FreeBSD/Linux Kernel Cross Reference
sys/sys/buf.h
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 * 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.
21 *
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
32 * SUCH DAMAGE.
33 *
34 * @(#)buf.h 8.9 (Berkeley) 3/30/95
35 * $FreeBSD: releng/6.4/sys/sys/buf.h 170553 2007-06-11 10:53:48Z kib $
36 */
37
38 #ifndef _SYS_BUF_H_
39 #define _SYS_BUF_H_
40
41 #include <sys/bufobj.h>
42 #include <sys/queue.h>
43 #include <sys/lock.h>
44 #include <sys/lockmgr.h>
45
46 struct bio;
47 struct buf;
48 struct bufobj;
49 struct mount;
50 struct vnode;
51 struct uio;
52
53 /*
54 * To avoid including <ufs/ffs/softdep.h>
55 */
56 LIST_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 */
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);
68 } bioops;
69
70 struct vm_object;
71
72 typedef unsigned char b_xflags_t;
73
74 /*
75 * The buffer header describes an I/O operation in the kernel.
76 *
77 * NOTES:
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.
82 *
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 ).
86 *
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.
89 *
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
94 */
95 struct buf {
96 struct bufobj *b_bufobj;
97 long b_bcount;
98 void *b_caller1;
99 caddr_t b_data;
100 int b_error;
101 uint8_t b_iocmd;
102 uint8_t b_ioflags;
103 off_t b_iooffset;
104 long b_resid;
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. */
128 union pager_info {
129 int pg_reqpage;
130 } b_pager;
131 union cluster_info {
132 TAILQ_HEAD(cluster_list_head, buf) cluster_head;
133 TAILQ_ENTRY(buf) cluster_entry;
134 } b_cluster;
135 struct vm_page *b_pages[btoc(MAXPHYS)];
136 int b_npages;
137 struct workhead b_dep; /* (D) List of filesystem dependencies. */
138 };
139
140 #define b_object b_bufobj->bo_object
141
142 /*
143 * These flags are kept in b_flags.
144 *
145 * Notes:
146 *
147 * B_ASYNC VOP calls on bp's are usually async whether or not
148 * B_ASYNC is set, but some subsystems, such as NFS, like
149 * to know what is best for the caller so they can
150 * optimize the I/O.
151 *
152 * B_PAGING Indicates that bp is being used by the paging system or
153 * some paging system and that the bp is not linked into
154 * the b_vp's clean/dirty linked lists or ref counts.
155 * Buffer vp reassignments are illegal in this case.
156 *
157 * B_CACHE This may only be set if the buffer is entirely valid.
158 * The situation where B_DELWRI is set and B_CACHE is
159 * clear MUST be committed to disk by getblk() so
160 * B_DELWRI can also be cleared. See the comments for
161 * getblk() in kern/vfs_bio.c. If B_CACHE is clear,
162 * the caller is expected to clear BIO_ERROR and B_INVAL,
163 * set BIO_READ, and initiate an I/O.
164 *
165 * The 'entire buffer' is defined to be the range from
166 * 0 through b_bcount.
167 *
168 * B_MALLOC Request that the buffer be allocated from the malloc
169 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
170 *
171 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers
172 * by filesystems that allow clustering when the buffer
173 * is fully dirty and indicates that it may be clustered
174 * with other adjacent dirty buffers. Note the clustering
175 * may not be used with the stage 1 data write under NFS
176 * but may be used for the commit rpc portion.
177 *
178 * B_VMIO Indicates that the buffer is tied into an VM object.
179 * The buffer's data is always PAGE_SIZE aligned even
180 * if b_bufsize and b_bcount are not. ( b_bufsize is
181 * always at least DEV_BSIZE aligned, though ).
182 *
183 * B_DIRECT Hint that we should attempt to completely free
184 * the pages underlying the buffer. B_DIRECT is
185 * sticky until the buffer is released and typically
186 * only has an effect when B_RELBUF is also set.
187 *
188 */
189
190 #define B_AGE 0x00000001 /* Move to age queue when I/O done. */
191 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */
192 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */
193 #define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */
194 #define B_DEFERRED 0x00000010 /* Skipped over for cleaning */
195 #define B_CACHE 0x00000020 /* Bread found us in the cache. */
196 #define B_VALIDSUSPWRT 0x00000040 /* Valid write during suspension. */
197 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */
198 #define B_PERSISTENT 0x00000100 /* Perm. ref'ed while EXT2FS mounted. */
199 #define B_DONE 0x00000200 /* I/O completed. */
200 #define B_EINTR 0x00000400 /* I/O was interrupted */
201 #define B_00000800 0x00000800 /* Available flag. */
202 #define B_00001000 0x00001000 /* Available flag. */
203 #define B_INVAL 0x00002000 /* Does not contain valid info. */
204 #define B_00004000 0x00004000 /* Available flag. */
205 #define B_NOCACHE 0x00008000 /* Do not cache block after use. */
206 #define B_MALLOC 0x00010000 /* malloced b_data */
207 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */
208 #define B_000400000 0x00040000 /* Available flag. */
209 #define B_000800000 0x00080000 /* Available flag. */
210 #define B_00100000 0x00100000 /* Available flag. */
211 #define B_DIRTY 0x00200000 /* Needs writing later (in EXT2FS). */
212 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */
213 #define B_00800000 0x00800000 /* Available flag. */
214 #define B_01000000 0x01000000 /* Available flag. */
215 #define B_02000000 0x02000000 /* Available flag. */
216 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */
217 #define B_08000000 0x08000000 /* Available flag. */
218 #define B_RAM 0x10000000 /* Read ahead mark (flag) */
219 #define B_VMIO 0x20000000 /* VMIO flag */
220 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */
221 #define B_REMFREE 0x80000000 /* Delayed bremfree */
222
223 #define PRINT_BUF_FLAGS "\2\40remfree\37cluster\36vmio\35ram\34b27" \
224 "\33paging\32b25\31b24\30b23\27relbuf\26dirty\25b20" \
225 "\24b19\23b18\22clusterok\21malloc\20nocache\17b14\16inval" \
226 "\15b12\14b11\13eintr\12done\11persist\10delwri\7validsuspwrt" \
227 "\6cache\5deferred\4direct\3async\2needcommit\1age"
228
229 /*
230 * These flags are kept in b_xflags.
231 */
232 #define BX_VNDIRTY 0x00000001 /* On vnode dirty list */
233 #define BX_VNCLEAN 0x00000002 /* On vnode clean list */
234 #define BX_BKGRDWRITE 0x00000010 /* Do writes in background */
235 #define BX_BKGRDMARKER 0x00000020 /* Mark buffer for splay tree */
236 #define BX_ALTDATA 0x00000040 /* Holds extended data */
237
238 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */
239
240 /*
241 * These flags are kept in b_vflags.
242 */
243 #define BV_SCANNED 0x00000001 /* VOP_FSYNC funcs mark written bufs */
244 #define BV_BKGRDINPROG 0x00000002 /* Background write in progress */
245 #define BV_BKGRDWAIT 0x00000004 /* Background write waiting */
246
247 #ifdef _KERNEL
248 /*
249 * Buffer locking
250 */
251 extern const char *buf_wmesg; /* Default buffer lock message */
252 #define BUF_WMESG "bufwait"
253 #include <sys/proc.h> /* XXX for curthread */
254 #include <sys/mutex.h>
255
256 /*
257 * Initialize a lock.
258 */
259 #define BUF_LOCKINIT(bp) \
260 lockinit(&(bp)->b_lock, PRIBIO + 4, buf_wmesg, 0, 0)
261 /*
262 *
263 * Get a lock sleeping non-interruptably until it becomes available.
264 */
265 static __inline int BUF_LOCK(struct buf *, int, struct mtx *);
266 static __inline int
267 BUF_LOCK(struct buf *bp, int locktype, struct mtx *interlock)
268 {
269 int s, ret;
270
271 s = splbio();
272 mtx_lock(bp->b_lock.lk_interlock);
273 locktype |= LK_INTERNAL;
274 bp->b_lock.lk_wmesg = buf_wmesg;
275 bp->b_lock.lk_prio = PRIBIO + 4;
276 ret = lockmgr(&(bp)->b_lock, locktype, interlock, curthread);
277 splx(s);
278 return ret;
279 }
280 /*
281 * Get a lock sleeping with specified interruptably and timeout.
282 */
283 static __inline int BUF_TIMELOCK(struct buf *, int, struct mtx *,
284 char *, int, int);
285 static __inline int
286 BUF_TIMELOCK(struct buf *bp, int locktype, struct mtx *interlock,
287 char *wmesg, int catch, int timo)
288 {
289 int s, ret;
290
291 s = splbio();
292 mtx_lock(bp->b_lock.lk_interlock);
293 locktype |= LK_INTERNAL | LK_TIMELOCK;
294 bp->b_lock.lk_wmesg = wmesg;
295 bp->b_lock.lk_prio = (PRIBIO + 4) | catch;
296 bp->b_lock.lk_timo = timo;
297 ret = lockmgr(&(bp)->b_lock, (locktype), interlock, curthread);
298 splx(s);
299 return ret;
300 }
301 /*
302 * Release a lock. Only the acquiring process may free the lock unless
303 * it has been handed off to biodone.
304 */
305 static __inline void BUF_UNLOCK(struct buf *);
306 static __inline void
307 BUF_UNLOCK(struct buf *bp)
308 {
309 int s;
310
311 s = splbio();
312 KASSERT((bp->b_flags & B_REMFREE) == 0,
313 ("BUF_UNLOCK %p while B_REMFREE is still set.", bp));
314 lockmgr(&(bp)->b_lock, LK_RELEASE, NULL, curthread);
315 splx(s);
316 }
317
318 /*
319 * Free a buffer lock.
320 */
321 #define BUF_LOCKFREE(bp) \
322 do { \
323 if (BUF_REFCNT(bp) > 0) \
324 panic("free locked buf"); \
325 lockdestroy(&(bp)->b_lock); \
326 } while (0)
327
328 #ifdef _SYS_PROC_H_ /* Avoid #include <sys/proc.h> pollution */
329 /*
330 * When initiating asynchronous I/O, change ownership of the lock to the
331 * kernel. Once done, the lock may legally released by biodone. The
332 * original owning process can no longer acquire it recursively, but must
333 * wait until the I/O is completed and the lock has been freed by biodone.
334 */
335 static __inline void BUF_KERNPROC(struct buf *);
336 static __inline void
337 BUF_KERNPROC(struct buf *bp)
338 {
339 struct thread *td = curthread;
340
341 if ((td != PCPU_GET(idlethread))
342 && bp->b_lock.lk_lockholder == td)
343 td->td_locks--;
344 bp->b_lock.lk_lockholder = LK_KERNPROC;
345 }
346 #endif
347 /*
348 * Find out the number of references to a lock.
349 */
350 static __inline int BUF_REFCNT(struct buf *);
351 static __inline int
352 BUF_REFCNT(struct buf *bp)
353 {
354 int s, ret;
355
356 /*
357 * When the system is panicing, the lock manager grants all lock
358 * requests whether or not the lock is available. To avoid "unlocked
359 * buffer" panics after a crash, we just claim that all buffers
360 * are locked when cleaning up after a system panic.
361 */
362 if (panicstr != NULL)
363 return (1);
364 s = splbio();
365 ret = lockcount(&(bp)->b_lock);
366 splx(s);
367 return ret;
368 }
369
370
371 /*
372 * Find out the number of waiters on a lock.
373 */
374 static __inline int BUF_LOCKWAITERS(struct buf *);
375 static __inline int
376 BUF_LOCKWAITERS(struct buf *bp)
377 {
378 return (lockwaiters(&bp->b_lock));
379 }
380
381 #endif /* _KERNEL */
382
383 struct buf_queue_head {
384 TAILQ_HEAD(buf_queue, buf) queue;
385 daddr_t last_pblkno;
386 struct buf *insert_point;
387 struct buf *switch_point;
388 };
389
390 /*
391 * This structure describes a clustered I/O. It is stored in the b_saveaddr
392 * field of the buffer on which I/O is done. At I/O completion, cluster
393 * callback uses the structure to parcel I/O's to individual buffers, and
394 * then free's this structure.
395 */
396 struct cluster_save {
397 long bs_bcount; /* Saved b_bcount. */
398 long bs_bufsize; /* Saved b_bufsize. */
399 void *bs_saveaddr; /* Saved b_addr. */
400 int bs_nchildren; /* Number of associated buffers. */
401 struct buf **bs_children; /* List of associated buffers. */
402 };
403
404 #ifdef _KERNEL
405
406 static __inline int
407 bwrite(struct buf *bp)
408 {
409
410 KASSERT(bp->b_bufobj != NULL, ("bwrite: no bufobj bp=%p", bp));
411 KASSERT(bp->b_bufobj->bo_ops != NULL, ("bwrite: no bo_ops bp=%p", bp));
412 KASSERT(bp->b_bufobj->bo_ops->bop_write != NULL,
413 ("bwrite: no bop_write bp=%p", bp));
414 return (BO_WRITE(bp->b_bufobj, bp));
415 }
416
417 static __inline void
418 bstrategy(struct buf *bp)
419 {
420
421 KASSERT(bp->b_bufobj != NULL, ("bstrategy: no bufobj bp=%p", bp));
422 KASSERT(bp->b_bufobj->bo_ops != NULL,
423 ("bstrategy: no bo_ops bp=%p", bp));
424 KASSERT(bp->b_bufobj->bo_ops->bop_strategy != NULL,
425 ("bstrategy: no bop_strategy bp=%p", bp));
426 BO_STRATEGY(bp->b_bufobj, bp);
427 }
428
429 static __inline void
430 buf_start(struct buf *bp)
431 {
432 if (bioops.io_start)
433 (*bioops.io_start)(bp);
434 }
435
436 static __inline void
437 buf_complete(struct buf *bp)
438 {
439 if (bioops.io_complete)
440 (*bioops.io_complete)(bp);
441 }
442
443 static __inline void
444 buf_deallocate(struct buf *bp)
445 {
446 if (bioops.io_deallocate)
447 (*bioops.io_deallocate)(bp);
448 BUF_LOCKFREE(bp);
449 }
450
451 static __inline int
452 buf_countdeps(struct buf *bp, int i)
453 {
454 if (bioops.io_countdeps)
455 return ((*bioops.io_countdeps)(bp, i));
456 else
457 return (0);
458 }
459
460 #endif /* _KERNEL */
461
462 /*
463 * Zero out the buffer's data area.
464 */
465 #define clrbuf(bp) { \
466 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \
467 (bp)->b_resid = 0; \
468 }
469
470 /*
471 * Flags for getblk's last parameter.
472 */
473 #define GB_LOCK_NOWAIT 0x0001 /* Fail if we block on a buf lock. */
474 #define GB_NOCREAT 0x0002 /* Don't create a buf if not found. */
475
476 #ifdef _KERNEL
477 extern int nbuf; /* The number of buffer headers */
478 extern int maxswzone; /* Max KVA for swap structures */
479 extern int maxbcache; /* Max KVA for buffer cache */
480 extern int runningbufspace;
481 extern int hibufspace;
482 extern int dirtybufthresh;
483 extern int bdwriteskip;
484 extern int dirtybufferflushes;
485 extern int altbufferflushes;
486 extern int buf_maxio; /* nominal maximum I/O for buffer */
487 extern struct buf *buf; /* The buffer headers. */
488 extern char *buffers; /* The buffer contents. */
489 extern int bufpages; /* Number of memory pages in the buffer pool. */
490 extern struct buf *swbuf; /* Swap I/O buffer headers. */
491 extern int nswbuf; /* Number of swap I/O buffer headers. */
492 extern int cluster_pbuf_freecnt; /* Number of pbufs for clusters */
493 extern int vnode_pbuf_freecnt; /* Number of pbufs for vnode pager */
494
495 void runningbufwakeup(struct buf *);
496 void waitrunningbufspace(void);
497 caddr_t kern_vfs_bio_buffer_alloc(caddr_t v, long physmem_est);
498 void bufinit(void);
499 void bwillwrite(void);
500 int buf_dirty_count_severe(void);
501 void bremfree(struct buf *);
502 void bremfreef(struct buf *); /* XXX Force bremfree, only for nfs. */
503 int bread(struct vnode *, daddr_t, int, struct ucred *, struct buf **);
504 int breadn(struct vnode *, daddr_t, int, daddr_t *, int *, int,
505 struct ucred *, struct buf **);
506 void bdwrite(struct buf *);
507 void bawrite(struct buf *);
508 void bdirty(struct buf *);
509 void bundirty(struct buf *);
510 void bufstrategy(struct bufobj *, struct buf *);
511 void brelse(struct buf *);
512 void bqrelse(struct buf *);
513 int vfs_bio_awrite(struct buf *);
514 struct buf * getpbuf(int *);
515 struct buf *incore(struct bufobj *, daddr_t);
516 struct buf *gbincore(struct bufobj *, daddr_t);
517 struct buf *getblk(struct vnode *, daddr_t, int, int, int, int);
518 struct buf *geteblk(int);
519 int bufwait(struct buf *);
520 int bufwrite(struct buf *);
521 void bufdone(struct buf *);
522
523 int cluster_read(struct vnode *, u_quad_t, daddr_t, long,
524 struct ucred *, long, int, struct buf **);
525 int cluster_wbuild(struct vnode *, long, daddr_t, int);
526 void cluster_write(struct vnode *, struct buf *, u_quad_t, int);
527 void vfs_bio_set_validclean(struct buf *, int base, int size);
528 void vfs_bio_clrbuf(struct buf *);
529 void vfs_busy_pages(struct buf *, int clear_modify);
530 void vfs_unbusy_pages(struct buf *);
531 int vmapbuf(struct buf *);
532 void vunmapbuf(struct buf *);
533 void relpbuf(struct buf *, int *);
534 void brelvp(struct buf *);
535 void bgetvp(struct vnode *, struct buf *);
536 void pbgetbo(struct bufobj *bo, struct buf *bp);
537 void pbgetvp(struct vnode *, struct buf *);
538 void pbrelbo(struct buf *);
539 void pbrelvp(struct buf *);
540 int allocbuf(struct buf *bp, int size);
541 void reassignbuf(struct buf *);
542 struct buf *trypbuf(int *);
543 void bwait(struct buf *, u_char, const char *);
544 void bdone(struct buf *);
545
546 #endif /* _KERNEL */
547
548 #endif /* !_SYS_BUF_H_ */
Cache object: 8e068d7aa25130b4446451af460e837f
|