Merge branch 'linux-next' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes...
[sfrench/cifs-2.6.git] / arch / x86 / kernel / setup.c
index 6f80b852a1961a6b496bc2404c5194769557d97a..531b55b8e81a1de1827eac5691d5f8aef1d8d10c 100644 (file)
-#include <linux/kernel.h>
+/*
+ *  Copyright (C) 1995  Linus Torvalds
+ *
+ *  Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
+ *
+ *  Memory region support
+ *     David Parsons <orc@pell.chi.il.us>, July-August 1999
+ *
+ *  Added E820 sanitization routine (removes overlapping memory regions);
+ *  Brian Moyle <bmoyle@mvista.com>, February 2001
+ *
+ * Moved CPU detection code to cpu/${cpu}.c
+ *    Patrick Mochel <mochel@osdl.org>, March 2002
+ *
+ *  Provisions for empty E820 memory regions (reported by certain BIOSes).
+ *  Alex Achenbach <xela@slit.de>, December 2002.
+ *
+ */
+
+/*
+ * This file handles the architecture-dependent parts of initialization
+ */
+
+#include <linux/sched.h>
+#include <linux/mm.h>
+#include <linux/mmzone.h>
+#include <linux/screen_info.h>
+#include <linux/ioport.h>
+#include <linux/acpi.h>
+#include <linux/apm_bios.h>
+#include <linux/initrd.h>
+#include <linux/bootmem.h>
+#include <linux/seq_file.h>
+#include <linux/console.h>
+#include <linux/mca.h>
+#include <linux/root_dev.h>
+#include <linux/highmem.h>
 #include <linux/module.h>
+#include <linux/efi.h>
 #include <linux/init.h>
-#include <linux/bootmem.h>
+#include <linux/edd.h>
+#include <linux/iscsi_ibft.h>
+#include <linux/nodemask.h>
+#include <linux/kexec.h>
+#include <linux/dmi.h>
+#include <linux/pfn.h>
+#include <linux/pci.h>
+#include <asm/pci-direct.h>
+#include <linux/init_ohci1394_dma.h>
+#include <linux/kvm_para.h>
+
+#include <linux/errno.h>
+#include <linux/kernel.h>
+#include <linux/stddef.h>
+#include <linux/unistd.h>
+#include <linux/ptrace.h>
+#include <linux/slab.h>
+#include <linux/user.h>
+#include <linux/delay.h>
+#include <linux/highmem.h>
+
+#include <linux/kallsyms.h>
+#include <linux/edd.h>
+#include <linux/iscsi_ibft.h>
+#include <linux/kexec.h>
+#include <linux/cpufreq.h>
+#include <linux/dma-mapping.h>
+#include <linux/ctype.h>
+#include <linux/uaccess.h>
+
 #include <linux/percpu.h>
+#include <linux/crash_dump.h>
+
+#include <video/edid.h>
+
+#include <asm/mtrr.h>
+#include <asm/apic.h>
+#include <asm/e820.h>
+#include <asm/mpspec.h>
+#include <asm/setup.h>
+#include <asm/arch_hooks.h>
+#include <asm/efi.h>
+#include <asm/sections.h>
+#include <asm/dmi.h>
+#include <asm/io_apic.h>
+#include <asm/ist.h>
+#include <asm/vmi.h>
+#include <setup_arch.h>
+#include <asm/bios_ebda.h>
+#include <asm/cacheflush.h>
+#include <asm/processor.h>
+#include <asm/bugs.h>
+
+#include <asm/system.h>
+#include <asm/vsyscall.h>
 #include <asm/smp.h>
+#include <asm/desc.h>
+#include <asm/dma.h>
+#include <asm/gart.h>
+#include <asm/mmu_context.h>
+#include <asm/proto.h>
+
+#include <mach_apic.h>
+#include <asm/paravirt.h>
+
 #include <asm/percpu.h>
 #include <asm/sections.h>
-#include <asm/processor.h>
-#include <asm/setup.h>
 #include <asm/topology.h>
-#include <asm/mpspec.h>
 #include <asm/apicdef.h>
+#ifdef CONFIG_X86_64
+#include <asm/numa_64.h>
+#endif
 
-#ifdef CONFIG_X86_LOCAL_APIC
-unsigned int num_processors;
-unsigned disabled_cpus __cpuinitdata;
-/* Processor that is doing the boot up */
-unsigned int boot_cpu_physical_apicid = -1U;
-EXPORT_SYMBOL(boot_cpu_physical_apicid);
+#ifndef ARCH_SETUP
+#define ARCH_SETUP
+#endif
+
+#ifndef CONFIG_DEBUG_BOOT_PARAMS
+struct boot_params __initdata boot_params;
+#else
+struct boot_params boot_params;
+#endif
+
+/*
+ * Machine setup..
+ */
+static struct resource data_resource = {
+       .name   = "Kernel data",
+       .start  = 0,
+       .end    = 0,
+       .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
+};
+
+static struct resource code_resource = {
+       .name   = "Kernel code",
+       .start  = 0,
+       .end    = 0,
+       .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
+};
+
+static struct resource bss_resource = {
+       .name   = "Kernel bss",
+       .start  = 0,
+       .end    = 0,
+       .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
+};
+
+
+#ifdef CONFIG_X86_32
+/* This value is set up by the early boot code to point to the value
+   immediately after the boot time page tables.  It contains a *physical*
+   address, and must not be in the .bss segment! */
+unsigned long init_pg_tables_start __initdata = ~0UL;
+unsigned long init_pg_tables_end __initdata = ~0UL;
+
+static struct resource video_ram_resource = {
+       .name   = "Video RAM area",
+       .start  = 0xa0000,
+       .end    = 0xbffff,
+       .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
+};
+
+/* cpu data as detected by the assembly code in head.S */
+struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
+/* common cpu data for all cpus */
+struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
+EXPORT_SYMBOL(boot_cpu_data);
+static void set_mca_bus(int x)
+{
+#ifdef CONFIG_MCA
+       MCA_bus = x;
+#endif
+}
+
+unsigned int def_to_bigsmp;
+
+/* for MCA, but anyone else can use it if they want */
+unsigned int machine_id;
+unsigned int machine_submodel_id;
+unsigned int BIOS_revision;
+
+struct apm_info apm_info;
+EXPORT_SYMBOL(apm_info);
+
+#if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
+       defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
+struct ist_info ist_info;
+EXPORT_SYMBOL(ist_info);
+#else
+struct ist_info ist_info;
+#endif
+
+#else
+struct cpuinfo_x86 boot_cpu_data __read_mostly;
+EXPORT_SYMBOL(boot_cpu_data);
+#endif
 
-DEFINE_PER_CPU(u16, x86_cpu_to_apicid) = BAD_APICID;
-EXPORT_PER_CPU_SYMBOL(x86_cpu_to_apicid);
 
-/* Bitmask of physically existing CPUs */
-physid_mask_t phys_cpu_present_map;
+#if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
+unsigned long mmu_cr4_features;
+#else
+unsigned long mmu_cr4_features = X86_CR4_PAE;
 #endif
 
-#if defined(CONFIG_HAVE_SETUP_PER_CPU_AREA) && defined(CONFIG_X86_SMP)
+/* Boot loader ID as an integer, for the benefit of proc_dointvec */
+int bootloader_type;
+
 /*
- * Copy data used in early init routines from the initial arrays to the
- * per cpu data areas.  These arrays then become expendable and the
- * *_early_ptr's are zeroed indicating that the static arrays are gone.
+ * Early DMI memory
+ */
+int dmi_alloc_index;
+char dmi_alloc_data[DMI_MAX_DATA];
+
+/*
+ * Setup options
+ */
+struct screen_info screen_info;
+EXPORT_SYMBOL(screen_info);
+struct edid_info edid_info;
+EXPORT_SYMBOL_GPL(edid_info);
+
+extern int root_mountflags;
+
+unsigned long saved_video_mode;
+
+#define RAMDISK_IMAGE_START_MASK       0x07FF
+#define RAMDISK_PROMPT_FLAG            0x8000
+#define RAMDISK_LOAD_FLAG              0x4000
+
+static char __initdata command_line[COMMAND_LINE_SIZE];
+
+#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
+struct edd edd;
+#ifdef CONFIG_EDD_MODULE
+EXPORT_SYMBOL(edd);
+#endif
+/**
+ * copy_edd() - Copy the BIOS EDD information
+ *              from boot_params into a safe place.
+ *
  */
-static void __init setup_per_cpu_maps(void)
+static inline void copy_edd(void)
+{
+     memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
+           sizeof(edd.mbr_signature));
+     memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
+     edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
+     edd.edd_info_nr = boot_params.eddbuf_entries;
+}
+#else
+static inline void copy_edd(void)
+{
+}
+#endif
+
+#ifdef CONFIG_BLK_DEV_INITRD
+
+#ifdef CONFIG_X86_32
+
+#define MAX_MAP_CHUNK  (NR_FIX_BTMAPS << PAGE_SHIFT)
+static void __init relocate_initrd(void)
 {
-       int cpu;
 
-       for_each_possible_cpu(cpu) {
-               per_cpu(x86_cpu_to_apicid, cpu) = x86_cpu_to_apicid_init[cpu];
-               per_cpu(x86_bios_cpu_apicid, cpu) =
-                                               x86_bios_cpu_apicid_init[cpu];
-#ifdef CONFIG_NUMA
-               per_cpu(x86_cpu_to_node_map, cpu) =
-                                               x86_cpu_to_node_map_init[cpu];
+       u64 ramdisk_image = boot_params.hdr.ramdisk_image;
+       u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
+       u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
+       u64 ramdisk_here;
+       unsigned long slop, clen, mapaddr;
+       char *p, *q;
+
+       /* We need to move the initrd down into lowmem */
+       ramdisk_here = find_e820_area(0, end_of_lowmem, ramdisk_size,
+                                        PAGE_SIZE);
+
+       if (ramdisk_here == -1ULL)
+               panic("Cannot find place for new RAMDISK of size %lld\n",
+                        ramdisk_size);
+
+       /* Note: this includes all the lowmem currently occupied by
+          the initrd, we rely on that fact to keep the data intact. */
+       reserve_early(ramdisk_here, ramdisk_here + ramdisk_size,
+                        "NEW RAMDISK");
+       initrd_start = ramdisk_here + PAGE_OFFSET;
+       initrd_end   = initrd_start + ramdisk_size;
+       printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
+                        ramdisk_here, ramdisk_here + ramdisk_size);
+
+       q = (char *)initrd_start;
+
+       /* Copy any lowmem portion of the initrd */
+       if (ramdisk_image < end_of_lowmem) {
+               clen = end_of_lowmem - ramdisk_image;
+               p = (char *)__va(ramdisk_image);
+               memcpy(q, p, clen);
+               q += clen;
+               ramdisk_image += clen;
+               ramdisk_size  -= clen;
+       }
+
+       /* Copy the highmem portion of the initrd */
+       while (ramdisk_size) {
+               slop = ramdisk_image & ~PAGE_MASK;
+               clen = ramdisk_size;
+               if (clen > MAX_MAP_CHUNK-slop)
+                       clen = MAX_MAP_CHUNK-slop;
+               mapaddr = ramdisk_image & PAGE_MASK;
+               p = early_ioremap(mapaddr, clen+slop);
+               memcpy(q, p+slop, clen);
+               early_iounmap(p, clen+slop);
+               q += clen;
+               ramdisk_image += clen;
+               ramdisk_size  -= clen;
+       }
+       /* high pages is not converted by early_res_to_bootmem */
+       ramdisk_image = boot_params.hdr.ramdisk_image;
+       ramdisk_size  = boot_params.hdr.ramdisk_size;
+       printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to"
+               " %08llx - %08llx\n",
+               ramdisk_image, ramdisk_image + ramdisk_size - 1,
+               ramdisk_here, ramdisk_here + ramdisk_size - 1);
+}
 #endif
+
+static void __init reserve_initrd(void)
+{
+       u64 ramdisk_image = boot_params.hdr.ramdisk_image;
+       u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
+       u64 ramdisk_end   = ramdisk_image + ramdisk_size;
+       u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
+
+       if (!boot_params.hdr.type_of_loader ||
+           !ramdisk_image || !ramdisk_size)
+               return;         /* No initrd provided by bootloader */
+
+       initrd_start = 0;
+
+       if (ramdisk_size >= (end_of_lowmem>>1)) {
+               free_early(ramdisk_image, ramdisk_end);
+               printk(KERN_ERR "initrd too large to handle, "
+                      "disabling initrd\n");
+               return;
+       }
+
+       printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image,
+                       ramdisk_end);
+
+
+       if (ramdisk_end <= end_of_lowmem) {
+               /* All in lowmem, easy case */
+               /*
+                * don't need to reserve again, already reserved early
+                * in i386_start_kernel
+                */
+               initrd_start = ramdisk_image + PAGE_OFFSET;
+               initrd_end = initrd_start + ramdisk_size;
+               return;
        }
 
-       /* indicate the early static arrays will soon be gone */
-       x86_cpu_to_apicid_early_ptr = NULL;
-       x86_bios_cpu_apicid_early_ptr = NULL;
-#ifdef CONFIG_NUMA
-       x86_cpu_to_node_map_early_ptr = NULL;
+#ifdef CONFIG_X86_32
+       relocate_initrd();
+#else
+       printk(KERN_ERR "initrd extends beyond end of memory "
+              "(0x%08llx > 0x%08llx)\ndisabling initrd\n",
+              ramdisk_end, end_of_lowmem);
+       initrd_start = 0;
 #endif
+       free_early(ramdisk_image, ramdisk_end);
 }
+#else
+static void __init reserve_initrd(void)
+{
+}
+#endif /* CONFIG_BLK_DEV_INITRD */
+
+static void __init parse_setup_data(void)
+{
+       struct setup_data *data;
+       u64 pa_data;
+
+       if (boot_params.hdr.version < 0x0209)
+               return;
+       pa_data = boot_params.hdr.setup_data;
+       while (pa_data) {
+               data = early_ioremap(pa_data, PAGE_SIZE);
+               switch (data->type) {
+               case SETUP_E820_EXT:
+                       parse_e820_ext(data, pa_data);
+                       break;
+               default:
+                       break;
+               }
+               pa_data = data->next;
+               early_iounmap(data, PAGE_SIZE);
+       }
+}
+
+static void __init e820_reserve_setup_data(void)
+{
+       struct setup_data *data;
+       u64 pa_data;
+       int found = 0;
+
+       if (boot_params.hdr.version < 0x0209)
+               return;
+       pa_data = boot_params.hdr.setup_data;
+       while (pa_data) {
+               data = early_ioremap(pa_data, sizeof(*data));
+               e820_update_range(pa_data, sizeof(*data)+data->len,
+                        E820_RAM, E820_RESERVED_KERN);
+               found = 1;
+               pa_data = data->next;
+               early_iounmap(data, sizeof(*data));
+       }
+       if (!found)
+               return;
 
-#ifdef CONFIG_HAVE_CPUMASK_OF_CPU_MAP
-cpumask_t *cpumask_of_cpu_map __read_mostly;
-EXPORT_SYMBOL(cpumask_of_cpu_map);
+       sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
+       memcpy(&e820_saved, &e820, sizeof(struct e820map));
+       printk(KERN_INFO "extended physical RAM map:\n");
+       e820_print_map("reserve setup_data");
+}
 
-/* requires nr_cpu_ids to be initialized */
-static void __init setup_cpumask_of_cpu(void)
+static void __init reserve_early_setup_data(void)
 {
-       int i;
+       struct setup_data *data;
+       u64 pa_data;
+       char buf[32];
+
+       if (boot_params.hdr.version < 0x0209)
+               return;
+       pa_data = boot_params.hdr.setup_data;
+       while (pa_data) {
+               data = early_ioremap(pa_data, sizeof(*data));
+               sprintf(buf, "setup data %x", data->type);
+               reserve_early(pa_data, pa_data+sizeof(*data)+data->len, buf);
+               pa_data = data->next;
+               early_iounmap(data, sizeof(*data));
+       }
+}
+
+/*
+ * --------- Crashkernel reservation ------------------------------
+ */
+
+#ifdef CONFIG_KEXEC
+
+/**
+ * Reserve @size bytes of crashkernel memory at any suitable offset.
+ *
+ * @size: Size of the crashkernel memory to reserve.
+ * Returns the base address on success, and -1ULL on failure.
+ */
+unsigned long long find_and_reserve_crashkernel(unsigned long long size)
+{
+       const unsigned long long alignment = 16<<20;    /* 16M */
+       unsigned long long start = 0LL;
+
+       while (1) {
+               int ret;
+
+               start = find_e820_area(start, ULONG_MAX, size, alignment);
+               if (start == -1ULL)
+                       return start;
+
+               /* try to reserve it */
+               ret = reserve_bootmem_generic(start, size, BOOTMEM_EXCLUSIVE);
+               if (ret >= 0)
+                       return start;
+
+               start += alignment;
+       }
+}
+
+static inline unsigned long long get_total_mem(void)
+{
+       unsigned long long total;
+
+       total = max_low_pfn - min_low_pfn;
+#ifdef CONFIG_HIGHMEM
+       total += highend_pfn - highstart_pfn;
+#endif
+
+       return total << PAGE_SHIFT;
+}
+
+static void __init reserve_crashkernel(void)
+{
+       unsigned long long total_mem;
+       unsigned long long crash_size, crash_base;
+       int ret;
+
+       total_mem = get_total_mem();
 
-       /* alloc_bootmem zeroes memory */
-       cpumask_of_cpu_map = alloc_bootmem_low(sizeof(cpumask_t) * nr_cpu_ids);
-       for (i = 0; i < nr_cpu_ids; i++)
-               cpu_set(i, cpumask_of_cpu_map[i]);
+       ret = parse_crashkernel(boot_command_line, total_mem,
+                       &crash_size, &crash_base);
+       if (ret != 0 || crash_size <= 0)
+               return;
+
+       /* 0 means: find the address automatically */
+       if (crash_base <= 0) {
+               crash_base = find_and_reserve_crashkernel(crash_size);
+               if (crash_base == -1ULL) {
+                       pr_info("crashkernel reservation failed. "
+                               "No suitable area found.\n");
+                       return;
+               }
+       } else {
+               ret = reserve_bootmem_generic(crash_base, crash_size,
+                                       BOOTMEM_EXCLUSIVE);
+               if (ret < 0) {
+                       pr_info("crashkernel reservation failed - "
+                               "memory is in use\n");
+                       return;
+               }
+       }
+
+       printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
+                       "for crashkernel (System RAM: %ldMB)\n",
+                       (unsigned long)(crash_size >> 20),
+                       (unsigned long)(crash_base >> 20),
+                       (unsigned long)(total_mem >> 20));
+
+       crashk_res.start = crash_base;
+       crashk_res.end   = crash_base + crash_size - 1;
+       insert_resource(&iomem_resource, &crashk_res);
 }
 #else
-static inline void setup_cpumask_of_cpu(void) { }
+static void __init reserve_crashkernel(void)
+{
+}
 #endif
 
-#ifdef CONFIG_X86_32
-/*
- * Great future not-so-futuristic plan: make i386 and x86_64 do it
- * the same way
+static struct resource standard_io_resources[] = {
+       { .name = "dma1", .start = 0x00, .end = 0x1f,
+               .flags = IORESOURCE_BUSY | IORESOURCE_IO },
+       { .name = "pic1", .start = 0x20, .end = 0x21,
+               .flags = IORESOURCE_BUSY | IORESOURCE_IO },
+       { .name = "timer0", .start = 0x40, .end = 0x43,
+               .flags = IORESOURCE_BUSY | IORESOURCE_IO },
+       { .name = "timer1", .start = 0x50, .end = 0x53,
+               .flags = IORESOURCE_BUSY | IORESOURCE_IO },
+       { .name = "keyboard", .start = 0x60, .end = 0x60,
+               .flags = IORESOURCE_BUSY | IORESOURCE_IO },
+       { .name = "keyboard", .start = 0x64, .end = 0x64,
+               .flags = IORESOURCE_BUSY | IORESOURCE_IO },
+       { .name = "dma page reg", .start = 0x80, .end = 0x8f,
+               .flags = IORESOURCE_BUSY | IORESOURCE_IO },
+       { .name = "pic2", .start = 0xa0, .end = 0xa1,
+               .flags = IORESOURCE_BUSY | IORESOURCE_IO },
+       { .name = "dma2", .start = 0xc0, .end = 0xdf,
+               .flags = IORESOURCE_BUSY | IORESOURCE_IO },
+       { .name = "fpu", .start = 0xf0, .end = 0xff,
+               .flags = IORESOURCE_BUSY | IORESOURCE_IO }
+};
+
+static void __init reserve_standard_io_resources(void)
+{
+       int i;
+
+       /* request I/O space for devices used on all i[345]86 PCs */
+       for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
+               request_resource(&ioport_resource, &standard_io_resources[i]);
+
+}
+
+#ifdef CONFIG_PROC_VMCORE
+/* elfcorehdr= specifies the location of elf core header
+ * stored by the crashed kernel. This option will be passed
+ * by kexec loader to the capture kernel.
  */
-unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
-EXPORT_SYMBOL(__per_cpu_offset);
+static int __init setup_elfcorehdr(char *arg)
+{
+       char *end;
+       if (!arg)
+               return -EINVAL;
+       elfcorehdr_addr = memparse(arg, &end);
+       return end > arg ? 0 : -EINVAL;
+}
+early_param("elfcorehdr", setup_elfcorehdr);
 #endif
 
 /*
- * Great future plan:
- * Declare PDA itself and support (irqstack,tss,pgd) as per cpu data.
- * Always point %gs to its beginning
+ * Determine if we were loaded by an EFI loader.  If so, then we have also been
+ * passed the efi memmap, systab, etc., so we should use these data structures
+ * for initialization.  Note, the efi init code path is determined by the
+ * global efi_enabled. This allows the same kernel image to be used on existing
+ * systems (with a traditional BIOS) as well as on EFI systems.
+ */
+/*
+ * setup_arch - architecture-specific boot-time initializations
+ *
+ * Note: On x86_64, fixmaps are ready for use even before this is called.
  */
-void __init setup_per_cpu_areas(void)
+
+void __init setup_arch(char **cmdline_p)
 {
-       int i, highest_cpu = 0;
-       unsigned long size;
+#ifdef CONFIG_X86_32
+       memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
+       visws_early_detect();
+       pre_setup_arch_hook();
+       early_cpu_init();
+#else
+       printk(KERN_INFO "Command line: %s\n", boot_command_line);
+#endif
 
-#ifdef CONFIG_HOTPLUG_CPU
-       prefill_possible_map();
+       early_ioremap_init();
+
+       ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
+       screen_info = boot_params.screen_info;
+       edid_info = boot_params.edid_info;
+#ifdef CONFIG_X86_32
+       apm_info.bios = boot_params.apm_bios_info;
+       ist_info = boot_params.ist_info;
+       if (boot_params.sys_desc_table.length != 0) {
+               set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
+               machine_id = boot_params.sys_desc_table.table[0];
+               machine_submodel_id = boot_params.sys_desc_table.table[1];
+               BIOS_revision = boot_params.sys_desc_table.table[2];
+       }
+#endif
+       saved_video_mode = boot_params.hdr.vid_mode;
+       bootloader_type = boot_params.hdr.type_of_loader;
+
+#ifdef CONFIG_BLK_DEV_RAM
+       rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
+       rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
+       rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
+#endif
+#ifdef CONFIG_EFI
+       if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
+#ifdef CONFIG_X86_32
+                    "EL32",
+#else
+                    "EL64",
+#endif
+        4)) {
+               efi_enabled = 1;
+               efi_reserve_early();
+       }
 #endif
 
-       /* Copy section for each CPU (we discard the original) */
-       size = PERCPU_ENOUGH_ROOM;
-       printk(KERN_INFO "PERCPU: Allocating %lu bytes of per cpu data\n",
-                         size);
+       ARCH_SETUP
 
-       for_each_possible_cpu(i) {
-               char *ptr;
-#ifndef CONFIG_NEED_MULTIPLE_NODES
-               ptr = alloc_bootmem_pages(size);
+       setup_memory_map();
+       parse_setup_data();
+       /* update the e820_saved too */
+       e820_reserve_setup_data();
+
+       copy_edd();
+
+       if (!boot_params.hdr.root_flags)
+               root_mountflags &= ~MS_RDONLY;
+       init_mm.start_code = (unsigned long) _text;
+       init_mm.end_code = (unsigned long) _etext;
+       init_mm.end_data = (unsigned long) _edata;
+#ifdef CONFIG_X86_32
+       init_mm.brk = init_pg_tables_end + PAGE_OFFSET;
 #else
-               int node = early_cpu_to_node(i);
-               if (!node_online(node) || !NODE_DATA(node)) {
-                       ptr = alloc_bootmem_pages(size);
-                       printk(KERN_INFO
-                              "cpu %d has no node or node-local memory\n", i);
-               }
-               else
-                       ptr = alloc_bootmem_pages_node(NODE_DATA(node), size);
+       init_mm.brk = (unsigned long) &_end;
 #endif
-               if (!ptr)
-                       panic("Cannot allocate cpu data for CPU %d\n", i);
+
+       code_resource.start = virt_to_phys(_text);
+       code_resource.end = virt_to_phys(_etext)-1;
+       data_resource.start = virt_to_phys(_etext);
+       data_resource.end = virt_to_phys(_edata)-1;
+       bss_resource.start = virt_to_phys(&__bss_start);
+       bss_resource.end = virt_to_phys(&__bss_stop)-1;
+
 #ifdef CONFIG_X86_64
-               cpu_pda(i)->data_offset = ptr - __per_cpu_start;
+       early_cpu_init();
+#endif
+       strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
+       *cmdline_p = command_line;
+
+       parse_early_param();
+
+       /* after early param, so could get panic from serial */
+       reserve_early_setup_data();
+
+       if (acpi_mps_check()) {
+#ifdef CONFIG_X86_LOCAL_APIC
+               disable_apic = 1;
+#endif
+               clear_cpu_cap(&boot_cpu_data, X86_FEATURE_APIC);
+       }
+
+#ifdef CONFIG_PCI
+       if (pci_early_dump_regs)
+               early_dump_pci_devices();
+#endif
+
+       finish_e820_parsing();
+
+#ifdef CONFIG_X86_32
+       probe_roms();
+#endif
+
+       /* after parse_early_param, so could debug it */
+       insert_resource(&iomem_resource, &code_resource);
+       insert_resource(&iomem_resource, &data_resource);
+       insert_resource(&iomem_resource, &bss_resource);
+
+       if (efi_enabled)
+               efi_init();
+
+#ifdef CONFIG_X86_32
+       if (ppro_with_ram_bug()) {
+               e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
+                                 E820_RESERVED);
+               sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
+               printk(KERN_INFO "fixed physical RAM map:\n");
+               e820_print_map("bad_ppro");
+       }
+#else
+       early_gart_iommu_check();
+#endif
+
+       /*
+        * partially used pages are not usable - thus
+        * we are rounding upwards:
+        */
+       max_pfn = e820_end_of_ram_pfn();
+
+       /* preallocate 4k for mptable mpc */
+       early_reserve_e820_mpc_new();
+       /* update e820 for memory not covered by WB MTRRs */
+       mtrr_bp_init();
+       if (mtrr_trim_uncached_memory(max_pfn))
+               max_pfn = e820_end_of_ram_pfn();
+
+#ifdef CONFIG_X86_32
+       /* max_low_pfn get updated here */
+       find_low_pfn_range();
 #else
-               __per_cpu_offset[i] = ptr - __per_cpu_start;
+       num_physpages = max_pfn;
+
+       check_efer();
+
+       /* How many end-of-memory variables you have, grandma! */
+       /* need this before calling reserve_initrd */
+       if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
+               max_low_pfn = e820_end_of_low_ram_pfn();
+       else
+               max_low_pfn = max_pfn;
+
+       high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
 #endif
-               memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 
-               highest_cpu = i;
+       /* max_pfn_mapped is updated here */
+       max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
+       max_pfn_mapped = max_low_pfn_mapped;
+
+#ifdef CONFIG_X86_64
+       if (max_pfn > max_low_pfn) {
+               max_pfn_mapped = init_memory_mapping(1UL<<32,
+                                                    max_pfn<<PAGE_SHIFT);
+               /* can we preseve max_low_pfn ?*/
+               max_low_pfn = max_pfn;
        }
+#endif
 
-       nr_cpu_ids = highest_cpu + 1;
-       printk(KERN_DEBUG "NR_CPUS: %d, nr_cpu_ids: %d\n", NR_CPUS, nr_cpu_ids);
+       /*
+        * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
+        */
 
-       /* Setup percpu data maps */
-       setup_per_cpu_maps();
+#ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
+       if (init_ohci1394_dma_early)
+               init_ohci1394_dma_on_all_controllers();
+#endif
 
-       /* Setup cpumask_of_cpu map */
-       setup_cpumask_of_cpu();
-}
+       reserve_initrd();
+
+#ifdef CONFIG_X86_64
+       vsmp_init();
+#endif
+
+       dmi_scan_machine();
+
+       io_delay_init();
+
+       /*
+        * Parse the ACPI tables for possible boot-time SMP configuration.
+        */
+       acpi_boot_table_init();
+
+#ifdef CONFIG_ACPI_NUMA
+       /*
+        * Parse SRAT to discover nodes.
+        */
+       acpi_numa_init();
+#endif
+
+       initmem_init(0, max_pfn);
+
+#ifdef CONFIG_X86_64
+       dma32_reserve_bootmem();
+#endif
+
+#ifdef CONFIG_ACPI_SLEEP
+       /*
+        * Reserve low memory region for sleep support.
+        */
+       acpi_reserve_bootmem();
+#endif
+#ifdef CONFIG_X86_FIND_SMP_CONFIG
+       /*
+        * Find and reserve possible boot-time SMP configuration:
+        */
+       find_smp_config();
+#endif
+       reserve_crashkernel();
+
+       reserve_ibft_region();
+
+#ifdef CONFIG_KVM_CLOCK
+       kvmclock_init();
+#endif
+
+#if defined(CONFIG_VMI) && defined(CONFIG_X86_32)
+       /*
+        * Must be after max_low_pfn is determined, and before kernel
+        * pagetables are setup.
+        */
+       vmi_init();
+#endif
+
+       paging_init();
+
+#ifdef CONFIG_X86_64
+       map_vsyscall();
+#endif
+
+#ifdef CONFIG_X86_GENERICARCH
+       generic_apic_probe();
+#endif
+
+       early_quirks();
+
+       /*
+        * Read APIC and some other early information from ACPI tables.
+        */
+       acpi_boot_init();
+
+#if defined(CONFIG_X86_MPPARSE) || defined(CONFIG_X86_VISWS)
+       /*
+        * get boot-time SMP configuration:
+        */
+       if (smp_found_config)
+               get_smp_config();
+#endif
 
+       prefill_possible_map();
+#ifdef CONFIG_X86_64
+       init_cpu_to_node();
+#endif
+
+#ifdef CONFIG_X86_NUMAQ
+       /*
+        * need to check online nodes num, call it
+        * here before time_init/tsc_init
+        */
+       numaq_tsc_disable();
+#endif
+
+       init_apic_mappings();
+       ioapic_init_mappings();
+
+#if defined(CONFIG_SMP) && defined(CONFIG_X86_PC) && defined(CONFIG_X86_32)
+       if (def_to_bigsmp)
+               printk(KERN_WARNING "More than 8 CPUs detected and "
+                       "CONFIG_X86_PC cannot handle it.\nUse "
+                       "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
+#endif
+       kvm_guest_init();
+
+       e820_reserve_resources();
+       e820_mark_nosave_regions(max_low_pfn);
+
+#ifdef CONFIG_X86_32
+       request_resource(&iomem_resource, &video_ram_resource);
+#endif
+       reserve_standard_io_resources();
+
+       e820_setup_gap();
+
+#ifdef CONFIG_VT
+#if defined(CONFIG_VGA_CONSOLE)
+       if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
+               conswitchp = &vga_con;
+#elif defined(CONFIG_DUMMY_CONSOLE)
+       conswitchp = &dummy_con;
 #endif
+#endif
+}