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 * 4. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * @(#)buf.h 8.9 (Berkeley) 3/30/95
41 #include <sys/bufobj.h>
42 #include <sys/queue.h>
44 #include <sys/lockmgr.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)(struct buf *);
65 void (*io_complete)(struct buf *);
66 void (*io_deallocate)(struct buf *);
67 int (*io_countdeps)(struct buf *, int);
72 typedef unsigned char b_xflags_t;
75 * The buffer header describes an I/O operation in the kernel.
78 * b_bufsize, b_bcount. b_bufsize is the allocation size of the
79 * buffer, either DEV_BSIZE or PAGE_SIZE aligned. b_bcount is the
80 * originally requested buffer size and can serve as a bounds check
81 * against EOF. For most, but not all uses, b_bcount == b_bufsize.
83 * b_dirtyoff, b_dirtyend. Buffers support piecemeal, unaligned
84 * ranges of dirty data that need to be written to backing store.
85 * The range is typically clipped at b_bcount ( not b_bufsize ).
87 * b_resid. Number of bytes remaining in I/O. After an I/O operation
88 * completes, b_resid is usually 0 indicating 100% success.
90 * All fields are protected by the buffer lock except those marked:
91 * V - Protected by owning vnode lock
92 * Q - Protected by the buf queue lock
93 * D - Protected by an dependency implementation specific lock
96 struct bufobj *b_bufobj;
105 void (*b_iodone)(struct buf *);
106 daddr_t b_blkno; /* Underlying physical block number. */
107 off_t b_offset; /* Offset into file. */
108 TAILQ_ENTRY(buf) b_bobufs; /* (V) Buffer's associated vnode. */
109 struct buf *b_left; /* (V) splay tree link */
110 struct buf *b_right; /* (V) splay tree link */
111 uint32_t b_vflags; /* (V) BV_* flags */
112 TAILQ_ENTRY(buf) b_freelist; /* (Q) Free list position inactive. */
113 unsigned short b_qindex; /* (Q) buffer queue index */
114 uint32_t b_flags; /* B_* flags. */
115 b_xflags_t b_xflags; /* extra flags */
116 struct lock b_lock; /* Buffer lock */
117 long b_bufsize; /* Allocated buffer size. */
118 long b_runningbufspace; /* when I/O is running, pipelining */
119 caddr_t b_kvabase; /* base kva for buffer */
120 int b_kvasize; /* size of kva for buffer */
121 daddr_t b_lblkno; /* Logical block number. */
122 struct vnode *b_vp; /* Device vnode. */
123 int b_dirtyoff; /* Offset in buffer of dirty region. */
124 int b_dirtyend; /* Offset of end of dirty region. */
125 struct ucred *b_rcred; /* Read credentials reference. */
126 struct ucred *b_wcred; /* Write credentials reference. */
127 void *b_saveaddr; /* Original b_addr for physio. */
132 TAILQ_HEAD(cluster_list_head, buf) cluster_head;
133 TAILQ_ENTRY(buf) cluster_entry;
135 struct vm_page *b_pages[btoc(MAXPHYS)];
137 struct workhead b_dep; /* (D) List of filesystem dependencies. */
144 #define b_object b_bufobj->bo_object
147 * These flags are kept in b_flags.
151 * B_ASYNC VOP calls on bp's are usually async whether or not
152 * B_ASYNC is set, but some subsystems, such as NFS, like
153 * to know what is best for the caller so they can
156 * B_PAGING Indicates that bp is being used by the paging system or
157 * some paging system and that the bp is not linked into
158 * the b_vp's clean/dirty linked lists or ref counts.
159 * Buffer vp reassignments are illegal in this case.
161 * B_CACHE This may only be set if the buffer is entirely valid.
162 * The situation where B_DELWRI is set and B_CACHE is
163 * clear MUST be committed to disk by getblk() so
164 * B_DELWRI can also be cleared. See the comments for
165 * getblk() in kern/vfs_bio.c. If B_CACHE is clear,
166 * the caller is expected to clear BIO_ERROR and B_INVAL,
167 * set BIO_READ, and initiate an I/O.
169 * The 'entire buffer' is defined to be the range from
170 * 0 through b_bcount.
172 * B_MALLOC Request that the buffer be allocated from the malloc
173 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
175 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers
176 * by filesystems that allow clustering when the buffer
177 * is fully dirty and indicates that it may be clustered
178 * with other adjacent dirty buffers. Note the clustering
179 * may not be used with the stage 1 data write under NFS
180 * but may be used for the commit rpc portion.
182 * B_VMIO Indicates that the buffer is tied into an VM object.
183 * The buffer's data is always PAGE_SIZE aligned even
184 * if b_bufsize and b_bcount are not. ( b_bufsize is
185 * always at least DEV_BSIZE aligned, though ).
187 * B_DIRECT Hint that we should attempt to completely free
188 * the pages underlying the buffer. B_DIRECT is
189 * sticky until the buffer is released and typically
190 * only has an effect when B_RELBUF is also set.
194 #define B_AGE 0x00000001 /* Move to age queue when I/O done. */
195 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */
196 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */
197 #define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */
198 #define B_DEFERRED 0x00000010 /* Skipped over for cleaning */
199 #define B_CACHE 0x00000020 /* Bread found us in the cache. */
200 #define B_VALIDSUSPWRT 0x00000040 /* Valid write during suspension. */
201 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */
202 #define B_PERSISTENT 0x00000100 /* Perm. ref'ed while EXT2FS mounted. */
203 #define B_DONE 0x00000200 /* I/O completed. */
204 #define B_EINTR 0x00000400 /* I/O was interrupted */
205 #define B_00000800 0x00000800 /* Available flag. */
206 #define B_00001000 0x00001000 /* Available flag. */
207 #define B_INVAL 0x00002000 /* Does not contain valid info. */
208 #define B_00004000 0x00004000 /* Available flag. */
209 #define B_NOCACHE 0x00008000 /* Do not cache block after use. */
210 #define B_MALLOC 0x00010000 /* malloced b_data */
211 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */
212 #define B_000400000 0x00040000 /* Available flag. */
213 #define B_000800000 0x00080000 /* Available flag. */
214 #define B_00100000 0x00100000 /* Available flag. */
215 #define B_DIRTY 0x00200000 /* Needs writing later (in EXT2FS). */
216 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */
217 #define B_00800000 0x00800000 /* Available flag. */
218 #define B_01000000 0x01000000 /* Available flag. */
219 #define B_NEEDSGIANT 0x02000000 /* Buffer's vnode needs giant. */
220 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */
221 #define B_MANAGED 0x08000000 /* Managed by FS. */
222 #define B_RAM 0x10000000 /* Read ahead mark (flag) */
223 #define B_VMIO 0x20000000 /* VMIO flag */
224 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */
225 #define B_REMFREE 0x80000000 /* Delayed bremfree */
227 #define PRINT_BUF_FLAGS "\20\40remfree\37cluster\36vmio\35ram\34b27" \
228 "\33paging\32b25\31b24\30b23\27relbuf\26dirty\25b20" \
229 "\24b19\23b18\22clusterok\21malloc\20nocache\17b14\16inval" \
230 "\15b12\14b11\13eintr\12done\11persist\10delwri\7validsuspwrt" \
231 "\6cache\5deferred\4direct\3async\2needcommit\1age"
234 * These flags are kept in b_xflags.
236 #define BX_VNDIRTY 0x00000001 /* On vnode dirty list */
237 #define BX_VNCLEAN 0x00000002 /* On vnode clean list */
238 #define BX_BKGRDWRITE 0x00000010 /* Do writes in background */
239 #define BX_BKGRDMARKER 0x00000020 /* Mark buffer for splay tree */
240 #define BX_ALTDATA 0x00000040 /* Holds extended data */
242 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */
245 * These flags are kept in b_vflags.
247 #define BV_SCANNED 0x00000001 /* VOP_FSYNC funcs mark written bufs */
248 #define BV_BKGRDINPROG 0x00000002 /* Background write in progress */
249 #define BV_BKGRDWAIT 0x00000004 /* Background write waiting */
255 extern const char *buf_wmesg; /* Default buffer lock message */
256 #define BUF_WMESG "bufwait"
257 #include <sys/proc.h> /* XXX for curthread */
258 #include <sys/mutex.h>
263 #define BUF_LOCKINIT(bp) \
264 lockinit(&(bp)->b_lock, PRIBIO + 4, buf_wmesg, 0, 0)
267 * Get a lock sleeping non-interruptably until it becomes available.
269 static __inline int BUF_LOCK(struct buf *, int, struct mtx *);
271 BUF_LOCK(struct buf *bp, int locktype, struct mtx *interlock)
276 mtx_lock(bp->b_lock.lk_interlock);
277 locktype |= LK_INTERNAL;
278 bp->b_lock.lk_wmesg = buf_wmesg;
279 bp->b_lock.lk_prio = PRIBIO + 4;
280 ret = lockmgr(&(bp)->b_lock, locktype, interlock, curthread);
285 * Get a lock sleeping with specified interruptably and timeout.
287 static __inline int BUF_TIMELOCK(struct buf *, int, struct mtx *,
290 BUF_TIMELOCK(struct buf *bp, int locktype, struct mtx *interlock,
291 char *wmesg, int catch, int timo)
296 mtx_lock(bp->b_lock.lk_interlock);
297 locktype |= LK_INTERNAL | LK_TIMELOCK;
298 bp->b_lock.lk_wmesg = wmesg;
299 bp->b_lock.lk_prio = (PRIBIO + 4) | catch;
300 bp->b_lock.lk_timo = timo;
301 ret = lockmgr(&(bp)->b_lock, (locktype), interlock, curthread);
306 * Release a lock. Only the acquiring process may free the lock unless
307 * it has been handed off to biodone.
309 static __inline void BUF_UNLOCK(struct buf *);
311 BUF_UNLOCK(struct buf *bp)
316 KASSERT((bp->b_flags & B_REMFREE) == 0,
317 ("BUF_UNLOCK %p while B_REMFREE is still set.", bp));
318 lockmgr(&(bp)->b_lock, LK_RELEASE, NULL, curthread);
323 * Free a buffer lock.
325 #define BUF_LOCKFREE(bp) \
327 if (BUF_REFCNT(bp) > 0) \
328 panic("free locked buf"); \
329 lockdestroy(&(bp)->b_lock); \
332 #ifdef _SYS_PROC_H_ /* Avoid #include <sys/proc.h> pollution */
334 * When initiating asynchronous I/O, change ownership of the lock to the
335 * kernel. Once done, the lock may legally released by biodone. The
336 * original owning process can no longer acquire it recursively, but must
337 * wait until the I/O is completed and the lock has been freed by biodone.
339 static __inline void BUF_KERNPROC(struct buf *);
341 BUF_KERNPROC(struct buf *bp)
343 struct thread *td = curthread;
345 if (!TD_IS_IDLETHREAD(td) && bp->b_lock.lk_lockholder == td)
347 bp->b_lock.lk_lockholder = LK_KERNPROC;
351 * Find out the number of references to a lock.
353 static __inline int BUF_REFCNT(struct buf *);
355 BUF_REFCNT(struct buf *bp)
360 * When the system is panicing, the lock manager grants all lock
361 * requests whether or not the lock is available. To avoid "unlocked
362 * buffer" panics after a crash, we just claim that all buffers
363 * are locked when cleaning up after a system panic.
365 if (panicstr != NULL)
368 ret = lockcount(&(bp)->b_lock);
375 * Find out the number of waiters on a lock.
377 static __inline int BUF_LOCKWAITERS(struct buf *);
379 BUF_LOCKWAITERS(struct buf *bp)
381 return (lockwaiters(&bp->b_lock));
386 struct buf_queue_head {
387 TAILQ_HEAD(buf_queue, buf) queue;
389 struct buf *insert_point;
390 struct buf *switch_point;
394 * This structure describes a clustered I/O. It is stored in the b_saveaddr
395 * field of the buffer on which I/O is done. At I/O completion, cluster
396 * callback uses the structure to parcel I/O's to individual buffers, and
397 * then free's this structure.
399 struct cluster_save {
400 long bs_bcount; /* Saved b_bcount. */
401 long bs_bufsize; /* Saved b_bufsize. */
402 void *bs_saveaddr; /* Saved b_addr. */
403 int bs_nchildren; /* Number of associated buffers. */
404 struct buf **bs_children; /* List of associated buffers. */
410 bwrite(struct buf *bp)
413 KASSERT(bp->b_bufobj != NULL, ("bwrite: no bufobj bp=%p", bp));
414 KASSERT(bp->b_bufobj->bo_ops != NULL, ("bwrite: no bo_ops bp=%p", bp));
415 KASSERT(bp->b_bufobj->bo_ops->bop_write != NULL,
416 ("bwrite: no bop_write bp=%p", bp));
417 return (BO_WRITE(bp->b_bufobj, bp));
421 bstrategy(struct buf *bp)
424 KASSERT(bp->b_bufobj != NULL, ("bstrategy: no bufobj bp=%p", bp));
425 KASSERT(bp->b_bufobj->bo_ops != NULL,
426 ("bstrategy: no bo_ops bp=%p", bp));
427 KASSERT(bp->b_bufobj->bo_ops->bop_strategy != NULL,
428 ("bstrategy: no bop_strategy bp=%p", bp));
429 BO_STRATEGY(bp->b_bufobj, bp);
433 buf_start(struct buf *bp)
436 (*bioops.io_start)(bp);
440 buf_complete(struct buf *bp)
442 if (bioops.io_complete)
443 (*bioops.io_complete)(bp);
447 buf_deallocate(struct buf *bp)
449 if (bioops.io_deallocate)
450 (*bioops.io_deallocate)(bp);
455 buf_countdeps(struct buf *bp, int i)
457 if (bioops.io_countdeps)
458 return ((*bioops.io_countdeps)(bp, i));
466 * Zero out the buffer's data area.
468 #define clrbuf(bp) { \
469 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \
474 * Flags for getblk's last parameter.
476 #define GB_LOCK_NOWAIT 0x0001 /* Fail if we block on a buf lock. */
477 #define GB_NOCREAT 0x0002 /* Don't create a buf if not found. */
480 extern int nbuf; /* The number of buffer headers */
481 extern int maxswzone; /* Max KVA for swap structures */
482 extern int maxbcache; /* Max KVA for buffer cache */
483 extern int runningbufspace;
484 extern int hibufspace;
485 extern int dirtybufthresh;
486 extern int bdwriteskip;
487 extern int dirtybufferflushes;
488 extern int altbufferflushes;
489 extern int buf_maxio; /* nominal maximum I/O for buffer */
490 extern struct buf *buf; /* The buffer headers. */
491 extern char *buffers; /* The buffer contents. */
492 extern int bufpages; /* Number of memory pages in the buffer pool. */
493 extern struct buf *swbuf; /* Swap I/O buffer headers. */
494 extern int nswbuf; /* Number of swap I/O buffer headers. */
495 extern int cluster_pbuf_freecnt; /* Number of pbufs for clusters */
496 extern int vnode_pbuf_freecnt; /* Number of pbufs for vnode pager */
498 void runningbufwakeup(struct buf *);
499 void waitrunningbufspace(void);
500 caddr_t kern_vfs_bio_buffer_alloc(caddr_t v, long physmem_est);
502 void bwillwrite(void);
503 int buf_dirty_count_severe(void);
504 void bremfree(struct buf *);
505 void bremfreef(struct buf *); /* XXX Force bremfree, only for nfs. */
506 int bread(struct vnode *, daddr_t, int, struct ucred *, struct buf **);
507 void breada(struct vnode *, daddr_t *, int *, int, struct ucred *);
508 int breadn(struct vnode *, daddr_t, int, daddr_t *, int *, int,
509 struct ucred *, struct buf **);
510 void bdwrite(struct buf *);
511 void bawrite(struct buf *);
512 void bdirty(struct buf *);
513 void bundirty(struct buf *);
514 void bufstrategy(struct bufobj *, struct buf *);
515 void brelse(struct buf *);
516 void bqrelse(struct buf *);
517 int vfs_bio_awrite(struct buf *);
518 struct buf * getpbuf(int *);
519 struct buf *incore(struct bufobj *, daddr_t);
520 struct buf *gbincore(struct bufobj *, daddr_t);
521 struct buf *getblk(struct vnode *, daddr_t, int, int, int, int);
522 struct buf *geteblk(int);
523 int bufwait(struct buf *);
524 int bufwrite(struct buf *);
525 void bufdone(struct buf *);
526 void bufdone_finish(struct buf *);
528 int cluster_read(struct vnode *, u_quad_t, daddr_t, long,
529 struct ucred *, long, int, struct buf **);
530 int cluster_wbuild(struct vnode *, long, daddr_t, int);
531 void cluster_write(struct vnode *, struct buf *, u_quad_t, int);
532 void vfs_bio_set_validclean(struct buf *, int base, int size);
533 void vfs_bio_clrbuf(struct buf *);
534 void vfs_busy_pages(struct buf *, int clear_modify);
535 void vfs_unbusy_pages(struct buf *);
536 int vmapbuf(struct buf *);
537 void vunmapbuf(struct buf *);
538 void relpbuf(struct buf *, int *);
539 void brelvp(struct buf *);
540 void bgetvp(struct vnode *, struct buf *);
541 void pbgetbo(struct bufobj *bo, struct buf *bp);
542 void pbgetvp(struct vnode *, struct buf *);
543 void pbrelbo(struct buf *);
544 void pbrelvp(struct buf *);
545 int allocbuf(struct buf *bp, int size);
546 void reassignbuf(struct buf *);
547 struct buf *trypbuf(int *);
548 void bwait(struct buf *, u_char, const char *);
549 void bdone(struct buf *);
550 void bpin(struct buf *);
551 void bunpin(struct buf *);
552 void bunpin_wait(struct buf *);
556 #endif /* !_SYS_BUF_H_ */