Add missing prototype (fixes warning).
[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 $
1de703da 40 * $DragonFly: src/sys/sys/buf.h,v 1.2 2003/06/17 04:28:58 dillon Exp $
984263bc
MD
41 */
42
43#ifndef _SYS_BUF_H_
44#define _SYS_BUF_H_
45
46#include <sys/queue.h>
47#include <sys/lock.h>
48
49struct buf;
50struct mount;
51struct vnode;
52
53/*
54 * To avoid including <ufs/ffs/softdep.h>
55 */
56LIST_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 */
63extern 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));
71} bioops;
72
73struct iodone_chain {
74 long ic_prev_flags;
75 void (*ic_prev_iodone) __P((struct buf *));
76 void *ic_prev_iodone_chain;
77 struct {
78 long ia_long;
79 void *ia_ptr;
80 } ic_args[5];
81};
82
83/*
84 * The buffer header describes an I/O operation in the kernel.
85 *
86 * NOTES:
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.
91 *
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 ).
95 *
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.
98 */
99struct buf {
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 */
135 union pager_info {
136 void *pg_spc;
137 int pg_reqpage;
138 } b_pager;
139 union cluster_info {
140 TAILQ_HEAD(cluster_list_head, buf) cluster_head;
141 TAILQ_ENTRY(buf) cluster_entry;
142 } b_cluster;
143 struct vm_page *b_pages[btoc(MAXPHYS)];
144 int b_npages;
145 struct workhead b_dep; /* List of filesystem dependencies. */
146 struct chain_info { /* buffer chaining */
147 struct buf *parent;
148 int count;
149 } b_chain;
150};
151
152#define b_spc b_pager.pg_spc
153
154/*
155 * These flags are kept in b_flags.
156 *
157 * Notes:
158 *
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
162 * optimize the I/O.
163 *
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.
168 *
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.
176 *
177 * The 'entire buffer' is defined to be the range from
178 * 0 through b_bcount.
179 *
180 * B_MALLOC Request that the buffer be allocated from the malloc
181 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
182 *
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.
189 *
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 ).
194 *
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.
199 *
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.
204 */
205
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 */
239
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"
245
246/*
247 * These flags are kept in b_xflags.
248 */
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 */
255
256#define NOOFFSET (-1LL) /* No buffer offset calculated yet */
257
258#ifdef _KERNEL
259/*
260 * Buffer locking
261 */
262struct simplelock buftimelock; /* Interlock on setting prio and timo */
263extern char *buf_wmesg; /* Default buffer lock message */
264#define BUF_WMESG "bufwait"
265#include <sys/proc.h> /* XXX for curproc */
266/*
267 * Initialize a lock.
268 */
269#define BUF_LOCKINIT(bp) \
270 lockinit(&(bp)->b_lock, PRIBIO + 4, buf_wmesg, 0, 0)
271/*
272 *
273 * Get a lock sleeping non-interruptably until it becomes available.
274 */
275static __inline int BUF_LOCK __P((struct buf *, int));
276static __inline int
277BUF_LOCK(struct buf *bp, int locktype)
278{
279 int s, ret;
280
281 s = splbio();
282 simple_lock(&buftimelock);
283 locktype |= LK_INTERLOCK;
284 bp->b_lock.lk_wmesg = buf_wmesg;
285 bp->b_lock.lk_prio = PRIBIO + 4;
286 /* bp->b_lock.lk_timo = 0; not necessary */
287 ret = lockmgr(&(bp)->b_lock, locktype, &buftimelock, curproc);
288 splx(s);
289 return ret;
290}
291/*
292 * Get a lock sleeping with specified interruptably and timeout.
293 */
294static __inline int BUF_TIMELOCK __P((struct buf *, int, char *, int, int));
295static __inline int
296BUF_TIMELOCK(struct buf *bp, int locktype, char *wmesg, int catch, int timo)
297{
298 int s, ret;
299
300 s = splbio();
301 simple_lock(&buftimelock);
302 locktype |= LK_INTERLOCK | LK_TIMELOCK;
303 bp->b_lock.lk_wmesg = wmesg;
304 bp->b_lock.lk_prio = (PRIBIO + 4) | catch;
305 bp->b_lock.lk_timo = timo;
306 ret = lockmgr(&(bp)->b_lock, (locktype), &buftimelock, curproc);
307 splx(s);
308 return ret;
309}
310/*
311 * Release a lock. Only the acquiring process may free the lock unless
312 * it has been handed off to biodone.
313 */
314static __inline void BUF_UNLOCK __P((struct buf *));
315static __inline void
316BUF_UNLOCK(struct buf *bp)
317{
318 int s;
319
320 s = splbio();
321 lockmgr(&(bp)->b_lock, LK_RELEASE, NULL, curproc);
322 splx(s);
323}
324
325/*
326 * Free a buffer lock.
327 */
328#define BUF_LOCKFREE(bp) \
329 if (BUF_REFCNT(bp) > 0) \
330 panic("free locked buf")
331/*
332 * When initiating asynchronous I/O, change ownership of the lock to the
333 * kernel. Once done, the lock may legally released by biodone. The
334 * original owning process can no longer acquire it recursively, but must
335 * wait until the I/O is completed and the lock has been freed by biodone.
336 */
337static __inline void BUF_KERNPROC __P((struct buf *));
338static __inline void
339BUF_KERNPROC(struct buf *bp)
340{
341 struct proc *p = curproc;
342
343 if (p != NULL && bp->b_lock.lk_lockholder == p->p_pid)
344 p->p_locks--;
345 bp->b_lock.lk_lockholder = LK_KERNPROC;
346}
347/*
348 * Find out the number of references to a lock.
349 */
350static __inline int BUF_REFCNT __P((struct buf *));
351static __inline int
352BUF_REFCNT(struct buf *bp)
353{
354 int s, ret;
355
356 s = splbio();
357 ret = lockcount(&(bp)->b_lock);
358 splx(s);
359 return ret;
360}
361
362#endif /* _KERNEL */
363
364struct buf_queue_head {
365 TAILQ_HEAD(buf_queue, buf) queue;
366 daddr_t last_pblkno;
367 struct buf *insert_point;
368 struct buf *switch_point;
369};
370
371/*
372 * This structure describes a clustered I/O. It is stored in the b_saveaddr
373 * field of the buffer on which I/O is done. At I/O completion, cluster
374 * callback uses the structure to parcel I/O's to individual buffers, and
375 * then free's this structure.
376 */
377struct cluster_save {
378 long bs_bcount; /* Saved b_bcount. */
379 long bs_bufsize; /* Saved b_bufsize. */
380 void *bs_saveaddr; /* Saved b_addr. */
381 int bs_nchildren; /* Number of associated buffers. */
382 struct buf **bs_children; /* List of associated buffers. */
383};
384
385#ifdef _KERNEL
386static __inline void bufq_init __P((struct buf_queue_head *head));
387
388static __inline void bufq_insert_tail __P((struct buf_queue_head *head,
389 struct buf *bp));
390
391static __inline void bufq_remove __P((struct buf_queue_head *head,
392 struct buf *bp));
393
394static __inline struct buf *bufq_first __P((struct buf_queue_head *head));
395
396static __inline void
397bufq_init(struct buf_queue_head *head)
398{
399 TAILQ_INIT(&head->queue);
400 head->last_pblkno = 0;
401 head->insert_point = NULL;
402 head->switch_point = NULL;
403}
404
405static __inline void
406bufq_insert_tail(struct buf_queue_head *head, struct buf *bp)
407{
408 if ((bp->b_flags & B_ORDERED) != 0) {
409 head->insert_point = bp;
410 head->switch_point = NULL;
411 }
412 TAILQ_INSERT_TAIL(&head->queue, bp, b_act);
413}
414
415static __inline void
416bufq_remove(struct buf_queue_head *head, struct buf *bp)
417{
418 if (bp == head->switch_point)
419 head->switch_point = TAILQ_NEXT(bp, b_act);
420 if (bp == head->insert_point) {
421 head->insert_point = TAILQ_PREV(bp, buf_queue, b_act);
422 if (head->insert_point == NULL)
423 head->last_pblkno = 0;
424 } else if (bp == TAILQ_FIRST(&head->queue))
425 head->last_pblkno = bp->b_pblkno;
426 TAILQ_REMOVE(&head->queue, bp, b_act);
427 if (TAILQ_FIRST(&head->queue) == head->switch_point)
428 head->switch_point = NULL;
429}
430
431static __inline struct buf *
432bufq_first(struct buf_queue_head *head)
433{
434 return (TAILQ_FIRST(&head->queue));
435}
436
437#endif /* _KERNEL */
438
439/*
440 * Definitions for the buffer free lists.
441 */
442#define BUFFER_QUEUES 6 /* number of free buffer queues */
443
444#define QUEUE_NONE 0 /* on no queue */
445#define QUEUE_LOCKED 1 /* locked buffers */
446#define QUEUE_CLEAN 2 /* non-B_DELWRI buffers */
447#define QUEUE_DIRTY 3 /* B_DELWRI buffers */
448#define QUEUE_EMPTYKVA 4 /* empty buffer headers w/KVA assignment */
449#define QUEUE_EMPTY 5 /* empty buffer headers */
450
451/*
452 * Zero out the buffer's data area.
453 */
454#define clrbuf(bp) { \
455 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \
456 (bp)->b_resid = 0; \
457}
458
459/*
460 * Flags to low-level bitmap allocation routines (balloc).
461 *
462 * Note: sequential_heuristic() in kern/vfs_vnops.c limits the count
463 * to 127.
464 */
465#define B_SEQMASK 0x7F000000 /* Sequential heuristic mask. */
466#define B_SEQSHIFT 24 /* Sequential heuristic shift. */
467#define B_SEQMAX 0x7F
468#define B_CLRBUF 0x01 /* Cleared invalid areas of buffer. */
469#define B_SYNC 0x02 /* Do all allocations synchronously. */
470
471#ifdef _KERNEL
472extern int nbuf; /* The number of buffer headers */
473extern int maxswzone; /* Max KVA for swap structures */
474extern int maxbcache; /* Max KVA for buffer cache */
475extern int runningbufspace;
476extern int buf_maxio; /* nominal maximum I/O for buffer */
477extern struct buf *buf; /* The buffer headers. */
478extern char *buffers; /* The buffer contents. */
479extern int bufpages; /* Number of memory pages in the buffer pool. */
480extern struct buf *swbuf; /* Swap I/O buffer headers. */
481extern int nswbuf; /* Number of swap I/O buffer headers. */
482extern TAILQ_HEAD(swqueue, buf) bswlist;
483extern TAILQ_HEAD(bqueues, buf) bufqueues[BUFFER_QUEUES];
484
485struct uio;
486
487caddr_t bufhashinit __P((caddr_t));
488void bufinit __P((void));
489void bwillwrite __P((void));
490int buf_dirty_count_severe __P((void));
491void bremfree __P((struct buf *));
492int bread __P((struct vnode *, daddr_t, int,
493 struct ucred *, struct buf **));
494int breadn __P((struct vnode *, daddr_t, int, daddr_t *, int *, int,
495 struct ucred *, struct buf **));
496int bwrite __P((struct buf *));
497void bdwrite __P((struct buf *));
498void bawrite __P((struct buf *));
499void bdirty __P((struct buf *));
500void bundirty __P((struct buf *));
501int bowrite __P((struct buf *));
502void brelse __P((struct buf *));
503void bqrelse __P((struct buf *));
504int vfs_bio_awrite __P((struct buf *));
505struct buf * getpbuf __P((int *));
506struct buf *incore __P((struct vnode *, daddr_t));
507struct buf *gbincore __P((struct vnode *, daddr_t));
508int inmem __P((struct vnode *, daddr_t));
509struct buf *getblk __P((struct vnode *, daddr_t, int, int, int));
510struct buf *geteblk __P((int));
511int biowait __P((struct buf *));
512void biodone __P((struct buf *));
513
514void cluster_callback __P((struct buf *));
515int cluster_read __P((struct vnode *, u_quad_t, daddr_t, long,
516 struct ucred *, long, int, struct buf **));
517int cluster_wbuild __P((struct vnode *, long, daddr_t, int));
518void cluster_write __P((struct buf *, u_quad_t, int));
519int physio __P((dev_t dev, struct uio *uio, int ioflag));
520#define physread physio
521#define physwrite physio
522void vfs_bio_set_validclean __P((struct buf *, int base, int size));
523void vfs_bio_clrbuf __P((struct buf *));
524void vfs_busy_pages __P((struct buf *, int clear_modify));
525void vfs_unbusy_pages __P((struct buf *));
526void vwakeup __P((struct buf *));
527int vmapbuf __P((struct buf *));
528void vunmapbuf __P((struct buf *));
529void relpbuf __P((struct buf *, int *));
530void brelvp __P((struct buf *));
531void bgetvp __P((struct vnode *, struct buf *));
532void pbgetvp __P((struct vnode *, struct buf *));
533void pbrelvp __P((struct buf *));
534int allocbuf __P((struct buf *bp, int size));
535void reassignbuf __P((struct buf *, struct vnode *));
536void pbreassignbuf __P((struct buf *, struct vnode *));
537struct buf *trypbuf __P((int *));
538
539#endif /* _KERNEL */
540
541#endif /* !_SYS_BUF_H_ */