usr/src/cmd/devfsadm/disk_link.c

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or http://www.opensolaris.org/os/licensing.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21
  22 /*
  23  * Copyright 2016 Toomas Soome <tsoome@me.com>
  24  * Copyright 2017 Nexenta Systems, Inc.
  25  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  26  * Use is subject to license terms.
  27  */
  28
  29 #include <devfsadm.h>
  30 #include <stdio.h>
  31 #include <strings.h>
  32 #include <stdlib.h>
  33 #include <limits.h>
  34 #include <ctype.h>
  35 #include <unistd.h>
  36 #include <sys/int_fmtio.h>
  37 #include <sys/stat.h>
  38 #include <sys/scsi/scsi_address.h>
  39 #include <sys/libdevid.h>
  40 #include <sys/lofi.h>
  41
  42 #define DISK_SUBPATH_MAX 100
  43 #define RM_STALE 0x01
  44 #define DISK_LINK_RE    "^r?dsk/c[0-9]+(t[0-9A-F]+)?d[0-9]+(((s|p))[0-9]+)?$"
  45 #define DISK_LINK_TO_UPPER(ch)\
  46         (((ch) >= 'a' && (ch) <= 'z') ? (ch - 'a' + 'A') : ch)
  47
  48 #define SLICE_SMI       "s7"
  49 #define SLICE_EFI       ""
  50
  51 #define MN_SMI          "h"
  52 #define MN_EFI          "wd"
  53 #define ASCIIWWNSIZE    255
  54 #if defined(__i386) || defined(__amd64)
  55 /*
  56  * The number of minor nodes per LUN is defined by the disk drivers.
  57  * Currently it is set to 64. Refer CMLBUNIT_SHIFT (cmlb_impl.h)
  58  */
  59 #define NUM_MINORS_PER_INSTANCE 64
  60 #endif
  61
  62
  63 static int disk_callback_chan(di_minor_t minor, di_node_t node);
  64 static int disk_callback_nchan(di_minor_t minor, di_node_t node);
  65 static int disk_callback_blkdev(di_minor_t minor, di_node_t node);
  66 static int disk_callback_wwn(di_minor_t minor, di_node_t node);
  67 static int disk_callback_xvmd(di_minor_t minor, di_node_t node);
  68 static int disk_callback_fabric(di_minor_t minor, di_node_t node);
  69 static int disk_callback_sas(di_minor_t minor, di_node_t node);
  70 static void disk_common(di_minor_t minor, di_node_t node, char *disk,
  71                                 int flags);
  72 static char *diskctrl(di_node_t node, di_minor_t minor);
  73 static int reserved_links_exist(di_node_t node, di_minor_t minor, int nflags);
  74 static void disk_rm_lofi_all(char *file);
  75
  76
  77 static devfsadm_create_t disk_cbt[] = {
  78         { "disk", DDI_NT_BLOCK, NULL,
  79             TYPE_EXACT, ILEVEL_0, disk_callback_nchan
  80         },
  81         { "disk", DDI_NT_BLOCK_CHAN, NULL,
  82             TYPE_EXACT, ILEVEL_0, disk_callback_chan
  83         },
  84         { "disk", DDI_NT_BLOCK_BLKDEV, NULL,
  85             TYPE_EXACT, ILEVEL_0, disk_callback_blkdev
  86         },
  87         { "disk", DDI_NT_BLOCK_FABRIC, NULL,
  88                 TYPE_EXACT, ILEVEL_0, disk_callback_fabric
  89         },
  90         { "disk", DDI_NT_BLOCK_WWN, NULL,
  91             TYPE_EXACT, ILEVEL_0, disk_callback_wwn
  92         },
  93         { "disk", DDI_NT_BLOCK_SAS, NULL,
  94             TYPE_EXACT, ILEVEL_0, disk_callback_sas
  95         },
  96         { "disk", DDI_NT_CD, NULL,
  97             TYPE_EXACT, ILEVEL_0, disk_callback_nchan
  98         },
  99         { "disk", DDI_NT_CD_CHAN, NULL,
 100             TYPE_EXACT, ILEVEL_0, disk_callback_chan
 101         },
 102         { "disk", DDI_NT_BLOCK_XVMD, NULL,
 103             TYPE_EXACT, ILEVEL_0, disk_callback_xvmd
 104         },
 105         { "disk", DDI_NT_CD_XVMD, NULL,
 106             TYPE_EXACT, ILEVEL_0, disk_callback_xvmd
 107         },
 108 };
 109
 110 DEVFSADM_CREATE_INIT_V0(disk_cbt);
 111
 112 /*
 113  * HOT auto cleanup of disks is done for lofi devices only.
 114  */
 115 static devfsadm_remove_t disk_remove_cbt[] = {
 116         { "disk", DISK_LINK_RE, RM_HOT | RM_POST | RM_ALWAYS,
 117                 ILEVEL_0, disk_rm_lofi_all
 118         },
 119         { "disk", DISK_LINK_RE, RM_POST,
 120                 ILEVEL_0, devfsadm_rm_all
 121         }
 122 };
 123
 124 DEVFSADM_REMOVE_INIT_V0(disk_remove_cbt);
 125
 126 static devlink_re_t disks_re_array[] = {
 127         {"^r?dsk/c([0-9]+)", 1},
 128         {"^cfg/c([0-9]+)$", 1},
 129         {"^scsi/.+/c([0-9]+)", 1},
 130         {NULL}
 131 };
 132
 133 static char *disk_mid = "disk_mid";
 134 static char *modname = "disk_link";
 135
 136 /*
 137  * Check if link is from lofi by checking path from readlink().
 138  */
 139 static int
 140 is_lofi_disk(char *file)
 141 {
 142         char buf[PATH_MAX + 1];
 143         char filepath[PATH_MAX];
 144         char *ptr;
 145         ssize_t size;
 146
 147         size = snprintf(filepath, sizeof (filepath), "%s/dev/%s",
 148             devfsadm_root_path(), file);
 149         if (size > sizeof (filepath))
 150                 return (0);
 151
 152         size = readlink(filepath, buf, sizeof (buf) - 1);
 153         if (size == -1)
 154                 return (0);
 155         buf[size] = '\0';
 156         ptr = strchr(buf, '@');
 157         if (ptr == NULL)
 158                 return (0);
 159         ptr[1] = '\0';
 160         if (strcmp(buf, "../../devices/pseudo/lofi@") != 0)
 161                 return (0);
 162         return (1);
 163 }
 164
 165 /*
 166  * Wrapper around devfsadm_rm_link() for lofi devices.
 167  */
 168 static void disk_rm_lofi_all(char *file)
 169 {
 170         if (is_lofi_disk(file))
 171                 devfsadm_rm_link(file);
 172 }
 173
 174 int
 175 minor_init()
 176 {
 177         devfsadm_print(disk_mid,
 178             "%s: minor_init(): Creating disks reserved ID cache\n",
 179             modname);
 180         return (devfsadm_reserve_id_cache(disks_re_array, NULL));
 181 }
 182
 183 static int
 184 disk_callback_chan(di_minor_t minor, di_node_t node)
 185 {
 186         char *addr;
 187         char disk[23];
 188         char *driver;
 189         uint_t targ = 0;
 190         uint_t lun = 0;
 191
 192         driver = di_driver_name(node);
 193         if (strcmp(driver, LOFI_DRIVER_NAME) != 0) {
 194                 addr = di_bus_addr(node);
 195                 (void) sscanf(addr, "%X,%X", &targ, &lun);
 196         } else {
 197                 targ = di_instance(node);
 198         }
 199
 200         (void) snprintf(disk, sizeof (disk), "t%dd%d", targ, lun);
 201         disk_common(minor, node, disk, 0);
 202         return (DEVFSADM_CONTINUE);
 203
 204 }
 205
 206 static int
 207 disk_callback_nchan(di_minor_t minor, di_node_t node)
 208 {
 209         char *addr;
 210         char disk[10];
 211         uint_t lun;
 212
 213         addr = di_bus_addr(node);
 214         (void) sscanf(addr, "%X", &lun);
 215         (void) sprintf(disk, "d%d", lun);
 216         disk_common(minor, node, disk, 0);
 217         return (DEVFSADM_CONTINUE);
 218
 219 }
 220
 221 static int
 222 disk_callback_blkdev(di_minor_t minor, di_node_t node)
 223 {
 224         char *addr;
 225         char disk[DISK_SUBPATH_MAX];
 226         uint64_t eui64;
 227         uint_t lun = 0;
 228
 229         addr = di_bus_addr(node);
 230         (void) sscanf(addr, "w%016"PRIx64",%X", &eui64, &lun);
 231         (void) snprintf(disk, DISK_SUBPATH_MAX, "t%016"PRIX64"d%d", eui64, lun);
 232         disk_common(minor, node, disk, RM_STALE);
 233         return (DEVFSADM_CONTINUE);
 234 }
 235
 236 static int
 237 disk_callback_wwn(di_minor_t minor, di_node_t node)
 238 {
 239         char disk[10];
 240         int lun;
 241         int targ;
 242         int *intp;
 243
 244         if (di_prop_lookup_ints(DDI_DEV_T_ANY, node, SCSI_ADDR_PROP_TARGET,
 245             &intp) <= 0) {
 246                 return (DEVFSADM_CONTINUE);
 247         }
 248         targ = *intp;
 249         if (di_prop_lookup_ints(DDI_DEV_T_ANY, node, SCSI_ADDR_PROP_LUN,
 250             &intp) <= 0) {
 251                 lun = 0;
 252         } else {
 253                 lun = *intp;
 254         }
 255         (void) sprintf(disk, "t%dd%d", targ, lun);
 256
 257         disk_common(minor, node, disk, RM_STALE);
 258
 259         return (DEVFSADM_CONTINUE);
 260 }
 261
 262 static int
 263 disk_callback_fabric(di_minor_t minor, di_node_t node)
 264 {
 265         char disk[DISK_SUBPATH_MAX];
 266         int lun;
 267         int count;
 268         int *intp;
 269         uchar_t *str;
 270         uchar_t *wwn;
 271         uchar_t ascii_wwn[ASCIIWWNSIZE];
 272
 273         if (di_prop_lookup_strings(DDI_DEV_T_ANY, node,
 274             "client-guid", (char **)&wwn) > 0) {
 275                 if (strlcpy((char *)ascii_wwn, (char *)wwn,
 276                     sizeof (ascii_wwn)) >= sizeof (ascii_wwn)) {
 277                         devfsadm_errprint("SUNW_disk_link: GUID too long:%d",
 278                             strlen((char *)wwn));
 279                         return (DEVFSADM_CONTINUE);
 280                 }
 281                 lun = 0;
 282         } else if (di_prop_lookup_bytes(DDI_DEV_T_ANY, node,
 283             "port-wwn", &wwn) > 0) {
 284                 if (di_prop_lookup_ints(DDI_DEV_T_ANY, node,
 285                     SCSI_ADDR_PROP_LUN, &intp) > 0) {
 286                         lun = *intp;
 287                 } else {
 288                         lun = 0;
 289                 }
 290
 291                 for (count = 0, str = ascii_wwn; count < 8; count++, str += 2) {
 292                         (void) sprintf((caddr_t)str, "%02x", wwn[count]);
 293                 }
 294                 *str = '\0';
 295         } else {
 296                 return (DEVFSADM_CONTINUE);
 297         }
 298
 299         for (str = ascii_wwn; *str != '\0'; str++) {
 300                 *str = DISK_LINK_TO_UPPER(*str);
 301         }
 302
 303         (void) snprintf(disk, DISK_SUBPATH_MAX, "t%sd%d", ascii_wwn, lun);
 304
 305         disk_common(minor, node, disk, RM_STALE);
 306
 307         return (DEVFSADM_CONTINUE);
 308 }
 309
 310 static int
 311 disk_callback_sas(di_minor_t minor, di_node_t node)
 312 {
 313         char disk[DISK_SUBPATH_MAX];
 314         int lun64_found = 0;
 315         scsi_lun64_t lun64, sl;
 316         scsi_lun_t lun;
 317         int64_t *lun64p;
 318         uint64_t wwn;
 319         int *intp;
 320         char *tgt_port;
 321         uchar_t addr_method;
 322
 323         /* Get lun property */
 324         if (di_prop_lookup_int64(DDI_DEV_T_ANY, node,
 325             SCSI_ADDR_PROP_LUN64, &lun64p) > 0) {
 326                 if (*lun64p != SCSI_LUN64_ILLEGAL) {
 327                         lun64_found = 1;
 328                         lun64 = (uint64_t)*lun64p;
 329                 }
 330         }
 331         if ((!lun64_found) && (di_prop_lookup_ints(DDI_DEV_T_ANY, node,
 332             SCSI_ADDR_PROP_LUN, &intp) > 0)) {
 333                 lun64 = (uint64_t)*intp;
 334         }
 335
 336         lun = scsi_lun64_to_lun(lun64);
 337
 338         addr_method = (lun.sl_lun1_msb & SCSI_LUN_AM_MASK);
 339
 340         if (di_prop_lookup_strings(DDI_DEV_T_ANY, node,
 341             SCSI_ADDR_PROP_TARGET_PORT, &tgt_port) > 0) {
 342                 (void) scsi_wwnstr_to_wwn(tgt_port, &wwn);
 343                 if ((addr_method == SCSI_LUN_AM_PDEV) &&
 344                     (lun.sl_lun2_msb == 0) && (lun.sl_lun2_lsb == 0) &&
 345                     (lun.sl_lun3_msb == 0) && (lun.sl_lun3_lsb == 0) &&
 346                     (lun.sl_lun4_msb == 0) && (lun.sl_lun4_lsb == 0)) {
 347                         (void) snprintf(disk, DISK_SUBPATH_MAX,
 348                             "t%"PRIX64"d%"PRId64, wwn, lun64);
 349                 } else if ((addr_method == SCSI_LUN_AM_FLAT) &&
 350                     (lun.sl_lun2_msb == 0) && (lun.sl_lun2_lsb == 0) &&
 351                     (lun.sl_lun3_msb == 0) && (lun.sl_lun3_lsb == 0) &&
 352                     (lun.sl_lun4_msb == 0) && (lun.sl_lun4_lsb == 0)) {
 353                         sl = (lun.sl_lun1_msb << 8) | lun.sl_lun1_lsb;
 354                         (void) snprintf(disk, DISK_SUBPATH_MAX,
 355                             "t%"PRIX64"d%"PRIX16, wwn, sl);
 356                 } else {
 357                         (void) snprintf(disk, DISK_SUBPATH_MAX,
 358                             "t%"PRIX64"d%"PRIX64, wwn, lun64);
 359                 }
 360         } else if (di_prop_lookup_ints(DDI_DEV_T_ANY, node,
 361             SCSI_ADDR_PROP_SATA_PHY, &intp) > 0) {
 362                 /* Use phy format naming, for SATA devices without wwn */
 363                 if ((addr_method == SCSI_LUN_AM_PDEV) &&
 364                     (lun.sl_lun2_msb == 0) && (lun.sl_lun2_lsb == 0) &&
 365                     (lun.sl_lun3_msb == 0) && (lun.sl_lun3_lsb == 0) &&
 366                     (lun.sl_lun4_msb == 0) && (lun.sl_lun4_lsb == 0)) {
 367                         (void) snprintf(disk, DISK_SUBPATH_MAX,
 368                             "t%dd%"PRId64, *intp, lun64);
 369                 } else if ((addr_method == SCSI_LUN_AM_FLAT) &&
 370                     (lun.sl_lun2_msb == 0) && (lun.sl_lun2_lsb == 0) &&
 371                     (lun.sl_lun3_msb == 0) && (lun.sl_lun3_lsb == 0) &&
 372                     (lun.sl_lun4_msb == 0) && (lun.sl_lun4_lsb == 0)) {
 373                         sl = (lun.sl_lun1_msb << 8) | lun.sl_lun1_lsb;
 374                         (void) snprintf(disk, DISK_SUBPATH_MAX,
 375                             "t%dd%"PRIX16, *intp, sl);
 376                 } else {
 377                         (void) snprintf(disk, DISK_SUBPATH_MAX,
 378                             "t%dd%"PRIX64, *intp, lun64);
 379                 }
 380         } else {
 381                 return (DEVFSADM_CONTINUE);
 382         }
 383
 384         disk_common(minor, node, disk, RM_STALE);
 385
 386         return (DEVFSADM_CONTINUE);
 387 }
 388
 389 /*
 390  * xVM virtual block device
 391  *
 392  * Xen passes device number in the following format:
 393  *
 394  *    1 << 28 | disk << 8 | partition      xvd, disks or partitions 16 onwards
 395  *  202 <<  8 | disk << 4 | partition      xvd, disks and partitions up to 15
 396  *    8 <<  8 | disk << 4 | partition      sd, disks and partitions up to 15
 397  *    3 <<  8 | disk << 6 | partition      hd, disks 0..1, partitions 0..63
 398  *   22 <<  8 | (disk-2) << 6 | partition  hd, disks 2..3, partitions 0..63
 399  *    2 << 28 onwards                      reserved for future use
 400  *   other values less than 1 << 28        deprecated / reserved
 401  *
 402  * The corresponding /dev/dsk name can be:
 403  *
 404  *          c0tYdX
 405  *
 406  * where Y,X >= 0.
 407  *
 408  * For PV guests using the legacy naming (0, 1, 2, ...)
 409  * the disk names created will be c0d[0..767].
 410  */
 411
 412 #define HD_BASE         (3 << 8)
 413 #define XEN_EXT_SHIFT   (28)
 414
 415 /*
 416  * Return: Number of parsed and written parameters
 417  */
 418 static int
 419 decode_xen_device(uint_t device, uint_t *disk, uint_t *plun)
 420 {
 421         uint_t dsk, lun = 0;
 422         int ret = 1;
 423
 424         if ((device >> XEN_EXT_SHIFT) > 1)
 425                 return (0);
 426
 427         if (device < HD_BASE) {
 428                 /* legacy device address */
 429                 dsk = device;
 430                 goto end;
 431         }
 432
 433         ret = 2;
 434         if (device & (1 << XEN_EXT_SHIFT)) {
 435                 /* extended */
 436                 dsk = device & (~0xff);
 437                 lun = device & 0xff;
 438                 goto end;
 439         }
 440
 441         switch (device >> 8) {
 442         case 202:                               /* xvd */
 443                 dsk = (device >> 4) & 0xf;
 444                 lun =  device & 0xf;
 445                 break;
 446         case 8:                                 /* sd */
 447                 dsk = device & (~0xf);
 448                 lun = device & 0xf;
 449                 break;
 450         case 3:                                 /* hd, disk 0..1 */
 451                 dsk = (device >> 6) & 0x1;
 452                 lun = device & 0x3f;
 453                 break;
 454         case 22:                                /* hd, disk 2..3 */
 455                 dsk = ((device >> 6) & 0x1) + 2;
 456                 lun = device & 0x3f;
 457                 break;
 458         default:
 459                 return (0);
 460         }
 461 end:
 462         *disk = dsk;
 463         *plun = lun;
 464         return (ret);
 465 }
 466
 467 static int
 468 disk_callback_xvmd(di_minor_t minor, di_node_t node)
 469 {
 470         char *addr;
 471         char disk[16];
 472         uint_t targ;
 473         uint_t dsk, lun;
 474         int res;
 475
 476         addr = di_bus_addr(node);
 477         targ = strtol(addr, (char **)NULL, 10);
 478
 479         res = decode_xen_device(targ, &dsk, &lun);
 480
 481         /* HVM device names are generated using the standard generator */
 482
 483         if (res == 1)
 484                 (void) snprintf(disk, sizeof (disk),  "d%d", dsk);
 485         else if (res == 2)
 486                 (void) snprintf(disk, sizeof (disk), "t%dd%d", dsk, lun);
 487         else {
 488                 devfsadm_errprint("%s: invalid disk device number (%s)\n",
 489                     modname, addr);
 490                 return (DEVFSADM_CONTINUE);
 491         }
 492         disk_common(minor, node, disk, 0);
 493         return (DEVFSADM_CONTINUE);
 494 }
 495
 496 /*
 497  * This function is called for every disk minor node.
 498  * Calls enumerate to assign a logical controller number, and
 499  * then devfsadm_mklink to make the link.
 500  */
 501 static void
 502 disk_common(di_minor_t minor, di_node_t node, char *disk, int flags)
 503 {
 504         char l_path[PATH_MAX + 1];
 505         char sec_path[PATH_MAX + 1];
 506         char stale_re[DISK_SUBPATH_MAX];
 507         char *dir;
 508         char slice[4];
 509         char *mn;
 510         char *ctrl;
 511         int *int_prop;
 512         int  nflags = 0;
 513 #if defined(__i386) || defined(__amd64)
 514         char mn_copy[4];
 515         char *part;
 516         int part_num;
 517 #endif
 518
 519         mn = di_minor_name(minor);
 520         if (strstr(mn, ",raw")) {
 521                 dir = "rdsk";
 522 #if defined(__i386) || defined(__amd64)
 523                 (void) strncpy(mn_copy, mn, 4);
 524                 part = strtok(mn_copy, ",");
 525 #endif
 526         } else {
 527                 dir = "dsk";
 528 #if defined(__i386) || defined(__amd64)
 529                 part = mn;
 530 #endif
 531         }
 532
 533 #if defined(__i386) || defined(__amd64)
 534         /*
 535          * The following is a table describing the allocation of
 536          * minor numbers, minor names and /dev/dsk names for partitions
 537          * and slices on x86 systems.
 538          *
 539          *      Minor Number    Minor Name      /dev/dsk name
 540          *      ---------------------------------------------
 541          *      0 to 15         "a" to "p"      s0 to s15
 542          *      16              "q"             p0
 543          *      17 to 20        "r" to "u"      p1 to p4
 544          *      21 to 52        "p5" to "p36"   p5 to p36
 545          *
 546          */
 547         part_num = atoi(part + 1);
 548
 549         if ((mn[0] == 'p') && (part_num >= 5)) {
 550                 /* logical drive */
 551                 (void) snprintf(slice, 4, "%s", part);
 552         } else {
 553 #endif
 554         if (mn[0] < 'q') {
 555                 (void) sprintf(slice, "s%d", mn[0] - 'a');
 556         } else if (strncmp(mn, MN_EFI, 2) != 0) {
 557                 (void) sprintf(slice, "p%d", mn[0] - 'q');
 558         } else {
 559                 /* For EFI label */
 560                 (void) sprintf(slice, SLICE_EFI);
 561         }
 562 #if defined(__i386) || defined(__amd64)
 563         }
 564 #endif
 565
 566         nflags = 0;
 567         if (reserved_links_exist(node, minor, nflags) == DEVFSADM_SUCCESS) {
 568                 devfsadm_print(disk_mid, "Reserved link exists. Not "
 569                     "creating links for slice %s\n", slice);
 570                 return;
 571         }
 572
 573         if (NULL == (ctrl = diskctrl(node, minor)))
 574                 return;
 575
 576         (void) strcpy(l_path, dir);
 577         (void) strcat(l_path, "/c");
 578         (void) strcat(l_path, ctrl);
 579         (void) strcat(l_path, disk);
 580
 581         /*
 582          * If switching between SMI and EFI label or vice versa
 583          * cleanup the previous label's devlinks.
 584          */
 585         if (*mn == *(MN_SMI) || (strncmp(mn, MN_EFI, 2) == 0)) {
 586                 char *s, tpath[PATH_MAX + 1];
 587                 struct stat sb;
 588
 589                 s = l_path + strlen(l_path);
 590                 (void) strcat(l_path, (*mn == *(MN_SMI))
 591                     ? SLICE_EFI : SLICE_SMI);
 592                 /*
 593                  * Attempt the remove only if the stale link exists
 594                  */
 595                 (void) snprintf(tpath, sizeof (tpath), "%s/dev/%s",
 596                     devfsadm_root_path(), l_path);
 597                 if (lstat(tpath, &sb) != -1)
 598                         devfsadm_rm_all(l_path);
 599                 *s = '\0';
 600         }
 601         (void) strcat(l_path, slice);
 602
 603         (void) devfsadm_mklink(l_path, node, minor, nflags);
 604
 605         /* secondary links for removable and hotpluggable devices */
 606         if (di_prop_lookup_ints(DDI_DEV_T_ANY, node, "removable-media",
 607             &int_prop) >= 0) {
 608                 (void) strcpy(sec_path, "removable-media/");
 609                 (void) strcat(sec_path, l_path);
 610                 (void) devfsadm_secondary_link(sec_path, l_path, 0);
 611         }
 612         if (di_prop_lookup_ints(DDI_DEV_T_ANY, node, "hotpluggable",
 613             &int_prop) >= 0) {
 614                 (void) strcpy(sec_path, "hotpluggable/");
 615                 (void) strcat(sec_path, l_path);
 616                 (void) devfsadm_secondary_link(sec_path, l_path, 0);
 617         }
 618
 619         if ((flags & RM_STALE) == RM_STALE) {
 620                 (void) strcpy(stale_re, "^");
 621                 (void) strcat(stale_re, dir);
 622                 (void) strcat(stale_re, "/c");
 623                 (void) strcat(stale_re, ctrl);
 624                 (void) strcat(stale_re, "t[0-9A-F]+d[0-9]+(s[0-9]+)?$");
 625                 /*
 626                  * optimizations are made inside of devfsadm_rm_stale_links
 627                  * instead of before calling the function, as it always
 628                  * needs to add the valid link to the cache.
 629                  */
 630                 devfsadm_rm_stale_links(stale_re, l_path, node, minor);
 631         }
 632
 633         free(ctrl);
 634 }
 635
 636
 637 /* index of enumeration rule applicable to this module */
 638 #define RULE_INDEX      0
 639
 640 static char *
 641 diskctrl(di_node_t node, di_minor_t minor)
 642 {
 643         char path[PATH_MAX + 1];
 644         char *devfspath;
 645         char *buf, *mn;
 646
 647         devfsadm_enumerate_t rules[3] = {
 648             {"^r?dsk$/^c([0-9]+)", 1, MATCH_PARENT},
 649             {"^cfg$/^c([0-9]+)$", 1, MATCH_ADDR},
 650             {"^scsi$/^.+$/^c([0-9]+)", 1, MATCH_PARENT}
 651         };
 652
 653         mn = di_minor_name(minor);
 654
 655         if ((devfspath = di_devfs_path(node)) == NULL) {
 656                 return (NULL);
 657         }
 658         (void) strcpy(path, devfspath);
 659         (void) strcat(path, ":");
 660         (void) strcat(path, mn);
 661         di_devfs_path_free(devfspath);
 662
 663         /*
 664          * Use controller component of disk path
 665          */
 666         if (disk_enumerate_int(path, RULE_INDEX, &buf, rules, 3) ==
 667             DEVFSADM_MULTIPLE) {
 668
 669                 /*
 670                  * We failed because there are multiple logical controller
 671                  * numbers for a single physical controller.  If we use node
 672                  * name also in the match it should fix this and only find one
 673                  * logical controller. (See 4045879).
 674                  * NOTE: Rules for controllers are not changed, as there is
 675                  * no unique controller number for them in this case.
 676                  *
 677                  * MATCH_UNCACHED flag is private to the "disks" and "sgen"
 678                  * modules. NOT to be used by other modules.
 679                  */
 680
 681                 rules[0].flags = MATCH_NODE | MATCH_UNCACHED; /* disks */
 682                 rules[2].flags = MATCH_NODE | MATCH_UNCACHED; /* generic scsi */
 683                 if (devfsadm_enumerate_int(path, RULE_INDEX, &buf, rules, 3)) {
 684                         return (NULL);
 685                 }
 686         }
 687
 688         return (buf);
 689 }
 690
 691 typedef struct dvlist {
 692         char *dv_link;
 693         struct dvlist *dv_next;
 694 } dvlist_t;
 695
 696 static void
 697 free_dvlist(dvlist_t **pp)
 698 {
 699         dvlist_t *entry;
 700
 701         while (*pp) {
 702                 entry = *pp;
 703                 *pp = entry->dv_next;
 704                 assert(entry->dv_link);
 705                 free(entry->dv_link);
 706                 free(entry);
 707         }
 708 }
 709 static int
 710 dvlink_cb(di_devlink_t devlink, void *arg)
 711 {
 712         char *path;
 713         char *can_path;
 714         dvlist_t **pp = (dvlist_t **)arg;
 715         dvlist_t *entry = NULL;
 716
 717         entry = calloc(1, sizeof (dvlist_t));
 718         if (entry == NULL) {
 719                 devfsadm_errprint("%s: calloc failed\n", modname);
 720                 goto error;
 721         }
 722
 723         path = (char *)di_devlink_path(devlink);
 724         assert(path);
 725         if (path == NULL) {
 726                 devfsadm_errprint("%s: di_devlink_path() returned NULL\n",
 727                     modname);
 728                 goto error;
 729         }
 730
 731         devfsadm_print(disk_mid, "%s: found link %s in reverse link cache\n",
 732             modname, path);
 733
 734         /*
 735          * Return linkname in canonical form i.e. without the
 736          * "/dev/" prefix
 737          */
 738         can_path = strstr(path, "/dev/");
 739         if (can_path == NULL) {
 740                 devfsadm_errprint("%s: devlink path %s has no /dev/\n",
 741                     modname, path);
 742                 goto error;
 743         }
 744
 745         entry->dv_link = s_strdup(can_path + strlen("/dev/"));
 746         entry->dv_next = *pp;
 747         *pp = entry;
 748
 749         return (DI_WALK_CONTINUE);
 750
 751 error:
 752         free(entry);
 753         free_dvlist(pp);
 754         *pp = NULL;
 755         return (DI_WALK_TERMINATE);
 756 }
 757
 758 /*
 759  * Returns success only if all goes well. If there is no matching reserved link
 760  * or if there is an error, we assume no match. It is better to err on the side
 761  * of caution by creating extra links than to miss out creating a required link.
 762  */
 763 static int
 764 reserved_links_exist(di_node_t node, di_minor_t minor, int nflags)
 765 {
 766         di_devlink_handle_t dvlink_cache = devfsadm_devlink_cache();
 767         char phys_path[PATH_MAX];
 768         char *minor_path;
 769         dvlist_t *head;
 770         dvlist_t *entry;
 771         char *s;
 772         char l[PATH_MAX];
 773         int switch_link = 0;
 774         char *mn = di_minor_name(minor);
 775
 776         if (dvlink_cache == NULL || mn == NULL) {
 777                 devfsadm_errprint("%s: No minor or devlink cache\n", modname);
 778                 return (DEVFSADM_FAILURE);
 779         }
 780
 781         if (!devfsadm_have_reserved()) {
 782                 devfsadm_print(disk_mid, "%s: No reserved links\n", modname);
 783                 return (DEVFSADM_FAILURE);
 784         }
 785
 786         minor_path = di_devfs_minor_path(minor);
 787         if (minor_path == NULL) {
 788                 devfsadm_errprint("%s: di_devfs_minor_path failed\n", modname);
 789                 return (DEVFSADM_FAILURE);
 790         }
 791
 792         (void) strlcpy(phys_path, minor_path, sizeof (phys_path));
 793
 794         di_devfs_path_free(minor_path);
 795
 796         head = NULL;
 797         (void) di_devlink_cache_walk(dvlink_cache, DISK_LINK_RE, phys_path,
 798             DI_PRIMARY_LINK, &head, dvlink_cb);
 799
 800         /*
 801          * We may be switching between EFI label and SMI label in which case
 802          * we only have minors of the other type.
 803          */
 804         if (head == NULL && (*mn == *(MN_SMI) ||
 805             (strncmp(mn, MN_EFI, 2) == 0))) {
 806                 devfsadm_print(disk_mid, "%s: No links for minor %s in /dev. "
 807                     "Trying another label\n", modname, mn);
 808                 s = strrchr(phys_path, ':');
 809                 if (s == NULL) {
 810                         devfsadm_errprint("%s: invalid minor path: %s\n",
 811                             modname, phys_path);
 812                         return (DEVFSADM_FAILURE);
 813                 }
 814                 (void) snprintf(s+1, sizeof (phys_path) - (s + 1 - phys_path),
 815                     "%s%s", *mn == *(MN_SMI) ? MN_EFI : MN_SMI,
 816                     strstr(s, ",raw") ? ",raw" : "");
 817                 (void) di_devlink_cache_walk(dvlink_cache, DISK_LINK_RE,
 818                     phys_path, DI_PRIMARY_LINK, &head, dvlink_cb);
 819         }
 820
 821         if (head == NULL) {
 822                 devfsadm_print(disk_mid, "%s: minor %s has no links in /dev\n",
 823                     modname, phys_path);
 824                 /* no links on disk */
 825                 return (DEVFSADM_FAILURE);
 826         }
 827
 828         /*
 829          * It suffices to use 1 link to this minor, since
 830          * we are matching with reserved IDs on the basis of
 831          * the controller number which will be the same for
 832          * all links to this minor.
 833          */
 834         if (!devfsadm_is_reserved(disks_re_array, head->dv_link)) {
 835                 /* not reserved links */
 836                 devfsadm_print(disk_mid, "%s: devlink %s and its minor "
 837                     "are NOT reserved\n", modname, head->dv_link);
 838                 free_dvlist(&head);
 839                 return (DEVFSADM_FAILURE);
 840         }
 841
 842         devfsadm_print(disk_mid, "%s: devlink %s and its minor are on "
 843             "reserved list\n", modname, head->dv_link);
 844
 845         /*
 846          * Switch between SMI and EFI labels if required
 847          */
 848         switch_link = 0;
 849         if (*mn == *(MN_SMI) || (strncmp(mn, MN_EFI, 2) == 0)) {
 850                 for (entry = head; entry; entry = entry->dv_next) {
 851                         s = strrchr(entry->dv_link, '/');
 852                         assert(s);
 853                         if (s == NULL) {
 854                                 devfsadm_errprint("%s: disk link %s has no "
 855                                     "directory\n", modname, entry->dv_link);
 856                                 continue;
 857                         }
 858                         if (*mn == *(MN_SMI) && strchr(s, 's') == NULL) {
 859                                 (void) snprintf(l, sizeof (l), "%s%s",
 860                                     entry->dv_link, SLICE_SMI);
 861                                 switch_link = 1;
 862                                 devfsadm_print(disk_mid, "%s: switching "
 863                                     "reserved link from EFI to SMI label. "
 864                                     "New link is %s\n", modname, l);
 865                         } else if (strncmp(mn, MN_EFI, 2) == 0 &&
 866                             (s = strchr(s, 's'))) {
 867                                 *s = '\0';
 868                                 (void) snprintf(l, sizeof (l), "%s",
 869                                     entry->dv_link);
 870                                 *s = 's';
 871                                 switch_link = 1;
 872                                 devfsadm_print(disk_mid, "%s: switching "
 873                                     "reserved link from SMI to EFI label. "
 874                                     "New link is %s\n", modname, l);
 875                         }
 876                         if (switch_link) {
 877                                 devfsadm_print(disk_mid, "%s: switching "
 878                                     "link: deleting %s and creating %s\n",
 879                                     modname, entry->dv_link, l);
 880                                 devfsadm_rm_link(entry->dv_link);
 881                                 (void) devfsadm_mklink(l, node, minor, nflags);
 882                         }
 883                 }
 884         }
 885         free_dvlist(&head);
 886
 887         /*
 888          * return SUCCESS to indicate that new links to this minor should not
 889          * be created so that only compatibility links to this minor remain.
 890          */
 891         return (DEVFSADM_SUCCESS);
 892 }