Add BUF_CMD_FLUSH support - issue flush command to mass storage device.
[dragonfly.git] / sys / sys / buf.h
CommitLineData
984263bc
MD
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: src/sys/sys/buf.h,v 1.88.2.10 2003/01/25 19:02:23 dillon Exp $
a9bf1b8c 40 * $DragonFly: src/sys/sys/buf.h,v 1.54 2008/08/29 20:08:37 dillon Exp $
984263bc
MD
41 */
42
43#ifndef _SYS_BUF_H_
44#define _SYS_BUF_H_
45
03d6a592
MD
46#if defined(_KERNEL) || defined(_KERNEL_STRUCTURES)
47
335dda38 48#ifndef _SYS_QUEUE_H_
984263bc 49#include <sys/queue.h>
335dda38
MD
50#endif
51#ifndef _SYS_LOCK_H_
984263bc 52#include <sys/lock.h>
335dda38
MD
53#endif
54#ifndef _SYS_DEVICE_H_
55#include <sys/device.h>
56#endif
984263bc 57
54f51aeb
HP
58#ifndef _SYS_XIO_H_
59#include <sys/xio.h>
60#endif
6bae6177
MD
61#ifndef _SYS_TREE_H_
62#include <sys/tree.h>
63#endif
1eabbf32
HP
64#ifndef _SYS_BIO_H_
65#include <sys/bio.h>
66#endif
16523a43
MD
67#ifndef _SYS_SPINLOCK_H_
68#include <sys/spinlock.h>
69#endif
1eabbf32 70
984263bc 71struct buf;
81b5c339 72struct bio;
984263bc
MD
73struct mount;
74struct vnode;
54f51aeb 75struct xio;
984263bc 76
81b5c339
MD
77#define NBUF_BIO 4
78
6bae6177 79struct buf_rb_tree;
1f1ea522 80struct buf_rb_hash;
e656fe10
MD
81RB_PROTOTYPE2(buf_rb_tree, buf, b_rbnode, rb_buf_compare, off_t);
82RB_PROTOTYPE2(buf_rb_hash, buf, b_rbhash, rb_buf_compare, off_t);
6bae6177 83
984263bc
MD
84/*
85 * To avoid including <ufs/ffs/softdep.h>
86 */
87LIST_HEAD(workhead, worklist);
984263bc 88
e92ca23a
MD
89#endif
90
10f3fee5
MD
91typedef enum buf_cmd {
92 BUF_CMD_DONE = 0,
93 BUF_CMD_READ,
94 BUF_CMD_WRITE,
95 BUF_CMD_FREEBLKS,
a9bf1b8c
MD
96 BUF_CMD_FORMAT,
97 BUF_CMD_FLUSH
10f3fee5
MD
98} buf_cmd_t;
99
e92ca23a
MD
100#if defined(_KERNEL) || defined(_KERNEL_STRUCTURES)
101
984263bc
MD
102/*
103 * The buffer header describes an I/O operation in the kernel.
104 *
105 * NOTES:
9a71d53f
MD
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.
984263bc 110 *
4414f2c9
MD
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).
984263bc 120 *
4414f2c9
MD
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
124 * start at 0.
9a71d53f
MD
125 *
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).
81b5c339
MD
129 *
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.
132 *
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).
137 *
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.
146 *
408357d8
MD
147 * b_ops - initialized if a buffer has a bio_ops
148 *
81b5c339
MD
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.
984263bc
MD
155 */
156struct buf {
1f1ea522
MD
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 */
984263bc 159 TAILQ_ENTRY(buf) b_freelist; /* Free list position if not active. */
81b5c339 160 struct buf *b_cluster_next; /* Next buffer (cluster code) */
54078292 161 struct vnode *b_vp; /* (vp, loffset) index */
81b5c339 162 struct bio b_bio_array[NBUF_BIO]; /* BIO translation layers */
54078292 163 u_int32_t b_flags; /* B_* flags. */
984263bc 164 unsigned short b_qindex; /* buffer queue index */
9e45bec7 165 unsigned short b_unused01;
984263bc 166 struct lock b_lock; /* Buffer lock */
10f3fee5 167 buf_cmd_t b_cmd; /* I/O command */
54078292
MD
168 int b_bufsize; /* Allocated buffer size. */
169 int b_runningbufspace; /* when I/O is running, pipelining */
170 int b_bcount; /* Valid bytes in buffer. */
171 int b_resid; /* Remaining I/O */
81b5c339 172 int b_error; /* Error return */
984263bc
MD
173 caddr_t b_data; /* Memory, superblocks, indirect etc. */
174 caddr_t b_kvabase; /* base kva for buffer */
175 int b_kvasize; /* size of kva for buffer */
984263bc
MD
176 int b_dirtyoff; /* Offset in buffer of dirty region. */
177 int b_dirtyend; /* Offset of end of dirty region. */
81b5c339 178 struct xio b_xio; /* data buffer page list management */
408357d8 179 struct bio_ops *b_ops; /* bio_ops used w/ b_dep */
984263bc 180 struct workhead b_dep; /* List of filesystem dependencies. */
984263bc
MD
181};
182
81b5c339
MD
183/*
184 * XXX temporary
185 */
186#define b_bio1 b_bio_array[0] /* logical layer */
187#define b_bio2 b_bio_array[1] /* (typically) the disk layer */
188#define b_loffset b_bio1.bio_offset
984263bc 189
4b958e7b
MD
190
191/*
192 * Flags passed to getblk()
193 *
194 * GETBLK_PCATCH - Allow signals to be caught. getblk() is allowed to return
195 * NULL if this flag is passed.
196 *
197 * GETBLK_BHEAVY - This is a heavy weight buffer, meaning that resolving
198 * writes can require additional buffers.
b77cfc40
MD
199 *
200 * GETBLK_SZMATCH- blksize must match pre-existing b_bcount. getblk() can
201 * return NULL.
202 *
203 * GETBLK_NOWAIT - Do not use a blocking lock. getblk() can return NULL.
4b958e7b
MD
204 */
205#define GETBLK_PCATCH 0x0001 /* catch signals */
206#define GETBLK_BHEAVY 0x0002 /* heavy weight buffer */
b77cfc40
MD
207#define GETBLK_SZMATCH 0x0004 /* pre-existing buffer must match */
208#define GETBLK_NOWAIT 0x0008 /* non-blocking */
4b958e7b 209
984263bc
MD
210/*
211 * These flags are kept in b_flags.
212 *
213 * Notes:
214 *
215 * B_ASYNC VOP calls on bp's are usually async whether or not
216 * B_ASYNC is set, but some subsystems, such as NFS, like
217 * to know what is best for the caller so they can
218 * optimize the I/O.
219 *
220 * B_PAGING Indicates that bp is being used by the paging system or
221 * some paging system and that the bp is not linked into
222 * the b_vp's clean/dirty linked lists or ref counts.
223 * Buffer vp reassignments are illegal in this case.
224 *
225 * B_CACHE This may only be set if the buffer is entirely valid.
226 * The situation where B_DELWRI is set and B_CACHE is
227 * clear MUST be committed to disk by getblk() so
228 * B_DELWRI can also be cleared. See the comments for
229 * getblk() in kern/vfs_bio.c. If B_CACHE is clear,
230 * the caller is expected to clear B_ERROR|B_INVAL,
10f3fee5 231 * set BUF_CMD_READ, and initiate an I/O.
984263bc
MD
232 *
233 * The 'entire buffer' is defined to be the range from
234 * 0 through b_bcount.
235 *
236 * B_MALLOC Request that the buffer be allocated from the malloc
237 * pool, DEV_BSIZE aligned instead of PAGE_SIZE aligned.
238 *
239 * B_CLUSTEROK This flag is typically set for B_DELWRI buffers
240 * by filesystems that allow clustering when the buffer
241 * is fully dirty and indicates that it may be clustered
242 * with other adjacent dirty buffers. Note the clustering
243 * may not be used with the stage 1 data write under NFS
244 * but may be used for the commit rpc portion.
245 *
246 * B_VMIO Indicates that the buffer is tied into an VM object.
247 * The buffer's data is always PAGE_SIZE aligned even
248 * if b_bufsize and b_bcount are not. ( b_bufsize is
249 * always at least DEV_BSIZE aligned, though ).
250 *
251 * B_DIRECT Hint that we should attempt to completely free
252 * the pages underlying the buffer. B_DIRECT is
253 * sticky until the buffer is released and typically
254 * only has an effect when B_RELBUF is also set.
255 *
27bc0cb1
MD
256 * B_LOCKED The buffer will be released to the locked queue
257 * regardless of its current state. Note that
258 * if B_DELWRI is set, no I/O occurs until the caller
259 * acquires the buffer, clears B_LOCKED, then releases
260 * it again.
b86460bf
MD
261 *
262 * B_AGE When allocating a new buffer any buffer encountered
263 * with B_AGE set will be reallocated more quickly then
264 * buffers encountered without it set. B_AGE is set
265 * as part of the loop so idle buffers should eventually
266 * wind up with B_AGE set. B_AGE explicitly does NOT
267 * cause the buffer to be instantly reallocated for
268 * other purposes.
269 *
270 * Standard buffer flushing routines leave B_AGE intact
271 * through the DIRTY queue and into the CLEAN queue.
272 * Setting B_AGE on a dirty buffer will not cause it
273 * to be flushed more quickly but will cause it to be
274 * reallocated more quickly after having been flushed.
e409de84
MD
275 *
276 * B_NOCACHE Request that the buffer and backing store be
277 * destroyed on completion. If B_DELWRI is set and the
278 * write fails, the buffer remains intact.
984263bc
MD
279 */
280
b86460bf 281#define B_AGE 0x00000001 /* Reuse more quickly */
984263bc
MD
282#define B_NEEDCOMMIT 0x00000002 /* Append-write in progress. */
283#define B_ASYNC 0x00000004 /* Start I/O, do not wait. */
284#define B_DIRECT 0x00000008 /* direct I/O flag (pls free vmio) */
b86460bf 285#define B_DEFERRED 0x00000010 /* vfs-controlled deferment */
984263bc 286#define B_CACHE 0x00000020 /* Bread found us in the cache. */
1f1ea522 287#define B_HASHED 0x00000040 /* Indexed via v_rbhash_tree */
984263bc 288#define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */
4414f2c9 289#define B_BNOCLIP 0x00000100 /* EOF clipping b_bcount not allowed */
10f3fee5 290#define B_UNUSED0200 0x00000200
984263bc
MD
291#define B_EINTR 0x00000400 /* I/O was interrupted */
292#define B_ERROR 0x00000800 /* I/O error occurred. */
71bc1d95 293#define B_UNUSED1000 0x00001000 /* Unused */
984263bc
MD
294#define B_INVAL 0x00002000 /* Does not contain valid info. */
295#define B_LOCKED 0x00004000 /* Locked in core (not reusable). */
135bd6a8 296#define B_NOCACHE 0x00008000 /* Destroy buffer AND backing store */
984263bc
MD
297#define B_MALLOC 0x00010000 /* malloced b_data */
298#define B_CLUSTEROK 0x00020000 /* Pagein op, so swap() can count it. */
9fd739f4 299#define B_UNUSED40000 0x00040000
984263bc 300#define B_RAW 0x00080000 /* Set by physio for raw transfers. */
4b958e7b 301#define B_HEAVY 0x00100000 /* Heavy-weight buffer */
984263bc
MD
302#define B_DIRTY 0x00200000 /* Needs writing later. */
303#define B_RELBUF 0x00400000 /* Release VMIO buffer. */
304#define B_WANT 0x00800000 /* Used by vm_pager.c */
9e45bec7
MD
305#define B_VNCLEAN 0x01000000 /* On vnode clean list */
306#define B_VNDIRTY 0x02000000 /* On vnode dirty list */
984263bc
MD
307#define B_PAGING 0x04000000 /* volatile paging I/O -- bypass VMIO */
308#define B_ORDERED 0x08000000 /* Must guarantee I/O ordering */
309#define B_RAM 0x10000000 /* Read ahead mark (flag) */
310#define B_VMIO 0x20000000 /* VMIO flag */
311#define B_CLUSTER 0x40000000 /* pagein op, so swap() can count it */
f2d7fcf0 312#define B_UNUSED80000000 0x80000000
984263bc 313
9e45bec7 314#define PRINT_BUF_FLAGS "\20" \
f2d7fcf0 315 "\40unused31\37cluster\36vmio\35ram\34ordered" \
9e45bec7
MD
316 "\33paging\32vndirty\31vnclean\30want\27relbuf\26dirty" \
317 "\25unused20\24raw\23unused18\22clusterok\21malloc\20nocache" \
1b30fbcc 318 "\17locked\16inval\15unused12\14error\13eintr\12unused9\11bnoclip" \
9e45bec7 319 "\10delwri\7hashed\6cache\5deferred\4direct\3async\2needcommit\1age"
984263bc
MD
320
321#define NOOFFSET (-1LL) /* No buffer offset calculated yet */
322
323#ifdef _KERNEL
324/*
3020e3be 325 * Buffer locking. See sys/buf2.h for inline functions.
984263bc 326 */
984263bc
MD
327extern char *buf_wmesg; /* Default buffer lock message */
328#define BUF_WMESG "bufwait"
984263bc
MD
329
330#endif /* _KERNEL */
331
81b5c339
MD
332struct bio_queue_head {
333 TAILQ_HEAD(bio_queue, bio) queue;
54078292 334 off_t last_offset;
81b5c339
MD
335 struct bio *insert_point;
336 struct bio *switch_point;
984263bc
MD
337};
338
339/*
54078292 340 * This structure describes a clustered I/O.
984263bc
MD
341 */
342struct cluster_save {
984263bc
MD
343 int bs_nchildren; /* Number of associated buffers. */
344 struct buf **bs_children; /* List of associated buffers. */
345};
346
984263bc
MD
347/*
348 * Zero out the buffer's data area.
349 */
350#define clrbuf(bp) { \
351 bzero((bp)->b_data, (u_int)(bp)->b_bcount); \
352 (bp)->b_resid = 0; \
353}
354
355/*
356 * Flags to low-level bitmap allocation routines (balloc).
357 *
358 * Note: sequential_heuristic() in kern/vfs_vnops.c limits the count
359 * to 127.
360 */
361#define B_SEQMASK 0x7F000000 /* Sequential heuristic mask. */
362#define B_SEQSHIFT 24 /* Sequential heuristic shift. */
363#define B_SEQMAX 0x7F
364#define B_CLRBUF 0x01 /* Cleared invalid areas of buffer. */
365#define B_SYNC 0x02 /* Do all allocations synchronously. */
366
367#ifdef _KERNEL
368extern int nbuf; /* The number of buffer headers */
369extern int maxswzone; /* Max KVA for swap structures */
370extern int maxbcache; /* Max KVA for buffer cache */
371extern int runningbufspace;
1b30fbcc 372extern int runningbufcount;
868d24af 373extern int hidirtybufspace;
984263bc 374extern int buf_maxio; /* nominal maximum I/O for buffer */
41cbf46c 375extern struct buf *buf; /* The buffer headers. */
984263bc
MD
376extern char *buffers; /* The buffer contents. */
377extern int bufpages; /* Number of memory pages in the buffer pool. */
378extern struct buf *swbuf; /* Swap I/O buffer headers. */
379extern int nswbuf; /* Number of swap I/O buffer headers. */
984263bc
MD
380
381struct uio;
382
b153f746 383void bufinit (void);
c4df9635
MD
384int bd_heatup (void);
385void bd_wait (int count);
b153f746 386int buf_dirty_count_severe (void);
81b5c339
MD
387void initbufbio(struct buf *);
388void reinitbufbio(struct buf *);
389void clearbiocache(struct bio *);
b153f746 390void bremfree (struct buf *);
54078292
MD
391int bread (struct vnode *, off_t, int, struct buf **);
392int breadn (struct vnode *, off_t, int, off_t *, int *, int,
b153f746
RG
393 struct buf **);
394int bwrite (struct buf *);
395void bdwrite (struct buf *);
396void bawrite (struct buf *);
397void bdirty (struct buf *);
4b958e7b 398void bheavy (struct buf *);
b153f746
RG
399void bundirty (struct buf *);
400int bowrite (struct buf *);
401void brelse (struct buf *);
402void bqrelse (struct buf *);
403int vfs_bio_awrite (struct buf *);
81b5c339 404struct buf *getpbuf (int *);
54078292
MD
405int inmem (struct vnode *, off_t);
406struct buf *findblk (struct vnode *, off_t);
407struct buf *getblk (struct vnode *, off_t, int, int, int);
b153f746 408struct buf *geteblk (int);
5e23ca53 409void regetblk(struct buf *bp);
81b5c339 410struct bio *push_bio(struct bio *);
b77cfc40 411struct bio *pop_bio(struct bio *);
b153f746 412int biowait (struct buf *);
81b5c339 413void biodone (struct bio *);
984263bc 414
81b5c339 415void cluster_append(struct bio *, struct buf *);
54078292
MD
416int cluster_read (struct vnode *, off_t, off_t, int,
417 int, int, struct buf **);
418int cluster_wbuild (struct vnode *, int, off_t, int);
e92ca23a 419void cluster_write (struct buf *, off_t, int, int);
fef8985e
MD
420int physread (struct dev_read_args *);
421int physwrite (struct dev_write_args *);
b153f746
RG
422void vfs_bio_set_validclean (struct buf *, int base, int size);
423void vfs_bio_clrbuf (struct buf *);
10f3fee5 424void vfs_busy_pages (struct vnode *, struct buf *);
b153f746 425void vfs_unbusy_pages (struct buf *);
3591bbc6 426int vmapbuf (struct buf *, caddr_t, int);
b153f746
RG
427void vunmapbuf (struct buf *);
428void relpbuf (struct buf *, int *);
429void brelvp (struct buf *);
430void bgetvp (struct vnode *, struct buf *);
b153f746 431int allocbuf (struct buf *bp, int size);
41cbf46c 432int scan_all_buffers (int (*)(struct buf *, void *), void *);
1f1ea522 433void reassignbuf (struct buf *);
b153f746 434struct buf *trypbuf (int *);
408357d8 435void bio_ops_sync(struct mount *mp);
6bb60d59
MD
436void vm_hold_free_pages(struct buf *bp, vm_offset_t from, vm_offset_t to);
437void vm_hold_load_pages(struct buf *bp, vm_offset_t from, vm_offset_t to);
984263bc 438
03d6a592
MD
439#endif /* _KERNEL */
440#endif /* _KERNEL || _KERNEL_STRUCTURES */
441#endif /* !_SYS_BUF_H_ */