wd.c revision 1.472
1 /* $NetBSD: wd.c,v 1.472 2025/02/23 02:07:35 jakllsch Exp $ */ 2 3 /* 4 * Copyright (c) 1998, 2001 Manuel Bouyer. All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 18 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 25 */ 26 27 /*- 28 * Copyright (c) 1998, 2003, 2004 The NetBSD Foundation, Inc. 29 * All rights reserved. 30 * 31 * This code is derived from software contributed to The NetBSD Foundation 32 * by Charles M. Hannum and by Onno van der Linden. 33 * 34 * Redistribution and use in source and binary forms, with or without 35 * modification, are permitted provided that the following conditions 36 * are met: 37 * 1. Redistributions of source code must retain the above copyright 38 * notice, this list of conditions and the following disclaimer. 39 * 2. Redistributions in binary form must reproduce the above copyright 40 * notice, this list of conditions and the following disclaimer in the 41 * documentation and/or other materials provided with the distribution. 42 * 43 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 44 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 45 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 46 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 47 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 48 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 49 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 50 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 51 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 52 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 53 * POSSIBILITY OF SUCH DAMAGE. 54 */ 55 56 #include <sys/cdefs.h> 57 __KERNEL_RCSID(0, "$NetBSD: wd.c,v 1.472 2025/02/23 02:07:35 jakllsch Exp $"); 58 59 #include "opt_ata.h" 60 #include "opt_wd.h" 61 62 #include <sys/param.h> 63 #include <sys/systm.h> 64 #include <sys/kernel.h> 65 #include <sys/conf.h> 66 #include <sys/file.h> 67 #include <sys/stat.h> 68 #include <sys/ioctl.h> 69 #include <sys/buf.h> 70 #include <sys/bufq.h> 71 #include <sys/uio.h> 72 #include <sys/device.h> 73 #include <sys/disklabel.h> 74 #include <sys/disk.h> 75 #include <sys/syslog.h> 76 #include <sys/proc.h> 77 #include <sys/reboot.h> 78 #include <sys/vnode.h> 79 #include <sys/rndsource.h> 80 81 #include <sys/intr.h> 82 #include <sys/bus.h> 83 84 #include <dev/ata/atareg.h> 85 #include <dev/ata/atavar.h> 86 #include <dev/ata/wdvar.h> 87 #include <dev/ic/wdcreg.h> 88 #include <sys/ataio.h> 89 #include "locators.h" 90 91 #include <prop/proplib.h> 92 93 #define WDIORETRIES_SINGLE 4 /* number of retries for single-sector */ 94 #define WDIORETRIES 5 /* number of retries before giving up */ 95 #define RECOVERYTIME hz/2 /* time to wait before retrying a cmd */ 96 97 #define WDUNIT(dev) DISKUNIT(dev) 98 #define WDPART(dev) DISKPART(dev) 99 #define WDMINOR(unit, part) DISKMINOR(unit, part) 100 #define MAKEWDDEV(maj, unit, part) MAKEDISKDEV(maj, unit, part) 101 102 #define WDLABELDEV(dev) (MAKEWDDEV(major(dev), WDUNIT(dev), RAW_PART)) 103 104 #define DEBUG_FUNCS 0x08 105 #define DEBUG_PROBE 0x10 106 #define DEBUG_DETACH 0x20 107 #define DEBUG_XFERS 0x40 108 #ifdef ATADEBUG 109 #ifndef ATADEBUG_WD_MASK 110 #define ATADEBUG_WD_MASK 0x0 111 #endif 112 int wdcdebug_wd_mask = ATADEBUG_WD_MASK; 113 #define ATADEBUG_PRINT(args, level) \ 114 if (wdcdebug_wd_mask & (level)) \ 115 printf args 116 #else 117 #define ATADEBUG_PRINT(args, level) 118 #endif 119 120 static int wdprobe(device_t, cfdata_t, void *); 121 static void wdattach(device_t, device_t, void *); 122 static int wddetach(device_t, int); 123 static void wdperror(const struct wd_softc *, struct ata_xfer *); 124 125 static void wdminphys(struct buf *); 126 127 static int wd_firstopen(device_t, dev_t, int, int); 128 static int wd_lastclose(device_t); 129 static bool wd_suspend(device_t, const pmf_qual_t *); 130 static int wd_standby(struct wd_softc *, int); 131 132 CFATTACH_DECL3_NEW(wd, sizeof(struct wd_softc), 133 wdprobe, wdattach, wddetach, NULL, NULL, NULL, DVF_DETACH_SHUTDOWN); 134 135 extern struct cfdriver wd_cd; 136 137 static dev_type_open(wdopen); 138 static dev_type_close(wdclose); 139 static dev_type_read(wdread); 140 static dev_type_write(wdwrite); 141 static dev_type_ioctl(wdioctl); 142 static dev_type_strategy(wdstrategy); 143 static dev_type_dump(wddump); 144 static dev_type_size(wdsize); 145 static dev_type_discard(wddiscard); 146 147 const struct bdevsw wd_bdevsw = { 148 .d_open = wdopen, 149 .d_close = wdclose, 150 .d_strategy = wdstrategy, 151 .d_ioctl = wdioctl, 152 .d_dump = wddump, 153 .d_psize = wdsize, 154 .d_discard = wddiscard, 155 .d_cfdriver = &wd_cd, 156 .d_devtounit = disklabel_dev_unit, 157 .d_flag = D_DISK 158 }; 159 160 const struct cdevsw wd_cdevsw = { 161 .d_open = wdopen, 162 .d_close = wdclose, 163 .d_read = wdread, 164 .d_write = wdwrite, 165 .d_ioctl = wdioctl, 166 .d_stop = nostop, 167 .d_tty = notty, 168 .d_poll = nopoll, 169 .d_mmap = nommap, 170 .d_kqfilter = nokqfilter, 171 .d_discard = wddiscard, 172 .d_cfdriver = &wd_cd, 173 .d_devtounit = disklabel_dev_unit, 174 .d_flag = D_DISK 175 }; 176 177 /* #define WD_DUMP_NOT_TRUSTED if you just want to watch */ 178 static int wddoingadump = 0; 179 static int wddumprecalibrated = 0; 180 181 /* 182 * Glue necessary to hook WDCIOCCOMMAND into physio 183 */ 184 185 struct wd_ioctl { 186 LIST_ENTRY(wd_ioctl) wi_list; 187 struct buf wi_bp; 188 struct uio wi_uio; 189 struct iovec wi_iov; 190 atareq_t wi_atareq; 191 struct wd_softc *wi_softc; 192 }; 193 194 static struct wd_ioctl *wi_find(struct buf *); 195 static void wi_free(struct wd_ioctl *); 196 static struct wd_ioctl *wi_get(struct wd_softc *); 197 static void wdioctlstrategy(struct buf *); 198 199 static void wdrestart(void *); 200 static void wdstart1(struct wd_softc *, struct buf *, struct ata_xfer *); 201 static int wd_diskstart(device_t, struct buf *); 202 static int wd_dumpblocks(device_t, void *, daddr_t, int); 203 static void wd_iosize(device_t, int *); 204 static int wd_discard(device_t, off_t, off_t); 205 static void wdbioretry(void *); 206 static void wdbiorequeue(void *); 207 static void wddone(device_t, struct ata_xfer *); 208 static int wd_get_params(struct wd_softc *, struct ataparams *); 209 static void wd_set_geometry(struct wd_softc *); 210 static int wd_flushcache(struct wd_softc *, int); 211 static int wd_trim(struct wd_softc *, daddr_t, long); 212 static bool wd_shutdown(device_t, int); 213 214 static int wd_getcache(struct wd_softc *, int *); 215 static int wd_setcache(struct wd_softc *, int); 216 217 static void wd_sysctl_attach(struct wd_softc *); 218 static void wd_sysctl_detach(struct wd_softc *); 219 220 static const struct dkdriver wddkdriver = { 221 .d_open = wdopen, 222 .d_close = wdclose, 223 .d_strategy = wdstrategy, 224 .d_minphys = wdminphys, 225 .d_diskstart = wd_diskstart, 226 .d_dumpblocks = wd_dumpblocks, 227 .d_iosize = wd_iosize, 228 .d_firstopen = wd_firstopen, 229 .d_lastclose = wd_lastclose, 230 .d_discard = wd_discard 231 }; 232 233 #ifdef HAS_BAD144_HANDLING 234 static void bad144intern(struct wd_softc *); 235 #endif 236 237 #define WD_QUIRK_SPLIT_MOD15_WRITE 0x0001 /* must split certain writes */ 238 239 #define WD_QUIRK_FMT "\20\1SPLIT_MOD15_WRITE" 240 241 /* 242 * Quirk table for IDE drives. Put more-specific matches first, since 243 * a simple globing routine is used for matching. 244 */ 245 static const struct wd_quirk { 246 const char *wdq_match; /* inquiry pattern to match */ 247 int wdq_quirks; /* drive quirks */ 248 } wd_quirk_table[] = { 249 /* 250 * Some Seagate S-ATA drives have a PHY which can get confused 251 * with the way data is packetized by some S-ATA controllers. 252 * 253 * The work-around is to split in two any write transfer whose 254 * sector count % 15 == 1 (assuming 512 byte sectors). 255 * 256 * XXX This is an incomplete list. There are at least a couple 257 * XXX more model numbers. If you have trouble with such transfers 258 * XXX (8K is the most common) on Seagate S-ATA drives, please 259 * XXX notify thorpej (at) NetBSD.org. 260 * 261 * The ST360015AS has not yet been confirmed to have this 262 * issue, however, it is the only other drive in the 263 * Seagate Barracuda Serial ATA V family. 264 * 265 */ 266 { "ST3120023AS", WD_QUIRK_SPLIT_MOD15_WRITE }, 267 { "ST380023AS", WD_QUIRK_SPLIT_MOD15_WRITE }, 268 { "ST360015AS", WD_QUIRK_SPLIT_MOD15_WRITE }, 269 { NULL, 270 0 } 271 }; 272 273 static const struct wd_quirk * 274 wd_lookup_quirks(const char *name) 275 { 276 const struct wd_quirk *wdq; 277 const char *estr; 278 279 for (wdq = wd_quirk_table; wdq->wdq_match != NULL; wdq++) { 280 /* 281 * We only want exact matches (which include matches 282 * against globbing characters). 283 */ 284 if (pmatch(name, wdq->wdq_match, &estr) == 2) 285 return (wdq); 286 } 287 return (NULL); 288 } 289 290 static int 291 wdprobe(device_t parent, cfdata_t match, void *aux) 292 { 293 struct ata_device *adev = aux; 294 295 if (adev == NULL) 296 return 0; 297 if (adev->adev_bustype->bustype_type != SCSIPI_BUSTYPE_ATA) 298 return 0; 299 300 if (match->cf_loc[ATA_HLCF_DRIVE] != ATA_HLCF_DRIVE_DEFAULT && 301 match->cf_loc[ATA_HLCF_DRIVE] != adev->adev_drv_data->drive) 302 return 0; 303 return 1; 304 } 305 306 static void 307 wdattach(device_t parent, device_t self, void *aux) 308 { 309 struct wd_softc *wd = device_private(self); 310 struct dk_softc *dksc = &wd->sc_dksc; 311 struct ata_device *adev= aux; 312 int i, blank; 313 char tbuf[41],pbuf[9], c, *p, *q; 314 const struct wd_quirk *wdq; 315 int dtype = DKTYPE_UNKNOWN; 316 317 dksc->sc_dev = self; 318 319 ATADEBUG_PRINT(("wdattach\n"), DEBUG_FUNCS | DEBUG_PROBE); 320 mutex_init(&wd->sc_lock, MUTEX_DEFAULT, IPL_BIO); 321 #ifdef WD_SOFTBADSECT 322 SLIST_INIT(&wd->sc_bslist); 323 cv_init(&wd->sc_bslist_cv, "wdbadsect"); 324 #endif 325 wd->atabus = adev->adev_bustype; 326 wd->inflight = 0; 327 wd->drvp = adev->adev_drv_data; 328 329 wd->drvp->drv_openings = 1; 330 wd->drvp->drv_done = wddone; 331 wd->drvp->drv_softc = dksc->sc_dev; /* done in atabusconfig_thread() 332 but too late */ 333 334 SLIST_INIT(&wd->sc_retry_list); 335 SLIST_INIT(&wd->sc_requeue_list); 336 callout_init(&wd->sc_retry_callout, 0); /* XXX MPSAFE */ 337 callout_init(&wd->sc_requeue_callout, 0); /* XXX MPSAFE */ 338 callout_init(&wd->sc_restart_diskqueue, 0); /* XXX MPSAFE */ 339 340 aprint_naive("\n"); 341 aprint_normal("\n"); 342 343 /* read our drive info */ 344 if (wd_get_params(wd, &wd->sc_params) != 0) { 345 aprint_error_dev(self, "IDENTIFY failed\n"); 346 goto out; 347 } 348 349 for (blank = 0, p = wd->sc_params.atap_model, q = tbuf, i = 0; 350 i < sizeof(wd->sc_params.atap_model); i++) { 351 c = *p++; 352 if (c == '\0') 353 break; 354 if (c != ' ') { 355 if (blank) { 356 *q++ = ' '; 357 blank = 0; 358 } 359 *q++ = c; 360 } else 361 blank = 1; 362 } 363 *q++ = '\0'; 364 365 wd->sc_typename = kmem_asprintf("%s", tbuf); 366 aprint_normal_dev(self, "<%s>\n", wd->sc_typename); 367 368 wdq = wd_lookup_quirks(tbuf); 369 if (wdq != NULL) 370 wd->sc_quirks = wdq->wdq_quirks; 371 372 if (wd->sc_quirks != 0) { 373 char sbuf[sizeof(WD_QUIRK_FMT) + 64]; 374 snprintb(sbuf, sizeof(sbuf), WD_QUIRK_FMT, wd->sc_quirks); 375 aprint_normal_dev(self, "quirks %s\n", sbuf); 376 377 if (wd->sc_quirks & WD_QUIRK_SPLIT_MOD15_WRITE) { 378 aprint_error_dev(self, "drive corrupts write transfers with certain controllers, consider replacing\n"); 379 } 380 } 381 382 if ((wd->sc_params.atap_multi & 0xff) > 1) { 383 wd->drvp->multi = wd->sc_params.atap_multi & 0xff; 384 } else { 385 wd->drvp->multi = 1; 386 } 387 388 aprint_verbose_dev(self, "drive supports %d-sector PIO transfers,", 389 wd->drvp->multi); 390 391 /* 48-bit LBA addressing */ 392 if ((wd->sc_params.atap_cmd2_en & ATA_CMD2_LBA48) != 0) 393 wd->sc_flags |= WDF_LBA48; 394 395 /* Prior to ATA-4, LBA was optional. */ 396 if ((wd->sc_params.atap_capabilities1 & WDC_CAP_LBA) != 0) 397 wd->sc_flags |= WDF_LBA; 398 #if 0 399 /* ATA-4 requires LBA. */ 400 if (wd->sc_params.atap_ataversion != 0xffff && 401 wd->sc_params.atap_ataversion >= WDC_VER_ATA4) 402 wd->sc_flags |= WDF_LBA; 403 #endif 404 405 if ((wd->sc_flags & WDF_LBA48) != 0) { 406 aprint_verbose(" LBA48 addressing\n"); 407 wd->sc_capacity = 408 ((uint64_t) wd->sc_params.atap_max_lba[3] << 48) | 409 ((uint64_t) wd->sc_params.atap_max_lba[2] << 32) | 410 ((uint64_t) wd->sc_params.atap_max_lba[1] << 16) | 411 ((uint64_t) wd->sc_params.atap_max_lba[0] << 0); 412 wd->sc_capacity28 = 413 (wd->sc_params.atap_capacity[1] << 16) | 414 wd->sc_params.atap_capacity[0]; 415 /* 416 * Force LBA48 addressing for invalid numbers. 417 */ 418 if (wd->sc_capacity28 > 0xfffffff) 419 wd->sc_capacity28 = 0xfffffff; 420 } else if ((wd->sc_flags & WDF_LBA) != 0) { 421 aprint_verbose(" LBA addressing\n"); 422 wd->sc_capacity28 = 423 (wd->sc_params.atap_capacity[1] << 16) | 424 wd->sc_params.atap_capacity[0]; 425 /* 426 * Limit capacity to LBA28 numbers to avoid overflow. 427 */ 428 if (wd->sc_capacity28 > 0xfffffff) 429 wd->sc_capacity28 = 0xfffffff; 430 wd->sc_capacity = wd->sc_capacity28; 431 } else { 432 aprint_verbose(" chs addressing\n"); 433 wd->sc_capacity = 434 wd->sc_params.atap_cylinders * 435 wd->sc_params.atap_heads * 436 wd->sc_params.atap_sectors; 437 /* 438 * LBA28 size is ignored for CHS addressing. Use a reasonable 439 * value for debugging. The CHS values may be artificial and 440 * are mostly ignored. 441 */ 442 if (wd->sc_capacity < 0xfffffff) 443 wd->sc_capacity28 = wd->sc_capacity; 444 else 445 wd->sc_capacity28 = 0xfffffff; 446 } 447 if ((wd->sc_params.atap_secsz & ATA_SECSZ_VALID_MASK) == ATA_SECSZ_VALID 448 && ((wd->sc_params.atap_secsz & ATA_SECSZ_LLS) != 0)) { 449 wd->sc_blksize = 2ULL * 450 ((uint32_t)((wd->sc_params.atap_lls_secsz[1] << 16) | 451 wd->sc_params.atap_lls_secsz[0])); 452 } else { 453 wd->sc_blksize = 512; 454 } 455 wd->sc_sectoralign.dsa_firstaligned = 0; 456 wd->sc_sectoralign.dsa_alignment = 1; 457 if ((wd->sc_params.atap_secsz & ATA_SECSZ_VALID_MASK) == ATA_SECSZ_VALID 458 && ((wd->sc_params.atap_secsz & ATA_SECSZ_LPS) != 0)) { 459 wd->sc_sectoralign.dsa_alignment = 1 << 460 (wd->sc_params.atap_secsz & ATA_SECSZ_LPS_SZMSK); 461 if ((wd->sc_params.atap_logical_align & ATA_LA_VALID_MASK) == 462 ATA_LA_VALID) { 463 wd->sc_sectoralign.dsa_firstaligned = 464 wd->sc_params.atap_logical_align & ATA_LA_MASK; 465 } 466 } 467 wd->sc_capacity512 = (wd->sc_capacity * wd->sc_blksize) / DEV_BSIZE; 468 format_bytes(pbuf, sizeof(pbuf), wd->sc_capacity * wd->sc_blksize); 469 aprint_normal_dev(self, "%s, %d cyl, %d head, %d sec, " 470 "%d bytes/sect x %llu sectors", 471 pbuf, 472 (wd->sc_flags & WDF_LBA) ? (int)(wd->sc_capacity / 473 (wd->sc_params.atap_heads * wd->sc_params.atap_sectors)) : 474 wd->sc_params.atap_cylinders, 475 wd->sc_params.atap_heads, wd->sc_params.atap_sectors, 476 wd->sc_blksize, (unsigned long long)wd->sc_capacity); 477 if (wd->sc_sectoralign.dsa_alignment != 1) { 478 aprint_normal(" (%d bytes/physsect", 479 wd->sc_sectoralign.dsa_alignment * wd->sc_blksize); 480 if (wd->sc_sectoralign.dsa_firstaligned != 0) { 481 aprint_normal("; first aligned sector: %jd", 482 (intmax_t)wd->sc_sectoralign.dsa_firstaligned); 483 } 484 aprint_normal(")"); 485 } 486 aprint_normal("\n"); 487 488 ATADEBUG_PRINT(("%s: atap_dmatiming_mimi=%d, atap_dmatiming_recom=%d\n", 489 device_xname(self), wd->sc_params.atap_dmatiming_mimi, 490 wd->sc_params.atap_dmatiming_recom), DEBUG_PROBE); 491 492 if (wd->sc_blksize <= 0 || !powerof2(wd->sc_blksize) || 493 wd->sc_blksize < DEV_BSIZE || wd->sc_blksize > MAXPHYS) { 494 aprint_normal_dev(self, "WARNING: block size %u " 495 "might not actually work\n", wd->sc_blksize); 496 } 497 498 if (strcmp(wd->sc_params.atap_model, "ST506") == 0) 499 dtype = DKTYPE_ST506; 500 else 501 dtype = DKTYPE_ESDI; 502 503 out: 504 /* 505 * Initialize and attach the disk structure. 506 */ 507 dk_init(dksc, self, dtype); 508 disk_init(&dksc->sc_dkdev, dksc->sc_xname, &wddkdriver); 509 510 /* Attach dk and disk subsystems */ 511 dk_attach(dksc); 512 disk_attach(&dksc->sc_dkdev); 513 wd_set_geometry(wd); 514 515 bufq_alloc(&dksc->sc_bufq, BUFQ_DISK_DEFAULT_STRAT, BUFQ_SORT_RAWBLOCK); 516 517 /* reference to label structure, used by ata code */ 518 wd->drvp->lp = dksc->sc_dkdev.dk_label; 519 520 /* Discover wedges on this disk. */ 521 dkwedge_discover(&dksc->sc_dkdev); 522 523 if (!pmf_device_register1(self, wd_suspend, NULL, wd_shutdown)) 524 aprint_error_dev(self, "couldn't establish power handler\n"); 525 526 wd_sysctl_attach(wd); 527 } 528 529 static bool 530 wd_suspend(device_t dv, const pmf_qual_t *qual) 531 { 532 struct wd_softc *sc = device_private(dv); 533 534 /* the adapter needs to be enabled */ 535 if (sc->atabus->ata_addref(sc->drvp)) 536 return true; /* no need to complain */ 537 538 wd_flushcache(sc, AT_WAIT); 539 wd_standby(sc, AT_WAIT); 540 541 sc->atabus->ata_delref(sc->drvp); 542 return true; 543 } 544 545 static int 546 wddetach(device_t self, int flags) 547 { 548 struct wd_softc *wd = device_private(self); 549 struct dk_softc *dksc = &wd->sc_dksc; 550 int bmaj, cmaj, i, mn, rc; 551 552 if ((rc = disk_begindetach(&dksc->sc_dkdev, wd_lastclose, self, flags)) != 0) 553 return rc; 554 555 /* locate the major number */ 556 bmaj = bdevsw_lookup_major(&wd_bdevsw); 557 cmaj = cdevsw_lookup_major(&wd_cdevsw); 558 559 /* Nuke the vnodes for any open instances. */ 560 for (i = 0; i < MAXPARTITIONS; i++) { 561 mn = WDMINOR(device_unit(self), i); 562 vdevgone(bmaj, mn, mn, VBLK); 563 vdevgone(cmaj, mn, mn, VCHR); 564 } 565 566 dk_drain(dksc); 567 568 /* Kill off any pending commands. */ 569 mutex_enter(&wd->sc_lock); 570 wd->atabus->ata_killpending(wd->drvp); 571 572 callout_halt(&wd->sc_retry_callout, &wd->sc_lock); 573 callout_destroy(&wd->sc_retry_callout); 574 callout_halt(&wd->sc_requeue_callout, &wd->sc_lock); 575 callout_destroy(&wd->sc_requeue_callout); 576 callout_halt(&wd->sc_restart_diskqueue, &wd->sc_lock); 577 callout_destroy(&wd->sc_restart_diskqueue); 578 579 mutex_exit(&wd->sc_lock); 580 581 bufq_free(dksc->sc_bufq); 582 583 /* Delete all of our wedges. */ 584 dkwedge_delall(&dksc->sc_dkdev); 585 586 if (flags & DETACH_POWEROFF) 587 wd_standby(wd, AT_POLL); 588 589 /* Detach from the disk list. */ 590 disk_detach(&dksc->sc_dkdev); 591 disk_destroy(&dksc->sc_dkdev); 592 593 dk_detach(dksc); 594 595 #ifdef WD_SOFTBADSECT 596 /* Clean out the bad sector list */ 597 while (!SLIST_EMPTY(&wd->sc_bslist)) { 598 struct disk_badsectors *dbs = SLIST_FIRST(&wd->sc_bslist); 599 SLIST_REMOVE_HEAD(&wd->sc_bslist, dbs_next); 600 kmem_free(dbs, sizeof(*dbs)); 601 } 602 wd->sc_bscount = 0; 603 #endif 604 if (wd->sc_typename != NULL) { 605 kmem_free(wd->sc_typename, strlen(wd->sc_typename) + 1); 606 wd->sc_typename = NULL; 607 } 608 609 pmf_device_deregister(self); 610 611 wd_sysctl_detach(wd); 612 613 #ifdef WD_SOFTBADSECT 614 KASSERT(SLIST_EMPTY(&wd->sc_bslist)); 615 cv_destroy(&wd->sc_bslist_cv); 616 #endif 617 618 mutex_destroy(&wd->sc_lock); 619 620 wd->drvp->drive_type = ATA_DRIVET_NONE; /* no drive any more here */ 621 wd->drvp->drive_flags = 0; 622 623 return (0); 624 } 625 626 /* 627 * Read/write routine for a buffer. Validates the arguments and schedules the 628 * transfer. Does not wait for the transfer to complete. 629 */ 630 static void 631 wdstrategy(struct buf *bp) 632 { 633 struct wd_softc *wd = 634 device_lookup_private(&wd_cd, WDUNIT(bp->b_dev)); 635 struct dk_softc *dksc = &wd->sc_dksc; 636 637 ATADEBUG_PRINT(("wdstrategy (%s)\n", dksc->sc_xname), 638 DEBUG_XFERS); 639 640 /* If device invalidated (e.g. media change, door open, 641 * device detachment), then error. 642 */ 643 if ((wd->sc_flags & WDF_LOADED) == 0 || 644 !device_is_enabled(dksc->sc_dev)) 645 goto err; 646 647 #ifdef WD_SOFTBADSECT 648 /* 649 * If the transfer about to be attempted contains only a block that 650 * is known to be bad then return an error for the transfer without 651 * even attempting to start a transfer up under the premis that we 652 * will just end up doing more retries for a transfer that will end 653 * up failing again. 654 */ 655 if (__predict_false(!SLIST_EMPTY(&wd->sc_bslist))) { 656 struct disklabel *lp = dksc->sc_dkdev.dk_label; 657 struct disk_badsectors *dbs; 658 daddr_t blkno, maxblk; 659 660 /* convert the block number to absolute */ 661 if (lp->d_secsize >= DEV_BSIZE) 662 blkno = bp->b_blkno / (lp->d_secsize / DEV_BSIZE); 663 else 664 blkno = bp->b_blkno * (DEV_BSIZE / lp->d_secsize); 665 if (WDPART(bp->b_dev) != RAW_PART) 666 blkno += lp->d_partitions[WDPART(bp->b_dev)].p_offset; 667 maxblk = blkno + (bp->b_bcount / wd->sc_blksize) - 1; 668 669 mutex_enter(&wd->sc_lock); 670 SLIST_FOREACH(dbs, &wd->sc_bslist, dbs_next) 671 if ((dbs->dbs_min <= bp->b_rawblkno && 672 bp->b_rawblkno <= dbs->dbs_max) || 673 (dbs->dbs_min <= maxblk && maxblk <= dbs->dbs_max)){ 674 mutex_exit(&wd->sc_lock); 675 goto err; 676 } 677 mutex_exit(&wd->sc_lock); 678 } 679 #endif 680 681 dk_strategy(dksc, bp); 682 return; 683 684 err: 685 bp->b_error = EIO; 686 bp->b_resid = bp->b_bcount; 687 biodone(bp); 688 } 689 690 static void 691 wdstart1(struct wd_softc *wd, struct buf *bp, struct ata_xfer *xfer) 692 { 693 struct dk_softc *dksc = &wd->sc_dksc; 694 const uint32_t secsize = dksc->sc_dkdev.dk_geom.dg_secsize; 695 696 KASSERT(bp == xfer->c_bio.bp || xfer->c_bio.bp == NULL); 697 KASSERT((xfer->c_flags & (C_WAITACT|C_FREE)) == 0); 698 KASSERT(mutex_owned(&wd->sc_lock)); 699 700 /* Reset state, so that retries don't use stale info */ 701 if (__predict_false(xfer->c_retries > 0)) { 702 xfer->c_flags = 0; 703 memset(&xfer->c_bio, 0, sizeof(xfer->c_bio)); 704 } 705 706 xfer->c_bio.blkno = bp->b_rawblkno; 707 xfer->c_bio.bcount = bp->b_bcount; 708 xfer->c_bio.databuf = bp->b_data; 709 xfer->c_bio.blkdone = 0; 710 xfer->c_bio.bp = bp; 711 712 /* Adjust blkno and bcount if xfer has been already partially done */ 713 if (__predict_false(xfer->c_skip > 0)) { 714 KASSERT(xfer->c_skip < xfer->c_bio.bcount); 715 KASSERT((xfer->c_skip % secsize) == 0); 716 xfer->c_bio.bcount -= xfer->c_skip; 717 xfer->c_bio.blkno += xfer->c_skip / secsize; 718 } 719 720 #ifdef WD_CHAOS_MONKEY 721 /* 722 * Override blkno to be over device capacity to trigger error, 723 * but only if it's read, to avoid trashing disk contents should 724 * the command be clipped, or otherwise misinterpreted, by the 725 * driver or controller. 726 */ 727 if (BUF_ISREAD(bp) && xfer->c_retries == 0 && wd->drv_chaos_freq > 0 && 728 (++wd->drv_chaos_cnt % wd->drv_chaos_freq) == 0) { 729 device_printf(dksc->sc_dev, "%s: chaos xfer %"PRIxPTR"\n", 730 __func__, (intptr_t)xfer & PAGE_MASK); 731 xfer->c_bio.blkno = 7777777 + wd->sc_capacity; 732 xfer->c_flags |= C_CHAOS; 733 } 734 #endif 735 736 /* 737 * If we're retrying, retry in single-sector mode. This will give us 738 * the sector number of the problem, and will eventually allow the 739 * transfer to succeed. If FUA is requested, we can't actually 740 * do this, as ATA_SINGLE is usually executed as PIO transfer by drivers 741 * which support it, and that isn't compatible with NCQ/FUA. 742 */ 743 if (xfer->c_retries >= WDIORETRIES_SINGLE && 744 (bp->b_flags & B_MEDIA_FUA) == 0) 745 xfer->c_bio.flags = ATA_SINGLE; 746 else 747 xfer->c_bio.flags = 0; 748 749 /* 750 * request LBA48 transfers when supported by the controller 751 * and needed by transfer offset or size. 752 */ 753 if (wd->sc_flags & WDF_LBA48 && 754 (((xfer->c_bio.blkno + xfer->c_bio.bcount / secsize) > 755 wd->sc_capacity28) || 756 ((xfer->c_bio.bcount / secsize) > 128))) 757 xfer->c_bio.flags |= ATA_LBA48; 758 759 /* 760 * If NCQ was negotiated, always use it for the first several attempts. 761 * Since device cancels all outstanding requests on error, downgrade 762 * to non-NCQ on retry, so that the retried transfer would not cause 763 * cascade failure for the other transfers if it fails again. 764 * If FUA was requested, we can't downgrade, as that would violate 765 * the semantics - FUA would not be honored. In that case, continue 766 * retrying with NCQ. 767 */ 768 if (WD_USE_NCQ(wd) && (xfer->c_retries < WDIORETRIES_SINGLE || 769 (bp->b_flags & B_MEDIA_FUA) != 0)) { 770 xfer->c_bio.flags |= ATA_LBA48; 771 xfer->c_flags |= C_NCQ; 772 773 if (WD_USE_NCQ_PRIO(wd) && 774 BIO_GETPRIO(bp) == BPRIO_TIMECRITICAL) 775 xfer->c_bio.flags |= ATA_PRIO_HIGH; 776 } 777 778 if (wd->sc_flags & WDF_LBA) 779 xfer->c_bio.flags |= ATA_LBA; 780 if (bp->b_flags & B_READ) { 781 xfer->c_bio.flags |= ATA_READ; 782 } else { 783 /* it's a write */ 784 wd->sc_flags |= WDF_DIRTY; 785 } 786 if (bp->b_flags & B_MEDIA_FUA) { 787 /* If not using NCQ, the command WRITE DMA FUA EXT is LBA48 */ 788 KASSERT((wd->sc_flags & WDF_LBA48) != 0); 789 if ((xfer->c_flags & C_NCQ) == 0) 790 xfer->c_bio.flags |= ATA_LBA48; 791 792 xfer->c_bio.flags |= ATA_FUA; 793 } 794 795 if (xfer->c_retries == 0) 796 wd->inflight++; 797 mutex_exit(&wd->sc_lock); 798 799 /* Queue the xfer */ 800 wd->atabus->ata_bio(wd->drvp, xfer); 801 802 mutex_enter(&wd->sc_lock); 803 } 804 805 static int 806 wd_diskstart(device_t dev, struct buf *bp) 807 { 808 struct wd_softc *wd = device_private(dev); 809 #ifdef ATADEBUG 810 struct dk_softc *dksc = &wd->sc_dksc; 811 #endif 812 struct ata_xfer *xfer; 813 struct ata_channel *chp; 814 unsigned openings; 815 int ticks; 816 817 mutex_enter(&wd->sc_lock); 818 819 chp = wd->drvp->chnl_softc; 820 821 ata_channel_lock(chp); 822 openings = ata_queue_openings(chp); 823 ata_channel_unlock(chp); 824 825 openings = uimin(openings, wd->drvp->drv_openings); 826 827 if (wd->inflight >= openings) { 828 /* 829 * pretend we run out of memory when the queue is full, 830 * so that the operation is retried after a minimal 831 * delay. 832 */ 833 xfer = NULL; 834 ticks = 1; 835 } else { 836 /* 837 * If there is no available memory, retry later. This 838 * happens very rarely and only under memory pressure, 839 * so wait relatively long before retry. 840 */ 841 xfer = ata_get_xfer(chp, false); 842 ticks = hz/2; 843 } 844 845 if (xfer == NULL) { 846 ATADEBUG_PRINT(("wd_diskstart %s no xfer\n", 847 dksc->sc_xname), DEBUG_XFERS); 848 849 /* 850 * The disk queue is pushed automatically when an I/O 851 * operation finishes or another one is queued. We 852 * need this extra timeout because an ATA channel 853 * might be shared by more than one disk queue and 854 * all queues need to be restarted when another slot 855 * becomes available. 856 */ 857 if (!callout_pending(&wd->sc_restart_diskqueue)) { 858 callout_reset(&wd->sc_restart_diskqueue, ticks, 859 wdrestart, dev); 860 } 861 862 mutex_exit(&wd->sc_lock); 863 return EAGAIN; 864 } 865 866 wdstart1(wd, bp, xfer); 867 868 mutex_exit(&wd->sc_lock); 869 870 return 0; 871 } 872 873 /* 874 * Queue a drive for I/O. 875 */ 876 static void 877 wdrestart(void *x) 878 { 879 device_t self = x; 880 struct wd_softc *wd = device_private(self); 881 struct dk_softc *dksc = &wd->sc_dksc; 882 883 ATADEBUG_PRINT(("wdstart %s\n", dksc->sc_xname), 884 DEBUG_XFERS); 885 886 if (!device_is_active(dksc->sc_dev)) 887 return; 888 889 dk_start(dksc, NULL); 890 } 891 892 static void 893 wddone(device_t self, struct ata_xfer *xfer) 894 { 895 struct wd_softc *wd = device_private(self); 896 struct dk_softc *dksc = &wd->sc_dksc; 897 const char *errmsg; 898 int do_perror = 0; 899 struct buf *bp; 900 901 ATADEBUG_PRINT(("wddone %s\n", dksc->sc_xname), 902 DEBUG_XFERS); 903 904 if (__predict_false(wddoingadump)) { 905 /* just drop it to the floor */ 906 ata_free_xfer(wd->drvp->chnl_softc, xfer); 907 return; 908 } 909 910 bp = xfer->c_bio.bp; 911 KASSERT(bp != NULL); 912 913 bp->b_resid = xfer->c_bio.bcount; 914 switch (xfer->c_bio.error) { 915 case ERR_DMA: 916 errmsg = "DMA error"; 917 goto retry; 918 case ERR_DF: 919 errmsg = "device fault"; 920 goto retry; 921 case TIMEOUT: 922 errmsg = "device timeout"; 923 goto retry; 924 case REQUEUE: 925 errmsg = "requeue"; 926 goto retry2; 927 case ERR_RESET: 928 errmsg = "channel reset"; 929 goto retry2; 930 case ERROR: 931 /* Don't care about media change bits */ 932 if (xfer->c_bio.r_error != 0 && 933 (xfer->c_bio.r_error & ~(WDCE_MC | WDCE_MCR)) == 0) 934 goto noerror; 935 errmsg = "error"; 936 do_perror = 1; 937 retry: /* Just reset and retry. Can we do more ? */ 938 if ((xfer->c_flags & C_RECOVERED) == 0) { 939 int wflags = (xfer->c_flags & C_POLL) ? AT_POLL : 0; 940 ata_channel_lock(wd->drvp->chnl_softc); 941 ata_thread_run(wd->drvp->chnl_softc, wflags, 942 ATACH_TH_DRIVE_RESET, wd->drvp->drive); 943 ata_channel_unlock(wd->drvp->chnl_softc); 944 } 945 retry2: 946 mutex_enter(&wd->sc_lock); 947 948 diskerr(bp, "wd", errmsg, LOG_PRINTF, 949 xfer->c_bio.blkdone, dksc->sc_dkdev.dk_label); 950 if (xfer->c_retries < WDIORETRIES) 951 printf(", xfer %"PRIxPTR", retry %d", 952 (intptr_t)xfer & PAGE_MASK, 953 xfer->c_retries); 954 printf("\n"); 955 if (do_perror) 956 wdperror(wd, xfer); 957 958 if (xfer->c_retries < WDIORETRIES) { 959 xfer->c_retries++; 960 961 /* Rerun ASAP if just requeued */ 962 if (xfer->c_bio.error == REQUEUE) { 963 SLIST_INSERT_HEAD(&wd->sc_requeue_list, xfer, 964 c_retrychain); 965 callout_reset(&wd->sc_requeue_callout, 966 1, wdbiorequeue, wd); 967 } else { 968 SLIST_INSERT_HEAD(&wd->sc_retry_list, xfer, 969 c_retrychain); 970 callout_reset(&wd->sc_retry_callout, 971 RECOVERYTIME, wdbioretry, wd); 972 } 973 974 mutex_exit(&wd->sc_lock); 975 return; 976 } 977 978 mutex_exit(&wd->sc_lock); 979 980 #ifdef WD_SOFTBADSECT 981 /* 982 * Not all errors indicate a failed block but those that do, 983 * put the block on the bad-block list for the device. Only 984 * do this for reads because the drive should do it for writes, 985 * itself, according to Manuel. 986 */ 987 if ((bp->b_flags & B_READ) && 988 ((wd->drvp->ata_vers >= 4 && xfer->c_bio.r_error & 64) || 989 (wd->drvp->ata_vers < 4 && xfer->c_bio.r_error & 192))) { 990 struct disk_badsectors *dbs; 991 992 dbs = kmem_zalloc(sizeof *dbs, KM_NOSLEEP); 993 if (dbs == NULL) { 994 device_printf(dksc->sc_dev, 995 "failed to add bad block to list\n"); 996 goto out; 997 } 998 999 dbs->dbs_min = bp->b_rawblkno; 1000 dbs->dbs_max = dbs->dbs_min + 1001 (bp->b_bcount /wd->sc_blksize) - 1; 1002 microtime(&dbs->dbs_failedat); 1003 1004 mutex_enter(&wd->sc_lock); 1005 SLIST_INSERT_HEAD(&wd->sc_bslist, dbs, dbs_next); 1006 wd->sc_bscount++; 1007 mutex_exit(&wd->sc_lock); 1008 } 1009 out: 1010 #endif 1011 bp->b_error = EIO; 1012 break; 1013 case NOERROR: 1014 #ifdef WD_CHAOS_MONKEY 1015 /* 1016 * For example Parallels AHCI emulation doesn't actually 1017 * return error for the invalid I/O, so just re-run 1018 * the request and do not panic. 1019 */ 1020 if (__predict_false(xfer->c_flags & C_CHAOS)) { 1021 xfer->c_bio.error = REQUEUE; 1022 errmsg = "chaos noerror"; 1023 goto retry2; 1024 } 1025 #endif 1026 1027 noerror: if ((xfer->c_bio.flags & ATA_CORR) || xfer->c_retries > 0) 1028 device_printf(dksc->sc_dev, 1029 "soft error (corrected) xfer %"PRIxPTR"\n", 1030 (intptr_t)xfer & PAGE_MASK); 1031 break; 1032 case ERR_NODEV: 1033 bp->b_error = EIO; 1034 break; 1035 } 1036 if (__predict_false(bp->b_error != 0) && bp->b_resid == 0) { 1037 /* 1038 * the disk or controller sometimes report a complete 1039 * xfer, when there has been an error. This is wrong, 1040 * assume nothing got transferred in this case 1041 */ 1042 bp->b_resid = bp->b_bcount; 1043 } 1044 1045 ata_free_xfer(wd->drvp->chnl_softc, xfer); 1046 1047 mutex_enter(&wd->sc_lock); 1048 wd->inflight--; 1049 mutex_exit(&wd->sc_lock); 1050 dk_done(dksc, bp); 1051 dk_start(dksc, NULL); 1052 } 1053 1054 static void 1055 wdbioretry(void *v) 1056 { 1057 struct wd_softc *wd = v; 1058 struct ata_xfer *xfer; 1059 1060 ATADEBUG_PRINT(("%s %s\n", __func__, wd->sc_dksc.sc_xname), 1061 DEBUG_XFERS); 1062 1063 mutex_enter(&wd->sc_lock); 1064 while ((xfer = SLIST_FIRST(&wd->sc_retry_list))) { 1065 SLIST_REMOVE_HEAD(&wd->sc_retry_list, c_retrychain); 1066 wdstart1(wd, xfer->c_bio.bp, xfer); 1067 } 1068 mutex_exit(&wd->sc_lock); 1069 } 1070 1071 static void 1072 wdbiorequeue(void *v) 1073 { 1074 struct wd_softc *wd = v; 1075 struct ata_xfer *xfer; 1076 1077 ATADEBUG_PRINT(("%s %s\n", __func__, wd->sc_dksc.sc_xname), 1078 DEBUG_XFERS); 1079 1080 mutex_enter(&wd->sc_lock); 1081 while ((xfer = SLIST_FIRST(&wd->sc_requeue_list))) { 1082 SLIST_REMOVE_HEAD(&wd->sc_requeue_list, c_retrychain); 1083 wdstart1(wd, xfer->c_bio.bp, xfer); 1084 } 1085 mutex_exit(&wd->sc_lock); 1086 } 1087 1088 static void 1089 wdminphys(struct buf *bp) 1090 { 1091 const struct wd_softc * const wd = 1092 device_lookup_private(&wd_cd, WDUNIT(bp->b_dev)); 1093 int maxsectors; 1094 1095 /* 1096 * The limit is actually 65536 for LBA48 and 256 for non-LBA48, 1097 * but that requires to set the count for the ATA command 1098 * to 0, which is somewhat error prone, so better stay safe. 1099 */ 1100 if (wd->sc_flags & WDF_LBA48) 1101 maxsectors = 65535; 1102 else 1103 maxsectors = 128; 1104 1105 if (bp->b_bcount > (wd->sc_blksize * maxsectors)) 1106 bp->b_bcount = (wd->sc_blksize * maxsectors); 1107 1108 minphys(bp); 1109 } 1110 1111 static void 1112 wd_iosize(device_t dev, int *count) 1113 { 1114 struct buf B; 1115 int bmaj; 1116 1117 bmaj = bdevsw_lookup_major(&wd_bdevsw); 1118 B.b_dev = MAKEWDDEV(bmaj,device_unit(dev),RAW_PART); 1119 B.b_bcount = *count; 1120 1121 wdminphys(&B); 1122 1123 *count = B.b_bcount; 1124 } 1125 1126 static int 1127 wdread(dev_t dev, struct uio *uio, int flags) 1128 { 1129 1130 ATADEBUG_PRINT(("wdread\n"), DEBUG_XFERS); 1131 return (physio(wdstrategy, NULL, dev, B_READ, wdminphys, uio)); 1132 } 1133 1134 static int 1135 wdwrite(dev_t dev, struct uio *uio, int flags) 1136 { 1137 1138 ATADEBUG_PRINT(("wdwrite\n"), DEBUG_XFERS); 1139 return (physio(wdstrategy, NULL, dev, B_WRITE, wdminphys, uio)); 1140 } 1141 1142 static int 1143 wdopen(dev_t dev, int flag, int fmt, struct lwp *l) 1144 { 1145 struct wd_softc *wd; 1146 struct dk_softc *dksc; 1147 int unit, part, error; 1148 1149 ATADEBUG_PRINT(("wdopen\n"), DEBUG_FUNCS); 1150 unit = WDUNIT(dev); 1151 wd = device_lookup_private(&wd_cd, unit); 1152 if (wd == NULL) 1153 return (ENXIO); 1154 dksc = &wd->sc_dksc; 1155 1156 if (! device_is_active(dksc->sc_dev)) 1157 return (ENODEV); 1158 1159 part = WDPART(dev); 1160 1161 if (wd->sc_capacity == 0) 1162 return (ENODEV); 1163 1164 /* 1165 * If any partition is open, but the disk has been invalidated, 1166 * disallow further opens. 1167 */ 1168 if ((wd->sc_flags & (WDF_OPEN | WDF_LOADED)) == WDF_OPEN) { 1169 if (part != RAW_PART || fmt != S_IFCHR) 1170 return EIO; 1171 } 1172 1173 error = dk_open(dksc, dev, flag, fmt, l); 1174 1175 return error; 1176 } 1177 1178 /* 1179 * Serialized by caller 1180 */ 1181 static int 1182 wd_firstopen(device_t self, dev_t dev, int flag, int fmt) 1183 { 1184 struct wd_softc *wd = device_private(self); 1185 struct dk_softc *dksc = &wd->sc_dksc; 1186 int error; 1187 1188 error = wd->atabus->ata_addref(wd->drvp); 1189 if (error) 1190 return error; 1191 1192 if ((wd->sc_flags & WDF_LOADED) == 0) { 1193 int param_error; 1194 1195 /* Load the physical device parameters. */ 1196 param_error = wd_get_params(wd, &wd->sc_params); 1197 if (param_error != 0) { 1198 aprint_error_dev(dksc->sc_dev, "IDENTIFY failed\n"); 1199 error = EIO; 1200 goto bad; 1201 } 1202 wd_set_geometry(wd); 1203 wd->sc_flags |= WDF_LOADED; 1204 } 1205 1206 wd->sc_flags |= WDF_OPEN; 1207 return 0; 1208 1209 bad: 1210 wd->atabus->ata_delref(wd->drvp); 1211 return error; 1212 } 1213 1214 /* 1215 * Caller must hold wd->sc_dk.dk_openlock. 1216 */ 1217 static int 1218 wd_lastclose(device_t self) 1219 { 1220 struct wd_softc *wd = device_private(self); 1221 1222 KASSERTMSG(bufq_peek(wd->sc_dksc.sc_bufq) == NULL, "bufq not empty"); 1223 1224 if (wd->sc_flags & WDF_DIRTY) 1225 wd_flushcache(wd, AT_WAIT); 1226 1227 wd->atabus->ata_delref(wd->drvp); 1228 wd->sc_flags &= ~WDF_OPEN; 1229 1230 return 0; 1231 } 1232 1233 static int 1234 wdclose(dev_t dev, int flag, int fmt, struct lwp *l) 1235 { 1236 struct wd_softc *wd; 1237 struct dk_softc *dksc; 1238 int unit; 1239 1240 unit = WDUNIT(dev); 1241 wd = device_lookup_private(&wd_cd, unit); 1242 dksc = &wd->sc_dksc; 1243 1244 return dk_close(dksc, dev, flag, fmt, l); 1245 } 1246 1247 void 1248 wdperror(const struct wd_softc *wd, struct ata_xfer *xfer) 1249 { 1250 static const char *const errstr0_3[] = {"address mark not found", 1251 "track 0 not found", "aborted command", "media change requested", 1252 "id not found", "media changed", "uncorrectable data error", 1253 "bad block detected"}; 1254 static const char *const errstr4_5[] = { 1255 "obsolete (address mark not found)", 1256 "no media/write protected", "aborted command", 1257 "media change requested", "id not found", "media changed", 1258 "uncorrectable data error", "interface CRC error"}; 1259 const char *const *errstr; 1260 int i; 1261 const char *sep = ""; 1262 1263 const struct dk_softc *dksc = &wd->sc_dksc; 1264 const char *devname = dksc->sc_xname; 1265 struct ata_drive_datas *drvp = wd->drvp; 1266 int errno = xfer->c_bio.r_error; 1267 1268 if (drvp->ata_vers >= 4) 1269 errstr = errstr4_5; 1270 else 1271 errstr = errstr0_3; 1272 1273 printf("%s: (", devname); 1274 1275 if (errno == 0) 1276 printf("error not notified"); 1277 1278 for (i = 0; i < 8; i++) { 1279 if (errno & (1 << i)) { 1280 printf("%s%s", sep, errstr[i]); 1281 sep = ", "; 1282 } 1283 } 1284 printf(")\n"); 1285 } 1286 1287 int 1288 wdioctl(dev_t dev, u_long cmd, void *addr, int flag, struct lwp *l) 1289 { 1290 struct wd_softc *wd = 1291 device_lookup_private(&wd_cd, WDUNIT(dev)); 1292 struct dk_softc *dksc = &wd->sc_dksc; 1293 1294 ATADEBUG_PRINT(("wdioctl\n"), DEBUG_FUNCS); 1295 1296 if ((wd->sc_flags & WDF_LOADED) == 0) 1297 return EIO; 1298 1299 switch (cmd) { 1300 #ifdef HAS_BAD144_HANDLING 1301 case DIOCSBAD: 1302 if ((flag & FWRITE) == 0) 1303 return EBADF; 1304 dksc->sc_dkdev.dk_cpulabel->bad = *(struct dkbad *)addr; 1305 dksc->sc_dkdev.dk_label->d_flags |= D_BADSECT; 1306 bad144intern(wd); 1307 return 0; 1308 #endif 1309 #ifdef WD_SOFTBADSECT 1310 case DIOCBSLIST: { 1311 uint32_t count, missing, skip; 1312 struct disk_badsecinfo dbsi; 1313 struct disk_badsectors *dbs, dbsbuf; 1314 size_t available; 1315 uint8_t *laddr; 1316 int error; 1317 1318 dbsi = *(struct disk_badsecinfo *)addr; 1319 missing = wd->sc_bscount; 1320 count = 0; 1321 available = dbsi.dbsi_bufsize; 1322 skip = dbsi.dbsi_skip; 1323 laddr = (uint8_t *)dbsi.dbsi_buffer; 1324 1325 /* 1326 * We start this loop with the expectation that all of the 1327 * entries will be missed and decrement this counter each 1328 * time we either skip over one (already copied out) or 1329 * we actually copy it back to user space. The structs 1330 * holding the bad sector information are copied directly 1331 * back to user space whilst the summary is returned via 1332 * the struct passed in via the ioctl. 1333 */ 1334 error = 0; 1335 mutex_enter(&wd->sc_lock); 1336 wd->sc_bslist_inuse++; 1337 SLIST_FOREACH(dbs, &wd->sc_bslist, dbs_next) { 1338 if (skip > 0) { 1339 missing--; 1340 skip--; 1341 continue; 1342 } 1343 if (available < sizeof(*dbs)) 1344 break; 1345 available -= sizeof(*dbs); 1346 memset(&dbsbuf, 0, sizeof(dbsbuf)); 1347 dbsbuf.dbs_min = dbs->dbs_min; 1348 dbsbuf.dbs_max = dbs->dbs_max; 1349 dbsbuf.dbs_failedat = dbs->dbs_failedat; 1350 mutex_exit(&wd->sc_lock); 1351 error = copyout(&dbsbuf, laddr, sizeof(dbsbuf)); 1352 mutex_enter(&wd->sc_lock); 1353 if (error) 1354 break; 1355 laddr += sizeof(*dbs); 1356 missing--; 1357 count++; 1358 } 1359 if (--wd->sc_bslist_inuse == 0) 1360 cv_broadcast(&wd->sc_bslist_cv); 1361 mutex_exit(&wd->sc_lock); 1362 dbsi.dbsi_left = missing; 1363 dbsi.dbsi_copied = count; 1364 *(struct disk_badsecinfo *)addr = dbsi; 1365 1366 /* 1367 * If we copied anything out, ignore error and return 1368 * success -- can't back it out. 1369 */ 1370 return count ? 0 : error; 1371 } 1372 1373 case DIOCBSFLUSH: { 1374 int error; 1375 1376 /* Clean out the bad sector list */ 1377 mutex_enter(&wd->sc_lock); 1378 while (wd->sc_bslist_inuse) { 1379 error = cv_wait_sig(&wd->sc_bslist_cv, &wd->sc_lock); 1380 if (error) { 1381 mutex_exit(&wd->sc_lock); 1382 return error; 1383 } 1384 } 1385 while (!SLIST_EMPTY(&wd->sc_bslist)) { 1386 struct disk_badsectors *dbs = 1387 SLIST_FIRST(&wd->sc_bslist); 1388 SLIST_REMOVE_HEAD(&wd->sc_bslist, dbs_next); 1389 mutex_exit(&wd->sc_lock); 1390 kmem_free(dbs, sizeof(*dbs)); 1391 mutex_enter(&wd->sc_lock); 1392 } 1393 mutex_exit(&wd->sc_lock); 1394 wd->sc_bscount = 0; 1395 return 0; 1396 } 1397 #endif 1398 1399 #ifdef notyet 1400 case DIOCWFORMAT: 1401 if ((flag & FWRITE) == 0) 1402 return EBADF; 1403 { 1404 register struct format_op *fop; 1405 struct iovec aiov; 1406 struct uio auio; 1407 int error1; 1408 1409 fop = (struct format_op *)addr; 1410 aiov.iov_base = fop->df_buf; 1411 aiov.iov_len = fop->df_count; 1412 auio.uio_iov = &aiov; 1413 auio.uio_iovcnt = 1; 1414 auio.uio_resid = fop->df_count; 1415 auio.uio_offset = 1416 fop->df_startblk * wd->sc_dk.dk_label->d_secsize; 1417 auio.uio_vmspace = l->l_proc->p_vmspace; 1418 error1 = physio(wdformat, NULL, dev, B_WRITE, wdminphys, 1419 &auio); 1420 fop->df_count -= auio.uio_resid; 1421 fop->df_reg[0] = wdc->sc_status; 1422 fop->df_reg[1] = wdc->sc_error; 1423 return error1; 1424 } 1425 #endif 1426 case DIOCGCACHE: 1427 return wd_getcache(wd, (int *)addr); 1428 1429 case DIOCSCACHE: 1430 return wd_setcache(wd, *(int *)addr); 1431 1432 case DIOCCACHESYNC: 1433 return wd_flushcache(wd, AT_WAIT); 1434 1435 case ATAIOCCOMMAND: 1436 /* 1437 * Make sure this command is (relatively) safe first 1438 */ 1439 if ((((atareq_t *) addr)->flags & ATACMD_READ) == 0 && 1440 (flag & FWRITE) == 0) 1441 return (EBADF); 1442 { 1443 struct wd_ioctl *wi; 1444 atareq_t *atareq = (atareq_t *) addr; 1445 int error1; 1446 1447 wi = wi_get(wd); 1448 wi->wi_atareq = *atareq; 1449 1450 if (atareq->datalen && atareq->flags & 1451 (ATACMD_READ | ATACMD_WRITE)) { 1452 void *tbuf; 1453 if (atareq->datalen < DEV_BSIZE 1454 && atareq->command == WDCC_IDENTIFY) { 1455 tbuf = kmem_zalloc(DEV_BSIZE, KM_SLEEP); 1456 wi->wi_iov.iov_base = tbuf; 1457 wi->wi_iov.iov_len = DEV_BSIZE; 1458 UIO_SETUP_SYSSPACE(&wi->wi_uio); 1459 } else { 1460 tbuf = NULL; 1461 wi->wi_iov.iov_base = atareq->databuf; 1462 wi->wi_iov.iov_len = atareq->datalen; 1463 wi->wi_uio.uio_vmspace = l->l_proc->p_vmspace; 1464 } 1465 wi->wi_uio.uio_iov = &wi->wi_iov; 1466 wi->wi_uio.uio_iovcnt = 1; 1467 wi->wi_uio.uio_resid = atareq->datalen; 1468 wi->wi_uio.uio_offset = 0; 1469 wi->wi_uio.uio_rw = 1470 (atareq->flags & ATACMD_READ) ? B_READ : B_WRITE; 1471 error1 = physio(wdioctlstrategy, &wi->wi_bp, dev, 1472 (atareq->flags & ATACMD_READ) ? B_READ : B_WRITE, 1473 wdminphys, &wi->wi_uio); 1474 if (tbuf != NULL && error1 == 0) { 1475 error1 = copyout(tbuf, atareq->databuf, 1476 atareq->datalen); 1477 kmem_free(tbuf, DEV_BSIZE); 1478 } 1479 } else { 1480 /* No need to call physio if we don't have any 1481 user data */ 1482 wi->wi_bp.b_flags = 0; 1483 wi->wi_bp.b_data = 0; 1484 wi->wi_bp.b_bcount = 0; 1485 wi->wi_bp.b_dev = dev; 1486 wi->wi_bp.b_proc = l->l_proc; 1487 wdioctlstrategy(&wi->wi_bp); 1488 error1 = wi->wi_bp.b_error; 1489 } 1490 *atareq = wi->wi_atareq; 1491 wi_free(wi); 1492 return(error1); 1493 } 1494 1495 case DIOCGSECTORALIGN: { 1496 struct disk_sectoralign *dsa = addr; 1497 int part = WDPART(dev); 1498 1499 *dsa = wd->sc_sectoralign; 1500 if (part != RAW_PART) { 1501 struct disklabel *lp = dksc->sc_dkdev.dk_label; 1502 daddr_t offset = lp->d_partitions[part].p_offset; 1503 uint32_t r = offset % dsa->dsa_alignment; 1504 1505 if (r < dsa->dsa_firstaligned) 1506 dsa->dsa_firstaligned = dsa->dsa_firstaligned 1507 - r; 1508 else 1509 dsa->dsa_firstaligned = (dsa->dsa_firstaligned 1510 + dsa->dsa_alignment) - r; 1511 } 1512 1513 return 0; 1514 } 1515 1516 default: 1517 return dk_ioctl(dksc, dev, cmd, addr, flag, l); 1518 } 1519 1520 #ifdef DIAGNOSTIC 1521 panic("wdioctl: impossible"); 1522 #endif 1523 } 1524 1525 static int 1526 wd_discard(device_t dev, off_t pos, off_t len) 1527 { 1528 struct wd_softc *wd = device_private(dev); 1529 daddr_t bno; 1530 long size, done; 1531 long maxatonce, amount; 1532 int result; 1533 1534 if (!(wd->sc_params.atap_ata_major & WDC_VER_ATA7) 1535 || !(wd->sc_params.support_dsm & ATA_SUPPORT_DSM_TRIM)) { 1536 /* not supported; ignore request */ 1537 ATADEBUG_PRINT(("wddiscard (unsupported)\n"), DEBUG_FUNCS); 1538 return 0; 1539 } 1540 maxatonce = 0xffff; /*wd->sc_params.max_dsm_blocks*/ 1541 1542 ATADEBUG_PRINT(("wddiscard\n"), DEBUG_FUNCS); 1543 1544 if ((wd->sc_flags & WDF_LOADED) == 0) 1545 return EIO; 1546 1547 /* round the start up and the end down */ 1548 bno = (pos + wd->sc_blksize - 1) / wd->sc_blksize; 1549 size = ((pos + len) / wd->sc_blksize) - bno; 1550 1551 done = 0; 1552 while (done < size) { 1553 amount = size - done; 1554 if (amount > maxatonce) { 1555 amount = maxatonce; 1556 } 1557 result = wd_trim(wd, bno + done, amount); 1558 if (result) { 1559 return result; 1560 } 1561 done += amount; 1562 } 1563 return 0; 1564 } 1565 1566 static int 1567 wddiscard(dev_t dev, off_t pos, off_t len) 1568 { 1569 struct wd_softc *wd; 1570 struct dk_softc *dksc; 1571 int unit; 1572 1573 unit = WDUNIT(dev); 1574 wd = device_lookup_private(&wd_cd, unit); 1575 dksc = &wd->sc_dksc; 1576 1577 return dk_discard(dksc, dev, pos, len); 1578 } 1579 1580 #ifdef B_FORMAT 1581 int 1582 wdformat(struct buf *bp) 1583 { 1584 1585 bp->b_flags |= B_FORMAT; 1586 return wdstrategy(bp); 1587 } 1588 #endif 1589 1590 int 1591 wdsize(dev_t dev) 1592 { 1593 struct wd_softc *wd; 1594 struct dk_softc *dksc; 1595 int unit; 1596 1597 ATADEBUG_PRINT(("wdsize\n"), DEBUG_FUNCS); 1598 1599 unit = WDUNIT(dev); 1600 wd = device_lookup_private(&wd_cd, unit); 1601 if (wd == NULL) 1602 return (-1); 1603 dksc = &wd->sc_dksc; 1604 1605 if (!device_is_active(dksc->sc_dev)) 1606 return (-1); 1607 1608 return dk_size(dksc, dev); 1609 } 1610 1611 /* 1612 * Dump core after a system crash. 1613 */ 1614 static int 1615 wddump(dev_t dev, daddr_t blkno, void *va, size_t size) 1616 { 1617 struct wd_softc *wd; 1618 struct dk_softc *dksc; 1619 int unit; 1620 1621 /* Check if recursive dump; if so, punt. */ 1622 if (wddoingadump) 1623 return EFAULT; 1624 wddoingadump = 1; 1625 1626 unit = WDUNIT(dev); 1627 wd = device_lookup_private(&wd_cd, unit); 1628 if (wd == NULL) 1629 return (ENXIO); 1630 dksc = &wd->sc_dksc; 1631 1632 return dk_dump(dksc, dev, blkno, va, size, 0); 1633 } 1634 1635 static int 1636 wd_dumpblocks(device_t dev, void *va, daddr_t blkno, int nblk) 1637 { 1638 struct wd_softc *wd = device_private(dev); 1639 struct dk_softc *dksc = &wd->sc_dksc; 1640 struct disk_geom *dg = &dksc->sc_dkdev.dk_geom; 1641 struct ata_xfer *xfer = &wd->dump_xfer; 1642 int err; 1643 1644 /* Recalibrate, if first dump transfer. */ 1645 if (wddumprecalibrated == 0) { 1646 wddumprecalibrated = 1; 1647 ata_channel_lock(wd->drvp->chnl_softc); 1648 /* This will directly execute the reset due to AT_POLL */ 1649 ata_thread_run(wd->drvp->chnl_softc, AT_POLL, 1650 ATACH_TH_DRIVE_RESET, wd->drvp->drive); 1651 1652 wd->drvp->state = RESET; 1653 ata_channel_unlock(wd->drvp->chnl_softc); 1654 } 1655 1656 memset(xfer, 0, sizeof(*xfer)); 1657 xfer->c_flags |= C_PRIVATE_ALLOC | C_SKIP_QUEUE; 1658 1659 xfer->c_bio.blkno = blkno; 1660 xfer->c_bio.flags = ATA_POLL; 1661 if (wd->sc_flags & WDF_LBA48 && 1662 (xfer->c_bio.blkno + nblk) > wd->sc_capacity28) 1663 xfer->c_bio.flags |= ATA_LBA48; 1664 if (wd->sc_flags & WDF_LBA) 1665 xfer->c_bio.flags |= ATA_LBA; 1666 xfer->c_bio.bcount = nblk * dg->dg_secsize; 1667 xfer->c_bio.databuf = va; 1668 #ifndef WD_DUMP_NOT_TRUSTED 1669 /* This will poll until the bio is complete */ 1670 wd->atabus->ata_bio(wd->drvp, xfer); 1671 1672 switch(err = xfer->c_bio.error) { 1673 case TIMEOUT: 1674 printf("wddump: device timed out"); 1675 err = EIO; 1676 break; 1677 case ERR_DF: 1678 printf("wddump: drive fault"); 1679 err = EIO; 1680 break; 1681 case ERR_DMA: 1682 printf("wddump: DMA error"); 1683 err = EIO; 1684 break; 1685 case ERROR: 1686 printf("wddump: "); 1687 wdperror(wd, xfer); 1688 err = EIO; 1689 break; 1690 case NOERROR: 1691 err = 0; 1692 break; 1693 default: 1694 panic("wddump: unknown error type %x", err); 1695 } 1696 1697 if (err != 0) { 1698 printf("\n"); 1699 return err; 1700 } 1701 #else /* WD_DUMP_NOT_TRUSTED */ 1702 /* Let's just talk about this first... */ 1703 printf("wd%d: dump addr 0x%x, cylin %d, head %d, sector %d\n", 1704 unit, va, cylin, head, sector); 1705 delay(500 * 1000); /* half a second */ 1706 #endif 1707 1708 wddoingadump = 0; 1709 return 0; 1710 } 1711 1712 #ifdef HAS_BAD144_HANDLING 1713 /* 1714 * Internalize the bad sector table. 1715 */ 1716 void 1717 bad144intern(struct wd_softc *wd) 1718 { 1719 struct dk_softc *dksc = &wd->sc_dksc; 1720 struct dkbad *bt = &dksc->sc_dkdev.dk_cpulabel->bad; 1721 struct disklabel *lp = dksc->sc_dkdev.dk_label; 1722 int i = 0; 1723 1724 ATADEBUG_PRINT(("bad144intern\n"), DEBUG_XFERS); 1725 1726 for (; i < NBT_BAD; i++) { 1727 if (bt->bt_bad[i].bt_cyl == 0xffff) 1728 break; 1729 wd->drvp->badsect[i] = 1730 bt->bt_bad[i].bt_cyl * lp->d_secpercyl + 1731 (bt->bt_bad[i].bt_trksec >> 8) * lp->d_nsectors + 1732 (bt->bt_bad[i].bt_trksec & 0xff); 1733 } 1734 for (; i < NBT_BAD+1; i++) 1735 wd->drvp->badsect[i] = -1; 1736 } 1737 #endif 1738 1739 static void 1740 wd_set_geometry(struct wd_softc *wd) 1741 { 1742 struct dk_softc *dksc = &wd->sc_dksc; 1743 struct disk_geom *dg = &dksc->sc_dkdev.dk_geom; 1744 1745 memset(dg, 0, sizeof(*dg)); 1746 1747 dg->dg_secperunit = wd->sc_capacity; 1748 dg->dg_secsize = wd->sc_blksize; 1749 dg->dg_nsectors = wd->sc_params.atap_sectors; 1750 dg->dg_ntracks = wd->sc_params.atap_heads; 1751 if ((wd->sc_flags & WDF_LBA) == 0) 1752 dg->dg_ncylinders = wd->sc_params.atap_cylinders; 1753 1754 disk_set_info(dksc->sc_dev, &dksc->sc_dkdev, wd->sc_typename); 1755 } 1756 1757 int 1758 wd_get_params(struct wd_softc *wd, struct ataparams *params) 1759 { 1760 int retry = 0; 1761 struct ata_channel *chp = wd->drvp->chnl_softc; 1762 const int flags = AT_WAIT; 1763 1764 again: 1765 switch (wd->atabus->ata_get_params(wd->drvp, flags, params)) { 1766 case CMD_AGAIN: 1767 return 1; 1768 case CMD_ERR: 1769 if (retry == 0) { 1770 retry++; 1771 ata_channel_lock(chp); 1772 (*wd->atabus->ata_reset_drive)(wd->drvp, flags, NULL); 1773 ata_channel_unlock(chp); 1774 goto again; 1775 } 1776 1777 if (wd->drvp->drive_type != ATA_DRIVET_OLD) 1778 return 1; 1779 /* 1780 * We `know' there's a drive here; just assume it's old. 1781 * This geometry is only used to read the MBR and print a 1782 * (false) attach message. 1783 */ 1784 strncpy(params->atap_model, "ST506", 1785 sizeof params->atap_model); 1786 params->atap_config = ATA_CFG_FIXED; 1787 params->atap_cylinders = 1024; 1788 params->atap_heads = 8; 1789 params->atap_sectors = 17; 1790 params->atap_multi = 1; 1791 params->atap_capabilities1 = params->atap_capabilities2 = 0; 1792 wd->drvp->ata_vers = -1; /* Mark it as pre-ATA */ 1793 /* FALLTHROUGH */ 1794 case CMD_OK: 1795 return 0; 1796 default: 1797 panic("wd_get_params: bad return code from ata_get_params"); 1798 /* NOTREACHED */ 1799 } 1800 } 1801 1802 int 1803 wd_getcache(struct wd_softc *wd, int *bitsp) 1804 { 1805 struct ataparams params; 1806 1807 if (wd_get_params(wd, ¶ms) != 0) 1808 return EIO; 1809 if (params.atap_cmd_set1 == 0x0000 || 1810 params.atap_cmd_set1 == 0xffff || 1811 (params.atap_cmd_set1 & WDC_CMD1_CACHE) == 0) { 1812 *bitsp = 0; 1813 return 0; 1814 } 1815 *bitsp = DKCACHE_WCHANGE | DKCACHE_READ; 1816 if (params.atap_cmd1_en & WDC_CMD1_CACHE) 1817 *bitsp |= DKCACHE_WRITE; 1818 1819 if (WD_USE_NCQ(wd) || (wd->drvp->drive_flags & ATA_DRIVE_WFUA)) 1820 *bitsp |= DKCACHE_FUA; 1821 1822 return 0; 1823 } 1824 1825 1826 static int 1827 wd_check_error(const struct dk_softc *dksc, const struct ata_xfer *xfer, 1828 const char *func) 1829 { 1830 static const char at_errbits[] = "\20\10ERROR\11TIMEOU\12DF"; 1831 1832 int flags = xfer->c_ata_c.flags; 1833 1834 if ((flags & AT_ERROR) != 0 && xfer->c_ata_c.r_error == WDCE_ABRT) { 1835 /* command not supported */ 1836 aprint_debug_dev(dksc->sc_dev, "%s: not supported\n", func); 1837 return ENODEV; 1838 } 1839 if (flags & (AT_ERROR | AT_TIMEOU | AT_DF)) { 1840 char sbuf[sizeof(at_errbits) + 64]; 1841 snprintb(sbuf, sizeof(sbuf), at_errbits, flags); 1842 device_printf(dksc->sc_dev, "%s: status=%s\n", func, sbuf); 1843 return EIO; 1844 } 1845 return 0; 1846 } 1847 1848 int 1849 wd_setcache(struct wd_softc *wd, int bits) 1850 { 1851 struct dk_softc *dksc = &wd->sc_dksc; 1852 struct ataparams params; 1853 struct ata_xfer *xfer; 1854 int error; 1855 1856 if (wd_get_params(wd, ¶ms) != 0) 1857 return EIO; 1858 1859 if (params.atap_cmd_set1 == 0x0000 || 1860 params.atap_cmd_set1 == 0xffff || 1861 (params.atap_cmd_set1 & WDC_CMD1_CACHE) == 0) 1862 return EOPNOTSUPP; 1863 1864 if ((bits & DKCACHE_READ) == 0 || 1865 (bits & DKCACHE_SAVE) != 0) 1866 return EOPNOTSUPP; 1867 1868 xfer = ata_get_xfer(wd->drvp->chnl_softc, true); 1869 1870 xfer->c_ata_c.r_command = SET_FEATURES; 1871 xfer->c_ata_c.r_st_bmask = 0; 1872 xfer->c_ata_c.r_st_pmask = 0; 1873 xfer->c_ata_c.timeout = 30000; /* 30s timeout */ 1874 xfer->c_ata_c.flags = AT_WAIT; 1875 if (bits & DKCACHE_WRITE) 1876 xfer->c_ata_c.r_features = WDSF_WRITE_CACHE_EN; 1877 else 1878 xfer->c_ata_c.r_features = WDSF_WRITE_CACHE_DS; 1879 1880 wd->atabus->ata_exec_command(wd->drvp, xfer); 1881 ata_wait_cmd(wd->drvp->chnl_softc, xfer); 1882 1883 error = wd_check_error(dksc, xfer, __func__); 1884 ata_free_xfer(wd->drvp->chnl_softc, xfer); 1885 return error; 1886 } 1887 1888 static int 1889 wd_standby(struct wd_softc *wd, int flags) 1890 { 1891 struct dk_softc *dksc = &wd->sc_dksc; 1892 struct ata_xfer *xfer; 1893 int error; 1894 1895 aprint_debug_dev(dksc->sc_dev, "standby immediate\n"); 1896 xfer = ata_get_xfer(wd->drvp->chnl_softc, true); 1897 1898 xfer->c_ata_c.r_command = WDCC_STANDBY_IMMED; 1899 xfer->c_ata_c.r_st_bmask = WDCS_DRDY; 1900 xfer->c_ata_c.r_st_pmask = WDCS_DRDY; 1901 xfer->c_ata_c.flags = flags; 1902 xfer->c_ata_c.timeout = 30000; /* 30s timeout */ 1903 1904 wd->atabus->ata_exec_command(wd->drvp, xfer); 1905 ata_wait_cmd(wd->drvp->chnl_softc, xfer); 1906 1907 error = wd_check_error(dksc, xfer, __func__); 1908 ata_free_xfer(wd->drvp->chnl_softc, xfer); 1909 return error; 1910 } 1911 1912 int 1913 wd_flushcache(struct wd_softc *wd, int flags) 1914 { 1915 struct dk_softc *dksc = &wd->sc_dksc; 1916 struct ata_xfer *xfer; 1917 int error; 1918 1919 /* 1920 * WDCC_FLUSHCACHE is here since ATA-4, but some drives report 1921 * only ATA-2 and still support it. 1922 */ 1923 if (wd->drvp->ata_vers < 4 && 1924 ((wd->sc_params.atap_cmd_set2 & WDC_CMD2_FC) == 0 || 1925 wd->sc_params.atap_cmd_set2 == 0xffff)) 1926 return ENODEV; 1927 1928 xfer = ata_get_xfer(wd->drvp->chnl_softc, true); 1929 1930 if ((wd->sc_params.atap_cmd2_en & ATA_CMD2_LBA48) != 0 && 1931 (wd->sc_params.atap_cmd2_en & ATA_CMD2_FCE) != 0) { 1932 xfer->c_ata_c.r_command = WDCC_FLUSHCACHE_EXT; 1933 flags |= AT_LBA48; 1934 } else 1935 xfer->c_ata_c.r_command = WDCC_FLUSHCACHE; 1936 xfer->c_ata_c.r_st_bmask = WDCS_DRDY; 1937 xfer->c_ata_c.r_st_pmask = WDCS_DRDY; 1938 xfer->c_ata_c.flags = flags | AT_READREG; 1939 xfer->c_ata_c.timeout = 300000; /* 5m timeout */ 1940 1941 wd->atabus->ata_exec_command(wd->drvp, xfer); 1942 ata_wait_cmd(wd->drvp->chnl_softc, xfer); 1943 1944 error = wd_check_error(dksc, xfer, __func__); 1945 wd->sc_flags &= ~WDF_DIRTY; 1946 ata_free_xfer(wd->drvp->chnl_softc, xfer); 1947 return error; 1948 } 1949 1950 /* 1951 * Execute TRIM command, assumes sleep context. 1952 */ 1953 static int 1954 wd_trim(struct wd_softc *wd, daddr_t bno, long size) 1955 { 1956 struct dk_softc *dksc = &wd->sc_dksc; 1957 struct ata_xfer *xfer; 1958 int error; 1959 unsigned char *req; 1960 1961 xfer = ata_get_xfer(wd->drvp->chnl_softc, true); 1962 1963 req = kmem_zalloc(512, KM_SLEEP); 1964 req[0] = bno & 0xff; 1965 req[1] = (bno >> 8) & 0xff; 1966 req[2] = (bno >> 16) & 0xff; 1967 req[3] = (bno >> 24) & 0xff; 1968 req[4] = (bno >> 32) & 0xff; 1969 req[5] = (bno >> 40) & 0xff; 1970 req[6] = size & 0xff; 1971 req[7] = (size >> 8) & 0xff; 1972 1973 /* 1974 * XXX We could possibly use NCQ TRIM, which supports executing 1975 * this command concurrently. It would need some investigation, some 1976 * early or not so early disk firmware caused data loss with NCQ TRIM. 1977 * atastart() et.al would need to be adjusted to allow and support 1978 * running several non-I/O ATA commands in parallel. 1979 */ 1980 1981 xfer->c_ata_c.r_command = ATA_DATA_SET_MANAGEMENT; 1982 xfer->c_ata_c.r_count = 1; 1983 xfer->c_ata_c.r_features = ATA_SUPPORT_DSM_TRIM; 1984 xfer->c_ata_c.r_st_bmask = WDCS_DRDY; 1985 xfer->c_ata_c.r_st_pmask = WDCS_DRDY; 1986 xfer->c_ata_c.timeout = 30000; /* 30s timeout */ 1987 xfer->c_ata_c.data = req; 1988 xfer->c_ata_c.bcount = 512; 1989 xfer->c_ata_c.flags |= AT_WRITE | AT_WAIT; 1990 1991 wd->atabus->ata_exec_command(wd->drvp, xfer); 1992 ata_wait_cmd(wd->drvp->chnl_softc, xfer); 1993 1994 kmem_free(req, 512); 1995 error = wd_check_error(dksc, xfer, __func__); 1996 ata_free_xfer(wd->drvp->chnl_softc, xfer); 1997 return error; 1998 } 1999 2000 bool 2001 wd_shutdown(device_t dev, int how) 2002 { 2003 struct wd_softc *wd = device_private(dev); 2004 2005 /* the adapter needs to be enabled */ 2006 if (wd->atabus->ata_addref(wd->drvp)) 2007 return true; /* no need to complain */ 2008 2009 wd_flushcache(wd, AT_POLL); 2010 if ((how & RB_POWERDOWN) == RB_POWERDOWN) 2011 wd_standby(wd, AT_POLL); 2012 return true; 2013 } 2014 2015 /* 2016 * Allocate space for a ioctl queue structure. Mostly taken from 2017 * scsipi_ioctl.c 2018 */ 2019 struct wd_ioctl * 2020 wi_get(struct wd_softc *wd) 2021 { 2022 struct wd_ioctl *wi; 2023 2024 wi = kmem_zalloc(sizeof(struct wd_ioctl), KM_SLEEP); 2025 wi->wi_softc = wd; 2026 buf_init(&wi->wi_bp); 2027 2028 return (wi); 2029 } 2030 2031 /* 2032 * Free an ioctl structure and remove it from our list 2033 */ 2034 2035 void 2036 wi_free(struct wd_ioctl *wi) 2037 { 2038 buf_destroy(&wi->wi_bp); 2039 kmem_free(wi, sizeof(*wi)); 2040 } 2041 2042 /* 2043 * Find a wd_ioctl structure based on the struct buf. 2044 */ 2045 2046 struct wd_ioctl * 2047 wi_find(struct buf *bp) 2048 { 2049 return container_of(bp, struct wd_ioctl, wi_bp); 2050 } 2051 2052 static uint 2053 wi_sector_size(const struct wd_ioctl * const wi) 2054 { 2055 switch (wi->wi_atareq.command) { 2056 case WDCC_READ: 2057 case WDCC_WRITE: 2058 case WDCC_READMULTI: 2059 case WDCC_WRITEMULTI: 2060 case WDCC_READDMA: 2061 case WDCC_WRITEDMA: 2062 case WDCC_READ_EXT: 2063 case WDCC_WRITE_EXT: 2064 case WDCC_READMULTI_EXT: 2065 case WDCC_WRITEMULTI_EXT: 2066 case WDCC_READDMA_EXT: 2067 case WDCC_WRITEDMA_EXT: 2068 case WDCC_READ_FPDMA_QUEUED: 2069 case WDCC_WRITE_FPDMA_QUEUED: 2070 return wi->wi_softc->sc_blksize; 2071 default: 2072 return 512; 2073 } 2074 } 2075 2076 /* 2077 * Ioctl pseudo strategy routine 2078 * 2079 * This is mostly stolen from scsipi_ioctl.c:scsistrategy(). What 2080 * happens here is: 2081 * 2082 * - wdioctl() queues a wd_ioctl structure. 2083 * 2084 * - wdioctl() calls physio/wdioctlstrategy based on whether or not 2085 * user space I/O is required. If physio() is called, physio() eventually 2086 * calls wdioctlstrategy(). 2087 * 2088 * - In either case, wdioctlstrategy() calls wd->atabus->ata_exec_command() 2089 * to perform the actual command 2090 * 2091 * The reason for the use of the pseudo strategy routine is because 2092 * when doing I/O to/from user space, physio _really_ wants to be in 2093 * the loop. We could put the entire buffer into the ioctl request 2094 * structure, but that won't scale if we want to do things like download 2095 * microcode. 2096 */ 2097 2098 void 2099 wdioctlstrategy(struct buf *bp) 2100 { 2101 struct wd_ioctl *wi; 2102 struct ata_xfer *xfer; 2103 int error = 0; 2104 2105 wi = wi_find(bp); 2106 if (wi == NULL) { 2107 printf("wdioctlstrategy: " 2108 "No matching ioctl request found in queue\n"); 2109 error = EINVAL; 2110 goto out2; 2111 } 2112 2113 xfer = ata_get_xfer(wi->wi_softc->drvp->chnl_softc, true); 2114 2115 /* 2116 * Abort if physio broke up the transfer 2117 */ 2118 2119 if (bp->b_bcount != wi->wi_atareq.datalen) { 2120 printf("physio split wd ioctl request... cannot proceed\n"); 2121 error = EIO; 2122 goto out; 2123 } 2124 2125 /* 2126 * Abort if we didn't get a buffer size that was a multiple of 2127 * our sector size (or overflows CHS/LBA28 sector count) 2128 */ 2129 2130 if ((bp->b_bcount % wi_sector_size(wi)) != 0 || 2131 (bp->b_bcount / wi_sector_size(wi)) >= 2132 (1 << NBBY)) { 2133 error = EINVAL; 2134 goto out; 2135 } 2136 2137 /* 2138 * Make sure a timeout was supplied in the ioctl request 2139 */ 2140 2141 if (wi->wi_atareq.timeout == 0) { 2142 error = EINVAL; 2143 goto out; 2144 } 2145 2146 if (wi->wi_atareq.flags & ATACMD_READ) 2147 xfer->c_ata_c.flags |= AT_READ; 2148 else if (wi->wi_atareq.flags & ATACMD_WRITE) 2149 xfer->c_ata_c.flags |= AT_WRITE; 2150 2151 if (wi->wi_atareq.flags & ATACMD_READREG) 2152 xfer->c_ata_c.flags |= AT_READREG; 2153 2154 if ((wi->wi_atareq.flags & ATACMD_LBA) != 0) 2155 xfer->c_ata_c.flags |= AT_LBA; 2156 2157 xfer->c_ata_c.flags |= AT_WAIT; 2158 2159 xfer->c_ata_c.timeout = wi->wi_atareq.timeout; 2160 xfer->c_ata_c.r_command = wi->wi_atareq.command; 2161 xfer->c_ata_c.r_lba = ((wi->wi_atareq.head & 0x0f) << 24) | 2162 (wi->wi_atareq.cylinder << 8) | 2163 wi->wi_atareq.sec_num; 2164 xfer->c_ata_c.r_count = wi->wi_atareq.sec_count; 2165 xfer->c_ata_c.r_features = wi->wi_atareq.features; 2166 xfer->c_ata_c.r_st_bmask = WDCS_DRDY; 2167 xfer->c_ata_c.r_st_pmask = WDCS_DRDY; 2168 xfer->c_ata_c.data = wi->wi_bp.b_data; 2169 xfer->c_ata_c.bcount = wi->wi_bp.b_bcount; 2170 2171 wi->wi_softc->atabus->ata_exec_command(wi->wi_softc->drvp, xfer); 2172 ata_wait_cmd(wi->wi_softc->drvp->chnl_softc, xfer); 2173 2174 if (xfer->c_ata_c.flags & (AT_ERROR | AT_TIMEOU | AT_DF)) { 2175 if (xfer->c_ata_c.flags & AT_ERROR) { 2176 wi->wi_atareq.retsts = ATACMD_ERROR; 2177 wi->wi_atareq.error = xfer->c_ata_c.r_error; 2178 } else if (xfer->c_ata_c.flags & AT_DF) 2179 wi->wi_atareq.retsts = ATACMD_DF; 2180 else 2181 wi->wi_atareq.retsts = ATACMD_TIMEOUT; 2182 } else { 2183 wi->wi_atareq.retsts = ATACMD_OK; 2184 if (wi->wi_atareq.flags & ATACMD_READREG) { 2185 wi->wi_atareq.command = xfer->c_ata_c.r_status; 2186 wi->wi_atareq.features = xfer->c_ata_c.r_error; 2187 wi->wi_atareq.sec_count = xfer->c_ata_c.r_count; 2188 wi->wi_atareq.sec_num = xfer->c_ata_c.r_lba & 0xff; 2189 wi->wi_atareq.head = (xfer->c_ata_c.r_device & 0xf0) | 2190 ((xfer->c_ata_c.r_lba >> 24) & 0x0f); 2191 wi->wi_atareq.cylinder = 2192 (xfer->c_ata_c.r_lba >> 8) & 0xffff; 2193 wi->wi_atareq.error = xfer->c_ata_c.r_error; 2194 } 2195 } 2196 2197 out: 2198 ata_free_xfer(wi->wi_softc->drvp->chnl_softc, xfer); 2199 out2: 2200 bp->b_error = error; 2201 if (error) 2202 bp->b_resid = bp->b_bcount; 2203 biodone(bp); 2204 } 2205 2206 static void 2207 wd_sysctl_attach(struct wd_softc *wd) 2208 { 2209 struct dk_softc *dksc = &wd->sc_dksc; 2210 const struct sysctlnode *node; 2211 int error; 2212 2213 /* sysctl set-up */ 2214 if (sysctl_createv(&wd->nodelog, 0, NULL, &node, 2215 0, CTLTYPE_NODE, dksc->sc_xname, 2216 SYSCTL_DESCR("wd driver settings"), 2217 NULL, 0, NULL, 0, 2218 CTL_HW, CTL_CREATE, CTL_EOL) != 0) { 2219 aprint_error_dev(dksc->sc_dev, 2220 "could not create %s.%s sysctl node\n", 2221 "hw", dksc->sc_xname); 2222 return; 2223 } 2224 2225 wd->drv_ncq = true; 2226 if ((error = sysctl_createv(&wd->nodelog, 0, NULL, NULL, 2227 CTLFLAG_READWRITE, CTLTYPE_BOOL, "use_ncq", 2228 SYSCTL_DESCR("use NCQ if supported"), 2229 NULL, 0, &wd->drv_ncq, 0, 2230 CTL_HW, node->sysctl_num, CTL_CREATE, CTL_EOL)) 2231 != 0) { 2232 aprint_error_dev(dksc->sc_dev, 2233 "could not create %s.%s.use_ncq sysctl - error %d\n", 2234 "hw", dksc->sc_xname, error); 2235 return; 2236 } 2237 2238 wd->drv_ncq_prio = false; 2239 if ((error = sysctl_createv(&wd->nodelog, 0, NULL, NULL, 2240 CTLFLAG_READWRITE, CTLTYPE_BOOL, "use_ncq_prio", 2241 SYSCTL_DESCR("use NCQ PRIORITY if supported"), 2242 NULL, 0, &wd->drv_ncq_prio, 0, 2243 CTL_HW, node->sysctl_num, CTL_CREATE, CTL_EOL)) 2244 != 0) { 2245 aprint_error_dev(dksc->sc_dev, 2246 "could not create %s.%s.use_ncq_prio sysctl - error %d\n", 2247 "hw", dksc->sc_xname, error); 2248 return; 2249 } 2250 2251 #ifdef WD_CHAOS_MONKEY 2252 wd->drv_chaos_freq = 0; 2253 if ((error = sysctl_createv(&wd->nodelog, 0, NULL, NULL, 2254 CTLFLAG_READWRITE, CTLTYPE_INT, "chaos_freq", 2255 SYSCTL_DESCR("simulated bio read error rate"), 2256 NULL, 0, &wd->drv_chaos_freq, 0, 2257 CTL_HW, node->sysctl_num, CTL_CREATE, CTL_EOL)) 2258 != 0) { 2259 aprint_error_dev(dksc->sc_dev, 2260 "could not create %s.%s.chaos_freq sysctl - error %d\n", 2261 "hw", dksc->sc_xname, error); 2262 return; 2263 } 2264 2265 wd->drv_chaos_cnt = 0; 2266 if ((error = sysctl_createv(&wd->nodelog, 0, NULL, NULL, 2267 CTLFLAG_READONLY, CTLTYPE_INT, "chaos_cnt", 2268 SYSCTL_DESCR("number of processed bio reads"), 2269 NULL, 0, &wd->drv_chaos_cnt, 0, 2270 CTL_HW, node->sysctl_num, CTL_CREATE, CTL_EOL)) 2271 != 0) { 2272 aprint_error_dev(dksc->sc_dev, 2273 "could not create %s.%s.chaos_cnt sysctl - error %d\n", 2274 "hw", dksc->sc_xname, error); 2275 return; 2276 } 2277 #endif 2278 2279 } 2280 2281 static void 2282 wd_sysctl_detach(struct wd_softc *wd) 2283 { 2284 sysctl_teardown(&wd->nodelog); 2285 } 2286 2287 #ifdef ATADEBUG 2288 int wddebug(void); 2289 2290 int 2291 wddebug(void) 2292 { 2293 struct wd_softc *wd; 2294 struct dk_softc *dksc; 2295 int unit; 2296 2297 for (unit = 0; unit <= 3; unit++) { 2298 wd = device_lookup_private(&wd_cd, unit); 2299 if (wd == NULL) 2300 continue; 2301 dksc = &wd->sc_dksc; 2302 printf("%s fl %x bufq %p:\n", 2303 dksc->sc_xname, wd->sc_flags, bufq_peek(dksc->sc_bufq)); 2304 2305 atachannel_debug(wd->drvp->chnl_softc); 2306 } 2307 return 0; 2308 } 2309 #endif /* ATADEBUG */ 2310
Indexes created Sat Sep 27 22:09:54 GMT 2025