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
31 #include <sys/param.h>
32 #include <sys/systm.h>
33 #include <sys/kernel.h>
34 #include <sys/sysctl.h>
35 #include <sys/module.h>
36 #include <sys/malloc.h>
40 #include <sys/vnode.h>
41 #include <sys/queue.h>
42 #include <sys/device.h>
44 #include <sys/syslink_rpc.h>
46 #include <sys/dsched.h>
47 #include <sys/devfs.h>
49 #include <machine/stdarg.h>
51 #include <sys/thread2.h>
52 #include <sys/mplock2.h>
55 * system link descriptors identify the command in the
56 * arguments structure.
58 #define DDESCNAME(name) __CONCAT(__CONCAT(dev_,name),_desc)
60 #define DEVOP_DESC_INIT(name) \
61 struct syslink_desc DDESCNAME(name) = { \
62 __offsetof(struct dev_ops, __CONCAT(d_, name)), \
65 DEVOP_DESC_INIT(default);
66 DEVOP_DESC_INIT(open);
67 DEVOP_DESC_INIT(close);
68 DEVOP_DESC_INIT(read);
69 DEVOP_DESC_INIT(write);
70 DEVOP_DESC_INIT(ioctl);
71 DEVOP_DESC_INIT(dump);
72 DEVOP_DESC_INIT(psize);
73 DEVOP_DESC_INIT(mmap);
74 DEVOP_DESC_INIT(mmap_single);
75 DEVOP_DESC_INIT(strategy);
76 DEVOP_DESC_INIT(kqfilter);
77 DEVOP_DESC_INIT(revoke);
78 DEVOP_DESC_INIT(clone);
83 struct dev_ops dead_dev_ops;
85 static d_open_t noopen;
86 static d_close_t noclose;
87 static d_read_t noread;
88 static d_write_t nowrite;
89 static d_ioctl_t noioctl;
90 static d_mmap_t nommap;
91 static d_mmap_single_t nommap_single;
92 static d_strategy_t nostrategy;
93 static d_dump_t nodump;
94 static d_psize_t nopsize;
95 static d_kqfilter_t nokqfilter;
96 static d_clone_t noclone;
97 static d_revoke_t norevoke;
99 struct dev_ops default_dev_ops = {
101 .d_default = NULL, /* must be NULL */
108 .d_mmap_single = nommap_single,
109 .d_strategy = nostrategy,
112 .d_kqfilter = nokqfilter,
113 .d_revoke = norevoke,
119 dev_needmplock(cdev_t dev)
121 return((dev->si_ops->head.flags & D_MPSAFE) == 0);
126 dev_nokvabio(cdev_t dev)
128 return((dev->si_ops->head.flags & D_KVABIO) == 0);
131 /************************************************************************
132 * GENERAL DEVICE API FUNCTIONS *
133 ************************************************************************
135 * The MPSAFEness of these depends on dev->si_ops->head.flags
138 dev_dopen(cdev_t dev, int oflags, int devtype, struct ucred *cred,
141 struct dev_open_args ap;
142 int needmplock = dev_needmplock(dev);
145 ap.a_head.a_desc = &dev_open_desc;
146 ap.a_head.a_dev = dev;
147 ap.a_oflags = oflags;
148 ap.a_devtype = devtype;
154 error = dev->si_ops->d_open(&ap);
161 dev_dclose(cdev_t dev, int fflag, int devtype, struct file *fp)
163 struct dev_close_args ap;
164 int needmplock = dev_needmplock(dev);
167 ap.a_head.a_desc = &dev_close_desc;
168 ap.a_head.a_dev = dev;
170 ap.a_devtype = devtype;
175 error = dev->si_ops->d_close(&ap);
182 dev_dread(cdev_t dev, struct uio *uio, int ioflag, struct file *fp)
184 struct dev_read_args ap;
185 int needmplock = dev_needmplock(dev);
188 ap.a_head.a_desc = &dev_read_desc;
189 ap.a_head.a_dev = dev;
191 ap.a_ioflag = ioflag;
196 error = dev->si_ops->d_read(&ap);
200 dev->si_lastread = time_uptime;
205 dev_dwrite(cdev_t dev, struct uio *uio, int ioflag, struct file *fp)
207 struct dev_write_args ap;
208 int needmplock = dev_needmplock(dev);
211 dev->si_lastwrite = time_uptime;
212 ap.a_head.a_desc = &dev_write_desc;
213 ap.a_head.a_dev = dev;
215 ap.a_ioflag = ioflag;
220 error = dev->si_ops->d_write(&ap);
227 dev_dioctl(cdev_t dev, u_long cmd, caddr_t data, int fflag, struct ucred *cred,
228 struct sysmsg *msg, struct file *fp)
230 struct dev_ioctl_args ap;
231 int needmplock = dev_needmplock(dev);
234 ap.a_head.a_desc = &dev_ioctl_desc;
235 ap.a_head.a_dev = dev;
245 error = dev->si_ops->d_ioctl(&ap);
252 dev_dmmap(cdev_t dev, vm_offset_t offset, int nprot, struct file *fp)
254 struct dev_mmap_args ap;
255 int needmplock = dev_needmplock(dev);
258 ap.a_head.a_desc = &dev_mmap_desc;
259 ap.a_head.a_dev = dev;
260 ap.a_offset = offset;
266 error = dev->si_ops->d_mmap(&ap);
276 dev_dmmap_single(cdev_t dev, vm_ooffset_t *offset, vm_size_t size,
277 struct vm_object **object, int nprot, struct file *fp)
279 struct dev_mmap_single_args ap;
280 int needmplock = dev_needmplock(dev);
283 ap.a_head.a_desc = &dev_mmap_single_desc;
284 ap.a_head.a_dev = dev;
285 ap.a_offset = offset;
287 ap.a_object = object;
293 error = dev->si_ops->d_mmap_single(&ap);
301 dev_dclone(cdev_t dev)
303 struct dev_clone_args ap;
304 int needmplock = dev_needmplock(dev);
307 ap.a_head.a_desc = &dev_clone_desc;
308 ap.a_head.a_dev = dev;
312 error = dev->si_ops->d_clone(&ap);
319 dev_drevoke(cdev_t dev)
321 struct dev_revoke_args ap;
322 int needmplock = dev_needmplock(dev);
325 ap.a_head.a_desc = &dev_revoke_desc;
326 ap.a_head.a_dev = dev;
330 error = dev->si_ops->d_revoke(&ap);
338 * Core device strategy call, used to issue I/O on a device. There are
339 * two versions, a non-chained version and a chained version. The chained
340 * version reuses a BIO set up by vn_strategy(). The only difference is
341 * that, for now, we do not push a new tracking structure when chaining
342 * from vn_strategy. XXX this will ultimately have to change.
345 dev_dstrategy(cdev_t dev, struct bio *bio)
347 struct dev_strategy_args ap;
348 struct bio_track *track;
349 struct buf *bp = bio->bio_buf;
350 int needmplock = dev_needmplock(dev);
353 * If the device doe snot support KVABIO and the buffer is using
354 * KVABIO, we must synchronize b_data to all cpus before dispatching.
356 if (dev_nokvabio(dev) && (bp->b_flags & B_KVABIO))
359 ap.a_head.a_desc = &dev_strategy_desc;
360 ap.a_head.a_dev = dev;
363 KKASSERT(bio->bio_track == NULL);
364 KKASSERT(bp->b_cmd != BUF_CMD_DONE);
365 if (bp->b_cmd == BUF_CMD_READ)
366 track = &dev->si_track_read;
368 track = &dev->si_track_write;
369 bio_track_ref(track);
370 bio->bio_track = track;
371 dsched_buf_enter(bp); /* might stack */
373 KKASSERT((bio->bio_flags & BIO_DONE) == 0);
376 (void)dev->si_ops->d_strategy(&ap);
382 dev_dstrategy_chain(cdev_t dev, struct bio *bio)
384 struct dev_strategy_args ap;
385 struct buf *bp = bio->bio_buf;
386 int needmplock = dev_needmplock(dev);
389 * If the device doe snot support KVABIO and the buffer is using
390 * KVABIO, we must synchronize b_data to all cpus before dispatching.
392 if (dev_nokvabio(dev) && (bp->b_flags & B_KVABIO))
395 ap.a_head.a_desc = &dev_strategy_desc;
396 ap.a_head.a_dev = dev;
399 KKASSERT(bio->bio_track != NULL);
400 KKASSERT((bio->bio_flags & BIO_DONE) == 0);
403 (void)dev->si_ops->d_strategy(&ap);
409 * note: the disk layer is expected to set count, blkno, and secsize before
410 * forwarding the message.
413 dev_ddump(cdev_t dev, void *virtual, vm_offset_t physical, off_t offset,
416 struct dev_dump_args ap;
417 int needmplock = dev_needmplock(dev);
420 ap.a_head.a_desc = &dev_dump_desc;
421 ap.a_head.a_dev = dev;
425 ap.a_virtual = virtual;
426 ap.a_physical = physical;
427 ap.a_offset = offset;
428 ap.a_length = length;
432 error = dev->si_ops->d_dump(&ap);
439 dev_dpsize(cdev_t dev)
441 struct dev_psize_args ap;
442 int needmplock = dev_needmplock(dev);
445 ap.a_head.a_desc = &dev_psize_desc;
446 ap.a_head.a_dev = dev;
450 error = dev->si_ops->d_psize(&ap);
455 return (ap.a_result);
460 * Pass-thru to the device kqfilter.
462 * NOTE: We explicitly preset a_result to 0 so d_kqfilter() functions
463 * which return 0 do not have to bother setting a_result.
466 dev_dkqfilter(cdev_t dev, struct knote *kn, struct file *fp)
468 struct dev_kqfilter_args ap;
469 int needmplock = dev_needmplock(dev);
472 ap.a_head.a_desc = &dev_kqfilter_desc;
473 ap.a_head.a_dev = dev;
480 error = dev->si_ops->d_kqfilter(&ap);
489 /************************************************************************
490 * DEVICE HELPER FUNCTIONS *
491 ************************************************************************/
497 dev_drefs(cdev_t dev)
499 return(dev->si_sysref.refcnt);
506 dev_dname(cdev_t dev)
508 return(dev->si_ops->head.name);
515 dev_dflags(cdev_t dev)
517 return(dev->si_ops->head.flags);
526 return(dev->si_ops->head.maj);
530 * Used when forwarding a request through layers. The caller adjusts
531 * ap->a_head.a_dev and then calls this function.
534 dev_doperate(struct dev_generic_args *ap)
536 int (*func)(struct dev_generic_args *);
537 int needmplock = dev_needmplock(ap->a_dev);
540 func = *(void **)((char *)ap->a_dev->si_ops + ap->a_desc->sd_offset);
552 * Used by the console intercept code only. Issue an operation through
553 * a foreign ops structure allowing the ops structure associated
554 * with the device to remain intact.
557 dev_doperate_ops(struct dev_ops *ops, struct dev_generic_args *ap)
559 int (*func)(struct dev_generic_args *);
560 int needmplock = ((ops->head.flags & D_MPSAFE) == 0);
563 func = *(void **)((char *)ops + ap->a_desc->sd_offset);
575 * Convert a template dev_ops into the real thing by filling in
576 * uninitialized fields.
579 compile_dev_ops(struct dev_ops *ops)
583 for (offset = offsetof(struct dev_ops, dev_ops_first_field);
584 offset <= offsetof(struct dev_ops, dev_ops_last_field);
585 offset += sizeof(void *)
587 void **func_p = (void **)((char *)ops + offset);
588 void **def_p = (void **)((char *)&default_dev_ops + offset);
589 if (*func_p == NULL) {
591 *func_p = ops->d_default;
598 /************************************************************************
599 * MAJOR/MINOR SPACE FUNCTION *
600 ************************************************************************/
603 * This makes a dev_ops entry visible to userland (e.g /dev/<blah>).
605 * Disk devices typically register their major, e.g. 'ad0', and then call
606 * into the disk label management code which overloads its own onto e.g. 'ad0'
607 * to support all the various slice and partition combinations.
609 * The mask/match supplied in this call are a full 32 bits and the same
610 * mask and match must be specified in a later dev_ops_remove() call to
611 * match this add. However, the match value for the minor number should never
612 * have any bits set in the major number's bit range (8-15). The mask value
613 * may be conveniently specified as -1 without creating any major number
619 rb_dev_ops_compare(struct dev_ops_maj *a, struct dev_ops_maj *b)
623 else if (a->maj > b->maj)
628 RB_GENERATE2(dev_ops_rb_tree, dev_ops_maj, rbnode, rb_dev_ops_compare, int, maj);
630 struct dev_ops_rb_tree dev_ops_rbhead = RB_INITIALIZER(dev_ops_rbhead);
633 dev_ops_remove_all(struct dev_ops *ops)
635 return devfs_destroy_dev_by_ops(ops, -1);
639 dev_ops_remove_minor(struct dev_ops *ops, int minor)
641 return devfs_destroy_dev_by_ops(ops, minor);
645 dev_ops_intercept(cdev_t dev, struct dev_ops *iops)
647 struct dev_ops *oops = dev->si_ops;
649 compile_dev_ops(iops);
650 iops->head.maj = oops->head.maj;
651 iops->head.data = oops->head.data;
652 iops->head.flags = oops->head.flags;
654 dev->si_flags |= SI_INTERCEPTED;
660 dev_ops_restore(cdev_t dev, struct dev_ops *oops)
662 struct dev_ops *iops = dev->si_ops;
665 dev->si_flags &= ~SI_INTERCEPTED;
667 iops->head.data = NULL;
668 iops->head.flags = 0;
671 /************************************************************************
672 * DEFAULT DEV OPS FUNCTIONS *
673 ************************************************************************/
677 * Unsupported devswitch functions (e.g. for writing to read-only device).
678 * XXX may belong elsewhere.
681 norevoke(struct dev_revoke_args *ap)
688 noclone(struct dev_clone_args *ap)
691 return (0); /* allow the clone */
695 noopen(struct dev_open_args *ap)
701 noclose(struct dev_close_args *ap)
707 noread(struct dev_read_args *ap)
713 nowrite(struct dev_write_args *ap)
719 noioctl(struct dev_ioctl_args *ap)
725 nokqfilter(struct dev_kqfilter_args *ap)
731 nommap(struct dev_mmap_args *ap)
737 nommap_single(struct dev_mmap_single_args *ap)
743 nostrategy(struct dev_strategy_args *ap)
745 struct bio *bio = ap->a_bio;
747 bio->bio_buf->b_flags |= B_ERROR;
748 bio->bio_buf->b_error = EOPNOTSUPP;
754 nopsize(struct dev_psize_args *ap)
761 nodump(struct dev_dump_args *ap)
767 * XXX this is probably bogus. Any device that uses it isn't checking the
771 nullopen(struct dev_open_args *ap)
777 nullclose(struct dev_close_args *ap)