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/5.1/sys/sys/buf.h 112183 2003-03-13 07:31:45Z jeff $
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 /* XXX: b_io must be the first element of struct buf for now /phk */
106 /* XXX: if you change this, fix BIOTOBUF macro below */
107 struct bio b_io; /* "Builtin" I/O request. */
108 #define BIOTOBUF(biop) ((struct buf *)(biop))
109 #define b_bcount b_io.bio_bcount
110 #define b_blkno b_io.bio_blkno
111 #define b_caller1 b_io.bio_caller1
112 #define b_data b_io.bio_data
113 #define b_dev b_io.bio_dev
114 #define b_driver1 b_io.bio_driver1
115 #define b_driver2 b_io.bio_driver2
116 #define b_error b_io.bio_error
117 #define b_iocmd b_io.bio_cmd
118 #define b_ioflags b_io.bio_flags
119 #define b_pblkno b_io.bio_pblkno
120 #define b_resid b_io.bio_resid
121 struct buf_ops *b_op;
122 unsigned b_magic;
123 #define B_MAGIC_BIO 0x10b10b10
124 #define B_MAGIC_NFS 0x67238234
125 void (*b_iodone)(struct buf *);
126 off_t b_offset; /* Offset into file. */
127 TAILQ_ENTRY(buf) b_vnbufs; /* (V) Buffer's associated vnode. */
128 struct buf *b_left; /* (V) splay tree link */
129 struct buf *b_right; /* (V) splay tree link */
130 uint32_t b_vflags; /* (V) BV_* flags */
131 TAILQ_ENTRY(buf) b_freelist; /* (Q) Free list position inactive. */
132 unsigned short b_qindex; /* (Q) buffer queue index */
133 uint32_t b_flags; /* B_* flags. */
134 b_xflags_t b_xflags; /* extra flags */
135 struct lock b_lock; /* Buffer lock */
136 long b_bufsize; /* Allocated buffer size. */
137 long b_runningbufspace; /* when I/O is running, pipelining */
138 caddr_t b_kvabase; /* base kva for buffer */
139 int b_kvasize; /* size of kva for buffer */
140 daddr_t b_lblkno; /* Logical block number. */
141 struct vnode *b_vp; /* Device vnode. */
142 struct vm_object *b_object; /* Object for vp */
143 int b_dirtyoff; /* Offset in buffer of dirty region. */
144 int b_dirtyend; /* Offset of end of dirty region. */
145 struct ucred *b_rcred; /* Read credentials reference. */
146 struct ucred *b_wcred; /* Write credentials reference. */
147 void *b_saveaddr; /* Original b_addr for physio. */
148 union pager_info {
149 void *pg_spc;
150 int pg_reqpage;
151 } b_pager;
152 union cluster_info {
153 TAILQ_HEAD(cluster_list_head, buf) cluster_head;
154 TAILQ_ENTRY(buf) cluster_entry;
155 } b_cluster;
156 struct vm_page *b_pages[btoc(MAXPHYS)];
157 int b_npages;
158 struct workhead b_dep; /* (D) List of filesystem dependencies. */
159 };
160
161 #define b_spc b_pager.pg_spc
162
163 /*
164 * These flags are kept in b_flags.
165 *
166 * Notes:
167 *
168 * B_ASYNC VOP calls on bp's are usually async whether or not
169 * B_ASYNC is set, but some subsystems, such as NFS, like
170 * to know what is best for the caller so they can
171 * optimize the I/O.
172 *
173 * B_PAGING Indicates that bp is being used by the paging system or
174 * some paging system and that the bp is not linked into
175 * the b_vp's clean/dirty linked lists or ref counts.
176 * Buffer vp reassignments are illegal in this case.
177 *
178 * B_CACHE This may only be set if the buffer is entirely valid.
179 * The situation where B_DELWRI is set and B_CACHE is
180 * clear MUST be committed to disk by getblk() so
181 * B_DELWRI can also be cleared. See the comments for
182 * getblk() in kern/vfs_bio.c. If B_CACHE is clear,
183 * the caller is expected to clear BIO_ERROR and B_INVAL,
184 * set BIO_READ, and initiate an I/O.
185 *
186 * The 'entire buffer' is defined to be the range from
187 * 0 through b_bcount.
188 *
189 * B_MALLOC Request that the buffer be allocated from the malloc
190 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
191 *
192 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers
193 * by filesystems that allow clustering when the buffer
194 * is fully dirty and indicates that it may be clustered
195 * with other adjacent dirty buffers. Note the clustering
196 * may not be used with the stage 1 data write under NFS
197 * but may be used for the commit rpc portion.
198 *
199 * B_VMIO Indicates that the buffer is tied into an VM object.
200 * The buffer's data is always PAGE_SIZE aligned even
201 * if b_bufsize and b_bcount are not. ( b_bufsize is
202 * always at least DEV_BSIZE aligned, though ).
203 *
204 * B_DIRECT Hint that we should attempt to completely free
205 * the pages underlying the buffer. B_DIRECT is
206 * sticky until the buffer is released and typically
207 * only has an effect when B_RELBUF is also set.
208 *
209 * B_NOWDRAIN This flag should be set when a device (like MD)
210 * does a turn-around VOP_WRITE from its strategy
211 * routine. This flag prevents bwrite() from blocking
212 * in wdrain, avoiding a deadlock situation.
213 */
214
215 #define B_AGE 0x00000001 /* Move to age queue when I/O done. */
216 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */
217 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */
218 #define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */
219 #define B_DEFERRED 0x00000010 /* Skipped over for cleaning */
220 #define B_CACHE 0x00000020 /* Bread found us in the cache. */
221 #define B_VALIDSUSPWRT 0x00000040 /* Valid write during suspension. */
222 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */
223 #define B_00000100 0x00000100 /* Available flag. */
224 #define B_DONE 0x00000200 /* I/O completed. */
225 #define B_EINTR 0x00000400 /* I/O was interrupted */
226 #define B_NOWDRAIN 0x00000800 /* Avoid wdrain deadlock */
227 #define B_00001000 0x00001000 /* Available flag. */
228 #define B_INVAL 0x00002000 /* Does not contain valid info. */
229 #define B_LOCKED 0x00004000 /* Locked in core (not reusable). */
230 #define B_NOCACHE 0x00008000 /* Do not cache block after use. */
231 #define B_MALLOC 0x00010000 /* malloced b_data */
232 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */
233 #define B_PHYS 0x00040000 /* I/O to user memory. */
234 #define B_00080000 0x00080000 /* Available flag. */
235 #define B_00100000 0x00100000 /* Available flag. */
236 #define B_DIRTY 0x00200000 /* Needs writing later (in EXT2FS). */
237 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */
238 #define B_00800000 0x00800000 /* Available flag. */
239 #define B_WRITEINPROG 0x01000000 /* Write in progress. */
240 #define B_02000000 0x02000000 /* Available flag. */
241 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */
242 #define B_08000000 0x08000000 /* Available flag. */
243 #define B_RAM 0x10000000 /* Read ahead mark (flag) */
244 #define B_VMIO 0x20000000 /* VMIO flag */
245 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */
246 #define B_80000000 0x80000000 /* Available flag. */
247
248 #define PRINT_BUF_FLAGS "\2\40b31\37cluster\36vmio\35ram\34b27" \
249 "\33paging\32b25\31writeinprog\30b23\27relbuf\26dirty\25b20" \
250 "\24b19\23phys\22clusterok\21malloc\20nocache\17locked\16inval" \
251 "\15scanned\14nowdrain\13eintr\12done\11b8\10delwri\7validsuspwrt" \
252 "\6cache\5deferred\4direct\3async\2needcommit\1age"
253
254 /*
255 * These flags are kept in b_xflags.
256 */
257 #define BX_VNDIRTY 0x00000001 /* On vnode dirty list */
258 #define BX_VNCLEAN 0x00000002 /* On vnode clean list */
259 #define BX_BKGRDWRITE 0x00000004 /* Do writes in background */
260 #define BX_BKGRDINPROG 0x00000008 /* Background write in progress */
261 #define BX_BKGRDWAIT 0x00000010 /* Background write waiting */
262 #define BX_BKGRDMARKER 0x00000020 /* Mark buffer for splay tree */
263 #define BX_ALTDATA 0x00000040 /* Holds extended data */
264
265 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */
266
267 #define BV_SCANNED 0x00001000 /* VOP_FSYNC funcs mark written bufs */
268
269 #ifdef _KERNEL
270 /*
271 * Buffer locking
272 */
273 extern const char *buf_wmesg; /* Default buffer lock message */
274 #define BUF_WMESG "bufwait"
275 #include <sys/proc.h> /* XXX for curthread */
276 #include <sys/mutex.h>
277
278 /*
279 * Initialize a lock.
280 */
281 #define BUF_LOCKINIT(bp) \
282 lockinit(&(bp)->b_lock, PRIBIO + 4, buf_wmesg, 0, 0)
283 /*
284 *
285 * Get a lock sleeping non-interruptably until it becomes available.
286 */
287 static __inline int BUF_LOCK(struct buf *, int, struct mtx *);
288 static __inline int
289 BUF_LOCK(struct buf *bp, int locktype, struct mtx *interlock)
290 {
291 int s, ret;
292
293 s = splbio();
294 mtx_lock(bp->b_lock.lk_interlock);
295 locktype |= LK_INTERNAL;
296 bp->b_lock.lk_wmesg = buf_wmesg;
297 bp->b_lock.lk_prio = PRIBIO + 4;
298 ret = lockmgr(&(bp)->b_lock, locktype, interlock, curthread);
299 splx(s);
300 return ret;
301 }
302 /*
303 * Get a lock sleeping with specified interruptably and timeout.
304 */
305 static __inline int BUF_TIMELOCK(struct buf *, int, struct mtx *,
306 char *, int, int);
307 static __inline int
308 BUF_TIMELOCK(struct buf *bp, int locktype, struct mtx *interlock,
309 char *wmesg, int catch, int timo)
310 {
311 int s, ret;
312
313 s = splbio();
314 mtx_lock(bp->b_lock.lk_interlock);
315 locktype |= LK_INTERNAL | LK_TIMELOCK;
316 bp->b_lock.lk_wmesg = wmesg;
317 bp->b_lock.lk_prio = (PRIBIO + 4) | catch;
318 bp->b_lock.lk_timo = timo;
319 ret = lockmgr(&(bp)->b_lock, (locktype), interlock, curthread);
320 splx(s);
321 return ret;
322 }
323 /*
324 * Release a lock. Only the acquiring process may free the lock unless
325 * it has been handed off to biodone.
326 */
327 static __inline void BUF_UNLOCK(struct buf *);
328 static __inline void
329 BUF_UNLOCK(struct buf *bp)
330 {
331 int s;
332
333 s = splbio();
334 lockmgr(&(bp)->b_lock, LK_RELEASE, NULL, curthread);
335 splx(s);
336 }
337
338 /*
339 * Free a buffer lock.
340 */
341 #define BUF_LOCKFREE(bp) \
342 do { \
343 if (BUF_REFCNT(bp) > 0) \
344 panic("free locked buf"); \
345 lockdestroy(&(bp)->b_lock); \
346 } while (0)
347
348 #ifdef _SYS_PROC_H_ /* Avoid #include <sys/proc.h> pollution */
349 /*
350 * When initiating asynchronous I/O, change ownership of the lock to the
351 * kernel. Once done, the lock may legally released by biodone. The
352 * original owning process can no longer acquire it recursively, but must
353 * wait until the I/O is completed and the lock has been freed by biodone.
354 */
355 static __inline void BUF_KERNPROC(struct buf *);
356 static __inline void
357 BUF_KERNPROC(struct buf *bp)
358 {
359 struct thread *td = curthread;
360
361 if ((td != PCPU_GET(idlethread))
362 && bp->b_lock.lk_lockholder == td)
363 td->td_locks--;
364 bp->b_lock.lk_lockholder = LK_KERNPROC;
365 }
366 #endif
367 /*
368 * Find out the number of references to a lock.
369 */
370 static __inline int BUF_REFCNT(struct buf *);
371 static __inline int
372 BUF_REFCNT(struct buf *bp)
373 {
374 int s, ret;
375
376 /*
377 * When the system is panicing, the lock manager grants all lock
378 * requests whether or not the lock is available. To avoid "unlocked
379 * buffer" panics after a crash, we just claim that all buffers
380 * are locked when cleaning up after a system panic.
381 */
382 if (panicstr != NULL)
383 return (1);
384 s = splbio();
385 ret = lockcount(&(bp)->b_lock);
386 splx(s);
387 return ret;
388 }
389
390 #endif /* _KERNEL */
391
392 struct buf_queue_head {
393 TAILQ_HEAD(buf_queue, buf) queue;
394 daddr_t last_pblkno;
395 struct buf *insert_point;
396 struct buf *switch_point;
397 };
398
399 /*
400 * This structure describes a clustered I/O. It is stored in the b_saveaddr
401 * field of the buffer on which I/O is done. At I/O completion, cluster
402 * callback uses the structure to parcel I/O's to individual buffers, and
403 * then free's this structure.
404 */
405 struct cluster_save {
406 long bs_bcount; /* Saved b_bcount. */
407 long bs_bufsize; /* Saved b_bufsize. */
408 void *bs_saveaddr; /* Saved b_addr. */
409 int bs_nchildren; /* Number of associated buffers. */
410 struct buf **bs_children; /* List of associated buffers. */
411 };
412
413 #ifdef _KERNEL
414
415 #define BUF_WRITE(bp) \
416 (bp)->b_op->bop_write(bp)
417
418 static __inline void
419 buf_start(struct buf *bp)
420 {
421 if (bioops.io_start)
422 (*bioops.io_start)(bp);
423 }
424
425 static __inline void
426 buf_complete(struct buf *bp)
427 {
428 if (bioops.io_complete)
429 (*bioops.io_complete)(bp);
430 }
431
432 static __inline void
433 buf_deallocate(struct buf *bp)
434 {
435 if (bioops.io_deallocate)
436 (*bioops.io_deallocate)(bp);
437 BUF_LOCKFREE(bp);
438 }
439
440 static __inline void
441 buf_movedeps(struct buf *bp, struct buf *bp2)
442 {
443 if (bioops.io_movedeps)
444 (*bioops.io_movedeps)(bp, bp2);
445 }
446
447 static __inline int
448 buf_countdeps(struct buf *bp, int i)
449 {
450 if (bioops.io_countdeps)
451 return ((*bioops.io_countdeps)(bp, i));
452 else
453 return (0);
454 }
455
456 #endif /* _KERNEL */
457
458 /*
459 * Zero out the buffer's data area.
460 */
461 #define clrbuf(bp) { \
462 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \
463 (bp)->b_resid = 0; \
464 }
465
466 /*
467 * Flags for getblk's last parameter.
468 */
469 #define GB_LOCK_NOWAIT 0x0001 /* Fail if we block on a buf lock. */
470
471 #ifdef _KERNEL
472 extern int nbuf; /* The number of buffer headers */
473 extern int maxswzone; /* Max KVA for swap structures */
474 extern int maxbcache; /* Max KVA for buffer cache */
475 extern int runningbufspace;
476 extern int buf_maxio; /* nominal maximum I/O for buffer */
477 extern struct buf *buf; /* The buffer headers. */
478 extern char *buffers; /* The buffer contents. */
479 extern int bufpages; /* Number of memory pages in the buffer pool. */
480 extern struct buf *swbuf; /* Swap I/O buffer headers. */
481 extern int nswbuf; /* Number of swap I/O buffer headers. */
482
483 struct uio;
484
485 caddr_t kern_vfs_bio_buffer_alloc(caddr_t v, long physmem_est);
486 void bufinit(void);
487 void bwillwrite(void);
488 int buf_dirty_count_severe(void);
489 void bremfree(struct buf *);
490 int bread(struct vnode *, daddr_t, int, struct ucred *, struct buf **);
491 int breadn(struct vnode *, daddr_t, int, daddr_t *, int *, int,
492 struct ucred *, struct buf **);
493 int bwrite(struct buf *);
494 void bdwrite(struct buf *);
495 void bawrite(struct buf *);
496 void bdirty(struct buf *);
497 void bundirty(struct buf *);
498 void brelse(struct buf *);
499 void bqrelse(struct buf *);
500 int vfs_bio_awrite(struct buf *);
501 struct buf * getpbuf(int *);
502 struct buf *incore(struct vnode *, daddr_t);
503 struct buf *gbincore(struct vnode *, daddr_t);
504 int inmem(struct vnode *, daddr_t);
505 struct buf *getblk(struct vnode *, daddr_t, int, int, int, int);
506 struct buf *geteblk(int);
507 int bufwait(struct buf *);
508 void bufdone(struct buf *);
509 void bufdonebio(struct bio *);
510
511 void cluster_callback(struct buf *);
512 int cluster_read(struct vnode *, u_quad_t, daddr_t, long,
513 struct ucred *, long, int, struct buf **);
514 int cluster_wbuild(struct vnode *, long, daddr_t, int);
515 void cluster_write(struct buf *, u_quad_t, int);
516 void vfs_bio_set_validclean(struct buf *, int base, int size);
517 void vfs_bio_clrbuf(struct buf *);
518 void vfs_busy_pages(struct buf *, int clear_modify);
519 void vfs_unbusy_pages(struct buf *);
520 void vwakeup(struct buf *);
521 int vmapbuf(struct buf *);
522 void vunmapbuf(struct buf *);
523 void relpbuf(struct buf *, int *);
524 void brelvp(struct buf *);
525 void bgetvp(struct vnode *, struct buf *);
526 void pbgetvp(struct vnode *, struct buf *);
527 void pbrelvp(struct buf *);
528 int allocbuf(struct buf *bp, int size);
529 void reassignbuf(struct buf *, struct vnode *);
530 struct buf *trypbuf(int *);
531 void bwait(struct buf *, u_char, const char *);
532 void bdone(struct buf *);
533
534 #endif /* _KERNEL */
535
536 #endif /* !_SYS_BUF_H_ */
Cache object: 539ef5bd601cc8ddd9ecfe5d0ae3f60b
|