1 /*-
2  * Copyright (c) 2004 Poul-Henning Kamp
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  *
26  * $FreeBSD: stable/9/sys/sys/bufobj.h 183754 2008-10-10 21:23:50Z attilio $
27  */
28 
29 /*
30  * Architectural notes:
31  *
32  * bufobj is a new object which is what buffers hang from in the buffer
33  * cache.
34  *
35  * This used to be vnodes, but we need non-vnode code to be able
36  * to use the buffer cache as well, specifically geom classes like gbde,
37  * raid3 and raid5.
38  *
39  * All vnodes will contain a bufobj initially, but down the road we may
40  * want to only allocate bufobjs when they are needed.  There could be a
41  * large number of vnodes in the system which wouldn't need a bufobj during
42  * their lifetime.
43  *
44  * The exact relationship to the vmobject is not determined at this point,
45  * it may in fact be that we find them to be two sides of the same object
46  * once things starts to crystalize.
47  */
48 
49 #ifndef _SYS_BUFOBJ_H_
50 #define _SYS_BUFOBJ_H_
51 
52 #if defined(_KERNEL) || defined(_KVM_VNODE)
53 
54 #include <sys/queue.h>
55 #include <sys/_lock.h>
56 #include <sys/_mutex.h>
57 
58 struct bufobj;
59 struct buf_ops;
60 
61 extern struct buf_ops buf_ops_bio;
62 
63 TAILQ_HEAD(buflists, buf);
64 
65 /* A Buffer splay list */
66 struct bufv {
67 	struct buflists	bv_hd;		/* Sorted blocklist */
68 	struct buf	*bv_root;	/* Buf splay tree */
69 	int		bv_cnt;		/* Number of buffers */
70 };
71 
72 typedef void b_strategy_t(struct bufobj *, struct buf *);
73 typedef int b_write_t(struct buf *);
74 typedef int b_sync_t(struct bufobj *, int waitfor);
75 typedef void b_bdflush_t(struct bufobj *, struct buf *);
76 
77 struct buf_ops {
78 	char		*bop_name;
79 	b_write_t	*bop_write;
80 	b_strategy_t	*bop_strategy;
81 	b_sync_t	*bop_sync;
82 	b_bdflush_t	*bop_bdflush;
83 };
84 
85 #define BO_STRATEGY(bo, bp)	((bo)->bo_ops->bop_strategy((bo), (bp)))
86 #define BO_SYNC(bo, w)		((bo)->bo_ops->bop_sync((bo), (w)))
87 #define BO_WRITE(bo, bp)	((bo)->bo_ops->bop_write((bp)))
88 #define BO_BDFLUSH(bo, bp)	((bo)->bo_ops->bop_bdflush((bo), (bp)))
89 
90 struct bufobj {
91 	struct mtx	bo_mtx;		/* Mutex which protects "i" things */
92 	struct bufv	bo_clean;	/* i Clean buffers */
93 	struct bufv	bo_dirty;	/* i Dirty buffers */
94 	long		bo_numoutput;	/* i Writes in progress */
95 	u_int		bo_flag;	/* i Flags */
96 	struct buf_ops	*bo_ops;	/* - Buffer operations */
97 	int		bo_bsize;	/* - Block size for i/o */
98 	struct vm_object *bo_object;	/* v Place to store VM object */
99 	LIST_ENTRY(bufobj) bo_synclist;	/* S dirty vnode list */
100 	void		*bo_private;	/* private pointer */
101 	struct vnode	*__bo_vnode;	/*
102 					 * XXX: This vnode pointer is here
103 					 * XXX: only to keep the syncer working
104 					 * XXX: for now.
105 					 */
106 };
107 
108 /*
109  * XXX BO_ONWORKLST could be replaced with a check for NULL list elements
110  * in v_synclist.
111  */
112 #define	BO_ONWORKLST	(1 << 0)	/* On syncer work-list */
113 #define	BO_WWAIT	(1 << 1)	/* Wait for output to complete */
114 #define	BO_NEEDSGIANT	(1 << 2)	/* Require giant for child buffers. */
115 
116 #define	BO_MTX(bo)		(&(bo)->bo_mtx)
117 #define	BO_LOCK(bo)		mtx_lock(BO_MTX((bo)))
118 #define	BO_UNLOCK(bo)		mtx_unlock(BO_MTX((bo)))
119 #define	ASSERT_BO_LOCKED(bo)	mtx_assert(BO_MTX((bo)), MA_OWNED)
120 #define	ASSERT_BO_UNLOCKED(bo)	mtx_assert(BO_MTX((bo)), MA_NOTOWNED)
121 
122 void bufobj_wdrop(struct bufobj *bo);
123 void bufobj_wref(struct bufobj *bo);
124 void bufobj_wrefl(struct bufobj *bo);
125 int bufobj_invalbuf(struct bufobj *bo, int flags, int slpflag, int slptimeo);
126 int bufobj_wwait(struct bufobj *bo, int slpflag, int timeo);
127 int bufsync(struct bufobj *bo, int waitfor);
128 void bufbdflush(struct bufobj *bo, struct buf *bp);
129 
130 #endif /* defined(_KERNEL) || defined(_KVM_VNODE) */
131 #endif /* _SYS_BUFOBJ_H_ */
132