1 /* $NetBSD: ccd.c,v 1.188 2022/03/28 12:33:20 riastradh Exp $ */ 2 3 /*- 4 * Copyright (c) 1996, 1997, 1998, 1999, 2007, 2009 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe, and by Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1988 University of Utah. 34 * Copyright (c) 1990, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * This code is derived from software contributed to Berkeley by 38 * the Systems Programming Group of the University of Utah Computer 39 * Science Department. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * from: Utah $Hdr$ 66 * 67 * @(#)cd.c 8.2 (Berkeley) 11/16/93 68 */ 69 70 /* 71 * "Concatenated" disk driver. 72 * 73 * Notes on concurrency: 74 * 75 * => sc_dvlock serializes access to the device nodes, excluding block I/O. 76 * 77 * => sc_iolock serializes access to (sc_flags & CCDF_INITED), disk stats, 78 * sc_stop, sc_bufq and b_resid from master buffers. 79 * 80 * => a combination of CCDF_INITED, sc_inflight, and sc_iolock is used to 81 * serialize I/O and configuration changes. 82 * 83 * => the in-core disk label does not change while the device is open. 84 * 85 * On memory consumption: ccd fans out I/O requests and so needs to 86 * allocate memory. If the system is desperately low on memory, we 87 * single thread I/O. 88 */ 89 90 #include <sys/cdefs.h> 91 __KERNEL_RCSID(0, "$NetBSD: ccd.c,v 1.188 2022/03/28 12:33:20 riastradh Exp $"); 92 93 #include <sys/param.h> 94 #include <sys/systm.h> 95 #include <sys/kernel.h> 96 #include <sys/proc.h> 97 #include <sys/errno.h> 98 #include <sys/buf.h> 99 #include <sys/kmem.h> 100 #include <sys/pool.h> 101 #include <sys/module.h> 102 #include <sys/namei.h> 103 #include <sys/stat.h> 104 #include <sys/ioctl.h> 105 #include <sys/disklabel.h> 106 #include <sys/device.h> 107 #include <sys/disk.h> 108 #include <sys/syslog.h> 109 #include <sys/fcntl.h> 110 #include <sys/vnode.h> 111 #include <sys/conf.h> 112 #include <sys/mutex.h> 113 #include <sys/queue.h> 114 #include <sys/kauth.h> 115 #include <sys/kthread.h> 116 #include <sys/bufq.h> 117 #include <sys/sysctl.h> 118 #include <sys/compat_stub.h> 119 120 #include <uvm/uvm_extern.h> 121 122 #include <dev/ccdvar.h> 123 #include <dev/dkvar.h> 124 125 #include <miscfs/specfs/specdev.h> /* for v_rdev */ 126 127 #include "ioconf.h" 128 129 #if defined(CCDDEBUG) && !defined(DEBUG) 130 #define DEBUG 131 #endif 132 133 #ifdef DEBUG 134 #define CCDB_FOLLOW 0x01 135 #define CCDB_INIT 0x02 136 #define CCDB_IO 0x04 137 #define CCDB_LABEL 0x08 138 #define CCDB_VNODE 0x10 139 int ccddebug = 0x00; 140 #endif 141 142 #define ccdunit(x) DISKUNIT(x) 143 144 struct ccdbuf { 145 struct buf cb_buf; /* new I/O buf */ 146 struct buf *cb_obp; /* ptr. to original I/O buf */ 147 struct ccd_softc *cb_sc; /* pointer to ccd softc */ 148 int cb_comp; /* target component */ 149 SIMPLEQ_ENTRY(ccdbuf) cb_q; /* fifo of component buffers */ 150 }; 151 152 /* component buffer pool */ 153 static pool_cache_t ccd_cache; 154 155 #define CCD_GETBUF() pool_cache_get(ccd_cache, PR_WAITOK) 156 #define CCD_PUTBUF(cbp) pool_cache_put(ccd_cache, cbp) 157 158 #define CCDLABELDEV(dev) \ 159 (MAKEDISKDEV(major((dev)), ccdunit((dev)), RAW_PART)) 160 161 /* called by main() at boot time */ 162 void ccddetach(void); 163 164 /* called by biodone() at interrupt time */ 165 static void ccdiodone(struct buf *); 166 167 static void ccdinterleave(struct ccd_softc *); 168 static int ccdinit(struct ccd_softc *, char **, struct vnode **, 169 struct lwp *); 170 static struct ccdbuf *ccdbuffer(struct ccd_softc *, struct buf *, 171 daddr_t, void *, long); 172 static void ccdgetdefaultlabel(struct ccd_softc *, struct disklabel *); 173 static void ccdgetdisklabel(dev_t); 174 static void ccdmakedisklabel(struct ccd_softc *); 175 static void ccdstart(struct ccd_softc *); 176 static void ccdthread(void *); 177 178 static dev_type_open(ccdopen); 179 static dev_type_close(ccdclose); 180 static dev_type_read(ccdread); 181 static dev_type_write(ccdwrite); 182 static dev_type_ioctl(ccdioctl); 183 static dev_type_strategy(ccdstrategy); 184 static dev_type_size(ccdsize); 185 186 const struct bdevsw ccd_bdevsw = { 187 .d_open = ccdopen, 188 .d_close = ccdclose, 189 .d_strategy = ccdstrategy, 190 .d_ioctl = ccdioctl, 191 .d_dump = nodump, 192 .d_psize = ccdsize, 193 .d_discard = nodiscard, 194 .d_flag = D_DISK | D_MPSAFE 195 }; 196 197 const struct cdevsw ccd_cdevsw = { 198 .d_open = ccdopen, 199 .d_close = ccdclose, 200 .d_read = ccdread, 201 .d_write = ccdwrite, 202 .d_ioctl = ccdioctl, 203 .d_stop = nostop, 204 .d_tty = notty, 205 .d_poll = nopoll, 206 .d_mmap = nommap, 207 .d_kqfilter = nokqfilter, 208 .d_discard = nodiscard, 209 .d_flag = D_DISK | D_MPSAFE 210 }; 211 212 static const struct dkdriver ccddkdriver = { 213 .d_strategy = ccdstrategy, 214 .d_minphys = minphys 215 }; 216 217 #ifdef DEBUG 218 static void printiinfo(struct ccdiinfo *); 219 #endif 220 221 static LIST_HEAD(, ccd_softc) ccds = LIST_HEAD_INITIALIZER(ccds); 222 static kmutex_t ccd_lock; 223 224 SYSCTL_SETUP_PROTO(sysctl_kern_ccd_setup); 225 226 static struct ccd_softc * 227 ccdcreate(int unit) { 228 struct ccd_softc *sc = kmem_zalloc(sizeof(*sc), KM_SLEEP); 229 230 /* Initialize per-softc structures. */ 231 snprintf(sc->sc_xname, sizeof(sc->sc_xname), "ccd%d", unit); 232 sc->sc_unit = unit; 233 mutex_init(&sc->sc_dvlock, MUTEX_DEFAULT, IPL_NONE); 234 sc->sc_iolock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 235 cv_init(&sc->sc_stop, "ccdstop"); 236 cv_init(&sc->sc_push, "ccdthr"); 237 disk_init(&sc->sc_dkdev, sc->sc_xname, &ccddkdriver); 238 return sc; 239 } 240 241 static void 242 ccddestroy(struct ccd_softc *sc) { 243 mutex_obj_free(sc->sc_iolock); 244 mutex_exit(&sc->sc_dvlock); 245 mutex_destroy(&sc->sc_dvlock); 246 cv_destroy(&sc->sc_stop); 247 cv_destroy(&sc->sc_push); 248 disk_destroy(&sc->sc_dkdev); 249 kmem_free(sc, sizeof(*sc)); 250 } 251 252 static struct ccd_softc * 253 ccdget(int unit, int make) { 254 struct ccd_softc *sc; 255 if (unit < 0) { 256 #ifdef DIAGNOSTIC 257 panic("%s: unit %d!", __func__, unit); 258 #endif 259 return NULL; 260 } 261 mutex_enter(&ccd_lock); 262 LIST_FOREACH(sc, &ccds, sc_link) { 263 if (sc->sc_unit == unit) { 264 mutex_exit(&ccd_lock); 265 return sc; 266 } 267 } 268 mutex_exit(&ccd_lock); 269 if (!make) 270 return NULL; 271 if ((sc = ccdcreate(unit)) == NULL) 272 return NULL; 273 mutex_enter(&ccd_lock); 274 LIST_INSERT_HEAD(&ccds, sc, sc_link); 275 mutex_exit(&ccd_lock); 276 return sc; 277 } 278 279 static void 280 ccdput(struct ccd_softc *sc) { 281 mutex_enter(&ccd_lock); 282 LIST_REMOVE(sc, sc_link); 283 mutex_exit(&ccd_lock); 284 ccddestroy(sc); 285 } 286 287 /* 288 * Called by main() during pseudo-device attachment. All we need 289 * to do is allocate enough space for devices to be configured later. 290 */ 291 void 292 ccdattach(int num) 293 { 294 mutex_init(&ccd_lock, MUTEX_DEFAULT, IPL_NONE); 295 296 /* Initialize the component buffer pool. */ 297 ccd_cache = pool_cache_init(sizeof(struct ccdbuf), 0, 298 0, 0, "ccdbuf", NULL, IPL_BIO, NULL, NULL, NULL); 299 } 300 301 void 302 ccddetach(void) 303 { 304 pool_cache_destroy(ccd_cache); 305 mutex_destroy(&ccd_lock); 306 } 307 308 static int 309 ccdinit(struct ccd_softc *cs, char **cpaths, struct vnode **vpp, 310 struct lwp *l) 311 { 312 struct ccdcinfo *ci = NULL; 313 int ix; 314 struct ccdgeom *ccg = &cs->sc_geom; 315 char *tmppath; 316 int error, path_alloced; 317 uint64_t psize, minsize; 318 unsigned secsize, maxsecsize; 319 struct disk_geom *dg; 320 321 #ifdef DEBUG 322 if (ccddebug & (CCDB_FOLLOW|CCDB_INIT)) 323 printf("%s: ccdinit\n", cs->sc_xname); 324 #endif 325 326 /* Allocate space for the component info. */ 327 cs->sc_cinfo = kmem_alloc(cs->sc_nccdisks * sizeof(*cs->sc_cinfo), 328 KM_SLEEP); 329 tmppath = kmem_alloc(MAXPATHLEN, KM_SLEEP); 330 331 cs->sc_size = 0; 332 333 /* 334 * Verify that each component piece exists and record 335 * relevant information about it. 336 */ 337 maxsecsize = 0; 338 minsize = 0; 339 for (ix = 0, path_alloced = 0; ix < cs->sc_nccdisks; ix++) { 340 ci = &cs->sc_cinfo[ix]; 341 ci->ci_vp = vpp[ix]; 342 343 /* 344 * Copy in the pathname of the component. 345 */ 346 memset(tmppath, 0, MAXPATHLEN); /* sanity */ 347 error = copyinstr(cpaths[ix], tmppath, 348 MAXPATHLEN, &ci->ci_pathlen); 349 if (ci->ci_pathlen == 0) 350 error = EINVAL; 351 if (error) { 352 #ifdef DEBUG 353 if (ccddebug & (CCDB_FOLLOW|CCDB_INIT)) 354 printf("%s: can't copy path, error = %d\n", 355 cs->sc_xname, error); 356 #endif 357 goto out; 358 } 359 ci->ci_path = kmem_alloc(ci->ci_pathlen, KM_SLEEP); 360 memcpy(ci->ci_path, tmppath, ci->ci_pathlen); 361 path_alloced++; 362 363 /* 364 * XXX: Cache the component's dev_t. 365 */ 366 ci->ci_dev = vpp[ix]->v_rdev; 367 368 /* 369 * Get partition information for the component. 370 */ 371 error = getdisksize(vpp[ix], &psize, &secsize); 372 if (error) { 373 #ifdef DEBUG 374 if (ccddebug & (CCDB_FOLLOW|CCDB_INIT)) 375 printf("%s: %s: disksize failed, error = %d\n", 376 cs->sc_xname, ci->ci_path, error); 377 #endif 378 goto out; 379 } 380 381 /* 382 * Calculate the size, truncating to an interleave 383 * boundary if necessary. 384 */ 385 maxsecsize = secsize > maxsecsize ? secsize : maxsecsize; 386 if (cs->sc_ileave > 1) 387 psize -= psize % cs->sc_ileave; 388 389 if (psize == 0) { 390 #ifdef DEBUG 391 if (ccddebug & (CCDB_FOLLOW|CCDB_INIT)) 392 printf("%s: %s: size == 0\n", 393 cs->sc_xname, ci->ci_path); 394 #endif 395 error = ENODEV; 396 goto out; 397 } 398 399 if (minsize == 0 || psize < minsize) 400 minsize = psize; 401 ci->ci_size = psize; 402 cs->sc_size += psize; 403 } 404 405 /* 406 * Don't allow the interleave to be smaller than 407 * the biggest component sector. 408 */ 409 if ((cs->sc_ileave > 0) && 410 (cs->sc_ileave < (maxsecsize / DEV_BSIZE))) { 411 #ifdef DEBUG 412 if (ccddebug & (CCDB_FOLLOW|CCDB_INIT)) 413 printf("%s: interleave must be at least %d\n", 414 cs->sc_xname, (maxsecsize / DEV_BSIZE)); 415 #endif 416 error = EINVAL; 417 goto out; 418 } 419 420 /* 421 * If uniform interleave is desired set all sizes to that of 422 * the smallest component. 423 */ 424 if (cs->sc_flags & CCDF_UNIFORM) { 425 for (ci = cs->sc_cinfo; 426 ci < &cs->sc_cinfo[cs->sc_nccdisks]; ci++) 427 ci->ci_size = minsize; 428 429 cs->sc_size = cs->sc_nccdisks * minsize; 430 } 431 432 /* 433 * Construct the interleave table. 434 */ 435 ccdinterleave(cs); 436 437 /* 438 * Create pseudo-geometry based on 1MB cylinders. It's 439 * pretty close. 440 */ 441 ccg->ccg_secsize = DEV_BSIZE; 442 ccg->ccg_ntracks = 1; 443 ccg->ccg_nsectors = 1024 * (1024 / ccg->ccg_secsize); 444 ccg->ccg_ncylinders = cs->sc_size / ccg->ccg_nsectors; 445 446 dg = &cs->sc_dkdev.dk_geom; 447 memset(dg, 0, sizeof(*dg)); 448 dg->dg_secperunit = cs->sc_size; 449 dg->dg_secsize = ccg->ccg_secsize; 450 dg->dg_nsectors = ccg->ccg_nsectors; 451 dg->dg_ntracks = ccg->ccg_ntracks; 452 dg->dg_ncylinders = ccg->ccg_ncylinders; 453 454 if (cs->sc_ileave > 0) 455 aprint_normal("%s: Interleaving %d component%s " 456 "(%d block interleave)\n", cs->sc_xname, 457 cs->sc_nccdisks, (cs->sc_nccdisks != 0 ? "s" : ""), 458 cs->sc_ileave); 459 else 460 aprint_normal("%s: Concatenating %d component%s\n", 461 cs->sc_xname, 462 cs->sc_nccdisks, (cs->sc_nccdisks != 0 ? "s" : "")); 463 for (ix = 0; ix < cs->sc_nccdisks; ix++) { 464 ci = &cs->sc_cinfo[ix]; 465 aprint_normal("%s: %s (%ju blocks)\n", cs->sc_xname, 466 ci->ci_path, (uintmax_t)ci->ci_size); 467 } 468 aprint_normal("%s: total %ju blocks\n", cs->sc_xname, cs->sc_size); 469 470 /* 471 * Create thread to handle deferred I/O. 472 */ 473 cs->sc_zap = false; 474 error = kthread_create(PRI_BIO, KTHREAD_MPSAFE, NULL, ccdthread, 475 cs, &cs->sc_thread, "%s", cs->sc_xname); 476 if (error) { 477 printf("ccdinit: can't create thread: %d\n", error); 478 goto out; 479 } 480 481 /* 482 * Only now that everything is set up can we enable the device. 483 */ 484 mutex_enter(cs->sc_iolock); 485 cs->sc_flags |= CCDF_INITED; 486 mutex_exit(cs->sc_iolock); 487 kmem_free(tmppath, MAXPATHLEN); 488 return (0); 489 490 out: 491 for (ix = 0; ix < path_alloced; ix++) { 492 kmem_free(cs->sc_cinfo[ix].ci_path, 493 cs->sc_cinfo[ix].ci_pathlen); 494 } 495 kmem_free(cs->sc_cinfo, cs->sc_nccdisks * sizeof(struct ccdcinfo)); 496 kmem_free(tmppath, MAXPATHLEN); 497 return (error); 498 } 499 500 static void 501 ccdinterleave(struct ccd_softc *cs) 502 { 503 struct ccdcinfo *ci, *smallci; 504 struct ccdiinfo *ii; 505 daddr_t bn, lbn; 506 int ix; 507 u_long size; 508 509 #ifdef DEBUG 510 if (ccddebug & CCDB_INIT) 511 printf("ccdinterleave(%p): ileave %d\n", cs, cs->sc_ileave); 512 #endif 513 /* 514 * Allocate an interleave table. 515 * Chances are this is too big, but we don't care. 516 */ 517 size = (cs->sc_nccdisks + 1) * sizeof(struct ccdiinfo); 518 cs->sc_itable = kmem_zalloc(size, KM_SLEEP); 519 520 /* 521 * Trivial case: no interleave (actually interleave of disk size). 522 * Each table entry represents a single component in its entirety. 523 */ 524 if (cs->sc_ileave == 0) { 525 bn = 0; 526 ii = cs->sc_itable; 527 528 for (ix = 0; ix < cs->sc_nccdisks; ix++) { 529 /* Allocate space for ii_index. */ 530 ii->ii_indexsz = sizeof(int); 531 ii->ii_index = kmem_alloc(ii->ii_indexsz, KM_SLEEP); 532 ii->ii_ndisk = 1; 533 ii->ii_startblk = bn; 534 ii->ii_startoff = 0; 535 ii->ii_index[0] = ix; 536 bn += cs->sc_cinfo[ix].ci_size; 537 ii++; 538 } 539 ii->ii_ndisk = 0; 540 #ifdef DEBUG 541 if (ccddebug & CCDB_INIT) 542 printiinfo(cs->sc_itable); 543 #endif 544 return; 545 } 546 547 /* 548 * The following isn't fast or pretty; it doesn't have to be. 549 */ 550 size = 0; 551 bn = lbn = 0; 552 for (ii = cs->sc_itable; ; ii++) { 553 /* Allocate space for ii_index. */ 554 ii->ii_indexsz = sizeof(int) * cs->sc_nccdisks; 555 ii->ii_index = kmem_alloc(ii->ii_indexsz, KM_SLEEP); 556 557 /* 558 * Locate the smallest of the remaining components 559 */ 560 smallci = NULL; 561 for (ci = cs->sc_cinfo; 562 ci < &cs->sc_cinfo[cs->sc_nccdisks]; ci++) 563 if (ci->ci_size > size && 564 (smallci == NULL || 565 ci->ci_size < smallci->ci_size)) 566 smallci = ci; 567 568 /* 569 * Nobody left, all done 570 */ 571 if (smallci == NULL) { 572 ii->ii_ndisk = 0; 573 break; 574 } 575 576 /* 577 * Record starting logical block and component offset 578 */ 579 ii->ii_startblk = bn / cs->sc_ileave; 580 ii->ii_startoff = lbn; 581 582 /* 583 * Determine how many disks take part in this interleave 584 * and record their indices. 585 */ 586 ix = 0; 587 for (ci = cs->sc_cinfo; 588 ci < &cs->sc_cinfo[cs->sc_nccdisks]; ci++) 589 if (ci->ci_size >= smallci->ci_size) 590 ii->ii_index[ix++] = ci - cs->sc_cinfo; 591 ii->ii_ndisk = ix; 592 bn += ix * (smallci->ci_size - size); 593 lbn = smallci->ci_size / cs->sc_ileave; 594 size = smallci->ci_size; 595 } 596 #ifdef DEBUG 597 if (ccddebug & CCDB_INIT) 598 printiinfo(cs->sc_itable); 599 #endif 600 } 601 602 /* ARGSUSED */ 603 static int 604 ccdopen(dev_t dev, int flags, int fmt, struct lwp *l) 605 { 606 int unit = ccdunit(dev); 607 struct ccd_softc *cs; 608 struct disklabel *lp; 609 int error = 0, part, pmask; 610 611 #ifdef DEBUG 612 if (ccddebug & CCDB_FOLLOW) 613 printf("ccdopen(0x%"PRIx64", 0x%x)\n", dev, flags); 614 #endif 615 if ((cs = ccdget(unit, 1)) == NULL) 616 return ENXIO; 617 618 mutex_enter(&cs->sc_dvlock); 619 620 lp = cs->sc_dkdev.dk_label; 621 622 part = DISKPART(dev); 623 pmask = (1 << part); 624 625 /* 626 * If we're initialized, check to see if there are any other 627 * open partitions. If not, then it's safe to update 628 * the in-core disklabel. Only read the disklabel if it is 629 * not already valid. 630 */ 631 if ((cs->sc_flags & (CCDF_INITED|CCDF_VLABEL)) == CCDF_INITED && 632 cs->sc_dkdev.dk_openmask == 0) 633 ccdgetdisklabel(dev); 634 635 /* Check that the partition exists. */ 636 if (part != RAW_PART) { 637 if (((cs->sc_flags & CCDF_INITED) == 0) || 638 ((part >= lp->d_npartitions) || 639 (lp->d_partitions[part].p_fstype == FS_UNUSED))) { 640 error = ENXIO; 641 goto done; 642 } 643 } 644 645 /* Prevent our unit from being unconfigured while open. */ 646 switch (fmt) { 647 case S_IFCHR: 648 cs->sc_dkdev.dk_copenmask |= pmask; 649 break; 650 651 case S_IFBLK: 652 cs->sc_dkdev.dk_bopenmask |= pmask; 653 break; 654 } 655 cs->sc_dkdev.dk_openmask = 656 cs->sc_dkdev.dk_copenmask | cs->sc_dkdev.dk_bopenmask; 657 658 done: 659 mutex_exit(&cs->sc_dvlock); 660 return (error); 661 } 662 663 /* ARGSUSED */ 664 static int 665 ccdclose(dev_t dev, int flags, int fmt, struct lwp *l) 666 { 667 int unit = ccdunit(dev); 668 struct ccd_softc *cs; 669 int part; 670 671 #ifdef DEBUG 672 if (ccddebug & CCDB_FOLLOW) 673 printf("ccdclose(0x%"PRIx64", 0x%x)\n", dev, flags); 674 #endif 675 676 if ((cs = ccdget(unit, 0)) == NULL) 677 return ENXIO; 678 679 mutex_enter(&cs->sc_dvlock); 680 681 part = DISKPART(dev); 682 683 /* ...that much closer to allowing unconfiguration... */ 684 switch (fmt) { 685 case S_IFCHR: 686 cs->sc_dkdev.dk_copenmask &= ~(1 << part); 687 break; 688 689 case S_IFBLK: 690 cs->sc_dkdev.dk_bopenmask &= ~(1 << part); 691 break; 692 } 693 cs->sc_dkdev.dk_openmask = 694 cs->sc_dkdev.dk_copenmask | cs->sc_dkdev.dk_bopenmask; 695 696 if (cs->sc_dkdev.dk_openmask == 0) { 697 if ((cs->sc_flags & CCDF_KLABEL) == 0) 698 cs->sc_flags &= ~CCDF_VLABEL; 699 } 700 701 mutex_exit(&cs->sc_dvlock); 702 return (0); 703 } 704 705 static bool 706 ccdbackoff(struct ccd_softc *cs) 707 { 708 709 /* XXX Arbitrary, should be a uvm call. */ 710 return uvm_availmem(true) < (uvmexp.freemin >> 1) && 711 disk_isbusy(&cs->sc_dkdev); 712 } 713 714 static void 715 ccdthread(void *cookie) 716 { 717 struct ccd_softc *cs; 718 719 cs = cookie; 720 721 #ifdef DEBUG 722 if (ccddebug & CCDB_FOLLOW) 723 printf("ccdthread: hello\n"); 724 #endif 725 726 mutex_enter(cs->sc_iolock); 727 while (__predict_true(!cs->sc_zap)) { 728 if (bufq_peek(cs->sc_bufq) == NULL) { 729 /* Nothing to do. */ 730 cv_wait(&cs->sc_push, cs->sc_iolock); 731 continue; 732 } 733 if (ccdbackoff(cs)) { 734 /* Wait for memory to become available. */ 735 (void)cv_timedwait(&cs->sc_push, cs->sc_iolock, 1); 736 continue; 737 } 738 #ifdef DEBUG 739 if (ccddebug & CCDB_FOLLOW) 740 printf("ccdthread: dispatching I/O\n"); 741 #endif 742 ccdstart(cs); 743 mutex_enter(cs->sc_iolock); 744 } 745 cs->sc_thread = NULL; 746 mutex_exit(cs->sc_iolock); 747 #ifdef DEBUG 748 if (ccddebug & CCDB_FOLLOW) 749 printf("ccdthread: goodbye\n"); 750 #endif 751 kthread_exit(0); 752 } 753 754 static void 755 ccdstrategy(struct buf *bp) 756 { 757 int unit = ccdunit(bp->b_dev); 758 struct ccd_softc *cs; 759 if ((cs = ccdget(unit, 0)) == NULL) 760 return; 761 762 /* Must be open or reading label. */ 763 KASSERT(cs->sc_dkdev.dk_openmask != 0 || 764 (cs->sc_flags & CCDF_RLABEL) != 0); 765 766 mutex_enter(cs->sc_iolock); 767 /* Synchronize with device init/uninit. */ 768 if (__predict_false((cs->sc_flags & CCDF_INITED) == 0)) { 769 mutex_exit(cs->sc_iolock); 770 #ifdef DEBUG 771 if (ccddebug & CCDB_FOLLOW) 772 printf("ccdstrategy: unit %d: not inited\n", unit); 773 #endif 774 bp->b_error = ENXIO; 775 bp->b_resid = bp->b_bcount; 776 biodone(bp); 777 return; 778 } 779 780 /* Defer to thread if system is low on memory. */ 781 bufq_put(cs->sc_bufq, bp); 782 if (__predict_false(ccdbackoff(cs))) { 783 mutex_exit(cs->sc_iolock); 784 #ifdef DEBUG 785 if (ccddebug & CCDB_FOLLOW) 786 printf("ccdstrategy: holding off on I/O\n"); 787 #endif 788 return; 789 } 790 ccdstart(cs); 791 } 792 793 static void 794 ccdstart(struct ccd_softc *cs) 795 { 796 daddr_t blkno; 797 int wlabel; 798 struct disklabel *lp; 799 long bcount, rcount; 800 struct ccdbuf *cbp; 801 char *addr; 802 daddr_t bn; 803 vnode_t *vp; 804 buf_t *bp; 805 806 KASSERT(mutex_owned(cs->sc_iolock)); 807 808 bp = bufq_get(cs->sc_bufq); 809 KASSERT(bp != NULL); 810 811 disk_busy(&cs->sc_dkdev); 812 813 #ifdef DEBUG 814 if (ccddebug & CCDB_FOLLOW) 815 printf("ccdstart(%s, %p)\n", cs->sc_xname, bp); 816 #endif 817 818 /* If it's a nil transfer, wake up the top half now. */ 819 if (bp->b_bcount == 0) 820 goto done; 821 822 lp = cs->sc_dkdev.dk_label; 823 824 /* 825 * Do bounds checking and adjust transfer. If there's an 826 * error, the bounds check will flag that for us. Convert 827 * the partition relative block number to an absolute. 828 */ 829 blkno = bp->b_blkno; 830 wlabel = cs->sc_flags & (CCDF_WLABEL|CCDF_LABELLING); 831 if (DISKPART(bp->b_dev) != RAW_PART) { 832 if (bounds_check_with_label(&cs->sc_dkdev, bp, wlabel) <= 0) 833 goto done; 834 blkno += lp->d_partitions[DISKPART(bp->b_dev)].p_offset; 835 } 836 mutex_exit(cs->sc_iolock); 837 bp->b_rawblkno = blkno; 838 839 /* Allocate the component buffers and start I/O! */ 840 bp->b_resid = bp->b_bcount; 841 bn = bp->b_rawblkno; 842 addr = bp->b_data; 843 for (bcount = bp->b_bcount; bcount > 0; bcount -= rcount) { 844 cbp = ccdbuffer(cs, bp, bn, addr, bcount); 845 rcount = cbp->cb_buf.b_bcount; 846 bn += btodb(rcount); 847 addr += rcount; 848 vp = cbp->cb_buf.b_vp; 849 if ((cbp->cb_buf.b_flags & B_READ) == 0) { 850 mutex_enter(vp->v_interlock); 851 vp->v_numoutput++; 852 mutex_exit(vp->v_interlock); 853 } 854 (void)VOP_STRATEGY(vp, &cbp->cb_buf); 855 } 856 return; 857 858 done: 859 disk_unbusy(&cs->sc_dkdev, 0, 0); 860 cv_broadcast(&cs->sc_stop); 861 cv_broadcast(&cs->sc_push); 862 mutex_exit(cs->sc_iolock); 863 bp->b_resid = bp->b_bcount; 864 biodone(bp); 865 } 866 867 /* 868 * Build a component buffer header. 869 */ 870 static struct ccdbuf * 871 ccdbuffer(struct ccd_softc *cs, struct buf *bp, daddr_t bn, void *addr, 872 long bcount) 873 { 874 struct ccdcinfo *ci; 875 struct ccdbuf *cbp; 876 daddr_t cbn, cboff; 877 u_int64_t cbc; 878 int ccdisk; 879 880 #ifdef DEBUG 881 if (ccddebug & CCDB_IO) 882 printf("ccdbuffer(%p, %p, %" PRId64 ", %p, %ld)\n", 883 cs, bp, bn, addr, bcount); 884 #endif 885 /* 886 * Determine which component bn falls in. 887 */ 888 cbn = bn; 889 cboff = 0; 890 891 /* 892 * Serially concatenated 893 */ 894 if (cs->sc_ileave == 0) { 895 daddr_t sblk; 896 897 sblk = 0; 898 for (ccdisk = 0, ci = &cs->sc_cinfo[ccdisk]; 899 cbn >= sblk + ci->ci_size; 900 ccdisk++, ci = &cs->sc_cinfo[ccdisk]) 901 sblk += ci->ci_size; 902 cbn -= sblk; 903 } 904 /* 905 * Interleaved 906 */ 907 else { 908 struct ccdiinfo *ii; 909 int off; 910 911 cboff = cbn % cs->sc_ileave; 912 cbn /= cs->sc_ileave; 913 for (ii = cs->sc_itable; ii->ii_ndisk; ii++) 914 if (ii->ii_startblk > cbn) 915 break; 916 ii--; 917 off = cbn - ii->ii_startblk; 918 if (ii->ii_ndisk == 1) { 919 ccdisk = ii->ii_index[0]; 920 cbn = ii->ii_startoff + off; 921 } else { 922 ccdisk = ii->ii_index[off % ii->ii_ndisk]; 923 cbn = ii->ii_startoff + off / ii->ii_ndisk; 924 } 925 cbn *= cs->sc_ileave; 926 ci = &cs->sc_cinfo[ccdisk]; 927 } 928 929 /* 930 * Fill in the component buf structure. 931 */ 932 cbp = CCD_GETBUF(); 933 KASSERT(cbp != NULL); 934 buf_init(&cbp->cb_buf); 935 cbp->cb_buf.b_flags = bp->b_flags; 936 cbp->cb_buf.b_oflags = bp->b_oflags; 937 cbp->cb_buf.b_cflags = bp->b_cflags; 938 cbp->cb_buf.b_iodone = ccdiodone; 939 cbp->cb_buf.b_proc = bp->b_proc; 940 cbp->cb_buf.b_dev = ci->ci_dev; 941 cbp->cb_buf.b_blkno = cbn + cboff; 942 cbp->cb_buf.b_data = addr; 943 cbp->cb_buf.b_vp = ci->ci_vp; 944 cbp->cb_buf.b_objlock = ci->ci_vp->v_interlock; 945 if (cs->sc_ileave == 0) 946 cbc = dbtob((u_int64_t)(ci->ci_size - cbn)); 947 else 948 cbc = dbtob((u_int64_t)(cs->sc_ileave - cboff)); 949 cbp->cb_buf.b_bcount = cbc < bcount ? cbc : bcount; 950 951 /* 952 * context for ccdiodone 953 */ 954 cbp->cb_obp = bp; 955 cbp->cb_sc = cs; 956 cbp->cb_comp = ccdisk; 957 958 BIO_COPYPRIO(&cbp->cb_buf, bp); 959 960 #ifdef DEBUG 961 if (ccddebug & CCDB_IO) 962 printf(" dev 0x%"PRIx64"(u%lu): cbp %p bn %" PRId64 " addr %p" 963 " bcnt %d\n", 964 ci->ci_dev, (unsigned long) (ci-cs->sc_cinfo), cbp, 965 cbp->cb_buf.b_blkno, cbp->cb_buf.b_data, 966 cbp->cb_buf.b_bcount); 967 #endif 968 969 return (cbp); 970 } 971 972 /* 973 * Called at interrupt time. 974 * Mark the component as done and if all components are done, 975 * take a ccd interrupt. 976 */ 977 static void 978 ccdiodone(struct buf *vbp) 979 { 980 struct ccdbuf *cbp = (struct ccdbuf *) vbp; 981 struct buf *bp = cbp->cb_obp; 982 struct ccd_softc *cs = cbp->cb_sc; 983 int count; 984 985 #ifdef DEBUG 986 if (ccddebug & CCDB_FOLLOW) 987 printf("ccdiodone(%p)\n", cbp); 988 if (ccddebug & CCDB_IO) { 989 printf("ccdiodone: bp %p bcount %d resid %d\n", 990 bp, bp->b_bcount, bp->b_resid); 991 printf(" dev 0x%"PRIx64"(u%d), cbp %p bn %" PRId64 " addr %p" 992 " bcnt %d\n", 993 cbp->cb_buf.b_dev, cbp->cb_comp, cbp, 994 cbp->cb_buf.b_blkno, cbp->cb_buf.b_data, 995 cbp->cb_buf.b_bcount); 996 } 997 #endif 998 999 if (cbp->cb_buf.b_error != 0) { 1000 bp->b_error = cbp->cb_buf.b_error; 1001 printf("%s: error %d on component %d\n", 1002 cs->sc_xname, bp->b_error, cbp->cb_comp); 1003 } 1004 count = cbp->cb_buf.b_bcount; 1005 buf_destroy(&cbp->cb_buf); 1006 CCD_PUTBUF(cbp); 1007 1008 /* 1009 * If all done, "interrupt". 1010 */ 1011 mutex_enter(cs->sc_iolock); 1012 bp->b_resid -= count; 1013 if (bp->b_resid < 0) 1014 panic("ccdiodone: count"); 1015 if (bp->b_resid == 0) { 1016 /* 1017 * Request is done for better or worse, wakeup the top half. 1018 */ 1019 if (bp->b_error != 0) 1020 bp->b_resid = bp->b_bcount; 1021 disk_unbusy(&cs->sc_dkdev, (bp->b_bcount - bp->b_resid), 1022 (bp->b_flags & B_READ)); 1023 if (!disk_isbusy(&cs->sc_dkdev)) { 1024 if (bufq_peek(cs->sc_bufq) != NULL) { 1025 cv_broadcast(&cs->sc_push); 1026 } 1027 cv_broadcast(&cs->sc_stop); 1028 } 1029 mutex_exit(cs->sc_iolock); 1030 biodone(bp); 1031 } else 1032 mutex_exit(cs->sc_iolock); 1033 } 1034 1035 /* ARGSUSED */ 1036 static int 1037 ccdread(dev_t dev, struct uio *uio, int flags) 1038 { 1039 int unit = ccdunit(dev); 1040 struct ccd_softc *cs; 1041 1042 #ifdef DEBUG 1043 if (ccddebug & CCDB_FOLLOW) 1044 printf("ccdread(0x%"PRIx64", %p)\n", dev, uio); 1045 #endif 1046 if ((cs = ccdget(unit, 0)) == NULL) 1047 return 0; 1048 1049 /* Unlocked advisory check, ccdstrategy check is synchronous. */ 1050 if ((cs->sc_flags & CCDF_INITED) == 0) 1051 return (ENXIO); 1052 1053 return (physio(ccdstrategy, NULL, dev, B_READ, minphys, uio)); 1054 } 1055 1056 /* ARGSUSED */ 1057 static int 1058 ccdwrite(dev_t dev, struct uio *uio, int flags) 1059 { 1060 int unit = ccdunit(dev); 1061 struct ccd_softc *cs; 1062 1063 #ifdef DEBUG 1064 if (ccddebug & CCDB_FOLLOW) 1065 printf("ccdwrite(0x%"PRIx64", %p)\n", dev, uio); 1066 #endif 1067 if ((cs = ccdget(unit, 0)) == NULL) 1068 return ENOENT; 1069 1070 /* Unlocked advisory check, ccdstrategy check is synchronous. */ 1071 if ((cs->sc_flags & CCDF_INITED) == 0) 1072 return (ENXIO); 1073 1074 return (physio(ccdstrategy, NULL, dev, B_WRITE, minphys, uio)); 1075 } 1076 1077 int (*compat_ccd_ioctl_60)(dev_t, u_long, void *, int, struct lwp *, 1078 int (*)(dev_t, u_long, void *, int, struct lwp *)) = (void *)enosys; 1079 1080 static int 1081 ccdioctl(dev_t dev, u_long cmd, void *data, int flag, struct lwp *l) 1082 { 1083 int unit = ccdunit(dev); 1084 int i, j, lookedup = 0, error = 0; 1085 int part, pmask, make, hook; 1086 struct ccd_softc *cs; 1087 struct ccd_ioctl *ccio = (struct ccd_ioctl *)data; 1088 kauth_cred_t uc; 1089 char **cpp; 1090 struct pathbuf *pb; 1091 struct vnode **vpp; 1092 #ifdef __HAVE_OLD_DISKLABEL 1093 struct disklabel newlabel; 1094 #endif 1095 1096 switch (cmd) { 1097 case CCDIOCSET: 1098 make = 1; 1099 break; 1100 default: 1101 MODULE_HOOK_CALL(ccd_ioctl_60_hook, 1102 (0, cmd, NULL, 0, NULL, NULL), 1103 enosys(), hook); 1104 if (hook == 0) 1105 make = 1; 1106 else 1107 make = 0; 1108 break; 1109 } 1110 1111 if ((cs = ccdget(unit, make)) == NULL) 1112 return ENOENT; 1113 uc = kauth_cred_get(); 1114 1115 MODULE_HOOK_CALL(ccd_ioctl_60_hook, 1116 (dev, cmd, data, flag, l, ccdioctl), 1117 enosys(), error); 1118 if (error != ENOSYS) 1119 return error; 1120 1121 /* Must be open for writes for these commands... */ 1122 switch (cmd) { 1123 case CCDIOCSET: 1124 case CCDIOCCLR: 1125 case DIOCSDINFO: 1126 case DIOCWDINFO: 1127 case DIOCCACHESYNC: 1128 case DIOCAWEDGE: 1129 case DIOCDWEDGE: 1130 case DIOCRMWEDGES: 1131 case DIOCMWEDGES: 1132 #ifdef __HAVE_OLD_DISKLABEL 1133 case ODIOCSDINFO: 1134 case ODIOCWDINFO: 1135 #endif 1136 case DIOCKLABEL: 1137 case DIOCWLABEL: 1138 if ((flag & FWRITE) == 0) 1139 return (EBADF); 1140 } 1141 1142 /* Must be initialized for these... */ 1143 switch (cmd) { 1144 case CCDIOCCLR: 1145 case DIOCGDINFO: 1146 case DIOCGSTRATEGY: 1147 case DIOCGCACHE: 1148 case DIOCCACHESYNC: 1149 case DIOCAWEDGE: 1150 case DIOCDWEDGE: 1151 case DIOCLWEDGES: 1152 case DIOCMWEDGES: 1153 case DIOCSDINFO: 1154 case DIOCWDINFO: 1155 case DIOCGPARTINFO: 1156 case DIOCWLABEL: 1157 case DIOCKLABEL: 1158 case DIOCGDEFLABEL: 1159 #ifdef __HAVE_OLD_DISKLABEL 1160 case ODIOCGDINFO: 1161 case ODIOCSDINFO: 1162 case ODIOCWDINFO: 1163 case ODIOCGDEFLABEL: 1164 #endif 1165 if ((cs->sc_flags & CCDF_INITED) == 0) 1166 return ENXIO; 1167 } 1168 1169 error = disk_ioctl(&cs->sc_dkdev, dev, cmd, data, flag, l); 1170 if (error != EPASSTHROUGH) 1171 return error; 1172 1173 switch (cmd) { 1174 case DIOCGSTRATEGY: 1175 { 1176 struct disk_strategy *dks = (void *)data; 1177 1178 mutex_enter(cs->sc_iolock); 1179 if (cs->sc_bufq != NULL) 1180 strlcpy(dks->dks_name, 1181 bufq_getstrategyname(cs->sc_bufq), 1182 sizeof(dks->dks_name)); 1183 else 1184 error = EINVAL; 1185 mutex_exit(cs->sc_iolock); 1186 dks->dks_paramlen = 0; 1187 break; 1188 } 1189 1190 case DIOCWDINFO: 1191 case DIOCSDINFO: 1192 #ifdef __HAVE_OLD_DISKLABEL 1193 case ODIOCWDINFO: 1194 case ODIOCSDINFO: 1195 #endif 1196 { 1197 struct disklabel *lp; 1198 #ifdef __HAVE_OLD_DISKLABEL 1199 if (cmd == ODIOCSDINFO || cmd == ODIOCWDINFO) { 1200 memset(&newlabel, 0, sizeof newlabel); 1201 memcpy(&newlabel, data, sizeof (struct olddisklabel)); 1202 lp = &newlabel; 1203 } else 1204 #endif 1205 lp = (struct disklabel *)data; 1206 1207 cs->sc_flags |= CCDF_LABELLING; 1208 1209 error = setdisklabel(cs->sc_dkdev.dk_label, 1210 lp, 0, cs->sc_dkdev.dk_cpulabel); 1211 if (error == 0) { 1212 if (cmd == DIOCWDINFO 1213 #ifdef __HAVE_OLD_DISKLABEL 1214 || cmd == ODIOCWDINFO 1215 #endif 1216 ) 1217 error = writedisklabel(CCDLABELDEV(dev), 1218 ccdstrategy, cs->sc_dkdev.dk_label, 1219 cs->sc_dkdev.dk_cpulabel); 1220 } 1221 1222 cs->sc_flags &= ~CCDF_LABELLING; 1223 break; 1224 } 1225 1226 case DIOCKLABEL: 1227 if (*(int *)data != 0) 1228 cs->sc_flags |= CCDF_KLABEL; 1229 else 1230 cs->sc_flags &= ~CCDF_KLABEL; 1231 break; 1232 1233 case DIOCWLABEL: 1234 if (*(int *)data != 0) 1235 cs->sc_flags |= CCDF_WLABEL; 1236 else 1237 cs->sc_flags &= ~CCDF_WLABEL; 1238 break; 1239 1240 case DIOCGDEFLABEL: 1241 ccdgetdefaultlabel(cs, (struct disklabel *)data); 1242 break; 1243 1244 #ifdef __HAVE_OLD_DISKLABEL 1245 case ODIOCGDEFLABEL: 1246 ccdgetdefaultlabel(cs, &newlabel); 1247 if (newlabel.d_npartitions > OLDMAXPARTITIONS) 1248 return ENOTTY; 1249 memcpy(data, &newlabel, sizeof (struct olddisklabel)); 1250 break; 1251 #endif 1252 default: 1253 error = ENOTTY; 1254 break; 1255 } 1256 1257 if (error != ENOTTY) 1258 return error; 1259 1260 mutex_enter(&cs->sc_dvlock); 1261 1262 error = 0; 1263 switch (cmd) { 1264 case CCDIOCSET: 1265 if (cs->sc_flags & CCDF_INITED) { 1266 error = EBUSY; 1267 goto out; 1268 } 1269 1270 /* Validate the flags. */ 1271 if ((ccio->ccio_flags & CCDF_USERMASK) != ccio->ccio_flags) { 1272 error = EINVAL; 1273 goto out; 1274 } 1275 1276 if (ccio->ccio_ndisks > CCD_MAXNDISKS || 1277 ccio->ccio_ndisks == 0) { 1278 error = EINVAL; 1279 goto out; 1280 } 1281 1282 /* Fill in some important bits. */ 1283 cs->sc_ileave = ccio->ccio_ileave; 1284 cs->sc_nccdisks = ccio->ccio_ndisks; 1285 cs->sc_flags = ccio->ccio_flags & CCDF_USERMASK; 1286 1287 /* 1288 * Allocate space for and copy in the array of 1289 * component pathnames and device numbers. 1290 */ 1291 cpp = kmem_alloc(ccio->ccio_ndisks * sizeof(*cpp), KM_SLEEP); 1292 vpp = kmem_alloc(ccio->ccio_ndisks * sizeof(*vpp), KM_SLEEP); 1293 error = copyin(ccio->ccio_disks, cpp, 1294 ccio->ccio_ndisks * sizeof(*cpp)); 1295 if (error) { 1296 kmem_free(vpp, ccio->ccio_ndisks * sizeof(*vpp)); 1297 kmem_free(cpp, ccio->ccio_ndisks * sizeof(*cpp)); 1298 goto out; 1299 } 1300 1301 #ifdef DEBUG 1302 if (ccddebug & CCDB_INIT) 1303 for (i = 0; i < ccio->ccio_ndisks; ++i) 1304 printf("ccdioctl: component %d: %p\n", 1305 i, cpp[i]); 1306 #endif 1307 1308 for (i = 0; i < ccio->ccio_ndisks; ++i) { 1309 #ifdef DEBUG 1310 if (ccddebug & CCDB_INIT) 1311 printf("ccdioctl: lookedup = %d\n", lookedup); 1312 #endif 1313 error = pathbuf_copyin(cpp[i], &pb); 1314 if (error == 0) { 1315 error = vn_bdev_openpath(pb, &vpp[i], l); 1316 pathbuf_destroy(pb); 1317 } 1318 if (error != 0) { 1319 for (j = 0; j < lookedup; ++j) 1320 (void)vn_close(vpp[j], FREAD|FWRITE, 1321 uc); 1322 kmem_free(vpp, ccio->ccio_ndisks * 1323 sizeof(*vpp)); 1324 kmem_free(cpp, ccio->ccio_ndisks * 1325 sizeof(*cpp)); 1326 1327 /* 1328 * No component data is allocated, 1329 * nothing is to be freed. 1330 */ 1331 cs->sc_nccdisks = 0; 1332 goto out; 1333 } 1334 ++lookedup; 1335 } 1336 1337 /* Attach the disk. */ 1338 disk_attach(&cs->sc_dkdev); 1339 bufq_alloc(&cs->sc_bufq, "fcfs", 0); 1340 1341 /* 1342 * Initialize the ccd. Fills in the softc for us. 1343 */ 1344 if ((error = ccdinit(cs, cpp, vpp, l)) != 0) { 1345 for (j = 0; j < lookedup; ++j) 1346 (void)vn_close(vpp[j], FREAD|FWRITE, 1347 uc); 1348 kmem_free(vpp, ccio->ccio_ndisks * sizeof(*vpp)); 1349 kmem_free(cpp, ccio->ccio_ndisks * sizeof(*cpp)); 1350 disk_detach(&cs->sc_dkdev); 1351 mutex_exit(&cs->sc_dvlock); 1352 bufq_free(cs->sc_bufq); 1353 return error; 1354 } 1355 1356 /* We can free the temporary variables now. */ 1357 kmem_free(vpp, ccio->ccio_ndisks * sizeof(*vpp)); 1358 kmem_free(cpp, ccio->ccio_ndisks * sizeof(*cpp)); 1359 1360 /* 1361 * The ccd has been successfully initialized, so 1362 * we can place it into the array. Don't try to 1363 * read the disklabel until the disk has been attached, 1364 * because space for the disklabel is allocated 1365 * in disk_attach(); 1366 */ 1367 ccio->ccio_unit = unit; 1368 ccio->ccio_size = cs->sc_size; 1369 1370 /* Try and read the disklabel. */ 1371 ccdgetdisklabel(dev); 1372 disk_set_info(NULL, &cs->sc_dkdev, NULL); 1373 1374 /* discover wedges */ 1375 mutex_exit(&cs->sc_dvlock); 1376 dkwedge_discover(&cs->sc_dkdev); 1377 return 0; 1378 1379 case CCDIOCCLR: 1380 /* 1381 * Don't unconfigure if any other partitions are open 1382 * or if both the character and block flavors of this 1383 * partition are open. 1384 */ 1385 part = DISKPART(dev); 1386 pmask = (1 << part); 1387 if ((cs->sc_dkdev.dk_openmask & ~pmask) || 1388 ((cs->sc_dkdev.dk_bopenmask & pmask) && 1389 (cs->sc_dkdev.dk_copenmask & pmask))) { 1390 error = EBUSY; 1391 goto out; 1392 } 1393 1394 /* Delete all of our wedges. */ 1395 dkwedge_delall(&cs->sc_dkdev); 1396 1397 /* Stop new I/O, wait for in-flight I/O to complete. */ 1398 mutex_enter(cs->sc_iolock); 1399 cs->sc_flags &= ~(CCDF_INITED|CCDF_VLABEL); 1400 cs->sc_zap = true; 1401 while (disk_isbusy(&cs->sc_dkdev) || 1402 bufq_peek(cs->sc_bufq) != NULL || 1403 cs->sc_thread != NULL) { 1404 cv_broadcast(&cs->sc_push); 1405 (void)cv_timedwait(&cs->sc_stop, cs->sc_iolock, hz); 1406 } 1407 mutex_exit(cs->sc_iolock); 1408 1409 /* 1410 * Free ccd_softc information and clear entry. 1411 */ 1412 1413 /* Close the components and free their pathnames. */ 1414 for (i = 0; i < cs->sc_nccdisks; ++i) { 1415 /* 1416 * XXX: this close could potentially fail and 1417 * cause Bad Things. Maybe we need to force 1418 * the close to happen? 1419 */ 1420 #ifdef DEBUG 1421 if (ccddebug & CCDB_VNODE) 1422 vprint("CCDIOCCLR: vnode info", 1423 cs->sc_cinfo[i].ci_vp); 1424 #endif 1425 (void)vn_close(cs->sc_cinfo[i].ci_vp, FREAD|FWRITE, 1426 uc); 1427 kmem_free(cs->sc_cinfo[i].ci_path, 1428 cs->sc_cinfo[i].ci_pathlen); 1429 } 1430 1431 if (cs->sc_nccdisks != 0) { 1432 /* Free interleave index. */ 1433 for (i = 0; cs->sc_itable[i].ii_ndisk; ++i) { 1434 kmem_free(cs->sc_itable[i].ii_index, 1435 cs->sc_itable[i].ii_indexsz); 1436 } 1437 /* Free component info and interleave table. */ 1438 kmem_free(cs->sc_cinfo, cs->sc_nccdisks * 1439 sizeof(struct ccdcinfo)); 1440 kmem_free(cs->sc_itable, (cs->sc_nccdisks + 1) * 1441 sizeof(struct ccdiinfo)); 1442 } 1443 1444 aprint_normal("%s: detached\n", cs->sc_xname); 1445 1446 /* Detach the disk. */ 1447 disk_detach(&cs->sc_dkdev); 1448 bufq_free(cs->sc_bufq); 1449 1450 /* also releases sc_dvlock */ 1451 ccdput(cs); 1452 1453 /* Don't break, otherwise cs is read again. */ 1454 return 0; 1455 1456 case DIOCGCACHE: 1457 { 1458 int dkcache = 0; 1459 1460 /* 1461 * We pass this call down to all components and report 1462 * intersection of the flags returned by the components. 1463 * If any errors out, we return error. CCD components 1464 * can not change unless the device is unconfigured, so 1465 * device feature flags will remain static. RCE/WCE can change 1466 * of course, if set directly on underlying device. 1467 */ 1468 for (error = 0, i = 0; i < cs->sc_nccdisks; i++) { 1469 error = VOP_IOCTL(cs->sc_cinfo[i].ci_vp, cmd, &j, 1470 flag, uc); 1471 if (error) 1472 break; 1473 1474 if (i == 0) 1475 dkcache = j; 1476 else 1477 dkcache = DKCACHE_COMBINE(dkcache, j); 1478 } 1479 1480 *((int *)data) = dkcache; 1481 break; 1482 } 1483 1484 case DIOCCACHESYNC: 1485 /* 1486 * We pass this call down to all components and report 1487 * the first error we encounter. 1488 */ 1489 for (error = 0, i = 0; i < cs->sc_nccdisks; i++) { 1490 j = VOP_IOCTL(cs->sc_cinfo[i].ci_vp, cmd, data, 1491 flag, uc); 1492 if (j != 0 && error == 0) 1493 error = j; 1494 } 1495 break; 1496 1497 default: 1498 error = ENOTTY; 1499 break; 1500 } 1501 1502 out: 1503 mutex_exit(&cs->sc_dvlock); 1504 return (error); 1505 } 1506 1507 static int 1508 ccdsize(dev_t dev) 1509 { 1510 struct ccd_softc *cs; 1511 struct disklabel *lp; 1512 int part, unit, omask, size; 1513 1514 unit = ccdunit(dev); 1515 if ((cs = ccdget(unit, 0)) == NULL) 1516 return -1; 1517 1518 if ((cs->sc_flags & CCDF_INITED) == 0) 1519 return (-1); 1520 1521 part = DISKPART(dev); 1522 omask = cs->sc_dkdev.dk_openmask & (1 << part); 1523 lp = cs->sc_dkdev.dk_label; 1524 1525 if (omask == 0 && ccdopen(dev, 0, S_IFBLK, curlwp)) 1526 return (-1); 1527 1528 if (lp->d_partitions[part].p_fstype != FS_SWAP) 1529 size = -1; 1530 else 1531 size = lp->d_partitions[part].p_size * 1532 (lp->d_secsize / DEV_BSIZE); 1533 1534 if (omask == 0 && ccdclose(dev, 0, S_IFBLK, curlwp)) 1535 return (-1); 1536 1537 return (size); 1538 } 1539 1540 static void 1541 ccdgetdefaultlabel(struct ccd_softc *cs, struct disklabel *lp) 1542 { 1543 struct ccdgeom *ccg = &cs->sc_geom; 1544 1545 memset(lp, 0, sizeof(*lp)); 1546 1547 if (cs->sc_size > UINT32_MAX) 1548 lp->d_secperunit = UINT32_MAX; 1549 else 1550 lp->d_secperunit = cs->sc_size; 1551 lp->d_secsize = ccg->ccg_secsize; 1552 lp->d_nsectors = ccg->ccg_nsectors; 1553 lp->d_ntracks = ccg->ccg_ntracks; 1554 lp->d_ncylinders = ccg->ccg_ncylinders; 1555 lp->d_secpercyl = lp->d_ntracks * lp->d_nsectors; 1556 1557 strncpy(lp->d_typename, "ccd", sizeof(lp->d_typename)); 1558 lp->d_type = DKTYPE_CCD; 1559 strncpy(lp->d_packname, "fictitious", sizeof(lp->d_packname)); 1560 lp->d_rpm = 3600; 1561 lp->d_interleave = 1; 1562 lp->d_flags = 0; 1563 1564 lp->d_partitions[RAW_PART].p_offset = 0; 1565 lp->d_partitions[RAW_PART].p_size = lp->d_secperunit; 1566 lp->d_partitions[RAW_PART].p_fstype = FS_UNUSED; 1567 lp->d_npartitions = RAW_PART + 1; 1568 1569 lp->d_magic = DISKMAGIC; 1570 lp->d_magic2 = DISKMAGIC; 1571 lp->d_checksum = dkcksum(cs->sc_dkdev.dk_label); 1572 } 1573 1574 /* 1575 * Read the disklabel from the ccd. If one is not present, fake one 1576 * up. 1577 */ 1578 static void 1579 ccdgetdisklabel(dev_t dev) 1580 { 1581 int unit = ccdunit(dev); 1582 struct ccd_softc *cs; 1583 const char *errstring; 1584 struct disklabel *lp; 1585 struct cpu_disklabel *clp; 1586 1587 if ((cs = ccdget(unit, 0)) == NULL) 1588 return; 1589 lp = cs->sc_dkdev.dk_label; 1590 clp = cs->sc_dkdev.dk_cpulabel; 1591 KASSERT(mutex_owned(&cs->sc_dvlock)); 1592 1593 memset(clp, 0, sizeof(*clp)); 1594 1595 ccdgetdefaultlabel(cs, lp); 1596 1597 /* 1598 * Call the generic disklabel extraction routine. 1599 */ 1600 cs->sc_flags |= CCDF_RLABEL; 1601 if ((cs->sc_flags & CCDF_NOLABEL) != 0) 1602 errstring = "CCDF_NOLABEL set; ignoring on-disk label"; 1603 else 1604 errstring = readdisklabel(CCDLABELDEV(dev), ccdstrategy, 1605 cs->sc_dkdev.dk_label, cs->sc_dkdev.dk_cpulabel); 1606 if (errstring) 1607 ccdmakedisklabel(cs); 1608 else { 1609 int i; 1610 struct partition *pp; 1611 1612 /* 1613 * Sanity check whether the found disklabel is valid. 1614 * 1615 * This is necessary since total size of ccd may vary 1616 * when an interleave is changed even though exactly 1617 * same componets are used, and old disklabel may used 1618 * if that is found. 1619 */ 1620 if (lp->d_secperunit < UINT32_MAX ? 1621 lp->d_secperunit != cs->sc_size : 1622 lp->d_secperunit > cs->sc_size) 1623 printf("WARNING: %s: " 1624 "total sector size in disklabel (%ju) != " 1625 "the size of ccd (%ju)\n", cs->sc_xname, 1626 (uintmax_t)lp->d_secperunit, 1627 (uintmax_t)cs->sc_size); 1628 for (i = 0; i < lp->d_npartitions; i++) { 1629 pp = &lp->d_partitions[i]; 1630 if (pp->p_offset + pp->p_size > cs->sc_size) 1631 printf("WARNING: %s: end of partition `%c' " 1632 "exceeds the size of ccd (%ju)\n", 1633 cs->sc_xname, 'a' + i, (uintmax_t)cs->sc_size); 1634 } 1635 } 1636 1637 #ifdef DEBUG 1638 /* It's actually extremely common to have unlabeled ccds. */ 1639 if (ccddebug & CCDB_LABEL) 1640 if (errstring != NULL) 1641 printf("%s: %s\n", cs->sc_xname, errstring); 1642 #endif 1643 1644 /* In-core label now valid. */ 1645 cs->sc_flags = (cs->sc_flags | CCDF_VLABEL) & ~CCDF_RLABEL; 1646 } 1647 1648 /* 1649 * Take care of things one might want to take care of in the event 1650 * that a disklabel isn't present. 1651 */ 1652 static void 1653 ccdmakedisklabel(struct ccd_softc *cs) 1654 { 1655 struct disklabel *lp = cs->sc_dkdev.dk_label; 1656 1657 /* 1658 * For historical reasons, if there's no disklabel present 1659 * the raw partition must be marked FS_BSDFFS. 1660 */ 1661 lp->d_partitions[RAW_PART].p_fstype = FS_BSDFFS; 1662 1663 strncpy(lp->d_packname, "default label", sizeof(lp->d_packname)); 1664 1665 lp->d_checksum = dkcksum(lp); 1666 } 1667 1668 #ifdef DEBUG 1669 static void 1670 printiinfo(struct ccdiinfo *ii) 1671 { 1672 int ix, i; 1673 1674 for (ix = 0; ii->ii_ndisk; ix++, ii++) { 1675 printf(" itab[%d]: #dk %d sblk %" PRId64 " soff %" PRId64, 1676 ix, ii->ii_ndisk, ii->ii_startblk, ii->ii_startoff); 1677 for (i = 0; i < ii->ii_ndisk; i++) 1678 printf(" %d", ii->ii_index[i]); 1679 printf("\n"); 1680 } 1681 } 1682 #endif 1683 1684 MODULE(MODULE_CLASS_DRIVER, ccd, "dk_subr,bufq_fcfs"); 1685 1686 static int 1687 ccd_modcmd(modcmd_t cmd, void *arg) 1688 { 1689 int error = 0; 1690 #ifdef _MODULE 1691 int bmajor = -1, cmajor = -1; 1692 #endif 1693 1694 1695 switch (cmd) { 1696 case MODULE_CMD_INIT: 1697 #ifdef _MODULE 1698 ccdattach(0); 1699 1700 error = devsw_attach("ccd", &ccd_bdevsw, &bmajor, 1701 &ccd_cdevsw, &cmajor); 1702 #endif 1703 break; 1704 1705 case MODULE_CMD_FINI: 1706 #ifdef _MODULE 1707 mutex_enter(&ccd_lock); 1708 if (!LIST_EMPTY(&ccds)) { 1709 mutex_exit(&ccd_lock); 1710 error = EBUSY; 1711 } else { 1712 mutex_exit(&ccd_lock); 1713 devsw_detach(&ccd_bdevsw, &ccd_cdevsw); 1714 ccddetach(); 1715 } 1716 #endif 1717 break; 1718 1719 case MODULE_CMD_STAT: 1720 return ENOTTY; 1721 1722 default: 1723 return ENOTTY; 1724 } 1725 1726 return error; 1727 } 1728 1729 static int 1730 ccd_units_sysctl(SYSCTLFN_ARGS) 1731 { 1732 struct sysctlnode node; 1733 struct ccd_softc *sc; 1734 int error, i, nccd, *units; 1735 size_t size; 1736 1737 nccd = 0; 1738 mutex_enter(&ccd_lock); 1739 LIST_FOREACH(sc, &ccds, sc_link) 1740 nccd++; 1741 mutex_exit(&ccd_lock); 1742 1743 if (nccd != 0) { 1744 size = nccd * sizeof(*units); 1745 units = kmem_zalloc(size, KM_SLEEP); 1746 i = 0; 1747 mutex_enter(&ccd_lock); 1748 LIST_FOREACH(sc, &ccds, sc_link) { 1749 if (i >= nccd) 1750 break; 1751 units[i] = sc->sc_unit; 1752 } 1753 mutex_exit(&ccd_lock); 1754 } else { 1755 units = NULL; 1756 size = 0; 1757 } 1758 1759 node = *rnode; 1760 node.sysctl_data = units; 1761 node.sysctl_size = size; 1762 1763 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1764 if (units) 1765 kmem_free(units, size); 1766 return error; 1767 } 1768 1769 static int 1770 ccd_info_sysctl(SYSCTLFN_ARGS) 1771 { 1772 struct sysctlnode node; 1773 struct ccddiskinfo ccd; 1774 struct ccd_softc *sc; 1775 int unit; 1776 1777 if (newp == NULL || newlen != sizeof(int)) 1778 return EINVAL; 1779 1780 unit = *(const int *)newp; 1781 newp = NULL; 1782 newlen = 0; 1783 ccd.ccd_ndisks = ~0; 1784 mutex_enter(&ccd_lock); 1785 LIST_FOREACH(sc, &ccds, sc_link) { 1786 if (sc->sc_unit == unit) { 1787 ccd.ccd_ileave = sc->sc_ileave; 1788 ccd.ccd_size = sc->sc_size; 1789 ccd.ccd_ndisks = sc->sc_nccdisks; 1790 ccd.ccd_flags = sc->sc_flags; 1791 break; 1792 } 1793 } 1794 mutex_exit(&ccd_lock); 1795 1796 if (ccd.ccd_ndisks == ~0) 1797 return ENOENT; 1798 1799 node = *rnode; 1800 node.sysctl_data = &ccd; 1801 node.sysctl_size = sizeof(ccd); 1802 1803 return sysctl_lookup(SYSCTLFN_CALL(&node)); 1804 } 1805 1806 static int 1807 ccd_components_sysctl(SYSCTLFN_ARGS) 1808 { 1809 struct sysctlnode node; 1810 int error, unit; 1811 size_t size; 1812 char *names, *p, *ep; 1813 struct ccd_softc *sc; 1814 1815 if (newp == NULL || newlen != sizeof(int)) 1816 return EINVAL; 1817 1818 size = 0; 1819 unit = *(const int *)newp; 1820 newp = NULL; 1821 newlen = 0; 1822 mutex_enter(&ccd_lock); 1823 LIST_FOREACH(sc, &ccds, sc_link) 1824 if (sc->sc_unit == unit) { 1825 for (size_t i = 0; i < sc->sc_nccdisks; i++) 1826 size += strlen(sc->sc_cinfo[i].ci_path) + 1; 1827 break; 1828 } 1829 mutex_exit(&ccd_lock); 1830 1831 if (size == 0) 1832 return ENOENT; 1833 names = kmem_zalloc(size, KM_SLEEP); 1834 p = names; 1835 ep = names + size; 1836 mutex_enter(&ccd_lock); 1837 LIST_FOREACH(sc, &ccds, sc_link) 1838 if (sc->sc_unit == unit) { 1839 for (size_t i = 0; i < sc->sc_nccdisks; i++) { 1840 char *d = sc->sc_cinfo[i].ci_path; 1841 while (p < ep && (*p++ = *d++) != '\0') 1842 continue; 1843 } 1844 break; 1845 } 1846 mutex_exit(&ccd_lock); 1847 1848 node = *rnode; 1849 node.sysctl_data = names; 1850 node.sysctl_size = ep - names; 1851 1852 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1853 kmem_free(names, size); 1854 return error; 1855 } 1856 1857 SYSCTL_SETUP(sysctl_kern_ccd_setup, "sysctl kern.ccd subtree setup") 1858 { 1859 const struct sysctlnode *node = NULL; 1860 1861 sysctl_createv(clog, 0, NULL, &node, 1862 CTLFLAG_PERMANENT, 1863 CTLTYPE_NODE, "ccd", 1864 SYSCTL_DESCR("ConCatenated Disk state"), 1865 NULL, 0, NULL, 0, 1866 CTL_KERN, CTL_CREATE, CTL_EOL); 1867 1868 if (node == NULL) 1869 return; 1870 1871 sysctl_createv(clog, 0, &node, NULL, 1872 CTLFLAG_PERMANENT | CTLFLAG_READONLY, 1873 CTLTYPE_STRUCT, "units", 1874 SYSCTL_DESCR("List of ccd unit numbers"), 1875 ccd_units_sysctl, 0, NULL, 0, 1876 CTL_CREATE, CTL_EOL); 1877 sysctl_createv(clog, 0, &node, NULL, 1878 CTLFLAG_PERMANENT | CTLFLAG_READWRITE, 1879 CTLTYPE_STRUCT, "info", 1880 SYSCTL_DESCR("Information about a CCD unit"), 1881 ccd_info_sysctl, 0, NULL, 0, 1882 CTL_CREATE, CTL_EOL); 1883 sysctl_createv(clog, 0, &node, NULL, 1884 CTLFLAG_PERMANENT | CTLFLAG_READWRITE, 1885 CTLTYPE_STRUCT, "components", 1886 SYSCTL_DESCR("Information about CCD components"), 1887 ccd_components_sysctl, 0, NULL, 0, 1888 CTL_CREATE, CTL_EOL); 1889 } 1890