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 <machine/stdarg.h>
47 #include <sys/devfs.h>
48 #include <sys/dsched.h>
50 #include <sys/thread2.h>
51 #include <sys/mplock2.h>
53 static int mpsafe_writes;
54 static int mplock_writes;
55 static int mpsafe_reads;
56 static int mplock_reads;
57 static int mpsafe_strategies;
58 static int mplock_strategies;
60 SYSCTL_INT(_kern, OID_AUTO, mpsafe_writes, CTLFLAG_RD, &mpsafe_writes,
62 SYSCTL_INT(_kern, OID_AUTO, mplock_writes, CTLFLAG_RD, &mplock_writes,
63 0, "non-mpsafe writes");
64 SYSCTL_INT(_kern, OID_AUTO, mpsafe_reads, CTLFLAG_RD, &mpsafe_reads,
66 SYSCTL_INT(_kern, OID_AUTO, mplock_reads, CTLFLAG_RD, &mplock_reads,
67 0, "non-mpsafe reads");
68 SYSCTL_INT(_kern, OID_AUTO, mpsafe_strategies, CTLFLAG_RD, &mpsafe_strategies,
69 0, "mpsafe strategies");
70 SYSCTL_INT(_kern, OID_AUTO, mplock_strategies, CTLFLAG_RD, &mplock_strategies,
71 0, "non-mpsafe strategies");
74 * system link descriptors identify the command in the
75 * arguments structure.
77 #define DDESCNAME(name) __CONCAT(__CONCAT(dev_,name),_desc)
79 #define DEVOP_DESC_INIT(name) \
80 struct syslink_desc DDESCNAME(name) = { \
81 __offsetof(struct dev_ops, __CONCAT(d_, name)), \
84 DEVOP_DESC_INIT(default);
85 DEVOP_DESC_INIT(open);
86 DEVOP_DESC_INIT(close);
87 DEVOP_DESC_INIT(read);
88 DEVOP_DESC_INIT(write);
89 DEVOP_DESC_INIT(ioctl);
90 DEVOP_DESC_INIT(dump);
91 DEVOP_DESC_INIT(psize);
92 DEVOP_DESC_INIT(mmap);
93 DEVOP_DESC_INIT(mmap_single);
94 DEVOP_DESC_INIT(strategy);
95 DEVOP_DESC_INIT(kqfilter);
96 DEVOP_DESC_INIT(revoke);
97 DEVOP_DESC_INIT(clone);
102 struct dev_ops dead_dev_ops;
104 struct dev_ops default_dev_ops = {
106 .d_default = NULL, /* must be NULL */
113 .d_mmap_single = nommap_single,
114 .d_strategy = nostrategy,
117 .d_kqfilter = nokqfilter,
118 .d_revoke = norevoke,
124 dev_needmplock(cdev_t dev)
126 return((dev->si_ops->head.flags & D_MPSAFE) == 0);
129 /************************************************************************
130 * GENERAL DEVICE API FUNCTIONS *
131 ************************************************************************
133 * The MPSAFEness of these depends on dev->si_ops->head.flags
136 dev_dopen(cdev_t dev, int oflags, int devtype, struct ucred *cred, struct file *fp)
138 struct dev_open_args ap;
139 int needmplock = dev_needmplock(dev);
142 ap.a_head.a_desc = &dev_open_desc;
143 ap.a_head.a_dev = dev;
144 ap.a_oflags = oflags;
145 ap.a_devtype = devtype;
151 error = dev->si_ops->d_open(&ap);
158 dev_dclose(cdev_t dev, int fflag, int devtype, struct file *fp)
160 struct dev_close_args ap;
161 int needmplock = dev_needmplock(dev);
164 ap.a_head.a_desc = &dev_close_desc;
165 ap.a_head.a_dev = dev;
167 ap.a_devtype = devtype;
172 error = dev->si_ops->d_close(&ap);
179 dev_dread(cdev_t dev, struct uio *uio, int ioflag, struct file *fp)
181 struct dev_read_args ap;
182 int needmplock = dev_needmplock(dev);
185 ap.a_head.a_desc = &dev_read_desc;
186 ap.a_head.a_dev = dev;
188 ap.a_ioflag = ioflag;
197 error = dev->si_ops->d_read(&ap);
201 dev->si_lastread = time_uptime;
206 dev_dwrite(cdev_t dev, struct uio *uio, int ioflag, struct file *fp)
208 struct dev_write_args ap;
209 int needmplock = dev_needmplock(dev);
212 dev->si_lastwrite = time_uptime;
213 ap.a_head.a_desc = &dev_write_desc;
214 ap.a_head.a_dev = dev;
216 ap.a_ioflag = ioflag;
225 error = dev->si_ops->d_write(&ap);
232 dev_dioctl(cdev_t dev, u_long cmd, caddr_t data, int fflag, struct ucred *cred,
233 struct sysmsg *msg, struct file *fp)
235 struct dev_ioctl_args ap;
236 int needmplock = dev_needmplock(dev);
239 ap.a_head.a_desc = &dev_ioctl_desc;
240 ap.a_head.a_dev = dev;
250 error = dev->si_ops->d_ioctl(&ap);
257 dev_dmmap(cdev_t dev, vm_offset_t offset, int nprot, struct file *fp)
259 struct dev_mmap_args ap;
260 int needmplock = dev_needmplock(dev);
263 ap.a_head.a_desc = &dev_mmap_desc;
264 ap.a_head.a_dev = dev;
265 ap.a_offset = offset;
271 error = dev->si_ops->d_mmap(&ap);
281 dev_dmmap_single(cdev_t dev, vm_ooffset_t *offset, vm_size_t size,
282 struct vm_object **object, int nprot, struct file *fp)
284 struct dev_mmap_single_args ap;
285 int needmplock = dev_needmplock(dev);
288 ap.a_head.a_desc = &dev_mmap_single_desc;
289 ap.a_head.a_dev = dev;
290 ap.a_offset = offset;
292 ap.a_object = object;
298 error = dev->si_ops->d_mmap_single(&ap);
306 dev_dclone(cdev_t dev)
308 struct dev_clone_args ap;
309 int needmplock = dev_needmplock(dev);
312 ap.a_head.a_desc = &dev_clone_desc;
313 ap.a_head.a_dev = dev;
317 error = dev->si_ops->d_clone(&ap);
324 dev_drevoke(cdev_t dev)
326 struct dev_revoke_args ap;
327 int needmplock = dev_needmplock(dev);
330 ap.a_head.a_desc = &dev_revoke_desc;
331 ap.a_head.a_dev = dev;
335 error = dev->si_ops->d_revoke(&ap);
343 * Core device strategy call, used to issue I/O on a device. There are
344 * two versions, a non-chained version and a chained version. The chained
345 * version reuses a BIO set up by vn_strategy(). The only difference is
346 * that, for now, we do not push a new tracking structure when chaining
347 * from vn_strategy. XXX this will ultimately have to change.
350 dev_dstrategy(cdev_t dev, struct bio *bio)
352 struct dev_strategy_args ap;
353 struct bio_track *track;
354 int needmplock = dev_needmplock(dev);
356 ap.a_head.a_desc = &dev_strategy_desc;
357 ap.a_head.a_dev = dev;
360 KKASSERT(bio->bio_track == NULL);
361 KKASSERT(bio->bio_buf->b_cmd != BUF_CMD_DONE);
362 if (bio->bio_buf->b_cmd == BUF_CMD_READ)
363 track = &dev->si_track_read;
365 track = &dev->si_track_write;
366 bio_track_ref(track);
367 bio->bio_track = track;
369 if (dsched_is_clear_buf_priv(bio->bio_buf))
370 dsched_new_buf(bio->bio_buf);
372 KKASSERT((bio->bio_flags & BIO_DONE) == 0);
379 (void)dev->si_ops->d_strategy(&ap);
385 dev_dstrategy_chain(cdev_t dev, struct bio *bio)
387 struct dev_strategy_args ap;
388 int needmplock = dev_needmplock(dev);
390 ap.a_head.a_desc = &dev_strategy_desc;
391 ap.a_head.a_dev = dev;
394 KKASSERT(bio->bio_track != NULL);
395 KKASSERT((bio->bio_flags & BIO_DONE) == 0);
398 (void)dev->si_ops->d_strategy(&ap);
404 * note: the disk layer is expected to set count, blkno, and secsize before
405 * forwarding the message.
408 dev_ddump(cdev_t dev, void *virtual, vm_offset_t physical, off_t offset,
411 struct dev_dump_args ap;
412 int needmplock = dev_needmplock(dev);
415 ap.a_head.a_desc = &dev_dump_desc;
416 ap.a_head.a_dev = dev;
420 ap.a_virtual = virtual;
421 ap.a_physical = physical;
422 ap.a_offset = offset;
423 ap.a_length = length;
427 error = dev->si_ops->d_dump(&ap);
434 dev_dpsize(cdev_t dev)
436 struct dev_psize_args ap;
437 int needmplock = dev_needmplock(dev);
440 ap.a_head.a_desc = &dev_psize_desc;
441 ap.a_head.a_dev = dev;
445 error = dev->si_ops->d_psize(&ap);
450 return (ap.a_result);
455 * Pass-thru to the device kqfilter.
457 * NOTE: We explicitly preset a_result to 0 so d_kqfilter() functions
458 * which return 0 do not have to bother setting a_result.
461 dev_dkqfilter(cdev_t dev, struct knote *kn, struct file *fp)
463 struct dev_kqfilter_args ap;
464 int needmplock = dev_needmplock(dev);
467 ap.a_head.a_desc = &dev_kqfilter_desc;
468 ap.a_head.a_dev = dev;
475 error = dev->si_ops->d_kqfilter(&ap);
484 /************************************************************************
485 * DEVICE HELPER FUNCTIONS *
486 ************************************************************************/
492 dev_drefs(cdev_t dev)
494 return(dev->si_sysref.refcnt);
501 dev_dname(cdev_t dev)
503 return(dev->si_ops->head.name);
510 dev_dflags(cdev_t dev)
512 return(dev->si_ops->head.flags);
521 return(dev->si_ops->head.maj);
525 * Used when forwarding a request through layers. The caller adjusts
526 * ap->a_head.a_dev and then calls this function.
529 dev_doperate(struct dev_generic_args *ap)
531 int (*func)(struct dev_generic_args *);
532 int needmplock = dev_needmplock(ap->a_dev);
535 func = *(void **)((char *)ap->a_dev->si_ops + ap->a_desc->sd_offset);
547 * Used by the console intercept code only. Issue an operation through
548 * a foreign ops structure allowing the ops structure associated
549 * with the device to remain intact.
552 dev_doperate_ops(struct dev_ops *ops, struct dev_generic_args *ap)
554 int (*func)(struct dev_generic_args *);
555 int needmplock = ((ops->head.flags & D_MPSAFE) == 0);
558 func = *(void **)((char *)ops + ap->a_desc->sd_offset);
570 * Convert a template dev_ops into the real thing by filling in
571 * uninitialized fields.
574 compile_dev_ops(struct dev_ops *ops)
578 for (offset = offsetof(struct dev_ops, dev_ops_first_field);
579 offset <= offsetof(struct dev_ops, dev_ops_last_field);
580 offset += sizeof(void *)
582 void **func_p = (void **)((char *)ops + offset);
583 void **def_p = (void **)((char *)&default_dev_ops + offset);
584 if (*func_p == NULL) {
586 *func_p = ops->d_default;
593 /************************************************************************
594 * MAJOR/MINOR SPACE FUNCTION *
595 ************************************************************************/
598 * This makes a dev_ops entry visible to userland (e.g /dev/<blah>).
600 * Disk devices typically register their major, e.g. 'ad0', and then call
601 * into the disk label management code which overloads its own onto e.g. 'ad0'
602 * to support all the various slice and partition combinations.
604 * The mask/match supplied in this call are a full 32 bits and the same
605 * mask and match must be specified in a later dev_ops_remove() call to
606 * match this add. However, the match value for the minor number should never
607 * have any bits set in the major number's bit range (8-15). The mask value
608 * may be conveniently specified as -1 without creating any major number
614 rb_dev_ops_compare(struct dev_ops_maj *a, struct dev_ops_maj *b)
618 else if (a->maj > b->maj)
623 RB_GENERATE2(dev_ops_rb_tree, dev_ops_maj, rbnode, rb_dev_ops_compare, int, maj);
625 struct dev_ops_rb_tree dev_ops_rbhead = RB_INITIALIZER(dev_ops_rbhead);
628 dev_ops_remove_all(struct dev_ops *ops)
630 return devfs_destroy_dev_by_ops(ops, -1);
634 dev_ops_remove_minor(struct dev_ops *ops, int minor)
636 return devfs_destroy_dev_by_ops(ops, minor);
640 dev_ops_intercept(cdev_t dev, struct dev_ops *iops)
642 struct dev_ops *oops = dev->si_ops;
644 compile_dev_ops(iops);
645 iops->head.maj = oops->head.maj;
646 iops->head.data = oops->head.data;
647 iops->head.flags = oops->head.flags;
649 dev->si_flags |= SI_INTERCEPTED;
655 dev_ops_restore(cdev_t dev, struct dev_ops *oops)
657 struct dev_ops *iops = dev->si_ops;
660 dev->si_flags &= ~SI_INTERCEPTED;
662 iops->head.data = NULL;
663 iops->head.flags = 0;
666 /************************************************************************
667 * DEFAULT DEV OPS FUNCTIONS *
668 ************************************************************************/
672 * Unsupported devswitch functions (e.g. for writing to read-only device).
673 * XXX may belong elsewhere.
676 norevoke(struct dev_revoke_args *ap)
683 noclone(struct dev_clone_args *ap)
686 return (0); /* allow the clone */
690 noopen(struct dev_open_args *ap)
696 noclose(struct dev_close_args *ap)
702 noread(struct dev_read_args *ap)
708 nowrite(struct dev_write_args *ap)
714 noioctl(struct dev_ioctl_args *ap)
720 nokqfilter(struct dev_kqfilter_args *ap)
726 nommap(struct dev_mmap_args *ap)
732 nommap_single(struct dev_mmap_single_args *ap)
738 nostrategy(struct dev_strategy_args *ap)
740 struct bio *bio = ap->a_bio;
742 bio->bio_buf->b_flags |= B_ERROR;
743 bio->bio_buf->b_error = EOPNOTSUPP;
749 nopsize(struct dev_psize_args *ap)
756 nodump(struct dev_dump_args *ap)
762 * XXX this is probably bogus. Any device that uses it isn't checking the
766 nullopen(struct dev_open_args *ap)
772 nullclose(struct dev_close_args *ap)