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>
47 #include <sys/lockmgr.h>
55 * To avoid including <ufs/ffs/softdep.h>
57 LIST_HEAD(workhead, worklist);
59 * These are currently used only by the soft dependency code, hence
60 * are stored once in a global variable. If other subsystems wanted
61 * to use these hooks, a pointer to a set of bio_ops could be added
64 extern struct bio_ops {
65 void (*io_start) __P((struct buf *));
66 void (*io_complete) __P((struct buf *));
67 void (*io_deallocate) __P((struct buf *));
68 void (*io_movedeps) __P((struct buf *, struct buf *));
69 int (*io_countdeps) __P((struct buf *, int));
74 int (*bop_write) __P((struct buf *));
77 extern struct buf_ops buf_ops_bio;
80 * The buffer header describes an I/O operation in the kernel.
83 * b_bufsize, b_bcount. b_bufsize is the allocation size of the
84 * buffer, either DEV_BSIZE or PAGE_SIZE aligned. b_bcount is the
85 * originally requested buffer size and can serve as a bounds check
86 * against EOF. For most, but not all uses, b_bcount == b_bufsize.
88 * b_dirtyoff, b_dirtyend. Buffers support piecemeal, unaligned
89 * ranges of dirty data that need to be written to backing store.
90 * The range is typically clipped at b_bcount ( not b_bufsize ).
92 * b_resid. Number of bytes remaining in I/O. After an I/O operation
93 * completes, b_resid is usually 0 indicating 100% success.
96 /* XXX: b_io must be the first element of struct buf for now /phk */
97 struct bio b_io; /* "Builtin" I/O request. */
98 #define b_bcount b_io.bio_bcount
99 #define b_blkno b_io.bio_blkno
100 #define b_caller1 b_io.bio_caller1
101 #define b_data b_io.bio_data
102 #define b_dev b_io.bio_dev
103 #define b_driver1 b_io.bio_driver1
104 #define b_driver2 b_io.bio_driver2
105 #define b_error b_io.bio_error
106 #define b_iocmd b_io.bio_cmd
107 #define b_ioflags b_io.bio_flags
108 #define b_pblkno b_io.bio_pblkno
109 #define b_resid b_io.bio_resid
110 struct buf_ops *b_op;
112 #define B_MAGIC_BIO 0x10b10b10
113 #define B_MAGIC_NFS 0x67238234
114 void (*b_iodone) __P((struct buf *));
115 off_t b_offset; /* Offset into file. */
116 LIST_ENTRY(buf) b_hash; /* Hash chain. */
117 TAILQ_ENTRY(buf) b_vnbufs; /* Buffer's associated vnode. */
118 TAILQ_ENTRY(buf) b_freelist; /* Free list position if not active. */
119 TAILQ_ENTRY(buf) b_act; /* Device driver queue when active. *new* */
120 long b_flags; /* B_* flags. */
121 unsigned short b_qindex; /* buffer queue index */
122 unsigned char b_xflags; /* extra flags */
123 struct lock b_lock; /* Buffer lock */
124 long b_bufsize; /* Allocated buffer size. */
125 long b_runningbufspace; /* when I/O is running, pipelining */
126 caddr_t b_kvabase; /* base kva for buffer */
127 int b_kvasize; /* size of kva for buffer */
128 daddr_t b_lblkno; /* Logical block number. */
129 struct vnode *b_vp; /* Device vnode. */
130 int b_dirtyoff; /* Offset in buffer of dirty region. */
131 int b_dirtyend; /* Offset of end of dirty region. */
132 struct ucred *b_rcred; /* Read credentials reference. */
133 struct ucred *b_wcred; /* Write credentials reference. */
134 void *b_saveaddr; /* Original b_addr for physio. */
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. */
148 #define b_spc b_pager.pg_spc
151 * These flags are kept in b_flags.
155 * B_ASYNC VOP calls on bp's are usually async whether or not
156 * B_ASYNC is set, but some subsystems, such as NFS, like
157 * to know what is best for the caller so they can
160 * B_PAGING Indicates that bp is being used by the paging system or
161 * some paging system and that the bp is not linked into
162 * the b_vp's clean/dirty linked lists or ref counts.
163 * Buffer vp reassignments are illegal in this case.
165 * B_CACHE This may only be set if the buffer is entirely valid.
166 * The situation where B_DELWRI is set and B_CACHE is
167 * clear MUST be committed to disk by getblk() so
168 * B_DELWRI can also be cleared. See the comments for
169 * getblk() in kern/vfs_bio.c. If B_CACHE is clear,
170 * the caller is expected to clear BIO_ERROR and B_INVAL,
171 * set BIO_READ, and initiate an I/O.
173 * The 'entire buffer' is defined to be the range from
174 * 0 through b_bcount.
176 * B_MALLOC Request that the buffer be allocated from the malloc
177 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
179 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers
180 * by filesystems that allow clustering when the buffer
181 * is fully dirty and indicates that it may be clustered
182 * with other adjacent dirty buffers. Note the clustering
183 * may not be used with the stage 1 data write under NFS
184 * but may be used for the commit rpc portion.
186 * B_VMIO Indicates that the buffer is tied into an VM object.
187 * The buffer's data is always PAGE_SIZE aligned even
188 * if b_bufsize and b_bcount are not. ( b_bufsize is
189 * always at least DEV_BSIZE aligned, though ).
193 #define B_AGE 0x00000001 /* Move to age queue when I/O done. */
194 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */
195 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */
196 #define B_UNUSED0 0x00000008 /* Old B_BAD */
197 #define B_DEFERRED 0x00000010 /* Skipped over for cleaning */
198 #define B_CACHE 0x00000020 /* Bread found us in the cache. */
199 #define B_VALIDSUSPWRT 0x00000040 /* Valid write during suspension. */
200 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */
201 #define B_DONE 0x00000200 /* I/O completed. */
202 #define B_EINTR 0x00000400 /* I/O was interrupted */
203 #define B_00000800 0x00000800 /* Available flag. */
204 #define B_SCANNED 0x00001000 /* VOP_FSYNC funcs mark written bufs */
205 #define B_INVAL 0x00002000 /* Does not contain valid info. */
206 #define B_LOCKED 0x00004000 /* Locked in core (not reusable). */
207 #define B_NOCACHE 0x00008000 /* Do not cache block after use. */
208 #define B_MALLOC 0x00010000 /* malloced b_data */
209 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */
210 #define B_PHYS 0x00040000 /* I/O to user memory. */
211 #define B_RAW 0x00080000 /* Set by physio for raw transfers. */
212 #define B_DIRTY 0x00200000 /* Needs writing later. */
213 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */
214 #define B_WANT 0x00800000 /* Used by vm_pager.c */
215 #define B_WRITEINPROG 0x01000000 /* Write in progress. */
216 #define B_XXX 0x02000000 /* Debugging flag. */
217 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */
218 #define B_08000000 0x08000000 /* Available flag. */
219 #define B_RAM 0x10000000 /* Read ahead mark (flag) */
220 #define B_VMIO 0x20000000 /* VMIO flag */
221 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */
222 #define B_80000000 0x80000000 /* Available flag. */
224 #define PRINT_BUF_FLAGS "\20\40autochain\37cluster\36vmio\35ram\34ordered" \
225 "\33paging\32xxx\31writeinprog\30want\27relbuf\26dirty" \
226 "\25read\24raw\23phys\22clusterok\21malloc\20nocache" \
227 "\17locked\16inval\15scanned\14error\13eintr\12done\11freebuf" \
228 "\10delwri\7call\6cache\4bad\3async\2needcommit\1age"
231 * These flags are kept in b_xflags.
233 #define BX_VNDIRTY 0x00000001 /* On vnode dirty list */
234 #define BX_VNCLEAN 0x00000002 /* On vnode clean list */
235 #define BX_BKGRDWRITE 0x00000004 /* Do writes in background */
236 #define BX_BKGRDINPROG 0x00000008 /* Background write in progress */
237 #define BX_BKGRDWAIT 0x00000010 /* Background write waiting */
239 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */
245 extern struct mtx buftimelock; /* Interlock on setting prio and timo */
246 extern char *buf_wmesg; /* Default buffer lock message */
247 #define BUF_WMESG "bufwait"
248 #include <sys/proc.h> /* XXX for curproc */
249 #include <sys/mutex.h>
254 #define BUF_LOCKINIT(bp) \
255 lockinit(&(bp)->b_lock, PRIBIO + 4, buf_wmesg, 0, 0)
258 * Get a lock sleeping non-interruptably until it becomes available.
260 static __inline int BUF_LOCK __P((struct buf *, int));
262 BUF_LOCK(struct buf *bp, int locktype)
267 mtx_lock(&buftimelock);
268 locktype |= LK_INTERLOCK;
269 bp->b_lock.lk_wmesg = buf_wmesg;
270 bp->b_lock.lk_prio = PRIBIO + 4;
271 bp->b_lock.lk_timo = 0;
272 ret = lockmgr(&(bp)->b_lock, locktype, &buftimelock, curproc);
277 * Get a lock sleeping with specified interruptably and timeout.
279 static __inline int BUF_TIMELOCK __P((struct buf *, int, char *, int, int));
281 BUF_TIMELOCK(struct buf *bp, int locktype, char *wmesg, int catch, int timo)
286 mtx_lock(&buftimelock);
287 locktype |= LK_INTERLOCK;
288 bp->b_lock.lk_wmesg = wmesg;
289 bp->b_lock.lk_prio = (PRIBIO + 4) | catch;
290 bp->b_lock.lk_timo = timo;
291 ret = lockmgr(&(bp)->b_lock, (locktype), &buftimelock, curproc);
296 * Release a lock. Only the acquiring process may free the lock unless
297 * it has been handed off to biodone.
299 static __inline void BUF_UNLOCK __P((struct buf *));
301 BUF_UNLOCK(struct buf *bp)
306 lockmgr(&(bp)->b_lock, LK_RELEASE, NULL, curproc);
311 * Free a buffer lock.
313 #define BUF_LOCKFREE(bp) \
315 if (BUF_REFCNT(bp) > 0) \
316 panic("free locked buf"); \
317 lockdestroy(&(bp)->b_lock); \
320 #ifdef _SYS_PROC_H_ /* Avoid #include <sys/proc.h> pollution */
322 * When initiating asynchronous I/O, change ownership of the lock to the
323 * kernel. Once done, the lock may legally released by biodone. The
324 * original owning process can no longer acquire it recursively, but must
325 * wait until the I/O is completed and the lock has been freed by biodone.
327 static __inline void BUF_KERNPROC __P((struct buf *));
329 BUF_KERNPROC(struct buf *bp)
331 struct proc *p = curproc;
333 if (p != PCPU_GET(idleproc) && bp->b_lock.lk_lockholder == p->p_pid)
335 bp->b_lock.lk_lockholder = LK_KERNPROC;
339 * Find out the number of references to a lock.
341 static __inline int BUF_REFCNT __P((struct buf *));
343 BUF_REFCNT(struct buf *bp)
348 ret = lockcount(&(bp)->b_lock);
355 struct buf_queue_head {
356 TAILQ_HEAD(buf_queue, buf) queue;
358 struct buf *insert_point;
359 struct buf *switch_point;
363 * This structure describes a clustered I/O. It is stored in the b_saveaddr
364 * field of the buffer on which I/O is done. At I/O completion, cluster
365 * callback uses the structure to parcel I/O's to individual buffers, and
366 * then free's this structure.
368 struct cluster_save {
369 long bs_bcount; /* Saved b_bcount. */
370 long bs_bufsize; /* Saved b_bufsize. */
371 void *bs_saveaddr; /* Saved b_addr. */
372 int bs_nchildren; /* Number of associated buffers. */
373 struct buf **bs_children; /* List of associated buffers. */
377 static __inline void bufq_init __P((struct buf_queue_head *head));
378 static __inline void bufq_insert_tail __P((struct buf_queue_head *head,
380 static __inline void bufq_remove __P((struct buf_queue_head *head,
382 static __inline struct buf *bufq_first __P((struct buf_queue_head *head));
385 bufq_init(struct buf_queue_head *head)
387 TAILQ_INIT(&head->queue);
388 head->last_pblkno = 0;
389 head->insert_point = NULL;
390 head->switch_point = NULL;
394 bufq_insert_tail(struct buf_queue_head *head, struct buf *bp)
396 if ((bp->b_ioflags & BIO_ORDERED) != 0) {
397 head->insert_point = bp;
398 head->switch_point = NULL;
400 TAILQ_INSERT_TAIL(&head->queue, bp, b_act);
404 bufq_remove(struct buf_queue_head *head, struct buf *bp)
406 if (bp == head->switch_point)
407 head->switch_point = TAILQ_NEXT(bp, b_act);
408 if (bp == head->insert_point) {
409 head->insert_point = TAILQ_PREV(bp, buf_queue, b_act);
410 if (head->insert_point == NULL)
411 head->last_pblkno = 0;
412 } else if (bp == TAILQ_FIRST(&head->queue))
413 head->last_pblkno = bp->b_pblkno;
414 TAILQ_REMOVE(&head->queue, bp, b_act);
415 if (TAILQ_FIRST(&head->queue) == head->switch_point)
416 head->switch_point = NULL;
419 static __inline struct buf *
420 bufq_first(struct buf_queue_head *head)
422 return (TAILQ_FIRST(&head->queue));
425 #define BUF_WRITE(bp) \
426 (bp)->b_op->bop_write(bp)
428 #define BUF_STRATEGY(bp) VOP_STRATEGY((bp)->b_vp, (bp))
431 buf_start(struct buf *bp)
434 (*bioops.io_start)(bp);
438 buf_complete(struct buf *bp)
440 if (bioops.io_complete)
441 (*bioops.io_complete)(bp);
445 buf_deallocate(struct buf *bp)
447 if (bioops.io_deallocate)
448 (*bioops.io_deallocate)(bp);
453 buf_movedeps(struct buf *bp, struct buf *bp2)
455 if (bioops.io_movedeps)
456 (*bioops.io_movedeps)(bp, bp2);
460 buf_countdeps(struct buf *bp, int i)
462 if (bioops.io_countdeps)
463 return ((*bioops.io_countdeps)(bp, i));
471 * Definitions for the buffer free lists.
473 #define BUFFER_QUEUES 6 /* number of free buffer queues */
475 #define QUEUE_NONE 0 /* on no queue */
476 #define QUEUE_LOCKED 1 /* locked buffers */
477 #define QUEUE_CLEAN 2 /* non-B_DELWRI buffers */
478 #define QUEUE_DIRTY 3 /* B_DELWRI buffers */
479 #define QUEUE_EMPTYKVA 4 /* empty buffer headers w/KVA assignment */
480 #define QUEUE_EMPTY 5 /* empty buffer headers */
483 * Zero out the buffer's data area.
485 #define clrbuf(bp) { \
486 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \
490 /* Flags to low-level allocation routines. */
491 #define B_CLRBUF 0x01 /* Request allocated buffer be cleared. */
492 #define B_SYNC 0x02 /* Do all allocations synchronously. */
493 #define B_METAONLY 0x04 /* Return indirect block buffer. */
494 #define B_NOWAIT 0x08 /* do not sleep to await lock */
497 extern int nbuf; /* The number of buffer headers */
498 extern int runningbufspace;
499 extern int buf_maxio; /* nominal maximum I/O for buffer */
500 extern struct buf *buf; /* The buffer headers. */
501 extern char *buffers; /* The buffer contents. */
502 extern int bufpages; /* Number of memory pages in the buffer pool. */
503 extern struct buf *swbuf; /* Swap I/O buffer headers. */
504 extern int nswbuf; /* Number of swap I/O buffer headers. */
505 extern TAILQ_HEAD(swqueue, buf) bswlist;
506 extern TAILQ_HEAD(bqueues, buf) bufqueues[BUFFER_QUEUES];
510 caddr_t bufhashinit __P((caddr_t));
511 void bufinit __P((void));
512 void bwillwrite __P((void));
513 int buf_dirty_count_severe __P((void));
514 void bremfree __P((struct buf *));
515 int bread __P((struct vnode *, daddr_t, int,
516 struct ucred *, struct buf **));
517 int breadn __P((struct vnode *, daddr_t, int, daddr_t *, int *, int,
518 struct ucred *, struct buf **));
519 int bwrite __P((struct buf *));
520 void bdwrite __P((struct buf *));
521 void bawrite __P((struct buf *));
522 void bdirty __P((struct buf *));
523 void bundirty __P((struct buf *));
524 int bowrite __P((struct buf *));
525 void brelse __P((struct buf *));
526 void bqrelse __P((struct buf *));
527 int vfs_bio_awrite __P((struct buf *));
528 struct buf * getpbuf __P((int *));
529 struct buf *incore __P((struct vnode *, daddr_t));
530 struct buf *gbincore __P((struct vnode *, daddr_t));
531 int inmem __P((struct vnode *, daddr_t));
532 struct buf *getblk __P((struct vnode *, daddr_t, int, int, int));
533 struct buf *geteblk __P((int));
534 int bufwait __P((struct buf *));
535 void bufdone __P((struct buf *));
536 void bufdonebio __P((struct bio *));
538 void cluster_callback __P((struct buf *));
539 int cluster_read __P((struct vnode *, u_quad_t, daddr_t, long,
540 struct ucred *, long, int, struct buf **));
541 int cluster_wbuild __P((struct vnode *, long, daddr_t, int));
542 void cluster_write __P((struct buf *, u_quad_t, int));
543 void vfs_bio_set_validclean __P((struct buf *, int base, int size));
544 void vfs_bio_clrbuf __P((struct buf *));
545 void vfs_busy_pages __P((struct buf *, int clear_modify));
546 void vfs_unbusy_pages __P((struct buf *));
547 void vwakeup __P((struct buf *));
548 void vmapbuf __P((struct buf *));
549 void vunmapbuf __P((struct buf *));
550 void relpbuf __P((struct buf *, int *));
551 void brelvp __P((struct buf *));
552 void bgetvp __P((struct vnode *, struct buf *));
553 void pbgetvp __P((struct vnode *, struct buf *));
554 void pbrelvp __P((struct buf *));
555 int allocbuf __P((struct buf *bp, int size));
556 void reassignbuf __P((struct buf *, struct vnode *));
557 void pbreassignbuf __P((struct buf *, struct vnode *));
558 struct buf *trypbuf __P((int *));
562 #endif /* !_SYS_BUF_H_ */