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
39 * $FreeBSD: src/sys/sys/buf.h,v 1.88.2.10 2003/01/25 19:02:23 dillon Exp $
40 * $DragonFly: src/sys/sys/buf.h,v 1.54 2008/08/29 20:08:37 dillon Exp $
46 #if defined(_KERNEL) || defined(_KERNEL_STRUCTURES)
49 #include <sys/queue.h>
54 #ifndef _SYS_DEVICE_H_
55 #include <sys/device.h>
67 #ifndef _SYS_SPINLOCK_H_
68 #include <sys/spinlock.h>
81 RB_PROTOTYPE2(buf_rb_tree, buf, b_rbnode, rb_buf_compare, off_t);
82 RB_PROTOTYPE2(buf_rb_hash, buf, b_rbhash, rb_buf_compare, off_t);
85 * To avoid including <ufs/ffs/softdep.h>
87 LIST_HEAD(workhead, worklist);
91 typedef enum buf_cmd {
100 #if defined(_KERNEL) || defined(_KERNEL_STRUCTURES)
103 * The buffer header describes an I/O operation in the kernel.
106 * b_bufsize represents the filesystem block size (for this particular
107 * block) and/or the allocation size or original request size. This
108 * field is NOT USED by lower device layers. VNode and device
109 * strategy routines WILL NEVER ACCESS THIS FIELD.
111 * b_bcount represents the I/O request size. Unless B_NOBCLIP is set,
112 * the device chain is allowed to clip b_bcount to accomodate the device
113 * EOF. Note that this is different from the byte oriented file EOF.
114 * If B_NOBCLIP is set, the device chain is required to generate an
115 * error if it would othrewise have to clip the request. Buffers
116 * obtained via getblk() automatically set B_NOBCLIP. It is important
117 * to note that EOF clipping via b_bcount is different from EOF clipping
118 * via returning a b_actual < b_bcount. B_NOBCLIP only effects block
119 * oriented EOF clipping (b_bcount modifications).
121 * b_actual represents the number of bytes of I/O that actually occured,
122 * whether an error occured or not. b_actual must be initialized to 0
123 * prior to initiating I/O as the device drivers will assume it to
126 * b_dirtyoff, b_dirtyend. Buffers support piecemeal, unaligned
127 * ranges of dirty data that need to be written to backing store.
128 * The range is typically clipped at b_bcount (not b_bufsize).
130 * b_bio1 and b_bio2 represent the two primary I/O layers. Additional
131 * I/O layers are allocated out of the object cache and may also exist.
133 * b_bio1 is the logical layer and contains offset or block number
134 * data for the primary vnode, b_vp. I/O operations are almost
135 * universally initiated from the logical layer, so you will often
136 * see things like: vn_strategy(bp->b_vp, &bp->b_bio1).
138 * b_bio2 is the first physical layer (typically the slice-relative
139 * layer) and contains the translated offset or block number for
140 * the block device underlying a filesystem. Filesystems such as UFS
141 * will maintain cached translations and you may see them initiate
142 * a 'physical' I/O using vn_strategy(devvp, &bp->b_bio2). BUT,
143 * remember that the layering is relative to bp->b_vp, so the
144 * device-relative block numbers for buffer cache operations that occur
145 * directly on a block device will be in the first BIO layer.
147 * b_ops - initialized if a buffer has a bio_ops
149 * NOTE!!! Only the BIO subsystem accesses b_bio1 and b_bio2 directly.
150 * ALL STRATEGY LAYERS FOR BOTH VNODES AND DEVICES ONLY ACCESS THE BIO
151 * PASSED TO THEM, AND WILL PUSH ANOTHER BIO LAYER IF FORWARDING THE
152 * I/O DEEPER. In particular, a vn_strategy() or dev_dstrategy()
153 * call should not ever access buf->b_vp as this vnode may be totally
154 * unrelated to the vnode/device whos strategy routine was called.
157 RB_ENTRY(buf) b_rbnode; /* RB node in vnode clean/dirty tree */
158 RB_ENTRY(buf) b_rbhash; /* RB node in vnode hash tree */
159 TAILQ_ENTRY(buf) b_freelist; /* Free list position if not active. */
160 struct buf *b_cluster_next; /* Next buffer (cluster code) */
161 struct vnode *b_vp; /* (vp, loffset) index */
162 struct bio b_bio_array[NBUF_BIO]; /* BIO translation layers */
163 u_int32_t b_flags; /* B_* flags. */
164 unsigned short b_qindex; /* buffer queue index */
165 unsigned char b_act_count; /* similar to vm_page act_count */
166 unsigned char b_unused01;
167 struct lock b_lock; /* Buffer lock */
168 void *b_iosched; /* I/O scheduler priv data */
169 buf_cmd_t b_cmd; /* I/O command */
170 int b_bufsize; /* Allocated buffer size. */
171 int b_runningbufspace; /* when I/O is running, pipelining */
172 int b_bcount; /* Valid bytes in buffer. */
173 int b_resid; /* Remaining I/O */
174 int b_error; /* Error return */
175 caddr_t b_data; /* Memory, superblocks, indirect etc. */
176 caddr_t b_kvabase; /* base kva for buffer */
177 int b_kvasize; /* size of kva for buffer */
178 int b_dirtyoff; /* Offset in buffer of dirty region. */
179 int b_dirtyend; /* Offset of end of dirty region. */
180 int b_refs; /* FINDBLK_REF/bqhold()/bqdrop() */
181 struct xio b_xio; /* data buffer page list management */
182 struct bio_ops *b_ops; /* bio_ops used w/ b_dep */
183 struct workhead b_dep; /* List of filesystem dependencies. */
189 #define b_bio1 b_bio_array[0] /* logical layer */
190 #define b_bio2 b_bio_array[1] /* (typically) the disk layer */
191 #define b_loffset b_bio1.bio_offset
195 * Flags passed to getblk()
197 * GETBLK_PCATCH - Allow signals to be caught. getblk() is allowed to return
198 * NULL if this flag is passed.
200 * GETBLK_BHEAVY - This is a heavy weight buffer, meaning that resolving
201 * writes can require additional buffers.
203 * GETBLK_SZMATCH- blksize must match pre-existing b_bcount. getblk() can
206 * GETBLK_NOWAIT - Do not use a blocking lock. getblk() can return NULL.
208 #define GETBLK_PCATCH 0x0001 /* catch signals */
209 #define GETBLK_BHEAVY 0x0002 /* heavy weight buffer */
210 #define GETBLK_SZMATCH 0x0004 /* pre-existing buffer must match */
211 #define GETBLK_NOWAIT 0x0008 /* non-blocking */
213 #define FINDBLK_TEST 0x0010 /* test only, do not lock */
214 #define FINDBLK_NBLOCK 0x0020 /* use non-blocking lock, can return NULL */
215 #define FINDBLK_REF 0x0040 /* ref the buf to prevent reuse */
218 * These flags are kept in b_flags.
222 * B_PAGING Indicates that bp is being used by the paging system or
223 * some paging system and that the bp is not linked into
224 * the b_vp's clean/dirty linked lists or ref counts.
225 * Buffer vp reassignments are illegal in this case.
227 * B_CACHE This may only be set if the buffer is entirely valid.
228 * The situation where B_DELWRI is set and B_CACHE is
229 * clear MUST be committed to disk by getblk() so
230 * B_DELWRI can also be cleared. See the comments for
231 * getblk() in kern/vfs_bio.c. If B_CACHE is clear,
232 * the caller is expected to clear B_ERROR|B_INVAL,
233 * set BUF_CMD_READ, and initiate an I/O.
235 * The 'entire buffer' is defined to be the range from
236 * 0 through b_bcount.
238 * B_MALLOC Request that the buffer be allocated from the malloc
239 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
241 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers
242 * by filesystems that allow clustering when the buffer
243 * is fully dirty and indicates that it may be clustered
244 * with other adjacent dirty buffers. Note the clustering
245 * may not be used with the stage 1 data write under NFS
246 * but may be used for the commit rpc portion.
248 * B_VMIO Indicates that the buffer is tied into an VM object.
249 * The buffer's data is always PAGE_SIZE aligned even
250 * if b_bufsize and b_bcount are not. ( b_bufsize is
251 * always at least DEV_BSIZE aligned, though ).
253 * B_DIRECT Hint that we should attempt to completely free
254 * the pages underlying the buffer. B_DIRECT is
255 * sticky until the buffer is released and typically
256 * only has an effect when B_RELBUF is also set.
258 * B_LOCKED The buffer will be released to the locked queue
259 * regardless of its current state. Note that
260 * if B_DELWRI is set, no I/O occurs until the caller
261 * acquires the buffer, clears B_LOCKED, then releases
264 * B_AGE When allocating a new buffer any buffer encountered
265 * with B_AGE set will be reallocated more quickly then
266 * buffers encountered without it set. B_AGE is set
267 * as part of the loop so idle buffers should eventually
268 * wind up with B_AGE set. B_AGE explicitly does NOT
269 * cause the buffer to be instantly reallocated for
272 * Standard buffer flushing routines leave B_AGE intact
273 * through the DIRTY queue and into the CLEAN queue.
274 * Setting B_AGE on a dirty buffer will not cause it
275 * to be flushed more quickly but will cause it to be
276 * reallocated more quickly after having been flushed.
278 * B_NOCACHE Request that the buffer and backing store be
279 * destroyed on completion. If B_DELWRI is set and the
280 * write fails, the buffer remains intact.
282 * B_NOTMETA May be set on block device buffers representing
283 * file data (i.e. which aren't really meta-data),
284 * which will cause the buffer cache to set PG_NOTMETA
285 * in the VM pages when releasing them and the
286 * swapcache to not try to cache them.
288 * B_MARKER Special marker buf, always skip.
291 #define B_AGE 0x00000001 /* Reuse more quickly */
292 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */
293 #define B_NOTMETA 0x00000004 /* This really isn't metadata */
294 #define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */
295 #define B_DEFERRED 0x00000010 /* vfs-controlled deferment */
296 #define B_CACHE 0x00000020 /* Bread found us in the cache. */
297 #define B_HASHED 0x00000040 /* Indexed via v_rbhash_tree */
298 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */
299 #define B_BNOCLIP 0x00000100 /* EOF clipping b_bcount not allowed */
300 #define B_HASBOGUS 0x00000200 /* Contains bogus pages */
301 #define B_EINTR 0x00000400 /* I/O was interrupted */
302 #define B_ERROR 0x00000800 /* I/O error occurred. */
303 #define B_IODEBUG 0x00001000 /* (Debugging only bread) */
304 #define B_INVAL 0x00002000 /* Does not contain valid info. */
305 #define B_LOCKED 0x00004000 /* Locked in core (not reusable). */
306 #define B_NOCACHE 0x00008000 /* Destroy buffer AND backing store */
307 #define B_MALLOC 0x00010000 /* malloced b_data */
308 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */
309 #define B_MARKER 0x00040000 /* Special marker buf in queue */
310 #define B_RAW 0x00080000 /* Set by physio for raw transfers. */
311 #define B_HEAVY 0x00100000 /* Heavy-weight buffer */
312 #define B_DIRTY 0x00200000 /* Needs writing later. */
313 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */
314 #define B_UNUSED23 0x00800000 /* Request wakeup on done */
315 #define B_VNCLEAN 0x01000000 /* On vnode clean list */
316 #define B_VNDIRTY 0x02000000 /* On vnode dirty list */
317 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */
318 #define B_ORDERED 0x08000000 /* Must guarantee I/O ordering */
319 #define B_RAM 0x10000000 /* Read ahead mark (flag) */
320 #define B_VMIO 0x20000000 /* VMIO flag */
321 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */
322 #define B_VFSFLAG1 0x80000000 /* VFSs can set this flag */
324 #define PRINT_BUF_FLAGS "\20" \
325 "\40unused31\37cluster\36vmio\35ram\34ordered" \
326 "\33paging\32vndirty\31vnclean\30unused23\27relbuf\26dirty" \
327 "\25unused20\24raw\23unused18\22clusterok\21malloc\20nocache" \
328 "\17locked\16inval\15unused12\14error\13eintr\12unused9\11bnoclip" \
329 "\10delwri\7hashed\6cache\5deferred\4direct\3unused2\2needcommit\1age"
331 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */
335 * Buffer locking. See sys/buf2.h for inline functions.
337 extern char *buf_wmesg; /* Default buffer lock message */
338 #define BUF_WMESG "bufwait"
342 struct bio_queue_head {
343 TAILQ_HEAD(bio_queue, bio) queue;
346 struct bio *transition;
347 struct bio *bio_unused;
351 * This structure describes a clustered I/O.
353 struct cluster_save {
354 int bs_nchildren; /* Number of associated buffers. */
355 struct buf **bs_children; /* List of associated buffers. */
359 * Zero out the buffer's data area.
361 #define clrbuf(bp) { \
362 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \
367 * Flags to low-level bitmap allocation routines (balloc).
369 * Note: sequential_heuristic() in kern/vfs_vnops.c limits the count
372 #define B_SEQMASK 0x7F000000 /* Sequential heuristic mask. */
373 #define B_SEQSHIFT 24 /* Sequential heuristic shift. */
374 #define B_SEQMAX 0x7F
375 #define B_CLRBUF 0x01 /* Cleared invalid areas of buffer. */
376 #define B_SYNC 0x02 /* Do all allocations synchronously. */
379 extern long nbuf; /* The number of buffer headers */
380 extern long maxswzone; /* Max KVA for swap structures */
381 extern long maxbcache; /* Max KVA for buffer cache */
382 extern long hidirtybufspace;
383 extern int buf_maxio; /* nominal maximum I/O for buffer */
384 extern struct buf *buf; /* The buffer headers. */
385 extern char *buffers; /* The buffer contents. */
386 extern int bufpages; /* Number of memory pages in the buffer pool. */
387 extern struct buf *swbuf; /* Swap I/O buffer headers. */
388 extern long nswbuf; /* Number of swap I/O buffer headers. */
389 extern int bioq_reorder_burst_interval;
390 extern int bioq_reorder_burst_bytes;
391 extern int bioq_reorder_minor_interval;
392 extern int bioq_reorder_minor_bytes;
398 long bd_heatup (void);
399 void bd_wait (long count);
400 void waitrunningbufspace(void);
401 int buf_dirty_count_severe (void);
402 int buf_runningbufspace_severe (void);
403 void initbufbio(struct buf *);
404 void uninitbufbio(struct buf *);
405 void reinitbufbio(struct buf *);
406 void clearbiocache(struct bio *);
407 void bremfree (struct buf *);
408 int breadx (struct vnode *, off_t, int, struct buf **);
409 int breadnx (struct vnode *, off_t, int, off_t *, int *, int,
411 void breadcb(struct vnode *, off_t, int,
412 void (*)(struct bio *), void *);
413 int bwrite (struct buf *);
414 void bdwrite (struct buf *);
415 void buwrite (struct buf *);
416 void bawrite (struct buf *);
417 void bdirty (struct buf *);
418 void bheavy (struct buf *);
419 void bundirty (struct buf *);
420 int bowrite (struct buf *);
421 void brelse (struct buf *);
422 void bqrelse (struct buf *);
423 int cluster_awrite (struct buf *);
424 struct buf *getpbuf (int *);
425 struct buf *getpbuf_kva (int *);
426 int inmem (struct vnode *, off_t);
427 struct buf *findblk (struct vnode *, off_t, int);
428 struct buf *getblk (struct vnode *, off_t, int, int, int);
429 struct buf *getcacheblk (struct vnode *, off_t, int, int);
430 struct buf *geteblk (int);
431 struct buf *getnewbuf(int, int, int, int);
432 void bqhold(struct buf *bp);
433 void bqdrop(struct buf *bp);
434 void regetblk(struct buf *bp);
435 struct bio *push_bio(struct bio *);
436 struct bio *pop_bio(struct bio *);
437 int biowait (struct bio *, const char *);
438 int biowait_timeout (struct bio *, const char *, int);
439 void bpdone (struct buf *, int);
440 void biodone (struct bio *);
441 void biodone_sync (struct bio *);
442 void pbuf_adjcount(int *pfreecnt, int n);
444 void cluster_append(struct bio *, struct buf *);
445 int cluster_readx (struct vnode *, off_t, off_t, int,
446 size_t, size_t, struct buf **);
447 void cluster_write (struct buf *, off_t, int, int);
448 int physread (struct dev_read_args *);
449 int physwrite (struct dev_write_args *);
450 void vfs_bio_clrbuf (struct buf *);
451 void vfs_busy_pages (struct vnode *, struct buf *);
452 void vfs_unbusy_pages (struct buf *);
453 int vmapbuf (struct buf *, caddr_t, int);
454 void vunmapbuf (struct buf *);
455 void relpbuf (struct buf *, int *);
456 void brelvp (struct buf *);
457 int bgetvp (struct vnode *, struct buf *, int);
458 void bsetrunningbufspace(struct buf *, int);
459 int allocbuf (struct buf *bp, int size);
460 int scan_all_buffers (int (*)(struct buf *, void *), void *);
461 void reassignbuf (struct buf *);
462 struct buf *trypbuf (int *);
463 struct buf *trypbuf_kva (int *);
464 void bio_ops_sync(struct mount *mp);
465 void vm_hold_free_pages(struct buf *bp, vm_offset_t from, vm_offset_t to);
466 void vm_hold_load_pages(struct buf *bp, vm_offset_t from, vm_offset_t to);
467 void nestiobuf_done(struct bio *mbio, int donebytes, int error, struct devstat *stats);
468 void nestiobuf_init(struct bio *mbio);
469 void nestiobuf_add(struct bio *mbio, struct buf *bp, int off, size_t size, struct devstat *stats);
470 void nestiobuf_start(struct bio *mbio);
471 void nestiobuf_error(struct bio *mbio, int error);
473 #endif /* _KERNEL || _KERNEL_STRUCTURES */
474 #endif /* !_SYS_BUF_H_ */