
68 } bioops;
69 
70 struct vm_object;
71 struct vm_page;
72 
73 typedef uint32_t b_xflags_t;
74 
75 /*
76  * The buffer header describes an I/O operation in the kernel.
77  *
78  * NOTES:
79  *	b_bufsize, b_bcount.  b_bufsize is the allocation size of the
80  *	buffer, either DEV_BSIZE or PAGE_SIZE aligned.  b_bcount is the
81  *	originally requested buffer size and can serve as a bounds check
82  *	against EOF.  For most, but not all uses, b_bcount == b_bufsize.
83  *
84  *	b_dirtyoff, b_dirtyend.  Buffers support piecemeal, unaligned
85  *	ranges of dirty data that need to be written to backing store.
86  *	The range is typically clipped at b_bcount ( not b_bufsize ).
87  *
88  *	b_resid.  Number of bytes remaining in I/O.  After an I/O operation
89  *	completes, b_resid is usually 0 indicating 100% success.
90  *
91  *	All fields are protected by the buffer lock except those marked:
92  *		V - Protected by owning bufobj lock
93  *		Q - Protected by the buf queue lock
94  *		D - Protected by an dependency implementation specific lock
95  */
96 struct buf {
97 	struct bufobj	*b_bufobj;
98 	long		b_bcount;
99 	void		*b_caller1;
100 	caddr_t		b_data;
101 	int		b_error;
102 	uint16_t	b_iocmd;	/* BIO_* bio_cmd from bio.h */
103 	uint16_t	b_ioflags;	/* BIO_* bio_flags from bio.h */
104 	off_t		b_iooffset;
105 	long		b_resid;
106 	void	(*b_iodone)(struct buf *);
107 	void	(*b_ckhashcalc)(struct buf *);
108 	uint64_t	b_ckhash;	/* B_CKHASH requested check-hash */
109 	daddr_t b_blkno;		/* Underlying physical block number. */
110 	off_t	b_offset;		/* Offset into file. */
111 	TAILQ_ENTRY(buf) b_bobufs;	/* (V) Buffer's associated vnode. */
112 	uint32_t	b_vflags;	/* (V) BV_* flags */
113 	uint8_t		b_qindex;	/* (Q) buffer queue index */
114 	uint8_t		b_domain;	/* (Q) buf domain this resides in */
115 	uint16_t	b_subqueue;	/* (Q) per-cpu q if any */
116 	uint32_t	b_flags;	/* B_* flags. */
117 	b_xflags_t b_xflags;		/* extra flags */
118 	struct lock b_lock;		/* Buffer lock */
119 	long	b_bufsize;		/* Allocated buffer size. */
120 	int	b_runningbufspace;	/* when I/O is running, pipelining */
121 	int	b_kvasize;		/* size of kva for buffer */
122 	int	b_dirtyoff;		/* Offset in buffer of dirty region. */
123 	int	b_dirtyend;		/* Offset of end of dirty region. */
124 	caddr_t	b_kvabase;		/* base kva for buffer */
125 	daddr_t b_lblkno;		/* Logical block number. */
126 	struct	vnode *b_vp;		/* Device vnode. */
127 	struct	ucred *b_rcred;		/* Read credentials reference. */
128 	struct	ucred *b_wcred;		/* Write credentials reference. */
129 	union {
130 		TAILQ_ENTRY(buf) b_freelist; /* (Q) */
131 		struct {
132 			void	(*b_pgiodone)(void *, struct vm_page **,
133 				    int, int);
134 			int	b_pgbefore;
135 			int	b_pgafter;
136 		};
137 	};
138 	union	cluster_info {
139 		TAILQ_HEAD(cluster_list_head, buf) cluster_head;
140 		TAILQ_ENTRY(buf) cluster_entry;
141 	} b_cluster;
142 	int		b_npages;
143 	struct	workhead b_dep;		/* (D) List of filesystem dependencies. */
144 	void	*b_fsprivate1;
145 	void	*b_fsprivate2;
146 	void	*b_fsprivate3;
147 
148 #if defined(FULL_BUF_TRACKING)
149 #define BUF_TRACKING_SIZE	32
150 #define BUF_TRACKING_ENTRY(x)	((x) & (BUF_TRACKING_SIZE - 1))
151 	const char	*b_io_tracking[BUF_TRACKING_SIZE];
152 	uint32_t	b_io_tcnt;
153 #elif defined(BUF_TRACKING)
154 	const char	*b_io_tracking;
155 #endif
156 	struct	vm_page *b_pages[];
157 };
158 
159 #define b_object	b_bufobj->bo_object
160 
161 /*
162  * These flags are kept in b_flags.
163  *
164  * Notes:
165  *
166  *	B_ASYNC		VOP calls on bp's are usually async whether or not
167  *			B_ASYNC is set, but some subsystems, such as NFS, like
168  *			to know what is best for the caller so they can
169  *			optimize the I/O.
170  *
171  *	B_PAGING	Indicates that bp is being used by the paging system or
172  *			some paging system and that the bp is not linked into
173  *			the b_vp's clean/dirty linked lists or ref counts.
174  *			Buffer vp reassignments are illegal in this case.
175  *
176  *	B_CACHE		This may only be set if the buffer is entirely valid.
177  *			The situation where B_DELWRI is set and B_CACHE is
178  *			clear MUST be committed to disk by getblk() so
179  *			B_DELWRI can also be cleared.  See the comments for
180  *			getblk() in kern/vfs_bio.c.  If B_CACHE is clear,
181  *			the caller is expected to clear BIO_ERROR and B_INVAL,
182  *			set BIO_READ, and initiate an I/O.
183  *
184  *			The 'entire buffer' is defined to be the range from
185  *			0 through b_bcount.
186  *
187  *	B_MALLOC	Request that the buffer be allocated from the malloc
188  *			pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
189  *
190  *	B_CLUSTEROK	This flag is typically set for B_DELWRI buffers
191  *			by filesystems that allow clustering when the buffer
192  *			is fully dirty and indicates that it may be clustered
193  *			with other adjacent dirty buffers.  Note the clustering
194  *			may not be used with the stage 1 data write under NFS
195  *			but may be used for the commit rpc portion.
196  *
197  *	B_INVALONERR	This flag is set on dirty buffers.  It specifies that a
198  *			write error should forcibly invalidate the buffer
199  *			contents.  This flag should be used with caution, as it
200  *			discards data.  It is incompatible with B_ASYNC.
201  *
202  *	B_VMIO		Indicates that the buffer is tied into an VM object.
203  *			The buffer's data is always PAGE_SIZE aligned even
204  *			if b_bufsize and b_bcount are not.  ( b_bufsize is
205  *			always at least DEV_BSIZE aligned, though ).
206  *
207  *	B_DIRECT	Hint that we should attempt to completely free
208  *			the pages underlying the buffer.  B_DIRECT is
209  *			sticky until the buffer is released and typically
210  *			only has an effect when B_RELBUF is also set.
211  *
212  */
213 
214 #define	B_AGE		0x00000001	/* Move to age queue when I/O done. */
215 #define	B_NEEDCOMMIT	0x00000002	/* Append-write in progress. */
216 #define	B_ASYNC		0x00000004	/* Start I/O, do not wait. */
217 #define	B_DIRECT	0x00000008	/* direct I/O flag (pls free vmio) */
218 #define	B_DEFERRED	0x00000010	/* Skipped over for cleaning */
219 #define	B_CACHE		0x00000020	/* Bread found us in the cache. */
220 #define	B_VALIDSUSPWRT	0x00000040	/* Valid write during suspension. */
221 #define	B_DELWRI	0x00000080	/* Delay I/O until buffer reused. */
222 #define	B_CKHASH	0x00000100	/* checksum hash calculated on read */
223 #define	B_DONE		0x00000200	/* I/O completed. */
224 #define	B_EINTR		0x00000400	/* I/O was interrupted */
225 #define	B_NOREUSE	0x00000800	/* Contents not reused once released. */
226 #define	B_REUSE		0x00001000	/* Contents reused, second chance. */
227 #define	B_INVAL		0x00002000	/* Does not contain valid info. */
228 #define	B_BARRIER	0x00004000	/* Write this and all preceding first. */
229 #define	B_NOCACHE	0x00008000	/* Do not cache block after use. */
230 #define	B_MALLOC	0x00010000	/* malloced b_data */
231 #define	B_CLUSTEROK	0x00020000	/* Pagein op, so swap() can count it. */
232 #define	B_INVALONERR	0x00040000	/* Invalidate on write error. */
233 #define	B_IOSTARTED	0x00080000	/* buf_start() called */
234 #define	B_00100000	0x00100000	/* Available flag. */
235 #define	B_MAXPHYS	0x00200000	/* nitems(b_pages[]) = atop(MAXPHYS). */
236 #define	B_RELBUF	0x00400000	/* Release VMIO buffer. */
237 #define	B_FS_FLAG1	0x00800000	/* Available flag for FS use. */
238 #define	B_NOCOPY	0x01000000	/* Don't copy-on-write this buf. */
239 #define	B_INFREECNT	0x02000000	/* buf is counted in numfreebufs */
240 #define	B_PAGING	0x04000000	/* volatile paging I/O -- bypass VMIO */
241 #define B_MANAGED	0x08000000	/* Managed by FS. */
242 #define B_RAM		0x10000000	/* Read ahead mark (flag) */
243 #define B_VMIO		0x20000000	/* VMIO flag */
244 #define B_CLUSTER	0x40000000	/* pagein op, so swap() can count it */
245 #define B_REMFREE	0x80000000	/* Delayed bremfree */
246 
247 #define PRINT_BUF_FLAGS "\20\40remfree\37cluster\36vmio\35ram\34managed" \
248 	"\33paging\32infreecnt\31nocopy\30b23\27relbuf\26maxphys\25b20" \
249 	"\24iostarted\23invalonerr\22clusterok\21malloc\20nocache\17b14" \
250 	"\16inval\15reuse\14noreuse\13eintr\12done\11b8\10delwri" \
251 	"\7validsuspwrt\6cache\5deferred\4direct\3async\2needcommit\1age"
252 
253 /*
254  * These flags are kept in b_xflags.
255  *
256  * BX_FSPRIV reserves a set of eight flags that may be used by individual
257  * filesystems for their own purpose. Their specific definitions are
258  * found in the header files for each filesystem that uses them.
259  */
260 #define	BX_VNDIRTY	0x00000001	/* On vnode dirty list */
261 #define	BX_VNCLEAN	0x00000002	/* On vnode clean list */
262 #define	BX_CVTENXIO	0x00000004	/* Convert errors to ENXIO */
263 #define	BX_BKGRDWRITE	0x00000010	/* Do writes in background */
264 #define	BX_BKGRDMARKER	0x00000020	/* Mark buffer for splay tree */
265 #define	BX_ALTDATA	0x00000040	/* Holds extended data */
266 #define	BX_FSPRIV	0x00FF0000	/* Filesystem-specific flags mask */
267 
268 #define	PRINT_BUF_XFLAGS "\20\7altdata\6bkgrdmarker\5bkgrdwrite\3cvtenxio" \
269 	"\2clean\1dirty"
270 
271 #define	NOOFFSET	(-1LL)		/* No buffer offset calculated yet */
272 
273 /*
274  * These flags are kept in b_vflags.
275  */
276 #define	BV_SCANNED	0x00000001	/* VOP_FSYNC funcs mark written bufs */
277 #define	BV_BKGRDINPROG	0x00000002	/* Background write in progress */
278 #define	BV_BKGRDWAIT	0x00000004	/* Background write waiting */
279 #define	BV_BKGRDERR	0x00000008	/* Error from background write */
280 
281 #define	PRINT_BUF_VFLAGS "\20\4bkgrderr\3bkgrdwait\2bkgrdinprog\1scanned"
282 
283 #ifdef _KERNEL
284 
285 #ifndef NSWBUF_MIN
286 #define	NSWBUF_MIN	16
287 #endif
288 
289 /*
290  * Buffer locking
291  */
292 #include <sys/proc.h>			/* XXX for curthread */
293 #include <sys/mutex.h>
294 
295 /*
296  * Initialize a lock.
297  */
298 #define BUF_LOCKINIT(bp, wmesg)						\
299 	lockinit(&(bp)->b_lock, PRIBIO + 4, wmesg, 0, LK_NEW)
300 /*
301  *
302  * Get a lock sleeping non-interruptably until it becomes available.
303  */
304 #define	BUF_LOCK(bp, locktype, interlock)				\
305 	_lockmgr_args_rw(&(bp)->b_lock, (locktype), (interlock),	\
306 	    LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT,		\
307 	    LOCK_FILE, LOCK_LINE)
308 
309 /*
310  * Get a lock sleeping with specified interruptably and timeout.
311  */
312 #define	BUF_TIMELOCK(bp, locktype, interlock, wmesg, catch, timo)	\
313 	_lockmgr_args_rw(&(bp)->b_lock, (locktype) | LK_TIMELOCK,	\
314 	    (interlock), (wmesg), (PRIBIO + 4) | (catch), (timo),	\
315 	    LOCK_FILE, LOCK_LINE)
316 
317 /*
318  * Release a lock. Only the acquiring process may free the lock unless
319  * it has been handed off to biodone.
320  */
321 #define	BUF_UNLOCK(bp) do {						\
322 	KASSERT(((bp)->b_flags & B_REMFREE) == 0,			\
323 	    ("BUF_UNLOCK %p while B_REMFREE is still set.", (bp)));	\
324 									\
325 	BUF_UNLOCK_RAW((bp));						\
326 } while (0)
327 #define	BUF_UNLOCK_RAW(bp) do {						\
328 	(void)_lockmgr_args(&(bp)->b_lock, LK_RELEASE, NULL,		\
329 	    LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT,		\
330 	    LOCK_FILE, LOCK_LINE);					\
331 } while (0)
332 
333 /*
334  * Check if a buffer lock is recursed.
335  */
336 #define	BUF_LOCKRECURSED(bp)						\
337 	lockmgr_recursed(&(bp)->b_lock)
338 
339 /*
340  * Check if a buffer lock is currently held.
341  */
342 #define	BUF_ISLOCKED(bp)						\
343 	lockstatus(&(bp)->b_lock)
344 
345 /*
346  * Check if a buffer lock is currently held by LK_KERNPROC.
347  */
348 #define	BUF_DISOWNED(bp)						\
349 	lockmgr_disowned(&(bp)->b_lock)
350 
351 /*
352  * Free a buffer lock.
353  */
354 #define BUF_LOCKFREE(bp) 						\
355 	lockdestroy(&(bp)->b_lock)
356 
357 /*
358  * Print informations on a buffer lock.
359  */
360 #define BUF_LOCKPRINTINFO(bp) 						\
361 	lockmgr_printinfo(&(bp)->b_lock)
362 
363 /*
364  * Buffer lock assertions.
365  */
366 #if defined(INVARIANTS) && defined(INVARIANT_SUPPORT)
367 #define	BUF_ASSERT_LOCKED(bp)						\
368 	_lockmgr_assert(&(bp)->b_lock, KA_LOCKED, LOCK_FILE, LOCK_LINE)
369 #define	BUF_ASSERT_SLOCKED(bp)						\
370 	_lockmgr_assert(&(bp)->b_lock, KA_SLOCKED, LOCK_FILE, LOCK_LINE)
371 #define	BUF_ASSERT_XLOCKED(bp)						\
372 	_lockmgr_assert(&(bp)->b_lock, KA_XLOCKED, LOCK_FILE, LOCK_LINE)
373 #define	BUF_ASSERT_UNLOCKED(bp)						\
374 	_lockmgr_assert(&(bp)->b_lock, KA_UNLOCKED, LOCK_FILE, LOCK_LINE)
375 #else
376 #define	BUF_ASSERT_LOCKED(bp)
377 #define	BUF_ASSERT_SLOCKED(bp)
378 #define	BUF_ASSERT_XLOCKED(bp)
379 #define	BUF_ASSERT_UNLOCKED(bp)
380 #endif
381 
382 #ifdef _SYS_PROC_H_	/* Avoid #include <sys/proc.h> pollution */
383 /*
384  * When initiating asynchronous I/O, change ownership of the lock to the
385  * kernel. Once done, the lock may legally released by biodone. The
386  * original owning process can no longer acquire it recursively, but must
387  * wait until the I/O is completed and the lock has been freed by biodone.
388  */
389 #define	BUF_KERNPROC(bp)						\
390 	_lockmgr_disown(&(bp)->b_lock, LOCK_FILE, LOCK_LINE)
391 #endif
392 
393 #endif /* _KERNEL */
394 
395 struct buf_queue_head {
396 	TAILQ_HEAD(buf_queue, buf) queue;
397 	daddr_t last_pblkno;
398 	struct	buf *insert_point;
399 	struct	buf *switch_point;
400 };
401 
402 /*
403  * This structure describes a clustered I/O.
404  */
405 struct cluster_save {
406 	long	bs_bcount;		/* Saved b_bcount. */
407 	long	bs_bufsize;		/* Saved b_bufsize. */
408 	int	bs_nchildren;		/* Number of associated buffers. */
409 	struct buf **bs_children;	/* List of associated buffers. */
410 };
411 
412 /*
413  * Vnode clustering tracker
414  */
415 struct vn_clusterw {
416 	daddr_t	v_cstart;			/* v start block of cluster */
417 	daddr_t	v_lasta;			/* v last allocation  */
418 	daddr_t	v_lastw;			/* v last write  */
419 	int	v_clen;				/* v length of cur. cluster */
420 };
421 
422 #ifdef _KERNEL
423 
424 static __inline int
bwrite(struct buf * bp)425 bwrite(struct buf *bp)
426 {
427 
428 	KASSERT(bp->b_bufobj != NULL, ("bwrite: no bufobj bp=%p", bp));
429 	KASSERT(bp->b_bufobj->bo_ops != NULL, ("bwrite: no bo_ops bp=%p", bp));
430 	KASSERT(bp->b_bufobj->bo_ops->bop_write != NULL,
431 	    ("bwrite: no bop_write bp=%p", bp));
432 	return (BO_WRITE(bp->b_bufobj, bp));
433 }
434 
435 static __inline void
bstrategy(struct buf * bp)436 bstrategy(struct buf *bp)
437 {
438 
439 	KASSERT(bp->b_bufobj != NULL, ("bstrategy: no bufobj bp=%p", bp));
440 	KASSERT(bp->b_bufobj->bo_ops != NULL,
441 	    ("bstrategy: no bo_ops bp=%p", bp));
442 	KASSERT(bp->b_bufobj->bo_ops->bop_strategy != NULL,
443 	    ("bstrategy: no bop_strategy bp=%p", bp));
444 	BO_STRATEGY(bp->b_bufobj, bp);
445 }
446 
447 static __inline void
buf_start(struct buf * bp)448 buf_start(struct buf *bp)
449 {
450 	KASSERT((bp->b_flags & B_IOSTARTED) == 0,
451 	    ("recursed buf_start %p", bp));
452 	bp->b_flags |= B_IOSTARTED;
453 	if (bioops.io_start)
454 		(*bioops.io_start)(bp);
455 }
456 
457 static __inline void
buf_complete(struct buf * bp)458 buf_complete(struct buf *bp)
459 {
460 	if ((bp->b_flags & B_IOSTARTED) != 0) {
461 		bp->b_flags &= ~B_IOSTARTED;
462 		if (bioops.io_complete)
463 			(*bioops.io_complete)(bp);
464 	}
465 }
466 
467 static __inline void
buf_deallocate(struct buf * bp)468 buf_deallocate(struct buf *bp)
469 {
470 	if (bioops.io_deallocate)
471 		(*bioops.io_deallocate)(bp);
472 }
473 
474 static __inline int
buf_countdeps(struct buf * bp,int i)475 buf_countdeps(struct buf *bp, int i)
476 {
477 	if (bioops.io_countdeps)
478 		return ((*bioops.io_countdeps)(bp, i));
479 	else
480 		return (0);
481 }
482 
483 static __inline void
buf_track(struct buf * bp __unused,const char * location __unused)484 buf_track(struct buf *bp __unused, const char *location __unused)
485 {
486 
487 #if defined(FULL_BUF_TRACKING)
488 	bp->b_io_tracking[BUF_TRACKING_ENTRY(bp->b_io_tcnt++)] = location;
489 #elif defined(BUF_TRACKING)
490 	bp->b_io_tracking = location;
491 #endif
492 }
493 
494 #endif /* _KERNEL */
495 
496 /*
497  * Zero out the buffer's data area.
498  */
499 #define	clrbuf(bp) {							\
500 	bzero((bp)->b_data, (u_int)(bp)->b_bcount);			\
501 	(bp)->b_resid = 0;						\
502 }
503 
504 /*
505  * Flags for getblk's last parameter.
506  */
507 #define	GB_LOCK_NOWAIT	0x0001		/* Fail if we block on a buf lock. */
508 #define	GB_NOCREAT	0x0002		/* Don't create a buf if not found. */
509 #define	GB_NOWAIT_BD	0x0004		/* Do not wait for bufdaemon. */
510 #define	GB_UNMAPPED	0x0008		/* Do not mmap buffer pages. */
511 #define	GB_KVAALLOC	0x0010		/* But allocate KVA. */
512 #define	GB_CKHASH	0x0020		/* If reading, calc checksum hash */
513 #define	GB_NOSPARSE	0x0040		/* Do not instantiate holes */
514 #define	GB_CVTENXIO	0x0080		/* Convert errors to ENXIO */
515 #define	GB_NOWITNESS	0x0100		/* Do not record for WITNESS */
516 
517 #ifdef _KERNEL
518 extern int	nbuf;			/* The number of buffer headers */
519 extern u_long	maxswzone;		/* Max KVA for swap structures */
520 extern u_long	maxbcache;		/* Max KVA for buffer cache */
521 extern int	maxbcachebuf;		/* Max buffer cache block size */
522 extern long	hibufspace;
523 extern int	dirtybufthresh;
524 extern int	bdwriteskip;
525 extern int	dirtybufferflushes;
526 extern int	altbufferflushes;
527 extern int	nswbuf;			/* Number of swap I/O buffer headers. */
528 extern caddr_t __read_mostly unmapped_buf; /* Data address for unmapped
529 					      buffers. */
530 
531 static inline int
buf_mapped(struct buf * bp)532 buf_mapped(struct buf *bp)
533 {
534 
535 	return (bp->b_data != unmapped_buf);
536 }
537 
538 long	runningbufclaim(struct buf *, int);
539 void	runningbufwakeup(struct buf *);
540 void	waitrunningbufspace(void);
541 caddr_t	kern_vfs_bio_buffer_alloc(caddr_t v, long physmem_est);
542 void	bufinit(void);
543 void	bufshutdown(int);
544 void	bdata2bio(struct buf *bp, struct bio *bip);
545 void	bwillwrite(void);
546 int	buf_dirty_count_severe(void);
547 void	bremfree(struct buf *);
548 void	bremfreef(struct buf *);	/* XXX Force bremfree, only for nfs. */
549 #define bread(vp, blkno, size, cred, bpp) \
550 	    breadn_flags(vp, blkno, blkno, size, NULL, NULL, 0, cred, 0, \
551 		NULL, bpp)
552 #define bread_gb(vp, blkno, size, cred, gbflags, bpp) \
553 	    breadn_flags(vp, blkno, blkno, size, NULL, NULL, 0, cred, \
554 		gbflags, NULL, bpp)
555 #define breadn(vp, blkno, size, rablkno, rabsize, cnt, cred, bpp) \
556 	    breadn_flags(vp, blkno, blkno, size, rablkno, rabsize, cnt, cred, \
557 		0, NULL, bpp)
558 int	breadn_flags(struct vnode *, daddr_t, daddr_t, int, daddr_t *, int *,
559 	    int, struct ucred *, int, void (*)(struct buf *), struct buf **);
560 void	bdwrite(struct buf *);
561 void	bawrite(struct buf *);
562 void	babarrierwrite(struct buf *);
563 int	bbarrierwrite(struct buf *);
564 void	bdirty(struct buf *);
565 void	bundirty(struct buf *);
566 void	bufstrategy(struct bufobj *, struct buf *);
567 void	brelse(struct buf *);
568 void	bqrelse(struct buf *);
569 int	vfs_bio_awrite(struct buf *);
570 void	vfs_busy_pages_acquire(struct buf *bp);
571 void	vfs_busy_pages_release(struct buf *bp);
572 struct buf *incore(struct bufobj *, daddr_t);
573 bool	inmem(struct vnode *, daddr_t);
574 struct buf *gbincore(struct bufobj *, daddr_t);
575 struct buf *gbincore_unlocked(struct bufobj *, daddr_t);
576 struct buf *getblk(struct vnode *, daddr_t, int, int, int, int);
577 int	getblkx(struct vnode *vp, daddr_t blkno, daddr_t dblkno, int size,
578 	    int slpflag, int slptimeo, int flags, struct buf **bpp);
579 struct buf *geteblk(int, int);
580 int	bufwait(struct buf *);
581 int	bufwrite(struct buf *);
582 void	bufdone(struct buf *);
583 void	bd_speedup(void);
584 
585 extern uma_zone_t pbuf_zone;
586 uma_zone_t pbuf_zsecond_create(const char *name, int max);
587 
588 struct vn_clusterw;
589 
590 void	cluster_init_vn(struct vn_clusterw *vnc);
591 int	cluster_read(struct vnode *, u_quad_t, daddr_t, long,
592 	    struct ucred *, long, int, int, struct buf **);
593 int	cluster_wbuild(struct vnode *, long, daddr_t, int, int);
594 void	cluster_write(struct vnode *, struct vn_clusterw *, struct buf *,
595 	    u_quad_t, int, int);
596 void	vfs_bio_brelse(struct buf *bp, int ioflags);
597 void	vfs_bio_bzero_buf(struct buf *bp, int base, int size);
598 void	vfs_bio_clrbuf(struct buf *);
599 void	vfs_bio_set_flags(struct buf *bp, int ioflags);
600 void	vfs_bio_set_valid(struct buf *, int base, int size);
601 void	vfs_busy_pages(struct buf *, int clear_modify);
602 void	vfs_unbusy_pages(struct buf *);
603 int	vmapbuf(struct buf *, void *, size_t, int);
604 void	vunmapbuf(struct buf *);
605 void	brelvp(struct buf *);
606 int	bgetvp(struct vnode *, struct buf *) __result_use_check;
607 void	pbgetbo(struct bufobj *bo, struct buf *bp);
608 void	pbgetvp(struct vnode *, struct buf *);
609 void	pbrelbo(struct buf *);
610 void	pbrelvp(struct buf *);
611 int	allocbuf(struct buf *bp, int size);
612 void	reassignbuf(struct buf *);
613 void	bwait(struct buf *, u_char, const char *);
614 void	bdone(struct buf *);
615 
616 typedef daddr_t (vbg_get_lblkno_t)(struct vnode *, vm_ooffset_t);
617 typedef int (vbg_get_blksize_t)(struct vnode *, daddr_t, long *);
618 int	vfs_bio_getpages(struct vnode *vp, struct vm_page **ma, int count,
619 	    int *rbehind, int *rahead, vbg_get_lblkno_t get_lblkno,
620 	    vbg_get_blksize_t get_blksize);
621 
622 #endif /* _KERNEL */
623 
624 #endif /* !_SYS_BUF_H_ */
625