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.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
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.
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
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 $
40 * $DragonFly: src/sys/sys/buf.h,v 1.3 2003/06/19 01:55:07 dillon Exp $
46 #include <sys/queue.h>
54 * To avoid including <ufs/ffs/softdep.h>
56 LIST_HEAD(workhead, worklist);
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
63 extern struct bio_ops {
64 void (*io_start) __P((struct buf *));
65 void (*io_complete) __P((struct buf *));
66 void (*io_deallocate) __P((struct buf *));
67 int (*io_fsync) __P((struct vnode *));
68 int (*io_sync) __P((struct mount *));
69 void (*io_movedeps) __P((struct buf *, struct buf *));
70 int (*io_countdeps) __P((struct buf *, int));
75 void (*ic_prev_iodone) __P((struct buf *));
76 void *ic_prev_iodone_chain;
84 * The buffer header describes an I/O operation in the kernel.
87 * b_bufsize, b_bcount. b_bufsize is the allocation size of the
88 * buffer, either DEV_BSIZE or PAGE_SIZE aligned. b_bcount is the
89 * originally requested buffer size and can serve as a bounds check
90 * against EOF. For most, but not all uses, b_bcount == b_bufsize.
92 * b_dirtyoff, b_dirtyend. Buffers support piecemeal, unaligned
93 * ranges of dirty data that need to be written to backing store.
94 * The range is typically clipped at b_bcount ( not b_bufsize ).
96 * b_resid. Number of bytes remaining in I/O. After an I/O operation
97 * completes, b_resid is usually 0 indicating 100% success.
100 LIST_ENTRY(buf) b_hash; /* Hash chain. */
101 TAILQ_ENTRY(buf) b_vnbufs; /* Buffer's associated vnode. */
102 TAILQ_ENTRY(buf) b_freelist; /* Free list position if not active. */
103 TAILQ_ENTRY(buf) b_act; /* Device driver queue when active. *new* */
104 long b_flags; /* B_* flags. */
105 unsigned short b_qindex; /* buffer queue index */
106 unsigned char b_xflags; /* extra flags */
107 struct lock b_lock; /* Buffer lock */
108 int b_error; /* Errno value. */
109 long b_bufsize; /* Allocated buffer size. */
110 long b_runningbufspace; /* when I/O is running, pipelining */
111 long b_bcount; /* Valid bytes in buffer. */
112 long b_resid; /* Remaining I/O. */
113 dev_t b_dev; /* Device associated with buffer. */
114 caddr_t b_data; /* Memory, superblocks, indirect etc. */
115 caddr_t b_kvabase; /* base kva for buffer */
116 int b_kvasize; /* size of kva for buffer */
117 daddr_t b_lblkno; /* Logical block number. */
118 daddr_t b_blkno; /* Underlying physical block number. */
119 off_t b_offset; /* Offset into file */
120 /* Function to call upon completion. */
121 void (*b_iodone) __P((struct buf *));
122 /* For nested b_iodone's. */
123 struct iodone_chain *b_iodone_chain;
124 struct vnode *b_vp; /* Device vnode. */
125 int b_dirtyoff; /* Offset in buffer of dirty region. */
126 int b_dirtyend; /* Offset of end of dirty region. */
127 struct ucred *b_rcred; /* Read credentials reference. */
128 struct ucred *b_wcred; /* Write credentials reference. */
129 daddr_t b_pblkno; /* physical block number */
130 void *b_saveaddr; /* Original b_addr for physio. */
131 void *b_driver1; /* for private use by the driver */
132 void *b_driver2; /* for private use by the driver */
133 void *b_caller1; /* for private use by the caller */
134 void *b_caller2; /* for private use by the caller */
140 TAILQ_HEAD(cluster_list_head, buf) cluster_head;
141 TAILQ_ENTRY(buf) cluster_entry;
143 struct vm_page *b_pages[btoc(MAXPHYS)];
145 struct workhead b_dep; /* List of filesystem dependencies. */
146 struct chain_info { /* buffer chaining */
152 #define b_spc b_pager.pg_spc
155 * These flags are kept in b_flags.
159 * B_ASYNC VOP calls on bp's are usually async whether or not
160 * B_ASYNC is set, but some subsystems, such as NFS, like
161 * to know what is best for the caller so they can
164 * B_PAGING Indicates that bp is being used by the paging system or
165 * some paging system and that the bp is not linked into
166 * the b_vp's clean/dirty linked lists or ref counts.
167 * Buffer vp reassignments are illegal in this case.
169 * B_CACHE This may only be set if the buffer is entirely valid.
170 * The situation where B_DELWRI is set and B_CACHE is
171 * clear MUST be committed to disk by getblk() so
172 * B_DELWRI can also be cleared. See the comments for
173 * getblk() in kern/vfs_bio.c. If B_CACHE is clear,
174 * the caller is expected to clear B_ERROR|B_INVAL,
175 * set B_READ, and initiate an I/O.
177 * The 'entire buffer' is defined to be the range from
178 * 0 through b_bcount.
180 * B_MALLOC Request that the buffer be allocated from the malloc
181 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
183 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers
184 * by filesystems that allow clustering when the buffer
185 * is fully dirty and indicates that it may be clustered
186 * with other adjacent dirty buffers. Note the clustering
187 * may not be used with the stage 1 data write under NFS
188 * but may be used for the commit rpc portion.
190 * B_VMIO Indicates that the buffer is tied into an VM object.
191 * The buffer's data is always PAGE_SIZE aligned even
192 * if b_bufsize and b_bcount are not. ( b_bufsize is
193 * always at least DEV_BSIZE aligned, though ).
195 * B_DIRECT Hint that we should attempt to completely free
196 * the pages underlying the buffer. B_DIRECT is
197 * sticky until the buffer is released and typically
198 * only has an effect when B_RELBUF is also set.
200 * B_NOWDRAIN This flag should be set when a device (like VN)
201 * does a turn-around VOP_WRITE from its strategy
202 * routine. This flag prevents bwrite() from blocking
203 * in wdrain, avoiding a deadlock situation.
206 #define B_AGE 0x00000001 /* Move to age queue when I/O done. */
207 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */
208 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */
209 #define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */
210 #define B_DEFERRED 0x00000010 /* Skipped over for cleaning */
211 #define B_CACHE 0x00000020 /* Bread found us in the cache. */
212 #define B_CALL 0x00000040 /* Call b_iodone from biodone. */
213 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */
214 #define B_FREEBUF 0x00000100 /* Instruct driver: free blocks */
215 #define B_DONE 0x00000200 /* I/O completed. */
216 #define B_EINTR 0x00000400 /* I/O was interrupted */
217 #define B_ERROR 0x00000800 /* I/O error occurred. */
218 #define B_SCANNED 0x00001000 /* VOP_FSYNC funcs mark written bufs */
219 #define B_INVAL 0x00002000 /* Does not contain valid info. */
220 #define B_LOCKED 0x00004000 /* Locked in core (not reusable). */
221 #define B_NOCACHE 0x00008000 /* Do not cache block after use. */
222 #define B_MALLOC 0x00010000 /* malloced b_data */
223 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */
224 #define B_PHYS 0x00040000 /* I/O to user memory. */
225 #define B_RAW 0x00080000 /* Set by physio for raw transfers. */
226 #define B_READ 0x00100000 /* Read buffer. */
227 #define B_DIRTY 0x00200000 /* Needs writing later. */
228 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */
229 #define B_WANT 0x00800000 /* Used by vm_pager.c */
230 #define B_WRITE 0x00000000 /* Write buffer (pseudo flag). */
231 #define B_WRITEINPROG 0x01000000 /* Write in progress. */
232 #define B_XXX 0x02000000 /* Debugging flag. */
233 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */
234 #define B_ORDERED 0x08000000 /* Must guarantee I/O ordering */
235 #define B_RAM 0x10000000 /* Read ahead mark (flag) */
236 #define B_VMIO 0x20000000 /* VMIO flag */
237 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */
238 #define B_NOWDRAIN 0x80000000 /* Avoid wdrain deadlock */
240 #define PRINT_BUF_FLAGS "\20\40nowdrain\37cluster\36vmio\35ram\34ordered" \
241 "\33paging\32xxx\31writeinprog\30want\27relbuf\26dirty" \
242 "\25read\24raw\23phys\22clusterok\21malloc\20nocache" \
243 "\17locked\16inval\15scanned\14error\13eintr\12done\11freebuf" \
244 "\10delwri\7call\6cache\4direct\3async\2needcommit\1age"
247 * These flags are kept in b_xflags.
249 #define BX_VNDIRTY 0x00000001 /* On vnode dirty list */
250 #define BX_VNCLEAN 0x00000002 /* On vnode clean list */
251 #define BX_BKGRDWRITE 0x00000004 /* Do writes in background */
252 #define BX_BKGRDINPROG 0x00000008 /* Background write in progress */
253 #define BX_BKGRDWAIT 0x00000010 /* Background write waiting */
254 #define BX_AUTOCHAINDONE 0x00000020 /* pager I/O chain auto mode */
256 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */
260 * Buffer locking. See sys/buf2.h for inline functions.
262 struct simplelock buftimelock; /* Interlock on setting prio and timo */
263 extern char *buf_wmesg; /* Default buffer lock message */
264 #define BUF_WMESG "bufwait"
268 struct buf_queue_head {
269 TAILQ_HEAD(buf_queue, buf) queue;
271 struct buf *insert_point;
272 struct buf *switch_point;
276 * This structure describes a clustered I/O. It is stored in the b_saveaddr
277 * field of the buffer on which I/O is done. At I/O completion, cluster
278 * callback uses the structure to parcel I/O's to individual buffers, and
279 * then free's this structure.
281 struct cluster_save {
282 long bs_bcount; /* Saved b_bcount. */
283 long bs_bufsize; /* Saved b_bufsize. */
284 void *bs_saveaddr; /* Saved b_addr. */
285 int bs_nchildren; /* Number of associated buffers. */
286 struct buf **bs_children; /* List of associated buffers. */
290 * Definitions for the buffer free lists.
292 #define BUFFER_QUEUES 6 /* number of free buffer queues */
294 #define QUEUE_NONE 0 /* on no queue */
295 #define QUEUE_LOCKED 1 /* locked buffers */
296 #define QUEUE_CLEAN 2 /* non-B_DELWRI buffers */
297 #define QUEUE_DIRTY 3 /* B_DELWRI buffers */
298 #define QUEUE_EMPTYKVA 4 /* empty buffer headers w/KVA assignment */
299 #define QUEUE_EMPTY 5 /* empty buffer headers */
302 * Zero out the buffer's data area.
304 #define clrbuf(bp) { \
305 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \
310 * Flags to low-level bitmap allocation routines (balloc).
312 * Note: sequential_heuristic() in kern/vfs_vnops.c limits the count
315 #define B_SEQMASK 0x7F000000 /* Sequential heuristic mask. */
316 #define B_SEQSHIFT 24 /* Sequential heuristic shift. */
317 #define B_SEQMAX 0x7F
318 #define B_CLRBUF 0x01 /* Cleared invalid areas of buffer. */
319 #define B_SYNC 0x02 /* Do all allocations synchronously. */
322 extern int nbuf; /* The number of buffer headers */
323 extern int maxswzone; /* Max KVA for swap structures */
324 extern int maxbcache; /* Max KVA for buffer cache */
325 extern int runningbufspace;
326 extern int buf_maxio; /* nominal maximum I/O for buffer */
327 extern struct buf *buf; /* The buffer headers. */
328 extern char *buffers; /* The buffer contents. */
329 extern int bufpages; /* Number of memory pages in the buffer pool. */
330 extern struct buf *swbuf; /* Swap I/O buffer headers. */
331 extern int nswbuf; /* Number of swap I/O buffer headers. */
332 extern TAILQ_HEAD(swqueue, buf) bswlist;
333 extern TAILQ_HEAD(bqueues, buf) bufqueues[BUFFER_QUEUES];
337 caddr_t bufhashinit __P((caddr_t));
338 void bufinit __P((void));
339 void bwillwrite __P((void));
340 int buf_dirty_count_severe __P((void));
341 void bremfree __P((struct buf *));
342 int bread __P((struct vnode *, daddr_t, int,
343 struct ucred *, struct buf **));
344 int breadn __P((struct vnode *, daddr_t, int, daddr_t *, int *, int,
345 struct ucred *, struct buf **));
346 int bwrite __P((struct buf *));
347 void bdwrite __P((struct buf *));
348 void bawrite __P((struct buf *));
349 void bdirty __P((struct buf *));
350 void bundirty __P((struct buf *));
351 int bowrite __P((struct buf *));
352 void brelse __P((struct buf *));
353 void bqrelse __P((struct buf *));
354 int vfs_bio_awrite __P((struct buf *));
355 struct buf * getpbuf __P((int *));
356 struct buf *incore __P((struct vnode *, daddr_t));
357 struct buf *gbincore __P((struct vnode *, daddr_t));
358 int inmem __P((struct vnode *, daddr_t));
359 struct buf *getblk __P((struct vnode *, daddr_t, int, int, int));
360 struct buf *geteblk __P((int));
361 int biowait __P((struct buf *));
362 void biodone __P((struct buf *));
364 void cluster_callback __P((struct buf *));
365 int cluster_read __P((struct vnode *, u_quad_t, daddr_t, long,
366 struct ucred *, long, int, struct buf **));
367 int cluster_wbuild __P((struct vnode *, long, daddr_t, int));
368 void cluster_write __P((struct buf *, u_quad_t, int));
369 int physio __P((dev_t dev, struct uio *uio, int ioflag));
370 #define physread physio
371 #define physwrite physio
372 void vfs_bio_set_validclean __P((struct buf *, int base, int size));
373 void vfs_bio_clrbuf __P((struct buf *));
374 void vfs_busy_pages __P((struct buf *, int clear_modify));
375 void vfs_unbusy_pages __P((struct buf *));
376 void vwakeup __P((struct buf *));
377 int vmapbuf __P((struct buf *));
378 void vunmapbuf __P((struct buf *));
379 void relpbuf __P((struct buf *, int *));
380 void brelvp __P((struct buf *));
381 void bgetvp __P((struct vnode *, struct buf *));
382 void pbgetvp __P((struct vnode *, struct buf *));
383 void pbrelvp __P((struct buf *));
384 int allocbuf __P((struct buf *bp, int size));
385 void reassignbuf __P((struct buf *, struct vnode *));
386 void pbreassignbuf __P((struct buf *, struct vnode *));
387 struct buf *trypbuf __P((int *));
391 #endif /* !_SYS_BUF_H_ */