md.c revision 1.64
1 /* $NetBSD: md.c,v 1.64 2010/11/22 21:10:10 pooka Exp $ */ 2 3 /* 4 * Copyright (c) 1995 Gordon W. Ross, Leo Weppelman. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 19 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 21 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 22 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 23 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 25 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 */ 27 28 /* 29 * This implements a general-purpose memory-disk. 30 * See md.h for notes on the config types. 31 * 32 * Note that this driver provides the same functionality 33 * as the MFS filesystem hack, but this is better because 34 * you can use this for any filesystem type you'd like! 35 * 36 * Credit for most of the kmem ramdisk code goes to: 37 * Leo Weppelman (atari) and Phil Nelson (pc532) 38 * Credit for the ideas behind the "user space memory" code goes 39 * to the authors of the MFS implementation. 40 */ 41 42 #include <sys/cdefs.h> 43 __KERNEL_RCSID(0, "$NetBSD: md.c,v 1.64 2010/11/22 21:10:10 pooka Exp $"); 44 45 #ifdef _KERNEL_OPT 46 #include "opt_md.h" 47 #else 48 #define MEMORY_DISK_SERVER 1 49 #endif 50 51 #include <sys/param.h> 52 #include <sys/kernel.h> 53 #include <sys/malloc.h> 54 #include <sys/systm.h> 55 #include <sys/buf.h> 56 #include <sys/bufq.h> 57 #include <sys/device.h> 58 #include <sys/disk.h> 59 #include <sys/stat.h> 60 #include <sys/proc.h> 61 #include <sys/conf.h> 62 #include <sys/disklabel.h> 63 64 #include <uvm/uvm_extern.h> 65 66 #include <dev/md.h> 67 68 /* 69 * The user-space functionality is included by default. 70 * Use `options MEMORY_DISK_SERVER=0' to turn it off. 71 */ 72 #ifndef MEMORY_DISK_SERVER 73 #error MEMORY_DISK_SERVER should be defined by opt_md.h 74 #endif /* MEMORY_DISK_SERVER */ 75 76 /* 77 * We should use the raw partition for ioctl. 78 */ 79 #define MD_UNIT(unit) DISKUNIT(unit) 80 81 /* autoconfig stuff... */ 82 83 struct md_softc { 84 device_t sc_dev; /* Self. */ 85 struct disk sc_dkdev; /* hook for generic disk handling */ 86 struct md_conf sc_md; 87 struct bufq_state *sc_buflist; 88 }; 89 /* shorthand for fields in sc_md: */ 90 #define sc_addr sc_md.md_addr 91 #define sc_size sc_md.md_size 92 #define sc_type sc_md.md_type 93 94 void mdattach(int); 95 96 static void md_attach(device_t, device_t, void *); 97 static int md_detach(device_t, int); 98 99 static dev_type_open(mdopen); 100 static dev_type_close(mdclose); 101 static dev_type_read(mdread); 102 static dev_type_write(mdwrite); 103 static dev_type_ioctl(mdioctl); 104 static dev_type_strategy(mdstrategy); 105 static dev_type_size(mdsize); 106 107 const struct bdevsw md_bdevsw = { 108 mdopen, mdclose, mdstrategy, mdioctl, nodump, mdsize, D_DISK 109 }; 110 111 const struct cdevsw md_cdevsw = { 112 mdopen, mdclose, mdread, mdwrite, mdioctl, 113 nostop, notty, nopoll, nommap, nokqfilter, D_DISK 114 }; 115 116 static struct dkdriver mddkdriver = { mdstrategy, NULL }; 117 118 extern struct cfdriver md_cd; 119 CFATTACH_DECL3_NEW(md, sizeof(struct md_softc), 120 0, md_attach, md_detach, NULL, NULL, NULL, DVF_DETACH_SHUTDOWN); 121 122 static kmutex_t md_device_lock; /* Protect unit creation / deletion. */ 123 extern size_t md_root_size; 124 125 static void md_set_disklabel(struct md_softc *); 126 127 /* 128 * This is called if we are configured as a pseudo-device 129 */ 130 void 131 mdattach(int n) 132 { 133 134 mutex_init(&md_device_lock, MUTEX_DEFAULT, IPL_NONE); 135 if (config_cfattach_attach(md_cd.cd_name, &md_ca)) { 136 aprint_error("%s: cfattach_attach failed\n", md_cd.cd_name); 137 return; 138 } 139 } 140 141 static void 142 md_attach(device_t parent, device_t self, void *aux) 143 { 144 struct md_softc *sc = device_private(self); 145 146 sc->sc_dev = self; 147 bufq_alloc(&sc->sc_buflist, "fcfs", 0); 148 149 /* XXX - Could accept aux info here to set the config. */ 150 #ifdef MEMORY_DISK_HOOKS 151 /* 152 * This external function might setup a pre-loaded disk. 153 * All it would need to do is setup the md_conf struct. 154 * See sys/dev/md_root.c for an example. 155 */ 156 md_attach_hook(device_unit(self), &sc->sc_md); 157 #endif 158 159 /* 160 * Initialize and attach the disk structure. 161 */ 162 disk_init(&sc->sc_dkdev, device_xname(self), &mddkdriver); 163 disk_attach(&sc->sc_dkdev); 164 165 if (sc->sc_type != MD_UNCONFIGURED) 166 md_set_disklabel(sc); 167 168 if (!pmf_device_register(self, NULL, NULL)) 169 aprint_error_dev(self, "couldn't establish power handler\n"); 170 } 171 172 static int 173 md_detach(device_t self, int flags) 174 { 175 struct md_softc *sc = device_private(self); 176 int rc; 177 178 rc = 0; 179 mutex_enter(&sc->sc_dkdev.dk_openlock); 180 if (sc->sc_dkdev.dk_openmask == 0) 181 ; /* nothing to do */ 182 else if ((flags & DETACH_FORCE) == 0) 183 rc = EBUSY; 184 mutex_exit(&sc->sc_dkdev.dk_openlock); 185 186 if (rc != 0) 187 return rc; 188 189 pmf_device_deregister(self); 190 disk_detach(&sc->sc_dkdev); 191 disk_destroy(&sc->sc_dkdev); 192 bufq_free(sc->sc_buflist); 193 return 0; 194 } 195 196 /* 197 * operational routines: 198 * open, close, read, write, strategy, 199 * ioctl, dump, size 200 */ 201 202 #if MEMORY_DISK_SERVER 203 static int md_server_loop(struct md_softc *sc); 204 static int md_ioctl_server(struct md_softc *sc, struct md_conf *umd, 205 struct lwp *l); 206 #endif /* MEMORY_DISK_SERVER */ 207 static int md_ioctl_kalloc(struct md_softc *sc, struct md_conf *umd, 208 struct lwp *l); 209 210 static int 211 mdsize(dev_t dev) 212 { 213 struct md_softc *sc; 214 215 sc = device_lookup_private(&md_cd, MD_UNIT(dev)); 216 if (sc == NULL) 217 return 0; 218 219 if (sc->sc_type == MD_UNCONFIGURED) 220 return 0; 221 222 return (sc->sc_size >> DEV_BSHIFT); 223 } 224 225 static int 226 mdopen(dev_t dev, int flag, int fmt, struct lwp *l) 227 { 228 int unit; 229 int part = DISKPART(dev); 230 int pmask = 1 << part; 231 cfdata_t cf; 232 struct md_softc *sc; 233 struct disk *dk; 234 235 mutex_enter(&md_device_lock); 236 unit = MD_UNIT(dev); 237 sc = device_lookup_private(&md_cd, unit); 238 if (sc == NULL) { 239 if (part != RAW_PART) { 240 mutex_exit(&md_device_lock); 241 return ENXIO; 242 } 243 cf = malloc(sizeof(*cf), M_DEVBUF, M_WAITOK); 244 cf->cf_name = md_cd.cd_name; 245 cf->cf_atname = md_cd.cd_name; 246 cf->cf_unit = unit; 247 cf->cf_fstate = FSTATE_STAR; 248 sc = device_private(config_attach_pseudo(cf)); 249 if (sc == NULL) { 250 mutex_exit(&md_device_lock); 251 return ENOMEM; 252 } 253 } 254 255 dk = &sc->sc_dkdev; 256 257 /* 258 * The raw partition is used for ioctl to configure. 259 */ 260 if (part == RAW_PART) 261 goto ok; 262 263 #ifdef MEMORY_DISK_HOOKS 264 /* Call the open hook to allow loading the device. */ 265 md_open_hook(unit, &sc->sc_md); 266 #endif 267 268 /* 269 * This is a normal, "slave" device, so 270 * enforce initialized. 271 */ 272 if (sc->sc_type == MD_UNCONFIGURED) { 273 mutex_exit(&md_device_lock); 274 return ENXIO; 275 } 276 277 ok: 278 /* XXX duplicates code in dk_open(). Call dk_open(), instead? */ 279 mutex_enter(&dk->dk_openlock); 280 /* Mark our unit as open. */ 281 switch (fmt) { 282 case S_IFCHR: 283 dk->dk_copenmask |= pmask; 284 break; 285 case S_IFBLK: 286 dk->dk_bopenmask |= pmask; 287 break; 288 } 289 290 dk->dk_openmask = dk->dk_copenmask | dk->dk_bopenmask; 291 292 mutex_exit(&dk->dk_openlock); 293 mutex_exit(&md_device_lock); 294 return 0; 295 } 296 297 static int 298 mdclose(dev_t dev, int flag, int fmt, struct lwp *l) 299 { 300 int part = DISKPART(dev); 301 int pmask = 1 << part; 302 int error; 303 cfdata_t cf; 304 struct md_softc *sc; 305 struct disk *dk; 306 307 sc = device_lookup_private(&md_cd, MD_UNIT(dev)); 308 if (sc == NULL) 309 return ENXIO; 310 311 dk = &sc->sc_dkdev; 312 313 mutex_enter(&dk->dk_openlock); 314 315 switch (fmt) { 316 case S_IFCHR: 317 dk->dk_copenmask &= ~pmask; 318 break; 319 case S_IFBLK: 320 dk->dk_bopenmask &= ~pmask; 321 break; 322 } 323 dk->dk_openmask = dk->dk_copenmask | dk->dk_bopenmask; 324 325 mutex_exit(&dk->dk_openlock); 326 327 mutex_enter(&md_device_lock); 328 cf = device_cfdata(sc->sc_dev); 329 error = config_detach(sc->sc_dev, DETACH_QUIET); 330 if (! error) 331 free(cf, M_DEVBUF); 332 mutex_exit(&md_device_lock); 333 return error; 334 } 335 336 static int 337 mdread(dev_t dev, struct uio *uio, int flags) 338 { 339 struct md_softc *sc; 340 341 sc = device_lookup_private(&md_cd, MD_UNIT(dev)); 342 343 if (sc == NULL || sc->sc_type == MD_UNCONFIGURED) 344 return ENXIO; 345 346 return (physio(mdstrategy, NULL, dev, B_READ, minphys, uio)); 347 } 348 349 static int 350 mdwrite(dev_t dev, struct uio *uio, int flags) 351 { 352 struct md_softc *sc; 353 354 sc = device_lookup_private(&md_cd, MD_UNIT(dev)); 355 356 if (sc == NULL || sc->sc_type == MD_UNCONFIGURED) 357 return ENXIO; 358 359 return (physio(mdstrategy, NULL, dev, B_WRITE, minphys, uio)); 360 } 361 362 /* 363 * Handle I/O requests, either directly, or 364 * by passing them to the server process. 365 */ 366 static void 367 mdstrategy(struct buf *bp) 368 { 369 struct md_softc *sc; 370 void * addr; 371 size_t off, xfer; 372 bool is_read; 373 374 sc = device_lookup_private(&md_cd, MD_UNIT(bp->b_dev)); 375 376 if (sc == NULL || sc->sc_type == MD_UNCONFIGURED) { 377 bp->b_error = ENXIO; 378 goto done; 379 } 380 381 switch (sc->sc_type) { 382 #if MEMORY_DISK_SERVER 383 case MD_UMEM_SERVER: 384 /* Just add this job to the server's queue. */ 385 bufq_put(sc->sc_buflist, bp); 386 wakeup((void *)sc); 387 /* see md_server_loop() */ 388 /* no biodone in this case */ 389 return; 390 #endif /* MEMORY_DISK_SERVER */ 391 392 case MD_KMEM_FIXED: 393 case MD_KMEM_ALLOCATED: 394 /* These are in kernel space. Access directly. */ 395 is_read = ((bp->b_flags & B_READ) == B_READ); 396 bp->b_resid = bp->b_bcount; 397 off = (bp->b_blkno << DEV_BSHIFT); 398 if (off >= sc->sc_size) { 399 if (is_read) 400 break; /* EOF */ 401 goto set_eio; 402 } 403 xfer = bp->b_resid; 404 if (xfer > (sc->sc_size - off)) 405 xfer = (sc->sc_size - off); 406 addr = (char *)sc->sc_addr + off; 407 disk_busy(&sc->sc_dkdev); 408 if (is_read) 409 memcpy(bp->b_data, addr, xfer); 410 else 411 memcpy(addr, bp->b_data, xfer); 412 disk_unbusy(&sc->sc_dkdev, xfer, is_read); 413 bp->b_resid -= xfer; 414 break; 415 416 default: 417 bp->b_resid = bp->b_bcount; 418 set_eio: 419 bp->b_error = EIO; 420 break; 421 } 422 done: 423 biodone(bp); 424 } 425 426 static int 427 mdioctl(dev_t dev, u_long cmd, void *data, int flag, struct lwp *l) 428 { 429 struct md_softc *sc; 430 struct md_conf *umd; 431 struct disklabel *lp; 432 struct partinfo *pp; 433 434 if ((sc = device_lookup_private(&md_cd, MD_UNIT(dev))) == NULL) 435 return ENXIO; 436 437 if (sc->sc_type != MD_UNCONFIGURED) { 438 switch (cmd) { 439 case DIOCGDINFO: 440 lp = (struct disklabel *)data; 441 *lp = *sc->sc_dkdev.dk_label; 442 return 0; 443 444 case DIOCGPART: 445 pp = (struct partinfo *)data; 446 pp->disklab = sc->sc_dkdev.dk_label; 447 pp->part = 448 &sc->sc_dkdev.dk_label->d_partitions[DISKPART(dev)]; 449 return 0; 450 } 451 } 452 453 /* If this is not the raw partition, punt! */ 454 if (DISKPART(dev) != RAW_PART) 455 return ENOTTY; 456 457 umd = (struct md_conf *)data; 458 switch (cmd) { 459 case MD_GETCONF: 460 *umd = sc->sc_md; 461 return 0; 462 463 case MD_SETCONF: 464 /* Can only set it once. */ 465 if (sc->sc_type != MD_UNCONFIGURED) 466 break; 467 switch (umd->md_type) { 468 case MD_KMEM_ALLOCATED: 469 return md_ioctl_kalloc(sc, umd, l); 470 #if MEMORY_DISK_SERVER 471 case MD_UMEM_SERVER: 472 return md_ioctl_server(sc, umd, l); 473 #endif /* MEMORY_DISK_SERVER */ 474 default: 475 break; 476 } 477 break; 478 } 479 return EINVAL; 480 } 481 482 static void 483 md_set_disklabel(struct md_softc *sc) 484 { 485 struct disklabel *lp = sc->sc_dkdev.dk_label; 486 struct partition *pp; 487 488 memset(lp, 0, sizeof(*lp)); 489 490 lp->d_secsize = DEV_BSIZE; 491 lp->d_secperunit = sc->sc_size / DEV_BSIZE; 492 if (lp->d_secperunit >= (32*64)) { 493 lp->d_nsectors = 32; 494 lp->d_ntracks = 64; 495 lp->d_ncylinders = lp->d_secperunit / (32*64); 496 } else { 497 lp->d_nsectors = 1; 498 lp->d_ntracks = 1; 499 lp->d_ncylinders = lp->d_secperunit; 500 } 501 lp->d_secpercyl = lp->d_ntracks*lp->d_nsectors; 502 503 strncpy(lp->d_typename, md_cd.cd_name, sizeof(lp->d_typename)); 504 lp->d_type = DTYPE_UNKNOWN; 505 strncpy(lp->d_packname, "fictitious", sizeof(lp->d_packname)); 506 lp->d_rpm = 3600; 507 lp->d_interleave = 1; 508 lp->d_flags = 0; 509 510 pp = &lp->d_partitions[0]; 511 pp->p_offset = 0; 512 pp->p_size = lp->d_secperunit; 513 pp->p_fstype = FS_BSDFFS; 514 515 pp = &lp->d_partitions[RAW_PART]; 516 pp->p_offset = 0; 517 pp->p_size = lp->d_secperunit; 518 pp->p_fstype = FS_UNUSED; 519 520 lp->d_npartitions = RAW_PART+1; 521 lp->d_magic = DISKMAGIC; 522 lp->d_magic2 = DISKMAGIC; 523 lp->d_checksum = dkcksum(lp); 524 } 525 526 /* 527 * Handle ioctl MD_SETCONF for (sc_type == MD_KMEM_ALLOCATED) 528 * Just allocate some kernel memory and return. 529 */ 530 static int 531 md_ioctl_kalloc(struct md_softc *sc, struct md_conf *umd, 532 struct lwp *l) 533 { 534 vaddr_t addr; 535 vsize_t size; 536 537 /* Sanity check the size. */ 538 size = umd->md_size; 539 addr = uvm_km_alloc(kernel_map, size, 0, UVM_KMF_WIRED|UVM_KMF_ZERO); 540 if (!addr) 541 return ENOMEM; 542 543 /* This unit is now configured. */ 544 sc->sc_addr = (void *)addr; /* kernel space */ 545 sc->sc_size = (size_t)size; 546 sc->sc_type = MD_KMEM_ALLOCATED; 547 md_set_disklabel(sc); 548 return 0; 549 } 550 551 #if MEMORY_DISK_SERVER 552 553 /* 554 * Handle ioctl MD_SETCONF for (sc_type == MD_UMEM_SERVER) 555 * Set config, then become the I/O server for this unit. 556 */ 557 static int 558 md_ioctl_server(struct md_softc *sc, struct md_conf *umd, 559 struct lwp *l) 560 { 561 vaddr_t end; 562 int error; 563 564 /* Sanity check addr, size. */ 565 end = (vaddr_t) ((char *)umd->md_addr + umd->md_size); 566 567 if ((end >= VM_MAXUSER_ADDRESS) || 568 (end < ((vaddr_t) umd->md_addr)) ) 569 return EINVAL; 570 571 /* This unit is now configured. */ 572 sc->sc_addr = umd->md_addr; /* user space */ 573 sc->sc_size = umd->md_size; 574 sc->sc_type = MD_UMEM_SERVER; 575 md_set_disklabel(sc); 576 577 /* Become the server daemon */ 578 error = md_server_loop(sc); 579 580 /* This server is now going away! */ 581 sc->sc_type = MD_UNCONFIGURED; 582 sc->sc_addr = 0; 583 sc->sc_size = 0; 584 585 return (error); 586 } 587 588 static int md_sleep_pri = PWAIT | PCATCH; 589 590 static int 591 md_server_loop(struct md_softc *sc) 592 { 593 struct buf *bp; 594 void *addr; /* user space address */ 595 size_t off; /* offset into "device" */ 596 size_t xfer; /* amount to transfer */ 597 int error; 598 bool is_read; 599 600 for (;;) { 601 /* Wait for some work to arrive. */ 602 while ((bp = bufq_get(sc->sc_buflist)) == NULL) { 603 error = tsleep((void *)sc, md_sleep_pri, "md_idle", 0); 604 if (error) 605 return error; 606 } 607 608 /* Do the transfer to/from user space. */ 609 error = 0; 610 is_read = ((bp->b_flags & B_READ) == B_READ); 611 bp->b_resid = bp->b_bcount; 612 off = (bp->b_blkno << DEV_BSHIFT); 613 if (off >= sc->sc_size) { 614 if (is_read) 615 goto done; /* EOF (not an error) */ 616 error = EIO; 617 goto done; 618 } 619 xfer = bp->b_resid; 620 if (xfer > (sc->sc_size - off)) 621 xfer = (sc->sc_size - off); 622 addr = (char *)sc->sc_addr + off; 623 disk_busy(&sc->sc_dkdev); 624 if (is_read) 625 error = copyin(addr, bp->b_data, xfer); 626 else 627 error = copyout(bp->b_data, addr, xfer); 628 disk_unbusy(&sc->sc_dkdev, (error ? 0 : xfer), is_read); 629 if (!error) 630 bp->b_resid -= xfer; 631 632 done: 633 if (error) { 634 bp->b_error = error; 635 } 636 biodone(bp); 637 } 638 } 639 #endif /* MEMORY_DISK_SERVER */ 640
Indexes created Thu Sep 25 16:09:42 GMT 2025