Add missing commit for the VM load heuristic and page allocation rate
[dragonfly.git] / sys / sys / buf.h
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1982, 1986, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, 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 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 *
38 * @(#)buf.h 8.9 (Berkeley) 3/30/95
39 * $FreeBSD: src/sys/sys/buf.h,v 1.88.2.10 2003/01/25 19:02:23 dillon Exp $
1f1ea522 40 * $DragonFly: src/sys/sys/buf.h,v 1.24 2006/03/05 18:38:36 dillon Exp $
984263bc
MD
41 */
42
43#ifndef _SYS_BUF_H_
44#define _SYS_BUF_H_
45
335dda38 46#ifndef _SYS_QUEUE_H_
984263bc 47#include <sys/queue.h>
335dda38
MD
48#endif
49#ifndef _SYS_LOCK_H_
984263bc 50#include <sys/lock.h>
335dda38
MD
51#endif
52#ifndef _SYS_DEVICE_H_
53#include <sys/device.h>
54#endif
984263bc 55
54f51aeb
HP
56#ifndef _SYS_XIO_H_
57#include <sys/xio.h>
58#endif
6bae6177
MD
59#ifndef _SYS_TREE_H_
60#include <sys/tree.h>
61#endif
1eabbf32
HP
62#ifndef _SYS_BIO_H_
63#include <sys/bio.h>
64#endif
16523a43
MD
65#ifndef _SYS_SPINLOCK_H_
66#include <sys/spinlock.h>
67#endif
1eabbf32 68
984263bc 69struct buf;
81b5c339 70struct bio;
984263bc
MD
71struct mount;
72struct vnode;
54f51aeb 73struct xio;
984263bc 74
81b5c339
MD
75#define NBUF_BIO 4
76
6bae6177 77struct buf_rb_tree;
1f1ea522
MD
78struct buf_rb_hash;
79RB_PROTOTYPE2(buf_rb_tree, buf, b_rbnode, rb_buf_compare, daddr_t, b_lblkno);
80RB_PROTOTYPE2(buf_rb_hash, buf, b_rbhash, rb_buf_compare, daddr_t, b_lblkno);
6bae6177 81
984263bc
MD
82/*
83 * To avoid including <ufs/ffs/softdep.h>
84 */
85LIST_HEAD(workhead, worklist);
86/*
87 * These are currently used only by the soft dependency code, hence
88 * are stored once in a global variable. If other subsystems wanted
89 * to use these hooks, a pointer to a set of bio_ops could be added
90 * to each buffer.
91 */
92extern struct bio_ops {
b153f746
RG
93 void (*io_start) (struct buf *);
94 void (*io_complete) (struct buf *);
95 void (*io_deallocate) (struct buf *);
1e5a0c40
HP
96 int (*io_fsync) (struct vnode *);
97 int (*io_sync) (struct mount *);
b153f746 98 void (*io_movedeps) (struct buf *, struct buf *);
1e5a0c40 99 int (*io_countdeps) (struct buf *, int);
984263bc
MD
100} bioops;
101
984263bc
MD
102/*
103 * The buffer header describes an I/O operation in the kernel.
104 *
105 * NOTES:
106 * b_bufsize, b_bcount. b_bufsize is the allocation size of the
107 * buffer, either DEV_BSIZE or PAGE_SIZE aligned. b_bcount is the
108 * originally requested buffer size and can serve as a bounds check
109 * against EOF. For most, but not all uses, b_bcount == b_bufsize.
110 *
111 * b_dirtyoff, b_dirtyend. Buffers support piecemeal, unaligned
112 * ranges of dirty data that need to be written to backing store.
113 * The range is typically clipped at b_bcount ( not b_bufsize ).
114 *
115 * b_resid. Number of bytes remaining in I/O. After an I/O operation
116 * completes, b_resid is usually 0 indicating 100% success.
81b5c339
MD
117 *
118 * b_bio1 and b_bio2 represent the two primary I/O layers. Additional
119 * I/O layers are allocated out of the object cache and may also exist.
120 *
121 * b_bio1 is the logical layer and contains offset or block number
122 * data for the primary vnode, b_vp. I/O operations are almost
123 * universally initiated from the logical layer, so you will often
124 * see things like: vn_strategy(bp->b_vp, &bp->b_bio1).
125 *
126 * b_bio2 is the first physical layer (typically the slice-relative
127 * layer) and contains the translated offset or block number for
128 * the block device underlying a filesystem. Filesystems such as UFS
129 * will maintain cached translations and you may see them initiate
130 * a 'physical' I/O using vn_strategy(devvp, &bp->b_bio2). BUT,
131 * remember that the layering is relative to bp->b_vp, so the
132 * device-relative block numbers for buffer cache operations that occur
133 * directly on a block device will be in the first BIO layer.
134 *
135 * NOTE!!! Only the BIO subsystem accesses b_bio1 and b_bio2 directly.
136 * ALL STRATEGY LAYERS FOR BOTH VNODES AND DEVICES ONLY ACCESS THE BIO
137 * PASSED TO THEM, AND WILL PUSH ANOTHER BIO LAYER IF FORWARDING THE
138 * I/O DEEPER. In particular, a vn_strategy() or dev_dstrategy()
139 * call should not ever access buf->b_vp as this vnode may be totally
140 * unrelated to the vnode/device whos strategy routine was called.
984263bc
MD
141 */
142struct buf {
1f1ea522
MD
143 RB_ENTRY(buf) b_rbnode; /* RB node in vnode clean/dirty tree */
144 RB_ENTRY(buf) b_rbhash; /* RB node in vnode hash tree */
984263bc 145 TAILQ_ENTRY(buf) b_freelist; /* Free list position if not active. */
81b5c339
MD
146 struct buf *b_cluster_next; /* Next buffer (cluster code) */
147 struct vnode *b_vp; /* (vp, lblkno) index */
148 struct bio b_bio_array[NBUF_BIO]; /* BIO translation layers */
984263bc
MD
149 long b_flags; /* B_* flags. */
150 unsigned short b_qindex; /* buffer queue index */
151 unsigned char b_xflags; /* extra flags */
152 struct lock b_lock; /* Buffer lock */
984263bc
MD
153 long b_bufsize; /* Allocated buffer size. */
154 long b_runningbufspace; /* when I/O is running, pipelining */
155 long b_bcount; /* Valid bytes in buffer. */
81b5c339
MD
156 long b_resid; /* Remaining I/O */
157 int b_error; /* Error return */
984263bc
MD
158 caddr_t b_data; /* Memory, superblocks, indirect etc. */
159 caddr_t b_kvabase; /* base kva for buffer */
160 int b_kvasize; /* size of kva for buffer */
984263bc
MD
161 int b_dirtyoff; /* Offset in buffer of dirty region. */
162 int b_dirtyend; /* Offset of end of dirty region. */
984263bc 163 void *b_saveaddr; /* Original b_addr for physio. */
81b5c339 164 struct xio b_xio; /* data buffer page list management */
984263bc 165 struct workhead b_dep; /* List of filesystem dependencies. */
984263bc
MD
166};
167
81b5c339
MD
168/*
169 * XXX temporary
170 */
171#define b_bio1 b_bio_array[0] /* logical layer */
172#define b_bio2 b_bio_array[1] /* (typically) the disk layer */
173#define b_loffset b_bio1.bio_offset
174#define b_lblkno b_bio1.bio_blkno
984263bc
MD
175
176/*
177 * These flags are kept in b_flags.
178 *
179 * Notes:
180 *
181 * B_ASYNC VOP calls on bp's are usually async whether or not
182 * B_ASYNC is set, but some subsystems, such as NFS, like
183 * to know what is best for the caller so they can
184 * optimize the I/O.
185 *
186 * B_PAGING Indicates that bp is being used by the paging system or
187 * some paging system and that the bp is not linked into
188 * the b_vp's clean/dirty linked lists or ref counts.
189 * Buffer vp reassignments are illegal in this case.
190 *
191 * B_CACHE This may only be set if the buffer is entirely valid.
192 * The situation where B_DELWRI is set and B_CACHE is
193 * clear MUST be committed to disk by getblk() so
194 * B_DELWRI can also be cleared. See the comments for
195 * getblk() in kern/vfs_bio.c. If B_CACHE is clear,
196 * the caller is expected to clear B_ERROR|B_INVAL,
197 * set B_READ, and initiate an I/O.
198 *
199 * The 'entire buffer' is defined to be the range from
200 * 0 through b_bcount.
201 *
202 * B_MALLOC Request that the buffer be allocated from the malloc
203 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
204 *
205 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers
206 * by filesystems that allow clustering when the buffer
207 * is fully dirty and indicates that it may be clustered
208 * with other adjacent dirty buffers. Note the clustering
209 * may not be used with the stage 1 data write under NFS
210 * but may be used for the commit rpc portion.
211 *
212 * B_VMIO Indicates that the buffer is tied into an VM object.
213 * The buffer's data is always PAGE_SIZE aligned even
214 * if b_bufsize and b_bcount are not. ( b_bufsize is
215 * always at least DEV_BSIZE aligned, though ).
216 *
217 * B_DIRECT Hint that we should attempt to completely free
218 * the pages underlying the buffer. B_DIRECT is
219 * sticky until the buffer is released and typically
220 * only has an effect when B_RELBUF is also set.
221 *
222 * B_NOWDRAIN This flag should be set when a device (like VN)
223 * does a turn-around VOP_WRITE from its strategy
224 * routine. This flag prevents bwrite() from blocking
225 * in wdrain, avoiding a deadlock situation.
226 */
227
228#define B_AGE 0x00000001 /* Move to age queue when I/O done. */
229#define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */
230#define B_ASYNC 0x00000004 /* Start I/O, do not wait. */
231#define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */
232#define B_DEFERRED 0x00000010 /* Skipped over for cleaning */
233#define B_CACHE 0x00000020 /* Bread found us in the cache. */
1f1ea522 234#define B_HASHED 0x00000040 /* Indexed via v_rbhash_tree */
984263bc
MD
235#define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */
236#define B_FREEBUF 0x00000100 /* Instruct driver: free blocks */
237#define B_DONE 0x00000200 /* I/O completed. */
238#define B_EINTR 0x00000400 /* I/O was interrupted */
239#define B_ERROR 0x00000800 /* I/O error occurred. */
71bc1d95 240#define B_UNUSED1000 0x00001000 /* Unused */
984263bc
MD
241#define B_INVAL 0x00002000 /* Does not contain valid info. */
242#define B_LOCKED 0x00004000 /* Locked in core (not reusable). */
243#define B_NOCACHE 0x00008000 /* Do not cache block after use. */
244#define B_MALLOC 0x00010000 /* malloced b_data */
245#define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */
246#define B_PHYS 0x00040000 /* I/O to user memory. */
247#define B_RAW 0x00080000 /* Set by physio for raw transfers. */
248#define B_READ 0x00100000 /* Read buffer. */
249#define B_DIRTY 0x00200000 /* Needs writing later. */
250#define B_RELBUF 0x00400000 /* Release VMIO buffer. */
251#define B_WANT 0x00800000 /* Used by vm_pager.c */
252#define B_WRITE 0x00000000 /* Write buffer (pseudo flag). */
71bc1d95 253#define B_UNUSED1000000 0x01000000 /* Unused */
984263bc
MD
254#define B_XXX 0x02000000 /* Debugging flag. */
255#define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */
256#define B_ORDERED 0x08000000 /* Must guarantee I/O ordering */
257#define B_RAM 0x10000000 /* Read ahead mark (flag) */
258#define B_VMIO 0x20000000 /* VMIO flag */
259#define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */
260#define B_NOWDRAIN 0x80000000 /* Avoid wdrain deadlock */
261
262#define PRINT_BUF_FLAGS "\20\40nowdrain\37cluster\36vmio\35ram\34ordered" \
263 "\33paging\32xxx\31writeinprog\30want\27relbuf\26dirty" \
264 "\25read\24raw\23phys\22clusterok\21malloc\20nocache" \
265 "\17locked\16inval\15scanned\14error\13eintr\12done\11freebuf" \
266 "\10delwri\7call\6cache\4direct\3async\2needcommit\1age"
267
268/*
269 * These flags are kept in b_xflags.
270 */
271#define BX_VNDIRTY 0x00000001 /* On vnode dirty list */
272#define BX_VNCLEAN 0x00000002 /* On vnode clean list */
273#define BX_BKGRDWRITE 0x00000004 /* Do writes in background */
274#define BX_BKGRDINPROG 0x00000008 /* Background write in progress */
275#define BX_BKGRDWAIT 0x00000010 /* Background write waiting */
276#define BX_AUTOCHAINDONE 0x00000020 /* pager I/O chain auto mode */
277
278#define NOOFFSET (-1LL) /* No buffer offset calculated yet */
279
280#ifdef _KERNEL
281/*
3020e3be 282 * Buffer locking. See sys/buf2.h for inline functions.
984263bc 283 */
984263bc
MD
284extern char *buf_wmesg; /* Default buffer lock message */
285#define BUF_WMESG "bufwait"
984263bc
MD
286
287#endif /* _KERNEL */
288
81b5c339
MD
289struct bio_queue_head {
290 TAILQ_HEAD(bio_queue, bio) queue;
291 daddr_t last_blkno;
292 struct bio *insert_point;
293 struct bio *switch_point;
984263bc
MD
294};
295
296/*
297 * This structure describes a clustered I/O. It is stored in the b_saveaddr
298 * field of the buffer on which I/O is done. At I/O completion, cluster
299 * callback uses the structure to parcel I/O's to individual buffers, and
300 * then free's this structure.
301 */
302struct cluster_save {
303 long bs_bcount; /* Saved b_bcount. */
304 long bs_bufsize; /* Saved b_bufsize. */
305 void *bs_saveaddr; /* Saved b_addr. */
306 int bs_nchildren; /* Number of associated buffers. */
307 struct buf **bs_children; /* List of associated buffers. */
308};
309
984263bc
MD
310/*
311 * Zero out the buffer's data area.
312 */
313#define clrbuf(bp) { \
314 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \
315 (bp)->b_resid = 0; \
316}
317
318/*
319 * Flags to low-level bitmap allocation routines (balloc).
320 *
321 * Note: sequential_heuristic() in kern/vfs_vnops.c limits the count
322 * to 127.
323 */
324#define B_SEQMASK 0x7F000000 /* Sequential heuristic mask. */
325#define B_SEQSHIFT 24 /* Sequential heuristic shift. */
326#define B_SEQMAX 0x7F
327#define B_CLRBUF 0x01 /* Cleared invalid areas of buffer. */
328#define B_SYNC 0x02 /* Do all allocations synchronously. */
329
330#ifdef _KERNEL
331extern int nbuf; /* The number of buffer headers */
332extern int maxswzone; /* Max KVA for swap structures */
333extern int maxbcache; /* Max KVA for buffer cache */
334extern int runningbufspace;
335extern int buf_maxio; /* nominal maximum I/O for buffer */
336extern struct buf *buf; /* The buffer headers. */
337extern char *buffers; /* The buffer contents. */
338extern int bufpages; /* Number of memory pages in the buffer pool. */
339extern struct buf *swbuf; /* Swap I/O buffer headers. */
340extern int nswbuf; /* Number of swap I/O buffer headers. */
984263bc
MD
341
342struct uio;
343
b153f746
RG
344void bufinit (void);
345void bwillwrite (void);
346int buf_dirty_count_severe (void);
81b5c339
MD
347void initbufbio(struct buf *);
348void reinitbufbio(struct buf *);
349void clearbiocache(struct bio *);
b153f746
RG
350void bremfree (struct buf *);
351int bread (struct vnode *, daddr_t, int, struct buf **);
352int breadn (struct vnode *, daddr_t, int, daddr_t *, int *, int,
353 struct buf **);
354int bwrite (struct buf *);
355void bdwrite (struct buf *);
356void bawrite (struct buf *);
357void bdirty (struct buf *);
358void bundirty (struct buf *);
359int bowrite (struct buf *);
360void brelse (struct buf *);
361void bqrelse (struct buf *);
362int vfs_bio_awrite (struct buf *);
81b5c339 363struct buf *getpbuf (int *);
b153f746 364int inmem (struct vnode *, daddr_t);
1f1ea522 365struct buf *findblk (struct vnode *, daddr_t);
b153f746
RG
366struct buf *getblk (struct vnode *, daddr_t, int, int, int);
367struct buf *geteblk (int);
81b5c339
MD
368struct bio *push_bio(struct bio *);
369void pop_bio(struct bio *);
b153f746 370int biowait (struct buf *);
81b5c339 371void biodone (struct bio *);
984263bc 372
81b5c339 373void cluster_append(struct bio *, struct buf *);
b153f746
RG
374int cluster_read (struct vnode *, u_quad_t, daddr_t, long,
375 long, int, struct buf **);
376int cluster_wbuild (struct vnode *, long, daddr_t, int);
377void cluster_write (struct buf *, u_quad_t, int);
378int physio (dev_t dev, struct uio *uio, int ioflag);
984263bc
MD
379#define physread physio
380#define physwrite physio
b153f746
RG
381void vfs_bio_set_validclean (struct buf *, int base, int size);
382void vfs_bio_clrbuf (struct buf *);
383void vfs_busy_pages (struct buf *, int clear_modify);
384void vfs_unbusy_pages (struct buf *);
b153f746
RG
385int vmapbuf (struct buf *);
386void vunmapbuf (struct buf *);
387void relpbuf (struct buf *, int *);
388void brelvp (struct buf *);
389void bgetvp (struct vnode *, struct buf *);
390void pbgetvp (struct vnode *, struct buf *);
391void pbrelvp (struct buf *);
392int allocbuf (struct buf *bp, int size);
1f1ea522 393void reassignbuf (struct buf *);
b153f746 394struct buf *trypbuf (int *);
984263bc
MD
395
396#endif /* _KERNEL */
397
398#endif /* !_SYS_BUF_H_ */