2 * Copyright (c) 2003 Matthew Dillon <dillon@backplane.com> All rights reserved.
3 * cdevsw from kern/kern_conf.c Copyright (c) 1995 Terrence R. Lambert
4 * cdevsw from kern/kern_conf.c Copyright (c) 1995 Julian R. Elishcer,
6 * Copyright (c) 1982, 1986, 1991, 1993
7 * The Regents of the University of California. All rights reserved.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * $DragonFly: src/sys/kern/kern_device.c,v 1.27 2007/07/23 18:59:50 dillon Exp $
32 #include <sys/param.h>
33 #include <sys/systm.h>
34 #include <sys/kernel.h>
35 #include <sys/sysctl.h>
36 #include <sys/systm.h>
37 #include <sys/module.h>
38 #include <sys/malloc.h>
42 #include <sys/vnode.h>
43 #include <sys/queue.h>
44 #include <sys/device.h>
46 #include <sys/syslink_rpc.h>
48 #include <machine/stdarg.h>
49 #include <sys/thread2.h>
50 #include <sys/devfs.h>
51 #include <sys/dsched.h>
54 * system link descriptors identify the command in the
55 * arguments structure.
57 #define DDESCNAME(name) __CONCAT(__CONCAT(dev_,name),_desc)
59 #define DEVOP_DESC_INIT(name) \
60 struct syslink_desc DDESCNAME(name) = { \
61 __offsetof(struct dev_ops, __CONCAT(d_, name)), \
64 DEVOP_DESC_INIT(default);
65 DEVOP_DESC_INIT(open);
66 DEVOP_DESC_INIT(close);
67 DEVOP_DESC_INIT(read);
68 DEVOP_DESC_INIT(write);
69 DEVOP_DESC_INIT(ioctl);
70 DEVOP_DESC_INIT(dump);
71 DEVOP_DESC_INIT(psize);
72 DEVOP_DESC_INIT(poll);
73 DEVOP_DESC_INIT(mmap);
74 DEVOP_DESC_INIT(strategy);
75 DEVOP_DESC_INIT(kqfilter);
76 DEVOP_DESC_INIT(revoke);
77 DEVOP_DESC_INIT(clone);
82 struct dev_ops dead_dev_ops;
84 struct dev_ops default_dev_ops = {
86 .d_default = NULL, /* must be NULL */
94 .d_strategy = nostrategy,
97 .d_kqfilter = nokqfilter,
102 /************************************************************************
103 * GENERAL DEVICE API FUNCTIONS *
104 ************************************************************************/
107 dev_dopen(cdev_t dev, int oflags, int devtype, struct ucred *cred)
109 struct dev_open_args ap;
111 ap.a_head.a_desc = &dev_open_desc;
112 ap.a_head.a_dev = dev;
113 ap.a_oflags = oflags;
114 ap.a_devtype = devtype;
116 return(dev->si_ops->d_open(&ap));
120 dev_dclose(cdev_t dev, int fflag, int devtype)
122 struct dev_close_args ap;
124 ap.a_head.a_desc = &dev_close_desc;
125 ap.a_head.a_dev = dev;
127 ap.a_devtype = devtype;
128 return(dev->si_ops->d_close(&ap));
132 dev_dread(cdev_t dev, struct uio *uio, int ioflag)
134 struct dev_read_args ap;
137 ap.a_head.a_desc = &dev_read_desc;
138 ap.a_head.a_dev = dev;
140 ap.a_ioflag = ioflag;
141 error = dev->si_ops->d_read(&ap);
143 dev->si_lastread = time_second;
148 dev_dwrite(cdev_t dev, struct uio *uio, int ioflag)
150 struct dev_write_args ap;
153 dev->si_lastwrite = time_second;
154 ap.a_head.a_desc = &dev_write_desc;
155 ap.a_head.a_dev = dev;
157 ap.a_ioflag = ioflag;
158 error = dev->si_ops->d_write(&ap);
163 dev_dioctl(cdev_t dev, u_long cmd, caddr_t data, int fflag, struct ucred *cred,
166 struct dev_ioctl_args ap;
168 ap.a_head.a_desc = &dev_ioctl_desc;
169 ap.a_head.a_dev = dev;
175 return(dev->si_ops->d_ioctl(&ap));
179 dev_dpoll(cdev_t dev, int events)
181 struct dev_poll_args ap;
184 ap.a_head.a_desc = &dev_poll_desc;
185 ap.a_head.a_dev = dev;
186 ap.a_events = events;
187 error = dev->si_ops->d_poll(&ap);
190 return (seltrue(dev, events));
194 dev_dmmap(cdev_t dev, vm_offset_t offset, int nprot)
196 struct dev_mmap_args ap;
199 ap.a_head.a_desc = &dev_mmap_desc;
200 ap.a_head.a_dev = dev;
201 ap.a_offset = offset;
203 error = dev->si_ops->d_mmap(&ap);
210 dev_dclone(cdev_t dev)
212 struct dev_clone_args ap;
214 ap.a_head.a_desc = &dev_clone_desc;
215 ap.a_head.a_dev = dev;
216 return (dev->si_ops->d_clone(&ap));
220 dev_drevoke(cdev_t dev)
222 struct dev_revoke_args ap;
224 ap.a_head.a_desc = &dev_revoke_desc;
225 ap.a_head.a_dev = dev;
226 return (dev->si_ops->d_revoke(&ap));
230 * Core device strategy call, used to issue I/O on a device. There are
231 * two versions, a non-chained version and a chained version. The chained
232 * version reuses a BIO set up by vn_strategy(). The only difference is
233 * that, for now, we do not push a new tracking structure when chaining
234 * from vn_strategy. XXX this will ultimately have to change.
237 dev_dstrategy(cdev_t dev, struct bio *bio)
239 struct dev_strategy_args ap;
240 struct bio_track *track;
242 ap.a_head.a_desc = &dev_strategy_desc;
243 ap.a_head.a_dev = dev;
246 KKASSERT(bio->bio_track == NULL);
247 KKASSERT(bio->bio_buf->b_cmd != BUF_CMD_DONE);
248 if (bio->bio_buf->b_cmd == BUF_CMD_READ)
249 track = &dev->si_track_read;
251 track = &dev->si_track_write;
252 bio_track_ref(track);
253 bio->bio_track = track;
255 if (dsched_is_clear_buf_priv(bio->bio_buf))
256 dsched_new_buf(bio->bio_buf);
258 KKASSERT((bio->bio_flags & BIO_DONE) == 0);
259 (void)dev->si_ops->d_strategy(&ap);
263 dev_dstrategy_chain(cdev_t dev, struct bio *bio)
265 struct dev_strategy_args ap;
267 ap.a_head.a_desc = &dev_strategy_desc;
268 ap.a_head.a_dev = dev;
271 KKASSERT(bio->bio_track != NULL);
272 KKASSERT((bio->bio_flags & BIO_DONE) == 0);
273 (void)dev->si_ops->d_strategy(&ap);
277 * note: the disk layer is expected to set count, blkno, and secsize before
278 * forwarding the message.
281 dev_ddump(cdev_t dev, void *virtual, vm_offset_t physical, off_t offset,
284 struct dev_dump_args ap;
286 ap.a_head.a_desc = &dev_dump_desc;
287 ap.a_head.a_dev = dev;
291 ap.a_virtual = virtual;
292 ap.a_physical = physical;
293 ap.a_offset = offset;
294 ap.a_length = length;
295 return(dev->si_ops->d_dump(&ap));
299 dev_dpsize(cdev_t dev)
301 struct dev_psize_args ap;
304 ap.a_head.a_desc = &dev_psize_desc;
305 ap.a_head.a_dev = dev;
306 error = dev->si_ops->d_psize(&ap);
308 return (ap.a_result);
313 dev_dkqfilter(cdev_t dev, struct knote *kn)
315 struct dev_kqfilter_args ap;
318 ap.a_head.a_desc = &dev_kqfilter_desc;
319 ap.a_head.a_dev = dev;
321 error = dev->si_ops->d_kqfilter(&ap);
327 /************************************************************************
328 * DEVICE HELPER FUNCTIONS *
329 ************************************************************************/
335 dev_drefs(cdev_t dev)
337 return(dev->si_sysref.refcnt);
344 dev_dname(cdev_t dev)
346 return(dev->si_ops->head.name);
353 dev_dflags(cdev_t dev)
355 return(dev->si_ops->head.flags);
364 return(dev->si_ops->head.maj);
368 * Used when forwarding a request through layers. The caller adjusts
369 * ap->a_head.a_dev and then calls this function.
372 dev_doperate(struct dev_generic_args *ap)
374 int (*func)(struct dev_generic_args *);
376 func = *(void **)((char *)ap->a_dev->si_ops + ap->a_desc->sd_offset);
381 * Used by the console intercept code only. Issue an operation through
382 * a foreign ops structure allowing the ops structure associated
383 * with the device to remain intact.
386 dev_doperate_ops(struct dev_ops *ops, struct dev_generic_args *ap)
388 int (*func)(struct dev_generic_args *);
390 func = *(void **)((char *)ops + ap->a_desc->sd_offset);
395 * Convert a template dev_ops into the real thing by filling in
396 * uninitialized fields.
399 compile_dev_ops(struct dev_ops *ops)
403 for (offset = offsetof(struct dev_ops, dev_ops_first_field);
404 offset <= offsetof(struct dev_ops, dev_ops_last_field);
405 offset += sizeof(void *)
407 void **func_p = (void **)((char *)ops + offset);
408 void **def_p = (void **)((char *)&default_dev_ops + offset);
409 if (*func_p == NULL) {
411 *func_p = ops->d_default;
418 /************************************************************************
419 * MAJOR/MINOR SPACE FUNCTION *
420 ************************************************************************/
423 * This makes a dev_ops entry visible to userland (e.g /dev/<blah>).
425 * Disk devices typically register their major, e.g. 'ad0', and then call
426 * into the disk label management code which overloads its own onto e.g. 'ad0'
427 * to support all the various slice and partition combinations.
429 * The mask/match supplied in this call are a full 32 bits and the same
430 * mask and match must be specified in a later dev_ops_remove() call to
431 * match this add. However, the match value for the minor number should never
432 * have any bits set in the major number's bit range (8-15). The mask value
433 * may be conveniently specified as -1 without creating any major number
439 rb_dev_ops_compare(struct dev_ops_maj *a, struct dev_ops_maj *b)
443 else if (a->maj > b->maj)
448 RB_GENERATE2(dev_ops_rb_tree, dev_ops_maj, rbnode, rb_dev_ops_compare, int, maj);
450 struct dev_ops_rb_tree dev_ops_rbhead = RB_INITIALIZER(dev_ops_rbhead);
453 dev_ops_remove_all(struct dev_ops *ops)
455 return devfs_destroy_dev_by_ops(ops, -1);
459 dev_ops_remove_minor(struct dev_ops *ops, int minor)
461 return devfs_destroy_dev_by_ops(ops, minor);
465 dev_ops_intercept(cdev_t dev, struct dev_ops *iops)
467 struct dev_ops *oops = dev->si_ops;
469 compile_dev_ops(iops);
470 iops->head.maj = oops->head.maj;
471 iops->head.data = oops->head.data;
472 iops->head.flags = oops->head.flags;
474 dev->si_flags |= SI_INTERCEPTED;
480 dev_ops_restore(cdev_t dev, struct dev_ops *oops)
482 struct dev_ops *iops = dev->si_ops;
485 dev->si_flags &= ~SI_INTERCEPTED;
487 iops->head.data = NULL;
488 iops->head.flags = 0;
491 /************************************************************************
492 * DEFAULT DEV OPS FUNCTIONS *
493 ************************************************************************/
497 * Unsupported devswitch functions (e.g. for writing to read-only device).
498 * XXX may belong elsewhere.
501 norevoke(struct dev_revoke_args *ap)
508 noclone(struct dev_clone_args *ap)
511 return (0); /* allow the clone */
515 noopen(struct dev_open_args *ap)
521 noclose(struct dev_close_args *ap)
527 noread(struct dev_read_args *ap)
533 nowrite(struct dev_write_args *ap)
539 noioctl(struct dev_ioctl_args *ap)
545 nokqfilter(struct dev_kqfilter_args *ap)
551 nommap(struct dev_mmap_args *ap)
557 nopoll(struct dev_poll_args *ap)
564 nostrategy(struct dev_strategy_args *ap)
566 struct bio *bio = ap->a_bio;
568 bio->bio_buf->b_flags |= B_ERROR;
569 bio->bio_buf->b_error = EOPNOTSUPP;
575 nopsize(struct dev_psize_args *ap)
582 nodump(struct dev_dump_args *ap)
588 * XXX this is probably bogus. Any device that uses it isn't checking the
592 nullopen(struct dev_open_args *ap)
598 nullclose(struct dev_close_args *ap)