osfmk/i386/i386_vm_init.c

   1 /*
   2  * Copyright (c) 2003-2012 Apple Inc. All rights reserved.
   3  *
   4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License. The rights granted to you under the License
  10  * may not be used to create, or enable the creation or redistribution of,
  11  * unlawful or unlicensed copies of an Apple operating system, or to
  12  * circumvent, violate, or enable the circumvention or violation of, any
  13  * terms of an Apple operating system software license agreement.
  14  *
  15  * Please obtain a copy of the License at
  16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
  17  *
  18  * The Original Code and all software distributed under the License are
  19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  23  * Please see the License for the specific language governing rights and
  24  * limitations under the License.
  25  *
  26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  27  */
  28 /*
  29  * @OSF_COPYRIGHT@
  30  */
  31 /*
  32  * Mach Operating System
  33  * Copyright (c) 1991,1990,1989, 1988 Carnegie Mellon University
  34  * All Rights Reserved.
  35  *
  36  * Permission to use, copy, modify and distribute this software and its
  37  * documentation is hereby granted, provided that both the copyright
  38  * notice and this permission notice appear in all copies of the
  39  * software, derivative works or modified versions, and any portions
  40  * thereof, and that both notices appear in supporting documentation.
  41  *
  42  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
  43  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
  44  * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
  45  *
  46  * Carnegie Mellon requests users of this software to return to
  47  *
  48  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
  49  *  School of Computer Science
  50  *  Carnegie Mellon University
  51  *  Pittsburgh PA 15213-3890
  52  *
  53  * any improvements or extensions that they make and grant Carnegie Mellon
  54  * the rights to redistribute these changes.
  55  */
  56
  57
  58 #include <mach/i386/vm_param.h>
  59
  60 #include <string.h>
  61 #include <mach/vm_param.h>
  62 #include <mach/vm_prot.h>
  63 #include <mach/machine.h>
  64 #include <mach/time_value.h>
  65 #include <kern/spl.h>
  66 #include <kern/assert.h>
  67 #include <kern/debug.h>
  68 #include <kern/misc_protos.h>
  69 #include <kern/cpu_data.h>
  70 #include <kern/processor.h>
  71 #include <vm/vm_page.h>
  72 #include <vm/pmap.h>
  73 #include <vm/vm_kern.h>
  74 #include <i386/pmap.h>
  75 #include <i386/misc_protos.h>
  76 #include <i386/cpuid.h>
  77 #include <mach/thread_status.h>
  78 #include <pexpert/i386/efi.h>
  79 #include <i386/i386_lowmem.h>
  80 #include <x86_64/lowglobals.h>
  81 #include <i386/pal_routines.h>
  82
  83 #include <mach-o/loader.h>
  84 #include <libkern/kernel_mach_header.h>
  85
  86
  87 vm_size_t       mem_size = 0;
  88 pmap_paddr_t    first_avail = 0;/* first after page tables */
  89
  90 uint64_t        max_mem;        /* Size of physical memory (bytes), adjusted by maxmem */
  91 uint64_t        mem_actual;
  92 uint64_t        sane_size = 0;  /* Memory size for defaults calculations */
  93
  94 /*
  95  * KASLR parameters
  96  */
  97 ppnum_t         vm_kernel_base_page;
  98 vm_offset_t     vm_kernel_base;
  99 vm_offset_t     vm_kernel_top;
 100 vm_offset_t     vm_kernel_stext;
 101 vm_offset_t     vm_kernel_etext;
 102 vm_offset_t     vm_kernel_slide;
 103 vm_offset_t     vm_kernel_slid_base;
 104 vm_offset_t     vm_kernel_slid_top;
 105 vm_offset_t vm_hib_base;
 106 vm_offset_t     vm_kext_base = VM_MIN_KERNEL_AND_KEXT_ADDRESS;
 107 vm_offset_t     vm_kext_top = VM_MIN_KERNEL_ADDRESS;
 108
 109 vm_offset_t vm_prelink_stext;
 110 vm_offset_t vm_prelink_etext;
 111 vm_offset_t vm_prelink_sinfo;
 112 vm_offset_t vm_prelink_einfo;
 113 vm_offset_t vm_slinkedit;
 114 vm_offset_t vm_elinkedit;
 115
 116 #define MAXLORESERVE    (32 * 1024 * 1024)
 117
 118 ppnum_t         max_ppnum = 0;
 119 ppnum_t         lowest_lo = 0;
 120 ppnum_t         lowest_hi = 0;
 121 ppnum_t         highest_hi = 0;
 122
 123 enum {PMAP_MAX_RESERVED_RANGES = 32};
 124 uint32_t pmap_reserved_pages_allocated = 0;
 125 uint32_t pmap_reserved_range_indices[PMAP_MAX_RESERVED_RANGES];
 126 uint32_t pmap_last_reserved_range_index = 0;
 127 uint32_t pmap_reserved_ranges = 0;
 128
 129 extern unsigned int bsd_mbuf_cluster_reserve(boolean_t *);
 130
 131 pmap_paddr_t     avail_start, avail_end;
 132 vm_offset_t     virtual_avail, virtual_end;
 133 static pmap_paddr_t     avail_remaining;
 134 vm_offset_t     static_memory_end = 0;
 135
 136 vm_offset_t     sHIB, eHIB, stext, etext, sdata, edata, end, sconst, econst;
 137
 138 /*
 139  * _mh_execute_header is the mach_header for the currently executing kernel
 140  */
 141 vm_offset_t segTEXTB; unsigned long segSizeTEXT;
 142 vm_offset_t segDATAB; unsigned long segSizeDATA;
 143 vm_offset_t segLINKB; unsigned long segSizeLINK;
 144 vm_offset_t segPRELINKTEXTB; unsigned long segSizePRELINKTEXT;
 145 vm_offset_t segPRELINKINFOB; unsigned long segSizePRELINKINFO;
 146 vm_offset_t segHIBB; unsigned long segSizeHIB;
 147 unsigned long segSizeConst;
 148
 149 static kernel_segment_command_t *segTEXT, *segDATA;
 150 static kernel_section_t *cursectTEXT, *lastsectTEXT;
 151 static kernel_segment_command_t *segCONST;
 152
 153 extern uint64_t firmware_Conventional_bytes;
 154 extern uint64_t firmware_RuntimeServices_bytes;
 155 extern uint64_t firmware_ACPIReclaim_bytes;
 156 extern uint64_t firmware_ACPINVS_bytes;
 157 extern uint64_t firmware_PalCode_bytes;
 158 extern uint64_t firmware_Reserved_bytes;
 159 extern uint64_t firmware_Unusable_bytes;
 160 extern uint64_t firmware_other_bytes;
 161 uint64_t firmware_MMIO_bytes;
 162
 163 /*
 164  * Linker magic to establish the highest address in the kernel.
 165  */
 166 extern void     *last_kernel_symbol;
 167
 168 boolean_t       memmap = FALSE;
 169 #if     DEBUG || DEVELOPMENT
 170 static void
 171 kprint_memmap(vm_offset_t maddr, unsigned int msize, unsigned int mcount) {
 172     unsigned int         i;
 173     unsigned int         j;
 174     pmap_memory_region_t *p = pmap_memory_regions;
 175     EfiMemoryRange       *mptr;
 176     addr64_t             region_start, region_end;
 177     addr64_t             efi_start, efi_end;
 178
 179     for (j = 0; j < pmap_memory_region_count; j++, p++) {
 180         kprintf("pmap region %d type %d base 0x%llx alloc_up 0x%llx alloc_down 0x%llx top 0x%llx\n",
 181             j, p->type,
 182             (addr64_t) p->base  << I386_PGSHIFT,
 183             (addr64_t) p->alloc_up << I386_PGSHIFT,
 184             (addr64_t) p->alloc_down << I386_PGSHIFT,
 185             (addr64_t) p->end   << I386_PGSHIFT);
 186         region_start = (addr64_t) p->base << I386_PGSHIFT;
 187         region_end = ((addr64_t) p->end << I386_PGSHIFT) - 1;
 188         mptr = (EfiMemoryRange *) maddr;
 189         for (i = 0;
 190              i < mcount;
 191              i++, mptr = (EfiMemoryRange *)(((vm_offset_t)mptr) + msize)) {
 192             if (mptr->Type != kEfiLoaderCode &&
 193                 mptr->Type != kEfiLoaderData &&
 194                 mptr->Type != kEfiBootServicesCode &&
 195                 mptr->Type != kEfiBootServicesData &&
 196                 mptr->Type != kEfiConventionalMemory) {
 197                 efi_start = (addr64_t)mptr->PhysicalStart;
 198                 efi_end = efi_start + ((vm_offset_t)mptr->NumberOfPages << I386_PGSHIFT) - 1;
 199                 if ((efi_start >= region_start && efi_start <= region_end) ||
 200                     (efi_end >= region_start && efi_end <= region_end)) {
 201                     kprintf(" *** Overlapping region with EFI runtime region %d\n", i);
 202                 }
 203             }
 204         }
 205     }
 206 }
 207 #define DPRINTF(x...)   do { if (memmap) kprintf(x); } while (0)
 208
 209 #else
 210
 211 static void
 212 kprint_memmap(vm_offset_t maddr, unsigned int msize, unsigned int mcount) {
 213 #pragma unused(maddr, msize, mcount)
 214 }
 215
 216 #define DPRINTF(x...)
 217 #endif /* DEBUG */
 218
 219 /*
 220  * Basic VM initialization.
 221  */
 222 void
 223 i386_vm_init(uint64_t   maxmem,
 224              boolean_t  IA32e,
 225              boot_args  *args)
 226 {
 227         pmap_memory_region_t *pmptr;
 228         pmap_memory_region_t *prev_pmptr;
 229         EfiMemoryRange *mptr;
 230         unsigned int mcount;
 231         unsigned int msize;
 232         vm_offset_t maddr;
 233         ppnum_t fap;
 234         unsigned int i;
 235         ppnum_t maxpg = 0;
 236         uint32_t pmap_type;
 237         uint32_t maxloreserve;
 238         uint32_t maxdmaaddr;
 239         uint32_t  mbuf_reserve = 0;
 240         boolean_t mbuf_override = FALSE;
 241         boolean_t coalescing_permitted;
 242         vm_kernel_base_page = i386_btop(args->kaddr);
 243         vm_offset_t base_address;
 244         vm_offset_t static_base_address;
 245
 246         PE_parse_boot_argn("memmap", &memmap, sizeof(memmap));
 247
 248         /*
 249          * Establish the KASLR parameters.
 250          */
 251         static_base_address = ml_static_ptovirt(KERNEL_BASE_OFFSET);
 252         base_address        = ml_static_ptovirt(args->kaddr);
 253         vm_kernel_slide     = base_address - static_base_address;
 254         if (args->kslide) {
 255                 kprintf("KASLR slide: 0x%016lx dynamic\n", vm_kernel_slide);
 256                 if (vm_kernel_slide != ((vm_offset_t)args->kslide))
 257                         panic("Kernel base inconsistent with slide - rebased?");
 258         } else {
 259                 /* No slide relative to on-disk symbols */
 260                 kprintf("KASLR slide: 0x%016lx static and ignored\n",
 261                         vm_kernel_slide);
 262                 vm_kernel_slide = 0;
 263         }
 264
 265         /*
 266          * Zero out local relocations to avoid confusing kxld.
 267          * TODO: might be better to move this code to OSKext::initialize
 268          */
 269         if (_mh_execute_header.flags & MH_PIE) {
 270                 struct load_command *loadcmd;
 271                 uint32_t cmd;
 272
 273                 loadcmd = (struct load_command *)((uintptr_t)&_mh_execute_header +
 274                                                   sizeof (_mh_execute_header));
 275
 276                 for (cmd = 0; cmd < _mh_execute_header.ncmds; cmd++) {
 277                         if (loadcmd->cmd == LC_DYSYMTAB) {
 278                                 struct dysymtab_command *dysymtab;
 279
 280                                 dysymtab = (struct dysymtab_command *)loadcmd;
 281                                 dysymtab->nlocrel = 0;
 282                                 dysymtab->locreloff = 0;
 283                                 kprintf("Hiding local relocations\n");
 284                                 break;
 285                         }
 286                         loadcmd = (struct load_command *)((uintptr_t)loadcmd + loadcmd->cmdsize);
 287                 }
 288         }
 289
 290         /*
 291          * Now retrieve addresses for end, edata, and etext
 292          * from MACH-O headers.
 293          */
 294         segTEXTB = (vm_offset_t) getsegdatafromheader(&_mh_execute_header,
 295                                         "__TEXT", &segSizeTEXT);
 296         segDATAB = (vm_offset_t) getsegdatafromheader(&_mh_execute_header,
 297                                         "__DATA", &segSizeDATA);
 298         segLINKB = (vm_offset_t) getsegdatafromheader(&_mh_execute_header,
 299                                         "__LINKEDIT", &segSizeLINK);
 300         segHIBB  = (vm_offset_t) getsegdatafromheader(&_mh_execute_header,
 301                                         "__HIB", &segSizeHIB);
 302         segPRELINKTEXTB = (vm_offset_t) getsegdatafromheader(&_mh_execute_header,
 303                                                 "__PRELINK_TEXT", &segSizePRELINKTEXT);
 304         segPRELINKINFOB = (vm_offset_t) getsegdatafromheader(&_mh_execute_header,
 305                                                 "__PRELINK_INFO", &segSizePRELINKINFO);
 306         segTEXT = getsegbynamefromheader(&_mh_execute_header,
 307                                         "__TEXT");
 308         segDATA = getsegbynamefromheader(&_mh_execute_header,
 309                                         "__DATA");
 310         segCONST = getsegbynamefromheader(&_mh_execute_header,
 311                                         "__CONST");
 312         cursectTEXT = lastsectTEXT = firstsect(segTEXT);
 313         /* Discover the last TEXT section within the TEXT segment */
 314         while ((cursectTEXT = nextsect(segTEXT, cursectTEXT)) != NULL) {
 315                 lastsectTEXT = cursectTEXT;
 316         }
 317
 318         sHIB  = segHIBB;
 319         eHIB  = segHIBB + segSizeHIB;
 320         vm_hib_base = sHIB;
 321         /* Zero-padded from ehib to stext if text is 2M-aligned */
 322         stext = segTEXTB;
 323         lowGlo.lgStext = stext;
 324         etext = (vm_offset_t) round_page_64(lastsectTEXT->addr + lastsectTEXT->size);
 325         /* Zero-padded from etext to sdata if text is 2M-aligned */
 326         sdata = segDATAB;
 327         edata = segDATAB + segSizeDATA;
 328
 329         sconst = segCONST->vmaddr;
 330         segSizeConst = segCONST->vmsize;
 331         econst = sconst + segSizeConst;
 332
 333         assert(((sconst|econst) & PAGE_MASK) == 0);
 334
 335         DPRINTF("segTEXTB    = %p\n", (void *) segTEXTB);
 336         DPRINTF("segDATAB    = %p\n", (void *) segDATAB);
 337         DPRINTF("segLINKB    = %p\n", (void *) segLINKB);
 338         DPRINTF("segHIBB     = %p\n", (void *) segHIBB);
 339         DPRINTF("segPRELINKTEXTB = %p\n", (void *) segPRELINKTEXTB);
 340         DPRINTF("segPRELINKINFOB = %p\n", (void *) segPRELINKINFOB);
 341         DPRINTF("sHIB        = %p\n", (void *) sHIB);
 342         DPRINTF("eHIB        = %p\n", (void *) eHIB);
 343         DPRINTF("stext       = %p\n", (void *) stext);
 344         DPRINTF("etext       = %p\n", (void *) etext);
 345         DPRINTF("sdata       = %p\n", (void *) sdata);
 346         DPRINTF("edata       = %p\n", (void *) edata);
 347         DPRINTF("sconst      = %p\n", (void *) sconst);
 348         DPRINTF("econst      = %p\n", (void *) econst);
 349         DPRINTF("kernel_top  = %p\n", (void *) &last_kernel_symbol);
 350
 351         vm_kernel_base  = sHIB;
 352         vm_kernel_top   = (vm_offset_t) &last_kernel_symbol;
 353         vm_kernel_stext = stext;
 354         vm_kernel_etext = etext;
 355         vm_prelink_stext = segPRELINKTEXTB;
 356         vm_prelink_etext = segPRELINKTEXTB + segSizePRELINKTEXT;
 357         vm_prelink_sinfo = segPRELINKINFOB;
 358         vm_prelink_einfo = segPRELINKINFOB + segSizePRELINKINFO;
 359         vm_slinkedit = segLINKB;
 360         vm_elinkedit = segLINKB + segSizeLINK;
 361         vm_kernel_slid_base = vm_kext_base + vm_kernel_slide;
 362         vm_kernel_slid_top = vm_prelink_einfo;
 363
 364         vm_set_page_size();
 365
 366         /*
 367          * Compute the memory size.
 368          */
 369
 370         avail_remaining = 0;
 371         avail_end = 0;
 372         pmptr = pmap_memory_regions;
 373         prev_pmptr = 0;
 374         pmap_memory_region_count = pmap_memory_region_current = 0;
 375         fap = (ppnum_t) i386_btop(first_avail);
 376
 377         maddr = ml_static_ptovirt((vm_offset_t)args->MemoryMap);
 378         mptr = (EfiMemoryRange *)maddr;
 379         if (args->MemoryMapDescriptorSize == 0)
 380                 panic("Invalid memory map descriptor size");
 381         msize = args->MemoryMapDescriptorSize;
 382         mcount = args->MemoryMapSize / msize;
 383
 384 #define FOURGIG 0x0000000100000000ULL
 385 #define ONEGIG  0x0000000040000000ULL
 386
 387         for (i = 0; i < mcount; i++, mptr = (EfiMemoryRange *)(((vm_offset_t)mptr) + msize)) {
 388                 ppnum_t base, top;
 389                 uint64_t region_bytes = 0;
 390
 391                 if (pmap_memory_region_count >= PMAP_MEMORY_REGIONS_SIZE) {
 392                         kprintf("WARNING: truncating memory region count at %d\n", pmap_memory_region_count);
 393                         break;
 394                 }
 395                 base = (ppnum_t) (mptr->PhysicalStart >> I386_PGSHIFT);
 396                 top = (ppnum_t) (((mptr->PhysicalStart) >> I386_PGSHIFT) + mptr->NumberOfPages - 1);
 397
 398                 if (base == 0) {
 399                         /*
 400                          * Avoid having to deal with the edge case of the
 401                          * very first possible physical page and the roll-over
 402                          * to -1; just ignore that page.
 403                          */
 404                         kprintf("WARNING: ignoring first page in [0x%llx:0x%llx]\n", (uint64_t) base, (uint64_t) top);
 405                         base++;
 406                 }
 407                 if (top + 1 == 0) {
 408                         /*
 409                          * Avoid having to deal with the edge case of the
 410                          * very last possible physical page and the roll-over
 411                          * to 0; just ignore that page.
 412                          */
 413                         kprintf("WARNING: ignoring last page in [0x%llx:0x%llx]\n", (uint64_t) base, (uint64_t) top);
 414                         top--;
 415                 }
 416                 if (top < base) {
 417                         /*
 418                          * That was the only page in that region, so
 419                          * ignore the whole region.
 420                          */
 421                         continue;
 422                 }
 423
 424 #if     MR_RSV_TEST
 425                 static uint32_t nmr = 0;
 426                 if ((base > 0x20000) && (nmr++ < 4))
 427                         mptr->Attribute |= EFI_MEMORY_KERN_RESERVED;
 428 #endif
 429                 region_bytes = (uint64_t)(mptr->NumberOfPages << I386_PGSHIFT);
 430                 pmap_type = mptr->Type;
 431
 432                 switch (mptr->Type) {
 433                 case kEfiLoaderCode:
 434                 case kEfiLoaderData:
 435                 case kEfiBootServicesCode:
 436                 case kEfiBootServicesData:
 437                 case kEfiConventionalMemory:
 438                         /*
 439                          * Consolidate usable memory types into one.
 440                          */
 441                         pmap_type = kEfiConventionalMemory;
 442                         sane_size += region_bytes;
 443                         firmware_Conventional_bytes += region_bytes;
 444                         break;
 445                         /*
 446                          * sane_size should reflect the total amount of physical
 447                          * RAM in the system, not just the amount that is
 448                          * available for the OS to use.
 449                          * We now get this value from SMBIOS tables
 450                          * rather than reverse engineering the memory map.
 451                          * But the legacy computation of "sane_size" is kept
 452                          * for diagnostic information.
 453                          */
 454
 455                 case kEfiRuntimeServicesCode:
 456                 case kEfiRuntimeServicesData:
 457                         firmware_RuntimeServices_bytes += region_bytes;
 458                         sane_size += region_bytes;
 459                         break;
 460                 case kEfiACPIReclaimMemory:
 461                         firmware_ACPIReclaim_bytes += region_bytes;
 462                         sane_size += region_bytes;
 463                         break;
 464                 case kEfiACPIMemoryNVS:
 465                         firmware_ACPINVS_bytes += region_bytes;
 466                         sane_size += region_bytes;
 467                         break;
 468                 case kEfiPalCode:
 469                         firmware_PalCode_bytes += region_bytes;
 470                         sane_size += region_bytes;
 471                         break;
 472
 473                 case kEfiReservedMemoryType:
 474                         firmware_Reserved_bytes += region_bytes;
 475                         break;
 476                 case kEfiUnusableMemory:
 477                         firmware_Unusable_bytes += region_bytes;
 478                         break;
 479                 case kEfiMemoryMappedIO:
 480                 case kEfiMemoryMappedIOPortSpace:
 481                         firmware_MMIO_bytes += region_bytes;
 482                         break;
 483                 default:
 484                         firmware_other_bytes += region_bytes;
 485                         break;
 486                 }
 487
 488                 DPRINTF("EFI region %d: type %u/%d, base 0x%x, top 0x%x %s\n",
 489                     i, mptr->Type, pmap_type, base, top,
 490                     (mptr->Attribute&EFI_MEMORY_KERN_RESERVED)? "RESERVED" :
 491                     (mptr->Attribute&EFI_MEMORY_RUNTIME)? "RUNTIME" : "");
 492
 493                 if (maxpg) {
 494                         if (base >= maxpg)
 495                                 break;
 496                         top = (top > maxpg) ? maxpg : top;
 497                 }
 498
 499                 /*
 500                  * handle each region
 501                  */
 502                 if ((mptr->Attribute & EFI_MEMORY_RUNTIME) == EFI_MEMORY_RUNTIME ||
 503                     pmap_type != kEfiConventionalMemory) {
 504                         prev_pmptr = 0;
 505                         continue;
 506                 } else {
 507                         /*
 508                          * Usable memory region
 509                          */
 510                         if (top < I386_LOWMEM_RESERVED ||
 511                             !pal_is_usable_memory(base, top)) {
 512                                 prev_pmptr = 0;
 513                                 continue;
 514                         }
 515                         /*
 516                          * A range may be marked with with the
 517                          * EFI_MEMORY_KERN_RESERVED attribute
 518                          * on some systems, to indicate that the range
 519                          * must not be made available to devices.
 520                          */
 521
 522                         if (mptr->Attribute & EFI_MEMORY_KERN_RESERVED) {
 523                                 if (++pmap_reserved_ranges > PMAP_MAX_RESERVED_RANGES) {
 524                                         panic("Too many reserved ranges %u\n", pmap_reserved_ranges);
 525                                 }
 526                         }
 527
 528                         if (top < fap) {
 529                                 /*
 530                                  * entire range below first_avail
 531                                  * salvage some low memory pages
 532                                  * we use some very low memory at startup
 533                                  * mark as already allocated here
 534                                  */
 535                                 if (base >= I386_LOWMEM_RESERVED)
 536                                         pmptr->base = base;
 537                                 else
 538                                         pmptr->base = I386_LOWMEM_RESERVED;
 539
 540                                 pmptr->end = top;
 541
 542
 543                                 if ((mptr->Attribute & EFI_MEMORY_KERN_RESERVED) &&
 544                                     (top < vm_kernel_base_page)) {
 545                                         pmptr->alloc_up = pmptr->base;
 546                                         pmptr->alloc_down = pmptr->end;
 547                                         pmap_reserved_range_indices[pmap_last_reserved_range_index++] = pmap_memory_region_count;
 548                                 }
 549                                 else {
 550                                         /*
 551                                          * mark as already mapped
 552                                          */
 553                                         pmptr->alloc_up = top + 1;
 554                                         pmptr->alloc_down = top;
 555                                 }
 556                                 pmptr->type = pmap_type;
 557                                 pmptr->attribute = mptr->Attribute;
 558                         }
 559                         else if ( (base < fap) && (top > fap) ) {
 560                                 /*
 561                                  * spans first_avail
 562                                  * put mem below first avail in table but
 563                                  * mark already allocated
 564                                  */
 565                                 pmptr->base = base;
 566                                 pmptr->end = (fap - 1);
 567                                 pmptr->alloc_up = pmptr->end + 1;
 568                                 pmptr->alloc_down = pmptr->end;
 569                                 pmptr->type = pmap_type;
 570                                 pmptr->attribute = mptr->Attribute;
 571                                 /*
 572                                  * we bump these here inline so the accounting
 573                                  * below works correctly
 574                                  */
 575                                 pmptr++;
 576                                 pmap_memory_region_count++;
 577
 578                                 pmptr->alloc_up = pmptr->base = fap;
 579                                 pmptr->type = pmap_type;
 580                                 pmptr->attribute = mptr->Attribute;
 581                                 pmptr->alloc_down = pmptr->end = top;
 582
 583                                 if (mptr->Attribute & EFI_MEMORY_KERN_RESERVED)
 584                                         pmap_reserved_range_indices[pmap_last_reserved_range_index++] = pmap_memory_region_count;
 585                         } else {
 586                                 /*
 587                                  * entire range useable
 588                                  */
 589                                 pmptr->alloc_up = pmptr->base = base;
 590                                 pmptr->type = pmap_type;
 591                                 pmptr->attribute = mptr->Attribute;
 592                                 pmptr->alloc_down = pmptr->end = top;
 593                                 if (mptr->Attribute & EFI_MEMORY_KERN_RESERVED)
 594                                         pmap_reserved_range_indices[pmap_last_reserved_range_index++] = pmap_memory_region_count;
 595                         }
 596
 597                         if (i386_ptob(pmptr->end) > avail_end )
 598                                 avail_end = i386_ptob(pmptr->end);
 599
 600                         avail_remaining += (pmptr->end - pmptr->base);
 601                         coalescing_permitted = (prev_pmptr && (pmptr->attribute == prev_pmptr->attribute) && ((pmptr->attribute & EFI_MEMORY_KERN_RESERVED) == 0));
 602                         /*
 603                          * Consolidate contiguous memory regions, if possible
 604                          */
 605                         if (prev_pmptr &&
 606                             (pmptr->type == prev_pmptr->type) &&
 607                             (coalescing_permitted) &&
 608                             (pmptr->base == pmptr->alloc_up) &&
 609                             (prev_pmptr->end == prev_pmptr->alloc_down) &&
 610                             (pmptr->base == (prev_pmptr->end + 1)))
 611                         {
 612                                 prev_pmptr->end = pmptr->end;
 613                                 prev_pmptr->alloc_down = pmptr->alloc_down;
 614                         } else {
 615                                 pmap_memory_region_count++;
 616                                 prev_pmptr = pmptr;
 617                                 pmptr++;
 618                         }
 619                 }
 620         }
 621
 622         if (memmap) {
 623                 kprint_memmap(maddr, msize, mcount);
 624         }
 625
 626         avail_start = first_avail;
 627         mem_actual = args->PhysicalMemorySize;
 628
 629         /*
 630          * For user visible memory size, round up to 128 Mb
 631          * - accounting for the various stolen memory not reported by EFI.
 632          * This is maintained for historical, comparison purposes but
 633          * we now use the memory size reported by EFI/Booter.
 634          */
 635         sane_size = (sane_size + 128 * MB - 1) & ~((uint64_t)(128 * MB - 1));
 636         if (sane_size != mem_actual)
 637                 printf("mem_actual: 0x%llx\n legacy sane_size: 0x%llx\n",
 638                         mem_actual, sane_size);
 639         sane_size = mem_actual;
 640
 641         /*
 642          * We cap at KERNEL_MAXMEM bytes (currently 32GB for K32, 96GB for K64).
 643          * Unless overriden by the maxmem= boot-arg
 644          * -- which is a non-zero maxmem argument to this function.
 645          */
 646         if (maxmem == 0 && sane_size > KERNEL_MAXMEM) {
 647                 maxmem = KERNEL_MAXMEM;
 648                 printf("Physical memory %lld bytes capped at %dGB\n",
 649                         sane_size, (uint32_t) (KERNEL_MAXMEM/GB));
 650         }
 651
 652         /*
 653          * if user set maxmem, reduce memory sizes
 654          */
 655         if ( (maxmem > (uint64_t)first_avail) && (maxmem < sane_size)) {
 656                 ppnum_t discarded_pages  = (ppnum_t)((sane_size - maxmem) >> I386_PGSHIFT);
 657                 ppnum_t highest_pn = 0;
 658                 ppnum_t cur_end  = 0;
 659                 uint64_t        pages_to_use;
 660                 unsigned        cur_region = 0;
 661
 662                 sane_size = maxmem;
 663
 664                 if (avail_remaining > discarded_pages)
 665                         avail_remaining -= discarded_pages;
 666                 else
 667                         avail_remaining = 0;
 668
 669                 pages_to_use = avail_remaining;
 670
 671                 while (cur_region < pmap_memory_region_count && pages_to_use) {
 672                         for (cur_end = pmap_memory_regions[cur_region].base;
 673                              cur_end < pmap_memory_regions[cur_region].end && pages_to_use;
 674                              cur_end++) {
 675                                 if (cur_end > highest_pn)
 676                                         highest_pn = cur_end;
 677                                 pages_to_use--;
 678                         }
 679                         if (pages_to_use == 0) {
 680                                 pmap_memory_regions[cur_region].end = cur_end;
 681                                 pmap_memory_regions[cur_region].alloc_down = cur_end;
 682                         }
 683
 684                         cur_region++;
 685                 }
 686                 pmap_memory_region_count = cur_region;
 687
 688                 avail_end = i386_ptob(highest_pn + 1);
 689         }
 690
 691         /*
 692          * mem_size is only a 32 bit container... follow the PPC route
 693          * and pin it to a 2 Gbyte maximum
 694          */
 695         if (sane_size > (FOURGIG >> 1))
 696                 mem_size = (vm_size_t)(FOURGIG >> 1);
 697         else
 698                 mem_size = (vm_size_t)sane_size;
 699         max_mem = sane_size;
 700
 701         kprintf("Physical memory %llu MB\n", sane_size/MB);
 702
 703         max_valid_low_ppnum = (2 * GB) / PAGE_SIZE;
 704
 705         if (!PE_parse_boot_argn("max_valid_dma_addr", &maxdmaaddr, sizeof (maxdmaaddr))) {
 706                 max_valid_dma_address = (uint64_t)4 * (uint64_t)GB;
 707         } else {
 708                 max_valid_dma_address = ((uint64_t) maxdmaaddr) * MB;
 709
 710                 if ((max_valid_dma_address / PAGE_SIZE) < max_valid_low_ppnum)
 711                         max_valid_low_ppnum = (ppnum_t)(max_valid_dma_address / PAGE_SIZE);
 712         }
 713         if (avail_end >= max_valid_dma_address) {
 714
 715                 if (!PE_parse_boot_argn("maxloreserve", &maxloreserve, sizeof (maxloreserve))) {
 716
 717                         if (sane_size >= (ONEGIG * 15))
 718                                 maxloreserve = (MAXLORESERVE / PAGE_SIZE) * 4;
 719                         else if (sane_size >= (ONEGIG * 7))
 720                                 maxloreserve = (MAXLORESERVE / PAGE_SIZE) * 2;
 721                         else
 722                                 maxloreserve = MAXLORESERVE / PAGE_SIZE;
 723
 724 #if SOCKETS
 725                         mbuf_reserve = bsd_mbuf_cluster_reserve(&mbuf_override) / PAGE_SIZE;
 726 #endif
 727                 } else
 728                         maxloreserve = (maxloreserve * (1024 * 1024)) / PAGE_SIZE;
 729
 730                 if (maxloreserve) {
 731                         vm_lopage_free_limit = maxloreserve;
 732
 733                         if (mbuf_override == TRUE) {
 734                                 vm_lopage_free_limit += mbuf_reserve;
 735                                 vm_lopage_lowater = 0;
 736                         } else
 737                                 vm_lopage_lowater = vm_lopage_free_limit / 16;
 738
 739                         vm_lopage_refill = TRUE;
 740                         vm_lopage_needed = TRUE;
 741                 }
 742         }
 743
 744         /*
 745          *      Initialize kernel physical map.
 746          *      Kernel virtual address starts at VM_KERNEL_MIN_ADDRESS.
 747          */
 748         kprintf("avail_remaining = 0x%lx\n", (unsigned long)avail_remaining);
 749         pmap_bootstrap(0, IA32e);
 750 }
 751
 752
 753 unsigned int
 754 pmap_free_pages(void)
 755 {
 756         return (unsigned int)avail_remaining;
 757 }
 758
 759
 760 boolean_t pmap_next_page_reserved(ppnum_t *);
 761
 762 /*
 763  * Pick a page from a "kernel private" reserved range; works around
 764  * errata on some hardware.
 765  */
 766 boolean_t
 767 pmap_next_page_reserved(ppnum_t *pn) {
 768         if (pmap_reserved_ranges) {
 769                 uint32_t n;
 770                 pmap_memory_region_t *region;
 771                 for (n = 0; n < pmap_last_reserved_range_index; n++) {
 772                         uint32_t reserved_index = pmap_reserved_range_indices[n];
 773                         region = &pmap_memory_regions[reserved_index];
 774                         if (region->alloc_up <= region->alloc_down) {
 775                                 *pn = region->alloc_up++;
 776                                 avail_remaining--;
 777
 778                                 if (*pn > max_ppnum)
 779                                         max_ppnum = *pn;
 780
 781                                 if (lowest_lo == 0 || *pn < lowest_lo)
 782                                         lowest_lo = *pn;
 783
 784                                 pmap_reserved_pages_allocated++;
 785 #if DEBUG
 786                                 if (region->alloc_up > region->alloc_down) {
 787                                         kprintf("Exhausted reserved range index: %u, base: 0x%x end: 0x%x, type: 0x%x, attribute: 0x%llx\n", reserved_index, region->base, region->end, region->type, region->attribute);
 788                                 }
 789 #endif
 790                                 return TRUE;
 791                         }
 792                 }
 793         }
 794         return FALSE;
 795 }
 796
 797
 798 boolean_t
 799 pmap_next_page_hi(
 800                   ppnum_t *pn)
 801 {
 802         pmap_memory_region_t *region;
 803         int     n;
 804
 805         if (pmap_next_page_reserved(pn))
 806                 return TRUE;
 807
 808         if (avail_remaining) {
 809                 for (n = pmap_memory_region_count - 1; n >= 0; n--) {
 810                         region = &pmap_memory_regions[n];
 811
 812                         if (region->alloc_down >= region->alloc_up) {
 813                                 *pn = region->alloc_down--;
 814                                 avail_remaining--;
 815
 816                                 if (*pn > max_ppnum)
 817                                         max_ppnum = *pn;
 818
 819                                 if (lowest_lo == 0 || *pn < lowest_lo)
 820                                         lowest_lo = *pn;
 821
 822                                 if (lowest_hi == 0 || *pn < lowest_hi)
 823                                         lowest_hi = *pn;
 824
 825                                 if (*pn > highest_hi)
 826                                         highest_hi = *pn;
 827
 828                                 return TRUE;
 829                         }
 830                 }
 831         }
 832         return FALSE;
 833 }
 834
 835
 836 boolean_t
 837 pmap_next_page(
 838                ppnum_t *pn)
 839 {
 840         if (avail_remaining) while (pmap_memory_region_current < pmap_memory_region_count) {
 841                 if (pmap_memory_regions[pmap_memory_region_current].alloc_up >
 842                     pmap_memory_regions[pmap_memory_region_current].alloc_down) {
 843                         pmap_memory_region_current++;
 844                         continue;
 845                 }
 846                 *pn = pmap_memory_regions[pmap_memory_region_current].alloc_up++;
 847                 avail_remaining--;
 848
 849                 if (*pn > max_ppnum)
 850                         max_ppnum = *pn;
 851
 852                 if (lowest_lo == 0 || *pn < lowest_lo)
 853                         lowest_lo = *pn;
 854
 855                 return TRUE;
 856         }
 857         return FALSE;
 858 }
 859
 860
 861 boolean_t
 862 pmap_valid_page(
 863         ppnum_t pn)
 864 {
 865         unsigned int i;
 866         pmap_memory_region_t *pmptr = pmap_memory_regions;
 867
 868         for (i = 0; i < pmap_memory_region_count; i++, pmptr++) {
 869                 if ( (pn >= pmptr->base) && (pn <= pmptr->end) )
 870                         return TRUE;
 871         }
 872         return FALSE;
 873 }
 874