sys/kern/kern_device.c

   1 /*
   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,
   5  *                                                      All rights reserved.
   6  * Copyright (c) 1982, 1986, 1991, 1993
   7  *      The Regents of the University of California.  All rights reserved.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  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.
  17  *
  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
  28  * SUCH DAMAGE.
  29  *
  30  * $DragonFly: src/sys/kern/kern_device.c,v 1.27 2007/07/23 18:59:50 dillon Exp $
  31  */
  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>
  39 #include <sys/conf.h>
  40 #include <sys/bio.h>
  41 #include <sys/buf.h>
  42 #include <sys/vnode.h>
  43 #include <sys/queue.h>
  44 #include <sys/device.h>
  45 #include <sys/tree.h>
  46 #include <sys/syslink_rpc.h>
  47 #include <sys/proc.h>
  48 #include <machine/stdarg.h>
  49 #include <sys/thread2.h>
  50 #include <sys/devfs.h>
  51
  52 /*
  53  * system link descriptors identify the command in the
  54  * arguments structure.
  55  */
  56 #define DDESCNAME(name) __CONCAT(__CONCAT(dev_,name),_desc)
  57
  58 #define DEVOP_DESC_INIT(name)                                           \
  59             struct syslink_desc DDESCNAME(name) = {                     \
  60                 __offsetof(struct dev_ops, __CONCAT(d_, name)), \
  61             #name }
  62
  63 DEVOP_DESC_INIT(default);
  64 DEVOP_DESC_INIT(open);
  65 DEVOP_DESC_INIT(close);
  66 DEVOP_DESC_INIT(read);
  67 DEVOP_DESC_INIT(write);
  68 DEVOP_DESC_INIT(ioctl);
  69 DEVOP_DESC_INIT(dump);
  70 DEVOP_DESC_INIT(psize);
  71 DEVOP_DESC_INIT(poll);
  72 DEVOP_DESC_INIT(mmap);
  73 DEVOP_DESC_INIT(strategy);
  74 DEVOP_DESC_INIT(kqfilter);
  75 DEVOP_DESC_INIT(revoke);
  76 DEVOP_DESC_INIT(clone);
  77
  78 /*
  79  * Misc default ops
  80  */
  81 struct dev_ops dead_dev_ops;
  82
  83 struct dev_ops default_dev_ops = {
  84         { "null" },
  85         .d_default = NULL,      /* must be NULL */
  86         .d_open = noopen,
  87         .d_close = noclose,
  88         .d_read = noread,
  89         .d_write = nowrite,
  90         .d_ioctl = noioctl,
  91         .d_poll = nopoll,
  92         .d_mmap = nommap,
  93         .d_strategy = nostrategy,
  94         .d_dump = nodump,
  95         .d_psize = nopsize,
  96         .d_kqfilter = nokqfilter,
  97         .d_revoke = norevoke,
  98         .d_clone = noclone
  99 };
 100
 101 /************************************************************************
 102  *                      GENERAL DEVICE API FUNCTIONS                    *
 103  ************************************************************************/
 104
 105 int
 106 dev_dopen(cdev_t dev, int oflags, int devtype, struct ucred *cred)
 107 {
 108         struct dev_open_args ap;
 109
 110         ap.a_head.a_desc = &dev_open_desc;
 111         ap.a_head.a_dev = dev;
 112         ap.a_oflags = oflags;
 113         ap.a_devtype = devtype;
 114         ap.a_cred = cred;
 115         return(dev->si_ops->d_open(&ap));
 116 }
 117
 118 int
 119 dev_dclose(cdev_t dev, int fflag, int devtype)
 120 {
 121         struct dev_close_args ap;
 122
 123         ap.a_head.a_desc = &dev_close_desc;
 124         ap.a_head.a_dev = dev;
 125         ap.a_fflag = fflag;
 126         ap.a_devtype = devtype;
 127         return(dev->si_ops->d_close(&ap));
 128 }
 129
 130 int
 131 dev_dread(cdev_t dev, struct uio *uio, int ioflag)
 132 {
 133         struct dev_read_args ap;
 134         int error;
 135
 136         ap.a_head.a_desc = &dev_read_desc;
 137         ap.a_head.a_dev = dev;
 138         ap.a_uio = uio;
 139         ap.a_ioflag = ioflag;
 140         error = dev->si_ops->d_read(&ap);
 141         if (error == 0)
 142                 dev->si_lastread = time_second;
 143         return (error);
 144 }
 145
 146 int
 147 dev_dwrite(cdev_t dev, struct uio *uio, int ioflag)
 148 {
 149         struct dev_write_args ap;
 150         int error;
 151
 152         dev->si_lastwrite = time_second;
 153         ap.a_head.a_desc = &dev_write_desc;
 154         ap.a_head.a_dev = dev;
 155         ap.a_uio = uio;
 156         ap.a_ioflag = ioflag;
 157         error = dev->si_ops->d_write(&ap);
 158         return (error);
 159 }
 160
 161 int
 162 dev_dioctl(cdev_t dev, u_long cmd, caddr_t data, int fflag, struct ucred *cred,
 163            struct sysmsg *msg)
 164 {
 165         struct dev_ioctl_args ap;
 166
 167         ap.a_head.a_desc = &dev_ioctl_desc;
 168         ap.a_head.a_dev = dev;
 169         ap.a_cmd = cmd;
 170         ap.a_data = data;
 171         ap.a_fflag = fflag;
 172         ap.a_cred = cred;
 173         ap.a_sysmsg = msg;
 174         return(dev->si_ops->d_ioctl(&ap));
 175 }
 176
 177 int
 178 dev_dpoll(cdev_t dev, int events)
 179 {
 180         struct dev_poll_args ap;
 181         int error;
 182
 183         ap.a_head.a_desc = &dev_poll_desc;
 184         ap.a_head.a_dev = dev;
 185         ap.a_events = events;
 186         error = dev->si_ops->d_poll(&ap);
 187         if (error == 0)
 188                 return(ap.a_events);
 189         return (seltrue(dev, events));
 190 }
 191
 192 int
 193 dev_dmmap(cdev_t dev, vm_offset_t offset, int nprot)
 194 {
 195         struct dev_mmap_args ap;
 196         int error;
 197
 198         ap.a_head.a_desc = &dev_mmap_desc;
 199         ap.a_head.a_dev = dev;
 200         ap.a_offset = offset;
 201         ap.a_nprot = nprot;
 202         error = dev->si_ops->d_mmap(&ap);
 203         if (error == 0)
 204                 return(ap.a_result);
 205         return(-1);
 206 }
 207
 208 int
 209 dev_dclone(cdev_t dev)
 210 {
 211         struct dev_clone_args ap;
 212
 213         ap.a_head.a_desc = &dev_clone_desc;
 214         ap.a_head.a_dev = dev;
 215         return (dev->si_ops->d_clone(&ap));
 216 }
 217
 218 int
 219 dev_drevoke(cdev_t dev)
 220 {
 221         struct dev_revoke_args ap;
 222
 223         ap.a_head.a_desc = &dev_revoke_desc;
 224         ap.a_head.a_dev = dev;
 225         return (dev->si_ops->d_revoke(&ap));
 226 }
 227
 228 /*
 229  * Core device strategy call, used to issue I/O on a device.  There are
 230  * two versions, a non-chained version and a chained version.  The chained
 231  * version reuses a BIO set up by vn_strategy().  The only difference is
 232  * that, for now, we do not push a new tracking structure when chaining
 233  * from vn_strategy.  XXX this will ultimately have to change.
 234  */
 235 void
 236 dev_dstrategy(cdev_t dev, struct bio *bio)
 237 {
 238         struct dev_strategy_args ap;
 239         struct bio_track *track;
 240
 241         ap.a_head.a_desc = &dev_strategy_desc;
 242         ap.a_head.a_dev = dev;
 243         ap.a_bio = bio;
 244
 245         KKASSERT(bio->bio_track == NULL);
 246         KKASSERT(bio->bio_buf->b_cmd != BUF_CMD_DONE);
 247         if (bio->bio_buf->b_cmd == BUF_CMD_READ)
 248             track = &dev->si_track_read;
 249         else
 250             track = &dev->si_track_write;
 251         bio_track_ref(track);
 252         bio->bio_track = track;
 253         KKASSERT((bio->bio_flags & BIO_DONE) == 0);
 254         (void)dev->si_ops->d_strategy(&ap);
 255 }
 256
 257 void
 258 dev_dstrategy_chain(cdev_t dev, struct bio *bio)
 259 {
 260         struct dev_strategy_args ap;
 261
 262         ap.a_head.a_desc = &dev_strategy_desc;
 263         ap.a_head.a_dev = dev;
 264         ap.a_bio = bio;
 265
 266         KKASSERT(bio->bio_track != NULL);
 267         KKASSERT((bio->bio_flags & BIO_DONE) == 0);
 268         (void)dev->si_ops->d_strategy(&ap);
 269 }
 270
 271 /*
 272  * note: the disk layer is expected to set count, blkno, and secsize before
 273  * forwarding the message.
 274  */
 275 int
 276 dev_ddump(cdev_t dev, void *virtual, vm_offset_t physical, off_t offset,
 277     size_t length)
 278 {
 279         struct dev_dump_args ap;
 280
 281         ap.a_head.a_desc = &dev_dump_desc;
 282         ap.a_head.a_dev = dev;
 283         ap.a_count = 0;
 284         ap.a_blkno = 0;
 285         ap.a_secsize = 0;
 286         ap.a_virtual = virtual;
 287         ap.a_physical = physical;
 288         ap.a_offset = offset;
 289         ap.a_length = length;
 290         return(dev->si_ops->d_dump(&ap));
 291 }
 292
 293 int64_t
 294 dev_dpsize(cdev_t dev)
 295 {
 296         struct dev_psize_args ap;
 297         int error;
 298
 299         ap.a_head.a_desc = &dev_psize_desc;
 300         ap.a_head.a_dev = dev;
 301         error = dev->si_ops->d_psize(&ap);
 302         if (error == 0)
 303                 return (ap.a_result);
 304         return(-1);
 305 }
 306
 307 int
 308 dev_dkqfilter(cdev_t dev, struct knote *kn)
 309 {
 310         struct dev_kqfilter_args ap;
 311         int error;
 312
 313         ap.a_head.a_desc = &dev_kqfilter_desc;
 314         ap.a_head.a_dev = dev;
 315         ap.a_kn = kn;
 316         error = dev->si_ops->d_kqfilter(&ap);
 317         if (error == 0)
 318                 return(ap.a_result);
 319         return(ENODEV);
 320 }
 321
 322 /************************************************************************
 323  *                      DEVICE HELPER FUNCTIONS                         *
 324  ************************************************************************/
 325
 326 /*
 327  * MPSAFE
 328  */
 329 int
 330 dev_drefs(cdev_t dev)
 331 {
 332     return(dev->si_sysref.refcnt);
 333 }
 334
 335 /*
 336  * MPSAFE
 337  */
 338 const char *
 339 dev_dname(cdev_t dev)
 340 {
 341     return(dev->si_ops->head.name);
 342 }
 343
 344 /*
 345  * MPSAFE
 346  */
 347 int
 348 dev_dflags(cdev_t dev)
 349 {
 350     return(dev->si_ops->head.flags);
 351 }
 352
 353 /*
 354  * MPSAFE
 355  */
 356 int
 357 dev_dmaj(cdev_t dev)
 358 {
 359     return(dev->si_ops->head.maj);
 360 }
 361
 362 /*
 363  * Used when forwarding a request through layers.  The caller adjusts
 364  * ap->a_head.a_dev and then calls this function.
 365  */
 366 int
 367 dev_doperate(struct dev_generic_args *ap)
 368 {
 369     int (*func)(struct dev_generic_args *);
 370
 371     func = *(void **)((char *)ap->a_dev->si_ops + ap->a_desc->sd_offset);
 372     return (func(ap));
 373 }
 374
 375 /*
 376  * Used by the console intercept code only.  Issue an operation through
 377  * a foreign ops structure allowing the ops structure associated
 378  * with the device to remain intact.
 379  */
 380 int
 381 dev_doperate_ops(struct dev_ops *ops, struct dev_generic_args *ap)
 382 {
 383     int (*func)(struct dev_generic_args *);
 384
 385     func = *(void **)((char *)ops + ap->a_desc->sd_offset);
 386     return (func(ap));
 387 }
 388
 389 /*
 390  * Convert a template dev_ops into the real thing by filling in
 391  * uninitialized fields.
 392  */
 393 void
 394 compile_dev_ops(struct dev_ops *ops)
 395 {
 396         int offset;
 397
 398         for (offset = offsetof(struct dev_ops, dev_ops_first_field);
 399              offset <= offsetof(struct dev_ops, dev_ops_last_field);
 400              offset += sizeof(void *)
 401         ) {
 402                 void **func_p = (void **)((char *)ops + offset);
 403                 void **def_p = (void **)((char *)&default_dev_ops + offset);
 404                 if (*func_p == NULL) {
 405                         if (ops->d_default)
 406                                 *func_p = ops->d_default;
 407                         else
 408                                 *func_p = *def_p;
 409                 }
 410         }
 411 }
 412
 413 /************************************************************************
 414  *                      MAJOR/MINOR SPACE FUNCTION                      *
 415  ************************************************************************/
 416
 417 /*
 418  * This makes a dev_ops entry visible to userland (e.g /dev/<blah>).
 419  *
 420  * Disk devices typically register their major, e.g. 'ad0', and then call
 421  * into the disk label management code which overloads its own onto e.g. 'ad0'
 422  * to support all the various slice and partition combinations.
 423  *
 424  * The mask/match supplied in this call are a full 32 bits and the same
 425  * mask and match must be specified in a later dev_ops_remove() call to
 426  * match this add.  However, the match value for the minor number should never
 427  * have any bits set in the major number's bit range (8-15).  The mask value
 428  * may be conveniently specified as -1 without creating any major number
 429  * interference.
 430  */
 431
 432 static
 433 int
 434 rb_dev_ops_compare(struct dev_ops_maj *a, struct dev_ops_maj *b)
 435 {
 436     if (a->maj < b->maj)
 437         return(-1);
 438     else if (a->maj > b->maj)
 439         return(1);
 440     return(0);
 441 }
 442
 443 RB_GENERATE2(dev_ops_rb_tree, dev_ops_maj, rbnode, rb_dev_ops_compare, int, maj);
 444
 445 struct dev_ops_rb_tree dev_ops_rbhead = RB_INITIALIZER(dev_ops_rbhead);
 446
 447 int
 448 dev_ops_remove_all(struct dev_ops *ops)
 449 {
 450         return devfs_destroy_dev_by_ops(ops, -1);
 451 }
 452
 453 int
 454 dev_ops_remove_minor(struct dev_ops *ops, int minor)
 455 {
 456         return devfs_destroy_dev_by_ops(ops, minor);
 457 }
 458
 459 struct dev_ops *
 460 dev_ops_intercept(cdev_t dev, struct dev_ops *iops)
 461 {
 462         struct dev_ops *oops = dev->si_ops;
 463
 464         compile_dev_ops(iops);
 465         iops->head.maj = oops->head.maj;
 466         iops->head.data = oops->head.data;
 467         iops->head.flags = oops->head.flags;
 468         dev->si_ops = iops;
 469         dev->si_flags |= SI_INTERCEPTED;
 470
 471         return (oops);
 472 }
 473
 474 void
 475 dev_ops_restore(cdev_t dev, struct dev_ops *oops)
 476 {
 477         struct dev_ops *iops = dev->si_ops;
 478
 479         dev->si_ops = oops;
 480         dev->si_flags &= ~SI_INTERCEPTED;
 481         iops->head.maj = 0;
 482         iops->head.data = NULL;
 483         iops->head.flags = 0;
 484 }
 485
 486 /************************************************************************
 487  *                      DEFAULT DEV OPS FUNCTIONS                       *
 488  ************************************************************************/
 489
 490
 491 /*
 492  * Unsupported devswitch functions (e.g. for writing to read-only device).
 493  * XXX may belong elsewhere.
 494  */
 495 int
 496 norevoke(struct dev_revoke_args *ap)
 497 {
 498         /* take no action */
 499         return(0);
 500 }
 501
 502 int
 503 noclone(struct dev_clone_args *ap)
 504 {
 505         /* take no action */
 506         return (0);     /* allow the clone */
 507 }
 508
 509 int
 510 noopen(struct dev_open_args *ap)
 511 {
 512         return (ENODEV);
 513 }
 514
 515 int
 516 noclose(struct dev_close_args *ap)
 517 {
 518         return (ENODEV);
 519 }
 520
 521 int
 522 noread(struct dev_read_args *ap)
 523 {
 524         return (ENODEV);
 525 }
 526
 527 int
 528 nowrite(struct dev_write_args *ap)
 529 {
 530         return (ENODEV);
 531 }
 532
 533 int
 534 noioctl(struct dev_ioctl_args *ap)
 535 {
 536         return (ENODEV);
 537 }
 538
 539 int
 540 nokqfilter(struct dev_kqfilter_args *ap)
 541 {
 542         return (ENODEV);
 543 }
 544
 545 int
 546 nommap(struct dev_mmap_args *ap)
 547 {
 548         return (ENODEV);
 549 }
 550
 551 int
 552 nopoll(struct dev_poll_args *ap)
 553 {
 554         ap->a_events = 0;
 555         return(0);
 556 }
 557
 558 int
 559 nostrategy(struct dev_strategy_args *ap)
 560 {
 561         struct bio *bio = ap->a_bio;
 562
 563         bio->bio_buf->b_flags |= B_ERROR;
 564         bio->bio_buf->b_error = EOPNOTSUPP;
 565         biodone(bio);
 566         return(0);
 567 }
 568
 569 int
 570 nopsize(struct dev_psize_args *ap)
 571 {
 572         ap->a_result = 0;
 573         return(0);
 574 }
 575
 576 int
 577 nodump(struct dev_dump_args *ap)
 578 {
 579         return (ENODEV);
 580 }
 581
 582 /*
 583  * XXX this is probably bogus.  Any device that uses it isn't checking the
 584  * minor number.
 585  */
 586 int
 587 nullopen(struct dev_open_args *ap)
 588 {
 589         return (0);
 590 }
 591
 592 int
 593 nullclose(struct dev_close_args *ap)
 594 {
 595         return (0);
 596 }
 597