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 * 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: releng/5.2/sys/sys/buf.h 122747 2003-11-15 09:28:09Z phk $
40 */
41
42 #ifndef _SYS_BUF_H_
43 #define _SYS_BUF_H_
44
45 #include <sys/queue.h>
46 #include <sys/lock.h>
47 #include <sys/lockmgr.h>
48
49 struct bio;
50 struct buf;
51 struct mount;
52 struct vnode;
53
54 /*
55 * To avoid including <ufs/ffs/softdep.h>
56 */
57 LIST_HEAD(workhead, worklist);
58 /*
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
62 * to each buffer.
63 */
64 extern struct bio_ops {
65 void (*io_start)(struct buf *);
66 void (*io_complete)(struct buf *);
67 void (*io_deallocate)(struct buf *);
68 void (*io_movedeps)(struct buf *, struct buf *);
69 int (*io_countdeps)(struct buf *, int);
70 } bioops;
71
72 struct buf_ops {
73 char *bop_name;
74 int (*bop_write)(struct buf *);
75 };
76
77 extern struct buf_ops buf_ops_bio;
78
79 struct vm_object;
80
81 typedef unsigned char b_xflags_t;
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 *
99 * All fields are protected by the buffer lock except those marked:
100 * V - Protected by owning vnode lock
101 * Q - Protected by the buf queue lock
102 * D - Protected by an dependency implementation specific lock
103 */
104 struct buf {
105 struct bio b_io; /* "Builtin" I/O request. */
106 #define b_bcount b_io.bio_bcount
107 #define b_caller1 b_io.bio_caller1
108 #define b_data b_io.bio_data
109 #define b_dev b_io.bio_dev
110 #define b_error b_io.bio_error
111 #define b_iocmd b_io.bio_cmd
112 #define b_ioflags b_io.bio_flags
113 #define b_iooffset b_io.bio_offset
114 #define b_resid b_io.bio_resid
115 struct buf_ops *b_op;
116 unsigned b_magic;
117 #define B_MAGIC_BIO 0x10b10b10
118 #define B_MAGIC_NFS 0x67238234
119 void (*b_iodone)(struct buf *);
120 daddr_t b_blkno; /* Underlying physical block number. */
121 off_t b_offset; /* Offset into file. */
122 TAILQ_ENTRY(buf) b_vnbufs; /* (V) Buffer's associated vnode. */
123 struct buf *b_left; /* (V) splay tree link */
124 struct buf *b_right; /* (V) splay tree link */
125 uint32_t b_vflags; /* (V) BV_* flags */
126 TAILQ_ENTRY(buf) b_freelist; /* (Q) Free list position inactive. */
127 unsigned short b_qindex; /* (Q) buffer queue index */
128 uint32_t b_flags; /* B_* flags. */
129 b_xflags_t b_xflags; /* extra flags */
130 struct lock b_lock; /* Buffer lock */
131 long b_bufsize; /* Allocated buffer size. */
132 long b_runningbufspace; /* when I/O is running, pipelining */
133 caddr_t b_kvabase; /* base kva for buffer */
134 int b_kvasize; /* size of kva for buffer */
135 daddr_t b_lblkno; /* Logical block number. */
136 struct vnode *b_vp; /* Device vnode. */
137 struct vm_object *b_object; /* Object for vp */
138 int b_dirtyoff; /* Offset in buffer of dirty region. */
139 int b_dirtyend; /* Offset of end of dirty region. */
140 struct ucred *b_rcred; /* Read credentials reference. */
141 struct ucred *b_wcred; /* Write credentials reference. */
142 void *b_saveaddr; /* Original b_addr for physio. */
143 union pager_info {
144 int pg_reqpage;
145 } b_pager;
146 union cluster_info {
147 TAILQ_HEAD(cluster_list_head, buf) cluster_head;
148 TAILQ_ENTRY(buf) cluster_entry;
149 } b_cluster;
150 struct vm_page *b_pages[btoc(MAXPHYS)];
151 int b_npages;
152 struct workhead b_dep; /* (D) List of filesystem dependencies. */
153 };
154
155 /*
156 * These flags are kept in b_flags.
157 *
158 * Notes:
159 *
160 * B_ASYNC VOP calls on bp's are usually async whether or not
161 * B_ASYNC is set, but some subsystems, such as NFS, like
162 * to know what is best for the caller so they can
163 * optimize the I/O.
164 *
165 * B_PAGING Indicates that bp is being used by the paging system or
166 * some paging system and that the bp is not linked into
167 * the b_vp's clean/dirty linked lists or ref counts.
168 * Buffer vp reassignments are illegal in this case.
169 *
170 * B_CACHE This may only be set if the buffer is entirely valid.
171 * The situation where B_DELWRI is set and B_CACHE is
172 * clear MUST be committed to disk by getblk() so
173 * B_DELWRI can also be cleared. See the comments for
174 * getblk() in kern/vfs_bio.c. If B_CACHE is clear,
175 * the caller is expected to clear BIO_ERROR and B_INVAL,
176 * set BIO_READ, and initiate an I/O.
177 *
178 * The 'entire buffer' is defined to be the range from
179 * 0 through b_bcount.
180 *
181 * B_MALLOC Request that the buffer be allocated from the malloc
182 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
183 *
184 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers
185 * by filesystems that allow clustering when the buffer
186 * is fully dirty and indicates that it may be clustered
187 * with other adjacent dirty buffers. Note the clustering
188 * may not be used with the stage 1 data write under NFS
189 * but may be used for the commit rpc portion.
190 *
191 * B_VMIO Indicates that the buffer is tied into an VM object.
192 * The buffer's data is always PAGE_SIZE aligned even
193 * if b_bufsize and b_bcount are not. ( b_bufsize is
194 * always at least DEV_BSIZE aligned, though ).
195 *
196 * B_DIRECT Hint that we should attempt to completely free
197 * the pages underlying the buffer. B_DIRECT is
198 * sticky until the buffer is released and typically
199 * only has an effect when B_RELBUF is also set.
200 *
201 */
202
203 #define B_AGE 0x00000001 /* Move to age queue when I/O done. */
204 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */
205 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */
206 #define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */
207 #define B_DEFERRED 0x00000010 /* Skipped over for cleaning */
208 #define B_CACHE 0x00000020 /* Bread found us in the cache. */
209 #define B_VALIDSUSPWRT 0x00000040 /* Valid write during suspension. */
210 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */
211 #define B_00000100 0x00000100 /* Available flag. */
212 #define B_DONE 0x00000200 /* I/O completed. */
213 #define B_EINTR 0x00000400 /* I/O was interrupted */
214 #define B_00000800 0x00000800 /* Available flag. */
215 #define B_00001000 0x00001000 /* Available flag. */
216 #define B_INVAL 0x00002000 /* Does not contain valid info. */
217 #define B_00004000 0x00004000 /* Available flag. */
218 #define B_NOCACHE 0x00008000 /* Do not cache block after use. */
219 #define B_MALLOC 0x00010000 /* malloced b_data */
220 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */
221 #define B_000400000 0x00040000 /* Available flag. */
222 #define B_000800000 0x00080000 /* Available flag. */
223 #define B_00100000 0x00100000 /* Available flag. */
224 #define B_DIRTY 0x00200000 /* Needs writing later (in EXT2FS). */
225 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */
226 #define B_00800000 0x00800000 /* Available flag. */
227 #define B_WRITEINPROG 0x01000000 /* Write in progress. */
228 #define B_02000000 0x02000000 /* Available flag. */
229 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */
230 #define B_08000000 0x08000000 /* Available flag. */
231 #define B_RAM 0x10000000 /* Read ahead mark (flag) */
232 #define B_VMIO 0x20000000 /* VMIO flag */
233 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */
234 #define B_80000000 0x80000000 /* Available flag. */
235
236 #define PRINT_BUF_FLAGS "\2\40b31\37cluster\36vmio\35ram\34b27" \
237 "\33paging\32b25\31writeinprog\30b23\27relbuf\26dirty\25b20" \
238 "\24b19\23phys\22clusterok\21malloc\20nocache\17locked\16inval" \
239 "\15scanned\14nowdrain\13eintr\12done\11b8\10delwri\7validsuspwrt" \
240 "\6cache\5deferred\4direct\3async\2needcommit\1age"
241
242 /*
243 * These flags are kept in b_xflags.
244 */
245 #define BX_VNDIRTY 0x00000001 /* On vnode dirty list */
246 #define BX_VNCLEAN 0x00000002 /* On vnode clean list */
247 #define BX_BKGRDWRITE 0x00000010 /* Do writes in background */
248 #define BX_BKGRDMARKER 0x00000020 /* Mark buffer for splay tree */
249 #define BX_ALTDATA 0x00000040 /* Holds extended data */
250
251 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */
252
253 /*
254 * These flags are kept in b_vflags.
255 */
256 #define BV_SCANNED 0x00000001 /* VOP_FSYNC funcs mark written bufs */
257 #define BV_BKGRDINPROG 0x00000002 /* Background write in progress */
258 #define BV_BKGRDWAIT 0x00000004 /* Background write waiting */
259
260 #ifdef _KERNEL
261 /*
262 * Buffer locking
263 */
264 extern const char *buf_wmesg; /* Default buffer lock message */
265 #define BUF_WMESG "bufwait"
266 #include <sys/proc.h> /* XXX for curthread */
267 #include <sys/mutex.h>
268
269 /*
270 * Initialize a lock.
271 */
272 #define BUF_LOCKINIT(bp) \
273 lockinit(&(bp)->b_lock, PRIBIO + 4, buf_wmesg, 0, 0)
274 /*
275 *
276 * Get a lock sleeping non-interruptably until it becomes available.
277 */
278 static __inline int BUF_LOCK(struct buf *, int, struct mtx *);
279 static __inline int
280 BUF_LOCK(struct buf *bp, int locktype, struct mtx *interlock)
281 {
282 int s, ret;
283
284 s = splbio();
285 mtx_lock(bp->b_lock.lk_interlock);
286 locktype |= LK_INTERNAL;
287 bp->b_lock.lk_wmesg = buf_wmesg;
288 bp->b_lock.lk_prio = PRIBIO + 4;
289 ret = lockmgr(&(bp)->b_lock, locktype, interlock, curthread);
290 splx(s);
291 return ret;
292 }
293 /*
294 * Get a lock sleeping with specified interruptably and timeout.
295 */
296 static __inline int BUF_TIMELOCK(struct buf *, int, struct mtx *,
297 char *, int, int);
298 static __inline int
299 BUF_TIMELOCK(struct buf *bp, int locktype, struct mtx *interlock,
300 char *wmesg, int catch, int timo)
301 {
302 int s, ret;
303
304 s = splbio();
305 mtx_lock(bp->b_lock.lk_interlock);
306 locktype |= LK_INTERNAL | LK_TIMELOCK;
307 bp->b_lock.lk_wmesg = wmesg;
308 bp->b_lock.lk_prio = (PRIBIO + 4) | catch;
309 bp->b_lock.lk_timo = timo;
310 ret = lockmgr(&(bp)->b_lock, (locktype), interlock, curthread);
311 splx(s);
312 return ret;
313 }
314 /*
315 * Release a lock. Only the acquiring process may free the lock unless
316 * it has been handed off to biodone.
317 */
318 static __inline void BUF_UNLOCK(struct buf *);
319 static __inline void
320 BUF_UNLOCK(struct buf *bp)
321 {
322 int s;
323
324 s = splbio();
325 lockmgr(&(bp)->b_lock, LK_RELEASE, NULL, curthread);
326 splx(s);
327 }
328
329 /*
330 * Free a buffer lock.
331 */
332 #define BUF_LOCKFREE(bp) \
333 do { \
334 if (BUF_REFCNT(bp) > 0) \
335 panic("free locked buf"); \
336 lockdestroy(&(bp)->b_lock); \
337 } while (0)
338
339 #ifdef _SYS_PROC_H_ /* Avoid #include <sys/proc.h> pollution */
340 /*
341 * When initiating asynchronous I/O, change ownership of the lock to the
342 * kernel. Once done, the lock may legally released by biodone. The
343 * original owning process can no longer acquire it recursively, but must
344 * wait until the I/O is completed and the lock has been freed by biodone.
345 */
346 static __inline void BUF_KERNPROC(struct buf *);
347 static __inline void
348 BUF_KERNPROC(struct buf *bp)
349 {
350 struct thread *td = curthread;
351
352 if ((td != PCPU_GET(idlethread))
353 && bp->b_lock.lk_lockholder == td)
354 td->td_locks--;
355 bp->b_lock.lk_lockholder = LK_KERNPROC;
356 }
357 #endif
358 /*
359 * Find out the number of references to a lock.
360 */
361 static __inline int BUF_REFCNT(struct buf *);
362 static __inline int
363 BUF_REFCNT(struct buf *bp)
364 {
365 int s, ret;
366
367 /*
368 * When the system is panicing, the lock manager grants all lock
369 * requests whether or not the lock is available. To avoid "unlocked
370 * buffer" panics after a crash, we just claim that all buffers
371 * are locked when cleaning up after a system panic.
372 */
373 if (panicstr != NULL)
374 return (1);
375 s = splbio();
376 ret = lockcount(&(bp)->b_lock);
377 splx(s);
378 return ret;
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 #define BUF_WRITE(bp) \
407 (bp)->b_op->bop_write(bp)
408
409 static __inline void
410 buf_start(struct buf *bp)
411 {
412 if (bioops.io_start)
413 (*bioops.io_start)(bp);
414 }
415
416 static __inline void
417 buf_complete(struct buf *bp)
418 {
419 if (bioops.io_complete)
420 (*bioops.io_complete)(bp);
421 }
422
423 static __inline void
424 buf_deallocate(struct buf *bp)
425 {
426 if (bioops.io_deallocate)
427 (*bioops.io_deallocate)(bp);
428 BUF_LOCKFREE(bp);
429 }
430
431 static __inline void
432 buf_movedeps(struct buf *bp, struct buf *bp2)
433 {
434 if (bioops.io_movedeps)
435 (*bioops.io_movedeps)(bp, bp2);
436 }
437
438 static __inline int
439 buf_countdeps(struct buf *bp, int i)
440 {
441 if (bioops.io_countdeps)
442 return ((*bioops.io_countdeps)(bp, i));
443 else
444 return (0);
445 }
446
447 #endif /* _KERNEL */
448
449 /*
450 * Zero out the buffer's data area.
451 */
452 #define clrbuf(bp) { \
453 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \
454 (bp)->b_resid = 0; \
455 }
456
457 /*
458 * Flags for getblk's last parameter.
459 */
460 #define GB_LOCK_NOWAIT 0x0001 /* Fail if we block on a buf lock. */
461 #define GB_NOCREAT 0x0002 /* Don't create a buf if not found. */
462
463 #ifdef _KERNEL
464 extern int nbuf; /* The number of buffer headers */
465 extern int maxswzone; /* Max KVA for swap structures */
466 extern int maxbcache; /* Max KVA for buffer cache */
467 extern int runningbufspace;
468 extern int buf_maxio; /* nominal maximum I/O for buffer */
469 extern struct buf *buf; /* The buffer headers. */
470 extern char *buffers; /* The buffer contents. */
471 extern int bufpages; /* Number of memory pages in the buffer pool. */
472 extern struct buf *swbuf; /* Swap I/O buffer headers. */
473 extern int nswbuf; /* Number of swap I/O buffer headers. */
474
475 struct uio;
476
477 caddr_t kern_vfs_bio_buffer_alloc(caddr_t v, long physmem_est);
478 void bufinit(void);
479 void bwillwrite(void);
480 int buf_dirty_count_severe(void);
481 void bremfree(struct buf *);
482 int bread(struct vnode *, daddr_t, int, struct ucred *, struct buf **);
483 int breadn(struct vnode *, daddr_t, int, daddr_t *, int *, int,
484 struct ucred *, struct buf **);
485 int bwrite(struct buf *);
486 void bdwrite(struct buf *);
487 void bawrite(struct buf *);
488 void bdirty(struct buf *);
489 void bundirty(struct buf *);
490 void brelse(struct buf *);
491 void bqrelse(struct buf *);
492 int vfs_bio_awrite(struct buf *);
493 struct buf * getpbuf(int *);
494 struct buf *incore(struct vnode *, daddr_t);
495 struct buf *gbincore(struct vnode *, daddr_t);
496 int inmem(struct vnode *, daddr_t);
497 struct buf *getblk(struct vnode *, daddr_t, int, int, int, int);
498 struct buf *geteblk(int);
499 int bufwait(struct buf *);
500 void bufdone(struct buf *);
501
502 void cluster_callback(struct buf *);
503 int cluster_read(struct vnode *, u_quad_t, daddr_t, long,
504 struct ucred *, long, int, struct buf **);
505 int cluster_wbuild(struct vnode *, long, daddr_t, int);
506 void cluster_write(struct buf *, u_quad_t, int);
507 void vfs_bio_set_validclean(struct buf *, int base, int size);
508 void vfs_bio_clrbuf(struct buf *);
509 void vfs_busy_pages(struct buf *, int clear_modify);
510 void vfs_unbusy_pages(struct buf *);
511 void vwakeup(struct buf *);
512 int vmapbuf(struct buf *);
513 void vunmapbuf(struct buf *);
514 void relpbuf(struct buf *, int *);
515 void brelvp(struct buf *);
516 void bgetvp(struct vnode *, struct buf *);
517 void pbgetvp(struct vnode *, struct buf *);
518 void pbrelvp(struct buf *);
519 int allocbuf(struct buf *bp, int size);
520 void reassignbuf(struct buf *, struct vnode *);
521 struct buf *trypbuf(int *);
522 void bwait(struct buf *, u_char, const char *);
523 void bdone(struct buf *);
524
525 #endif /* _KERNEL */
526
527 #endif /* !_SYS_BUF_H_ */
Cache object: 1eed083aad041f1e78d0a8197274219b
|