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
45 #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 void (*io_movedeps) __P((struct buf *, struct buf *));
68 int (*io_countdeps) __P((struct buf *, int));
72 * The buffer header describes an I/O operation in the kernel.
75 * b_bufsize, b_bcount. b_bufsize is the allocation size of the
76 * buffer, either DEV_BSIZE or PAGE_SIZE aligned. b_bcount is the
77 * originally requested buffer size and can serve as a bounds check
78 * against EOF. For most, but not all uses, b_bcount == b_bufsize.
80 * b_dirtyoff, b_dirtyend. Buffers support piecemeal, unaligned
81 * ranges of dirty data that need to be written to backing store.
82 * The range is typically clipped at b_bcount ( not b_bufsize ).
84 * b_resid. Number of bytes remaining in I/O. After an I/O operation
85 * completes, b_resid is usually 0 indicating 100% success.
88 /* XXX: b_io must be the first element of struct buf for now /phk */
89 struct bio b_io; /* "Builtin" I/O request. */
90 #define b_bcount b_io.bio_bcount
91 #define b_blkno b_io.bio_blkno
92 #define b_caller1 b_io.bio_caller1
93 #define b_data b_io.bio_data
94 #define b_dev b_io.bio_dev
95 #define b_driver1 b_io.bio_driver1
96 #define b_driver2 b_io.bio_driver2
97 #define b_error b_io.bio_error
98 #define b_iocmd b_io.bio_cmd
99 #define b_ioflags b_io.bio_flags
100 #define b_pblkno b_io.bio_pblkno
101 #define b_resid b_io.bio_resid
102 void (*b_iodone) __P((struct buf *));
103 off_t b_offset; /* Offset into file. */
104 LIST_ENTRY(buf) b_hash; /* Hash chain. */
105 TAILQ_ENTRY(buf) b_vnbufs; /* Buffer's associated vnode. */
106 TAILQ_ENTRY(buf) b_freelist; /* Free list position if not active. */
107 TAILQ_ENTRY(buf) b_act; /* Device driver queue when active. *new* */
108 long b_flags; /* B_* flags. */
109 unsigned short b_qindex; /* buffer queue index */
110 unsigned char b_xflags; /* extra flags */
111 struct lock b_lock; /* Buffer lock */
112 long b_bufsize; /* Allocated buffer size. */
113 long b_runningbufspace; /* when I/O is running, pipelining */
114 caddr_t b_kvabase; /* base kva for buffer */
115 int b_kvasize; /* size of kva for buffer */
116 daddr_t b_lblkno; /* Logical block number. */
117 struct vnode *b_vp; /* Device vnode. */
118 int b_dirtyoff; /* Offset in buffer of dirty region. */
119 int b_dirtyend; /* Offset of end of dirty region. */
120 struct ucred *b_rcred; /* Read credentials reference. */
121 struct ucred *b_wcred; /* Write credentials reference. */
122 void *b_saveaddr; /* Original b_addr for physio. */
128 TAILQ_HEAD(cluster_list_head, buf) cluster_head;
129 TAILQ_ENTRY(buf) cluster_entry;
131 struct vm_page *b_pages[btoc(MAXPHYS)];
133 struct workhead b_dep; /* List of filesystem dependencies. */
136 #define b_spc b_pager.pg_spc
139 * These flags are kept in b_flags.
143 * B_ASYNC VOP calls on bp's are usually async whether or not
144 * B_ASYNC is set, but some subsystems, such as NFS, like
145 * to know what is best for the caller so they can
148 * B_PAGING Indicates that bp is being used by the paging system or
149 * some paging system and that the bp is not linked into
150 * the b_vp's clean/dirty linked lists or ref counts.
151 * Buffer vp reassignments are illegal in this case.
153 * B_CACHE This may only be set if the buffer is entirely valid.
154 * The situation where B_DELWRI is set and B_CACHE is
155 * clear MUST be committed to disk by getblk() so
156 * B_DELWRI can also be cleared. See the comments for
157 * getblk() in kern/vfs_bio.c. If B_CACHE is clear,
158 * the caller is expected to clear BIO_ERROR and B_INVAL,
159 * set BIO_READ, and initiate an I/O.
161 * The 'entire buffer' is defined to be the range from
162 * 0 through b_bcount.
164 * B_MALLOC Request that the buffer be allocated from the malloc
165 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
167 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers
168 * by filesystems that allow clustering when the buffer
169 * is fully dirty and indicates that it may be clustered
170 * with other adjacent dirty buffers. Note the clustering
171 * may not be used with the stage 1 data write under NFS
172 * but may be used for the commit rpc portion.
174 * B_VMIO Indicates that the buffer is tied into an VM object.
175 * The buffer's data is always PAGE_SIZE aligned even
176 * if b_bufsize and b_bcount are not. ( b_bufsize is
177 * always at least DEV_BSIZE aligned, though ).
181 #define B_AGE 0x00000001 /* Move to age queue when I/O done. */
182 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */
183 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */
184 #define B_UNUSED0 0x00000008 /* Old B_BAD */
185 #define B_DEFERRED 0x00000010 /* Skipped over for cleaning */
186 #define B_CACHE 0x00000020 /* Bread found us in the cache. */
187 #define B_VALIDSUSPWRT 0x00000040 /* Valid write during suspension. */
188 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */
189 #define B_DONE 0x00000200 /* I/O completed. */
190 #define B_EINTR 0x00000400 /* I/O was interrupted */
191 #define B_00000800 0x00000800 /* Available flag. */
192 #define B_SCANNED 0x00001000 /* VOP_FSYNC funcs mark written bufs */
193 #define B_INVAL 0x00002000 /* Does not contain valid info. */
194 #define B_LOCKED 0x00004000 /* Locked in core (not reusable). */
195 #define B_NOCACHE 0x00008000 /* Do not cache block after use. */
196 #define B_MALLOC 0x00010000 /* malloced b_data */
197 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */
198 #define B_PHYS 0x00040000 /* I/O to user memory. */
199 #define B_RAW 0x00080000 /* Set by physio for raw transfers. */
200 #define B_DIRTY 0x00200000 /* Needs writing later. */
201 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */
202 #define B_WANT 0x00800000 /* Used by vm_pager.c */
203 #define B_WRITEINPROG 0x01000000 /* Write in progress. */
204 #define B_XXX 0x02000000 /* Debugging flag. */
205 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */
206 #define B_08000000 0x08000000 /* Available flag. */
207 #define B_RAM 0x10000000 /* Read ahead mark (flag) */
208 #define B_VMIO 0x20000000 /* VMIO flag */
209 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */
210 #define B_80000000 0x80000000 /* Available flag. */
212 #define PRINT_BUF_FLAGS "\20\40autochain\37cluster\36vmio\35ram\34ordered" \
213 "\33paging\32xxx\31writeinprog\30want\27relbuf\26dirty" \
214 "\25read\24raw\23phys\22clusterok\21malloc\20nocache" \
215 "\17locked\16inval\15scanned\14error\13eintr\12done\11freebuf" \
216 "\10delwri\7call\6cache\4bad\3async\2needcommit\1age"
219 * These flags are kept in b_xflags.
221 #define BX_VNDIRTY 0x00000001 /* On vnode dirty list */
222 #define BX_VNCLEAN 0x00000002 /* On vnode clean list */
223 #define BX_BKGRDWRITE 0x00000004 /* Do writes in background */
224 #define BX_BKGRDINPROG 0x00000008 /* Background write in progress */
225 #define BX_BKGRDWAIT 0x00000010 /* Background write waiting */
227 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */
233 extern struct mtx buftimelock; /* Interlock on setting prio and timo */
234 extern char *buf_wmesg; /* Default buffer lock message */
235 #define BUF_WMESG "bufwait"
236 #include <sys/proc.h> /* XXX for curproc */
237 #include <sys/mutex.h>
242 #define BUF_LOCKINIT(bp) \
243 lockinit(&(bp)->b_lock, PRIBIO + 4, buf_wmesg, 0, 0)
246 * Get a lock sleeping non-interruptably until it becomes available.
248 static __inline int BUF_LOCK __P((struct buf *, int));
250 BUF_LOCK(struct buf *bp, int locktype)
255 mtx_lock(&buftimelock);
256 locktype |= LK_INTERLOCK;
257 bp->b_lock.lk_wmesg = buf_wmesg;
258 bp->b_lock.lk_prio = PRIBIO + 4;
259 bp->b_lock.lk_timo = 0;
260 ret = lockmgr(&(bp)->b_lock, locktype, &buftimelock, curproc);
265 * Get a lock sleeping with specified interruptably and timeout.
267 static __inline int BUF_TIMELOCK __P((struct buf *, int, char *, int, int));
269 BUF_TIMELOCK(struct buf *bp, int locktype, char *wmesg, int catch, int timo)
274 mtx_lock(&buftimelock);
275 locktype |= LK_INTERLOCK;
276 bp->b_lock.lk_wmesg = wmesg;
277 bp->b_lock.lk_prio = (PRIBIO + 4) | catch;
278 bp->b_lock.lk_timo = timo;
279 ret = lockmgr(&(bp)->b_lock, (locktype), &buftimelock, curproc);
284 * Release a lock. Only the acquiring process may free the lock unless
285 * it has been handed off to biodone.
287 static __inline void BUF_UNLOCK __P((struct buf *));
289 BUF_UNLOCK(struct buf *bp)
294 lockmgr(&(bp)->b_lock, LK_RELEASE, NULL, curproc);
299 * Free a buffer lock.
301 #define BUF_LOCKFREE(bp) \
303 if (BUF_REFCNT(bp) > 0) \
304 panic("free locked buf"); \
305 lockdestroy(&(bp)->b_lock); \
308 #ifdef _SYS_PROC_H_ /* Avoid #include <sys/proc.h> pollution */
310 * When initiating asynchronous I/O, change ownership of the lock to the
311 * kernel. Once done, the lock may legally released by biodone. The
312 * original owning process can no longer acquire it recursively, but must
313 * wait until the I/O is completed and the lock has been freed by biodone.
315 static __inline void BUF_KERNPROC __P((struct buf *));
317 BUF_KERNPROC(struct buf *bp)
319 struct proc *p = curproc;
321 if (p != PCPU_GET(idleproc) && bp->b_lock.lk_lockholder == p->p_pid)
323 bp->b_lock.lk_lockholder = LK_KERNPROC;
327 * Find out the number of references to a lock.
329 static __inline int BUF_REFCNT __P((struct buf *));
331 BUF_REFCNT(struct buf *bp)
336 ret = lockcount(&(bp)->b_lock);
343 struct buf_queue_head {
344 TAILQ_HEAD(buf_queue, buf) queue;
346 struct buf *insert_point;
347 struct buf *switch_point;
351 * This structure describes a clustered I/O. It is stored in the b_saveaddr
352 * field of the buffer on which I/O is done. At I/O completion, cluster
353 * callback uses the structure to parcel I/O's to individual buffers, and
354 * then free's this structure.
356 struct cluster_save {
357 long bs_bcount; /* Saved b_bcount. */
358 long bs_bufsize; /* Saved b_bufsize. */
359 void *bs_saveaddr; /* Saved b_addr. */
360 int bs_nchildren; /* Number of associated buffers. */
361 struct buf **bs_children; /* List of associated buffers. */
365 static __inline void bufq_init __P((struct buf_queue_head *head));
366 static __inline void bufq_insert_tail __P((struct buf_queue_head *head,
368 static __inline void bufq_remove __P((struct buf_queue_head *head,
370 static __inline struct buf *bufq_first __P((struct buf_queue_head *head));
373 bufq_init(struct buf_queue_head *head)
375 TAILQ_INIT(&head->queue);
376 head->last_pblkno = 0;
377 head->insert_point = NULL;
378 head->switch_point = NULL;
382 bufq_insert_tail(struct buf_queue_head *head, struct buf *bp)
384 if ((bp->b_ioflags & BIO_ORDERED) != 0) {
385 head->insert_point = bp;
386 head->switch_point = NULL;
388 TAILQ_INSERT_TAIL(&head->queue, bp, b_act);
392 bufq_remove(struct buf_queue_head *head, struct buf *bp)
394 if (bp == head->switch_point)
395 head->switch_point = TAILQ_NEXT(bp, b_act);
396 if (bp == head->insert_point) {
397 head->insert_point = TAILQ_PREV(bp, buf_queue, b_act);
398 if (head->insert_point == NULL)
399 head->last_pblkno = 0;
400 } else if (bp == TAILQ_FIRST(&head->queue))
401 head->last_pblkno = bp->b_pblkno;
402 TAILQ_REMOVE(&head->queue, bp, b_act);
403 if (TAILQ_FIRST(&head->queue) == head->switch_point)
404 head->switch_point = NULL;
407 static __inline struct buf *
408 bufq_first(struct buf_queue_head *head)
410 return (TAILQ_FIRST(&head->queue));
413 #define BUF_WRITE(bp) VOP_BWRITE((bp)->b_vp, (bp))
414 #define BUF_STRATEGY(bp) VOP_STRATEGY((bp)->b_vp, (bp))
417 buf_start(struct buf *bp)
420 (*bioops.io_start)(bp);
424 buf_complete(struct buf *bp)
426 if (bioops.io_complete)
427 (*bioops.io_complete)(bp);
431 buf_deallocate(struct buf *bp)
433 if (bioops.io_deallocate)
434 (*bioops.io_deallocate)(bp);
439 buf_movedeps(struct buf *bp, struct buf *bp2)
441 if (bioops.io_movedeps)
442 (*bioops.io_movedeps)(bp, bp2);
446 buf_countdeps(struct buf *bp, int i)
448 if (bioops.io_countdeps)
449 return ((*bioops.io_countdeps)(bp, i));
457 * Definitions for the buffer free lists.
459 #define BUFFER_QUEUES 6 /* number of free buffer queues */
461 #define QUEUE_NONE 0 /* on no queue */
462 #define QUEUE_LOCKED 1 /* locked buffers */
463 #define QUEUE_CLEAN 2 /* non-B_DELWRI buffers */
464 #define QUEUE_DIRTY 3 /* B_DELWRI buffers */
465 #define QUEUE_EMPTYKVA 4 /* empty buffer headers w/KVA assignment */
466 #define QUEUE_EMPTY 5 /* empty buffer headers */
469 * Zero out the buffer's data area.
471 #define clrbuf(bp) { \
472 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \
476 /* Flags to low-level allocation routines. */
477 #define B_CLRBUF 0x01 /* Request allocated buffer be cleared. */
478 #define B_SYNC 0x02 /* Do all allocations synchronously. */
479 #define B_METAONLY 0x04 /* Return indirect block buffer. */
480 #define B_NOWAIT 0x08 /* do not sleep to await lock */
483 extern int nbuf; /* The number of buffer headers */
484 extern int runningbufspace;
485 extern int buf_maxio; /* nominal maximum I/O for buffer */
486 extern struct buf *buf; /* The buffer headers. */
487 extern char *buffers; /* The buffer contents. */
488 extern int bufpages; /* Number of memory pages in the buffer pool. */
489 extern struct buf *swbuf; /* Swap I/O buffer headers. */
490 extern int nswbuf; /* Number of swap I/O buffer headers. */
491 extern TAILQ_HEAD(swqueue, buf) bswlist;
492 extern TAILQ_HEAD(bqueues, buf) bufqueues[BUFFER_QUEUES];
496 caddr_t bufhashinit __P((caddr_t));
497 void bufinit __P((void));
498 void bwillwrite __P((void));
499 int buf_dirty_count_severe __P((void));
500 void bremfree __P((struct buf *));
501 int bread __P((struct vnode *, daddr_t, int,
502 struct ucred *, struct buf **));
503 int breadn __P((struct vnode *, daddr_t, int, daddr_t *, int *, int,
504 struct ucred *, struct buf **));
505 int bwrite __P((struct buf *));
506 void bdwrite __P((struct buf *));
507 void bawrite __P((struct buf *));
508 void bdirty __P((struct buf *));
509 void bundirty __P((struct buf *));
510 int bowrite __P((struct buf *));
511 void brelse __P((struct buf *));
512 void bqrelse __P((struct buf *));
513 int vfs_bio_awrite __P((struct buf *));
514 struct buf * getpbuf __P((int *));
515 struct buf *incore __P((struct vnode *, daddr_t));
516 struct buf *gbincore __P((struct vnode *, daddr_t));
517 int inmem __P((struct vnode *, daddr_t));
518 struct buf *getblk __P((struct vnode *, daddr_t, int, int, int));
519 struct buf *geteblk __P((int));
520 int bufwait __P((struct buf *));
521 void bufdone __P((struct buf *));
522 void bufdonebio __P((struct bio *));
524 void cluster_callback __P((struct buf *));
525 int cluster_read __P((struct vnode *, u_quad_t, daddr_t, long,
526 struct ucred *, long, int, struct buf **));
527 int cluster_wbuild __P((struct vnode *, long, daddr_t, int));
528 void cluster_write __P((struct buf *, u_quad_t, int));
529 void vfs_bio_set_validclean __P((struct buf *, int base, int size));
530 void vfs_bio_clrbuf __P((struct buf *));
531 void vfs_busy_pages __P((struct buf *, int clear_modify));
532 void vfs_unbusy_pages __P((struct buf *));
533 void vwakeup __P((struct buf *));
534 void vmapbuf __P((struct buf *));
535 void vunmapbuf __P((struct buf *));
536 void relpbuf __P((struct buf *, int *));
537 void brelvp __P((struct buf *));
538 void bgetvp __P((struct vnode *, struct buf *));
539 void pbgetvp __P((struct vnode *, struct buf *));
540 void pbrelvp __P((struct buf *));
541 int allocbuf __P((struct buf *bp, int size));
542 void reassignbuf __P((struct buf *, struct vnode *));
543 void pbreassignbuf __P((struct buf *, struct vnode *));
544 struct buf *trypbuf __P((int *));
548 #endif /* !_SYS_BUF_H_ */