Linux Audio

Check our new training course

Loading...
v4.17
   1/*
   2 *  Copyright (C) 1995  Linus Torvalds
   3 *
   4 *  Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
   5 *
   6 *  Memory region support
   7 *	David Parsons <orc@pell.chi.il.us>, July-August 1999
   8 *
   9 *  Added E820 sanitization routine (removes overlapping memory regions);
  10 *  Brian Moyle <bmoyle@mvista.com>, February 2001
  11 *
  12 * Moved CPU detection code to cpu/${cpu}.c
  13 *    Patrick Mochel <mochel@osdl.org>, March 2002
  14 *
  15 *  Provisions for empty E820 memory regions (reported by certain BIOSes).
  16 *  Alex Achenbach <xela@slit.de>, December 2002.
  17 *
  18 */
  19
  20/*
  21 * This file handles the architecture-dependent parts of initialization
  22 */
  23
  24#include <linux/sched.h>
  25#include <linux/mm.h>
  26#include <linux/mmzone.h>
  27#include <linux/screen_info.h>
  28#include <linux/ioport.h>
  29#include <linux/acpi.h>
  30#include <linux/sfi.h>
  31#include <linux/apm_bios.h>
  32#include <linux/initrd.h>
  33#include <linux/bootmem.h>
  34#include <linux/memblock.h>
  35#include <linux/seq_file.h>
  36#include <linux/console.h>
  37#include <linux/root_dev.h>
  38#include <linux/highmem.h>
  39#include <linux/export.h>
  40#include <linux/efi.h>
  41#include <linux/init.h>
  42#include <linux/edd.h>
  43#include <linux/iscsi_ibft.h>
  44#include <linux/nodemask.h>
  45#include <linux/kexec.h>
  46#include <linux/dmi.h>
  47#include <linux/pfn.h>
  48#include <linux/pci.h>
  49#include <asm/pci-direct.h>
  50#include <linux/init_ohci1394_dma.h>
  51#include <linux/kvm_para.h>
  52#include <linux/dma-contiguous.h>
  53#include <xen/xen.h>
  54
  55#include <linux/errno.h>
  56#include <linux/kernel.h>
  57#include <linux/stddef.h>
  58#include <linux/unistd.h>
  59#include <linux/ptrace.h>
  60#include <linux/user.h>
  61#include <linux/delay.h>
  62
  63#include <linux/kallsyms.h>
  64#include <linux/cpufreq.h>
  65#include <linux/dma-mapping.h>
  66#include <linux/ctype.h>
  67#include <linux/uaccess.h>
  68
  69#include <linux/percpu.h>
  70#include <linux/crash_dump.h>
  71#include <linux/tboot.h>
  72#include <linux/jiffies.h>
  73#include <linux/mem_encrypt.h>
  74
  75#include <linux/usb/xhci-dbgp.h>
  76#include <video/edid.h>
  77
  78#include <asm/mtrr.h>
  79#include <asm/apic.h>
  80#include <asm/realmode.h>
  81#include <asm/e820/api.h>
  82#include <asm/mpspec.h>
  83#include <asm/setup.h>
  84#include <asm/efi.h>
  85#include <asm/timer.h>
  86#include <asm/i8259.h>
  87#include <asm/sections.h>
  88#include <asm/io_apic.h>
  89#include <asm/ist.h>
  90#include <asm/setup_arch.h>
  91#include <asm/bios_ebda.h>
  92#include <asm/cacheflush.h>
  93#include <asm/processor.h>
  94#include <asm/bugs.h>
  95#include <asm/kasan.h>
  96
  97#include <asm/vsyscall.h>
  98#include <asm/cpu.h>
  99#include <asm/desc.h>
 100#include <asm/dma.h>
 101#include <asm/iommu.h>
 102#include <asm/gart.h>
 103#include <asm/mmu_context.h>
 104#include <asm/proto.h>
 105
 106#include <asm/paravirt.h>
 107#include <asm/hypervisor.h>
 108#include <asm/olpc_ofw.h>
 109
 110#include <asm/percpu.h>
 111#include <asm/topology.h>
 112#include <asm/apicdef.h>
 113#include <asm/amd_nb.h>
 114#include <asm/mce.h>
 115#include <asm/alternative.h>
 116#include <asm/prom.h>
 117#include <asm/microcode.h>
 118#include <asm/kaslr.h>
 119#include <asm/unwind.h>
 120
 121/*
 122 * max_low_pfn_mapped: highest direct mapped pfn under 4GB
 123 * max_pfn_mapped:     highest direct mapped pfn over 4GB
 124 *
 125 * The direct mapping only covers E820_TYPE_RAM regions, so the ranges and gaps are
 126 * represented by pfn_mapped
 127 */
 128unsigned long max_low_pfn_mapped;
 129unsigned long max_pfn_mapped;
 130
 131#ifdef CONFIG_DMI
 132RESERVE_BRK(dmi_alloc, 65536);
 133#endif
 134
 135
 136static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
 137unsigned long _brk_end = (unsigned long)__brk_base;
 138
 
 
 
 
 
 
 
 
 
 
 
 
 139struct boot_params boot_params;
 140
 141/*
 142 * Machine setup..
 143 */
 144static struct resource data_resource = {
 145	.name	= "Kernel data",
 146	.start	= 0,
 147	.end	= 0,
 148	.flags	= IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
 149};
 150
 151static struct resource code_resource = {
 152	.name	= "Kernel code",
 153	.start	= 0,
 154	.end	= 0,
 155	.flags	= IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
 156};
 157
 158static struct resource bss_resource = {
 159	.name	= "Kernel bss",
 160	.start	= 0,
 161	.end	= 0,
 162	.flags	= IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
 163};
 164
 165
 166#ifdef CONFIG_X86_32
 167/* cpu data as detected by the assembly code in head_32.S */
 168struct cpuinfo_x86 new_cpu_data;
 169
 
 170/* common cpu data for all cpus */
 171struct cpuinfo_x86 boot_cpu_data __read_mostly;
 
 
 172EXPORT_SYMBOL(boot_cpu_data);
 173
 174unsigned int def_to_bigsmp;
 175
 176/* for MCA, but anyone else can use it if they want */
 177unsigned int machine_id;
 178unsigned int machine_submodel_id;
 179unsigned int BIOS_revision;
 180
 181struct apm_info apm_info;
 182EXPORT_SYMBOL(apm_info);
 183
 184#if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
 185	defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
 186struct ist_info ist_info;
 187EXPORT_SYMBOL(ist_info);
 188#else
 189struct ist_info ist_info;
 190#endif
 191
 192#else
 193struct cpuinfo_x86 boot_cpu_data __read_mostly;
 
 
 194EXPORT_SYMBOL(boot_cpu_data);
 195#endif
 196
 197
 198#if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
 199__visible unsigned long mmu_cr4_features __ro_after_init;
 200#else
 201__visible unsigned long mmu_cr4_features __ro_after_init = X86_CR4_PAE;
 202#endif
 203
 204/* Boot loader ID and version as integers, for the benefit of proc_dointvec */
 205int bootloader_type, bootloader_version;
 206
 207/*
 208 * Setup options
 209 */
 210struct screen_info screen_info;
 211EXPORT_SYMBOL(screen_info);
 212struct edid_info edid_info;
 213EXPORT_SYMBOL_GPL(edid_info);
 214
 215extern int root_mountflags;
 216
 217unsigned long saved_video_mode;
 218
 219#define RAMDISK_IMAGE_START_MASK	0x07FF
 220#define RAMDISK_PROMPT_FLAG		0x8000
 221#define RAMDISK_LOAD_FLAG		0x4000
 222
 223static char __initdata command_line[COMMAND_LINE_SIZE];
 224#ifdef CONFIG_CMDLINE_BOOL
 225static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
 226#endif
 227
 228#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
 229struct edd edd;
 230#ifdef CONFIG_EDD_MODULE
 231EXPORT_SYMBOL(edd);
 232#endif
 233/**
 234 * copy_edd() - Copy the BIOS EDD information
 235 *              from boot_params into a safe place.
 236 *
 237 */
 238static inline void __init copy_edd(void)
 239{
 240     memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
 241	    sizeof(edd.mbr_signature));
 242     memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
 243     edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
 244     edd.edd_info_nr = boot_params.eddbuf_entries;
 245}
 246#else
 247static inline void __init copy_edd(void)
 248{
 249}
 250#endif
 251
 252void * __init extend_brk(size_t size, size_t align)
 253{
 254	size_t mask = align - 1;
 255	void *ret;
 256
 257	BUG_ON(_brk_start == 0);
 258	BUG_ON(align & mask);
 259
 260	_brk_end = (_brk_end + mask) & ~mask;
 261	BUG_ON((char *)(_brk_end + size) > __brk_limit);
 262
 263	ret = (void *)_brk_end;
 264	_brk_end += size;
 265
 266	memset(ret, 0, size);
 267
 268	return ret;
 269}
 270
 271#ifdef CONFIG_X86_32
 272static void __init cleanup_highmap(void)
 273{
 274}
 275#endif
 276
 277static void __init reserve_brk(void)
 278{
 279	if (_brk_end > _brk_start)
 280		memblock_reserve(__pa_symbol(_brk_start),
 281				 _brk_end - _brk_start);
 282
 283	/* Mark brk area as locked down and no longer taking any
 284	   new allocations */
 285	_brk_start = 0;
 286}
 287
 288u64 relocated_ramdisk;
 289
 290#ifdef CONFIG_BLK_DEV_INITRD
 291
 292static u64 __init get_ramdisk_image(void)
 293{
 294	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 295
 296	ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
 297
 298	return ramdisk_image;
 299}
 300static u64 __init get_ramdisk_size(void)
 301{
 302	u64 ramdisk_size = boot_params.hdr.ramdisk_size;
 303
 304	ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
 305
 306	return ramdisk_size;
 307}
 308
 
 309static void __init relocate_initrd(void)
 310{
 311	/* Assume only end is not page aligned */
 312	u64 ramdisk_image = get_ramdisk_image();
 313	u64 ramdisk_size  = get_ramdisk_size();
 314	u64 area_size     = PAGE_ALIGN(ramdisk_size);
 
 
 315
 316	/* We need to move the initrd down into directly mapped mem */
 317	relocated_ramdisk = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
 318						   area_size, PAGE_SIZE);
 319
 320	if (!relocated_ramdisk)
 321		panic("Cannot find place for new RAMDISK of size %lld\n",
 322		      ramdisk_size);
 323
 324	/* Note: this includes all the mem currently occupied by
 325	   the initrd, we rely on that fact to keep the data intact. */
 326	memblock_reserve(relocated_ramdisk, area_size);
 327	initrd_start = relocated_ramdisk + PAGE_OFFSET;
 328	initrd_end   = initrd_start + ramdisk_size;
 329	printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
 330	       relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
 331
 332	copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size);
 333
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 334	printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
 335		" [mem %#010llx-%#010llx]\n",
 336		ramdisk_image, ramdisk_image + ramdisk_size - 1,
 337		relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
 338}
 339
 340static void __init early_reserve_initrd(void)
 341{
 342	/* Assume only end is not page aligned */
 343	u64 ramdisk_image = get_ramdisk_image();
 344	u64 ramdisk_size  = get_ramdisk_size();
 345	u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 346
 347	if (!boot_params.hdr.type_of_loader ||
 348	    !ramdisk_image || !ramdisk_size)
 349		return;		/* No initrd provided by bootloader */
 350
 351	memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
 352}
 353static void __init reserve_initrd(void)
 354{
 355	/* Assume only end is not page aligned */
 356	u64 ramdisk_image = get_ramdisk_image();
 357	u64 ramdisk_size  = get_ramdisk_size();
 358	u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 359	u64 mapped_size;
 360
 361	if (!boot_params.hdr.type_of_loader ||
 362	    !ramdisk_image || !ramdisk_size)
 363		return;		/* No initrd provided by bootloader */
 364
 365	initrd_start = 0;
 366
 367	mapped_size = memblock_mem_size(max_pfn_mapped);
 368	if (ramdisk_size >= (mapped_size>>1))
 369		panic("initrd too large to handle, "
 370		       "disabling initrd (%lld needed, %lld available)\n",
 371		       ramdisk_size, mapped_size>>1);
 372
 373	printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
 374			ramdisk_end - 1);
 375
 376	if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
 377				PFN_DOWN(ramdisk_end))) {
 378		/* All are mapped, easy case */
 379		initrd_start = ramdisk_image + PAGE_OFFSET;
 380		initrd_end = initrd_start + ramdisk_size;
 381		return;
 382	}
 383
 384	relocate_initrd();
 385
 386	memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
 387}
 388
 389#else
 390static void __init early_reserve_initrd(void)
 391{
 392}
 393static void __init reserve_initrd(void)
 394{
 395}
 396#endif /* CONFIG_BLK_DEV_INITRD */
 397
 398static void __init parse_setup_data(void)
 399{
 400	struct setup_data *data;
 401	u64 pa_data, pa_next;
 402
 403	pa_data = boot_params.hdr.setup_data;
 404	while (pa_data) {
 405		u32 data_len, data_type;
 406
 407		data = early_memremap(pa_data, sizeof(*data));
 
 
 408		data_len = data->len + sizeof(struct setup_data);
 409		data_type = data->type;
 410		pa_next = data->next;
 411		early_memunmap(data, sizeof(*data));
 412
 413		switch (data_type) {
 414		case SETUP_E820_EXT:
 415			e820__memory_setup_extended(pa_data, data_len);
 416			break;
 417		case SETUP_DTB:
 418			add_dtb(pa_data);
 419			break;
 420		case SETUP_EFI:
 421			parse_efi_setup(pa_data, data_len);
 422			break;
 423		default:
 424			break;
 425		}
 426		pa_data = pa_next;
 427	}
 428}
 429
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 430static void __init memblock_x86_reserve_range_setup_data(void)
 431{
 432	struct setup_data *data;
 433	u64 pa_data;
 434
 435	pa_data = boot_params.hdr.setup_data;
 436	while (pa_data) {
 437		data = early_memremap(pa_data, sizeof(*data));
 438		memblock_reserve(pa_data, sizeof(*data) + data->len);
 439		pa_data = data->next;
 440		early_memunmap(data, sizeof(*data));
 441	}
 442}
 443
 444/*
 445 * --------- Crashkernel reservation ------------------------------
 446 */
 447
 448#ifdef CONFIG_KEXEC_CORE
 449
 450/* 16M alignment for crash kernel regions */
 451#define CRASH_ALIGN		(16 << 20)
 452
 453/*
 454 * Keep the crash kernel below this limit.  On 32 bits earlier kernels
 455 * would limit the kernel to the low 512 MiB due to mapping restrictions.
 456 * On 64bit, old kexec-tools need to under 896MiB.
 457 */
 458#ifdef CONFIG_X86_32
 459# define CRASH_ADDR_LOW_MAX	(512 << 20)
 460# define CRASH_ADDR_HIGH_MAX	(512 << 20)
 461#else
 462# define CRASH_ADDR_LOW_MAX	(896UL << 20)
 463# define CRASH_ADDR_HIGH_MAX	MAXMEM
 464#endif
 465
 466static int __init reserve_crashkernel_low(void)
 467{
 468#ifdef CONFIG_X86_64
 469	unsigned long long base, low_base = 0, low_size = 0;
 
 470	unsigned long total_low_mem;
 
 
 471	int ret;
 472
 473	total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT));
 474
 475	/* crashkernel=Y,low */
 476	ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base);
 477	if (ret) {
 
 478		/*
 479		 * two parts from lib/swiotlb.c:
 480		 * -swiotlb size: user-specified with swiotlb= or default.
 481		 *
 482		 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
 483		 * to 8M for other buffers that may need to stay low too. Also
 484		 * make sure we allocate enough extra low memory so that we
 485		 * don't run out of DMA buffers for 32-bit devices.
 486		 */
 487		low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
 
 488	} else {
 489		/* passed with crashkernel=0,low ? */
 490		if (!low_size)
 491			return 0;
 492	}
 493
 494	low_base = memblock_find_in_range(0, 1ULL << 32, low_size, CRASH_ALIGN);
 
 
 495	if (!low_base) {
 496		pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
 497		       (unsigned long)(low_size >> 20));
 498		return -ENOMEM;
 499	}
 500
 501	ret = memblock_reserve(low_base, low_size);
 502	if (ret) {
 503		pr_err("%s: Error reserving crashkernel low memblock.\n", __func__);
 504		return ret;
 505	}
 506
 
 507	pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
 508		(unsigned long)(low_size >> 20),
 509		(unsigned long)(low_base >> 20),
 510		(unsigned long)(total_low_mem >> 20));
 511
 512	crashk_low_res.start = low_base;
 513	crashk_low_res.end   = low_base + low_size - 1;
 514	insert_resource(&iomem_resource, &crashk_low_res);
 515#endif
 516	return 0;
 517}
 518
 519static void __init reserve_crashkernel(void)
 520{
 521	unsigned long long crash_size, crash_base, total_mem;
 
 
 522	bool high = false;
 523	int ret;
 524
 525	total_mem = memblock_phys_mem_size();
 526
 527	/* crashkernel=XM */
 528	ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base);
 
 529	if (ret != 0 || crash_size <= 0) {
 530		/* crashkernel=X,high */
 531		ret = parse_crashkernel_high(boot_command_line, total_mem,
 532					     &crash_size, &crash_base);
 533		if (ret != 0 || crash_size <= 0)
 534			return;
 535		high = true;
 536	}
 537
 538	if (xen_pv_domain()) {
 539		pr_info("Ignoring crashkernel for a Xen PV domain\n");
 540		return;
 541	}
 542
 543	/* 0 means: find the address automatically */
 544	if (crash_base <= 0) {
 545		/*
 546		 * Set CRASH_ADDR_LOW_MAX upper bound for crash memory,
 547		 * as old kexec-tools loads bzImage below that, unless
 548		 * "crashkernel=size[KMG],high" is specified.
 549		 */
 550		crash_base = memblock_find_in_range(CRASH_ALIGN,
 551						    high ? CRASH_ADDR_HIGH_MAX
 552							 : CRASH_ADDR_LOW_MAX,
 553						    crash_size, CRASH_ALIGN);
 
 554		if (!crash_base) {
 555			pr_info("crashkernel reservation failed - No suitable area found.\n");
 556			return;
 557		}
 558
 559	} else {
 560		unsigned long long start;
 561
 562		start = memblock_find_in_range(crash_base,
 563					       crash_base + crash_size,
 564					       crash_size, 1 << 20);
 565		if (start != crash_base) {
 566			pr_info("crashkernel reservation failed - memory is in use.\n");
 567			return;
 568		}
 569	}
 570	ret = memblock_reserve(crash_base, crash_size);
 571	if (ret) {
 572		pr_err("%s: Error reserving crashkernel memblock.\n", __func__);
 573		return;
 574	}
 575
 576	if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) {
 577		memblock_free(crash_base, crash_size);
 578		return;
 579	}
 580
 581	pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
 582		(unsigned long)(crash_size >> 20),
 583		(unsigned long)(crash_base >> 20),
 584		(unsigned long)(total_mem >> 20));
 585
 586	crashk_res.start = crash_base;
 587	crashk_res.end   = crash_base + crash_size - 1;
 588	insert_resource(&iomem_resource, &crashk_res);
 
 
 
 589}
 590#else
 591static void __init reserve_crashkernel(void)
 592{
 593}
 594#endif
 595
 596static struct resource standard_io_resources[] = {
 597	{ .name = "dma1", .start = 0x00, .end = 0x1f,
 598		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 599	{ .name = "pic1", .start = 0x20, .end = 0x21,
 600		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 601	{ .name = "timer0", .start = 0x40, .end = 0x43,
 602		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 603	{ .name = "timer1", .start = 0x50, .end = 0x53,
 604		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 605	{ .name = "keyboard", .start = 0x60, .end = 0x60,
 606		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 607	{ .name = "keyboard", .start = 0x64, .end = 0x64,
 608		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 609	{ .name = "dma page reg", .start = 0x80, .end = 0x8f,
 610		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 611	{ .name = "pic2", .start = 0xa0, .end = 0xa1,
 612		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 613	{ .name = "dma2", .start = 0xc0, .end = 0xdf,
 614		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 615	{ .name = "fpu", .start = 0xf0, .end = 0xff,
 616		.flags = IORESOURCE_BUSY | IORESOURCE_IO }
 617};
 618
 619void __init reserve_standard_io_resources(void)
 620{
 621	int i;
 622
 623	/* request I/O space for devices used on all i[345]86 PCs */
 624	for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
 625		request_resource(&ioport_resource, &standard_io_resources[i]);
 626
 627}
 628
 629static __init void reserve_ibft_region(void)
 630{
 631	unsigned long addr, size = 0;
 632
 633	addr = find_ibft_region(&size);
 634
 635	if (size)
 636		memblock_reserve(addr, size);
 637}
 638
 639static bool __init snb_gfx_workaround_needed(void)
 640{
 641#ifdef CONFIG_PCI
 642	int i;
 643	u16 vendor, devid;
 644	static const __initconst u16 snb_ids[] = {
 645		0x0102,
 646		0x0112,
 647		0x0122,
 648		0x0106,
 649		0x0116,
 650		0x0126,
 651		0x010a,
 652	};
 653
 654	/* Assume no if something weird is going on with PCI */
 655	if (!early_pci_allowed())
 656		return false;
 657
 658	vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
 659	if (vendor != 0x8086)
 660		return false;
 661
 662	devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
 663	for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
 664		if (devid == snb_ids[i])
 665			return true;
 666#endif
 667
 668	return false;
 669}
 670
 671/*
 672 * Sandy Bridge graphics has trouble with certain ranges, exclude
 673 * them from allocation.
 674 */
 675static void __init trim_snb_memory(void)
 676{
 677	static const __initconst unsigned long bad_pages[] = {
 678		0x20050000,
 679		0x20110000,
 680		0x20130000,
 681		0x20138000,
 682		0x40004000,
 683	};
 684	int i;
 685
 686	if (!snb_gfx_workaround_needed())
 687		return;
 688
 689	printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
 690
 691	/*
 692	 * Reserve all memory below the 1 MB mark that has not
 693	 * already been reserved.
 694	 */
 695	memblock_reserve(0, 1<<20);
 696	
 697	for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
 698		if (memblock_reserve(bad_pages[i], PAGE_SIZE))
 699			printk(KERN_WARNING "failed to reserve 0x%08lx\n",
 700			       bad_pages[i]);
 701	}
 702}
 703
 704/*
 705 * Here we put platform-specific memory range workarounds, i.e.
 706 * memory known to be corrupt or otherwise in need to be reserved on
 707 * specific platforms.
 708 *
 709 * If this gets used more widely it could use a real dispatch mechanism.
 710 */
 711static void __init trim_platform_memory_ranges(void)
 712{
 713	trim_snb_memory();
 714}
 715
 716static void __init trim_bios_range(void)
 717{
 718	/*
 719	 * A special case is the first 4Kb of memory;
 720	 * This is a BIOS owned area, not kernel ram, but generally
 721	 * not listed as such in the E820 table.
 722	 *
 723	 * This typically reserves additional memory (64KiB by default)
 724	 * since some BIOSes are known to corrupt low memory.  See the
 725	 * Kconfig help text for X86_RESERVE_LOW.
 726	 */
 727	e820__range_update(0, PAGE_SIZE, E820_TYPE_RAM, E820_TYPE_RESERVED);
 728
 729	/*
 730	 * special case: Some BIOSen report the PC BIOS
 731	 * area (640->1Mb) as ram even though it is not.
 732	 * take them out.
 733	 */
 734	e820__range_remove(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_TYPE_RAM, 1);
 735
 736	e820__update_table(e820_table);
 737}
 738
 739/* called before trim_bios_range() to spare extra sanitize */
 740static void __init e820_add_kernel_range(void)
 741{
 742	u64 start = __pa_symbol(_text);
 743	u64 size = __pa_symbol(_end) - start;
 744
 745	/*
 746	 * Complain if .text .data and .bss are not marked as E820_TYPE_RAM and
 747	 * attempt to fix it by adding the range. We may have a confused BIOS,
 748	 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
 749	 * exclude kernel range. If we really are running on top non-RAM,
 750	 * we will crash later anyways.
 751	 */
 752	if (e820__mapped_all(start, start + size, E820_TYPE_RAM))
 753		return;
 754
 755	pr_warn(".text .data .bss are not marked as E820_TYPE_RAM!\n");
 756	e820__range_remove(start, size, E820_TYPE_RAM, 0);
 757	e820__range_add(start, size, E820_TYPE_RAM);
 758}
 759
 760static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
 761
 762static int __init parse_reservelow(char *p)
 763{
 764	unsigned long long size;
 765
 766	if (!p)
 767		return -EINVAL;
 768
 769	size = memparse(p, &p);
 770
 771	if (size < 4096)
 772		size = 4096;
 773
 774	if (size > 640*1024)
 775		size = 640*1024;
 776
 777	reserve_low = size;
 778
 779	return 0;
 780}
 781
 782early_param("reservelow", parse_reservelow);
 783
 784static void __init trim_low_memory_range(void)
 785{
 786	memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
 787}
 788	
 789/*
 790 * Dump out kernel offset information on panic.
 791 */
 792static int
 793dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
 794{
 795	if (kaslr_enabled()) {
 796		pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
 797			 kaslr_offset(),
 798			 __START_KERNEL,
 799			 __START_KERNEL_map,
 800			 MODULES_VADDR-1);
 801	} else {
 802		pr_emerg("Kernel Offset: disabled\n");
 803	}
 804
 805	return 0;
 806}
 807
 808/*
 809 * Determine if we were loaded by an EFI loader.  If so, then we have also been
 810 * passed the efi memmap, systab, etc., so we should use these data structures
 811 * for initialization.  Note, the efi init code path is determined by the
 812 * global efi_enabled. This allows the same kernel image to be used on existing
 813 * systems (with a traditional BIOS) as well as on EFI systems.
 814 */
 815/*
 816 * setup_arch - architecture-specific boot-time initializations
 817 *
 818 * Note: On x86_64, fixmaps are ready for use even before this is called.
 819 */
 820
 821void __init setup_arch(char **cmdline_p)
 822{
 823	memblock_reserve(__pa_symbol(_text),
 824			 (unsigned long)__bss_stop - (unsigned long)_text);
 825
 826	early_reserve_initrd();
 827
 828	/*
 829	 * At this point everything still needed from the boot loader
 830	 * or BIOS or kernel text should be early reserved or marked not
 831	 * RAM in e820. All other memory is free game.
 832	 */
 833
 834#ifdef CONFIG_X86_32
 835	memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
 836
 837	/*
 838	 * copy kernel address range established so far and switch
 839	 * to the proper swapper page table
 840	 */
 841	clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
 842			initial_page_table + KERNEL_PGD_BOUNDARY,
 843			KERNEL_PGD_PTRS);
 844
 845	load_cr3(swapper_pg_dir);
 846	/*
 847	 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
 848	 * a cr3 based tlb flush, so the following __flush_tlb_all()
 849	 * will not flush anything because the cpu quirk which clears
 850	 * X86_FEATURE_PGE has not been invoked yet. Though due to the
 851	 * load_cr3() above the TLB has been flushed already. The
 852	 * quirk is invoked before subsequent calls to __flush_tlb_all()
 853	 * so proper operation is guaranteed.
 854	 */
 855	__flush_tlb_all();
 856#else
 857	printk(KERN_INFO "Command line: %s\n", boot_command_line);
 858	boot_cpu_data.x86_phys_bits = MAX_PHYSMEM_BITS;
 859#endif
 860
 861	/*
 862	 * If we have OLPC OFW, we might end up relocating the fixmap due to
 863	 * reserve_top(), so do this before touching the ioremap area.
 864	 */
 865	olpc_ofw_detect();
 866
 867	idt_setup_early_traps();
 868	early_cpu_init();
 869	early_ioremap_init();
 870
 871	setup_olpc_ofw_pgd();
 872
 873	ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
 874	screen_info = boot_params.screen_info;
 875	edid_info = boot_params.edid_info;
 876#ifdef CONFIG_X86_32
 877	apm_info.bios = boot_params.apm_bios_info;
 878	ist_info = boot_params.ist_info;
 
 
 
 
 
 879#endif
 880	saved_video_mode = boot_params.hdr.vid_mode;
 881	bootloader_type = boot_params.hdr.type_of_loader;
 882	if ((bootloader_type >> 4) == 0xe) {
 883		bootloader_type &= 0xf;
 884		bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
 885	}
 886	bootloader_version  = bootloader_type & 0xf;
 887	bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
 888
 889#ifdef CONFIG_BLK_DEV_RAM
 890	rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
 891	rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
 892	rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 893#endif
 894#ifdef CONFIG_EFI
 895	if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 896		     EFI32_LOADER_SIGNATURE, 4)) {
 897		set_bit(EFI_BOOT, &efi.flags);
 898	} else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 899		     EFI64_LOADER_SIGNATURE, 4)) {
 900		set_bit(EFI_BOOT, &efi.flags);
 901		set_bit(EFI_64BIT, &efi.flags);
 902	}
 
 
 
 903#endif
 904
 905	x86_init.oem.arch_setup();
 906
 907	iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
 908	e820__memory_setup();
 909	parse_setup_data();
 910
 911	copy_edd();
 912
 913	if (!boot_params.hdr.root_flags)
 914		root_mountflags &= ~MS_RDONLY;
 915	init_mm.start_code = (unsigned long) _text;
 916	init_mm.end_code = (unsigned long) _etext;
 917	init_mm.end_data = (unsigned long) _edata;
 918	init_mm.brk = _brk_end;
 919
 920	mpx_mm_init(&init_mm);
 921
 922	code_resource.start = __pa_symbol(_text);
 923	code_resource.end = __pa_symbol(_etext)-1;
 924	data_resource.start = __pa_symbol(_etext);
 925	data_resource.end = __pa_symbol(_edata)-1;
 926	bss_resource.start = __pa_symbol(__bss_start);
 927	bss_resource.end = __pa_symbol(__bss_stop)-1;
 928
 929#ifdef CONFIG_CMDLINE_BOOL
 930#ifdef CONFIG_CMDLINE_OVERRIDE
 931	strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 932#else
 933	if (builtin_cmdline[0]) {
 934		/* append boot loader cmdline to builtin */
 935		strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
 936		strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 937		strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 938	}
 939#endif
 940#endif
 941
 942	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 943	*cmdline_p = command_line;
 944
 945	/*
 946	 * x86_configure_nx() is called before parse_early_param() to detect
 947	 * whether hardware doesn't support NX (so that the early EHCI debug
 948	 * console setup can safely call set_fixmap()). It may then be called
 949	 * again from within noexec_setup() during parsing early parameters
 950	 * to honor the respective command line option.
 951	 */
 952	x86_configure_nx();
 953
 954	parse_early_param();
 955
 956	if (efi_enabled(EFI_BOOT))
 957		efi_memblock_x86_reserve_range();
 958#ifdef CONFIG_MEMORY_HOTPLUG
 959	/*
 960	 * Memory used by the kernel cannot be hot-removed because Linux
 961	 * cannot migrate the kernel pages. When memory hotplug is
 962	 * enabled, we should prevent memblock from allocating memory
 963	 * for the kernel.
 964	 *
 965	 * ACPI SRAT records all hotpluggable memory ranges. But before
 966	 * SRAT is parsed, we don't know about it.
 967	 *
 968	 * The kernel image is loaded into memory at very early time. We
 969	 * cannot prevent this anyway. So on NUMA system, we set any
 970	 * node the kernel resides in as un-hotpluggable.
 971	 *
 972	 * Since on modern servers, one node could have double-digit
 973	 * gigabytes memory, we can assume the memory around the kernel
 974	 * image is also un-hotpluggable. So before SRAT is parsed, just
 975	 * allocate memory near the kernel image to try the best to keep
 976	 * the kernel away from hotpluggable memory.
 977	 */
 978	if (movable_node_is_enabled())
 979		memblock_set_bottom_up(true);
 980#endif
 981
 982	x86_report_nx();
 983
 984	/* after early param, so could get panic from serial */
 985	memblock_x86_reserve_range_setup_data();
 986
 987	if (acpi_mps_check()) {
 988#ifdef CONFIG_X86_LOCAL_APIC
 989		disable_apic = 1;
 990#endif
 991		setup_clear_cpu_cap(X86_FEATURE_APIC);
 992	}
 993
 994#ifdef CONFIG_PCI
 995	if (pci_early_dump_regs)
 996		early_dump_pci_devices();
 997#endif
 998
 999	e820__reserve_setup_data();
1000	e820__finish_early_params();
 
1001
1002	if (efi_enabled(EFI_BOOT))
1003		efi_init();
1004
1005	dmi_scan_machine();
1006	dmi_memdev_walk();
1007	dmi_set_dump_stack_arch_desc();
1008
1009	/*
1010	 * VMware detection requires dmi to be available, so this
1011	 * needs to be done after dmi_scan_machine(), for the boot CPU.
1012	 */
1013	init_hypervisor_platform();
1014
1015	x86_init.resources.probe_roms();
1016
1017	/* after parse_early_param, so could debug it */
1018	insert_resource(&iomem_resource, &code_resource);
1019	insert_resource(&iomem_resource, &data_resource);
1020	insert_resource(&iomem_resource, &bss_resource);
1021
1022	e820_add_kernel_range();
1023	trim_bios_range();
1024#ifdef CONFIG_X86_32
1025	if (ppro_with_ram_bug()) {
1026		e820__range_update(0x70000000ULL, 0x40000ULL, E820_TYPE_RAM,
1027				  E820_TYPE_RESERVED);
1028		e820__update_table(e820_table);
1029		printk(KERN_INFO "fixed physical RAM map:\n");
1030		e820__print_table("bad_ppro");
1031	}
1032#else
1033	early_gart_iommu_check();
1034#endif
1035
1036	/*
1037	 * partially used pages are not usable - thus
1038	 * we are rounding upwards:
1039	 */
1040	max_pfn = e820__end_of_ram_pfn();
1041
1042	/* update e820 for memory not covered by WB MTRRs */
1043	mtrr_bp_init();
1044	if (mtrr_trim_uncached_memory(max_pfn))
1045		max_pfn = e820__end_of_ram_pfn();
1046
1047	max_possible_pfn = max_pfn;
1048
1049	/*
1050	 * This call is required when the CPU does not support PAT. If
1051	 * mtrr_bp_init() invoked it already via pat_init() the call has no
1052	 * effect.
1053	 */
1054	init_cache_modes();
1055
1056	/*
1057	 * Define random base addresses for memory sections after max_pfn is
1058	 * defined and before each memory section base is used.
1059	 */
1060	kernel_randomize_memory();
1061
1062#ifdef CONFIG_X86_32
1063	/* max_low_pfn get updated here */
1064	find_low_pfn_range();
1065#else
1066	check_x2apic();
1067
1068	/* How many end-of-memory variables you have, grandma! */
1069	/* need this before calling reserve_initrd */
1070	if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1071		max_low_pfn = e820__end_of_low_ram_pfn();
1072	else
1073		max_low_pfn = max_pfn;
1074
1075	high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1076#endif
1077
1078	/*
1079	 * Find and reserve possible boot-time SMP configuration:
1080	 */
1081	find_smp_config();
1082
1083	reserve_ibft_region();
1084
1085	early_alloc_pgt_buf();
1086
1087	/*
1088	 * Need to conclude brk, before e820__memblock_setup()
1089	 *  it could use memblock_find_in_range, could overlap with
1090	 *  brk area.
1091	 */
1092	reserve_brk();
1093
1094	cleanup_highmap();
1095
1096	memblock_set_current_limit(ISA_END_ADDRESS);
1097	e820__memblock_setup();
1098
1099	reserve_bios_regions();
1100
1101	if (efi_enabled(EFI_MEMMAP)) {
1102		efi_fake_memmap();
1103		efi_find_mirror();
1104		efi_esrt_init();
1105
1106		/*
1107		 * The EFI specification says that boot service code won't be
1108		 * called after ExitBootServices(). This is, in fact, a lie.
1109		 */
 
1110		efi_reserve_boot_services();
1111	}
1112
1113	/* preallocate 4k for mptable mpc */
1114	e820__memblock_alloc_reserved_mpc_new();
1115
1116#ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1117	setup_bios_corruption_check();
1118#endif
1119
1120#ifdef CONFIG_X86_32
1121	printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1122			(max_pfn_mapped<<PAGE_SHIFT) - 1);
1123#endif
1124
1125	reserve_real_mode();
1126
1127	trim_platform_memory_ranges();
1128	trim_low_memory_range();
1129
1130	init_mem_mapping();
1131
1132	idt_setup_early_pf();
1133
1134	/*
1135	 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1136	 * with the current CR4 value.  This may not be necessary, but
1137	 * auditing all the early-boot CR4 manipulation would be needed to
1138	 * rule it out.
1139	 *
1140	 * Mask off features that don't work outside long mode (just
1141	 * PCIDE for now).
1142	 */
1143	mmu_cr4_features = __read_cr4() & ~X86_CR4_PCIDE;
1144
1145	memblock_set_current_limit(get_max_mapped());
 
1146
1147	/*
1148	 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1149	 */
1150
1151#ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1152	if (init_ohci1394_dma_early)
1153		init_ohci1394_dma_on_all_controllers();
1154#endif
1155	/* Allocate bigger log buffer */
1156	setup_log_buf(1);
1157
1158	if (efi_enabled(EFI_BOOT)) {
1159		switch (boot_params.secure_boot) {
1160		case efi_secureboot_mode_disabled:
1161			pr_info("Secure boot disabled\n");
1162			break;
1163		case efi_secureboot_mode_enabled:
1164			pr_info("Secure boot enabled\n");
1165			break;
1166		default:
1167			pr_info("Secure boot could not be determined\n");
1168			break;
1169		}
1170	}
1171
1172	reserve_initrd();
1173
1174	acpi_table_upgrade();
 
 
1175
1176	vsmp_init();
1177
1178	io_delay_init();
1179
1180	early_platform_quirks();
1181
1182	/*
1183	 * Parse the ACPI tables for possible boot-time SMP configuration.
1184	 */
1185	acpi_boot_table_init();
1186
1187	early_acpi_boot_init();
1188
1189	initmem_init();
1190	dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1191
1192	/*
1193	 * Reserve memory for crash kernel after SRAT is parsed so that it
1194	 * won't consume hotpluggable memory.
1195	 */
1196	reserve_crashkernel();
1197
1198	memblock_find_dma_reserve();
1199
1200#ifdef CONFIG_KVM_GUEST
1201	kvmclock_init();
1202#endif
1203
1204	tsc_early_delay_calibrate();
1205	if (!early_xdbc_setup_hardware())
1206		early_xdbc_register_console();
1207
1208	x86_init.paging.pagetable_init();
1209
1210	kasan_init();
 
 
 
 
 
1211
1212	/*
1213	 * Sync back kernel address range.
1214	 *
1215	 * FIXME: Can the later sync in setup_cpu_entry_areas() replace
1216	 * this call?
1217	 */
1218	sync_initial_page_table();
1219
1220	tboot_probe();
1221
 
1222	map_vsyscall();
 
1223
1224	generic_apic_probe();
1225
1226	early_quirks();
1227
1228	/*
1229	 * Read APIC and some other early information from ACPI tables.
1230	 */
1231	acpi_boot_init();
1232	sfi_init();
1233	x86_dtb_init();
1234
1235	/*
1236	 * get boot-time SMP configuration:
1237	 */
1238	get_smp_config();
1239
1240	/*
1241	 * Systems w/o ACPI and mptables might not have it mapped the local
1242	 * APIC yet, but prefill_possible_map() might need to access it.
1243	 */
1244	init_apic_mappings();
1245
1246	prefill_possible_map();
1247
1248	init_cpu_to_node();
1249
1250	io_apic_init_mappings();
 
 
1251
1252	x86_init.hyper.guest_late_init();
1253
1254	e820__reserve_resources();
1255	e820__register_nosave_regions(max_low_pfn);
1256
1257	x86_init.resources.reserve_resources();
1258
1259	e820__setup_pci_gap();
1260
1261#ifdef CONFIG_VT
1262#if defined(CONFIG_VGA_CONSOLE)
1263	if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1264		conswitchp = &vga_con;
1265#elif defined(CONFIG_DUMMY_CONSOLE)
1266	conswitchp = &dummy_con;
1267#endif
1268#endif
1269	x86_init.oem.banner();
1270
1271	x86_init.timers.wallclock_init();
1272
1273	mcheck_init();
1274
1275	arch_init_ideal_nops();
1276
1277	register_refined_jiffies(CLOCK_TICK_RATE);
1278
1279#ifdef CONFIG_EFI
1280	if (efi_enabled(EFI_BOOT))
1281		efi_apply_memmap_quirks();
1282#endif
1283
1284	unwind_init();
1285}
1286
1287#ifdef CONFIG_X86_32
1288
1289static struct resource video_ram_resource = {
1290	.name	= "Video RAM area",
1291	.start	= 0xa0000,
1292	.end	= 0xbffff,
1293	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
1294};
1295
1296void __init i386_reserve_resources(void)
1297{
1298	request_resource(&iomem_resource, &video_ram_resource);
1299	reserve_standard_io_resources();
1300}
1301
1302#endif /* CONFIG_X86_32 */
1303
1304static struct notifier_block kernel_offset_notifier = {
1305	.notifier_call = dump_kernel_offset
1306};
1307
1308static int __init register_kernel_offset_dumper(void)
1309{
1310	atomic_notifier_chain_register(&panic_notifier_list,
1311					&kernel_offset_notifier);
1312	return 0;
1313}
1314__initcall(register_kernel_offset_dumper);
1315
1316void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
1317{
1318	if (!boot_cpu_has(X86_FEATURE_OSPKE))
1319		return;
1320
1321	seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
1322}
v3.15
   1/*
   2 *  Copyright (C) 1995  Linus Torvalds
   3 *
   4 *  Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
   5 *
   6 *  Memory region support
   7 *	David Parsons <orc@pell.chi.il.us>, July-August 1999
   8 *
   9 *  Added E820 sanitization routine (removes overlapping memory regions);
  10 *  Brian Moyle <bmoyle@mvista.com>, February 2001
  11 *
  12 * Moved CPU detection code to cpu/${cpu}.c
  13 *    Patrick Mochel <mochel@osdl.org>, March 2002
  14 *
  15 *  Provisions for empty E820 memory regions (reported by certain BIOSes).
  16 *  Alex Achenbach <xela@slit.de>, December 2002.
  17 *
  18 */
  19
  20/*
  21 * This file handles the architecture-dependent parts of initialization
  22 */
  23
  24#include <linux/sched.h>
  25#include <linux/mm.h>
  26#include <linux/mmzone.h>
  27#include <linux/screen_info.h>
  28#include <linux/ioport.h>
  29#include <linux/acpi.h>
  30#include <linux/sfi.h>
  31#include <linux/apm_bios.h>
  32#include <linux/initrd.h>
  33#include <linux/bootmem.h>
  34#include <linux/memblock.h>
  35#include <linux/seq_file.h>
  36#include <linux/console.h>
  37#include <linux/root_dev.h>
  38#include <linux/highmem.h>
  39#include <linux/module.h>
  40#include <linux/efi.h>
  41#include <linux/init.h>
  42#include <linux/edd.h>
  43#include <linux/iscsi_ibft.h>
  44#include <linux/nodemask.h>
  45#include <linux/kexec.h>
  46#include <linux/dmi.h>
  47#include <linux/pfn.h>
  48#include <linux/pci.h>
  49#include <asm/pci-direct.h>
  50#include <linux/init_ohci1394_dma.h>
  51#include <linux/kvm_para.h>
  52#include <linux/dma-contiguous.h>
 
  53
  54#include <linux/errno.h>
  55#include <linux/kernel.h>
  56#include <linux/stddef.h>
  57#include <linux/unistd.h>
  58#include <linux/ptrace.h>
  59#include <linux/user.h>
  60#include <linux/delay.h>
  61
  62#include <linux/kallsyms.h>
  63#include <linux/cpufreq.h>
  64#include <linux/dma-mapping.h>
  65#include <linux/ctype.h>
  66#include <linux/uaccess.h>
  67
  68#include <linux/percpu.h>
  69#include <linux/crash_dump.h>
  70#include <linux/tboot.h>
  71#include <linux/jiffies.h>
 
  72
 
  73#include <video/edid.h>
  74
  75#include <asm/mtrr.h>
  76#include <asm/apic.h>
  77#include <asm/realmode.h>
  78#include <asm/e820.h>
  79#include <asm/mpspec.h>
  80#include <asm/setup.h>
  81#include <asm/efi.h>
  82#include <asm/timer.h>
  83#include <asm/i8259.h>
  84#include <asm/sections.h>
  85#include <asm/io_apic.h>
  86#include <asm/ist.h>
  87#include <asm/setup_arch.h>
  88#include <asm/bios_ebda.h>
  89#include <asm/cacheflush.h>
  90#include <asm/processor.h>
  91#include <asm/bugs.h>
 
  92
  93#include <asm/vsyscall.h>
  94#include <asm/cpu.h>
  95#include <asm/desc.h>
  96#include <asm/dma.h>
  97#include <asm/iommu.h>
  98#include <asm/gart.h>
  99#include <asm/mmu_context.h>
 100#include <asm/proto.h>
 101
 102#include <asm/paravirt.h>
 103#include <asm/hypervisor.h>
 104#include <asm/olpc_ofw.h>
 105
 106#include <asm/percpu.h>
 107#include <asm/topology.h>
 108#include <asm/apicdef.h>
 109#include <asm/amd_nb.h>
 110#include <asm/mce.h>
 111#include <asm/alternative.h>
 112#include <asm/prom.h>
 
 
 
 113
 114/*
 115 * max_low_pfn_mapped: highest direct mapped pfn under 4GB
 116 * max_pfn_mapped:     highest direct mapped pfn over 4GB
 117 *
 118 * The direct mapping only covers E820_RAM regions, so the ranges and gaps are
 119 * represented by pfn_mapped
 120 */
 121unsigned long max_low_pfn_mapped;
 122unsigned long max_pfn_mapped;
 123
 124#ifdef CONFIG_DMI
 125RESERVE_BRK(dmi_alloc, 65536);
 126#endif
 127
 128
 129static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
 130unsigned long _brk_end = (unsigned long)__brk_base;
 131
 132#ifdef CONFIG_X86_64
 133int default_cpu_present_to_apicid(int mps_cpu)
 134{
 135	return __default_cpu_present_to_apicid(mps_cpu);
 136}
 137
 138int default_check_phys_apicid_present(int phys_apicid)
 139{
 140	return __default_check_phys_apicid_present(phys_apicid);
 141}
 142#endif
 143
 144struct boot_params boot_params;
 145
 146/*
 147 * Machine setup..
 148 */
 149static struct resource data_resource = {
 150	.name	= "Kernel data",
 151	.start	= 0,
 152	.end	= 0,
 153	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
 154};
 155
 156static struct resource code_resource = {
 157	.name	= "Kernel code",
 158	.start	= 0,
 159	.end	= 0,
 160	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
 161};
 162
 163static struct resource bss_resource = {
 164	.name	= "Kernel bss",
 165	.start	= 0,
 166	.end	= 0,
 167	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
 168};
 169
 170
 171#ifdef CONFIG_X86_32
 172/* cpu data as detected by the assembly code in head.S */
 173struct cpuinfo_x86 new_cpu_data = {
 174	.wp_works_ok = -1,
 175};
 176/* common cpu data for all cpus */
 177struct cpuinfo_x86 boot_cpu_data __read_mostly = {
 178	.wp_works_ok = -1,
 179};
 180EXPORT_SYMBOL(boot_cpu_data);
 181
 182unsigned int def_to_bigsmp;
 183
 184/* for MCA, but anyone else can use it if they want */
 185unsigned int machine_id;
 186unsigned int machine_submodel_id;
 187unsigned int BIOS_revision;
 188
 189struct apm_info apm_info;
 190EXPORT_SYMBOL(apm_info);
 191
 192#if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
 193	defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
 194struct ist_info ist_info;
 195EXPORT_SYMBOL(ist_info);
 196#else
 197struct ist_info ist_info;
 198#endif
 199
 200#else
 201struct cpuinfo_x86 boot_cpu_data __read_mostly = {
 202	.x86_phys_bits = MAX_PHYSMEM_BITS,
 203};
 204EXPORT_SYMBOL(boot_cpu_data);
 205#endif
 206
 207
 208#if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
 209__visible unsigned long mmu_cr4_features;
 210#else
 211__visible unsigned long mmu_cr4_features = X86_CR4_PAE;
 212#endif
 213
 214/* Boot loader ID and version as integers, for the benefit of proc_dointvec */
 215int bootloader_type, bootloader_version;
 216
 217/*
 218 * Setup options
 219 */
 220struct screen_info screen_info;
 221EXPORT_SYMBOL(screen_info);
 222struct edid_info edid_info;
 223EXPORT_SYMBOL_GPL(edid_info);
 224
 225extern int root_mountflags;
 226
 227unsigned long saved_video_mode;
 228
 229#define RAMDISK_IMAGE_START_MASK	0x07FF
 230#define RAMDISK_PROMPT_FLAG		0x8000
 231#define RAMDISK_LOAD_FLAG		0x4000
 232
 233static char __initdata command_line[COMMAND_LINE_SIZE];
 234#ifdef CONFIG_CMDLINE_BOOL
 235static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
 236#endif
 237
 238#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
 239struct edd edd;
 240#ifdef CONFIG_EDD_MODULE
 241EXPORT_SYMBOL(edd);
 242#endif
 243/**
 244 * copy_edd() - Copy the BIOS EDD information
 245 *              from boot_params into a safe place.
 246 *
 247 */
 248static inline void __init copy_edd(void)
 249{
 250     memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
 251	    sizeof(edd.mbr_signature));
 252     memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
 253     edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
 254     edd.edd_info_nr = boot_params.eddbuf_entries;
 255}
 256#else
 257static inline void __init copy_edd(void)
 258{
 259}
 260#endif
 261
 262void * __init extend_brk(size_t size, size_t align)
 263{
 264	size_t mask = align - 1;
 265	void *ret;
 266
 267	BUG_ON(_brk_start == 0);
 268	BUG_ON(align & mask);
 269
 270	_brk_end = (_brk_end + mask) & ~mask;
 271	BUG_ON((char *)(_brk_end + size) > __brk_limit);
 272
 273	ret = (void *)_brk_end;
 274	_brk_end += size;
 275
 276	memset(ret, 0, size);
 277
 278	return ret;
 279}
 280
 281#ifdef CONFIG_X86_32
 282static void __init cleanup_highmap(void)
 283{
 284}
 285#endif
 286
 287static void __init reserve_brk(void)
 288{
 289	if (_brk_end > _brk_start)
 290		memblock_reserve(__pa_symbol(_brk_start),
 291				 _brk_end - _brk_start);
 292
 293	/* Mark brk area as locked down and no longer taking any
 294	   new allocations */
 295	_brk_start = 0;
 296}
 297
 298u64 relocated_ramdisk;
 299
 300#ifdef CONFIG_BLK_DEV_INITRD
 301
 302static u64 __init get_ramdisk_image(void)
 303{
 304	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 305
 306	ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
 307
 308	return ramdisk_image;
 309}
 310static u64 __init get_ramdisk_size(void)
 311{
 312	u64 ramdisk_size = boot_params.hdr.ramdisk_size;
 313
 314	ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
 315
 316	return ramdisk_size;
 317}
 318
 319#define MAX_MAP_CHUNK	(NR_FIX_BTMAPS << PAGE_SHIFT)
 320static void __init relocate_initrd(void)
 321{
 322	/* Assume only end is not page aligned */
 323	u64 ramdisk_image = get_ramdisk_image();
 324	u64 ramdisk_size  = get_ramdisk_size();
 325	u64 area_size     = PAGE_ALIGN(ramdisk_size);
 326	unsigned long slop, clen, mapaddr;
 327	char *p, *q;
 328
 329	/* We need to move the initrd down into directly mapped mem */
 330	relocated_ramdisk = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
 331						   area_size, PAGE_SIZE);
 332
 333	if (!relocated_ramdisk)
 334		panic("Cannot find place for new RAMDISK of size %lld\n",
 335		      ramdisk_size);
 336
 337	/* Note: this includes all the mem currently occupied by
 338	   the initrd, we rely on that fact to keep the data intact. */
 339	memblock_reserve(relocated_ramdisk, area_size);
 340	initrd_start = relocated_ramdisk + PAGE_OFFSET;
 341	initrd_end   = initrd_start + ramdisk_size;
 342	printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
 343	       relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
 344
 345	q = (char *)initrd_start;
 346
 347	/* Copy the initrd */
 348	while (ramdisk_size) {
 349		slop = ramdisk_image & ~PAGE_MASK;
 350		clen = ramdisk_size;
 351		if (clen > MAX_MAP_CHUNK-slop)
 352			clen = MAX_MAP_CHUNK-slop;
 353		mapaddr = ramdisk_image & PAGE_MASK;
 354		p = early_memremap(mapaddr, clen+slop);
 355		memcpy(q, p+slop, clen);
 356		early_iounmap(p, clen+slop);
 357		q += clen;
 358		ramdisk_image += clen;
 359		ramdisk_size  -= clen;
 360	}
 361
 362	ramdisk_image = get_ramdisk_image();
 363	ramdisk_size  = get_ramdisk_size();
 364	printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
 365		" [mem %#010llx-%#010llx]\n",
 366		ramdisk_image, ramdisk_image + ramdisk_size - 1,
 367		relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
 368}
 369
 370static void __init early_reserve_initrd(void)
 371{
 372	/* Assume only end is not page aligned */
 373	u64 ramdisk_image = get_ramdisk_image();
 374	u64 ramdisk_size  = get_ramdisk_size();
 375	u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 376
 377	if (!boot_params.hdr.type_of_loader ||
 378	    !ramdisk_image || !ramdisk_size)
 379		return;		/* No initrd provided by bootloader */
 380
 381	memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
 382}
 383static void __init reserve_initrd(void)
 384{
 385	/* Assume only end is not page aligned */
 386	u64 ramdisk_image = get_ramdisk_image();
 387	u64 ramdisk_size  = get_ramdisk_size();
 388	u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 389	u64 mapped_size;
 390
 391	if (!boot_params.hdr.type_of_loader ||
 392	    !ramdisk_image || !ramdisk_size)
 393		return;		/* No initrd provided by bootloader */
 394
 395	initrd_start = 0;
 396
 397	mapped_size = memblock_mem_size(max_pfn_mapped);
 398	if (ramdisk_size >= (mapped_size>>1))
 399		panic("initrd too large to handle, "
 400		       "disabling initrd (%lld needed, %lld available)\n",
 401		       ramdisk_size, mapped_size>>1);
 402
 403	printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
 404			ramdisk_end - 1);
 405
 406	if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
 407				PFN_DOWN(ramdisk_end))) {
 408		/* All are mapped, easy case */
 409		initrd_start = ramdisk_image + PAGE_OFFSET;
 410		initrd_end = initrd_start + ramdisk_size;
 411		return;
 412	}
 413
 414	relocate_initrd();
 415
 416	memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
 417}
 
 418#else
 419static void __init early_reserve_initrd(void)
 420{
 421}
 422static void __init reserve_initrd(void)
 423{
 424}
 425#endif /* CONFIG_BLK_DEV_INITRD */
 426
 427static void __init parse_setup_data(void)
 428{
 429	struct setup_data *data;
 430	u64 pa_data, pa_next;
 431
 432	pa_data = boot_params.hdr.setup_data;
 433	while (pa_data) {
 434		u32 data_len, map_len, data_type;
 435
 436		map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
 437			      (u64)sizeof(struct setup_data));
 438		data = early_memremap(pa_data, map_len);
 439		data_len = data->len + sizeof(struct setup_data);
 440		data_type = data->type;
 441		pa_next = data->next;
 442		early_iounmap(data, map_len);
 443
 444		switch (data_type) {
 445		case SETUP_E820_EXT:
 446			parse_e820_ext(pa_data, data_len);
 447			break;
 448		case SETUP_DTB:
 449			add_dtb(pa_data);
 450			break;
 451		case SETUP_EFI:
 452			parse_efi_setup(pa_data, data_len);
 453			break;
 454		default:
 455			break;
 456		}
 457		pa_data = pa_next;
 458	}
 459}
 460
 461static void __init e820_reserve_setup_data(void)
 462{
 463	struct setup_data *data;
 464	u64 pa_data;
 465	int found = 0;
 466
 467	pa_data = boot_params.hdr.setup_data;
 468	while (pa_data) {
 469		data = early_memremap(pa_data, sizeof(*data));
 470		e820_update_range(pa_data, sizeof(*data)+data->len,
 471			 E820_RAM, E820_RESERVED_KERN);
 472		found = 1;
 473		pa_data = data->next;
 474		early_iounmap(data, sizeof(*data));
 475	}
 476	if (!found)
 477		return;
 478
 479	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 480	memcpy(&e820_saved, &e820, sizeof(struct e820map));
 481	printk(KERN_INFO "extended physical RAM map:\n");
 482	e820_print_map("reserve setup_data");
 483}
 484
 485static void __init memblock_x86_reserve_range_setup_data(void)
 486{
 487	struct setup_data *data;
 488	u64 pa_data;
 489
 490	pa_data = boot_params.hdr.setup_data;
 491	while (pa_data) {
 492		data = early_memremap(pa_data, sizeof(*data));
 493		memblock_reserve(pa_data, sizeof(*data) + data->len);
 494		pa_data = data->next;
 495		early_iounmap(data, sizeof(*data));
 496	}
 497}
 498
 499/*
 500 * --------- Crashkernel reservation ------------------------------
 501 */
 502
 503#ifdef CONFIG_KEXEC
 
 
 
 504
 505/*
 506 * Keep the crash kernel below this limit.  On 32 bits earlier kernels
 507 * would limit the kernel to the low 512 MiB due to mapping restrictions.
 508 * On 64bit, old kexec-tools need to under 896MiB.
 509 */
 510#ifdef CONFIG_X86_32
 511# define CRASH_KERNEL_ADDR_LOW_MAX	(512 << 20)
 512# define CRASH_KERNEL_ADDR_HIGH_MAX	(512 << 20)
 513#else
 514# define CRASH_KERNEL_ADDR_LOW_MAX	(896UL<<20)
 515# define CRASH_KERNEL_ADDR_HIGH_MAX	MAXMEM
 516#endif
 517
 518static void __init reserve_crashkernel_low(void)
 519{
 520#ifdef CONFIG_X86_64
 521	const unsigned long long alignment = 16<<20;	/* 16M */
 522	unsigned long long low_base = 0, low_size = 0;
 523	unsigned long total_low_mem;
 524	unsigned long long base;
 525	bool auto_set = false;
 526	int ret;
 527
 528	total_low_mem = memblock_mem_size(1UL<<(32-PAGE_SHIFT));
 
 529	/* crashkernel=Y,low */
 530	ret = parse_crashkernel_low(boot_command_line, total_low_mem,
 531						&low_size, &base);
 532	if (ret != 0) {
 533		/*
 534		 * two parts from lib/swiotlb.c:
 535		 *	swiotlb size: user specified with swiotlb= or default.
 536		 *	swiotlb overflow buffer: now is hardcoded to 32k.
 537		 *		We round it to 8M for other buffers that
 538		 *		may need to stay low too.
 
 
 539		 */
 540		low_size = swiotlb_size_or_default() + (8UL<<20);
 541		auto_set = true;
 542	} else {
 543		/* passed with crashkernel=0,low ? */
 544		if (!low_size)
 545			return;
 546	}
 547
 548	low_base = memblock_find_in_range(low_size, (1ULL<<32),
 549					low_size, alignment);
 550
 551	if (!low_base) {
 552		if (!auto_set)
 553			pr_info("crashkernel low reservation failed - No suitable area found.\n");
 
 
 554
 555		return;
 
 
 
 556	}
 557
 558	memblock_reserve(low_base, low_size);
 559	pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
 560			(unsigned long)(low_size >> 20),
 561			(unsigned long)(low_base >> 20),
 562			(unsigned long)(total_low_mem >> 20));
 
 563	crashk_low_res.start = low_base;
 564	crashk_low_res.end   = low_base + low_size - 1;
 565	insert_resource(&iomem_resource, &crashk_low_res);
 566#endif
 
 567}
 568
 569static void __init reserve_crashkernel(void)
 570{
 571	const unsigned long long alignment = 16<<20;	/* 16M */
 572	unsigned long long total_mem;
 573	unsigned long long crash_size, crash_base;
 574	bool high = false;
 575	int ret;
 576
 577	total_mem = memblock_phys_mem_size();
 578
 579	/* crashkernel=XM */
 580	ret = parse_crashkernel(boot_command_line, total_mem,
 581			&crash_size, &crash_base);
 582	if (ret != 0 || crash_size <= 0) {
 583		/* crashkernel=X,high */
 584		ret = parse_crashkernel_high(boot_command_line, total_mem,
 585				&crash_size, &crash_base);
 586		if (ret != 0 || crash_size <= 0)
 587			return;
 588		high = true;
 589	}
 590
 
 
 
 
 
 591	/* 0 means: find the address automatically */
 592	if (crash_base <= 0) {
 593		/*
 594		 *  kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
 
 
 595		 */
 596		crash_base = memblock_find_in_range(alignment,
 597					high ? CRASH_KERNEL_ADDR_HIGH_MAX :
 598					       CRASH_KERNEL_ADDR_LOW_MAX,
 599					crash_size, alignment);
 600
 601		if (!crash_base) {
 602			pr_info("crashkernel reservation failed - No suitable area found.\n");
 603			return;
 604		}
 605
 606	} else {
 607		unsigned long long start;
 608
 609		start = memblock_find_in_range(crash_base,
 610				 crash_base + crash_size, crash_size, 1<<20);
 
 611		if (start != crash_base) {
 612			pr_info("crashkernel reservation failed - memory is in use.\n");
 613			return;
 614		}
 615	}
 616	memblock_reserve(crash_base, crash_size);
 
 
 
 
 617
 618	printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
 619			"for crashkernel (System RAM: %ldMB)\n",
 620			(unsigned long)(crash_size >> 20),
 621			(unsigned long)(crash_base >> 20),
 622			(unsigned long)(total_mem >> 20));
 
 
 
 
 623
 624	crashk_res.start = crash_base;
 625	crashk_res.end   = crash_base + crash_size - 1;
 626	insert_resource(&iomem_resource, &crashk_res);
 627
 628	if (crash_base >= (1ULL<<32))
 629		reserve_crashkernel_low();
 630}
 631#else
 632static void __init reserve_crashkernel(void)
 633{
 634}
 635#endif
 636
 637static struct resource standard_io_resources[] = {
 638	{ .name = "dma1", .start = 0x00, .end = 0x1f,
 639		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 640	{ .name = "pic1", .start = 0x20, .end = 0x21,
 641		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 642	{ .name = "timer0", .start = 0x40, .end = 0x43,
 643		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 644	{ .name = "timer1", .start = 0x50, .end = 0x53,
 645		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 646	{ .name = "keyboard", .start = 0x60, .end = 0x60,
 647		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 648	{ .name = "keyboard", .start = 0x64, .end = 0x64,
 649		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 650	{ .name = "dma page reg", .start = 0x80, .end = 0x8f,
 651		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 652	{ .name = "pic2", .start = 0xa0, .end = 0xa1,
 653		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 654	{ .name = "dma2", .start = 0xc0, .end = 0xdf,
 655		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 656	{ .name = "fpu", .start = 0xf0, .end = 0xff,
 657		.flags = IORESOURCE_BUSY | IORESOURCE_IO }
 658};
 659
 660void __init reserve_standard_io_resources(void)
 661{
 662	int i;
 663
 664	/* request I/O space for devices used on all i[345]86 PCs */
 665	for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
 666		request_resource(&ioport_resource, &standard_io_resources[i]);
 667
 668}
 669
 670static __init void reserve_ibft_region(void)
 671{
 672	unsigned long addr, size = 0;
 673
 674	addr = find_ibft_region(&size);
 675
 676	if (size)
 677		memblock_reserve(addr, size);
 678}
 679
 680static bool __init snb_gfx_workaround_needed(void)
 681{
 682#ifdef CONFIG_PCI
 683	int i;
 684	u16 vendor, devid;
 685	static const __initconst u16 snb_ids[] = {
 686		0x0102,
 687		0x0112,
 688		0x0122,
 689		0x0106,
 690		0x0116,
 691		0x0126,
 692		0x010a,
 693	};
 694
 695	/* Assume no if something weird is going on with PCI */
 696	if (!early_pci_allowed())
 697		return false;
 698
 699	vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
 700	if (vendor != 0x8086)
 701		return false;
 702
 703	devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
 704	for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
 705		if (devid == snb_ids[i])
 706			return true;
 707#endif
 708
 709	return false;
 710}
 711
 712/*
 713 * Sandy Bridge graphics has trouble with certain ranges, exclude
 714 * them from allocation.
 715 */
 716static void __init trim_snb_memory(void)
 717{
 718	static const __initconst unsigned long bad_pages[] = {
 719		0x20050000,
 720		0x20110000,
 721		0x20130000,
 722		0x20138000,
 723		0x40004000,
 724	};
 725	int i;
 726
 727	if (!snb_gfx_workaround_needed())
 728		return;
 729
 730	printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
 731
 732	/*
 733	 * Reserve all memory below the 1 MB mark that has not
 734	 * already been reserved.
 735	 */
 736	memblock_reserve(0, 1<<20);
 737	
 738	for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
 739		if (memblock_reserve(bad_pages[i], PAGE_SIZE))
 740			printk(KERN_WARNING "failed to reserve 0x%08lx\n",
 741			       bad_pages[i]);
 742	}
 743}
 744
 745/*
 746 * Here we put platform-specific memory range workarounds, i.e.
 747 * memory known to be corrupt or otherwise in need to be reserved on
 748 * specific platforms.
 749 *
 750 * If this gets used more widely it could use a real dispatch mechanism.
 751 */
 752static void __init trim_platform_memory_ranges(void)
 753{
 754	trim_snb_memory();
 755}
 756
 757static void __init trim_bios_range(void)
 758{
 759	/*
 760	 * A special case is the first 4Kb of memory;
 761	 * This is a BIOS owned area, not kernel ram, but generally
 762	 * not listed as such in the E820 table.
 763	 *
 764	 * This typically reserves additional memory (64KiB by default)
 765	 * since some BIOSes are known to corrupt low memory.  See the
 766	 * Kconfig help text for X86_RESERVE_LOW.
 767	 */
 768	e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED);
 769
 770	/*
 771	 * special case: Some BIOSen report the PC BIOS
 772	 * area (640->1Mb) as ram even though it is not.
 773	 * take them out.
 774	 */
 775	e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
 776
 777	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 778}
 779
 780/* called before trim_bios_range() to spare extra sanitize */
 781static void __init e820_add_kernel_range(void)
 782{
 783	u64 start = __pa_symbol(_text);
 784	u64 size = __pa_symbol(_end) - start;
 785
 786	/*
 787	 * Complain if .text .data and .bss are not marked as E820_RAM and
 788	 * attempt to fix it by adding the range. We may have a confused BIOS,
 789	 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
 790	 * exclude kernel range. If we really are running on top non-RAM,
 791	 * we will crash later anyways.
 792	 */
 793	if (e820_all_mapped(start, start + size, E820_RAM))
 794		return;
 795
 796	pr_warn(".text .data .bss are not marked as E820_RAM!\n");
 797	e820_remove_range(start, size, E820_RAM, 0);
 798	e820_add_region(start, size, E820_RAM);
 799}
 800
 801static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
 802
 803static int __init parse_reservelow(char *p)
 804{
 805	unsigned long long size;
 806
 807	if (!p)
 808		return -EINVAL;
 809
 810	size = memparse(p, &p);
 811
 812	if (size < 4096)
 813		size = 4096;
 814
 815	if (size > 640*1024)
 816		size = 640*1024;
 817
 818	reserve_low = size;
 819
 820	return 0;
 821}
 822
 823early_param("reservelow", parse_reservelow);
 824
 825static void __init trim_low_memory_range(void)
 826{
 827	memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
 828}
 829	
 830/*
 831 * Dump out kernel offset information on panic.
 832 */
 833static int
 834dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
 835{
 836	pr_emerg("Kernel Offset: 0x%lx from 0x%lx "
 837		 "(relocation range: 0x%lx-0x%lx)\n",
 838		 (unsigned long)&_text - __START_KERNEL, __START_KERNEL,
 839		 __START_KERNEL_map, MODULES_VADDR-1);
 
 
 
 
 
 840
 841	return 0;
 842}
 843
 844/*
 845 * Determine if we were loaded by an EFI loader.  If so, then we have also been
 846 * passed the efi memmap, systab, etc., so we should use these data structures
 847 * for initialization.  Note, the efi init code path is determined by the
 848 * global efi_enabled. This allows the same kernel image to be used on existing
 849 * systems (with a traditional BIOS) as well as on EFI systems.
 850 */
 851/*
 852 * setup_arch - architecture-specific boot-time initializations
 853 *
 854 * Note: On x86_64, fixmaps are ready for use even before this is called.
 855 */
 856
 857void __init setup_arch(char **cmdline_p)
 858{
 859	memblock_reserve(__pa_symbol(_text),
 860			 (unsigned long)__bss_stop - (unsigned long)_text);
 861
 862	early_reserve_initrd();
 863
 864	/*
 865	 * At this point everything still needed from the boot loader
 866	 * or BIOS or kernel text should be early reserved or marked not
 867	 * RAM in e820. All other memory is free game.
 868	 */
 869
 870#ifdef CONFIG_X86_32
 871	memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
 872
 873	/*
 874	 * copy kernel address range established so far and switch
 875	 * to the proper swapper page table
 876	 */
 877	clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
 878			initial_page_table + KERNEL_PGD_BOUNDARY,
 879			KERNEL_PGD_PTRS);
 880
 881	load_cr3(swapper_pg_dir);
 
 
 
 
 
 
 
 
 
 882	__flush_tlb_all();
 883#else
 884	printk(KERN_INFO "Command line: %s\n", boot_command_line);
 
 885#endif
 886
 887	/*
 888	 * If we have OLPC OFW, we might end up relocating the fixmap due to
 889	 * reserve_top(), so do this before touching the ioremap area.
 890	 */
 891	olpc_ofw_detect();
 892
 893	early_trap_init();
 894	early_cpu_init();
 895	early_ioremap_init();
 896
 897	setup_olpc_ofw_pgd();
 898
 899	ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
 900	screen_info = boot_params.screen_info;
 901	edid_info = boot_params.edid_info;
 902#ifdef CONFIG_X86_32
 903	apm_info.bios = boot_params.apm_bios_info;
 904	ist_info = boot_params.ist_info;
 905	if (boot_params.sys_desc_table.length != 0) {
 906		machine_id = boot_params.sys_desc_table.table[0];
 907		machine_submodel_id = boot_params.sys_desc_table.table[1];
 908		BIOS_revision = boot_params.sys_desc_table.table[2];
 909	}
 910#endif
 911	saved_video_mode = boot_params.hdr.vid_mode;
 912	bootloader_type = boot_params.hdr.type_of_loader;
 913	if ((bootloader_type >> 4) == 0xe) {
 914		bootloader_type &= 0xf;
 915		bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
 916	}
 917	bootloader_version  = bootloader_type & 0xf;
 918	bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
 919
 920#ifdef CONFIG_BLK_DEV_RAM
 921	rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
 922	rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
 923	rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 924#endif
 925#ifdef CONFIG_EFI
 926	if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 927		     "EL32", 4)) {
 928		set_bit(EFI_BOOT, &efi.flags);
 929	} else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 930		     "EL64", 4)) {
 931		set_bit(EFI_BOOT, &efi.flags);
 932		set_bit(EFI_64BIT, &efi.flags);
 933	}
 934
 935	if (efi_enabled(EFI_BOOT))
 936		efi_memblock_x86_reserve_range();
 937#endif
 938
 939	x86_init.oem.arch_setup();
 940
 941	iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
 942	setup_memory_map();
 943	parse_setup_data();
 944
 945	copy_edd();
 946
 947	if (!boot_params.hdr.root_flags)
 948		root_mountflags &= ~MS_RDONLY;
 949	init_mm.start_code = (unsigned long) _text;
 950	init_mm.end_code = (unsigned long) _etext;
 951	init_mm.end_data = (unsigned long) _edata;
 952	init_mm.brk = _brk_end;
 953
 
 
 954	code_resource.start = __pa_symbol(_text);
 955	code_resource.end = __pa_symbol(_etext)-1;
 956	data_resource.start = __pa_symbol(_etext);
 957	data_resource.end = __pa_symbol(_edata)-1;
 958	bss_resource.start = __pa_symbol(__bss_start);
 959	bss_resource.end = __pa_symbol(__bss_stop)-1;
 960
 961#ifdef CONFIG_CMDLINE_BOOL
 962#ifdef CONFIG_CMDLINE_OVERRIDE
 963	strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 964#else
 965	if (builtin_cmdline[0]) {
 966		/* append boot loader cmdline to builtin */
 967		strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
 968		strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 969		strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 970	}
 971#endif
 972#endif
 973
 974	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 975	*cmdline_p = command_line;
 976
 977	/*
 978	 * x86_configure_nx() is called before parse_early_param() to detect
 979	 * whether hardware doesn't support NX (so that the early EHCI debug
 980	 * console setup can safely call set_fixmap()). It may then be called
 981	 * again from within noexec_setup() during parsing early parameters
 982	 * to honor the respective command line option.
 983	 */
 984	x86_configure_nx();
 985
 986	parse_early_param();
 987
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 988	x86_report_nx();
 989
 990	/* after early param, so could get panic from serial */
 991	memblock_x86_reserve_range_setup_data();
 992
 993	if (acpi_mps_check()) {
 994#ifdef CONFIG_X86_LOCAL_APIC
 995		disable_apic = 1;
 996#endif
 997		setup_clear_cpu_cap(X86_FEATURE_APIC);
 998	}
 999
1000#ifdef CONFIG_PCI
1001	if (pci_early_dump_regs)
1002		early_dump_pci_devices();
1003#endif
1004
1005	/* update the e820_saved too */
1006	e820_reserve_setup_data();
1007	finish_e820_parsing();
1008
1009	if (efi_enabled(EFI_BOOT))
1010		efi_init();
1011
1012	dmi_scan_machine();
1013	dmi_memdev_walk();
1014	dmi_set_dump_stack_arch_desc();
1015
1016	/*
1017	 * VMware detection requires dmi to be available, so this
1018	 * needs to be done after dmi_scan_machine, for the BP.
1019	 */
1020	init_hypervisor_platform();
1021
1022	x86_init.resources.probe_roms();
1023
1024	/* after parse_early_param, so could debug it */
1025	insert_resource(&iomem_resource, &code_resource);
1026	insert_resource(&iomem_resource, &data_resource);
1027	insert_resource(&iomem_resource, &bss_resource);
1028
1029	e820_add_kernel_range();
1030	trim_bios_range();
1031#ifdef CONFIG_X86_32
1032	if (ppro_with_ram_bug()) {
1033		e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
1034				  E820_RESERVED);
1035		sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
1036		printk(KERN_INFO "fixed physical RAM map:\n");
1037		e820_print_map("bad_ppro");
1038	}
1039#else
1040	early_gart_iommu_check();
1041#endif
1042
1043	/*
1044	 * partially used pages are not usable - thus
1045	 * we are rounding upwards:
1046	 */
1047	max_pfn = e820_end_of_ram_pfn();
1048
1049	/* update e820 for memory not covered by WB MTRRs */
1050	mtrr_bp_init();
1051	if (mtrr_trim_uncached_memory(max_pfn))
1052		max_pfn = e820_end_of_ram_pfn();
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1053
1054#ifdef CONFIG_X86_32
1055	/* max_low_pfn get updated here */
1056	find_low_pfn_range();
1057#else
1058	check_x2apic();
1059
1060	/* How many end-of-memory variables you have, grandma! */
1061	/* need this before calling reserve_initrd */
1062	if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1063		max_low_pfn = e820_end_of_low_ram_pfn();
1064	else
1065		max_low_pfn = max_pfn;
1066
1067	high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1068#endif
1069
1070	/*
1071	 * Find and reserve possible boot-time SMP configuration:
1072	 */
1073	find_smp_config();
1074
1075	reserve_ibft_region();
1076
1077	early_alloc_pgt_buf();
1078
1079	/*
1080	 * Need to conclude brk, before memblock_x86_fill()
1081	 *  it could use memblock_find_in_range, could overlap with
1082	 *  brk area.
1083	 */
1084	reserve_brk();
1085
1086	cleanup_highmap();
1087
1088	memblock_set_current_limit(ISA_END_ADDRESS);
1089	memblock_x86_fill();
 
 
 
 
 
 
 
1090
1091	/*
1092	 * The EFI specification says that boot service code won't be called
1093	 * after ExitBootServices(). This is, in fact, a lie.
1094	 */
1095	if (efi_enabled(EFI_MEMMAP))
1096		efi_reserve_boot_services();
 
1097
1098	/* preallocate 4k for mptable mpc */
1099	early_reserve_e820_mpc_new();
1100
1101#ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1102	setup_bios_corruption_check();
1103#endif
1104
1105#ifdef CONFIG_X86_32
1106	printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1107			(max_pfn_mapped<<PAGE_SHIFT) - 1);
1108#endif
1109
1110	reserve_real_mode();
1111
1112	trim_platform_memory_ranges();
1113	trim_low_memory_range();
1114
1115	init_mem_mapping();
1116
1117	early_trap_pf_init();
1118
1119	setup_real_mode();
 
 
 
 
 
 
 
 
 
1120
1121	memblock_set_current_limit(get_max_mapped());
1122	dma_contiguous_reserve(0);
1123
1124	/*
1125	 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1126	 */
1127
1128#ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1129	if (init_ohci1394_dma_early)
1130		init_ohci1394_dma_on_all_controllers();
1131#endif
1132	/* Allocate bigger log buffer */
1133	setup_log_buf(1);
1134
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1135	reserve_initrd();
1136
1137#if defined(CONFIG_ACPI) && defined(CONFIG_BLK_DEV_INITRD)
1138	acpi_initrd_override((void *)initrd_start, initrd_end - initrd_start);
1139#endif
1140
1141	vsmp_init();
1142
1143	io_delay_init();
1144
 
 
1145	/*
1146	 * Parse the ACPI tables for possible boot-time SMP configuration.
1147	 */
1148	acpi_boot_table_init();
1149
1150	early_acpi_boot_init();
1151
1152	initmem_init();
 
1153
1154	/*
1155	 * Reserve memory for crash kernel after SRAT is parsed so that it
1156	 * won't consume hotpluggable memory.
1157	 */
1158	reserve_crashkernel();
1159
1160	memblock_find_dma_reserve();
1161
1162#ifdef CONFIG_KVM_GUEST
1163	kvmclock_init();
1164#endif
1165
 
 
 
 
1166	x86_init.paging.pagetable_init();
1167
1168	if (boot_cpu_data.cpuid_level >= 0) {
1169		/* A CPU has %cr4 if and only if it has CPUID */
1170		mmu_cr4_features = read_cr4();
1171		if (trampoline_cr4_features)
1172			*trampoline_cr4_features = mmu_cr4_features;
1173	}
1174
1175#ifdef CONFIG_X86_32
1176	/* sync back kernel address range */
1177	clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1178			swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
1179			KERNEL_PGD_PTRS);
1180#endif
 
1181
1182	tboot_probe();
1183
1184#ifdef CONFIG_X86_64
1185	map_vsyscall();
1186#endif
1187
1188	generic_apic_probe();
1189
1190	early_quirks();
1191
1192	/*
1193	 * Read APIC and some other early information from ACPI tables.
1194	 */
1195	acpi_boot_init();
1196	sfi_init();
1197	x86_dtb_init();
1198
1199	/*
1200	 * get boot-time SMP configuration:
1201	 */
1202	if (smp_found_config)
1203		get_smp_config();
 
 
 
 
 
1204
1205	prefill_possible_map();
1206
1207	init_cpu_to_node();
1208
1209	init_apic_mappings();
1210	if (x86_io_apic_ops.init)
1211		x86_io_apic_ops.init();
1212
1213	kvm_guest_init();
1214
1215	e820_reserve_resources();
1216	e820_mark_nosave_regions(max_low_pfn);
1217
1218	x86_init.resources.reserve_resources();
1219
1220	e820_setup_gap();
1221
1222#ifdef CONFIG_VT
1223#if defined(CONFIG_VGA_CONSOLE)
1224	if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1225		conswitchp = &vga_con;
1226#elif defined(CONFIG_DUMMY_CONSOLE)
1227	conswitchp = &dummy_con;
1228#endif
1229#endif
1230	x86_init.oem.banner();
1231
1232	x86_init.timers.wallclock_init();
1233
1234	mcheck_init();
1235
1236	arch_init_ideal_nops();
1237
1238	register_refined_jiffies(CLOCK_TICK_RATE);
1239
1240#ifdef CONFIG_EFI
1241	if (efi_enabled(EFI_BOOT))
1242		efi_apply_memmap_quirks();
1243#endif
 
 
1244}
1245
1246#ifdef CONFIG_X86_32
1247
1248static struct resource video_ram_resource = {
1249	.name	= "Video RAM area",
1250	.start	= 0xa0000,
1251	.end	= 0xbffff,
1252	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
1253};
1254
1255void __init i386_reserve_resources(void)
1256{
1257	request_resource(&iomem_resource, &video_ram_resource);
1258	reserve_standard_io_resources();
1259}
1260
1261#endif /* CONFIG_X86_32 */
1262
1263static struct notifier_block kernel_offset_notifier = {
1264	.notifier_call = dump_kernel_offset
1265};
1266
1267static int __init register_kernel_offset_dumper(void)
1268{
1269	atomic_notifier_chain_register(&panic_notifier_list,
1270					&kernel_offset_notifier);
1271	return 0;
1272}
1273__initcall(register_kernel_offset_dumper);