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 uint32_t b_vflags; /* (V) BV_* flags */
110 TAILQ_ENTRY(buf) b_freelist; /* (Q) Free list position inactive. */
111 unsigned short b_qindex; /* (Q) buffer queue index */
112 uint32_t b_flags; /* B_* flags. */
113 b_xflags_t b_xflags; /* extra flags */
114 struct lock b_lock; /* Buffer lock */
115 long b_bufsize; /* Allocated buffer size. */
116 long b_runningbufspace; /* when I/O is running, pipelining */
117 caddr_t b_kvabase; /* base kva for buffer */
118 caddr_t b_kvaalloc; /* allocated kva for B_KVAALLOC */
119 int b_kvasize; /* size of kva for buffer */
120 daddr_t b_lblkno; /* Logical block number. */
121 struct vnode *b_vp; /* Device vnode. */
122 int b_dirtyoff; /* Offset in buffer of dirty region. */
123 int b_dirtyend; /* Offset of end of dirty region. */
124 struct ucred *b_rcred; /* Read credentials reference. */
125 struct ucred *b_wcred; /* Write credentials reference. */
126 void *b_saveaddr; /* Original b_addr for physio. */
131 TAILQ_HEAD(cluster_list_head, buf) cluster_head;
132 TAILQ_ENTRY(buf) cluster_entry;
134 struct vm_page *b_pages[btoc(MAXPHYS)];
136 struct workhead b_dep; /* (D) List of filesystem dependencies. */
143 #define b_object b_bufobj->bo_object
146 * These flags are kept in b_flags.
150 * B_ASYNC VOP calls on bp's are usually async whether or not
151 * B_ASYNC is set, but some subsystems, such as NFS, like
152 * to know what is best for the caller so they can
155 * B_PAGING Indicates that bp is being used by the paging system or
156 * some paging system and that the bp is not linked into
157 * the b_vp's clean/dirty linked lists or ref counts.
158 * Buffer vp reassignments are illegal in this case.
160 * B_CACHE This may only be set if the buffer is entirely valid.
161 * The situation where B_DELWRI is set and B_CACHE is
162 * clear MUST be committed to disk by getblk() so
163 * B_DELWRI can also be cleared. See the comments for
164 * getblk() in kern/vfs_bio.c. If B_CACHE is clear,
165 * the caller is expected to clear BIO_ERROR and B_INVAL,
166 * set BIO_READ, and initiate an I/O.
168 * The 'entire buffer' is defined to be the range from
169 * 0 through b_bcount.
171 * B_MALLOC Request that the buffer be allocated from the malloc
172 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
174 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers
175 * by filesystems that allow clustering when the buffer
176 * is fully dirty and indicates that it may be clustered
177 * with other adjacent dirty buffers. Note the clustering
178 * may not be used with the stage 1 data write under NFS
179 * but may be used for the commit rpc portion.
181 * B_VMIO Indicates that the buffer is tied into an VM object.
182 * The buffer's data is always PAGE_SIZE aligned even
183 * if b_bufsize and b_bcount are not. ( b_bufsize is
184 * always at least DEV_BSIZE aligned, though ).
186 * B_DIRECT Hint that we should attempt to completely free
187 * the pages underlying the buffer. B_DIRECT is
188 * sticky until the buffer is released and typically
189 * only has an effect when B_RELBUF is also set.
193 #define B_AGE 0x00000001 /* Move to age queue when I/O done. */
194 #define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */
195 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */
196 #define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */
197 #define B_DEFERRED 0x00000010 /* Skipped over for cleaning */
198 #define B_CACHE 0x00000020 /* Bread found us in the cache. */
199 #define B_VALIDSUSPWRT 0x00000040 /* Valid write during suspension. */
200 #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */
201 #define B_PERSISTENT 0x00000100 /* Perm. ref'ed while EXT2FS mounted. */
202 #define B_DONE 0x00000200 /* I/O completed. */
203 #define B_EINTR 0x00000400 /* I/O was interrupted */
204 #define B_UNMAPPED 0x00000800 /* KVA is not mapped. */
205 #define B_KVAALLOC 0x00001000 /* But allocated. */
206 #define B_INVAL 0x00002000 /* Does not contain valid info. */
207 #define B_BARRIER 0x00004000 /* Write this and all preceeding first. */
208 #define B_NOCACHE 0x00008000 /* Do not cache block after use. */
209 #define B_MALLOC 0x00010000 /* malloced b_data */
210 #define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */
211 #define B_000400000 0x00040000 /* Available flag. */
212 #define B_000800000 0x00080000 /* Available flag. */
213 #define B_00100000 0x00100000 /* Available flag. */
214 #define B_DIRTY 0x00200000 /* Needs writing later (in EXT2FS). */
215 #define B_RELBUF 0x00400000 /* Release VMIO buffer. */
216 #define B_00800000 0x00800000 /* Available flag. */
217 #define B_NOCOPY 0x01000000 /* Don't copy-on-write this buf. */
218 #define B_INFREECNT 0x02000000 /* buf is counted in numfreebufs */
219 #define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */
220 #define B_MANAGED 0x08000000 /* Managed by FS. */
221 #define B_RAM 0x10000000 /* Read ahead mark (flag) */
222 #define B_VMIO 0x20000000 /* VMIO flag */
223 #define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */
224 #define B_REMFREE 0x80000000 /* Delayed bremfree */
226 #define PRINT_BUF_FLAGS "\20\40remfree\37cluster\36vmio\35ram\34managed" \
227 "\33paging\32infreecnt\31nocopy\30b23\27relbuf\26dirty\25b20" \
228 "\24b19\23b18\22clusterok\21malloc\20nocache\17b14\16inval" \
229 "\15b12\14b11\13eintr\12done\11persist\10delwri\7validsuspwrt" \
230 "\6cache\5deferred\4direct\3async\2needcommit\1age"
233 * These flags are kept in b_xflags.
235 #define BX_VNDIRTY 0x00000001 /* On vnode dirty list */
236 #define BX_VNCLEAN 0x00000002 /* On vnode clean list */
237 #define BX_BKGRDWRITE 0x00000010 /* Do writes in background */
238 #define BX_BKGRDMARKER 0x00000020 /* Mark buffer for splay tree */
239 #define BX_ALTDATA 0x00000040 /* Holds extended data */
241 #define PRINT_BUF_XFLAGS "\20\7altdata\6bkgrdmarker\5bkgrdwrite\2clean\1dirty"
243 #define NOOFFSET (-1LL) /* No buffer offset calculated yet */
246 * These flags are kept in b_vflags.
248 #define BV_SCANNED 0x00000001 /* VOP_FSYNC funcs mark written bufs */
249 #define BV_BKGRDINPROG 0x00000002 /* Background write in progress */
250 #define BV_BKGRDWAIT 0x00000004 /* Background write waiting */
252 #define PRINT_BUF_VFLAGS "\20\3bkgrdwait\2bkgrdinprog\1scanned"
258 extern const char *buf_wmesg; /* Default buffer lock message */
259 #define BUF_WMESG "bufwait"
260 #include <sys/proc.h> /* XXX for curthread */
261 #include <sys/mutex.h>
266 #define BUF_LOCKINIT(bp) \
267 lockinit(&(bp)->b_lock, PRIBIO + 4, buf_wmesg, 0, 0)
270 * Get a lock sleeping non-interruptably until it becomes available.
272 #define BUF_LOCK(bp, locktype, interlock) \
273 _lockmgr_args_rw(&(bp)->b_lock, (locktype), (interlock), \
274 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, \
275 LOCK_FILE, LOCK_LINE)
278 * Get a lock sleeping with specified interruptably and timeout.
280 #define BUF_TIMELOCK(bp, locktype, interlock, wmesg, catch, timo) \
281 _lockmgr_args_rw(&(bp)->b_lock, (locktype) | LK_TIMELOCK, \
282 (interlock), (wmesg), (PRIBIO + 4) | (catch), (timo), \
283 LOCK_FILE, LOCK_LINE)
286 * Release a lock. Only the acquiring process may free the lock unless
287 * it has been handed off to biodone.
289 #define BUF_UNLOCK(bp) do { \
290 KASSERT(((bp)->b_flags & B_REMFREE) == 0, \
291 ("BUF_UNLOCK %p while B_REMFREE is still set.", (bp))); \
293 (void)_lockmgr_args(&(bp)->b_lock, LK_RELEASE, NULL, \
294 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, \
295 LOCK_FILE, LOCK_LINE); \
299 * Check if a buffer lock is recursed.
301 #define BUF_LOCKRECURSED(bp) \
302 lockmgr_recursed(&(bp)->b_lock)
305 * Check if a buffer lock is currently held.
307 #define BUF_ISLOCKED(bp) \
308 lockstatus(&(bp)->b_lock)
310 * Free a buffer lock.
312 #define BUF_LOCKFREE(bp) \
313 lockdestroy(&(bp)->b_lock)
316 * Print informations on a buffer lock.
318 #define BUF_LOCKPRINTINFO(bp) \
319 lockmgr_printinfo(&(bp)->b_lock)
322 * Buffer lock assertions.
324 #if defined(INVARIANTS) && defined(INVARIANT_SUPPORT)
325 #define BUF_ASSERT_LOCKED(bp) \
326 _lockmgr_assert(&(bp)->b_lock, KA_LOCKED, LOCK_FILE, LOCK_LINE)
327 #define BUF_ASSERT_SLOCKED(bp) \
328 _lockmgr_assert(&(bp)->b_lock, KA_SLOCKED, LOCK_FILE, LOCK_LINE)
329 #define BUF_ASSERT_XLOCKED(bp) \
330 _lockmgr_assert(&(bp)->b_lock, KA_XLOCKED, LOCK_FILE, LOCK_LINE)
331 #define BUF_ASSERT_UNLOCKED(bp) \
332 _lockmgr_assert(&(bp)->b_lock, KA_UNLOCKED, LOCK_FILE, LOCK_LINE)
333 #define BUF_ASSERT_HELD(bp)
334 #define BUF_ASSERT_UNHELD(bp)
336 #define BUF_ASSERT_LOCKED(bp)
337 #define BUF_ASSERT_SLOCKED(bp)
338 #define BUF_ASSERT_XLOCKED(bp)
339 #define BUF_ASSERT_UNLOCKED(bp)
340 #define BUF_ASSERT_HELD(bp)
341 #define BUF_ASSERT_UNHELD(bp)
344 #ifdef _SYS_PROC_H_ /* Avoid #include <sys/proc.h> pollution */
346 * When initiating asynchronous I/O, change ownership of the lock to the
347 * kernel. Once done, the lock may legally released by biodone. The
348 * original owning process can no longer acquire it recursively, but must
349 * wait until the I/O is completed and the lock has been freed by biodone.
351 #define BUF_KERNPROC(bp) \
352 _lockmgr_disown(&(bp)->b_lock, LOCK_FILE, LOCK_LINE)
356 * Find out if the lock has waiters or not.
358 #define BUF_LOCKWAITERS(bp) \
359 lockmgr_waiters(&(bp)->b_lock)
363 struct buf_queue_head {
364 TAILQ_HEAD(buf_queue, buf) queue;
366 struct buf *insert_point;
367 struct buf *switch_point;
371 * This structure describes a clustered I/O. It is stored in the b_saveaddr
372 * field of the buffer on which I/O is done. At I/O completion, cluster
373 * callback uses the structure to parcel I/O's to individual buffers, and
374 * then free's this structure.
376 struct cluster_save {
377 long bs_bcount; /* Saved b_bcount. */
378 long bs_bufsize; /* Saved b_bufsize. */
379 void *bs_saveaddr; /* Saved b_addr. */
380 int bs_nchildren; /* Number of associated buffers. */
381 struct buf **bs_children; /* List of associated buffers. */
387 bwrite(struct buf *bp)
390 KASSERT(bp->b_bufobj != NULL, ("bwrite: no bufobj bp=%p", bp));
391 KASSERT(bp->b_bufobj->bo_ops != NULL, ("bwrite: no bo_ops bp=%p", bp));
392 KASSERT(bp->b_bufobj->bo_ops->bop_write != NULL,
393 ("bwrite: no bop_write bp=%p", bp));
394 return (BO_WRITE(bp->b_bufobj, bp));
398 bstrategy(struct buf *bp)
401 KASSERT(bp->b_bufobj != NULL, ("bstrategy: no bufobj bp=%p", bp));
402 KASSERT(bp->b_bufobj->bo_ops != NULL,
403 ("bstrategy: no bo_ops bp=%p", bp));
404 KASSERT(bp->b_bufobj->bo_ops->bop_strategy != NULL,
405 ("bstrategy: no bop_strategy bp=%p", bp));
406 BO_STRATEGY(bp->b_bufobj, bp);
410 buf_start(struct buf *bp)
413 (*bioops.io_start)(bp);
417 buf_complete(struct buf *bp)
419 if (bioops.io_complete)
420 (*bioops.io_complete)(bp);
424 buf_deallocate(struct buf *bp)
426 if (bioops.io_deallocate)
427 (*bioops.io_deallocate)(bp);
431 buf_countdeps(struct buf *bp, int i)
433 if (bioops.io_countdeps)
434 return ((*bioops.io_countdeps)(bp, i));
442 * Zero out the buffer's data area.
444 #define clrbuf(bp) { \
445 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \
450 * Flags for getblk's last parameter.
452 #define GB_LOCK_NOWAIT 0x0001 /* Fail if we block on a buf lock. */
453 #define GB_NOCREAT 0x0002 /* Don't create a buf if not found. */
454 #define GB_NOWAIT_BD 0x0004 /* Do not wait for bufdaemon. */
455 #define GB_UNMAPPED 0x0008 /* Do not mmap buffer pages. */
456 #define GB_KVAALLOC 0x0010 /* But allocate KVA. */
459 extern int nbuf; /* The number of buffer headers */
460 extern long maxswzone; /* Max KVA for swap structures */
461 extern long maxbcache; /* Max KVA for buffer cache */
462 extern long runningbufspace;
463 extern long hibufspace;
464 extern int dirtybufthresh;
465 extern int bdwriteskip;
466 extern int dirtybufferflushes;
467 extern int altbufferflushes;
468 extern struct buf *buf; /* The buffer headers. */
469 extern struct buf *swbuf; /* Swap I/O buffer headers. */
470 extern int nswbuf; /* Number of swap I/O buffer headers. */
471 extern int cluster_pbuf_freecnt; /* Number of pbufs for clusters */
472 extern int vnode_pbuf_freecnt; /* Number of pbufs for vnode pager */
473 extern caddr_t unmapped_buf;
475 void runningbufwakeup(struct buf *);
476 void waitrunningbufspace(void);
477 caddr_t kern_vfs_bio_buffer_alloc(caddr_t v, long physmem_est);
479 void bdata2bio(struct buf *bp, struct bio *bip);
480 void bwillwrite(void);
481 int buf_dirty_count_severe(void);
482 void bremfree(struct buf *);
483 void bremfreef(struct buf *); /* XXX Force bremfree, only for nfs. */
484 #define bread(vp, blkno, size, cred, bpp) \
485 breadn_flags(vp, blkno, size, NULL, NULL, 0, cred, 0, bpp)
486 #define bread_gb(vp, blkno, size, cred, gbflags, bpp) \
487 breadn_flags(vp, blkno, size, NULL, NULL, 0, cred, \
489 #define breadn(vp, blkno, size, rablkno, rabsize, cnt, cred, bpp) \
490 breadn_flags(vp, blkno, size, rablkno, rabsize, cnt, cred, 0, bpp)
491 int breadn_flags(struct vnode *, daddr_t, int, daddr_t *, int *, int,
492 struct ucred *, int, struct buf **);
493 void breada(struct vnode *, daddr_t *, int *, int, struct ucred *);
494 void bdwrite(struct buf *);
495 void bawrite(struct buf *);
496 void babarrierwrite(struct buf *);
497 int bbarrierwrite(struct buf *);
498 void bdirty(struct buf *);
499 void bundirty(struct buf *);
500 void bufstrategy(struct bufobj *, struct buf *);
501 void brelse(struct buf *);
502 void bqrelse(struct buf *);
503 int vfs_bio_awrite(struct buf *);
504 void vfs_drain_busy_pages(struct buf *bp);
505 struct buf * getpbuf(int *);
506 struct buf *incore(struct bufobj *, daddr_t);
507 struct buf *gbincore(struct bufobj *, daddr_t);
508 struct buf *getblk(struct vnode *, daddr_t, int, int, int, int);
509 struct buf *geteblk(int, int);
510 int bufwait(struct buf *);
511 int bufwrite(struct buf *);
512 void bufdone(struct buf *);
513 void bufdone_finish(struct buf *);
514 void bd_speedup(void);
516 int cluster_read(struct vnode *, u_quad_t, daddr_t, long,
517 struct ucred *, long, int, int, struct buf **);
518 int cluster_wbuild(struct vnode *, long, daddr_t, int, int);
519 void cluster_write(struct vnode *, struct buf *, u_quad_t, int, int);
520 void vfs_bio_bzero_buf(struct buf *bp, int base, int size);
521 void vfs_bio_set_valid(struct buf *, int base, int size);
522 void vfs_bio_clrbuf(struct buf *);
523 void vfs_busy_pages(struct buf *, int clear_modify);
524 void vfs_unbusy_pages(struct buf *);
525 int vmapbuf(struct buf *, int);
526 void vunmapbuf(struct buf *);
527 void relpbuf(struct buf *, int *);
528 void brelvp(struct buf *);
529 void bgetvp(struct vnode *, struct buf *);
530 void pbgetbo(struct bufobj *bo, struct buf *bp);
531 void pbgetvp(struct vnode *, struct buf *);
532 void pbrelbo(struct buf *);
533 void pbrelvp(struct buf *);
534 int allocbuf(struct buf *bp, int size);
535 void reassignbuf(struct buf *);
536 struct buf *trypbuf(int *);
537 void bwait(struct buf *, u_char, const char *);
538 void bdone(struct buf *);
539 void bpin(struct buf *);
540 void bunpin(struct buf *);
541 void bunpin_wait(struct buf *);
545 #endif /* !_SYS_BUF_H_ */