Linux Audio

Check our new training course

Linux kernel drivers training

May 6-19, 2025
Register
Loading...
v3.5.6
 
   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
  72#include <video/edid.h>
 
 
  73
  74#include <asm/mtrr.h>
  75#include <asm/apic.h>
  76#include <asm/realmode.h>
  77#include <asm/e820.h>
  78#include <asm/mpspec.h>
  79#include <asm/setup.h>
  80#include <asm/efi.h>
  81#include <asm/timer.h>
  82#include <asm/i8259.h>
  83#include <asm/sections.h>
  84#include <asm/dmi.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#ifdef CONFIG_X86_64
 111#include <asm/numa_64.h>
 112#endif
 113#include <asm/mce.h>
 114#include <asm/alternative.h>
 
 
 
 115#include <asm/prom.h>
 
 
 
 
 116
 117/*
 118 * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
 119 * The direct mapping extends to max_pfn_mapped, so that we can directly access
 120 * apertures, ACPI and other tables without having to play with fixmaps.
 
 
 121 */
 122unsigned long max_low_pfn_mapped;
 123unsigned long max_pfn_mapped;
 124
 125#ifdef CONFIG_DMI
 126RESERVE_BRK(dmi_alloc, 65536);
 127#endif
 128
 129
 130static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
 131unsigned long _brk_end = (unsigned long)__brk_base;
 132
 133#ifdef CONFIG_X86_64
 134int default_cpu_present_to_apicid(int mps_cpu)
 135{
 136	return __default_cpu_present_to_apicid(mps_cpu);
 137}
 138
 139int default_check_phys_apicid_present(int phys_apicid)
 140{
 141	return __default_check_phys_apicid_present(phys_apicid);
 142}
 143#endif
 144
 145#ifndef CONFIG_DEBUG_BOOT_PARAMS
 146struct boot_params __initdata boot_params;
 147#else
 148struct boot_params boot_params;
 149#endif
 150
 151/*
 152 * Machine setup..
 
 
 153 */
 
 
 
 
 
 
 
 
 154static struct resource data_resource = {
 155	.name	= "Kernel data",
 156	.start	= 0,
 157	.end	= 0,
 158	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
 159};
 160
 161static struct resource code_resource = {
 162	.name	= "Kernel code",
 163	.start	= 0,
 164	.end	= 0,
 165	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
 166};
 167
 168static struct resource bss_resource = {
 169	.name	= "Kernel bss",
 170	.start	= 0,
 171	.end	= 0,
 172	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
 173};
 174
 175
 176#ifdef CONFIG_X86_32
 177/* cpu data as detected by the assembly code in head.S */
 178struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
 179/* common cpu data for all cpus */
 180struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
 
 181EXPORT_SYMBOL(boot_cpu_data);
 182
 183unsigned int def_to_bigsmp;
 184
 185/* for MCA, but anyone else can use it if they want */
 186unsigned int machine_id;
 187unsigned int machine_submodel_id;
 188unsigned int BIOS_revision;
 189
 190struct apm_info apm_info;
 191EXPORT_SYMBOL(apm_info);
 192
 193#if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
 194	defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
 195struct ist_info ist_info;
 196EXPORT_SYMBOL(ist_info);
 197#else
 198struct ist_info ist_info;
 199#endif
 200
 201#else
 202struct cpuinfo_x86 boot_cpu_data __read_mostly = {
 203	.x86_phys_bits = MAX_PHYSMEM_BITS,
 204};
 205EXPORT_SYMBOL(boot_cpu_data);
 206#endif
 207
 208
 209#if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
 210unsigned long mmu_cr4_features;
 211#else
 212unsigned long mmu_cr4_features = X86_CR4_PAE;
 213#endif
 214
 215/* Boot loader ID and version as integers, for the benefit of proc_dointvec */
 216int bootloader_type, bootloader_version;
 217
 218/*
 219 * Setup options
 220 */
 221struct screen_info screen_info;
 222EXPORT_SYMBOL(screen_info);
 223struct edid_info edid_info;
 224EXPORT_SYMBOL_GPL(edid_info);
 225
 226extern int root_mountflags;
 227
 228unsigned long saved_video_mode;
 229
 230#define RAMDISK_IMAGE_START_MASK	0x07FF
 231#define RAMDISK_PROMPT_FLAG		0x8000
 232#define RAMDISK_LOAD_FLAG		0x4000
 233
 234static char __initdata command_line[COMMAND_LINE_SIZE];
 235#ifdef CONFIG_CMDLINE_BOOL
 236static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
 237#endif
 238
 239#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
 240struct edd edd;
 241#ifdef CONFIG_EDD_MODULE
 242EXPORT_SYMBOL(edd);
 243#endif
 244/**
 245 * copy_edd() - Copy the BIOS EDD information
 246 *              from boot_params into a safe place.
 247 *
 248 */
 249static inline void __init copy_edd(void)
 250{
 251     memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
 252	    sizeof(edd.mbr_signature));
 253     memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
 254     edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
 255     edd.edd_info_nr = boot_params.eddbuf_entries;
 256}
 257#else
 258static inline void __init copy_edd(void)
 259{
 260}
 261#endif
 262
 263void * __init extend_brk(size_t size, size_t align)
 264{
 265	size_t mask = align - 1;
 266	void *ret;
 267
 268	BUG_ON(_brk_start == 0);
 269	BUG_ON(align & mask);
 270
 271	_brk_end = (_brk_end + mask) & ~mask;
 272	BUG_ON((char *)(_brk_end + size) > __brk_limit);
 273
 274	ret = (void *)_brk_end;
 275	_brk_end += size;
 276
 277	memset(ret, 0, size);
 278
 279	return ret;
 280}
 281
 282#ifdef CONFIG_X86_64
 283static void __init init_gbpages(void)
 284{
 285	if (direct_gbpages && cpu_has_gbpages)
 286		printk(KERN_INFO "Using GB pages for direct mapping\n");
 287	else
 288		direct_gbpages = 0;
 289}
 290#else
 291static inline void init_gbpages(void)
 292{
 293}
 294static void __init cleanup_highmap(void)
 295{
 296}
 297#endif
 298
 299static void __init reserve_brk(void)
 300{
 301	if (_brk_end > _brk_start)
 302		memblock_reserve(__pa(_brk_start),
 303				 __pa(_brk_end) - __pa(_brk_start));
 304
 305	/* Mark brk area as locked down and no longer taking any
 306	   new allocations */
 307	_brk_start = 0;
 308}
 309
 
 
 310#ifdef CONFIG_BLK_DEV_INITRD
 311
 312#define MAX_MAP_CHUNK	(NR_FIX_BTMAPS << PAGE_SHIFT)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 313static void __init relocate_initrd(void)
 314{
 315	/* Assume only end is not page aligned */
 316	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 317	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 318	u64 area_size     = PAGE_ALIGN(ramdisk_size);
 319	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 320	u64 ramdisk_here;
 321	unsigned long slop, clen, mapaddr;
 322	char *p, *q;
 323
 324	/* We need to move the initrd down into lowmem */
 325	ramdisk_here = memblock_find_in_range(0, end_of_lowmem, area_size,
 326					 PAGE_SIZE);
 327
 328	if (!ramdisk_here)
 
 
 
 
 329		panic("Cannot find place for new RAMDISK of size %lld\n",
 330			 ramdisk_size);
 331
 332	/* Note: this includes all the lowmem currently occupied by
 333	   the initrd, we rely on that fact to keep the data intact. */
 334	memblock_reserve(ramdisk_here, area_size);
 335	initrd_start = ramdisk_here + PAGE_OFFSET;
 336	initrd_end   = initrd_start + ramdisk_size;
 337	printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
 338			 ramdisk_here, ramdisk_here + ramdisk_size - 1);
 339
 340	q = (char *)initrd_start;
 341
 342	/* Copy any lowmem portion of the initrd */
 343	if (ramdisk_image < end_of_lowmem) {
 344		clen = end_of_lowmem - ramdisk_image;
 345		p = (char *)__va(ramdisk_image);
 346		memcpy(q, p, clen);
 347		q += clen;
 348		ramdisk_image += clen;
 349		ramdisk_size  -= clen;
 350	}
 351
 352	/* Copy the highmem portion of the initrd */
 353	while (ramdisk_size) {
 354		slop = ramdisk_image & ~PAGE_MASK;
 355		clen = ramdisk_size;
 356		if (clen > MAX_MAP_CHUNK-slop)
 357			clen = MAX_MAP_CHUNK-slop;
 358		mapaddr = ramdisk_image & PAGE_MASK;
 359		p = early_memremap(mapaddr, clen+slop);
 360		memcpy(q, p+slop, clen);
 361		early_iounmap(p, clen+slop);
 362		q += clen;
 363		ramdisk_image += clen;
 364		ramdisk_size  -= clen;
 365	}
 366	/* high pages is not converted by early_res_to_bootmem */
 367	ramdisk_image = boot_params.hdr.ramdisk_image;
 368	ramdisk_size  = boot_params.hdr.ramdisk_size;
 369	printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
 370		" [mem %#010llx-%#010llx]\n",
 371		ramdisk_image, ramdisk_image + ramdisk_size - 1,
 372		ramdisk_here, ramdisk_here + ramdisk_size - 1);
 373}
 374
 
 
 
 
 
 
 
 
 
 
 
 
 
 375static void __init reserve_initrd(void)
 376{
 377	/* Assume only end is not page aligned */
 378	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 379	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 380	u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 381	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 382
 383	if (!boot_params.hdr.type_of_loader ||
 384	    !ramdisk_image || !ramdisk_size)
 385		return;		/* No initrd provided by bootloader */
 386
 387	initrd_start = 0;
 388
 389	if (ramdisk_size >= (end_of_lowmem>>1)) {
 
 390		panic("initrd too large to handle, "
 391		       "disabling initrd (%lld needed, %lld available)\n",
 392		       ramdisk_size, end_of_lowmem>>1);
 393	}
 394
 395	printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
 396			ramdisk_end - 1);
 397
 398
 399	if (ramdisk_end <= end_of_lowmem) {
 400		/* All in lowmem, easy case */
 401		/*
 402		 * don't need to reserve again, already reserved early
 403		 * in i386_start_kernel
 404		 */
 405		initrd_start = ramdisk_image + PAGE_OFFSET;
 406		initrd_end = initrd_start + ramdisk_size;
 407		return;
 408	}
 409
 410	relocate_initrd();
 411
 412	memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
 413}
 
 414#else
 
 
 
 415static void __init reserve_initrd(void)
 416{
 417}
 418#endif /* CONFIG_BLK_DEV_INITRD */
 419
 420static void __init parse_setup_data(void)
 421{
 422	struct setup_data *data;
 423	u64 pa_data;
 424
 425	if (boot_params.hdr.version < 0x0209)
 426		return;
 427	pa_data = boot_params.hdr.setup_data;
 428	while (pa_data) {
 429		u32 data_len, map_len;
 430
 431		map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
 432			      (u64)sizeof(struct setup_data));
 433		data = early_memremap(pa_data, map_len);
 434		data_len = data->len + sizeof(struct setup_data);
 435		if (data_len > map_len) {
 436			early_iounmap(data, map_len);
 437			data = early_memremap(pa_data, data_len);
 438			map_len = data_len;
 439		}
 440
 441		switch (data->type) {
 442		case SETUP_E820_EXT:
 443			parse_e820_ext(data);
 444			break;
 445		case SETUP_DTB:
 446			add_dtb(pa_data);
 447			break;
 
 
 
 448		default:
 449			break;
 450		}
 451		pa_data = data->next;
 452		early_iounmap(data, map_len);
 453	}
 454}
 455
 456static void __init e820_reserve_setup_data(void)
 457{
 458	struct setup_data *data;
 459	u64 pa_data;
 460	int found = 0;
 461
 462	if (boot_params.hdr.version < 0x0209)
 463		return;
 464	pa_data = boot_params.hdr.setup_data;
 465	while (pa_data) {
 466		data = early_memremap(pa_data, sizeof(*data));
 467		e820_update_range(pa_data, sizeof(*data)+data->len,
 468			 E820_RAM, E820_RESERVED_KERN);
 469		found = 1;
 470		pa_data = data->next;
 471		early_iounmap(data, sizeof(*data));
 472	}
 473	if (!found)
 474		return;
 475
 476	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 477	memcpy(&e820_saved, &e820, sizeof(struct e820map));
 478	printk(KERN_INFO "extended physical RAM map:\n");
 479	e820_print_map("reserve setup_data");
 480}
 481
 482static void __init memblock_x86_reserve_range_setup_data(void)
 483{
 484	struct setup_data *data;
 485	u64 pa_data;
 486
 487	if (boot_params.hdr.version < 0x0209)
 488		return;
 489	pa_data = boot_params.hdr.setup_data;
 490	while (pa_data) {
 491		data = early_memremap(pa_data, sizeof(*data));
 492		memblock_reserve(pa_data, sizeof(*data) + data->len);
 
 
 
 
 
 
 493		pa_data = data->next;
 494		early_iounmap(data, sizeof(*data));
 495	}
 496}
 497
 498/*
 499 * --------- Crashkernel reservation ------------------------------
 500 */
 501
 502#ifdef CONFIG_KEXEC
 
 
 
 503
 504/*
 505 * Keep the crash kernel below this limit.  On 32 bits earlier kernels
 506 * would limit the kernel to the low 512 MiB due to mapping restrictions.
 507 * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this
 508 * limit once kexec-tools are fixed.
 
 
 
 
 
 
 
 509 */
 510#ifdef CONFIG_X86_32
 511# define CRASH_KERNEL_ADDR_MAX	(512 << 20)
 
 512#else
 513# define CRASH_KERNEL_ADDR_MAX	(896 << 20)
 
 514#endif
 515
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 516static void __init reserve_crashkernel(void)
 517{
 518	unsigned long long total_mem;
 519	unsigned long long crash_size, crash_base;
 520	int ret;
 521
 522	total_mem = memblock_phys_mem_size();
 523
 524	ret = parse_crashkernel(boot_command_line, total_mem,
 525			&crash_size, &crash_base);
 526	if (ret != 0 || crash_size <= 0)
 
 
 
 
 
 
 
 
 
 
 527		return;
 
 528
 529	/* 0 means: find the address automatically */
 530	if (crash_base <= 0) {
 531		const unsigned long long alignment = 16<<20;	/* 16M */
 532
 533		/*
 534		 *  kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
 
 
 
 
 
 535		 */
 536		crash_base = memblock_find_in_range(alignment,
 537			       CRASH_KERNEL_ADDR_MAX, crash_size, alignment);
 538
 
 
 
 
 
 539		if (!crash_base) {
 540			pr_info("crashkernel reservation failed - No suitable area found.\n");
 541			return;
 542		}
 543	} else {
 544		unsigned long long start;
 545
 546		start = memblock_find_in_range(crash_base,
 547				 crash_base + crash_size, crash_size, 1<<20);
 
 548		if (start != crash_base) {
 549			pr_info("crashkernel reservation failed - memory is in use.\n");
 550			return;
 551		}
 552	}
 553	memblock_reserve(crash_base, crash_size);
 
 
 
 
 
 
 
 
 
 554
 555	printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
 556			"for crashkernel (System RAM: %ldMB)\n",
 557			(unsigned long)(crash_size >> 20),
 558			(unsigned long)(crash_base >> 20),
 559			(unsigned long)(total_mem >> 20));
 560
 561	crashk_res.start = crash_base;
 562	crashk_res.end   = crash_base + crash_size - 1;
 563	insert_resource(&iomem_resource, &crashk_res);
 564}
 565#else
 566static void __init reserve_crashkernel(void)
 567{
 568}
 569#endif
 570
 571static struct resource standard_io_resources[] = {
 572	{ .name = "dma1", .start = 0x00, .end = 0x1f,
 573		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 574	{ .name = "pic1", .start = 0x20, .end = 0x21,
 575		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 576	{ .name = "timer0", .start = 0x40, .end = 0x43,
 577		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 578	{ .name = "timer1", .start = 0x50, .end = 0x53,
 579		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 580	{ .name = "keyboard", .start = 0x60, .end = 0x60,
 581		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 582	{ .name = "keyboard", .start = 0x64, .end = 0x64,
 583		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 584	{ .name = "dma page reg", .start = 0x80, .end = 0x8f,
 585		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 586	{ .name = "pic2", .start = 0xa0, .end = 0xa1,
 587		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 588	{ .name = "dma2", .start = 0xc0, .end = 0xdf,
 589		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 590	{ .name = "fpu", .start = 0xf0, .end = 0xff,
 591		.flags = IORESOURCE_BUSY | IORESOURCE_IO }
 592};
 593
 594void __init reserve_standard_io_resources(void)
 595{
 596	int i;
 597
 598	/* request I/O space for devices used on all i[345]86 PCs */
 599	for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
 600		request_resource(&ioport_resource, &standard_io_resources[i]);
 601
 602}
 603
 604static __init void reserve_ibft_region(void)
 605{
 606	unsigned long addr, size = 0;
 607
 608	addr = find_ibft_region(&size);
 609
 610	if (size)
 611		memblock_reserve(addr, size);
 612}
 613
 614static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 615
 616static void __init trim_bios_range(void)
 617{
 618	/*
 619	 * A special case is the first 4Kb of memory;
 620	 * This is a BIOS owned area, not kernel ram, but generally
 621	 * not listed as such in the E820 table.
 622	 *
 623	 * This typically reserves additional memory (64KiB by default)
 624	 * since some BIOSes are known to corrupt low memory.  See the
 625	 * Kconfig help text for X86_RESERVE_LOW.
 626	 */
 627	e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
 628			  E820_RAM, E820_RESERVED);
 629
 630	/*
 631	 * special case: Some BIOSen report the PC BIOS
 632	 * area (640->1Mb) as ram even though it is not.
 633	 * take them out.
 634	 */
 635	e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
 636	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 
 637}
 638
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 639static int __init parse_reservelow(char *p)
 640{
 641	unsigned long long size;
 642
 643	if (!p)
 644		return -EINVAL;
 645
 646	size = memparse(p, &p);
 647
 648	if (size < 4096)
 649		size = 4096;
 650
 651	if (size > 640*1024)
 652		size = 640*1024;
 653
 654	reserve_low = size;
 655
 656	return 0;
 657}
 658
 659early_param("reservelow", parse_reservelow);
 660
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 661/*
 662 * Determine if we were loaded by an EFI loader.  If so, then we have also been
 663 * passed the efi memmap, systab, etc., so we should use these data structures
 664 * for initialization.  Note, the efi init code path is determined by the
 665 * global efi_enabled. This allows the same kernel image to be used on existing
 666 * systems (with a traditional BIOS) as well as on EFI systems.
 667 */
 668/*
 669 * setup_arch - architecture-specific boot-time initializations
 670 *
 671 * Note: On x86_64, fixmaps are ready for use even before this is called.
 672 */
 673
 674void __init setup_arch(char **cmdline_p)
 675{
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 676#ifdef CONFIG_X86_32
 677	memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
 678	visws_early_detect();
 679
 680	/*
 681	 * copy kernel address range established so far and switch
 682	 * to the proper swapper page table
 683	 */
 684	clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
 685			initial_page_table + KERNEL_PGD_BOUNDARY,
 686			KERNEL_PGD_PTRS);
 687
 688	load_cr3(swapper_pg_dir);
 
 
 
 
 
 
 
 
 
 689	__flush_tlb_all();
 690#else
 691	printk(KERN_INFO "Command line: %s\n", boot_command_line);
 
 692#endif
 693
 694	/*
 695	 * If we have OLPC OFW, we might end up relocating the fixmap due to
 696	 * reserve_top(), so do this before touching the ioremap area.
 697	 */
 698	olpc_ofw_detect();
 699
 700	early_trap_init();
 701	early_cpu_init();
 
 
 702	early_ioremap_init();
 703
 704	setup_olpc_ofw_pgd();
 705
 706	ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
 707	screen_info = boot_params.screen_info;
 708	edid_info = boot_params.edid_info;
 709#ifdef CONFIG_X86_32
 710	apm_info.bios = boot_params.apm_bios_info;
 711	ist_info = boot_params.ist_info;
 712	if (boot_params.sys_desc_table.length != 0) {
 713		machine_id = boot_params.sys_desc_table.table[0];
 714		machine_submodel_id = boot_params.sys_desc_table.table[1];
 715		BIOS_revision = boot_params.sys_desc_table.table[2];
 716	}
 717#endif
 718	saved_video_mode = boot_params.hdr.vid_mode;
 719	bootloader_type = boot_params.hdr.type_of_loader;
 720	if ((bootloader_type >> 4) == 0xe) {
 721		bootloader_type &= 0xf;
 722		bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
 723	}
 724	bootloader_version  = bootloader_type & 0xf;
 725	bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
 726
 727#ifdef CONFIG_BLK_DEV_RAM
 728	rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
 729	rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
 730	rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 731#endif
 732#ifdef CONFIG_EFI
 733	if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 734		     "EL32", 4)) {
 735		efi_enabled = 1;
 736		efi_64bit = false;
 737	} else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 738		     "EL64", 4)) {
 739		efi_enabled = 1;
 740		efi_64bit = true;
 741	}
 742	if (efi_enabled && efi_memblock_x86_reserve_range())
 743		efi_enabled = 0;
 744#endif
 745
 746	x86_init.oem.arch_setup();
 747
 748	iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
 749	setup_memory_map();
 750	parse_setup_data();
 751	/* update the e820_saved too */
 752	e820_reserve_setup_data();
 753
 754	copy_edd();
 755
 756	if (!boot_params.hdr.root_flags)
 757		root_mountflags &= ~MS_RDONLY;
 758	init_mm.start_code = (unsigned long) _text;
 759	init_mm.end_code = (unsigned long) _etext;
 760	init_mm.end_data = (unsigned long) _edata;
 761	init_mm.brk = _brk_end;
 762
 763	code_resource.start = virt_to_phys(_text);
 764	code_resource.end = virt_to_phys(_etext)-1;
 765	data_resource.start = virt_to_phys(_etext);
 766	data_resource.end = virt_to_phys(_edata)-1;
 767	bss_resource.start = virt_to_phys(&__bss_start);
 768	bss_resource.end = virt_to_phys(&__bss_stop)-1;
 
 
 769
 770#ifdef CONFIG_CMDLINE_BOOL
 771#ifdef CONFIG_CMDLINE_OVERRIDE
 772	strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 773#else
 774	if (builtin_cmdline[0]) {
 775		/* append boot loader cmdline to builtin */
 776		strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
 777		strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 778		strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 779	}
 780#endif
 781#endif
 782
 783	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 784	*cmdline_p = command_line;
 785
 786	/*
 787	 * x86_configure_nx() is called before parse_early_param() to detect
 788	 * whether hardware doesn't support NX (so that the early EHCI debug
 789	 * console setup can safely call set_fixmap()). It may then be called
 790	 * again from within noexec_setup() during parsing early parameters
 791	 * to honor the respective command line option.
 792	 */
 793	x86_configure_nx();
 794
 795	parse_early_param();
 796
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 797	x86_report_nx();
 798
 799	/* after early param, so could get panic from serial */
 800	memblock_x86_reserve_range_setup_data();
 801
 802	if (acpi_mps_check()) {
 803#ifdef CONFIG_X86_LOCAL_APIC
 804		disable_apic = 1;
 805#endif
 806		setup_clear_cpu_cap(X86_FEATURE_APIC);
 807	}
 808
 809#ifdef CONFIG_PCI
 810	if (pci_early_dump_regs)
 811		early_dump_pci_devices();
 812#endif
 813
 814	finish_e820_parsing();
 815
 816	if (efi_enabled)
 817		efi_init();
 818
 819	dmi_scan_machine();
 820
 821	/*
 822	 * VMware detection requires dmi to be available, so this
 823	 * needs to be done after dmi_scan_machine, for the BP.
 824	 */
 825	init_hypervisor_platform();
 826
 
 827	x86_init.resources.probe_roms();
 828
 829	/* after parse_early_param, so could debug it */
 830	insert_resource(&iomem_resource, &code_resource);
 
 831	insert_resource(&iomem_resource, &data_resource);
 832	insert_resource(&iomem_resource, &bss_resource);
 833
 
 834	trim_bios_range();
 835#ifdef CONFIG_X86_32
 836	if (ppro_with_ram_bug()) {
 837		e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
 838				  E820_RESERVED);
 839		sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 840		printk(KERN_INFO "fixed physical RAM map:\n");
 841		e820_print_map("bad_ppro");
 842	}
 843#else
 844	early_gart_iommu_check();
 845#endif
 846
 847	/*
 848	 * partially used pages are not usable - thus
 849	 * we are rounding upwards:
 850	 */
 851	max_pfn = e820_end_of_ram_pfn();
 852
 853	/* update e820 for memory not covered by WB MTRRs */
 854	mtrr_bp_init();
 855	if (mtrr_trim_uncached_memory(max_pfn))
 856		max_pfn = e820_end_of_ram_pfn();
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 857
 858#ifdef CONFIG_X86_32
 859	/* max_low_pfn get updated here */
 860	find_low_pfn_range();
 861#else
 862	num_physpages = max_pfn;
 863
 864	check_x2apic();
 865
 866	/* How many end-of-memory variables you have, grandma! */
 867	/* need this before calling reserve_initrd */
 868	if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
 869		max_low_pfn = e820_end_of_low_ram_pfn();
 870	else
 871		max_low_pfn = max_pfn;
 872
 873	high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
 874#endif
 875
 876	/*
 877	 * Find and reserve possible boot-time SMP configuration:
 878	 */
 879	find_smp_config();
 880
 881	reserve_ibft_region();
 882
 
 
 883	/*
 884	 * Need to conclude brk, before memblock_x86_fill()
 885	 *  it could use memblock_find_in_range, could overlap with
 886	 *  brk area.
 887	 */
 888	reserve_brk();
 889
 890	cleanup_highmap();
 891
 892	memblock.current_limit = get_max_mapped();
 893	memblock_x86_fill();
 
 
 
 
 
 
 894
 895	/*
 896	 * The EFI specification says that boot service code won't be called
 897	 * after ExitBootServices(). This is, in fact, a lie.
 898	 */
 899	if (efi_enabled)
 900		efi_reserve_boot_services();
 901
 902	/* preallocate 4k for mptable mpc */
 903	early_reserve_e820_mpc_new();
 904
 905#ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
 906	setup_bios_corruption_check();
 907#endif
 908
 
 909	printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
 910			(max_pfn_mapped<<PAGE_SHIFT) - 1);
 
 911
 912	setup_real_mode();
 913
 914	init_gbpages();
 
 915
 916	/* max_pfn_mapped is updated here */
 917	max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
 918	max_pfn_mapped = max_low_pfn_mapped;
 919
 920#ifdef CONFIG_X86_64
 921	if (max_pfn > max_low_pfn) {
 922		max_pfn_mapped = init_memory_mapping(1UL<<32,
 923						     max_pfn<<PAGE_SHIFT);
 924		/* can we preseve max_low_pfn ?*/
 925		max_low_pfn = max_pfn;
 926	}
 927#endif
 928	memblock.current_limit = get_max_mapped();
 929	dma_contiguous_reserve(0);
 
 
 
 
 930
 931	/*
 932	 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
 933	 */
 934
 935#ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
 936	if (init_ohci1394_dma_early)
 937		init_ohci1394_dma_on_all_controllers();
 938#endif
 939	/* Allocate bigger log buffer */
 940	setup_log_buf(1);
 941
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 942	reserve_initrd();
 943
 944	reserve_crashkernel();
 945
 946	vsmp_init();
 947
 948	io_delay_init();
 949
 
 
 950	/*
 951	 * Parse the ACPI tables for possible boot-time SMP configuration.
 952	 */
 953	acpi_boot_table_init();
 954
 955	early_acpi_boot_init();
 956
 957	initmem_init();
 
 
 
 
 
 
 
 
 
 
 
 958	memblock_find_dma_reserve();
 959
 960#ifdef CONFIG_KVM_CLOCK
 961	kvmclock_init();
 962#endif
 963
 964	x86_init.paging.pagetable_setup_start(swapper_pg_dir);
 965	paging_init();
 966	x86_init.paging.pagetable_setup_done(swapper_pg_dir);
 967
 968	if (boot_cpu_data.cpuid_level >= 0) {
 969		/* A CPU has %cr4 if and only if it has CPUID */
 970		mmu_cr4_features = read_cr4();
 971		if (trampoline_cr4_features)
 972			*trampoline_cr4_features = mmu_cr4_features;
 973	}
 974
 975#ifdef CONFIG_X86_32
 976	/* sync back kernel address range */
 977	clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
 978			swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
 979			KERNEL_PGD_PTRS);
 980#endif
 
 
 
 981
 982	tboot_probe();
 983
 984#ifdef CONFIG_X86_64
 985	map_vsyscall();
 986#endif
 987
 988	generic_apic_probe();
 989
 990	early_quirks();
 991
 992	/*
 993	 * Read APIC and some other early information from ACPI tables.
 994	 */
 995	acpi_boot_init();
 996	sfi_init();
 997	x86_dtb_init();
 998
 999	/*
1000	 * get boot-time SMP configuration:
1001	 */
1002	if (smp_found_config)
1003		get_smp_config();
 
 
 
 
 
1004
1005	prefill_possible_map();
1006
1007	init_cpu_to_node();
1008
1009	init_apic_mappings();
1010	if (x86_io_apic_ops.init)
1011		x86_io_apic_ops.init();
1012
1013	kvm_guest_init();
1014
1015	e820_reserve_resources();
1016	e820_mark_nosave_regions(max_low_pfn);
1017
1018	x86_init.resources.reserve_resources();
1019
1020	e820_setup_gap();
1021
1022#ifdef CONFIG_VT
1023#if defined(CONFIG_VGA_CONSOLE)
1024	if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1025		conswitchp = &vga_con;
1026#elif defined(CONFIG_DUMMY_CONSOLE)
1027	conswitchp = &dummy_con;
1028#endif
1029#endif
1030	x86_init.oem.banner();
1031
1032	x86_init.timers.wallclock_init();
1033
1034	x86_platform.wallclock_init();
1035
1036	mcheck_init();
1037
1038	arch_init_ideal_nops();
 
 
 
 
 
 
 
1039}
1040
1041#ifdef CONFIG_X86_32
1042
1043static struct resource video_ram_resource = {
1044	.name	= "Video RAM area",
1045	.start	= 0xa0000,
1046	.end	= 0xbffff,
1047	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
1048};
1049
1050void __init i386_reserve_resources(void)
1051{
1052	request_resource(&iomem_resource, &video_ram_resource);
1053	reserve_standard_io_resources();
1054}
1055
1056#endif /* CONFIG_X86_32 */
v5.9
   1// SPDX-License-Identifier: GPL-2.0-only
   2/*
   3 *  Copyright (C) 1995  Linus Torvalds
   4 *
   5 * This file contains the setup_arch() code, which handles the architecture-dependent
   6 * parts of early kernel initialization.
 
 
 
 
 
 
 
 
 
 
 
 
   7 */
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
   8#include <linux/console.h>
   9#include <linux/crash_dump.h>
  10#include <linux/dmi.h>
 
  11#include <linux/efi.h>
  12#include <linux/init_ohci1394_dma.h>
  13#include <linux/initrd.h>
  14#include <linux/iscsi_ibft.h>
  15#include <linux/memblock.h>
 
 
 
  16#include <linux/pci.h>
  17#include <linux/root_dev.h>
  18#include <linux/sfi.h>
  19#include <linux/hugetlb.h>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  20#include <linux/tboot.h>
  21#include <linux/usb/xhci-dbgp.h>
  22
  23#include <uapi/linux/mount.h>
  24
  25#include <xen/xen.h>
  26
 
  27#include <asm/apic.h>
  28#include <asm/numa.h>
 
 
 
 
 
 
 
 
 
 
 
  29#include <asm/bios_ebda.h>
 
 
  30#include <asm/bugs.h>
 
 
  31#include <asm/cpu.h>
  32#include <asm/efi.h>
 
 
  33#include <asm/gart.h>
 
 
 
 
  34#include <asm/hypervisor.h>
  35#include <asm/io_apic.h>
  36#include <asm/kasan.h>
  37#include <asm/kaslr.h>
 
 
 
 
 
 
  38#include <asm/mce.h>
  39#include <asm/mtrr.h>
  40#include <asm/realmode.h>
  41#include <asm/olpc_ofw.h>
  42#include <asm/pci-direct.h>
  43#include <asm/prom.h>
  44#include <asm/proto.h>
  45#include <asm/unwind.h>
  46#include <asm/vsyscall.h>
  47#include <linux/vmalloc.h>
  48
  49/*
  50 * max_low_pfn_mapped: highest directly mapped pfn < 4 GB
  51 * max_pfn_mapped:     highest directly mapped pfn > 4 GB
  52 *
  53 * The direct mapping only covers E820_TYPE_RAM regions, so the ranges and gaps are
  54 * represented by pfn_mapped[].
  55 */
  56unsigned long max_low_pfn_mapped;
  57unsigned long max_pfn_mapped;
  58
  59#ifdef CONFIG_DMI
  60RESERVE_BRK(dmi_alloc, 65536);
  61#endif
  62
  63
  64/*
  65 * Range of the BSS area. The size of the BSS area is determined
  66 * at link time, with RESERVE_BRK*() facility reserving additional
  67 * chunks.
  68 */
  69unsigned long _brk_start = (unsigned long)__brk_base;
  70unsigned long _brk_end   = (unsigned long)__brk_base;
 
 
 
 
 
 
 
  71
 
 
 
  72struct boot_params boot_params;
 
  73
  74/*
  75 * These are the four main kernel memory regions, we put them into
  76 * the resource tree so that kdump tools and other debugging tools
  77 * recover it:
  78 */
  79
  80static struct resource rodata_resource = {
  81	.name	= "Kernel rodata",
  82	.start	= 0,
  83	.end	= 0,
  84	.flags	= IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
  85};
  86
  87static struct resource data_resource = {
  88	.name	= "Kernel data",
  89	.start	= 0,
  90	.end	= 0,
  91	.flags	= IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
  92};
  93
  94static struct resource code_resource = {
  95	.name	= "Kernel code",
  96	.start	= 0,
  97	.end	= 0,
  98	.flags	= IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
  99};
 100
 101static struct resource bss_resource = {
 102	.name	= "Kernel bss",
 103	.start	= 0,
 104	.end	= 0,
 105	.flags	= IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
 106};
 107
 108
 109#ifdef CONFIG_X86_32
 110/* CPU data as detected by the assembly code in head_32.S */
 111struct cpuinfo_x86 new_cpu_data;
 112
 113/* Common CPU data for all CPUs */
 114struct cpuinfo_x86 boot_cpu_data __read_mostly;
 115EXPORT_SYMBOL(boot_cpu_data);
 116
 117unsigned int def_to_bigsmp;
 118
 119/* For MCA, but anyone else can use it if they want */
 120unsigned int machine_id;
 121unsigned int machine_submodel_id;
 122unsigned int BIOS_revision;
 123
 124struct apm_info apm_info;
 125EXPORT_SYMBOL(apm_info);
 126
 127#if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
 128	defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
 129struct ist_info ist_info;
 130EXPORT_SYMBOL(ist_info);
 131#else
 132struct ist_info ist_info;
 133#endif
 134
 135#else
 136struct cpuinfo_x86 boot_cpu_data __read_mostly;
 
 
 137EXPORT_SYMBOL(boot_cpu_data);
 138#endif
 139
 140
 141#if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
 142__visible unsigned long mmu_cr4_features __ro_after_init;
 143#else
 144__visible unsigned long mmu_cr4_features __ro_after_init = X86_CR4_PAE;
 145#endif
 146
 147/* Boot loader ID and version as integers, for the benefit of proc_dointvec */
 148int bootloader_type, bootloader_version;
 149
 150/*
 151 * Setup options
 152 */
 153struct screen_info screen_info;
 154EXPORT_SYMBOL(screen_info);
 155struct edid_info edid_info;
 156EXPORT_SYMBOL_GPL(edid_info);
 157
 158extern int root_mountflags;
 159
 160unsigned long saved_video_mode;
 161
 162#define RAMDISK_IMAGE_START_MASK	0x07FF
 163#define RAMDISK_PROMPT_FLAG		0x8000
 164#define RAMDISK_LOAD_FLAG		0x4000
 165
 166static char __initdata command_line[COMMAND_LINE_SIZE];
 167#ifdef CONFIG_CMDLINE_BOOL
 168static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
 169#endif
 170
 171#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
 172struct edd edd;
 173#ifdef CONFIG_EDD_MODULE
 174EXPORT_SYMBOL(edd);
 175#endif
 176/**
 177 * copy_edd() - Copy the BIOS EDD information
 178 *              from boot_params into a safe place.
 179 *
 180 */
 181static inline void __init copy_edd(void)
 182{
 183     memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
 184	    sizeof(edd.mbr_signature));
 185     memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
 186     edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
 187     edd.edd_info_nr = boot_params.eddbuf_entries;
 188}
 189#else
 190static inline void __init copy_edd(void)
 191{
 192}
 193#endif
 194
 195void * __init extend_brk(size_t size, size_t align)
 196{
 197	size_t mask = align - 1;
 198	void *ret;
 199
 200	BUG_ON(_brk_start == 0);
 201	BUG_ON(align & mask);
 202
 203	_brk_end = (_brk_end + mask) & ~mask;
 204	BUG_ON((char *)(_brk_end + size) > __brk_limit);
 205
 206	ret = (void *)_brk_end;
 207	_brk_end += size;
 208
 209	memset(ret, 0, size);
 210
 211	return ret;
 212}
 213
 214#ifdef CONFIG_X86_32
 
 
 
 
 
 
 
 
 
 
 
 215static void __init cleanup_highmap(void)
 216{
 217}
 218#endif
 219
 220static void __init reserve_brk(void)
 221{
 222	if (_brk_end > _brk_start)
 223		memblock_reserve(__pa_symbol(_brk_start),
 224				 _brk_end - _brk_start);
 225
 226	/* Mark brk area as locked down and no longer taking any
 227	   new allocations */
 228	_brk_start = 0;
 229}
 230
 231u64 relocated_ramdisk;
 232
 233#ifdef CONFIG_BLK_DEV_INITRD
 234
 235static u64 __init get_ramdisk_image(void)
 236{
 237	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 238
 239	ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
 240
 241	if (ramdisk_image == 0)
 242		ramdisk_image = phys_initrd_start;
 243
 244	return ramdisk_image;
 245}
 246static u64 __init get_ramdisk_size(void)
 247{
 248	u64 ramdisk_size = boot_params.hdr.ramdisk_size;
 249
 250	ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
 251
 252	if (ramdisk_size == 0)
 253		ramdisk_size = phys_initrd_size;
 254
 255	return ramdisk_size;
 256}
 257
 258static void __init relocate_initrd(void)
 259{
 260	/* Assume only end is not page aligned */
 261	u64 ramdisk_image = get_ramdisk_image();
 262	u64 ramdisk_size  = get_ramdisk_size();
 263	u64 area_size     = PAGE_ALIGN(ramdisk_size);
 
 
 
 
 
 
 
 
 264
 265	/* We need to move the initrd down into directly mapped mem */
 266	relocated_ramdisk = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
 267						   area_size, PAGE_SIZE);
 268
 269	if (!relocated_ramdisk)
 270		panic("Cannot find place for new RAMDISK of size %lld\n",
 271		      ramdisk_size);
 272
 273	/* Note: this includes all the mem currently occupied by
 274	   the initrd, we rely on that fact to keep the data intact. */
 275	memblock_reserve(relocated_ramdisk, area_size);
 276	initrd_start = relocated_ramdisk + PAGE_OFFSET;
 277	initrd_end   = initrd_start + ramdisk_size;
 278	printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
 279	       relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
 
 
 280
 281	copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size);
 
 
 
 
 
 
 
 
 282
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 283	printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
 284		" [mem %#010llx-%#010llx]\n",
 285		ramdisk_image, ramdisk_image + ramdisk_size - 1,
 286		relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
 287}
 288
 289static void __init early_reserve_initrd(void)
 290{
 291	/* Assume only end is not page aligned */
 292	u64 ramdisk_image = get_ramdisk_image();
 293	u64 ramdisk_size  = get_ramdisk_size();
 294	u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 295
 296	if (!boot_params.hdr.type_of_loader ||
 297	    !ramdisk_image || !ramdisk_size)
 298		return;		/* No initrd provided by bootloader */
 299
 300	memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
 301}
 302static void __init reserve_initrd(void)
 303{
 304	/* Assume only end is not page aligned */
 305	u64 ramdisk_image = get_ramdisk_image();
 306	u64 ramdisk_size  = get_ramdisk_size();
 307	u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 308	u64 mapped_size;
 309
 310	if (!boot_params.hdr.type_of_loader ||
 311	    !ramdisk_image || !ramdisk_size)
 312		return;		/* No initrd provided by bootloader */
 313
 314	initrd_start = 0;
 315
 316	mapped_size = memblock_mem_size(max_pfn_mapped);
 317	if (ramdisk_size >= (mapped_size>>1))
 318		panic("initrd too large to handle, "
 319		       "disabling initrd (%lld needed, %lld available)\n",
 320		       ramdisk_size, mapped_size>>1);
 
 321
 322	printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
 323			ramdisk_end - 1);
 324
 325	if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
 326				PFN_DOWN(ramdisk_end))) {
 327		/* All are mapped, easy case */
 
 
 
 
 328		initrd_start = ramdisk_image + PAGE_OFFSET;
 329		initrd_end = initrd_start + ramdisk_size;
 330		return;
 331	}
 332
 333	relocate_initrd();
 334
 335	memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
 336}
 337
 338#else
 339static void __init early_reserve_initrd(void)
 340{
 341}
 342static void __init reserve_initrd(void)
 343{
 344}
 345#endif /* CONFIG_BLK_DEV_INITRD */
 346
 347static void __init parse_setup_data(void)
 348{
 349	struct setup_data *data;
 350	u64 pa_data, pa_next;
 351
 
 
 352	pa_data = boot_params.hdr.setup_data;
 353	while (pa_data) {
 354		u32 data_len, data_type;
 355
 356		data = early_memremap(pa_data, sizeof(*data));
 
 
 357		data_len = data->len + sizeof(struct setup_data);
 358		data_type = data->type;
 359		pa_next = data->next;
 360		early_memunmap(data, sizeof(*data));
 
 
 361
 362		switch (data_type) {
 363		case SETUP_E820_EXT:
 364			e820__memory_setup_extended(pa_data, data_len);
 365			break;
 366		case SETUP_DTB:
 367			add_dtb(pa_data);
 368			break;
 369		case SETUP_EFI:
 370			parse_efi_setup(pa_data, data_len);
 371			break;
 372		default:
 373			break;
 374		}
 375		pa_data = pa_next;
 
 376	}
 377}
 378
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 379static void __init memblock_x86_reserve_range_setup_data(void)
 380{
 381	struct setup_data *data;
 382	u64 pa_data;
 383
 
 
 384	pa_data = boot_params.hdr.setup_data;
 385	while (pa_data) {
 386		data = early_memremap(pa_data, sizeof(*data));
 387		memblock_reserve(pa_data, sizeof(*data) + data->len);
 388
 389		if (data->type == SETUP_INDIRECT &&
 390		    ((struct setup_indirect *)data->data)->type != SETUP_INDIRECT)
 391			memblock_reserve(((struct setup_indirect *)data->data)->addr,
 392					 ((struct setup_indirect *)data->data)->len);
 393
 394		pa_data = data->next;
 395		early_memunmap(data, sizeof(*data));
 396	}
 397}
 398
 399/*
 400 * --------- Crashkernel reservation ------------------------------
 401 */
 402
 403#ifdef CONFIG_KEXEC_CORE
 404
 405/* 16M alignment for crash kernel regions */
 406#define CRASH_ALIGN		SZ_16M
 407
 408/*
 409 * Keep the crash kernel below this limit.
 410 *
 411 * Earlier 32-bits kernels would limit the kernel to the low 512 MB range
 412 * due to mapping restrictions.
 413 *
 414 * 64-bit kdump kernels need to be restricted to be under 64 TB, which is
 415 * the upper limit of system RAM in 4-level paging mode. Since the kdump
 416 * jump could be from 5-level paging to 4-level paging, the jump will fail if
 417 * the kernel is put above 64 TB, and during the 1st kernel bootup there's
 418 * no good way to detect the paging mode of the target kernel which will be
 419 * loaded for dumping.
 420 */
 421#ifdef CONFIG_X86_32
 422# define CRASH_ADDR_LOW_MAX	SZ_512M
 423# define CRASH_ADDR_HIGH_MAX	SZ_512M
 424#else
 425# define CRASH_ADDR_LOW_MAX	SZ_4G
 426# define CRASH_ADDR_HIGH_MAX	SZ_64T
 427#endif
 428
 429static int __init reserve_crashkernel_low(void)
 430{
 431#ifdef CONFIG_X86_64
 432	unsigned long long base, low_base = 0, low_size = 0;
 433	unsigned long total_low_mem;
 434	int ret;
 435
 436	total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT));
 437
 438	/* crashkernel=Y,low */
 439	ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base);
 440	if (ret) {
 441		/*
 442		 * two parts from kernel/dma/swiotlb.c:
 443		 * -swiotlb size: user-specified with swiotlb= or default.
 444		 *
 445		 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
 446		 * to 8M for other buffers that may need to stay low too. Also
 447		 * make sure we allocate enough extra low memory so that we
 448		 * don't run out of DMA buffers for 32-bit devices.
 449		 */
 450		low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
 451	} else {
 452		/* passed with crashkernel=0,low ? */
 453		if (!low_size)
 454			return 0;
 455	}
 456
 457	low_base = memblock_find_in_range(0, 1ULL << 32, low_size, CRASH_ALIGN);
 458	if (!low_base) {
 459		pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
 460		       (unsigned long)(low_size >> 20));
 461		return -ENOMEM;
 462	}
 463
 464	ret = memblock_reserve(low_base, low_size);
 465	if (ret) {
 466		pr_err("%s: Error reserving crashkernel low memblock.\n", __func__);
 467		return ret;
 468	}
 469
 470	pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
 471		(unsigned long)(low_size >> 20),
 472		(unsigned long)(low_base >> 20),
 473		(unsigned long)(total_low_mem >> 20));
 474
 475	crashk_low_res.start = low_base;
 476	crashk_low_res.end   = low_base + low_size - 1;
 477	insert_resource(&iomem_resource, &crashk_low_res);
 478#endif
 479	return 0;
 480}
 481
 482static void __init reserve_crashkernel(void)
 483{
 484	unsigned long long crash_size, crash_base, total_mem;
 485	bool high = false;
 486	int ret;
 487
 488	total_mem = memblock_phys_mem_size();
 489
 490	/* crashkernel=XM */
 491	ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base);
 492	if (ret != 0 || crash_size <= 0) {
 493		/* crashkernel=X,high */
 494		ret = parse_crashkernel_high(boot_command_line, total_mem,
 495					     &crash_size, &crash_base);
 496		if (ret != 0 || crash_size <= 0)
 497			return;
 498		high = true;
 499	}
 500
 501	if (xen_pv_domain()) {
 502		pr_info("Ignoring crashkernel for a Xen PV domain\n");
 503		return;
 504	}
 505
 506	/* 0 means: find the address automatically */
 507	if (!crash_base) {
 
 
 508		/*
 509		 * Set CRASH_ADDR_LOW_MAX upper bound for crash memory,
 510		 * crashkernel=x,high reserves memory over 4G, also allocates
 511		 * 256M extra low memory for DMA buffers and swiotlb.
 512		 * But the extra memory is not required for all machines.
 513		 * So try low memory first and fall back to high memory
 514		 * unless "crashkernel=size[KMG],high" is specified.
 515		 */
 516		if (!high)
 517			crash_base = memblock_find_in_range(CRASH_ALIGN,
 518						CRASH_ADDR_LOW_MAX,
 519						crash_size, CRASH_ALIGN);
 520		if (!crash_base)
 521			crash_base = memblock_find_in_range(CRASH_ALIGN,
 522						CRASH_ADDR_HIGH_MAX,
 523						crash_size, CRASH_ALIGN);
 524		if (!crash_base) {
 525			pr_info("crashkernel reservation failed - No suitable area found.\n");
 526			return;
 527		}
 528	} else {
 529		unsigned long long start;
 530
 531		start = memblock_find_in_range(crash_base,
 532					       crash_base + crash_size,
 533					       crash_size, 1 << 20);
 534		if (start != crash_base) {
 535			pr_info("crashkernel reservation failed - memory is in use.\n");
 536			return;
 537		}
 538	}
 539	ret = memblock_reserve(crash_base, crash_size);
 540	if (ret) {
 541		pr_err("%s: Error reserving crashkernel memblock.\n", __func__);
 542		return;
 543	}
 544
 545	if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) {
 546		memblock_free(crash_base, crash_size);
 547		return;
 548	}
 549
 550	pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
 551		(unsigned long)(crash_size >> 20),
 552		(unsigned long)(crash_base >> 20),
 553		(unsigned long)(total_mem >> 20));
 
 554
 555	crashk_res.start = crash_base;
 556	crashk_res.end   = crash_base + crash_size - 1;
 557	insert_resource(&iomem_resource, &crashk_res);
 558}
 559#else
 560static void __init reserve_crashkernel(void)
 561{
 562}
 563#endif
 564
 565static struct resource standard_io_resources[] = {
 566	{ .name = "dma1", .start = 0x00, .end = 0x1f,
 567		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 568	{ .name = "pic1", .start = 0x20, .end = 0x21,
 569		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 570	{ .name = "timer0", .start = 0x40, .end = 0x43,
 571		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 572	{ .name = "timer1", .start = 0x50, .end = 0x53,
 573		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 574	{ .name = "keyboard", .start = 0x60, .end = 0x60,
 575		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 576	{ .name = "keyboard", .start = 0x64, .end = 0x64,
 577		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 578	{ .name = "dma page reg", .start = 0x80, .end = 0x8f,
 579		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 580	{ .name = "pic2", .start = 0xa0, .end = 0xa1,
 581		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 582	{ .name = "dma2", .start = 0xc0, .end = 0xdf,
 583		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
 584	{ .name = "fpu", .start = 0xf0, .end = 0xff,
 585		.flags = IORESOURCE_BUSY | IORESOURCE_IO }
 586};
 587
 588void __init reserve_standard_io_resources(void)
 589{
 590	int i;
 591
 592	/* request I/O space for devices used on all i[345]86 PCs */
 593	for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
 594		request_resource(&ioport_resource, &standard_io_resources[i]);
 595
 596}
 597
 598static __init void reserve_ibft_region(void)
 599{
 600	unsigned long addr, size = 0;
 601
 602	addr = find_ibft_region(&size);
 603
 604	if (size)
 605		memblock_reserve(addr, size);
 606}
 607
 608static bool __init snb_gfx_workaround_needed(void)
 609{
 610#ifdef CONFIG_PCI
 611	int i;
 612	u16 vendor, devid;
 613	static const __initconst u16 snb_ids[] = {
 614		0x0102,
 615		0x0112,
 616		0x0122,
 617		0x0106,
 618		0x0116,
 619		0x0126,
 620		0x010a,
 621	};
 622
 623	/* Assume no if something weird is going on with PCI */
 624	if (!early_pci_allowed())
 625		return false;
 626
 627	vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
 628	if (vendor != 0x8086)
 629		return false;
 630
 631	devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
 632	for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
 633		if (devid == snb_ids[i])
 634			return true;
 635#endif
 636
 637	return false;
 638}
 639
 640/*
 641 * Sandy Bridge graphics has trouble with certain ranges, exclude
 642 * them from allocation.
 643 */
 644static void __init trim_snb_memory(void)
 645{
 646	static const __initconst unsigned long bad_pages[] = {
 647		0x20050000,
 648		0x20110000,
 649		0x20130000,
 650		0x20138000,
 651		0x40004000,
 652	};
 653	int i;
 654
 655	if (!snb_gfx_workaround_needed())
 656		return;
 657
 658	printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
 659
 660	/*
 661	 * Reserve all memory below the 1 MB mark that has not
 662	 * already been reserved.
 663	 */
 664	memblock_reserve(0, 1<<20);
 665	
 666	for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
 667		if (memblock_reserve(bad_pages[i], PAGE_SIZE))
 668			printk(KERN_WARNING "failed to reserve 0x%08lx\n",
 669			       bad_pages[i]);
 670	}
 671}
 672
 673/*
 674 * Here we put platform-specific memory range workarounds, i.e.
 675 * memory known to be corrupt or otherwise in need to be reserved on
 676 * specific platforms.
 677 *
 678 * If this gets used more widely it could use a real dispatch mechanism.
 679 */
 680static void __init trim_platform_memory_ranges(void)
 681{
 682	trim_snb_memory();
 683}
 684
 685static void __init trim_bios_range(void)
 686{
 687	/*
 688	 * A special case is the first 4Kb of memory;
 689	 * This is a BIOS owned area, not kernel ram, but generally
 690	 * not listed as such in the E820 table.
 691	 *
 692	 * This typically reserves additional memory (64KiB by default)
 693	 * since some BIOSes are known to corrupt low memory.  See the
 694	 * Kconfig help text for X86_RESERVE_LOW.
 695	 */
 696	e820__range_update(0, PAGE_SIZE, E820_TYPE_RAM, E820_TYPE_RESERVED);
 
 697
 698	/*
 699	 * special case: Some BIOSes report the PC BIOS
 700	 * area (640Kb -> 1Mb) as RAM even though it is not.
 701	 * take them out.
 702	 */
 703	e820__range_remove(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_TYPE_RAM, 1);
 704
 705	e820__update_table(e820_table);
 706}
 707
 708/* called before trim_bios_range() to spare extra sanitize */
 709static void __init e820_add_kernel_range(void)
 710{
 711	u64 start = __pa_symbol(_text);
 712	u64 size = __pa_symbol(_end) - start;
 713
 714	/*
 715	 * Complain if .text .data and .bss are not marked as E820_TYPE_RAM and
 716	 * attempt to fix it by adding the range. We may have a confused BIOS,
 717	 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
 718	 * exclude kernel range. If we really are running on top non-RAM,
 719	 * we will crash later anyways.
 720	 */
 721	if (e820__mapped_all(start, start + size, E820_TYPE_RAM))
 722		return;
 723
 724	pr_warn(".text .data .bss are not marked as E820_TYPE_RAM!\n");
 725	e820__range_remove(start, size, E820_TYPE_RAM, 0);
 726	e820__range_add(start, size, E820_TYPE_RAM);
 727}
 728
 729static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
 730
 731static int __init parse_reservelow(char *p)
 732{
 733	unsigned long long size;
 734
 735	if (!p)
 736		return -EINVAL;
 737
 738	size = memparse(p, &p);
 739
 740	if (size < 4096)
 741		size = 4096;
 742
 743	if (size > 640*1024)
 744		size = 640*1024;
 745
 746	reserve_low = size;
 747
 748	return 0;
 749}
 750
 751early_param("reservelow", parse_reservelow);
 752
 753static void __init trim_low_memory_range(void)
 754{
 755	memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
 756}
 757	
 758/*
 759 * Dump out kernel offset information on panic.
 760 */
 761static int
 762dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
 763{
 764	if (kaslr_enabled()) {
 765		pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
 766			 kaslr_offset(),
 767			 __START_KERNEL,
 768			 __START_KERNEL_map,
 769			 MODULES_VADDR-1);
 770	} else {
 771		pr_emerg("Kernel Offset: disabled\n");
 772	}
 773
 774	return 0;
 775}
 776
 777/*
 778 * Determine if we were loaded by an EFI loader.  If so, then we have also been
 779 * passed the efi memmap, systab, etc., so we should use these data structures
 780 * for initialization.  Note, the efi init code path is determined by the
 781 * global efi_enabled. This allows the same kernel image to be used on existing
 782 * systems (with a traditional BIOS) as well as on EFI systems.
 783 */
 784/*
 785 * setup_arch - architecture-specific boot-time initializations
 786 *
 787 * Note: On x86_64, fixmaps are ready for use even before this is called.
 788 */
 789
 790void __init setup_arch(char **cmdline_p)
 791{
 792	/*
 793	 * Reserve the memory occupied by the kernel between _text and
 794	 * __end_of_kernel_reserve symbols. Any kernel sections after the
 795	 * __end_of_kernel_reserve symbol must be explicitly reserved with a
 796	 * separate memblock_reserve() or they will be discarded.
 797	 */
 798	memblock_reserve(__pa_symbol(_text),
 799			 (unsigned long)__end_of_kernel_reserve - (unsigned long)_text);
 800
 801	/*
 802	 * Make sure page 0 is always reserved because on systems with
 803	 * L1TF its contents can be leaked to user processes.
 804	 */
 805	memblock_reserve(0, PAGE_SIZE);
 806
 807	early_reserve_initrd();
 808
 809	/*
 810	 * At this point everything still needed from the boot loader
 811	 * or BIOS or kernel text should be early reserved or marked not
 812	 * RAM in e820. All other memory is free game.
 813	 */
 814
 815#ifdef CONFIG_X86_32
 816	memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
 
 817
 818	/*
 819	 * copy kernel address range established so far and switch
 820	 * to the proper swapper page table
 821	 */
 822	clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
 823			initial_page_table + KERNEL_PGD_BOUNDARY,
 824			KERNEL_PGD_PTRS);
 825
 826	load_cr3(swapper_pg_dir);
 827	/*
 828	 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
 829	 * a cr3 based tlb flush, so the following __flush_tlb_all()
 830	 * will not flush anything because the CPU quirk which clears
 831	 * X86_FEATURE_PGE has not been invoked yet. Though due to the
 832	 * load_cr3() above the TLB has been flushed already. The
 833	 * quirk is invoked before subsequent calls to __flush_tlb_all()
 834	 * so proper operation is guaranteed.
 835	 */
 836	__flush_tlb_all();
 837#else
 838	printk(KERN_INFO "Command line: %s\n", boot_command_line);
 839	boot_cpu_data.x86_phys_bits = MAX_PHYSMEM_BITS;
 840#endif
 841
 842	/*
 843	 * If we have OLPC OFW, we might end up relocating the fixmap due to
 844	 * reserve_top(), so do this before touching the ioremap area.
 845	 */
 846	olpc_ofw_detect();
 847
 848	idt_setup_early_traps();
 849	early_cpu_init();
 850	arch_init_ideal_nops();
 851	jump_label_init();
 852	early_ioremap_init();
 853
 854	setup_olpc_ofw_pgd();
 855
 856	ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
 857	screen_info = boot_params.screen_info;
 858	edid_info = boot_params.edid_info;
 859#ifdef CONFIG_X86_32
 860	apm_info.bios = boot_params.apm_bios_info;
 861	ist_info = boot_params.ist_info;
 
 
 
 
 
 862#endif
 863	saved_video_mode = boot_params.hdr.vid_mode;
 864	bootloader_type = boot_params.hdr.type_of_loader;
 865	if ((bootloader_type >> 4) == 0xe) {
 866		bootloader_type &= 0xf;
 867		bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
 868	}
 869	bootloader_version  = bootloader_type & 0xf;
 870	bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
 871
 872#ifdef CONFIG_BLK_DEV_RAM
 873	rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
 
 
 874#endif
 875#ifdef CONFIG_EFI
 876	if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 877		     EFI32_LOADER_SIGNATURE, 4)) {
 878		set_bit(EFI_BOOT, &efi.flags);
 
 879	} else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 880		     EFI64_LOADER_SIGNATURE, 4)) {
 881		set_bit(EFI_BOOT, &efi.flags);
 882		set_bit(EFI_64BIT, &efi.flags);
 883	}
 
 
 884#endif
 885
 886	x86_init.oem.arch_setup();
 887
 888	iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
 889	e820__memory_setup();
 890	parse_setup_data();
 
 
 891
 892	copy_edd();
 893
 894	if (!boot_params.hdr.root_flags)
 895		root_mountflags &= ~MS_RDONLY;
 896	init_mm.start_code = (unsigned long) _text;
 897	init_mm.end_code = (unsigned long) _etext;
 898	init_mm.end_data = (unsigned long) _edata;
 899	init_mm.brk = _brk_end;
 900
 901	code_resource.start = __pa_symbol(_text);
 902	code_resource.end = __pa_symbol(_etext)-1;
 903	rodata_resource.start = __pa_symbol(__start_rodata);
 904	rodata_resource.end = __pa_symbol(__end_rodata)-1;
 905	data_resource.start = __pa_symbol(_sdata);
 906	data_resource.end = __pa_symbol(_edata)-1;
 907	bss_resource.start = __pa_symbol(__bss_start);
 908	bss_resource.end = __pa_symbol(__bss_stop)-1;
 909
 910#ifdef CONFIG_CMDLINE_BOOL
 911#ifdef CONFIG_CMDLINE_OVERRIDE
 912	strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 913#else
 914	if (builtin_cmdline[0]) {
 915		/* append boot loader cmdline to builtin */
 916		strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
 917		strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 918		strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 919	}
 920#endif
 921#endif
 922
 923	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 924	*cmdline_p = command_line;
 925
 926	/*
 927	 * x86_configure_nx() is called before parse_early_param() to detect
 928	 * whether hardware doesn't support NX (so that the early EHCI debug
 929	 * console setup can safely call set_fixmap()). It may then be called
 930	 * again from within noexec_setup() during parsing early parameters
 931	 * to honor the respective command line option.
 932	 */
 933	x86_configure_nx();
 934
 935	parse_early_param();
 936
 937	if (efi_enabled(EFI_BOOT))
 938		efi_memblock_x86_reserve_range();
 939#ifdef CONFIG_MEMORY_HOTPLUG
 940	/*
 941	 * Memory used by the kernel cannot be hot-removed because Linux
 942	 * cannot migrate the kernel pages. When memory hotplug is
 943	 * enabled, we should prevent memblock from allocating memory
 944	 * for the kernel.
 945	 *
 946	 * ACPI SRAT records all hotpluggable memory ranges. But before
 947	 * SRAT is parsed, we don't know about it.
 948	 *
 949	 * The kernel image is loaded into memory at very early time. We
 950	 * cannot prevent this anyway. So on NUMA system, we set any
 951	 * node the kernel resides in as un-hotpluggable.
 952	 *
 953	 * Since on modern servers, one node could have double-digit
 954	 * gigabytes memory, we can assume the memory around the kernel
 955	 * image is also un-hotpluggable. So before SRAT is parsed, just
 956	 * allocate memory near the kernel image to try the best to keep
 957	 * the kernel away from hotpluggable memory.
 958	 */
 959	if (movable_node_is_enabled())
 960		memblock_set_bottom_up(true);
 961#endif
 962
 963	x86_report_nx();
 964
 965	/* after early param, so could get panic from serial */
 966	memblock_x86_reserve_range_setup_data();
 967
 968	if (acpi_mps_check()) {
 969#ifdef CONFIG_X86_LOCAL_APIC
 970		disable_apic = 1;
 971#endif
 972		setup_clear_cpu_cap(X86_FEATURE_APIC);
 973	}
 974
 975	e820__reserve_setup_data();
 976	e820__finish_early_params();
 
 
 977
 978	if (efi_enabled(EFI_BOOT))
 
 
 979		efi_init();
 980
 981	dmi_setup();
 982
 983	/*
 984	 * VMware detection requires dmi to be available, so this
 985	 * needs to be done after dmi_setup(), for the boot CPU.
 986	 */
 987	init_hypervisor_platform();
 988
 989	tsc_early_init();
 990	x86_init.resources.probe_roms();
 991
 992	/* after parse_early_param, so could debug it */
 993	insert_resource(&iomem_resource, &code_resource);
 994	insert_resource(&iomem_resource, &rodata_resource);
 995	insert_resource(&iomem_resource, &data_resource);
 996	insert_resource(&iomem_resource, &bss_resource);
 997
 998	e820_add_kernel_range();
 999	trim_bios_range();
1000#ifdef CONFIG_X86_32
1001	if (ppro_with_ram_bug()) {
1002		e820__range_update(0x70000000ULL, 0x40000ULL, E820_TYPE_RAM,
1003				  E820_TYPE_RESERVED);
1004		e820__update_table(e820_table);
1005		printk(KERN_INFO "fixed physical RAM map:\n");
1006		e820__print_table("bad_ppro");
1007	}
1008#else
1009	early_gart_iommu_check();
1010#endif
1011
1012	/*
1013	 * partially used pages are not usable - thus
1014	 * we are rounding upwards:
1015	 */
1016	max_pfn = e820__end_of_ram_pfn();
1017
1018	/* update e820 for memory not covered by WB MTRRs */
1019	mtrr_bp_init();
1020	if (mtrr_trim_uncached_memory(max_pfn))
1021		max_pfn = e820__end_of_ram_pfn();
1022
1023	max_possible_pfn = max_pfn;
1024
1025	/*
1026	 * This call is required when the CPU does not support PAT. If
1027	 * mtrr_bp_init() invoked it already via pat_init() the call has no
1028	 * effect.
1029	 */
1030	init_cache_modes();
1031
1032	/*
1033	 * Define random base addresses for memory sections after max_pfn is
1034	 * defined and before each memory section base is used.
1035	 */
1036	kernel_randomize_memory();
1037
1038#ifdef CONFIG_X86_32
1039	/* max_low_pfn get updated here */
1040	find_low_pfn_range();
1041#else
 
 
1042	check_x2apic();
1043
1044	/* How many end-of-memory variables you have, grandma! */
1045	/* need this before calling reserve_initrd */
1046	if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1047		max_low_pfn = e820__end_of_low_ram_pfn();
1048	else
1049		max_low_pfn = max_pfn;
1050
1051	high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1052#endif
1053
1054	/*
1055	 * Find and reserve possible boot-time SMP configuration:
1056	 */
1057	find_smp_config();
1058
1059	reserve_ibft_region();
1060
1061	early_alloc_pgt_buf();
1062
1063	/*
1064	 * Need to conclude brk, before e820__memblock_setup()
1065	 *  it could use memblock_find_in_range, could overlap with
1066	 *  brk area.
1067	 */
1068	reserve_brk();
1069
1070	cleanup_highmap();
1071
1072	memblock_set_current_limit(ISA_END_ADDRESS);
1073	e820__memblock_setup();
1074
1075	reserve_bios_regions();
1076
1077	efi_fake_memmap();
1078	efi_find_mirror();
1079	efi_esrt_init();
1080
1081	/*
1082	 * The EFI specification says that boot service code won't be
1083	 * called after ExitBootServices(). This is, in fact, a lie.
1084	 */
1085	efi_reserve_boot_services();
 
1086
1087	/* preallocate 4k for mptable mpc */
1088	e820__memblock_alloc_reserved_mpc_new();
1089
1090#ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1091	setup_bios_corruption_check();
1092#endif
1093
1094#ifdef CONFIG_X86_32
1095	printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1096			(max_pfn_mapped<<PAGE_SHIFT) - 1);
1097#endif
1098
1099	reserve_real_mode();
1100
1101	trim_platform_memory_ranges();
1102	trim_low_memory_range();
1103
1104	init_mem_mapping();
 
 
1105
1106	idt_setup_early_pf();
1107
1108	/*
1109	 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1110	 * with the current CR4 value.  This may not be necessary, but
1111	 * auditing all the early-boot CR4 manipulation would be needed to
1112	 * rule it out.
1113	 *
1114	 * Mask off features that don't work outside long mode (just
1115	 * PCIDE for now).
1116	 */
1117	mmu_cr4_features = __read_cr4() & ~X86_CR4_PCIDE;
1118
1119	memblock_set_current_limit(get_max_mapped());
1120
1121	/*
1122	 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1123	 */
1124
1125#ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1126	if (init_ohci1394_dma_early)
1127		init_ohci1394_dma_on_all_controllers();
1128#endif
1129	/* Allocate bigger log buffer */
1130	setup_log_buf(1);
1131
1132	if (efi_enabled(EFI_BOOT)) {
1133		switch (boot_params.secure_boot) {
1134		case efi_secureboot_mode_disabled:
1135			pr_info("Secure boot disabled\n");
1136			break;
1137		case efi_secureboot_mode_enabled:
1138			pr_info("Secure boot enabled\n");
1139			break;
1140		default:
1141			pr_info("Secure boot could not be determined\n");
1142			break;
1143		}
1144	}
1145
1146	reserve_initrd();
1147
1148	acpi_table_upgrade();
1149
1150	vsmp_init();
1151
1152	io_delay_init();
1153
1154	early_platform_quirks();
1155
1156	/*
1157	 * Parse the ACPI tables for possible boot-time SMP configuration.
1158	 */
1159	acpi_boot_table_init();
1160
1161	early_acpi_boot_init();
1162
1163	initmem_init();
1164	dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1165
1166	if (boot_cpu_has(X86_FEATURE_GBPAGES))
1167		hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1168
1169	/*
1170	 * Reserve memory for crash kernel after SRAT is parsed so that it
1171	 * won't consume hotpluggable memory.
1172	 */
1173	reserve_crashkernel();
1174
1175	memblock_find_dma_reserve();
1176
1177	if (!early_xdbc_setup_hardware())
1178		early_xdbc_register_console();
 
1179
1180	x86_init.paging.pagetable_init();
 
 
 
 
 
 
 
 
 
1181
1182	kasan_init();
1183
1184	/*
1185	 * Sync back kernel address range.
1186	 *
1187	 * FIXME: Can the later sync in setup_cpu_entry_areas() replace
1188	 * this call?
1189	 */
1190	sync_initial_page_table();
1191
1192	tboot_probe();
1193
 
1194	map_vsyscall();
 
1195
1196	generic_apic_probe();
1197
1198	early_quirks();
1199
1200	/*
1201	 * Read APIC and some other early information from ACPI tables.
1202	 */
1203	acpi_boot_init();
1204	sfi_init();
1205	x86_dtb_init();
1206
1207	/*
1208	 * get boot-time SMP configuration:
1209	 */
1210	get_smp_config();
1211
1212	/*
1213	 * Systems w/o ACPI and mptables might not have it mapped the local
1214	 * APIC yet, but prefill_possible_map() might need to access it.
1215	 */
1216	init_apic_mappings();
1217
1218	prefill_possible_map();
1219
1220	init_cpu_to_node();
1221
1222	io_apic_init_mappings();
 
 
1223
1224	x86_init.hyper.guest_late_init();
1225
1226	e820__reserve_resources();
1227	e820__register_nosave_regions(max_pfn);
1228
1229	x86_init.resources.reserve_resources();
1230
1231	e820__setup_pci_gap();
1232
1233#ifdef CONFIG_VT
1234#if defined(CONFIG_VGA_CONSOLE)
1235	if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1236		conswitchp = &vga_con;
 
 
1237#endif
1238#endif
1239	x86_init.oem.banner();
1240
1241	x86_init.timers.wallclock_init();
1242
 
 
1243	mcheck_init();
1244
1245	register_refined_jiffies(CLOCK_TICK_RATE);
1246
1247#ifdef CONFIG_EFI
1248	if (efi_enabled(EFI_BOOT))
1249		efi_apply_memmap_quirks();
1250#endif
1251
1252	unwind_init();
1253}
1254
1255#ifdef CONFIG_X86_32
1256
1257static struct resource video_ram_resource = {
1258	.name	= "Video RAM area",
1259	.start	= 0xa0000,
1260	.end	= 0xbffff,
1261	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
1262};
1263
1264void __init i386_reserve_resources(void)
1265{
1266	request_resource(&iomem_resource, &video_ram_resource);
1267	reserve_standard_io_resources();
1268}
1269
1270#endif /* CONFIG_X86_32 */
1271
1272static struct notifier_block kernel_offset_notifier = {
1273	.notifier_call = dump_kernel_offset
1274};
1275
1276static int __init register_kernel_offset_dumper(void)
1277{
1278	atomic_notifier_chain_register(&panic_notifier_list,
1279					&kernel_offset_notifier);
1280	return 0;
1281}
1282__initcall(register_kernel_offset_dumper);