1 /*- 2 * SPDX-License-Identifier: (BSD-2-Clause-NetBSD AND BSD-3-Clause) 3 * 4 * Copyright (c) 2003 Poul-Henning Kamp. 5 * Copyright (c) 1996, 1997 The NetBSD Foundation, Inc. 6 * All rights reserved. 7 * 8 * This code is derived from software contributed to The NetBSD Foundation 9 * by Jason R. Thorpe. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 * 32 * $NetBSD: ccd.c,v 1.22 1995/12/08 19:13:26 thorpej Exp $ 33 */ 34 35 /*- 36 * Copyright (c) 1988 University of Utah. 37 * Copyright (c) 1990, 1993 38 * The Regents of the University of California. All rights reserved. 39 * 40 * This code is derived from software contributed to Berkeley by 41 * the Systems Programming Group of the University of Utah Computer 42 * Science Department. 43 * 44 * Redistribution and use in source and binary forms, with or without 45 * modification, are permitted provided that the following conditions 46 * are met: 47 * 1. Redistributions of source code must retain the above copyright 48 * notice, this list of conditions and the following disclaimer. 49 * 2. Redistributions in binary form must reproduce the above copyright 50 * notice, this list of conditions and the following disclaimer in the 51 * documentation and/or other materials provided with the distribution. 52 * 3. Neither the name of the University nor the names of its contributors 53 * may be used to endorse or promote products derived from this software 54 * without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 57 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 58 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 59 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 60 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 61 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 62 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 63 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 64 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 65 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 66 * SUCH DAMAGE. 67 * 68 * from: Utah $Hdr: cd.c 1.6 90/11/28$ 69 * 70 * @(#)cd.c 8.2 (Berkeley) 11/16/93 71 */ 72 73 /* 74 * Dynamic configuration and disklabel support by: 75 * Jason R. Thorpe <thorpej@nas.nasa.gov> 76 * Numerical Aerodynamic Simulation Facility 77 * Mail Stop 258-6 78 * NASA Ames Research Center 79 * Moffett Field, CA 94035 80 */ 81 82 #include <sys/cdefs.h> 83 __FBSDID("$FreeBSD$"); 84 85 #include <sys/param.h> 86 #include <sys/systm.h> 87 #include <sys/kernel.h> 88 #include <sys/module.h> 89 #include <sys/bio.h> 90 #include <sys/malloc.h> 91 #include <sys/sbuf.h> 92 #include <geom/geom.h> 93 94 /* 95 * Number of blocks to untouched in front of a component partition. 96 * This is to avoid violating its disklabel area when it starts at the 97 * beginning of the slice. 98 */ 99 #if !defined(CCD_OFFSET) 100 #define CCD_OFFSET 16 101 #endif 102 103 /* sc_flags */ 104 #define CCDF_UNIFORM 0x02 /* use LCCD of sizes for uniform interleave */ 105 #define CCDF_MIRROR 0x04 /* use mirroring */ 106 #define CCDF_NO_OFFSET 0x08 /* do not leave space in front */ 107 #define CCDF_LINUX 0x10 /* use Linux compatibility mode */ 108 109 /* Mask of user-settable ccd flags. */ 110 #define CCDF_USERMASK (CCDF_UNIFORM|CCDF_MIRROR) 111 112 /* 113 * Interleave description table. 114 * Computed at boot time to speed irregular-interleave lookups. 115 * The idea is that we interleave in "groups". First we interleave 116 * evenly over all component disks up to the size of the smallest 117 * component (the first group), then we interleave evenly over all 118 * remaining disks up to the size of the next-smallest (second group), 119 * and so on. 120 * 121 * Each table entry describes the interleave characteristics of one 122 * of these groups. For example if a concatenated disk consisted of 123 * three components of 5, 3, and 7 DEV_BSIZE blocks interleaved at 124 * DEV_BSIZE (1), the table would have three entries: 125 * 126 * ndisk startblk startoff dev 127 * 3 0 0 0, 1, 2 128 * 2 9 3 0, 2 129 * 1 13 5 2 130 * 0 - - - 131 * 132 * which says that the first nine blocks (0-8) are interleaved over 133 * 3 disks (0, 1, 2) starting at block offset 0 on any component disk, 134 * the next 4 blocks (9-12) are interleaved over 2 disks (0, 2) starting 135 * at component block 3, and the remaining blocks (13-14) are on disk 136 * 2 starting at offset 5. 137 */ 138 struct ccdiinfo { 139 int ii_ndisk; /* # of disks range is interleaved over */ 140 daddr_t ii_startblk; /* starting scaled block # for range */ 141 daddr_t ii_startoff; /* starting component offset (block #) */ 142 int *ii_index; /* ordered list of components in range */ 143 }; 144 145 /* 146 * Component info table. 147 * Describes a single component of a concatenated disk. 148 */ 149 struct ccdcinfo { 150 daddr_t ci_size; /* size */ 151 struct g_provider *ci_provider; /* provider */ 152 struct g_consumer *ci_consumer; /* consumer */ 153 }; 154 155 /* 156 * A concatenated disk is described by this structure. 157 */ 158 159 struct ccd_s { 160 LIST_ENTRY(ccd_s) list; 161 162 int sc_unit; /* logical unit number */ 163 int sc_flags; /* flags */ 164 daddr_t sc_size; /* size of ccd */ 165 int sc_ileave; /* interleave */ 166 u_int sc_ndisks; /* number of components */ 167 struct ccdcinfo *sc_cinfo; /* component info */ 168 struct ccdiinfo *sc_itable; /* interleave table */ 169 u_int32_t sc_secsize; /* # bytes per sector */ 170 int sc_pick; /* side of mirror picked */ 171 daddr_t sc_blk[2]; /* mirror localization */ 172 u_int32_t sc_offset; /* actual offset used */ 173 }; 174 175 static g_start_t g_ccd_start; 176 static void ccdiodone(struct bio *bp); 177 static void ccdinterleave(struct ccd_s *); 178 static int ccdinit(struct gctl_req *req, struct ccd_s *); 179 static int ccdbuffer(struct bio **ret, struct ccd_s *, 180 struct bio *, daddr_t, caddr_t, long); 181 182 static void 183 g_ccd_orphan(struct g_consumer *cp) 184 { 185 /* 186 * XXX: We don't do anything here. It is not obvious 187 * XXX: what DTRT would be, so we do what the previous 188 * XXX: code did: ignore it and let the user cope. 189 */ 190 } 191 192 static int 193 g_ccd_access(struct g_provider *pp, int dr, int dw, int de) 194 { 195 struct g_geom *gp; 196 struct g_consumer *cp1, *cp2; 197 int error; 198 199 de += dr; 200 de += dw; 201 202 gp = pp->geom; 203 error = ENXIO; 204 LIST_FOREACH(cp1, &gp->consumer, consumer) { 205 error = g_access(cp1, dr, dw, de); 206 if (error) { 207 LIST_FOREACH(cp2, &gp->consumer, consumer) { 208 if (cp1 == cp2) 209 break; 210 g_access(cp2, -dr, -dw, -de); 211 } 212 break; 213 } 214 } 215 return (error); 216 } 217 218 /* 219 * Free the softc and its substructures. 220 */ 221 static void 222 g_ccd_freesc(struct ccd_s *sc) 223 { 224 struct ccdiinfo *ii; 225 226 g_free(sc->sc_cinfo); 227 if (sc->sc_itable != NULL) { 228 for (ii = sc->sc_itable; ii->ii_ndisk > 0; ii++) 229 if (ii->ii_index != NULL) 230 g_free(ii->ii_index); 231 g_free(sc->sc_itable); 232 } 233 g_free(sc); 234 } 235 236 237 static int 238 ccdinit(struct gctl_req *req, struct ccd_s *cs) 239 { 240 struct ccdcinfo *ci; 241 daddr_t size; 242 int ix; 243 daddr_t minsize; 244 int maxsecsize; 245 off_t mediasize; 246 u_int sectorsize; 247 248 cs->sc_size = 0; 249 250 maxsecsize = 0; 251 minsize = 0; 252 253 if (cs->sc_flags & CCDF_LINUX) { 254 cs->sc_offset = 0; 255 cs->sc_ileave *= 2; 256 if (cs->sc_flags & CCDF_MIRROR && cs->sc_ndisks != 2) 257 gctl_error(req, "Mirror mode for Linux raids is " 258 "only supported with 2 devices"); 259 } else { 260 if (cs->sc_flags & CCDF_NO_OFFSET) 261 cs->sc_offset = 0; 262 else 263 cs->sc_offset = CCD_OFFSET; 264 265 } 266 for (ix = 0; ix < cs->sc_ndisks; ix++) { 267 ci = &cs->sc_cinfo[ix]; 268 269 mediasize = ci->ci_provider->mediasize; 270 sectorsize = ci->ci_provider->sectorsize; 271 if (sectorsize > maxsecsize) 272 maxsecsize = sectorsize; 273 size = mediasize / DEV_BSIZE - cs->sc_offset; 274 275 /* Truncate to interleave boundary */ 276 277 if (cs->sc_ileave > 1) 278 size -= size % cs->sc_ileave; 279 280 if (size == 0) { 281 gctl_error(req, "Component %s has effective size zero", 282 ci->ci_provider->name); 283 return(ENODEV); 284 } 285 286 if (minsize == 0 || size < minsize) 287 minsize = size; 288 ci->ci_size = size; 289 cs->sc_size += size; 290 } 291 292 /* 293 * Don't allow the interleave to be smaller than 294 * the biggest component sector. 295 */ 296 if ((cs->sc_ileave > 0) && 297 (cs->sc_ileave < (maxsecsize / DEV_BSIZE))) { 298 gctl_error(req, "Interleave to small for sector size"); 299 return(EINVAL); 300 } 301 302 /* 303 * If uniform interleave is desired set all sizes to that of 304 * the smallest component. This will guarantee that a single 305 * interleave table is generated. 306 * 307 * Lost space must be taken into account when calculating the 308 * overall size. Half the space is lost when CCDF_MIRROR is 309 * specified. 310 */ 311 if (cs->sc_flags & CCDF_UNIFORM) { 312 for (ix = 0; ix < cs->sc_ndisks; ix++) { 313 ci = &cs->sc_cinfo[ix]; 314 ci->ci_size = minsize; 315 } 316 cs->sc_size = cs->sc_ndisks * minsize; 317 } 318 319 if (cs->sc_flags & CCDF_MIRROR) { 320 /* 321 * Check to see if an even number of components 322 * have been specified. The interleave must also 323 * be non-zero in order for us to be able to 324 * guarantee the topology. 325 */ 326 if (cs->sc_ndisks % 2) { 327 gctl_error(req, 328 "Mirroring requires an even number of disks"); 329 return(EINVAL); 330 } 331 if (cs->sc_ileave == 0) { 332 gctl_error(req, 333 "An interleave must be specified when mirroring"); 334 return(EINVAL); 335 } 336 cs->sc_size = (cs->sc_ndisks/2) * minsize; 337 } 338 339 /* 340 * Construct the interleave table. 341 */ 342 ccdinterleave(cs); 343 344 /* 345 * Create pseudo-geometry based on 1MB cylinders. It's 346 * pretty close. 347 */ 348 cs->sc_secsize = maxsecsize; 349 350 return (0); 351 } 352 353 static void 354 ccdinterleave(struct ccd_s *cs) 355 { 356 struct ccdcinfo *ci, *smallci; 357 struct ccdiinfo *ii; 358 daddr_t bn, lbn; 359 int ix; 360 daddr_t size; 361 362 363 /* 364 * Allocate an interleave table. The worst case occurs when each 365 * of N disks is of a different size, resulting in N interleave 366 * tables. 367 * 368 * Chances are this is too big, but we don't care. 369 */ 370 size = (cs->sc_ndisks + 1) * sizeof(struct ccdiinfo); 371 cs->sc_itable = g_malloc(size, M_WAITOK | M_ZERO); 372 373 /* 374 * Trivial case: no interleave (actually interleave of disk size). 375 * Each table entry represents a single component in its entirety. 376 * 377 * An interleave of 0 may not be used with a mirror setup. 378 */ 379 if (cs->sc_ileave == 0) { 380 bn = 0; 381 ii = cs->sc_itable; 382 383 for (ix = 0; ix < cs->sc_ndisks; ix++) { 384 /* Allocate space for ii_index. */ 385 ii->ii_index = g_malloc(sizeof(int), M_WAITOK); 386 ii->ii_ndisk = 1; 387 ii->ii_startblk = bn; 388 ii->ii_startoff = 0; 389 ii->ii_index[0] = ix; 390 bn += cs->sc_cinfo[ix].ci_size; 391 ii++; 392 } 393 ii->ii_ndisk = 0; 394 return; 395 } 396 397 /* 398 * The following isn't fast or pretty; it doesn't have to be. 399 */ 400 size = 0; 401 bn = lbn = 0; 402 for (ii = cs->sc_itable; ; ii++) { 403 /* 404 * Allocate space for ii_index. We might allocate more then 405 * we use. 406 */ 407 ii->ii_index = g_malloc((sizeof(int) * cs->sc_ndisks), 408 M_WAITOK); 409 410 /* 411 * Locate the smallest of the remaining components 412 */ 413 smallci = NULL; 414 for (ci = cs->sc_cinfo; ci < &cs->sc_cinfo[cs->sc_ndisks]; 415 ci++) { 416 if (ci->ci_size > size && 417 (smallci == NULL || 418 ci->ci_size < smallci->ci_size)) { 419 smallci = ci; 420 } 421 } 422 423 /* 424 * Nobody left, all done 425 */ 426 if (smallci == NULL) { 427 ii->ii_ndisk = 0; 428 g_free(ii->ii_index); 429 ii->ii_index = NULL; 430 break; 431 } 432 433 /* 434 * Record starting logical block using an sc_ileave blocksize. 435 */ 436 ii->ii_startblk = bn / cs->sc_ileave; 437 438 /* 439 * Record starting component block using an sc_ileave 440 * blocksize. This value is relative to the beginning of 441 * a component disk. 442 */ 443 ii->ii_startoff = lbn; 444 445 /* 446 * Determine how many disks take part in this interleave 447 * and record their indices. 448 */ 449 ix = 0; 450 for (ci = cs->sc_cinfo; 451 ci < &cs->sc_cinfo[cs->sc_ndisks]; ci++) { 452 if (ci->ci_size >= smallci->ci_size) { 453 ii->ii_index[ix++] = ci - cs->sc_cinfo; 454 } 455 } 456 ii->ii_ndisk = ix; 457 bn += ix * (smallci->ci_size - size); 458 lbn = smallci->ci_size / cs->sc_ileave; 459 size = smallci->ci_size; 460 } 461 } 462 463 static void 464 g_ccd_start(struct bio *bp) 465 { 466 long bcount, rcount; 467 struct bio *cbp[2]; 468 caddr_t addr; 469 daddr_t bn; 470 int err; 471 struct ccd_s *cs; 472 473 cs = bp->bio_to->geom->softc; 474 475 /* 476 * Block all GETATTR requests, we wouldn't know which of our 477 * subdevices we should ship it off to. 478 * XXX: this may not be the right policy. 479 */ 480 if(bp->bio_cmd == BIO_GETATTR) { 481 g_io_deliver(bp, EINVAL); 482 return; 483 } 484 485 /* 486 * Translate the partition-relative block number to an absolute. 487 */ 488 bn = bp->bio_offset / cs->sc_secsize; 489 490 /* 491 * Allocate component buffers and fire off the requests 492 */ 493 addr = bp->bio_data; 494 for (bcount = bp->bio_length; bcount > 0; bcount -= rcount) { 495 err = ccdbuffer(cbp, cs, bp, bn, addr, bcount); 496 if (err) { 497 bp->bio_completed += bcount; 498 if (bp->bio_error == 0) 499 bp->bio_error = err; 500 if (bp->bio_completed == bp->bio_length) 501 g_io_deliver(bp, bp->bio_error); 502 return; 503 } 504 rcount = cbp[0]->bio_length; 505 506 if (cs->sc_flags & CCDF_MIRROR) { 507 /* 508 * Mirroring. Writes go to both disks, reads are 509 * taken from whichever disk seems most appropriate. 510 * 511 * We attempt to localize reads to the disk whos arm 512 * is nearest the read request. We ignore seeks due 513 * to writes when making this determination and we 514 * also try to avoid hogging. 515 */ 516 if (cbp[0]->bio_cmd != BIO_READ) { 517 g_io_request(cbp[0], cbp[0]->bio_from); 518 g_io_request(cbp[1], cbp[1]->bio_from); 519 } else { 520 int pick = cs->sc_pick; 521 daddr_t range = cs->sc_size / 16; 522 523 if (bn < cs->sc_blk[pick] - range || 524 bn > cs->sc_blk[pick] + range 525 ) { 526 cs->sc_pick = pick = 1 - pick; 527 } 528 cs->sc_blk[pick] = bn + btodb(rcount); 529 g_io_request(cbp[pick], cbp[pick]->bio_from); 530 } 531 } else { 532 /* 533 * Not mirroring 534 */ 535 g_io_request(cbp[0], cbp[0]->bio_from); 536 } 537 bn += btodb(rcount); 538 addr += rcount; 539 } 540 } 541 542 /* 543 * Build a component buffer header. 544 */ 545 static int 546 ccdbuffer(struct bio **cb, struct ccd_s *cs, struct bio *bp, daddr_t bn, caddr_t addr, long bcount) 547 { 548 struct ccdcinfo *ci, *ci2 = NULL; 549 struct bio *cbp; 550 daddr_t cbn, cboff; 551 off_t cbc; 552 553 /* 554 * Determine which component bn falls in. 555 */ 556 cbn = bn; 557 cboff = 0; 558 559 if (cs->sc_ileave == 0) { 560 /* 561 * Serially concatenated and neither a mirror nor a parity 562 * config. This is a special case. 563 */ 564 daddr_t sblk; 565 566 sblk = 0; 567 for (ci = cs->sc_cinfo; cbn >= sblk + ci->ci_size; ci++) 568 sblk += ci->ci_size; 569 cbn -= sblk; 570 } else { 571 struct ccdiinfo *ii; 572 int ccdisk, off; 573 574 /* 575 * Calculate cbn, the logical superblock (sc_ileave chunks), 576 * and cboff, a normal block offset (DEV_BSIZE chunks) relative 577 * to cbn. 578 */ 579 cboff = cbn % cs->sc_ileave; /* DEV_BSIZE gran */ 580 cbn = cbn / cs->sc_ileave; /* DEV_BSIZE * ileave gran */ 581 582 /* 583 * Figure out which interleave table to use. 584 */ 585 for (ii = cs->sc_itable; ii->ii_ndisk; ii++) { 586 if (ii->ii_startblk > cbn) 587 break; 588 } 589 ii--; 590 591 /* 592 * off is the logical superblock relative to the beginning 593 * of this interleave block. 594 */ 595 off = cbn - ii->ii_startblk; 596 597 /* 598 * We must calculate which disk component to use (ccdisk), 599 * and recalculate cbn to be the superblock relative to 600 * the beginning of the component. This is typically done by 601 * adding 'off' and ii->ii_startoff together. However, 'off' 602 * must typically be divided by the number of components in 603 * this interleave array to be properly convert it from a 604 * CCD-relative logical superblock number to a 605 * component-relative superblock number. 606 */ 607 if (ii->ii_ndisk == 1) { 608 /* 609 * When we have just one disk, it can't be a mirror 610 * or a parity config. 611 */ 612 ccdisk = ii->ii_index[0]; 613 cbn = ii->ii_startoff + off; 614 } else { 615 if (cs->sc_flags & CCDF_MIRROR) { 616 /* 617 * We have forced a uniform mapping, resulting 618 * in a single interleave array. We double 619 * up on the first half of the available 620 * components and our mirror is in the second 621 * half. This only works with a single 622 * interleave array because doubling up 623 * doubles the number of sectors, so there 624 * cannot be another interleave array because 625 * the next interleave array's calculations 626 * would be off. 627 */ 628 int ndisk2 = ii->ii_ndisk / 2; 629 ccdisk = ii->ii_index[off % ndisk2]; 630 cbn = ii->ii_startoff + off / ndisk2; 631 ci2 = &cs->sc_cinfo[ccdisk + ndisk2]; 632 } else { 633 ccdisk = ii->ii_index[off % ii->ii_ndisk]; 634 cbn = ii->ii_startoff + off / ii->ii_ndisk; 635 } 636 } 637 638 ci = &cs->sc_cinfo[ccdisk]; 639 640 /* 641 * Convert cbn from a superblock to a normal block so it 642 * can be used to calculate (along with cboff) the normal 643 * block index into this particular disk. 644 */ 645 cbn *= cs->sc_ileave; 646 } 647 648 /* 649 * Fill in the component buf structure. 650 */ 651 cbp = g_clone_bio(bp); 652 if (cbp == NULL) 653 return (ENOMEM); 654 cbp->bio_done = g_std_done; 655 cbp->bio_offset = dbtob(cbn + cboff + cs->sc_offset); 656 cbp->bio_data = addr; 657 if (cs->sc_ileave == 0) 658 cbc = dbtob((off_t)(ci->ci_size - cbn)); 659 else 660 cbc = dbtob((off_t)(cs->sc_ileave - cboff)); 661 cbp->bio_length = (cbc < bcount) ? cbc : bcount; 662 663 cbp->bio_from = ci->ci_consumer; 664 cb[0] = cbp; 665 666 if (cs->sc_flags & CCDF_MIRROR) { 667 cbp = g_clone_bio(bp); 668 if (cbp == NULL) 669 return (ENOMEM); 670 cbp->bio_done = cb[0]->bio_done = ccdiodone; 671 cbp->bio_offset = cb[0]->bio_offset; 672 cbp->bio_data = cb[0]->bio_data; 673 cbp->bio_length = cb[0]->bio_length; 674 cbp->bio_from = ci2->ci_consumer; 675 cbp->bio_caller1 = cb[0]; 676 cb[0]->bio_caller1 = cbp; 677 cb[1] = cbp; 678 } 679 return (0); 680 } 681 682 /* 683 * Called only for mirrored operations. 684 */ 685 static void 686 ccdiodone(struct bio *cbp) 687 { 688 struct bio *mbp, *pbp; 689 690 mbp = cbp->bio_caller1; 691 pbp = cbp->bio_parent; 692 693 if (pbp->bio_cmd == BIO_READ) { 694 if (cbp->bio_error == 0) { 695 /* We will not be needing the partner bio */ 696 if (mbp != NULL) { 697 pbp->bio_inbed++; 698 g_destroy_bio(mbp); 699 } 700 g_std_done(cbp); 701 return; 702 } 703 if (mbp != NULL) { 704 /* Try partner the bio instead */ 705 mbp->bio_caller1 = NULL; 706 pbp->bio_inbed++; 707 g_destroy_bio(cbp); 708 g_io_request(mbp, mbp->bio_from); 709 /* 710 * XXX: If this comes back OK, we should actually 711 * try to write the good data on the failed mirror 712 */ 713 return; 714 } 715 g_std_done(cbp); 716 return; 717 } 718 if (mbp != NULL) { 719 mbp->bio_caller1 = NULL; 720 pbp->bio_inbed++; 721 if (cbp->bio_error != 0 && pbp->bio_error == 0) 722 pbp->bio_error = cbp->bio_error; 723 g_destroy_bio(cbp); 724 return; 725 } 726 g_std_done(cbp); 727 } 728 729 static void 730 g_ccd_create(struct gctl_req *req, struct g_class *mp) 731 { 732 int *unit, *ileave, *nprovider; 733 struct g_geom *gp; 734 struct g_consumer *cp; 735 struct g_provider *pp; 736 struct ccd_s *sc; 737 struct sbuf *sb; 738 char buf[20]; 739 int i, error; 740 741 g_topology_assert(); 742 unit = gctl_get_paraml(req, "unit", sizeof (*unit)); 743 if (unit == NULL) { 744 gctl_error(req, "unit parameter not given"); 745 return; 746 } 747 ileave = gctl_get_paraml(req, "ileave", sizeof (*ileave)); 748 if (ileave == NULL) { 749 gctl_error(req, "ileave parameter not given"); 750 return; 751 } 752 nprovider = gctl_get_paraml(req, "nprovider", sizeof (*nprovider)); 753 if (nprovider == NULL) { 754 gctl_error(req, "nprovider parameter not given"); 755 return; 756 } 757 758 /* Check for duplicate unit */ 759 LIST_FOREACH(gp, &mp->geom, geom) { 760 sc = gp->softc; 761 if (sc != NULL && sc->sc_unit == *unit) { 762 gctl_error(req, "Unit %d already configured", *unit); 763 return; 764 } 765 } 766 767 if (*nprovider <= 0) { 768 gctl_error(req, "Bogus nprovider argument (= %d)", *nprovider); 769 return; 770 } 771 772 /* Check all providers are valid */ 773 for (i = 0; i < *nprovider; i++) { 774 sprintf(buf, "provider%d", i); 775 pp = gctl_get_provider(req, buf); 776 if (pp == NULL) 777 return; 778 } 779 780 gp = g_new_geomf(mp, "ccd%d", *unit); 781 sc = g_malloc(sizeof *sc, M_WAITOK | M_ZERO); 782 gp->softc = sc; 783 sc->sc_ndisks = *nprovider; 784 785 /* Allocate space for the component info. */ 786 sc->sc_cinfo = g_malloc(sc->sc_ndisks * sizeof(struct ccdcinfo), 787 M_WAITOK | M_ZERO); 788 789 /* Create consumers and attach to all providers */ 790 for (i = 0; i < *nprovider; i++) { 791 sprintf(buf, "provider%d", i); 792 pp = gctl_get_provider(req, buf); 793 cp = g_new_consumer(gp); 794 error = g_attach(cp, pp); 795 KASSERT(error == 0, ("attach to %s failed", pp->name)); 796 sc->sc_cinfo[i].ci_consumer = cp; 797 sc->sc_cinfo[i].ci_provider = pp; 798 } 799 800 sc->sc_unit = *unit; 801 sc->sc_ileave = *ileave; 802 803 if (gctl_get_param(req, "no_offset", NULL)) 804 sc->sc_flags |= CCDF_NO_OFFSET; 805 if (gctl_get_param(req, "linux", NULL)) 806 sc->sc_flags |= CCDF_LINUX; 807 808 if (gctl_get_param(req, "uniform", NULL)) 809 sc->sc_flags |= CCDF_UNIFORM; 810 if (gctl_get_param(req, "mirror", NULL)) 811 sc->sc_flags |= CCDF_MIRROR; 812 813 if (sc->sc_ileave == 0 && (sc->sc_flags & CCDF_MIRROR)) { 814 printf("%s: disabling mirror, interleave is 0\n", gp->name); 815 sc->sc_flags &= ~(CCDF_MIRROR); 816 } 817 818 if ((sc->sc_flags & CCDF_MIRROR) && !(sc->sc_flags & CCDF_UNIFORM)) { 819 printf("%s: mirror/parity forces uniform flag\n", gp->name); 820 sc->sc_flags |= CCDF_UNIFORM; 821 } 822 823 error = ccdinit(req, sc); 824 if (error != 0) { 825 g_ccd_freesc(sc); 826 gp->softc = NULL; 827 g_wither_geom(gp, ENXIO); 828 return; 829 } 830 831 pp = g_new_providerf(gp, "%s", gp->name); 832 pp->mediasize = sc->sc_size * (off_t)sc->sc_secsize; 833 pp->sectorsize = sc->sc_secsize; 834 g_error_provider(pp, 0); 835 836 sb = sbuf_new_auto(); 837 sbuf_printf(sb, "ccd%d: %d components ", sc->sc_unit, *nprovider); 838 for (i = 0; i < *nprovider; i++) { 839 sbuf_printf(sb, "%s%s", 840 i == 0 ? "(" : ", ", 841 sc->sc_cinfo[i].ci_provider->name); 842 } 843 sbuf_printf(sb, "), %jd blocks ", (off_t)pp->mediasize / DEV_BSIZE); 844 if (sc->sc_ileave != 0) 845 sbuf_printf(sb, "interleaved at %d blocks\n", 846 sc->sc_ileave); 847 else 848 sbuf_printf(sb, "concatenated\n"); 849 sbuf_finish(sb); 850 gctl_set_param_err(req, "output", sbuf_data(sb), sbuf_len(sb) + 1); 851 sbuf_delete(sb); 852 } 853 854 static int 855 g_ccd_destroy_geom(struct gctl_req *req, struct g_class *mp, struct g_geom *gp) 856 { 857 struct g_provider *pp; 858 struct ccd_s *sc; 859 860 g_topology_assert(); 861 sc = gp->softc; 862 pp = LIST_FIRST(&gp->provider); 863 if (sc == NULL || pp == NULL) 864 return (EBUSY); 865 if (pp->acr != 0 || pp->acw != 0 || pp->ace != 0) { 866 gctl_error(req, "%s is open(r%dw%de%d)", gp->name, 867 pp->acr, pp->acw, pp->ace); 868 return (EBUSY); 869 } 870 g_ccd_freesc(sc); 871 gp->softc = NULL; 872 g_wither_geom(gp, ENXIO); 873 return (0); 874 } 875 876 static void 877 g_ccd_list(struct gctl_req *req, struct g_class *mp) 878 { 879 struct sbuf *sb; 880 struct ccd_s *cs; 881 struct g_geom *gp; 882 int i, unit, *up; 883 884 up = gctl_get_paraml(req, "unit", sizeof (*up)); 885 if (up == NULL) { 886 gctl_error(req, "unit parameter not given"); 887 return; 888 } 889 unit = *up; 890 sb = sbuf_new_auto(); 891 LIST_FOREACH(gp, &mp->geom, geom) { 892 cs = gp->softc; 893 if (cs == NULL || (unit >= 0 && unit != cs->sc_unit)) 894 continue; 895 sbuf_printf(sb, "ccd%d\t\t%d\t%d\t", 896 cs->sc_unit, cs->sc_ileave, cs->sc_flags & CCDF_USERMASK); 897 898 for (i = 0; i < cs->sc_ndisks; ++i) { 899 sbuf_printf(sb, "%s/dev/%s", i == 0 ? "" : " ", 900 cs->sc_cinfo[i].ci_provider->name); 901 } 902 sbuf_printf(sb, "\n"); 903 } 904 sbuf_finish(sb); 905 gctl_set_param_err(req, "output", sbuf_data(sb), sbuf_len(sb) + 1); 906 sbuf_delete(sb); 907 } 908 909 static void 910 g_ccd_config(struct gctl_req *req, struct g_class *mp, char const *verb) 911 { 912 struct g_geom *gp; 913 914 g_topology_assert(); 915 if (!strcmp(verb, "create geom")) { 916 g_ccd_create(req, mp); 917 } else if (!strcmp(verb, "destroy geom")) { 918 gp = gctl_get_geom(req, mp, "geom"); 919 if (gp != NULL) 920 g_ccd_destroy_geom(req, mp, gp); 921 } else if (!strcmp(verb, "list")) { 922 g_ccd_list(req, mp); 923 } else { 924 gctl_error(req, "unknown verb"); 925 } 926 } 927 928 static struct g_class g_ccd_class = { 929 .name = "CCD", 930 .version = G_VERSION, 931 .ctlreq = g_ccd_config, 932 .destroy_geom = g_ccd_destroy_geom, 933 .start = g_ccd_start, 934 .orphan = g_ccd_orphan, 935 .access = g_ccd_access, 936 }; 937 938 DECLARE_GEOM_CLASS(g_ccd_class, g_ccd); 939