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 bufobj 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 caddr_t b_kvaalloc; /* allocated kva for B_KVAALLOC */
121 int b_kvasize; /* size of kva for buffer */
122 daddr_t b_lblkno; /* Logical block number. */
123 struct vnode *b_vp; /* Device vnode. */
124 int b_dirtyoff; /* Offset in buffer of dirty region. */
125 int b_dirtyend; /* Offset of end of dirty region. */
126 struct ucred *b_rcred; /* Read credentials reference. */
127 struct ucred *b_wcred; /* Write credentials reference. */
128 void *b_saveaddr; /* Original b_addr for physio. */
133 TAILQ_HEAD(cluster_list_head, buf) cluster_head;
134 TAILQ_ENTRY(buf) cluster_entry;
136 struct vm_page *b_pages[btoc(MAXPHYS)];
138 struct workhead b_dep; /* (D) List of filesystem dependencies. */
145 #define b_object b_bufobj->bo_object
148 * These flags are kept in b_flags.
152 * B_ASYNC VOP calls on bp's are usually async whether or not
153 * B_ASYNC is set, but some subsystems, such as NFS, like
154 * to know what is best for the caller so they can
157 * B_PAGING Indicates that bp is being used by the paging system or
158 * some paging system and that the bp is not linked into
159 * the b_vp's clean/dirty linked lists or ref counts.
160 * Buffer vp reassignments are illegal in this case.
162 * B_CACHE This may only be set if the buffer is entirely valid.
163 * The situation where B_DELWRI is set and B_CACHE is
164 * clear MUST be committed to disk by getblk() so
165 * B_DELWRI can also be cleared. See the comments for
166 * getblk() in kern/vfs_bio.c. If B_CACHE is clear,
167 * the caller is expected to clear BIO_ERROR and B_INVAL,
168 * set BIO_READ, and initiate an I/O.
170 * The 'entire buffer' is defined to be the range from
171 * 0 through b_bcount.
173 * B_MALLOC Request that the buffer be allocated from the malloc
174 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
176 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers
177 * by filesystems that allow clustering when the buffer
178 * is fully dirty and indicates that it may be clustered
179 * with other adjacent dirty buffers. Note the clustering
180 * may not be used with the stage 1 data write under NFS
181 * but may be used for the commit rpc portion.
183 * B_VMIO Indicates that the buffer is tied into an VM object.
184 * The buffer's data is always PAGE_SIZE aligned even
185 * if b_bufsize and b_bcount are not. ( b_bufsize is
186 * always at least DEV_BSIZE aligned, though ).
188 * B_DIRECT Hint that we should attempt to completely free
189 * the pages underlying the buffer. B_DIRECT is
190 * sticky until the buffer is released and typically
191 * only has an effect when B_RELBUF is also set.
195 #define B_AGE 0x00000001 /* Move to age queue when I/O done. */
196 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */
197 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */
198 #define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */
199 #define B_DEFERRED 0x00000010 /* Skipped over for cleaning */
200 #define B_CACHE 0x00000020 /* Bread found us in the cache. */
201 #define B_VALIDSUSPWRT 0x00000040 /* Valid write during suspension. */
202 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */
203 #define B_PERSISTENT 0x00000100 /* Perm. ref'ed while EXT2FS mounted. */
204 #define B_DONE 0x00000200 /* I/O completed. */
205 #define B_EINTR 0x00000400 /* I/O was interrupted */
206 #define B_UNMAPPED 0x00000800 /* KVA is not mapped. */
207 #define B_KVAALLOC 0x00001000 /* But allocated. */
208 #define B_INVAL 0x00002000 /* Does not contain valid info. */
209 #define B_BARRIER 0x00004000 /* Write this and all preceeding first. */
210 #define B_NOCACHE 0x00008000 /* Do not cache block after use. */
211 #define B_MALLOC 0x00010000 /* malloced b_data */
212 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */
213 #define B_000400000 0x00040000 /* Available flag. */
214 #define B_000800000 0x00080000 /* Available flag. */
215 #define B_00100000 0x00100000 /* Available flag. */
216 #define B_DIRTY 0x00200000 /* Needs writing later (in EXT2FS). */
217 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */
218 #define B_00800000 0x00800000 /* Available flag. */
219 #define B_NOCOPY 0x01000000 /* Don't copy-on-write this buf. */
220 #define B_NEEDSGIANT 0x02000000 /* Buffer's vnode needs giant. */
221 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */
222 #define B_MANAGED 0x08000000 /* Managed by FS. */
223 #define B_RAM 0x10000000 /* Read ahead mark (flag) */
224 #define B_VMIO 0x20000000 /* VMIO flag */
225 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */
226 #define B_REMFREE 0x80000000 /* Delayed bremfree */
228 #define PRINT_BUF_FLAGS "\20\40remfree\37cluster\36vmio\35ram\34managed" \
229 "\33paging\32needsgiant\31nocopy\30b23\27relbuf\26dirty\25b20" \
230 "\24b19\23b18\22clusterok\21malloc\20nocache\17b14\16inval" \
231 "\15b12\14b11\13eintr\12done\11persist\10delwri\7validsuspwrt" \
232 "\6cache\5deferred\4direct\3async\2needcommit\1age"
235 * These flags are kept in b_xflags.
237 #define BX_VNDIRTY 0x00000001 /* On vnode dirty list */
238 #define BX_VNCLEAN 0x00000002 /* On vnode clean list */
239 #define BX_BKGRDWRITE 0x00000010 /* Do writes in background */
240 #define BX_BKGRDMARKER 0x00000020 /* Mark buffer for splay tree */
241 #define BX_ALTDATA 0x00000040 /* Holds extended data */
243 #define PRINT_BUF_XFLAGS "\20\7altdata\6bkgrdmarker\5bkgrdwrite\2clean\1dirty"
245 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */
248 * These flags are kept in b_vflags.
250 #define BV_SCANNED 0x00000001 /* VOP_FSYNC funcs mark written bufs */
251 #define BV_BKGRDINPROG 0x00000002 /* Background write in progress */
252 #define BV_BKGRDWAIT 0x00000004 /* Background write waiting */
253 #define BV_INFREECNT 0x80000000 /* buf is counted in numfreebufs */
255 #define PRINT_BUF_VFLAGS "\20\40infreecnt\3bkgrdwait\2bkgrdinprog\1scanned"
261 extern const char *buf_wmesg; /* Default buffer lock message */
262 #define BUF_WMESG "bufwait"
263 #include <sys/proc.h> /* XXX for curthread */
264 #include <sys/mutex.h>
269 #define BUF_LOCKINIT(bp) \
270 lockinit(&(bp)->b_lock, PRIBIO + 4, buf_wmesg, 0, 0)
273 * Get a lock sleeping non-interruptably until it becomes available.
275 #define BUF_LOCK(bp, locktype, interlock) \
276 _lockmgr_args(&(bp)->b_lock, (locktype), (interlock), \
277 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, \
278 LOCK_FILE, LOCK_LINE)
281 * Get a lock sleeping with specified interruptably and timeout.
283 #define BUF_TIMELOCK(bp, locktype, interlock, wmesg, catch, timo) \
284 _lockmgr_args(&(bp)->b_lock, (locktype) | LK_TIMELOCK, \
285 (interlock), (wmesg), (PRIBIO + 4) | (catch), (timo), \
286 LOCK_FILE, LOCK_LINE)
289 * Release a lock. Only the acquiring process may free the lock unless
290 * it has been handed off to biodone.
292 #define BUF_UNLOCK(bp) do { \
293 KASSERT(((bp)->b_flags & B_REMFREE) == 0, \
294 ("BUF_UNLOCK %p while B_REMFREE is still set.", (bp))); \
296 (void)_lockmgr_args(&(bp)->b_lock, LK_RELEASE, NULL, \
297 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, \
298 LOCK_FILE, LOCK_LINE); \
302 * Check if a buffer lock is recursed.
304 #define BUF_LOCKRECURSED(bp) \
305 lockmgr_recursed(&(bp)->b_lock)
308 * Check if a buffer lock is currently held.
310 #define BUF_ISLOCKED(bp) \
311 lockstatus(&(bp)->b_lock)
313 * Free a buffer lock.
315 #define BUF_LOCKFREE(bp) \
316 lockdestroy(&(bp)->b_lock)
319 * Print informations on a buffer lock.
321 #define BUF_LOCKPRINTINFO(bp) \
322 lockmgr_printinfo(&(bp)->b_lock)
325 * Buffer lock assertions.
327 #if defined(INVARIANTS) && defined(INVARIANT_SUPPORT)
328 #define BUF_ASSERT_LOCKED(bp) \
329 _lockmgr_assert(&(bp)->b_lock, KA_LOCKED, LOCK_FILE, LOCK_LINE)
330 #define BUF_ASSERT_SLOCKED(bp) \
331 _lockmgr_assert(&(bp)->b_lock, KA_SLOCKED, LOCK_FILE, LOCK_LINE)
332 #define BUF_ASSERT_XLOCKED(bp) \
333 _lockmgr_assert(&(bp)->b_lock, KA_XLOCKED, LOCK_FILE, LOCK_LINE)
334 #define BUF_ASSERT_UNLOCKED(bp) \
335 _lockmgr_assert(&(bp)->b_lock, KA_UNLOCKED, LOCK_FILE, LOCK_LINE)
336 #define BUF_ASSERT_HELD(bp)
337 #define BUF_ASSERT_UNHELD(bp)
339 #define BUF_ASSERT_LOCKED(bp)
340 #define BUF_ASSERT_SLOCKED(bp)
341 #define BUF_ASSERT_XLOCKED(bp)
342 #define BUF_ASSERT_UNLOCKED(bp)
343 #define BUF_ASSERT_HELD(bp)
344 #define BUF_ASSERT_UNHELD(bp)
347 #ifdef _SYS_PROC_H_ /* Avoid #include <sys/proc.h> pollution */
349 * When initiating asynchronous I/O, change ownership of the lock to the
350 * kernel. Once done, the lock may legally released by biodone. The
351 * original owning process can no longer acquire it recursively, but must
352 * wait until the I/O is completed and the lock has been freed by biodone.
354 #define BUF_KERNPROC(bp) \
355 _lockmgr_disown(&(bp)->b_lock, LOCK_FILE, LOCK_LINE)
359 * Find out if the lock has waiters or not.
361 #define BUF_LOCKWAITERS(bp) \
362 lockmgr_waiters(&(bp)->b_lock)
366 struct buf_queue_head {
367 TAILQ_HEAD(buf_queue, buf) queue;
369 struct buf *insert_point;
370 struct buf *switch_point;
374 * This structure describes a clustered I/O. It is stored in the b_saveaddr
375 * field of the buffer on which I/O is done. At I/O completion, cluster
376 * callback uses the structure to parcel I/O's to individual buffers, and
377 * then free's this structure.
379 struct cluster_save {
380 long bs_bcount; /* Saved b_bcount. */
381 long bs_bufsize; /* Saved b_bufsize. */
382 void *bs_saveaddr; /* Saved b_addr. */
383 int bs_nchildren; /* Number of associated buffers. */
384 struct buf **bs_children; /* List of associated buffers. */
390 bwrite(struct buf *bp)
393 KASSERT(bp->b_bufobj != NULL, ("bwrite: no bufobj bp=%p", bp));
394 KASSERT(bp->b_bufobj->bo_ops != NULL, ("bwrite: no bo_ops bp=%p", bp));
395 KASSERT(bp->b_bufobj->bo_ops->bop_write != NULL,
396 ("bwrite: no bop_write bp=%p", bp));
397 return (BO_WRITE(bp->b_bufobj, bp));
401 bstrategy(struct buf *bp)
404 KASSERT(bp->b_bufobj != NULL, ("bstrategy: no bufobj bp=%p", bp));
405 KASSERT(bp->b_bufobj->bo_ops != NULL,
406 ("bstrategy: no bo_ops bp=%p", bp));
407 KASSERT(bp->b_bufobj->bo_ops->bop_strategy != NULL,
408 ("bstrategy: no bop_strategy bp=%p", bp));
409 BO_STRATEGY(bp->b_bufobj, bp);
413 buf_start(struct buf *bp)
416 (*bioops.io_start)(bp);
420 buf_complete(struct buf *bp)
422 if (bioops.io_complete)
423 (*bioops.io_complete)(bp);
427 buf_deallocate(struct buf *bp)
429 if (bioops.io_deallocate)
430 (*bioops.io_deallocate)(bp);
435 buf_countdeps(struct buf *bp, int i)
437 if (bioops.io_countdeps)
438 return ((*bioops.io_countdeps)(bp, i));
446 * Zero out the buffer's data area.
448 #define clrbuf(bp) { \
449 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \
454 * Flags for getblk's last parameter.
456 #define GB_LOCK_NOWAIT 0x0001 /* Fail if we block on a buf lock. */
457 #define GB_NOCREAT 0x0002 /* Don't create a buf if not found. */
458 #define GB_NOWAIT_BD 0x0004 /* Do not wait for bufdaemon. */
459 #define GB_UNMAPPED 0x0008 /* Do not mmap buffer pages. */
460 #define GB_KVAALLOC 0x0010 /* But allocate KVA. */
463 extern int nbuf; /* The number of buffer headers */
464 extern long maxswzone; /* Max KVA for swap structures */
465 extern long maxbcache; /* Max KVA for buffer cache */
466 extern long runningbufspace;
467 extern long hibufspace;
468 extern int dirtybufthresh;
469 extern int bdwriteskip;
470 extern int dirtybufferflushes;
471 extern int altbufferflushes;
472 extern int buf_maxio; /* nominal maximum I/O for buffer */
473 extern struct buf *buf; /* The buffer headers. */
474 extern char *buffers; /* The buffer contents. */
475 extern int bufpages; /* Number of memory pages in the buffer pool. */
476 extern struct buf *swbuf; /* Swap I/O buffer headers. */
477 extern int nswbuf; /* Number of swap I/O buffer headers. */
478 extern int cluster_pbuf_freecnt; /* Number of pbufs for clusters */
479 extern int vnode_pbuf_freecnt; /* Number of pbufs for vnode pager */
480 extern caddr_t unmapped_buf;
482 void runningbufwakeup(struct buf *);
483 void waitrunningbufspace(void);
484 caddr_t kern_vfs_bio_buffer_alloc(caddr_t v, long physmem_est);
486 void bdata2bio(struct buf *bp, struct bio *bip);
487 void bwillwrite(void);
488 int buf_dirty_count_severe(void);
489 void bremfree(struct buf *);
490 void bremfreef(struct buf *); /* XXX Force bremfree, only for nfs. */
491 int bread(struct vnode *, daddr_t, int, struct ucred *, struct buf **);
492 int bread_gb(struct vnode *, daddr_t, int, struct ucred *,
493 int gbflags, struct buf **);
494 void breada(struct vnode *, daddr_t *, int *, int, struct ucred *);
495 int breadn(struct vnode *, daddr_t, int, daddr_t *, int *, int,
496 struct ucred *, struct buf **);
497 int breadn_flags(struct vnode *, daddr_t, int, daddr_t *, int *, int,
498 struct ucred *, int, struct buf **);
499 void bdwrite(struct buf *);
500 void bawrite(struct buf *);
501 void babarrierwrite(struct buf *);
502 int bbarrierwrite(struct buf *);
503 void bdirty(struct buf *);
504 void bundirty(struct buf *);
505 void bufstrategy(struct bufobj *, struct buf *);
506 void brelse(struct buf *);
507 void bqrelse(struct buf *);
508 int vfs_bio_awrite(struct buf *);
509 struct buf * getpbuf(int *);
510 struct buf *incore(struct bufobj *, daddr_t);
511 struct buf *gbincore(struct bufobj *, daddr_t);
512 struct buf *getblk(struct vnode *, daddr_t, int, int, int, int);
513 struct buf *geteblk(int, int);
514 int bufwait(struct buf *);
515 int bufwrite(struct buf *);
516 void bufdone(struct buf *);
517 void bufdone_finish(struct buf *);
518 void bd_speedup(void);
520 int cluster_read(struct vnode *, u_quad_t, daddr_t, long,
521 struct ucred *, long, int, struct buf **);
522 int cluster_wbuild(struct vnode *, long, daddr_t, int);
523 void cluster_write(struct vnode *, struct buf *, u_quad_t, int);
524 int cluster_read_gb(struct vnode *, u_quad_t, daddr_t, long,
525 struct ucred *, long, int, int, struct buf **);
526 int cluster_wbuild_gb(struct vnode *, long, daddr_t, int, int);
527 void cluster_write_gb(struct vnode *, struct buf *, u_quad_t, int, int);
528 void vfs_bio_bzero_buf(struct buf *bp, int base, int size);
529 void vfs_bio_set_valid(struct buf *, int base, int size);
530 void vfs_bio_clrbuf(struct buf *);
531 void vfs_busy_pages(struct buf *, int clear_modify);
532 void vfs_unbusy_pages(struct buf *);
533 int vmapbuf(struct buf *, int);
534 void vunmapbuf(struct buf *);
535 void relpbuf(struct buf *, int *);
536 void brelvp(struct buf *);
537 void bgetvp(struct vnode *, struct buf *);
538 void pbgetbo(struct bufobj *bo, struct buf *bp);
539 void pbgetvp(struct vnode *, struct buf *);
540 void pbrelbo(struct buf *);
541 void pbrelvp(struct buf *);
542 int allocbuf(struct buf *bp, int size);
543 void reassignbuf(struct buf *);
544 struct buf *trypbuf(int *);
545 void bwait(struct buf *, u_char, const char *);
546 void bdone(struct buf *);
547 void bpin(struct buf *);
548 void bunpin(struct buf *);
549 void bunpin_wait(struct buf *);
553 #endif /* !_SYS_BUF_H_ */