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 2010 Sun Microsystems, Inc.  All rights reserved.
  24  * Use is subject to license terms.
  25  *
  26  * Copyright (c) 2010, Intel Corporation.
  27  * All rights reserved.
  28  *
  29  * Copyright (c) 2019, Joyent, Inc.
  30  */
  31 
  32 /*
  33  * This file contains the functionality that mimics the boot operations
  34  * on SPARC systems or the old boot.bin/multiboot programs on x86 systems.
  35  * The x86 kernel now does everything on its own.
  36  */
  37 
  38 #include <sys/types.h>
  39 #include <sys/bootconf.h>
  40 #include <sys/bootsvcs.h>
  41 #include <sys/bootinfo.h>
  42 #include <sys/multiboot.h>
  43 #include <sys/multiboot2.h>
  44 #include <sys/multiboot2_impl.h>
  45 #include <sys/bootvfs.h>
  46 #include <sys/bootprops.h>
  47 #include <sys/varargs.h>
  48 #include <sys/param.h>
  49 #include <sys/machparam.h>
  50 #include <sys/machsystm.h>
  51 #include <sys/archsystm.h>
  52 #include <sys/boot_console.h>
  53 #include <sys/framebuffer.h>
  54 #include <sys/cmn_err.h>
  55 #include <sys/systm.h>
  56 #include <sys/promif.h>
  57 #include <sys/archsystm.h>
  58 #include <sys/x86_archext.h>
  59 #include <sys/kobj.h>
  60 #include <sys/privregs.h>
  61 #include <sys/sysmacros.h>
  62 #include <sys/ctype.h>
  63 #include <sys/fastboot.h>
  64 #ifdef __xpv
  65 #include <sys/hypervisor.h>
  66 #include <net/if.h>
  67 #endif
  68 #include <vm/kboot_mmu.h>
  69 #include <vm/hat_pte.h>
  70 #include <sys/kobj.h>
  71 #include <sys/kobj_lex.h>
  72 #include <sys/pci_cfgspace_impl.h>
  73 #include <sys/fastboot_impl.h>
  74 #include <sys/acpi/acconfig.h>
  75 #include <sys/acpi/acpi.h>
  76 #include <sys/ddipropdefs.h>      /* For DDI prop types */
  77 
  78 static int have_console = 0;    /* set once primitive console is initialized */
  79 static char *boot_args = "";
  80 
  81 /*
  82  * Debugging macros
  83  */
  84 static uint_t kbm_debug = 0;
  85 #define DBG_MSG(s)      { if (kbm_debug) bop_printf(NULL, "%s", s); }
  86 #define DBG(x)          { if (kbm_debug)                        \
  87         bop_printf(NULL, "%s is %" PRIx64 "\n", #x, (uint64_t)(x));     \
  88         }
  89 
  90 #define PUT_STRING(s) {                         \
  91         char *cp;                               \
  92         for (cp = (s); *cp; ++cp)               \
  93                 bcons_putchar(*cp);             \
  94         }
  95 
  96 /* callback to boot_fb to set shadow frame buffer */
  97 extern void boot_fb_shadow_init(bootops_t *);
  98 
  99 bootops_t bootop;       /* simple bootops we'll pass on to kernel */
 100 struct bsys_mem bm;
 101 
 102 /*
 103  * Boot info from "glue" code in low memory. xbootp is used by:
 104  *      do_bop_phys_alloc(), do_bsys_alloc() and boot_prop_finish().
 105  */
 106 static struct xboot_info *xbootp;
 107 static uintptr_t next_virt;     /* next available virtual address */
 108 static paddr_t next_phys;       /* next available physical address from dboot */
 109 static paddr_t high_phys = -(paddr_t)1; /* last used physical address */
 110 
 111 /*
 112  * buffer for vsnprintf for console I/O
 113  */
 114 #define BUFFERSIZE      512
 115 static char buffer[BUFFERSIZE];
 116 
 117 /*
 118  * stuff to store/report/manipulate boot property settings.
 119  */
 120 typedef struct bootprop {
 121         struct bootprop *bp_next;
 122         char *bp_name;
 123         int bp_flags;                   /* DDI prop type */
 124         uint_t bp_vlen;                 /* 0 for boolean */
 125         char *bp_value;
 126 } bootprop_t;
 127 
 128 static bootprop_t *bprops = NULL;
 129 static char *curr_page = NULL;          /* ptr to avail bprop memory */
 130 static int curr_space = 0;              /* amount of memory at curr_page */
 131 
 132 #ifdef __xpv
 133 start_info_t *xen_info;
 134 shared_info_t *HYPERVISOR_shared_info;
 135 #endif
 136 
 137 /*
 138  * some allocator statistics
 139  */
 140 static ulong_t total_bop_alloc_scratch = 0;
 141 static ulong_t total_bop_alloc_kernel = 0;
 142 
 143 static void build_firmware_properties(struct xboot_info *);
 144 
 145 static int early_allocation = 1;
 146 
 147 int force_fastreboot = 0;
 148 volatile int fastreboot_onpanic = 0;
 149 int post_fastreboot = 0;
 150 #ifdef  __xpv
 151 volatile int fastreboot_capable = 0;
 152 #else
 153 volatile int fastreboot_capable = 1;
 154 #endif
 155 
 156 /*
 157  * Information saved from current boot for fast reboot.
 158  * If the information size exceeds what we have allocated, fast reboot
 159  * will not be supported.
 160  */
 161 multiboot_info_t saved_mbi;
 162 mb_memory_map_t saved_mmap[FASTBOOT_SAVED_MMAP_COUNT];
 163 uint8_t saved_drives[FASTBOOT_SAVED_DRIVES_SIZE];
 164 char saved_cmdline[FASTBOOT_SAVED_CMDLINE_LEN];
 165 int saved_cmdline_len = 0;
 166 size_t saved_file_size[FASTBOOT_MAX_FILES_MAP];
 167 
 168 /*
 169  * Turn off fastreboot_onpanic to avoid panic loop.
 170  */
 171 char fastreboot_onpanic_cmdline[FASTBOOT_SAVED_CMDLINE_LEN];
 172 static const char fastreboot_onpanic_args[] = " -B fastreboot_onpanic=0";
 173 
 174 /*
 175  * Pointers to where System Resource Affinity Table (SRAT), System Locality
 176  * Information Table (SLIT) and Maximum System Capability Table (MSCT)
 177  * are mapped into virtual memory
 178  */
 179 ACPI_TABLE_SRAT *srat_ptr = NULL;
 180 ACPI_TABLE_SLIT *slit_ptr = NULL;
 181 ACPI_TABLE_MSCT *msct_ptr = NULL;
 182 
 183 /*
 184  * Arbitrary limit on number of localities we handle; if
 185  * this limit is raised to more than UINT16_MAX, make sure
 186  * process_slit() knows how to handle it.
 187  */
 188 #define SLIT_LOCALITIES_MAX     (4096)
 189 
 190 #define SLIT_NUM_PROPNAME       "acpi-slit-localities"
 191 #define SLIT_PROPNAME           "acpi-slit"
 192 
 193 /*
 194  * Allocate aligned physical memory at boot time. This allocator allocates
 195  * from the highest possible addresses. This avoids exhausting memory that
 196  * would be useful for DMA buffers.
 197  */
 198 paddr_t
 199 do_bop_phys_alloc(uint64_t size, uint64_t align)
 200 {
 201         paddr_t pa = 0;
 202         paddr_t start;
 203         paddr_t end;
 204         struct memlist  *ml = (struct memlist *)xbootp->bi_phys_install;
 205 
 206         /*
 207          * Be careful if high memory usage is limited in startup.c
 208          * Since there are holes in the low part of the physical address
 209          * space we can treat physmem as a pfn (not just a pgcnt) and
 210          * get a conservative upper limit.
 211          */
 212         if (physmem != 0 && high_phys > pfn_to_pa(physmem))
 213                 high_phys = pfn_to_pa(physmem);
 214 
 215         /*
 216          * find the highest available memory in physinstalled
 217          */
 218         size = P2ROUNDUP(size, align);
 219         for (; ml; ml = ml->ml_next) {
 220                 start = P2ROUNDUP(ml->ml_address, align);
 221                 end = P2ALIGN(ml->ml_address + ml->ml_size, align);
 222                 if (start < next_phys)
 223                         start = P2ROUNDUP(next_phys, align);
 224                 if (end > high_phys)
 225                         end = P2ALIGN(high_phys, align);
 226 
 227                 if (end <= start)
 228                         continue;
 229                 if (end - start < size)
 230                         continue;
 231 
 232                 /*
 233                  * Early allocations need to use low memory, since
 234                  * physmem might be further limited by bootenv.rc
 235                  */
 236                 if (early_allocation) {
 237                         if (pa == 0 || start < pa)
 238                                 pa = start;
 239                 } else {
 240                         if (end - size > pa)
 241                                 pa = end - size;
 242                 }
 243         }
 244         if (pa != 0) {
 245                 if (early_allocation)
 246                         next_phys = pa + size;
 247                 else
 248                         high_phys = pa;
 249                 return (pa);
 250         }
 251         bop_panic("do_bop_phys_alloc(0x%" PRIx64 ", 0x%" PRIx64
 252             ") Out of memory\n", size, align);
 253         /*NOTREACHED*/
 254 }
 255 
 256 uintptr_t
 257 alloc_vaddr(size_t size, paddr_t align)
 258 {
 259         uintptr_t rv;
 260 
 261         next_virt = P2ROUNDUP(next_virt, (uintptr_t)align);
 262         rv = (uintptr_t)next_virt;
 263         next_virt += size;
 264         return (rv);
 265 }
 266 
 267 /*
 268  * Allocate virtual memory. The size is always rounded up to a multiple
 269  * of base pagesize.
 270  */
 271 
 272 /*ARGSUSED*/
 273 static caddr_t
 274 do_bsys_alloc(bootops_t *bop, caddr_t virthint, size_t size, int align)
 275 {
 276         paddr_t a = align;      /* same type as pa for masking */
 277         uint_t pgsize;
 278         paddr_t pa;
 279         uintptr_t va;
 280         ssize_t s;              /* the aligned size */
 281         uint_t level;
 282         uint_t is_kernel = (virthint != 0);
 283 
 284         if (a < MMU_PAGESIZE)
 285                 a = MMU_PAGESIZE;
 286         else if (!ISP2(a))
 287                 prom_panic("do_bsys_alloc() incorrect alignment");
 288         size = P2ROUNDUP(size, MMU_PAGESIZE);
 289 
 290         /*
 291          * Use the next aligned virtual address if we weren't given one.
 292          */
 293         if (virthint == NULL) {
 294                 virthint = (caddr_t)alloc_vaddr(size, a);
 295                 total_bop_alloc_scratch += size;
 296         } else {
 297                 total_bop_alloc_kernel += size;
 298         }
 299 
 300         /*
 301          * allocate the physical memory
 302          */
 303         pa = do_bop_phys_alloc(size, a);
 304 
 305         /*
 306          * Add the mappings to the page tables, try large pages first.
 307          */
 308         va = (uintptr_t)virthint;
 309         s = size;
 310         level = 1;
 311         pgsize = xbootp->bi_use_pae ? TWO_MEG : FOUR_MEG;
 312         if (xbootp->bi_use_largepage && a == pgsize) {
 313                 while (IS_P2ALIGNED(pa, pgsize) && IS_P2ALIGNED(va, pgsize) &&
 314                     s >= pgsize) {
 315                         kbm_map(va, pa, level, is_kernel);
 316                         va += pgsize;
 317                         pa += pgsize;
 318                         s -= pgsize;
 319                 }
 320         }
 321 
 322         /*
 323          * Map remaining pages use small mappings
 324          */
 325         level = 0;
 326         pgsize = MMU_PAGESIZE;
 327         while (s > 0) {
 328                 kbm_map(va, pa, level, is_kernel);
 329                 va += pgsize;
 330                 pa += pgsize;
 331                 s -= pgsize;
 332         }
 333         return (virthint);
 334 }
 335 
 336 /*
 337  * Free virtual memory - we'll just ignore these.
 338  */
 339 /*ARGSUSED*/
 340 static void
 341 do_bsys_free(bootops_t *bop, caddr_t virt, size_t size)
 342 {
 343         bop_printf(NULL, "do_bsys_free(virt=0x%p, size=0x%lx) ignored\n",
 344             (void *)virt, size);
 345 }
 346 
 347 /*
 348  * Old interface
 349  */
 350 /*ARGSUSED*/
 351 static caddr_t
 352 do_bsys_ealloc(bootops_t *bop, caddr_t virthint, size_t size,
 353     int align, int flags)
 354 {
 355         prom_panic("unsupported call to BOP_EALLOC()\n");
 356         return (0);
 357 }
 358 
 359 
 360 static void
 361 bsetprop(int flags, char *name, int nlen, void *value, int vlen)
 362 {
 363         uint_t size;
 364         uint_t need_size;
 365         bootprop_t *b;
 366 
 367         /*
 368          * align the size to 16 byte boundary
 369          */
 370         size = sizeof (bootprop_t) + nlen + 1 + vlen;
 371         size = (size + 0xf) & ~0xf;
 372         if (size > curr_space) {
 373                 need_size = (size + (MMU_PAGEOFFSET)) & MMU_PAGEMASK;
 374                 curr_page = do_bsys_alloc(NULL, 0, need_size, MMU_PAGESIZE);
 375                 curr_space = need_size;
 376         }
 377 
 378         /*
 379          * use a bootprop_t at curr_page and link into list
 380          */
 381         b = (bootprop_t *)curr_page;
 382         curr_page += sizeof (bootprop_t);
 383         curr_space -=  sizeof (bootprop_t);
 384         b->bp_next = bprops;
 385         bprops = b;
 386 
 387         /*
 388          * follow by name and ending zero byte
 389          */
 390         b->bp_name = curr_page;
 391         bcopy(name, curr_page, nlen);
 392         curr_page += nlen;
 393         *curr_page++ = 0;
 394         curr_space -= nlen + 1;
 395 
 396         /*
 397          * set the property type
 398          */
 399         b->bp_flags = flags & DDI_PROP_TYPE_MASK;
 400 
 401         /*
 402          * copy in value, but no ending zero byte
 403          */
 404         b->bp_value = curr_page;
 405         b->bp_vlen = vlen;
 406         if (vlen > 0) {
 407                 bcopy(value, curr_page, vlen);
 408                 curr_page += vlen;
 409                 curr_space -= vlen;
 410         }
 411 
 412         /*
 413          * align new values of curr_page, curr_space
 414          */
 415         while (curr_space & 0xf) {
 416                 ++curr_page;
 417                 --curr_space;
 418         }
 419 }
 420 
 421 static void
 422 bsetprops(char *name, char *value)
 423 {
 424         bsetprop(DDI_PROP_TYPE_STRING, name, strlen(name),
 425             value, strlen(value) + 1);
 426 }
 427 
 428 static void
 429 bsetprop32(char *name, uint32_t value)
 430 {
 431         bsetprop(DDI_PROP_TYPE_INT, name, strlen(name),
 432             (void *)&value, sizeof (value));
 433 }
 434 
 435 static void
 436 bsetprop64(char *name, uint64_t value)
 437 {
 438         bsetprop(DDI_PROP_TYPE_INT64, name, strlen(name),
 439             (void *)&value, sizeof (value));
 440 }
 441 
 442 static void
 443 bsetpropsi(char *name, int value)
 444 {
 445         char prop_val[32];
 446 
 447         (void) snprintf(prop_val, sizeof (prop_val), "%d", value);
 448         bsetprops(name, prop_val);
 449 }
 450 
 451 /*
 452  * to find the type of the value associated with this name
 453  */
 454 /*ARGSUSED*/
 455 int
 456 do_bsys_getproptype(bootops_t *bop, const char *name)
 457 {
 458         bootprop_t *b;
 459 
 460         for (b = bprops; b != NULL; b = b->bp_next) {
 461                 if (strcmp(name, b->bp_name) != 0)
 462                         continue;
 463                 return (b->bp_flags);
 464         }
 465         return (-1);
 466 }
 467 
 468 /*
 469  * to find the size of the buffer to allocate
 470  */
 471 /*ARGSUSED*/
 472 int
 473 do_bsys_getproplen(bootops_t *bop, const char *name)
 474 {
 475         bootprop_t *b;
 476 
 477         for (b = bprops; b; b = b->bp_next) {
 478                 if (strcmp(name, b->bp_name) != 0)
 479                         continue;
 480                 return (b->bp_vlen);
 481         }
 482         return (-1);
 483 }
 484 
 485 /*
 486  * get the value associated with this name
 487  */
 488 /*ARGSUSED*/
 489 int
 490 do_bsys_getprop(bootops_t *bop, const char *name, void *value)
 491 {
 492         bootprop_t *b;
 493 
 494         for (b = bprops; b; b = b->bp_next) {
 495                 if (strcmp(name, b->bp_name) != 0)
 496                         continue;
 497                 bcopy(b->bp_value, value, b->bp_vlen);
 498                 return (0);
 499         }
 500         return (-1);
 501 }
 502 
 503 /*
 504  * get the name of the next property in succession from the standalone
 505  */
 506 /*ARGSUSED*/
 507 static char *
 508 do_bsys_nextprop(bootops_t *bop, char *name)
 509 {
 510         bootprop_t *b;
 511 
 512         /*
 513          * A null name is a special signal for the 1st boot property
 514          */
 515         if (name == NULL || strlen(name) == 0) {
 516                 if (bprops == NULL)
 517                         return (NULL);
 518                 return (bprops->bp_name);
 519         }
 520 
 521         for (b = bprops; b; b = b->bp_next) {
 522                 if (name != b->bp_name)
 523                         continue;
 524                 b = b->bp_next;
 525                 if (b == NULL)
 526                         return (NULL);
 527                 return (b->bp_name);
 528         }
 529         return (NULL);
 530 }
 531 
 532 /*
 533  * Parse numeric value from a string. Understands decimal, hex, octal, - and ~
 534  */
 535 static int
 536 parse_value(char *p, uint64_t *retval)
 537 {
 538         int adjust = 0;
 539         uint64_t tmp = 0;
 540         int digit;
 541         int radix = 10;
 542 
 543         *retval = 0;
 544         if (*p == '-' || *p == '~')
 545                 adjust = *p++;
 546 
 547         if (*p == '0') {
 548                 ++p;
 549                 if (*p == 0)
 550                         return (0);
 551                 if (*p == 'x' || *p == 'X') {
 552                         radix = 16;
 553                         ++p;
 554                 } else {
 555                         radix = 8;
 556                         ++p;
 557                 }
 558         }
 559         while (*p) {
 560                 if ('0' <= *p && *p <= '9')
 561                         digit = *p - '0';
 562                 else if ('a' <= *p && *p <= 'f')
 563                         digit = 10 + *p - 'a';
 564                 else if ('A' <= *p && *p <= 'F')
 565                         digit = 10 + *p - 'A';
 566                 else
 567                         return (-1);
 568                 if (digit >= radix)
 569                         return (-1);
 570                 tmp = tmp * radix + digit;
 571                 ++p;
 572         }
 573         if (adjust == '-')
 574                 tmp = -tmp;
 575         else if (adjust == '~')
 576                 tmp = ~tmp;
 577         *retval = tmp;
 578         return (0);
 579 }
 580 
 581 static boolean_t
 582 unprintable(char *value, int size)
 583 {
 584         int i;
 585 
 586         if (size <= 0 || value[0] == '\0')
 587                 return (B_TRUE);
 588 
 589         for (i = 0; i < size; i++) {
 590                 if (value[i] == '\0')
 591                         return (i != (size - 1));
 592 
 593                 if (!isprint(value[i]))
 594                         return (B_TRUE);
 595         }
 596         return (B_FALSE);
 597 }
 598 
 599 /*
 600  * Print out information about all boot properties.
 601  * buffer is pointer to pre-allocated space to be used as temporary
 602  * space for property values.
 603  */
 604 static void
 605 boot_prop_display(char *buffer)
 606 {
 607         char *name = "";
 608         int i, len, flags, *buf32;
 609         int64_t *buf64;
 610 
 611         bop_printf(NULL, "\nBoot properties:\n");
 612 
 613         while ((name = do_bsys_nextprop(NULL, name)) != NULL) {
 614                 bop_printf(NULL, "\t0x%p %s = ", (void *)name, name);
 615                 (void) do_bsys_getprop(NULL, name, buffer);
 616                 len = do_bsys_getproplen(NULL, name);
 617                 flags = do_bsys_getproptype(NULL, name);
 618                 bop_printf(NULL, "len=%d ", len);
 619 
 620                 switch (flags) {
 621                 case DDI_PROP_TYPE_INT:
 622                         len = len / sizeof (int);
 623                         buf32 = (int *)buffer;
 624                         for (i = 0; i < len; i++) {
 625                                 bop_printf(NULL, "%08x", buf32[i]);
 626                                 if (i < len - 1)
 627                                         bop_printf(NULL, ".");
 628                         }
 629                         break;
 630                 case DDI_PROP_TYPE_STRING:
 631                         bop_printf(NULL, "%s", buffer);
 632                         break;
 633                 case DDI_PROP_TYPE_INT64:
 634                         len = len / sizeof (int64_t);
 635                         buf64 = (int64_t *)buffer;
 636                         for (i = 0; i < len; i++) {
 637                                 bop_printf(NULL, "%016" PRIx64, buf64[i]);
 638                                 if (i < len - 1)
 639                                         bop_printf(NULL, ".");
 640                         }
 641                         break;
 642                 default:
 643                         if (!unprintable(buffer, len)) {
 644                                 buffer[len] = 0;
 645                                 bop_printf(NULL, "%s", buffer);
 646                                 break;
 647                         }
 648                         for (i = 0; i < len; i++) {
 649                                 bop_printf(NULL, "%02x", buffer[i] & 0xff);
 650                                 if (i < len - 1)
 651                                         bop_printf(NULL, ".");
 652                         }
 653                         break;
 654                 }
 655                 bop_printf(NULL, "\n");
 656         }
 657 }
 658 
 659 /*
 660  * 2nd part of building the table of boot properties. This includes:
 661  * - values from /boot/solaris/bootenv.rc (ie. eeprom(1m) values)
 662  *
 663  * lines look like one of:
 664  * ^$
 665  * ^# comment till end of line
 666  * setprop name 'value'
 667  * setprop name value
 668  * setprop name "value"
 669  *
 670  * we do single character I/O since this is really just looking at memory
 671  */
 672 void
 673 boot_prop_finish(void)
 674 {
 675         int fd;
 676         char *line;
 677         int c;
 678         int bytes_read;
 679         char *name;
 680         int n_len;
 681         char *value;
 682         int v_len;
 683         char *inputdev; /* these override the command line if serial ports */
 684         char *outputdev;
 685         char *consoledev;
 686         uint64_t lvalue;
 687         int use_xencons = 0;
 688         extern int bootrd_debug;
 689 
 690 #ifdef __xpv
 691         if (!DOMAIN_IS_INITDOMAIN(xen_info))
 692                 use_xencons = 1;
 693 #endif /* __xpv */
 694 
 695         DBG_MSG("Opening /boot/solaris/bootenv.rc\n");
 696         fd = BRD_OPEN(bfs_ops, "/boot/solaris/bootenv.rc", 0);
 697         DBG(fd);
 698 
 699         line = do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, MMU_PAGESIZE);
 700         while (fd >= 0) {
 701 
 702                 /*
 703                  * get a line
 704                  */
 705                 for (c = 0; ; ++c) {
 706                         bytes_read = BRD_READ(bfs_ops, fd, line + c, 1);
 707                         if (bytes_read == 0) {
 708                                 if (c == 0)
 709                                         goto done;
 710                                 break;
 711                         }
 712                         if (line[c] == '\n')
 713                                 break;
 714                 }
 715                 line[c] = 0;
 716 
 717                 /*
 718                  * ignore comment lines
 719                  */
 720                 c = 0;
 721                 while (ISSPACE(line[c]))
 722                         ++c;
 723                 if (line[c] == '#' || line[c] == 0)
 724                         continue;
 725 
 726                 /*
 727                  * must have "setprop " or "setprop\t"
 728                  */
 729                 if (strncmp(line + c, "setprop ", 8) != 0 &&
 730                     strncmp(line + c, "setprop\t", 8) != 0)
 731                         continue;
 732                 c += 8;
 733                 while (ISSPACE(line[c]))
 734                         ++c;
 735                 if (line[c] == 0)
 736                         continue;
 737 
 738                 /*
 739                  * gather up the property name
 740                  */
 741                 name = line + c;
 742                 n_len = 0;
 743                 while (line[c] && !ISSPACE(line[c]))
 744                         ++n_len, ++c;
 745 
 746                 /*
 747                  * gather up the value, if any
 748                  */
 749                 value = "";
 750                 v_len = 0;
 751                 while (ISSPACE(line[c]))
 752                         ++c;
 753                 if (line[c] != 0) {
 754                         value = line + c;
 755                         while (line[c] && !ISSPACE(line[c]))
 756                                 ++v_len, ++c;
 757                 }
 758 
 759                 if (v_len >= 2 && value[0] == value[v_len - 1] &&
 760                     (value[0] == '\'' || value[0] == '"')) {
 761                         ++value;
 762                         v_len -= 2;
 763                 }
 764                 name[n_len] = 0;
 765                 if (v_len > 0)
 766                         value[v_len] = 0;
 767                 else
 768                         continue;
 769 
 770                 /*
 771                  * ignore "boot-file" property, it's now meaningless
 772                  */
 773                 if (strcmp(name, "boot-file") == 0)
 774                         continue;
 775                 if (strcmp(name, "boot-args") == 0 &&
 776                     strlen(boot_args) > 0)
 777                         continue;
 778 
 779                 /*
 780                  * If a property was explicitly set on the command line
 781                  * it will override a setting in bootenv.rc
 782                  */
 783                 if (do_bsys_getproplen(NULL, name) >= 0)
 784                         continue;
 785 
 786                 bsetprops(name, value);
 787         }
 788 done:
 789         if (fd >= 0)
 790                 (void) BRD_CLOSE(bfs_ops, fd);
 791 
 792         /*
 793          * Check if we have to limit the boot time allocator
 794          */
 795         if (do_bsys_getproplen(NULL, "physmem") != -1 &&
 796             do_bsys_getprop(NULL, "physmem", line) >= 0 &&
 797             parse_value(line, &lvalue) != -1) {
 798                 if (0 < lvalue && (lvalue < physmem || physmem == 0)) {
 799                         physmem = (pgcnt_t)lvalue;
 800                         DBG(physmem);
 801                 }
 802         }
 803         early_allocation = 0;
 804 
 805         /*
 806          * Check for bootrd_debug.
 807          */
 808         if (find_boot_prop("bootrd_debug"))
 809                 bootrd_debug = 1;
 810 
 811         /*
 812          * check to see if we have to override the default value of the console
 813          */
 814         if (!use_xencons) {
 815                 inputdev = line;
 816                 v_len = do_bsys_getproplen(NULL, "input-device");
 817                 if (v_len > 0)
 818                         (void) do_bsys_getprop(NULL, "input-device", inputdev);
 819                 else
 820                         v_len = 0;
 821                 inputdev[v_len] = 0;
 822 
 823                 outputdev = inputdev + v_len + 1;
 824                 v_len = do_bsys_getproplen(NULL, "output-device");
 825                 if (v_len > 0)
 826                         (void) do_bsys_getprop(NULL, "output-device",
 827                             outputdev);
 828                 else
 829                         v_len = 0;
 830                 outputdev[v_len] = 0;
 831 
 832                 consoledev = outputdev + v_len + 1;
 833                 v_len = do_bsys_getproplen(NULL, "console");
 834                 if (v_len > 0) {
 835                         (void) do_bsys_getprop(NULL, "console", consoledev);
 836                         if (post_fastreboot &&
 837                             strcmp(consoledev, "graphics") == 0) {
 838                                 bsetprops("console", "text");
 839                                 v_len = strlen("text");
 840                                 bcopy("text", consoledev, v_len);
 841                         }
 842                 } else {
 843                         v_len = 0;
 844                 }
 845                 consoledev[v_len] = 0;
 846                 bcons_init2(inputdev, outputdev, consoledev);
 847         } else {
 848                 /*
 849                  * Ensure console property exists
 850                  * If not create it as "hypervisor"
 851                  */
 852                 v_len = do_bsys_getproplen(NULL, "console");
 853                 if (v_len < 0)
 854                         bsetprops("console", "hypervisor");
 855                 inputdev = outputdev = consoledev = "hypervisor";
 856                 bcons_init2(inputdev, outputdev, consoledev);
 857         }
 858 
 859         if (find_boot_prop("prom_debug") || kbm_debug)
 860                 boot_prop_display(line);
 861 }
 862 
 863 /*
 864  * print formatted output
 865  */
 866 /*ARGSUSED*/
 867 void
 868 vbop_printf(void *ptr, const char *fmt, va_list ap)
 869 {
 870         if (have_console == 0)
 871                 return;
 872 
 873         (void) vsnprintf(buffer, BUFFERSIZE, fmt, ap);
 874         PUT_STRING(buffer);
 875 }
 876 
 877 /*PRINTFLIKE2*/
 878 void
 879 bop_printf(void *bop, const char *fmt, ...)
 880 {
 881         va_list ap;
 882 
 883         va_start(ap, fmt);
 884         vbop_printf(bop, fmt, ap);
 885         va_end(ap);
 886 }
 887 
 888 /*
 889  * Another panic() variant; this one can be used even earlier during boot than
 890  * prom_panic().
 891  */
 892 /*PRINTFLIKE1*/
 893 void
 894 bop_panic(const char *fmt, ...)
 895 {
 896         va_list ap;
 897 
 898         va_start(ap, fmt);
 899         bop_printf(NULL, fmt, ap);
 900         va_end(ap);
 901 
 902         bop_printf(NULL, "\nPress any key to reboot.\n");
 903         (void) bcons_getchar();
 904         bop_printf(NULL, "Resetting...\n");
 905         pc_reset();
 906 }
 907 
 908 /*
 909  * Do a real mode interrupt BIOS call
 910  */
 911 typedef struct bios_regs {
 912         unsigned short ax, bx, cx, dx, si, di, bp, es, ds;
 913 } bios_regs_t;
 914 typedef int (*bios_func_t)(int, bios_regs_t *);
 915 
 916 /*ARGSUSED*/
 917 static void
 918 do_bsys_doint(bootops_t *bop, int intnum, struct bop_regs *rp)
 919 {
 920 #if defined(__xpv)
 921         prom_panic("unsupported call to BOP_DOINT()\n");
 922 #else   /* __xpv */
 923         static int firsttime = 1;
 924         bios_func_t bios_func = (bios_func_t)(void *)(uintptr_t)0x5000;
 925         bios_regs_t br;
 926 
 927         /*
 928          * We're about to disable paging; we shouldn't be PCID enabled.
 929          */
 930         if (getcr4() & CR4_PCIDE)
 931                 prom_panic("do_bsys_doint() with PCID enabled\n");
 932 
 933         /*
 934          * The first time we do this, we have to copy the pre-packaged
 935          * low memory bios call code image into place.
 936          */
 937         if (firsttime) {
 938                 extern char bios_image[];
 939                 extern uint32_t bios_size;
 940 
 941                 bcopy(bios_image, (void *)bios_func, bios_size);
 942                 firsttime = 0;
 943         }
 944 
 945         br.ax = rp->eax.word.ax;
 946         br.bx = rp->ebx.word.bx;
 947         br.cx = rp->ecx.word.cx;
 948         br.dx = rp->edx.word.dx;
 949         br.bp = rp->ebp.word.bp;
 950         br.si = rp->esi.word.si;
 951         br.di = rp->edi.word.di;
 952         br.ds = rp->ds;
 953         br.es = rp->es;
 954 
 955         DBG_MSG("Doing BIOS call...");
 956         DBG(br.ax);
 957         DBG(br.bx);
 958         DBG(br.dx);
 959         rp->eflags = bios_func(intnum, &br);
 960         DBG_MSG("done\n");
 961 
 962         rp->eax.word.ax = br.ax;
 963         rp->ebx.word.bx = br.bx;
 964         rp->ecx.word.cx = br.cx;
 965         rp->edx.word.dx = br.dx;
 966         rp->ebp.word.bp = br.bp;
 967         rp->esi.word.si = br.si;
 968         rp->edi.word.di = br.di;
 969         rp->ds = br.ds;
 970         rp->es = br.es;
 971 #endif /* __xpv */
 972 }
 973 
 974 static struct boot_syscalls bop_sysp = {
 975         bcons_getchar,
 976         bcons_putchar,
 977         bcons_ischar,
 978 };
 979 
 980 static char *whoami;
 981 
 982 #define BUFLEN  64
 983 
 984 #if defined(__xpv)
 985 
 986 static char namebuf[32];
 987 
 988 static void
 989 xen_parse_props(char *s, char *prop_map[], int n_prop)
 990 {
 991         char **prop_name = prop_map;
 992         char *cp = s, *scp;
 993 
 994         do {
 995                 scp = cp;
 996                 while ((*cp != NULL) && (*cp != ':'))
 997                         cp++;
 998 
 999                 if ((scp != cp) && (*prop_name != NULL)) {
1000                         *cp = NULL;
1001                         bsetprops(*prop_name, scp);
1002                 }
1003 
1004                 cp++;
1005                 prop_name++;
1006                 n_prop--;
1007         } while (n_prop > 0);
1008 }
1009 
1010 #define VBDPATHLEN      64
1011 
1012 /*
1013  * parse the 'xpv-root' property to create properties used by
1014  * ufs_mountroot.
1015  */
1016 static void
1017 xen_vbdroot_props(char *s)
1018 {
1019         char vbdpath[VBDPATHLEN] = "/xpvd/xdf@";
1020         const char lnamefix[] = "/dev/dsk/c0d";
1021         char *pnp;
1022         char *prop_p;
1023         char mi;
1024         short minor;
1025         long addr = 0;
1026 
1027         pnp = vbdpath + strlen(vbdpath);
1028         prop_p = s + strlen(lnamefix);
1029         while ((*prop_p != '\0') && (*prop_p != 's') && (*prop_p != 'p'))
1030                 addr = addr * 10 + *prop_p++ - '0';
1031         (void) snprintf(pnp, VBDPATHLEN, "%lx", addr);
1032         pnp = vbdpath + strlen(vbdpath);
1033         if (*prop_p == 's')
1034                 mi = 'a';
1035         else if (*prop_p == 'p')
1036                 mi = 'q';
1037         else
1038                 ASSERT(0); /* shouldn't be here */
1039         prop_p++;
1040         ASSERT(*prop_p != '\0');
1041         if (ISDIGIT(*prop_p)) {
1042                 minor = *prop_p - '0';
1043                 prop_p++;
1044                 if (ISDIGIT(*prop_p)) {
1045                         minor = minor * 10 + *prop_p - '0';
1046                 }
1047         } else {
1048                 /* malformed root path, use 0 as default */
1049                 minor = 0;
1050         }
1051         ASSERT(minor < 16); /* at most 16 partitions */
1052         mi += minor;
1053         *pnp++ = ':';
1054         *pnp++ = mi;
1055         *pnp++ = '\0';
1056         bsetprops("fstype", "ufs");
1057         bsetprops("bootpath", vbdpath);
1058 
1059         DBG_MSG("VBD bootpath set to ");
1060         DBG_MSG(vbdpath);
1061         DBG_MSG("\n");
1062 }
1063 
1064 /*
1065  * parse the xpv-nfsroot property to create properties used by
1066  * nfs_mountroot.
1067  */
1068 static void
1069 xen_nfsroot_props(char *s)
1070 {
1071         char *prop_map[] = {
1072                 BP_SERVER_IP,   /* server IP address */
1073                 BP_SERVER_NAME, /* server hostname */
1074                 BP_SERVER_PATH, /* root path */
1075         };
1076         int n_prop = sizeof (prop_map) / sizeof (prop_map[0]);
1077 
1078         bsetprops("fstype", "nfs");
1079 
1080         xen_parse_props(s, prop_map, n_prop);
1081 
1082         /*
1083          * If a server name wasn't specified, use a default.
1084          */
1085         if (do_bsys_getproplen(NULL, BP_SERVER_NAME) == -1)
1086                 bsetprops(BP_SERVER_NAME, "unknown");
1087 }
1088 
1089 /*
1090  * Extract our IP address, etc. from the "xpv-ip" property.
1091  */
1092 static void
1093 xen_ip_props(char *s)
1094 {
1095         char *prop_map[] = {
1096                 BP_HOST_IP,             /* IP address */
1097                 NULL,                   /* NFS server IP address (ignored in */
1098                                         /* favour of xpv-nfsroot) */
1099                 BP_ROUTER_IP,           /* IP gateway */
1100                 BP_SUBNET_MASK,         /* IP subnet mask */
1101                 "xpv-hostname",         /* hostname (ignored) */
1102                 BP_NETWORK_INTERFACE,   /* interface name */
1103                 "xpv-hcp",              /* host configuration protocol */
1104         };
1105         int n_prop = sizeof (prop_map) / sizeof (prop_map[0]);
1106         char ifname[IFNAMSIZ];
1107 
1108         xen_parse_props(s, prop_map, n_prop);
1109 
1110         /*
1111          * A Linux dom0 administrator expects all interfaces to be
1112          * called "ethX", which is not the case here.
1113          *
1114          * If the interface name specified is "eth0", presume that
1115          * this is really intended to be "xnf0" (the first domU ->
1116          * dom0 interface for this domain).
1117          */
1118         if ((do_bsys_getprop(NULL, BP_NETWORK_INTERFACE, ifname) == 0) &&
1119             (strcmp("eth0", ifname) == 0)) {
1120                 bsetprops(BP_NETWORK_INTERFACE, "xnf0");
1121                 bop_printf(NULL,
1122                     "network interface name 'eth0' replaced with 'xnf0'\n");
1123         }
1124 }
1125 
1126 #else   /* __xpv */
1127 
1128 static void
1129 setup_rarp_props(struct sol_netinfo *sip)
1130 {
1131         char buf[BUFLEN];       /* to hold ip/mac addrs */
1132         uint8_t *val;
1133 
1134         val = (uint8_t *)&sip->sn_ciaddr;
1135         (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1136             val[0], val[1], val[2], val[3]);
1137         bsetprops(BP_HOST_IP, buf);
1138 
1139         val = (uint8_t *)&sip->sn_siaddr;
1140         (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1141             val[0], val[1], val[2], val[3]);
1142         bsetprops(BP_SERVER_IP, buf);
1143 
1144         if (sip->sn_giaddr != 0) {
1145                 val = (uint8_t *)&sip->sn_giaddr;
1146                 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1147                     val[0], val[1], val[2], val[3]);
1148                 bsetprops(BP_ROUTER_IP, buf);
1149         }
1150 
1151         if (sip->sn_netmask != 0) {
1152                 val = (uint8_t *)&sip->sn_netmask;
1153                 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1154                     val[0], val[1], val[2], val[3]);
1155                 bsetprops(BP_SUBNET_MASK, buf);
1156         }
1157 
1158         if (sip->sn_mactype != 4 || sip->sn_maclen != 6) {
1159                 bop_printf(NULL, "unsupported mac type %d, mac len %d\n",
1160                     sip->sn_mactype, sip->sn_maclen);
1161         } else {
1162                 val = sip->sn_macaddr;
1163                 (void) snprintf(buf, BUFLEN, "%x:%x:%x:%x:%x:%x",
1164                     val[0], val[1], val[2], val[3], val[4], val[5]);
1165                 bsetprops(BP_BOOT_MAC, buf);
1166         }
1167 }
1168 
1169 #endif  /* __xpv */
1170 
1171 static void
1172 build_panic_cmdline(const char *cmd, int cmdlen)
1173 {
1174         int proplen;
1175         size_t arglen;
1176 
1177         arglen = sizeof (fastreboot_onpanic_args);
1178         /*
1179          * If we allready have fastreboot-onpanic set to zero,
1180          * don't add them again.
1181          */
1182         if ((proplen = do_bsys_getproplen(NULL, FASTREBOOT_ONPANIC)) > 0 &&
1183             proplen <=  sizeof (fastreboot_onpanic_cmdline)) {
1184                 (void) do_bsys_getprop(NULL, FASTREBOOT_ONPANIC,
1185                     fastreboot_onpanic_cmdline);
1186                 if (FASTREBOOT_ONPANIC_NOTSET(fastreboot_onpanic_cmdline))
1187                         arglen = 1;
1188         }
1189 
1190         /*
1191          * construct fastreboot_onpanic_cmdline
1192          */
1193         if (cmdlen + arglen > sizeof (fastreboot_onpanic_cmdline)) {
1194                 DBG_MSG("Command line too long: clearing "
1195                     FASTREBOOT_ONPANIC "\n");
1196                 fastreboot_onpanic = 0;
1197         } else {
1198                 bcopy(cmd, fastreboot_onpanic_cmdline, cmdlen);
1199                 if (arglen != 1)
1200                         bcopy(fastreboot_onpanic_args,
1201                             fastreboot_onpanic_cmdline + cmdlen, arglen);
1202                 else
1203                         fastreboot_onpanic_cmdline[cmdlen] = 0;
1204         }
1205 }
1206 
1207 
1208 #ifndef __xpv
1209 /*
1210  * Construct boot command line for Fast Reboot. The saved_cmdline
1211  * is also reported by "eeprom bootcmd".
1212  */
1213 static void
1214 build_fastboot_cmdline(struct xboot_info *xbp)
1215 {
1216         saved_cmdline_len =  strlen(xbp->bi_cmdline) + 1;
1217         if (saved_cmdline_len > FASTBOOT_SAVED_CMDLINE_LEN) {
1218                 DBG(saved_cmdline_len);
1219                 DBG_MSG("Command line too long: clearing fastreboot_capable\n");
1220                 fastreboot_capable = 0;
1221         } else {
1222                 bcopy((void *)(xbp->bi_cmdline), (void *)saved_cmdline,
1223                     saved_cmdline_len);
1224                 saved_cmdline[saved_cmdline_len - 1] = '\0';
1225                 build_panic_cmdline(saved_cmdline, saved_cmdline_len - 1);
1226         }
1227 }
1228 
1229 /*
1230  * Save memory layout, disk drive information, unix and boot archive sizes for
1231  * Fast Reboot.
1232  */
1233 static void
1234 save_boot_info(struct xboot_info *xbi)
1235 {
1236         multiboot_info_t *mbi = xbi->bi_mb_info;
1237         struct boot_modules *modp;
1238         int i;
1239 
1240         bcopy(mbi, &saved_mbi, sizeof (multiboot_info_t));
1241         if (mbi->mmap_length > sizeof (saved_mmap)) {
1242                 DBG_MSG("mbi->mmap_length too big: clearing "
1243                     "fastreboot_capable\n");
1244                 fastreboot_capable = 0;
1245         } else {
1246                 bcopy((void *)(uintptr_t)mbi->mmap_addr, (void *)saved_mmap,
1247                     mbi->mmap_length);
1248         }
1249 
1250         if ((mbi->flags & MB_INFO_DRIVE_INFO) != 0) {
1251                 if (mbi->drives_length > sizeof (saved_drives)) {
1252                         DBG(mbi->drives_length);
1253                         DBG_MSG("mbi->drives_length too big: clearing "
1254                             "fastreboot_capable\n");
1255                         fastreboot_capable = 0;
1256                 } else {
1257                         bcopy((void *)(uintptr_t)mbi->drives_addr,
1258                             (void *)saved_drives, mbi->drives_length);
1259                 }
1260         } else {
1261                 saved_mbi.drives_length = 0;
1262                 saved_mbi.drives_addr = NULL;
1263         }
1264 
1265         /*
1266          * Current file sizes.  Used by fastboot.c to figure out how much
1267          * memory to reserve for panic reboot.
1268          * Use the module list from the dboot-constructed xboot_info
1269          * instead of the list referenced by the multiboot structure
1270          * because that structure may not be addressable now.
1271          */
1272         saved_file_size[FASTBOOT_NAME_UNIX] = FOUR_MEG - PAGESIZE;
1273         for (i = 0, modp = (struct boot_modules *)(uintptr_t)xbi->bi_modules;
1274             i < xbi->bi_module_cnt; i++, modp++) {
1275                 saved_file_size[FASTBOOT_NAME_BOOTARCHIVE] += modp->bm_size;
1276         }
1277 }
1278 #endif  /* __xpv */
1279 
1280 /*
1281  * Import boot environment module variables as properties, applying
1282  * blacklist filter for variables we know we will not use.
1283  *
1284  * Since the environment can be relatively large, containing many variables
1285  * used only for boot loader purposes, we will use a blacklist based filter.
1286  * To keep the blacklist from growing too large, we use prefix based filtering.
1287  * This is possible because in many cases, the loader variable names are
1288  * using a structured layout.
1289  *
1290  * We will not overwrite already set properties.
1291  *
1292  * Note that the menu items in particular can contain characters not
1293  * well-handled as bootparams, such as spaces, brackets, and the like, so that's
1294  * another reason.
1295  */
1296 static struct bop_blacklist {
1297         const char *bl_name;
1298         int bl_name_len;
1299 } bop_prop_blacklist[] = {
1300         { "ISADIR", sizeof ("ISADIR") },
1301         { "acpi", sizeof ("acpi") },
1302         { "autoboot_delay", sizeof ("autoboot_delay") },
1303         { "beansi_", sizeof ("beansi_") },
1304         { "beastie", sizeof ("beastie") },
1305         { "bemenu", sizeof ("bemenu") },
1306         { "boot.", sizeof ("boot.") },
1307         { "bootenv", sizeof ("bootenv") },
1308         { "currdev", sizeof ("currdev") },
1309         { "dhcp.", sizeof ("dhcp.") },
1310         { "interpret", sizeof ("interpret") },
1311         { "kernel", sizeof ("kernel") },
1312         { "loaddev", sizeof ("loaddev") },
1313         { "loader_", sizeof ("loader_") },
1314         { "mainansi_", sizeof ("mainansi_") },
1315         { "mainmenu_", sizeof ("mainmenu_") },
1316         { "maintoggled_", sizeof ("maintoggled_") },
1317         { "menu_timeout_command", sizeof ("menu_timeout_command") },
1318         { "menuset_", sizeof ("menuset_") },
1319         { "module_path", sizeof ("module_path") },
1320         { "nfs.", sizeof ("nfs.") },
1321         { "optionsansi_", sizeof ("optionsansi_") },
1322         { "optionsmenu_", sizeof ("optionsmenu_") },
1323         { "optionstoggled_", sizeof ("optionstoggled_") },
1324         { "pcibios", sizeof ("pcibios") },
1325         { "prompt", sizeof ("prompt") },
1326         { "smbios", sizeof ("smbios") },
1327         { "tem", sizeof ("tem") },
1328         { "twiddle_divisor", sizeof ("twiddle_divisor") },
1329         { "zfs_be", sizeof ("zfs_be") },
1330 };
1331 
1332 /*
1333  * Match the name against prefixes in above blacklist. If the match was
1334  * found, this name is blacklisted.
1335  */
1336 static boolean_t
1337 name_is_blacklisted(const char *name)
1338 {
1339         int i, n;
1340 
1341         n = sizeof (bop_prop_blacklist) / sizeof (bop_prop_blacklist[0]);
1342         for (i = 0; i < n; i++) {
1343                 if (strncmp(bop_prop_blacklist[i].bl_name, name,
1344                     bop_prop_blacklist[i].bl_name_len - 1) == 0) {
1345                         return (B_TRUE);
1346                 }
1347         }
1348         return (B_FALSE);
1349 }
1350 
1351 static void
1352 process_boot_environment(struct boot_modules *benv)
1353 {
1354         char *env, *ptr, *name, *value;
1355         uint32_t size, name_len, value_len;
1356 
1357         if (benv == NULL || benv->bm_type != BMT_ENV)
1358                 return;
1359         ptr = env = benv->bm_addr;
1360         size = benv->bm_size;
1361         do {
1362                 name = ptr;
1363                 /* find '=' */
1364                 while (*ptr != '=') {
1365                         ptr++;
1366                         if (ptr > env + size) /* Something is very wrong. */
1367                                 return;
1368                 }
1369                 name_len = ptr - name;
1370                 if (sizeof (buffer) <= name_len)
1371                         continue;
1372 
1373                 (void) strncpy(buffer, name, sizeof (buffer));
1374                 buffer[name_len] = '\0';
1375                 name = buffer;
1376 
1377                 value_len = 0;
1378                 value = ++ptr;
1379                 while ((uintptr_t)ptr - (uintptr_t)env < size) {
1380                         if (*ptr == '\0') {
1381                                 ptr++;
1382                                 value_len = (uintptr_t)ptr - (uintptr_t)env;
1383                                 break;
1384                         }
1385                         ptr++;
1386                 }
1387 
1388                 /* Did we reach the end of the module? */
1389                 if (value_len == 0)
1390                         return;
1391 
1392                 if (*value == '\0')
1393                         continue;
1394 
1395                 /* Is this property already set? */
1396                 if (do_bsys_getproplen(NULL, name) >= 0)
1397                         continue;
1398 
1399                 /* Translate netboot variables */
1400                 if (strcmp(name, "boot.netif.gateway") == 0) {
1401                         bsetprops(BP_ROUTER_IP, value);
1402                         continue;
1403                 }
1404                 if (strcmp(name, "boot.netif.hwaddr") == 0) {
1405                         bsetprops(BP_BOOT_MAC, value);
1406                         continue;
1407                 }
1408                 if (strcmp(name, "boot.netif.ip") == 0) {
1409                         bsetprops(BP_HOST_IP, value);
1410                         continue;
1411                 }
1412                 if (strcmp(name, "boot.netif.netmask") == 0) {
1413                         bsetprops(BP_SUBNET_MASK, value);
1414                         continue;
1415                 }
1416                 if (strcmp(name, "boot.netif.server") == 0) {
1417                         bsetprops(BP_SERVER_IP, value);
1418                         continue;
1419                 }
1420                 if (strcmp(name, "boot.netif.server") == 0) {
1421                         if (do_bsys_getproplen(NULL, BP_SERVER_IP) < 0)
1422                                 bsetprops(BP_SERVER_IP, value);
1423                         continue;
1424                 }
1425                 if (strcmp(name, "boot.nfsroot.server") == 0) {
1426                         if (do_bsys_getproplen(NULL, BP_SERVER_IP) < 0)
1427                                 bsetprops(BP_SERVER_IP, value);
1428                         continue;
1429                 }
1430                 if (strcmp(name, "boot.nfsroot.path") == 0) {
1431                         bsetprops(BP_SERVER_PATH, value);
1432                         continue;
1433                 }
1434 
1435                 if (name_is_blacklisted(name) == B_TRUE)
1436                         continue;
1437 
1438                 /* Create new property. */
1439                 bsetprops(name, value);
1440 
1441                 /* Avoid reading past the module end. */
1442                 if (size <= (uintptr_t)ptr - (uintptr_t)env)
1443                         return;
1444         } while (*ptr != '\0');
1445 }
1446 
1447 /*
1448  * 1st pass at building the table of boot properties. This includes:
1449  * - values set on the command line: -B a=x,b=y,c=z ....
1450  * - known values we just compute (ie. from xbp)
1451  * - values from /boot/solaris/bootenv.rc (ie. eeprom(1m) values)
1452  *
1453  * the grub command line looked like:
1454  * kernel boot-file [-B prop=value[,prop=value]...] [boot-args]
1455  *
1456  * whoami is the same as boot-file
1457  */
1458 static void
1459 build_boot_properties(struct xboot_info *xbp)
1460 {
1461         char *name;
1462         int name_len;
1463         char *value;
1464         int value_len;
1465         struct boot_modules *bm, *rdbm, *benv = NULL;
1466         char *propbuf;
1467         int quoted = 0;
1468         int boot_arg_len;
1469         uint_t i, midx;
1470         char modid[32];
1471 #ifndef __xpv
1472         static int stdout_val = 0;
1473         uchar_t boot_device;
1474         char str[3];
1475 #endif
1476 
1477         /*
1478          * These have to be done first, so that kobj_mount_root() works
1479          */
1480         DBG_MSG("Building boot properties\n");
1481         propbuf = do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, 0);
1482         DBG((uintptr_t)propbuf);
1483         if (xbp->bi_module_cnt > 0) {
1484                 bm = xbp->bi_modules;
1485                 rdbm = NULL;
1486                 for (midx = i = 0; i < xbp->bi_module_cnt; i++) {
1487                         if (bm[i].bm_type == BMT_ROOTFS) {
1488                                 rdbm = &bm[i];
1489                                 continue;
1490                         }
1491                         if (bm[i].bm_type == BMT_HASH || bm[i].bm_name == NULL)
1492                                 continue;
1493 
1494                         if (bm[i].bm_type == BMT_ENV) {
1495                                 if (benv == NULL)
1496                                         benv = &bm[i];
1497                                 else
1498                                         continue;
1499                         }
1500 
1501                         (void) snprintf(modid, sizeof (modid),
1502                             "module-name-%u", midx);
1503                         bsetprops(modid, (char *)bm[i].bm_name);
1504                         (void) snprintf(modid, sizeof (modid),
1505                             "module-addr-%u", midx);
1506                         bsetprop64(modid, (uint64_t)(uintptr_t)bm[i].bm_addr);
1507                         (void) snprintf(modid, sizeof (modid),
1508                             "module-size-%u", midx);
1509                         bsetprop64(modid, (uint64_t)bm[i].bm_size);
1510                         ++midx;
1511                 }
1512                 if (rdbm != NULL) {
1513                         bsetprop64("ramdisk_start",
1514                             (uint64_t)(uintptr_t)rdbm->bm_addr);
1515                         bsetprop64("ramdisk_end",
1516                             (uint64_t)(uintptr_t)rdbm->bm_addr + rdbm->bm_size);
1517                 }
1518         }
1519 
1520         /*
1521          * If there are any boot time modules or hashes present, then disable
1522          * fast reboot.
1523          */
1524         if (xbp->bi_module_cnt > 1) {
1525                 fastreboot_disable(FBNS_BOOTMOD);
1526         }
1527 
1528 #ifndef __xpv
1529         /*
1530          * Disable fast reboot if we're using the Multiboot 2 boot protocol,
1531          * since we don't currently support MB2 info and module relocation.
1532          * Note that fast reboot will have already been disabled if multiple
1533          * modules are present, since the current implementation assumes that
1534          * we only have a single module, the boot_archive.
1535          */
1536         if (xbp->bi_mb_version != 1) {
1537                 fastreboot_disable(FBNS_MULTIBOOT2);
1538         }
1539 #endif
1540 
1541         DBG_MSG("Parsing command line for boot properties\n");
1542         value = xbp->bi_cmdline;
1543 
1544         /*
1545          * allocate memory to collect boot_args into
1546          */
1547         boot_arg_len = strlen(xbp->bi_cmdline) + 1;
1548         boot_args = do_bsys_alloc(NULL, NULL, boot_arg_len, MMU_PAGESIZE);
1549         boot_args[0] = 0;
1550         boot_arg_len = 0;
1551 
1552 #ifdef __xpv
1553         /*
1554          * Xen puts a lot of device information in front of the kernel name
1555          * let's grab them and make them boot properties.  The first
1556          * string w/o an "=" in it will be the boot-file property.
1557          */
1558         (void) strcpy(namebuf, "xpv-");
1559         for (;;) {
1560                 /*
1561                  * get to next property
1562                  */
1563                 while (ISSPACE(*value))
1564                         ++value;
1565                 name = value;
1566                 /*
1567                  * look for an "="
1568                  */
1569                 while (*value && !ISSPACE(*value) && *value != '=') {
1570                         value++;
1571                 }
1572                 if (*value != '=') { /* no "=" in the property */
1573                         value = name;
1574                         break;
1575                 }
1576                 name_len = value - name;
1577                 value_len = 0;
1578                 /*
1579                  * skip over the "="
1580                  */
1581                 value++;
1582                 while (value[value_len] && !ISSPACE(value[value_len])) {
1583                         ++value_len;
1584                 }
1585                 /*
1586                  * build property name with "xpv-" prefix
1587                  */
1588                 if (name_len + 4 > 32) { /* skip if name too long */
1589                         value += value_len;
1590                         continue;
1591                 }
1592                 bcopy(name, &namebuf[4], name_len);
1593                 name_len += 4;
1594                 namebuf[name_len] = 0;
1595                 bcopy(value, propbuf, value_len);
1596                 propbuf[value_len] = 0;
1597                 bsetprops(namebuf, propbuf);
1598 
1599                 /*
1600                  * xpv-root is set to the logical disk name of the xen
1601                  * VBD when booting from a disk-based filesystem.
1602                  */
1603                 if (strcmp(namebuf, "xpv-root") == 0)
1604                         xen_vbdroot_props(propbuf);
1605                 /*
1606                  * While we're here, if we have a "xpv-nfsroot" property
1607                  * then we need to set "fstype" to "nfs" so we mount
1608                  * our root from the nfs server.  Also parse the xpv-nfsroot
1609                  * property to create the properties that nfs_mountroot will
1610                  * need to find the root and mount it.
1611                  */
1612                 if (strcmp(namebuf, "xpv-nfsroot") == 0)
1613                         xen_nfsroot_props(propbuf);
1614 
1615                 if (strcmp(namebuf, "xpv-ip") == 0)
1616                         xen_ip_props(propbuf);
1617                 value += value_len;
1618         }
1619 #endif
1620 
1621         while (ISSPACE(*value))
1622                 ++value;
1623         /*
1624          * value now points at the boot-file
1625          */
1626         value_len = 0;
1627         while (value[value_len] && !ISSPACE(value[value_len]))
1628                 ++value_len;
1629         if (value_len > 0) {
1630                 whoami = propbuf;
1631                 bcopy(value, whoami, value_len);
1632                 whoami[value_len] = 0;
1633                 bsetprops("boot-file", whoami);
1634                 /*
1635                  * strip leading path stuff from whoami, so running from
1636                  * PXE/miniroot makes sense.
1637                  */
1638                 if (strstr(whoami, "/platform/") != NULL)
1639                         whoami = strstr(whoami, "/platform/");
1640                 bsetprops("whoami", whoami);
1641         }
1642 
1643         /*
1644          * Values forcibly set boot properties on the command line via -B.
1645          * Allow use of quotes in values. Other stuff goes on kernel
1646          * command line.
1647          */
1648         name = value + value_len;
1649         while (*name != 0) {
1650                 /*
1651                  * anything not " -B" is copied to the command line
1652                  */
1653                 if (!ISSPACE(name[0]) || name[1] != '-' || name[2] != 'B') {
1654                         boot_args[boot_arg_len++] = *name;
1655                         boot_args[boot_arg_len] = 0;
1656                         ++name;
1657                         continue;
1658                 }
1659 
1660                 /*
1661                  * skip the " -B" and following white space
1662                  */
1663                 name += 3;
1664                 while (ISSPACE(*name))
1665                         ++name;
1666                 while (*name && !ISSPACE(*name)) {
1667                         value = strstr(name, "=");
1668                         if (value == NULL)
1669                                 break;
1670                         name_len = value - name;
1671                         ++value;
1672                         value_len = 0;
1673                         quoted = 0;
1674                         for (; ; ++value_len) {
1675                                 if (!value[value_len])
1676                                         break;
1677 
1678                                 /*
1679                                  * is this value quoted?
1680                                  */
1681                                 if (value_len == 0 &&
1682                                     (value[0] == '\'' || value[0] == '"')) {
1683                                         quoted = value[0];
1684                                         ++value_len;
1685                                 }
1686 
1687                                 /*
1688                                  * In the quote accept any character,
1689                                  * but look for ending quote.
1690                                  */
1691                                 if (quoted) {
1692                                         if (value[value_len] == quoted)
1693                                                 quoted = 0;
1694                                         continue;
1695                                 }
1696 
1697                                 /*
1698                                  * a comma or white space ends the value
1699                                  */
1700                                 if (value[value_len] == ',' ||
1701                                     ISSPACE(value[value_len]))
1702                                         break;
1703                         }
1704 
1705                         if (value_len == 0) {
1706                                 bsetprop(DDI_PROP_TYPE_ANY, name, name_len,
1707                                     NULL, 0);
1708                         } else {
1709                                 char *v = value;
1710                                 int l = value_len;
1711                                 if (v[0] == v[l - 1] &&
1712                                     (v[0] == '\'' || v[0] == '"')) {
1713                                         ++v;
1714                                         l -= 2;
1715                                 }
1716                                 bcopy(v, propbuf, l);
1717                                 propbuf[l] = '\0';
1718                                 bsetprop(DDI_PROP_TYPE_STRING, name, name_len,
1719                                     propbuf, l + 1);
1720                         }
1721                         name = value + value_len;
1722                         while (*name == ',')
1723                                 ++name;
1724                 }
1725         }
1726 
1727         /*
1728          * set boot-args property
1729          * 1275 name is bootargs, so set
1730          * that too
1731          */
1732         bsetprops("boot-args", boot_args);
1733         bsetprops("bootargs", boot_args);
1734 
1735         process_boot_environment(benv);
1736 
1737 #ifndef __xpv
1738         /*
1739          * Build boot command line for Fast Reboot
1740          */
1741         build_fastboot_cmdline(xbp);
1742 
1743         if (xbp->bi_mb_version == 1) {
1744                 multiboot_info_t *mbi = xbp->bi_mb_info;
1745                 int netboot;
1746                 struct sol_netinfo *sip;
1747 
1748                 /*
1749                  * set the BIOS boot device from GRUB
1750                  */
1751                 netboot = 0;
1752 
1753                 /*
1754                  * Save various boot information for Fast Reboot
1755                  */
1756                 save_boot_info(xbp);
1757 
1758                 if (mbi != NULL && mbi->flags & MB_INFO_BOOTDEV) {
1759                         boot_device = mbi->boot_device >> 24;
1760                         if (boot_device == 0x20)
1761                                 netboot++;
1762                         str[0] = (boot_device >> 4) + '0';
1763                         str[1] = (boot_device & 0xf) + '0';
1764                         str[2] = 0;
1765                         bsetprops("bios-boot-device", str);
1766                 } else {
1767                         netboot = 1;
1768                 }
1769 
1770                 /*
1771                  * In the netboot case, drives_info is overloaded with the
1772                  * dhcp ack. This is not multiboot compliant and requires
1773                  * special pxegrub!
1774                  */
1775                 if (netboot && mbi->drives_length != 0) {
1776                         sip = (struct sol_netinfo *)(uintptr_t)mbi->drives_addr;
1777                         if (sip->sn_infotype == SN_TYPE_BOOTP)
1778                                 bsetprop(DDI_PROP_TYPE_BYTE,
1779                                     "bootp-response",
1780                                     sizeof ("bootp-response"),
1781                                     (void *)(uintptr_t)mbi->drives_addr,
1782                                     mbi->drives_length);
1783                         else if (sip->sn_infotype == SN_TYPE_RARP)
1784                                 setup_rarp_props(sip);
1785                 }
1786         } else {
1787                 multiboot2_info_header_t *mbi = xbp->bi_mb_info;
1788                 multiboot_tag_bootdev_t *bootdev = NULL;
1789                 multiboot_tag_network_t *netdev = NULL;
1790 
1791                 if (mbi != NULL) {
1792                         bootdev = dboot_multiboot2_find_tag(mbi,
1793                             MULTIBOOT_TAG_TYPE_BOOTDEV);
1794                         netdev = dboot_multiboot2_find_tag(mbi,
1795                             MULTIBOOT_TAG_TYPE_NETWORK);
1796                 }
1797                 if (bootdev != NULL) {
1798                         DBG(bootdev->mb_biosdev);
1799                         boot_device = bootdev->mb_biosdev;
1800                         str[0] = (boot_device >> 4) + '0';
1801                         str[1] = (boot_device & 0xf) + '0';
1802                         str[2] = 0;
1803                         bsetprops("bios-boot-device", str);
1804                 }
1805                 if (netdev != NULL) {
1806                         bsetprop(DDI_PROP_TYPE_BYTE,
1807                             "bootp-response", sizeof ("bootp-response"),
1808                             (void *)(uintptr_t)netdev->mb_dhcpack,
1809                             netdev->mb_size -
1810                             sizeof (multiboot_tag_network_t));
1811                 }
1812         }
1813 
1814         bsetprop32("stdout", stdout_val);
1815 #endif /* __xpv */
1816 
1817         /*
1818          * more conjured up values for made up things....
1819          */
1820 #if defined(__xpv)
1821         bsetprops("mfg-name", "i86xpv");
1822         bsetprops("impl-arch-name", "i86xpv");
1823 #else
1824         bsetprops("mfg-name", "i86pc");
1825         bsetprops("impl-arch-name", "i86pc");
1826 #endif
1827 
1828         /*
1829          * Build firmware-provided system properties
1830          */
1831         build_firmware_properties(xbp);
1832 
1833         /*
1834          * XXPV
1835          *
1836          * Find out what these are:
1837          * - cpuid_feature_ecx_include
1838          * - cpuid_feature_ecx_exclude
1839          * - cpuid_feature_edx_include
1840          * - cpuid_feature_edx_exclude
1841          *
1842          * Find out what these are in multiboot:
1843          * - netdev-path
1844          * - fstype
1845          */
1846 }
1847 
1848 #ifdef __xpv
1849 /*
1850  * Under the Hypervisor, memory usable for DMA may be scarce. One
1851  * very likely large pool of DMA friendly memory is occupied by
1852  * the boot_archive, as it was loaded by grub into low MFNs.
1853  *
1854  * Here we free up that memory by copying the boot archive to what are
1855  * likely higher MFN pages and then swapping the mfn/pfn mappings.
1856  */
1857 #define PFN_2GIG        0x80000
1858 static void
1859 relocate_boot_archive(struct xboot_info *xbp)
1860 {
1861         mfn_t max_mfn = HYPERVISOR_memory_op(XENMEM_maximum_ram_page, NULL);
1862         struct boot_modules *bm = xbp->bi_modules;
1863         uintptr_t va;
1864         pfn_t va_pfn;
1865         mfn_t va_mfn;
1866         caddr_t copy;
1867         pfn_t copy_pfn;
1868         mfn_t copy_mfn;
1869         size_t  len;
1870         int slop;
1871         int total = 0;
1872         int relocated = 0;
1873         int mmu_update_return;
1874         mmu_update_t t[2];
1875         x86pte_t pte;
1876 
1877         /*
1878          * If all MFN's are below 2Gig, don't bother doing this.
1879          */
1880         if (max_mfn < PFN_2GIG)
1881                 return;
1882         if (xbp->bi_module_cnt < 1) {
1883                 DBG_MSG("no boot_archive!");
1884                 return;
1885         }
1886 
1887         DBG_MSG("moving boot_archive to high MFN memory\n");
1888         va = (uintptr_t)bm->bm_addr;
1889         len = bm->bm_size;
1890         slop = va & MMU_PAGEOFFSET;
1891         if (slop) {
1892                 va += MMU_PAGESIZE - slop;
1893                 len -= MMU_PAGESIZE - slop;
1894         }
1895         len = P2ALIGN(len, MMU_PAGESIZE);
1896 
1897         /*
1898          * Go through all boot_archive pages, swapping any low MFN pages
1899          * with memory at next_phys.
1900          */
1901         while (len != 0) {
1902                 ++total;
1903                 va_pfn = mmu_btop(va - ONE_GIG);
1904                 va_mfn = mfn_list[va_pfn];
1905                 if (mfn_list[va_pfn] < PFN_2GIG) {
1906                         copy = kbm_remap_window(next_phys, 1);
1907                         bcopy((void *)va, copy, MMU_PAGESIZE);
1908                         copy_pfn = mmu_btop(next_phys);
1909                         copy_mfn = mfn_list[copy_pfn];
1910 
1911                         pte = mfn_to_ma(copy_mfn) | PT_NOCONSIST | PT_VALID;
1912                         if (HYPERVISOR_update_va_mapping(va, pte,
1913                             UVMF_INVLPG | UVMF_LOCAL))
1914                                 bop_panic("relocate_boot_archive():  "
1915                                     "HYPERVISOR_update_va_mapping() failed");
1916 
1917                         mfn_list[va_pfn] = copy_mfn;
1918                         mfn_list[copy_pfn] = va_mfn;
1919 
1920                         t[0].ptr = mfn_to_ma(copy_mfn) | MMU_MACHPHYS_UPDATE;
1921                         t[0].val = va_pfn;
1922                         t[1].ptr = mfn_to_ma(va_mfn) | MMU_MACHPHYS_UPDATE;
1923                         t[1].val = copy_pfn;
1924                         if (HYPERVISOR_mmu_update(t, 2, &mmu_update_return,
1925                             DOMID_SELF) != 0 || mmu_update_return != 2)
1926                                 bop_panic("relocate_boot_archive():  "
1927                                     "HYPERVISOR_mmu_update() failed");
1928 
1929                         next_phys += MMU_PAGESIZE;
1930                         ++relocated;
1931                 }
1932                 len -= MMU_PAGESIZE;
1933                 va += MMU_PAGESIZE;
1934         }
1935         DBG_MSG("Relocated pages:\n");
1936         DBG(relocated);
1937         DBG_MSG("Out of total pages:\n");
1938         DBG(total);
1939 }
1940 #endif /* __xpv */
1941 
1942 #if !defined(__xpv)
1943 /*
1944  * simple description of a stack frame (args are 32 bit only currently)
1945  */
1946 typedef struct bop_frame {
1947         struct bop_frame *old_frame;
1948         pc_t retaddr;
1949         long arg[1];
1950 } bop_frame_t;
1951 
1952 void
1953 bop_traceback(bop_frame_t *frame)
1954 {
1955         pc_t pc;
1956         int cnt;
1957         char *ksym;
1958         ulong_t off;
1959 
1960         bop_printf(NULL, "Stack traceback:\n");
1961         for (cnt = 0; cnt < 30; ++cnt) {     /* up to 30 frames */
1962                 pc = frame->retaddr;
1963                 if (pc == 0)
1964                         break;
1965                 ksym = kobj_getsymname(pc, &off);
1966                 if (ksym)
1967                         bop_printf(NULL, "  %s+%lx", ksym, off);
1968                 else
1969                         bop_printf(NULL, "  0x%lx", pc);
1970 
1971                 frame = frame->old_frame;
1972                 if (frame == 0) {
1973                         bop_printf(NULL, "\n");
1974                         break;
1975                 }
1976                 bop_printf(NULL, "\n");
1977         }
1978 }
1979 
1980 struct trapframe {
1981         ulong_t error_code;     /* optional */
1982         ulong_t inst_ptr;
1983         ulong_t code_seg;
1984         ulong_t flags_reg;
1985         ulong_t stk_ptr;
1986         ulong_t stk_seg;
1987 };
1988 
1989 void
1990 bop_trap(ulong_t *tfp)
1991 {
1992         struct trapframe *tf = (struct trapframe *)tfp;
1993         bop_frame_t fakeframe;
1994         static int depth = 0;
1995 
1996         /*
1997          * Check for an infinite loop of traps.
1998          */
1999         if (++depth > 2)
2000                 bop_panic("Nested trap");
2001 
2002         bop_printf(NULL, "Unexpected trap\n");
2003 
2004         /*
2005          * adjust the tf for optional error_code by detecting the code selector
2006          */
2007         if (tf->code_seg != B64CODE_SEL)
2008                 tf = (struct trapframe *)(tfp - 1);
2009         else
2010                 bop_printf(NULL, "error code           0x%lx\n",
2011                     tf->error_code & 0xffffffff);
2012 
2013         bop_printf(NULL, "instruction pointer  0x%lx\n", tf->inst_ptr);
2014         bop_printf(NULL, "code segment         0x%lx\n", tf->code_seg & 0xffff);
2015         bop_printf(NULL, "flags register       0x%lx\n", tf->flags_reg);
2016         bop_printf(NULL, "return %%rsp          0x%lx\n", tf->stk_ptr);
2017         bop_printf(NULL, "return %%ss           0x%lx\n", tf->stk_seg & 0xffff);
2018 
2019         /* grab %[er]bp pushed by our code from the stack */
2020         fakeframe.old_frame = (bop_frame_t *)*(tfp - 3);
2021         fakeframe.retaddr = (pc_t)tf->inst_ptr;
2022         bop_printf(NULL, "Attempting stack backtrace:\n");
2023         bop_traceback(&fakeframe);
2024         bop_panic("unexpected trap in early boot");
2025 }
2026 
2027 extern void bop_trap_handler(void);
2028 
2029 static gate_desc_t *bop_idt;
2030 
2031 static desctbr_t bop_idt_info;
2032 
2033 /*
2034  * Install a temporary IDT that lets us catch errors in the boot time code.
2035  * We shouldn't get any faults at all while this is installed, so we'll
2036  * just generate a traceback and exit.
2037  */
2038 static void
2039 bop_idt_init(void)
2040 {
2041         int t;
2042 
2043         bop_idt = (gate_desc_t *)
2044             do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, MMU_PAGESIZE);
2045         bzero(bop_idt, MMU_PAGESIZE);
2046         for (t = 0; t < NIDT; ++t) {
2047                 /*
2048                  * Note that since boot runs without a TSS, the
2049                  * double fault handler cannot use an alternate stack (64-bit).
2050                  */
2051                 set_gatesegd(&bop_idt[t], &bop_trap_handler, B64CODE_SEL,
2052                     SDT_SYSIGT, TRP_KPL, 0);
2053         }
2054         bop_idt_info.dtr_limit = (NIDT * sizeof (gate_desc_t)) - 1;
2055         bop_idt_info.dtr_base = (uintptr_t)bop_idt;
2056         wr_idtr(&bop_idt_info);
2057 }
2058 #endif  /* !defined(__xpv) */
2059 
2060 /*
2061  * This is where we enter the kernel. It dummies up the boot_ops and
2062  * boot_syscalls vectors and jumps off to _kobj_boot()
2063  */
2064 void
2065 _start(struct xboot_info *xbp)
2066 {
2067         bootops_t *bops = &bootop;
2068         extern void _kobj_boot();
2069 
2070         /*
2071          * 1st off - initialize the console for any error messages
2072          */
2073         xbootp = xbp;
2074 #ifdef __xpv
2075         HYPERVISOR_shared_info = (void *)xbp->bi_shared_info;
2076         xen_info = xbp->bi_xen_start_info;
2077 #endif
2078 
2079 #ifndef __xpv
2080         if (*((uint32_t *)(FASTBOOT_SWTCH_PA + FASTBOOT_STACK_OFFSET)) ==
2081             FASTBOOT_MAGIC) {
2082                 post_fastreboot = 1;
2083                 *((uint32_t *)(FASTBOOT_SWTCH_PA + FASTBOOT_STACK_OFFSET)) = 0;
2084         }
2085 #endif
2086 
2087         bcons_init(xbp);
2088         have_console = 1;
2089 
2090         /*
2091          * enable debugging
2092          */
2093         if (find_boot_prop("kbm_debug") != NULL)
2094                 kbm_debug = 1;
2095 
2096         DBG_MSG("\n\n*** Entered Solaris in _start() cmdline is: ");
2097         DBG_MSG((char *)xbp->bi_cmdline);
2098         DBG_MSG("\n\n\n");
2099 
2100         /*
2101          * physavail is no longer used by startup
2102          */
2103         bm.physinstalled = xbp->bi_phys_install;
2104         bm.pcimem = xbp->bi_pcimem;
2105         bm.rsvdmem = xbp->bi_rsvdmem;
2106         bm.physavail = NULL;
2107 
2108         /*
2109          * initialize the boot time allocator
2110          */
2111         next_phys = xbp->bi_next_paddr;
2112         DBG(next_phys);
2113         next_virt = (uintptr_t)xbp->bi_next_vaddr;
2114         DBG(next_virt);
2115         DBG_MSG("Initializing boot time memory management...");
2116 #ifdef __xpv
2117         {
2118                 xen_platform_parameters_t p;
2119 
2120                 /* This call shouldn't fail, dboot already did it once. */
2121                 (void) HYPERVISOR_xen_version(XENVER_platform_parameters, &p);
2122                 mfn_to_pfn_mapping = (pfn_t *)(xen_virt_start = p.virt_start);
2123                 DBG(xen_virt_start);
2124         }
2125 #endif
2126         kbm_init(xbp);
2127         DBG_MSG("done\n");
2128 
2129         /*
2130          * Fill in the bootops vector
2131          */
2132         bops->bsys_version = BO_VERSION;
2133         bops->boot_mem = &bm;
2134         bops->bsys_alloc = do_bsys_alloc;
2135         bops->bsys_free = do_bsys_free;
2136         bops->bsys_getproplen = do_bsys_getproplen;
2137         bops->bsys_getprop = do_bsys_getprop;
2138         bops->bsys_nextprop = do_bsys_nextprop;
2139         bops->bsys_printf = bop_printf;
2140         bops->bsys_doint = do_bsys_doint;
2141 
2142         /*
2143          * BOP_EALLOC() is no longer needed
2144          */
2145         bops->bsys_ealloc = do_bsys_ealloc;
2146 
2147 #ifdef __xpv
2148         /*
2149          * On domain 0 we need to free up some physical memory that is
2150          * usable for DMA. Since GRUB loaded the boot_archive, it is
2151          * sitting in low MFN memory. We'll relocated the boot archive
2152          * pages to high PFN memory.
2153          */
2154         if (DOMAIN_IS_INITDOMAIN(xen_info))
2155                 relocate_boot_archive(xbp);
2156 #endif
2157 
2158 #ifndef __xpv
2159         /*
2160          * Install an IDT to catch early pagefaults (shouldn't have any).
2161          * Also needed for kmdb.
2162          */
2163         bop_idt_init();
2164 #endif
2165         /* Set up the shadow fb for framebuffer console */
2166         boot_fb_shadow_init(bops);
2167 
2168         /*
2169          * Start building the boot properties from the command line
2170          */
2171         DBG_MSG("Initializing boot properties:\n");
2172         build_boot_properties(xbp);
2173 
2174         if (find_boot_prop("prom_debug") || kbm_debug) {
2175                 char *value;
2176 
2177                 value = do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, MMU_PAGESIZE);
2178                 boot_prop_display(value);
2179         }
2180 
2181         /*
2182          * jump into krtld...
2183          */
2184         _kobj_boot(&bop_sysp, NULL, bops, NULL);
2185 }
2186 
2187 
2188 /*ARGSUSED*/
2189 static caddr_t
2190 no_more_alloc(bootops_t *bop, caddr_t virthint, size_t size, int align)
2191 {
2192         panic("Attempt to bsys_alloc() too late\n");
2193         return (NULL);
2194 }
2195 
2196 /*ARGSUSED*/
2197 static void
2198 no_more_free(bootops_t *bop, caddr_t virt, size_t size)
2199 {
2200         panic("Attempt to bsys_free() too late\n");
2201 }
2202 
2203 void
2204 bop_no_more_mem(void)
2205 {
2206         DBG(total_bop_alloc_scratch);
2207         DBG(total_bop_alloc_kernel);
2208         bootops->bsys_alloc = no_more_alloc;
2209         bootops->bsys_free = no_more_free;
2210 }
2211 
2212 
2213 /*
2214  * Set ACPI firmware properties
2215  */
2216 
2217 static caddr_t
2218 vmap_phys(size_t length, paddr_t pa)
2219 {
2220         paddr_t start, end;
2221         caddr_t va;
2222         size_t  len, page;
2223 
2224 #ifdef __xpv
2225         pa = pfn_to_pa(xen_assign_pfn(mmu_btop(pa))) | (pa & MMU_PAGEOFFSET);
2226 #endif
2227         start = P2ALIGN(pa, MMU_PAGESIZE);
2228         end = P2ROUNDUP(pa + length, MMU_PAGESIZE);
2229         len = end - start;
2230         va = (caddr_t)alloc_vaddr(len, MMU_PAGESIZE);
2231         for (page = 0; page < len; page += MMU_PAGESIZE)
2232                 kbm_map((uintptr_t)va + page, start + page, 0, 0);
2233         return (va + (pa & MMU_PAGEOFFSET));
2234 }
2235 
2236 static uint8_t
2237 checksum_table(uint8_t *tp, size_t len)
2238 {
2239         uint8_t sum = 0;
2240 
2241         while (len-- > 0)
2242                 sum += *tp++;
2243 
2244         return (sum);
2245 }
2246 
2247 static int
2248 valid_rsdp(ACPI_TABLE_RSDP *rp)
2249 {
2250 
2251         /* validate the V1.x checksum */
2252         if (checksum_table((uint8_t *)rp, ACPI_RSDP_CHECKSUM_LENGTH) != 0)
2253                 return (0);
2254 
2255         /* If pre-ACPI 2.0, this is a valid RSDP */
2256         if (rp->Revision < 2)
2257                 return (1);
2258 
2259         /* validate the V2.x checksum */
2260         if (checksum_table((uint8_t *)rp, ACPI_RSDP_XCHECKSUM_LENGTH) != 0)
2261                 return (0);
2262 
2263         return (1);
2264 }
2265 
2266 /*
2267  * Scan memory range for an RSDP;
2268  * see ACPI 3.0 Spec, 5.2.5.1
2269  */
2270 static ACPI_TABLE_RSDP *
2271 scan_rsdp(paddr_t start, paddr_t end)
2272 {
2273         ssize_t len  = end - start;
2274         caddr_t ptr;
2275 
2276         ptr = vmap_phys(len, start);
2277         while (len > 0) {
2278                 if (strncmp(ptr, ACPI_SIG_RSDP, strlen(ACPI_SIG_RSDP)) == 0 &&
2279                     valid_rsdp((ACPI_TABLE_RSDP *)ptr))
2280                         return ((ACPI_TABLE_RSDP *)ptr);
2281 
2282                 ptr += ACPI_RSDP_SCAN_STEP;
2283                 len -= ACPI_RSDP_SCAN_STEP;
2284         }
2285 
2286         return (NULL);
2287 }
2288 
2289 /*
2290  * Refer to ACPI 3.0 Spec, section 5.2.5.1 to understand this function
2291  */
2292 static ACPI_TABLE_RSDP *
2293 find_rsdp()
2294 {
2295         ACPI_TABLE_RSDP *rsdp;
2296         uint64_t rsdp_val = 0;
2297         uint16_t *ebda_seg;
2298         paddr_t  ebda_addr;
2299 
2300         /* check for "acpi-root-tab" property */
2301         if (do_bsys_getproplen(NULL, "acpi-root-tab") == sizeof (uint64_t)) {
2302                 (void) do_bsys_getprop(NULL, "acpi-root-tab", &rsdp_val);
2303                 if (rsdp_val != 0) {
2304                         rsdp = scan_rsdp(rsdp_val, rsdp_val + sizeof (*rsdp));
2305                         if (rsdp != NULL) {
2306                                 if (kbm_debug) {
2307                                         bop_printf(NULL,
2308                                             "Using RSDP from bootloader: "
2309                                             "0x%p\n", (void *)rsdp);
2310                                 }
2311                                 return (rsdp);
2312                         }
2313                 }
2314         }
2315 
2316         /*
2317          * Get the EBDA segment and scan the first 1K
2318          */
2319         ebda_seg = (uint16_t *)vmap_phys(sizeof (uint16_t),
2320             ACPI_EBDA_PTR_LOCATION);
2321         ebda_addr = *ebda_seg << 4;
2322         rsdp = scan_rsdp(ebda_addr, ebda_addr + ACPI_EBDA_WINDOW_SIZE);
2323         if (rsdp == NULL)
2324                 /* if EBDA doesn't contain RSDP, look in BIOS memory */
2325                 rsdp = scan_rsdp(ACPI_HI_RSDP_WINDOW_BASE,
2326                     ACPI_HI_RSDP_WINDOW_BASE + ACPI_HI_RSDP_WINDOW_SIZE);
2327         return (rsdp);
2328 }
2329 
2330 static ACPI_TABLE_HEADER *
2331 map_fw_table(paddr_t table_addr)
2332 {
2333         ACPI_TABLE_HEADER *tp;
2334         size_t len = MAX(sizeof (*tp), MMU_PAGESIZE);
2335 
2336         /*
2337          * Map at least a page; if the table is larger than this, remap it
2338          */
2339         tp = (ACPI_TABLE_HEADER *)vmap_phys(len, table_addr);
2340         if (tp->Length > len)
2341                 tp = (ACPI_TABLE_HEADER *)vmap_phys(tp->Length, table_addr);
2342         return (tp);
2343 }
2344 
2345 static ACPI_TABLE_HEADER *
2346 find_fw_table(char *signature)
2347 {
2348         static int revision = 0;
2349         static ACPI_TABLE_XSDT *xsdt;
2350         static int len;
2351         paddr_t xsdt_addr;
2352         ACPI_TABLE_RSDP *rsdp;
2353         ACPI_TABLE_HEADER *tp;
2354         paddr_t table_addr;
2355         int     n;
2356 
2357         if (strlen(signature) != ACPI_NAME_SIZE)
2358                 return (NULL);
2359 
2360         /*
2361          * Reading the ACPI 3.0 Spec, section 5.2.5.3 will help
2362          * understand this code.  If we haven't already found the RSDT/XSDT,
2363          * revision will be 0. Find the RSDP and check the revision
2364          * to find out whether to use the RSDT or XSDT.  If revision is
2365          * 0 or 1, use the RSDT and set internal revision to 1; if it is 2,
2366          * use the XSDT.  If the XSDT address is 0, though, fall back to
2367          * revision 1 and use the RSDT.
2368          */
2369         if (revision == 0) {
2370                 if ((rsdp = find_rsdp()) != NULL) {
2371                         revision = rsdp->Revision;
2372                         /*
2373                          * ACPI 6.0 states that current revision is 2
2374                          * from acpi_table_rsdp definition:
2375                          * Must be (0) for ACPI 1.0 or (2) for ACPI 2.0+
2376                          */
2377                         if (revision > 2)
2378                                 revision = 2;
2379                         switch (revision) {
2380                         case 2:
2381                                 /*
2382                                  * Use the XSDT unless BIOS is buggy and
2383                                  * claims to be rev 2 but has a null XSDT
2384                                  * address
2385                                  */
2386                                 xsdt_addr = rsdp->XsdtPhysicalAddress;
2387                                 if (xsdt_addr != 0)
2388                                         break;
2389                                 /* FALLTHROUGH */
2390                         case 0:
2391                                 /* treat RSDP rev 0 as revision 1 internally */
2392                                 revision = 1;
2393                                 /* FALLTHROUGH */
2394                         case 1:
2395                                 /* use the RSDT for rev 0/1 */
2396                                 xsdt_addr = rsdp->RsdtPhysicalAddress;
2397                                 break;
2398                         default:
2399                                 /* unknown revision */
2400                                 revision = 0;
2401                                 break;
2402                         }
2403                 }
2404                 if (revision == 0)
2405                         return (NULL);
2406 
2407                 /* cache the XSDT info */
2408                 xsdt = (ACPI_TABLE_XSDT *)map_fw_table(xsdt_addr);
2409                 len = (xsdt->Header.Length - sizeof (xsdt->Header)) /
2410                     ((revision == 1) ? sizeof (uint32_t) : sizeof (uint64_t));
2411         }
2412 
2413         /*
2414          * Scan the table headers looking for a signature match
2415          */
2416         for (n = 0; n < len; n++) {
2417                 ACPI_TABLE_RSDT *rsdt = (ACPI_TABLE_RSDT *)xsdt;
2418                 table_addr = (revision == 1) ? rsdt->TableOffsetEntry[n] :
2419                     xsdt->TableOffsetEntry[n];
2420 
2421                 if (table_addr == 0)
2422                         continue;
2423                 tp = map_fw_table(table_addr);
2424                 if (strncmp(tp->Signature, signature, ACPI_NAME_SIZE) == 0) {
2425                         return (tp);
2426                 }
2427         }
2428         return (NULL);
2429 }
2430 
2431 static void
2432 process_mcfg(ACPI_TABLE_MCFG *tp)
2433 {
2434         ACPI_MCFG_ALLOCATION *cfg_baap;
2435         char *cfg_baa_endp;
2436         int64_t ecfginfo[4];
2437 
2438         cfg_baap = (ACPI_MCFG_ALLOCATION *)((uintptr_t)tp + sizeof (*tp));
2439         cfg_baa_endp = ((char *)tp) + tp->Header.Length;
2440         while ((char *)cfg_baap < cfg_baa_endp) {
2441                 if (cfg_baap->Address != 0 && cfg_baap->PciSegment == 0) {
2442                         ecfginfo[0] = cfg_baap->Address;
2443                         ecfginfo[1] = cfg_baap->PciSegment;
2444                         ecfginfo[2] = cfg_baap->StartBusNumber;
2445                         ecfginfo[3] = cfg_baap->EndBusNumber;
2446                         bsetprop(DDI_PROP_TYPE_INT64,
2447                             MCFG_PROPNAME, strlen(MCFG_PROPNAME),
2448                             ecfginfo, sizeof (ecfginfo));
2449                         break;
2450                 }
2451                 cfg_baap++;
2452         }
2453 }
2454 
2455 #ifndef __xpv
2456 static void
2457 process_madt_entries(ACPI_TABLE_MADT *tp, uint32_t *cpu_countp,
2458     uint32_t *cpu_possible_countp, uint32_t *cpu_apicid_array)
2459 {
2460         ACPI_SUBTABLE_HEADER *item, *end;
2461         uint32_t cpu_count = 0;
2462         uint32_t cpu_possible_count = 0;
2463 
2464         /*
2465          * Determine number of CPUs and keep track of "final" APIC ID
2466          * for each CPU by walking through ACPI MADT processor list
2467          */
2468         end = (ACPI_SUBTABLE_HEADER *)(tp->Header.Length + (uintptr_t)tp);
2469         item = (ACPI_SUBTABLE_HEADER *)((uintptr_t)tp + sizeof (*tp));
2470 
2471         while (item < end) {
2472                 switch (item->Type) {
2473                 case ACPI_MADT_TYPE_LOCAL_APIC: {
2474                         ACPI_MADT_LOCAL_APIC *cpu =
2475                             (ACPI_MADT_LOCAL_APIC *) item;
2476 
2477                         if (cpu->LapicFlags & ACPI_MADT_ENABLED) {
2478                                 if (cpu_apicid_array != NULL)
2479                                         cpu_apicid_array[cpu_count] = cpu->Id;
2480                                 cpu_count++;
2481                         }
2482                         cpu_possible_count++;
2483                         break;
2484                 }
2485                 case ACPI_MADT_TYPE_LOCAL_X2APIC: {
2486                         ACPI_MADT_LOCAL_X2APIC *cpu =
2487                             (ACPI_MADT_LOCAL_X2APIC *) item;
2488 
2489                         if (cpu->LapicFlags & ACPI_MADT_ENABLED) {
2490                                 if (cpu_apicid_array != NULL)
2491                                         cpu_apicid_array[cpu_count] =
2492                                             cpu->LocalApicId;
2493                                 cpu_count++;
2494                         }
2495                         cpu_possible_count++;
2496                         break;
2497                 }
2498                 default:
2499                         if (kbm_debug)
2500                                 bop_printf(NULL, "MADT type %d\n", item->Type);
2501                         break;
2502                 }
2503 
2504                 item = (ACPI_SUBTABLE_HEADER *)((uintptr_t)item + item->Length);
2505         }
2506         if (cpu_countp)
2507                 *cpu_countp = cpu_count;
2508         if (cpu_possible_countp)
2509                 *cpu_possible_countp = cpu_possible_count;
2510 }
2511 
2512 static void
2513 process_madt(ACPI_TABLE_MADT *tp)
2514 {
2515         uint32_t cpu_count = 0;
2516         uint32_t cpu_possible_count = 0;
2517         uint32_t *cpu_apicid_array; /* x2APIC ID is 32bit! */
2518 
2519         if (tp != NULL) {
2520                 /* count cpu's */
2521                 process_madt_entries(tp, &cpu_count, &cpu_possible_count, NULL);
2522 
2523                 cpu_apicid_array = (uint32_t *)do_bsys_alloc(NULL, NULL,
2524                     cpu_count * sizeof (*cpu_apicid_array), MMU_PAGESIZE);
2525                 if (cpu_apicid_array == NULL)
2526                         bop_panic("Not enough memory for APIC ID array");
2527 
2528                 /* copy IDs */
2529                 process_madt_entries(tp, NULL, NULL, cpu_apicid_array);
2530 
2531                 /*
2532                  * Make boot property for array of "final" APIC IDs for each
2533                  * CPU
2534                  */
2535                 bsetprop(DDI_PROP_TYPE_INT,
2536                     BP_CPU_APICID_ARRAY, strlen(BP_CPU_APICID_ARRAY),
2537                     cpu_apicid_array, cpu_count * sizeof (*cpu_apicid_array));
2538         }
2539 
2540         /*
2541          * Check whether property plat-max-ncpus is already set.
2542          */
2543         if (do_bsys_getproplen(NULL, PLAT_MAX_NCPUS_NAME) < 0) {
2544                 /*
2545                  * Set plat-max-ncpus to number of maximum possible CPUs given
2546                  * in MADT if it hasn't been set.
2547                  * There's no formal way to detect max possible CPUs supported
2548                  * by platform according to ACPI spec3.0b. So current CPU
2549                  * hotplug implementation expects that all possible CPUs will
2550                  * have an entry in MADT table and set plat-max-ncpus to number
2551                  * of entries in MADT.
2552                  * With introducing of ACPI4.0, Maximum System Capability Table
2553                  * (MSCT) provides maximum number of CPUs supported by platform.
2554                  * If MSCT is unavailable, fall back to old way.
2555                  */
2556                 if (tp != NULL)
2557                         bsetpropsi(PLAT_MAX_NCPUS_NAME, cpu_possible_count);
2558         }
2559 
2560         /*
2561          * Set boot property boot-max-ncpus to number of CPUs existing at
2562          * boot time. boot-max-ncpus is mainly used for optimization.
2563          */
2564         if (tp != NULL)
2565                 bsetpropsi(BOOT_MAX_NCPUS_NAME, cpu_count);
2566 
2567         /*
2568          * User-set boot-ncpus overrides firmware count
2569          */
2570         if (do_bsys_getproplen(NULL, BOOT_NCPUS_NAME) >= 0)
2571                 return;
2572 
2573         /*
2574          * Set boot property boot-ncpus to number of active CPUs given in MADT
2575          * if it hasn't been set yet.
2576          */
2577         if (tp != NULL)
2578                 bsetpropsi(BOOT_NCPUS_NAME, cpu_count);
2579 }
2580 
2581 static void
2582 process_srat(ACPI_TABLE_SRAT *tp)
2583 {
2584         ACPI_SUBTABLE_HEADER *item, *end;
2585         int i;
2586         int proc_num, mem_num;
2587 #pragma pack(1)
2588         struct {
2589                 uint32_t domain;
2590                 uint32_t apic_id;
2591                 uint32_t sapic_id;
2592         } processor;
2593         struct {
2594                 uint32_t domain;
2595                 uint32_t x2apic_id;
2596         } x2apic;
2597         struct {
2598                 uint32_t domain;
2599                 uint64_t addr;
2600                 uint64_t length;
2601                 uint32_t flags;
2602         } memory;
2603 #pragma pack()
2604         char prop_name[30];
2605         uint64_t maxmem = 0;
2606 
2607         if (tp == NULL)
2608                 return;
2609 
2610         proc_num = mem_num = 0;
2611         end = (ACPI_SUBTABLE_HEADER *)(tp->Header.Length + (uintptr_t)tp);
2612         item = (ACPI_SUBTABLE_HEADER *)((uintptr_t)tp + sizeof (*tp));
2613         while (item < end) {
2614                 switch (item->Type) {
2615                 case ACPI_SRAT_TYPE_CPU_AFFINITY: {
2616                         ACPI_SRAT_CPU_AFFINITY *cpu =
2617                             (ACPI_SRAT_CPU_AFFINITY *) item;
2618 
2619                         if (!(cpu->Flags & ACPI_SRAT_CPU_ENABLED))
2620                                 break;
2621                         processor.domain = cpu->ProximityDomainLo;
2622                         for (i = 0; i < 3; i++)
2623                                 processor.domain +=
2624                                     cpu->ProximityDomainHi[i] << ((i + 1) * 8);
2625                         processor.apic_id = cpu->ApicId;
2626                         processor.sapic_id = cpu->LocalSapicEid;
2627                         (void) snprintf(prop_name, 30, "acpi-srat-processor-%d",
2628                             proc_num);
2629                         bsetprop(DDI_PROP_TYPE_INT,
2630                             prop_name, strlen(prop_name), &processor,
2631                             sizeof (processor));
2632                         proc_num++;
2633                         break;
2634                 }
2635                 case ACPI_SRAT_TYPE_MEMORY_AFFINITY: {
2636                         ACPI_SRAT_MEM_AFFINITY *mem =
2637                             (ACPI_SRAT_MEM_AFFINITY *)item;
2638 
2639                         if (!(mem->Flags & ACPI_SRAT_MEM_ENABLED))
2640                                 break;
2641                         memory.domain = mem->ProximityDomain;
2642                         memory.addr = mem->BaseAddress;
2643                         memory.length = mem->Length;
2644                         memory.flags = mem->Flags;
2645                         (void) snprintf(prop_name, 30, "acpi-srat-memory-%d",
2646                             mem_num);
2647                         bsetprop(DDI_PROP_TYPE_INT,
2648                             prop_name, strlen(prop_name), &memory,
2649                             sizeof (memory));
2650                         if ((mem->Flags & ACPI_SRAT_MEM_HOT_PLUGGABLE) &&
2651                             (memory.addr + memory.length > maxmem)) {
2652                                 maxmem = memory.addr + memory.length;
2653                         }
2654                         mem_num++;
2655                         break;
2656                 }
2657                 case ACPI_SRAT_TYPE_X2APIC_CPU_AFFINITY: {
2658                         ACPI_SRAT_X2APIC_CPU_AFFINITY *x2cpu =
2659                             (ACPI_SRAT_X2APIC_CPU_AFFINITY *) item;
2660 
2661                         if (!(x2cpu->Flags & ACPI_SRAT_CPU_ENABLED))
2662                                 break;
2663                         x2apic.domain = x2cpu->ProximityDomain;
2664                         x2apic.x2apic_id = x2cpu->ApicId;
2665                         (void) snprintf(prop_name, 30, "acpi-srat-processor-%d",
2666                             proc_num);
2667                         bsetprop(DDI_PROP_TYPE_INT,
2668                             prop_name, strlen(prop_name), &x2apic,
2669                             sizeof (x2apic));
2670                         proc_num++;
2671                         break;
2672                 }
2673                 default:
2674                         if (kbm_debug)
2675                                 bop_printf(NULL, "SRAT type %d\n", item->Type);
2676                         break;
2677                 }
2678 
2679                 item = (ACPI_SUBTABLE_HEADER *)
2680                     (item->Length + (uintptr_t)item);
2681         }
2682 
2683         /*
2684          * The maximum physical address calculated from the SRAT table is more
2685          * accurate than that calculated from the MSCT table.
2686          */
2687         if (maxmem != 0) {
2688                 plat_dr_physmax = btop(maxmem);
2689         }
2690 }
2691 
2692 static void
2693 process_slit(ACPI_TABLE_SLIT *tp)
2694 {
2695 
2696         /*
2697          * Check the number of localities; if it's too huge, we just
2698          * return and locality enumeration code will handle this later,
2699          * if possible.
2700          *
2701          * Note that the size of the table is the square of the
2702          * number of localities; if the number of localities exceeds
2703          * UINT16_MAX, the table size may overflow an int when being
2704          * passed to bsetprop() below.
2705          */
2706         if (tp->LocalityCount >= SLIT_LOCALITIES_MAX)
2707                 return;
2708 
2709         bsetprop64(SLIT_NUM_PROPNAME, tp->LocalityCount);
2710         bsetprop(DDI_PROP_TYPE_BYTE,
2711             SLIT_PROPNAME, strlen(SLIT_PROPNAME), &tp->Entry,
2712             tp->LocalityCount * tp->LocalityCount);
2713 }
2714 
2715 static ACPI_TABLE_MSCT *
2716 process_msct(ACPI_TABLE_MSCT *tp)
2717 {
2718         int last_seen = 0;
2719         int proc_num = 0;
2720         ACPI_MSCT_PROXIMITY *item, *end;
2721         extern uint64_t plat_dr_options;
2722 
2723         ASSERT(tp != NULL);
2724 
2725         end = (ACPI_MSCT_PROXIMITY *)(tp->Header.Length + (uintptr_t)tp);
2726         for (item = (void *)((uintptr_t)tp + tp->ProximityOffset);
2727             item < end;
2728             item = (void *)(item->Length + (uintptr_t)item)) {
2729                 /*
2730                  * Sanity check according to section 5.2.19.1 of ACPI 4.0.
2731                  * Revision     1
2732                  * Length       22
2733                  */
2734                 if (item->Revision != 1 || item->Length != 22) {
2735                         cmn_err(CE_CONT,
2736                             "?boot: unknown proximity domain structure in MSCT "
2737                             "with Revision(%d), Length(%d).\n",
2738                             (int)item->Revision, (int)item->Length);
2739                         return (NULL);
2740                 } else if (item->RangeStart > item->RangeEnd) {
2741                         cmn_err(CE_CONT,
2742                             "?boot: invalid proximity domain structure in MSCT "
2743                             "with RangeStart(%u), RangeEnd(%u).\n",
2744                             item->RangeStart, item->RangeEnd);
2745                         return (NULL);
2746                 } else if (item->RangeStart != last_seen) {
2747                         /*
2748                          * Items must be organized in ascending order of the
2749                          * proximity domain enumerations.
2750                          */
2751                         cmn_err(CE_CONT,
2752                             "?boot: invalid proximity domain structure in MSCT,"
2753                             " items are not orginized in ascending order.\n");
2754                         return (NULL);
2755                 }
2756 
2757                 /*
2758                  * If ProcessorCapacity is 0 then there would be no CPUs in this
2759                  * domain.
2760                  */
2761                 if (item->ProcessorCapacity != 0) {
2762                         proc_num += (item->RangeEnd - item->RangeStart + 1) *
2763                             item->ProcessorCapacity;
2764                 }
2765 
2766                 last_seen = item->RangeEnd - item->RangeStart + 1;
2767                 /*
2768                  * Break out if all proximity domains have been processed.
2769                  * Some BIOSes may have unused items at the end of MSCT table.
2770                  */
2771                 if (last_seen > tp->MaxProximityDomains) {
2772                         break;
2773                 }
2774         }
2775         if (last_seen != tp->MaxProximityDomains + 1) {
2776                 cmn_err(CE_CONT,
2777                     "?boot: invalid proximity domain structure in MSCT, "
2778                     "proximity domain count doesn't match.\n");
2779                 return (NULL);
2780         }
2781 
2782         /*
2783          * Set plat-max-ncpus property if it hasn't been set yet.
2784          */
2785         if (do_bsys_getproplen(NULL, PLAT_MAX_NCPUS_NAME) < 0) {
2786                 if (proc_num != 0) {
2787                         bsetpropsi(PLAT_MAX_NCPUS_NAME, proc_num);
2788                 }
2789         }
2790 
2791         /*
2792          * Use Maximum Physical Address from the MSCT table as upper limit for
2793          * memory hot-adding by default. It may be overridden by value from
2794          * the SRAT table or the "plat-dr-physmax" boot option.
2795          */
2796         plat_dr_physmax = btop(tp->MaxAddress + 1);
2797 
2798         /*
2799          * Existence of MSCT implies CPU/memory hotplug-capability for the
2800          * platform.
2801          */
2802         plat_dr_options |= PLAT_DR_FEATURE_CPU;
2803         plat_dr_options |= PLAT_DR_FEATURE_MEMORY;
2804 
2805         return (tp);
2806 }
2807 
2808 #else /* __xpv */
2809 static void
2810 enumerate_xen_cpus()
2811 {
2812         processorid_t   id, max_id;
2813 
2814         /*
2815          * User-set boot-ncpus overrides enumeration
2816          */
2817         if (do_bsys_getproplen(NULL, BOOT_NCPUS_NAME) >= 0)
2818                 return;
2819 
2820         /*
2821          * Probe every possible virtual CPU id and remember the
2822          * highest id present; the count of CPUs is one greater
2823          * than this.  This tacitly assumes at least cpu 0 is present.
2824          */
2825         max_id = 0;
2826         for (id = 0; id < MAX_VIRT_CPUS; id++)
2827                 if (HYPERVISOR_vcpu_op(VCPUOP_is_up, id, NULL) == 0)
2828                         max_id = id;
2829 
2830         bsetpropsi(BOOT_NCPUS_NAME, max_id+1);
2831 
2832 }
2833 #endif /* __xpv */
2834 
2835 /*ARGSUSED*/
2836 static void
2837 build_firmware_properties(struct xboot_info *xbp)
2838 {
2839         ACPI_TABLE_HEADER *tp = NULL;
2840 
2841 #ifndef __xpv
2842         if (xbp->bi_uefi_arch == XBI_UEFI_ARCH_64) {
2843                 bsetprops("efi-systype", "64");
2844                 bsetprop64("efi-systab",
2845                     (uint64_t)(uintptr_t)xbp->bi_uefi_systab);
2846                 if (kbm_debug)
2847                         bop_printf(NULL, "64-bit UEFI detected.\n");
2848         } else if (xbp->bi_uefi_arch == XBI_UEFI_ARCH_32) {
2849                 bsetprops("efi-systype", "32");
2850                 bsetprop64("efi-systab",
2851                     (uint64_t)(uintptr_t)xbp->bi_uefi_systab);
2852                 if (kbm_debug)
2853                         bop_printf(NULL, "32-bit UEFI detected.\n");
2854         }
2855 
2856         if (xbp->bi_acpi_rsdp != NULL) {
2857                 bsetprop64("acpi-root-tab",
2858                     (uint64_t)(uintptr_t)xbp->bi_acpi_rsdp);
2859         }
2860 
2861         if (xbp->bi_smbios != NULL) {
2862                 bsetprop64("smbios-address",
2863                     (uint64_t)(uintptr_t)xbp->bi_smbios);
2864         }
2865 
2866         if ((tp = find_fw_table(ACPI_SIG_MSCT)) != NULL)
2867                 msct_ptr = process_msct((ACPI_TABLE_MSCT *)tp);
2868         else
2869                 msct_ptr = NULL;
2870 
2871         if ((tp = find_fw_table(ACPI_SIG_MADT)) != NULL)
2872                 process_madt((ACPI_TABLE_MADT *)tp);
2873 
2874         if ((srat_ptr = (ACPI_TABLE_SRAT *)
2875             find_fw_table(ACPI_SIG_SRAT)) != NULL)
2876                 process_srat(srat_ptr);
2877 
2878         if (slit_ptr = (ACPI_TABLE_SLIT *)find_fw_table(ACPI_SIG_SLIT))
2879                 process_slit(slit_ptr);
2880 
2881         tp = find_fw_table(ACPI_SIG_MCFG);
2882 #else /* __xpv */
2883         enumerate_xen_cpus();
2884         if (DOMAIN_IS_INITDOMAIN(xen_info))
2885                 tp = find_fw_table(ACPI_SIG_MCFG);
2886 #endif /* __xpv */
2887         if (tp != NULL)
2888                 process_mcfg((ACPI_TABLE_MCFG *)tp);
2889 }
2890 
2891 /*
2892  * fake up a boot property for deferred early console output
2893  * this is used by both graphical boot and the (developer only)
2894  * USB serial console
2895  */
2896 void *
2897 defcons_init(size_t size)
2898 {
2899         static char *p = NULL;
2900 
2901         p = do_bsys_alloc(NULL, NULL, size, MMU_PAGESIZE);
2902         *p = 0;
2903         bsetprop32("deferred-console-buf", (uint32_t)((uintptr_t)&p));
2904         return (p);
2905 }
2906 
2907 /*ARGSUSED*/
2908 int
2909 boot_compinfo(int fd, struct compinfo *cbp)
2910 {
2911         cbp->iscmp = 0;
2912         cbp->blksize = MAXBSIZE;
2913         return (0);
2914 }
2915 
2916 #define BP_MAX_STRLEN   32
2917 
2918 /*
2919  * Get value for given boot property
2920  */
2921 int
2922 bootprop_getval(const char *prop_name, u_longlong_t *prop_value)
2923 {
2924         int             boot_prop_len;
2925         char            str[BP_MAX_STRLEN];
2926         u_longlong_t    value;
2927 
2928         boot_prop_len = BOP_GETPROPLEN(bootops, prop_name);
2929         if (boot_prop_len < 0 || boot_prop_len > sizeof (str) ||
2930             BOP_GETPROP(bootops, prop_name, str) < 0 ||
2931             kobj_getvalue(str, &value) == -1)
2932                 return (-1);
2933 
2934         if (prop_value)
2935                 *prop_value = value;
2936 
2937         return (0);
2938 }