Fedora kernel-2.6.17-1.2142_FC4 patched with stable patch-2.6.17.4-vs2.0.2-rc26.diff
[linux-2.6.git] / arch / x86_64 / kernel / setup.c
index 99f038e..655b919 100644 (file)
 #include <asm/processor.h>
 #include <linux/console.h>
 #include <linux/seq_file.h>
+#include <linux/crash_dump.h>
 #include <linux/root_dev.h>
 #include <linux/pci.h>
 #include <linux/acpi.h>
 #include <linux/kallsyms.h>
 #include <linux/edd.h>
+#include <linux/mmzone.h>
+#include <linux/kexec.h>
+#include <linux/cpufreq.h>
+#include <linux/dmi.h>
+#include <linux/dma-mapping.h>
+#include <linux/ctype.h>
+
 #include <asm/mtrr.h>
 #include <asm/uaccess.h>
 #include <asm/system.h>
 #include <asm/setup.h>
 #include <asm/mach_apic.h>
 #include <asm/numa.h>
+#include <asm/swiotlb.h>
+#include <asm/sections.h>
+#include <asm/gart-mapping.h>
+#include <asm/dmi.h>
 
 /*
  * Machine setup..
  */
 
-struct cpuinfo_x86 boot_cpu_data;
+struct cpuinfo_x86 boot_cpu_data __read_mostly;
 
 unsigned long mmu_cr4_features;
 
 int acpi_disabled;
 EXPORT_SYMBOL(acpi_disabled);
-#ifdef CONFIG_ACPI_BOOT
+#ifdef CONFIG_ACPI
 extern int __initdata acpi_ht;
 extern acpi_interrupt_flags    acpi_sci_flags;
 int __initdata acpi_force = 0;
@@ -81,15 +93,15 @@ int bootloader_type;
 
 unsigned long saved_video_mode;
 
-#ifdef CONFIG_SWIOTLB
-int swiotlb;
-EXPORT_SYMBOL(swiotlb);
-#endif
+/* 
+ * Early DMI memory
+ */
+int dmi_alloc_index;
+char dmi_alloc_data[DMI_MAX_DATA];
 
 /*
  * Setup options
  */
-struct drive_info_struct { char dummy[32]; } drive_info;
 struct screen_info screen_info;
 struct sys_desc_table_struct {
        unsigned short length;
@@ -100,7 +112,6 @@ struct edid_info edid_info;
 struct e820map e820;
 
 extern int root_mountflags;
-extern char _text, _etext, _edata, _end;
 
 char command_line[COMMAND_LINE_SIZE];
 
@@ -267,14 +278,18 @@ static void __init probe_roms(void)
        }
 }
 
+/* Check for full argument with no trailing characters */
+static int fullarg(char *p, char *arg)
+{
+       int l = strlen(arg);
+       return !memcmp(p, arg, l) && (p[l] == 0 || isspace(p[l]));
+}
+
 static __init void parse_cmdline_early (char ** cmdline_p)
 {
        char c = ' ', *to = command_line, *from = COMMAND_LINE;
        int len = 0;
-
-       /* Save unparsed command line copy for /proc/cmdline */
-       memcpy(saved_command_line, COMMAND_LINE, COMMAND_LINE_SIZE);
-       saved_command_line[COMMAND_LINE_SIZE-1] = '\0';
+       int userdef = 0;
 
        for (;;) {
                if (c != ' ') 
@@ -291,12 +306,12 @@ static __init void parse_cmdline_early (char ** cmdline_p)
                        maxcpus = simple_strtoul(from + 8, NULL, 0);
                }
 #endif
-#ifdef CONFIG_ACPI_BOOT
+#ifdef CONFIG_ACPI
                /* "acpi=off" disables both ACPI table parsing and interpreter init */
-               if (!memcmp(from, "acpi=off", 8))
+               if (fullarg(from,"acpi=off"))
                        disable_acpi();
 
-               if (!memcmp(from, "acpi=force", 10)) { 
+               if (fullarg(from, "acpi=force")) { 
                        /* add later when we do DMI horrors: */
                        acpi_force = 1;
                        acpi_disabled = 0;
@@ -304,43 +319,49 @@ static __init void parse_cmdline_early (char ** cmdline_p)
 
                /* acpi=ht just means: do ACPI MADT parsing 
                   at bootup, but don't enable the full ACPI interpreter */
-               if (!memcmp(from, "acpi=ht", 7)) { 
+               if (fullarg(from, "acpi=ht")) { 
                        if (!acpi_force)
                                disable_acpi();
                        acpi_ht = 1; 
                }
-                else if (!memcmp(from, "pci=noacpi", 10)) 
+                else if (fullarg(from, "pci=noacpi")) 
                        acpi_disable_pci();
-               else if (!memcmp(from, "acpi=noirq", 10))
+               else if (fullarg(from, "acpi=noirq"))
                        acpi_noirq_set();
 
-               else if (!memcmp(from, "acpi_sci=edge", 13))
+               else if (fullarg(from, "acpi_sci=edge"))
                        acpi_sci_flags.trigger =  1;
-               else if (!memcmp(from, "acpi_sci=level", 14))
+               else if (fullarg(from, "acpi_sci=level"))
                        acpi_sci_flags.trigger = 3;
-               else if (!memcmp(from, "acpi_sci=high", 13))
+               else if (fullarg(from, "acpi_sci=high"))
                        acpi_sci_flags.polarity = 1;
-               else if (!memcmp(from, "acpi_sci=low", 12))
+               else if (fullarg(from, "acpi_sci=low"))
                        acpi_sci_flags.polarity = 3;
 
                /* acpi=strict disables out-of-spec workarounds */
-               else if (!memcmp(from, "acpi=strict", 11)) {
+               else if (fullarg(from, "acpi=strict")) {
                        acpi_strict = 1;
                }
 #ifdef CONFIG_X86_IO_APIC
-               else if (!memcmp(from, "acpi_skip_timer_override", 24))
+               else if (fullarg(from, "acpi_skip_timer_override"))
                        acpi_skip_timer_override = 1;
 #endif
 #endif
 
-               if (!memcmp(from, "nolapic", 7) ||
-                   !memcmp(from, "disableapic", 11))
+               if (fullarg(from, "disable_timer_pin_1"))
+                       disable_timer_pin_1 = 1;
+               if (fullarg(from, "enable_timer_pin_1"))
+                       disable_timer_pin_1 = -1;
+
+               if (fullarg(from, "nolapic") || fullarg(from, "disableapic")) {
+                       clear_bit(X86_FEATURE_APIC, boot_cpu_data.x86_capability);
                        disable_apic = 1;
+               }
 
-               if (!memcmp(from, "noapic", 6)) 
+               if (fullarg(from, "noapic"))
                        skip_ioapic_setup = 1;
 
-               if (!memcmp(from, "apic", 4)) { 
+               if (fullarg(from,"apic")) {
                        skip_ioapic_setup = 0;
                        ioapic_force = 1;
                }
@@ -348,23 +369,78 @@ static __init void parse_cmdline_early (char ** cmdline_p)
                if (!memcmp(from, "mem=", 4))
                        parse_memopt(from+4, &from); 
 
-#ifdef CONFIG_DISCONTIGMEM
+               if (!memcmp(from, "memmap=", 7)) {
+                       /* exactmap option is for used defined memory */
+                       if (!memcmp(from+7, "exactmap", 8)) {
+#ifdef CONFIG_CRASH_DUMP
+                               /* If we are doing a crash dump, we
+                                * still need to know the real mem
+                                * size before original memory map is
+                                * reset.
+                                */
+                               saved_max_pfn = e820_end_of_ram();
+#endif
+                               from += 8+7;
+                               end_pfn_map = 0;
+                               e820.nr_map = 0;
+                               userdef = 1;
+                       }
+                       else {
+                               parse_memmapopt(from+7, &from);
+                               userdef = 1;
+                       }
+               }
+
+#ifdef CONFIG_NUMA
                if (!memcmp(from, "numa=", 5))
                        numa_setup(from+5); 
 #endif
 
-#ifdef CONFIG_GART_IOMMU 
                if (!memcmp(from,"iommu=",6)) { 
                        iommu_setup(from+6); 
                }
-#endif
 
-               if (!memcmp(from,"oops=panic", 10))
+               if (fullarg(from,"oops=panic"))
                        panic_on_oops = 1;
 
                if (!memcmp(from, "noexec=", 7))
                        nonx_setup(from + 7);
 
+#ifdef CONFIG_KEXEC
+               /* crashkernel=size@addr specifies the location to reserve for
+                * a crash kernel.  By reserving this memory we guarantee
+                * that linux never set's it up as a DMA target.
+                * Useful for holding code to do something appropriate
+                * after a kernel panic.
+                */
+               else if (!memcmp(from, "crashkernel=", 12)) {
+                       unsigned long size, base;
+                       size = memparse(from+12, &from);
+                       if (*from == '@') {
+                               base = memparse(from+1, &from);
+                               /* FIXME: Do I want a sanity check
+                                * to validate the memory range?
+                                */
+                               crashk_res.start = base;
+                               crashk_res.end   = base + size - 1;
+                       }
+               }
+#endif
+
+#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.
+                */
+               else if(!memcmp(from, "elfcorehdr=", 11))
+                       elfcorehdr_addr = memparse(from+11, &from);
+#endif
+
+#ifdef CONFIG_HOTPLUG_CPU
+               else if (!memcmp(from, "additional_cpus=", 16))
+                       setup_additional_cpus(from+16);
+#endif
+
        next_char:
                c = *(from++);
                if (!c)
@@ -373,21 +449,27 @@ static __init void parse_cmdline_early (char ** cmdline_p)
                        break;
                *(to++) = c;
        }
+       if (userdef) {
+               printk(KERN_INFO "user-defined physical RAM map:\n");
+               e820_print_map("user");
+       }
        *to = '\0';
        *cmdline_p = command_line;
 }
 
-#ifndef CONFIG_DISCONTIGMEM
-static void __init contig_initmem_init(void)
+#ifndef CONFIG_NUMA
+static void __init
+contig_initmem_init(unsigned long start_pfn, unsigned long end_pfn)
 {
-        unsigned long bootmap_size, bootmap; 
-        bootmap_size = bootmem_bootmap_pages(end_pfn)<<PAGE_SHIFT;
-        bootmap = find_e820_area(0, end_pfn<<PAGE_SHIFT, bootmap_size);
-        if (bootmap == -1L) 
-                panic("Cannot find bootmem map of size %ld\n",bootmap_size);
-        bootmap_size = init_bootmem(bootmap >> PAGE_SHIFT, end_pfn);
-        e820_bootmem_free(&contig_page_data, 0, end_pfn << PAGE_SHIFT); 
-        reserve_bootmem(bootmap, bootmap_size);
+       unsigned long bootmap_size, bootmap;
+
+       bootmap_size = bootmem_bootmap_pages(end_pfn)<<PAGE_SHIFT;
+       bootmap = find_e820_area(0, end_pfn<<PAGE_SHIFT, bootmap_size);
+       if (bootmap == -1L)
+               panic("Cannot find bootmem map of size %ld\n",bootmap_size);
+       bootmap_size = init_bootmem(bootmap >> PAGE_SHIFT, end_pfn);
+       e820_bootmem_free(NODE_DATA(0), 0, end_pfn << PAGE_SHIFT);
+       reserve_bootmem(bootmap, bootmap_size);
 } 
 #endif
 
@@ -411,6 +493,8 @@ static unsigned char *k8_nops[ASM_NOP_MAX+1] = {
      k8nops + 1 + 2 + 3 + 4 + 5 + 6 + 7,
 }; 
 
+extern char __vsyscall_0;
+
 /* Replace instructions with better alternatives for this CPU type.
 
    This runs before SMP is initialized to avoid SMP problems with
@@ -422,11 +506,17 @@ void apply_alternatives(void *start, void *end)
        struct alt_instr *a; 
        int diff, i, k;
        for (a = start; (void *)a < end; a++) { 
+               u8 *instr;
+
                if (!boot_cpu_has(a->cpuid))
                        continue;
 
                BUG_ON(a->replacementlen > a->instrlen); 
-               __inline_memcpy(a->instr, a->replacement, a->replacementlen); 
+               instr = a->instr;
+               /* vsyscall code is not mapped yet. resolve it manually. */
+               if (instr >= (u8 *)VSYSCALL_START && instr < (u8*)VSYSCALL_END)
+                       instr = __va(instr - (u8*)VSYSCALL_START + (u8*)__pa_symbol(&__vsyscall_0));
+               __inline_memcpy(instr, a->replacement, a->replacementlen);
                diff = a->instrlen - a->replacementlen; 
 
                /* Pad the rest with nops */
@@ -434,7 +524,7 @@ void apply_alternatives(void *start, void *end)
                        k = diff;
                        if (k > ASM_NOP_MAX)
                                k = ASM_NOP_MAX;
-                       __inline_memcpy(a->instr + i, k8_nops[k], k); 
+                       __inline_memcpy(instr + i, k8_nops[k], k);
                } 
        }
 } 
@@ -452,7 +542,7 @@ void __init alternative_instructions(void)
 static int __init noreplacement_setup(char *s)
 { 
      no_replacement = 1; 
-     return 0; 
+     return 1;
 } 
 
 __setup("noreplacement", noreplacement_setup); 
@@ -481,17 +571,28 @@ static inline void copy_edd(void)
 #endif
 
 #define EBDA_ADDR_POINTER 0x40E
-static void __init reserve_ebda_region(void)
+
+unsigned __initdata ebda_addr;
+unsigned __initdata ebda_size;
+
+static void discover_ebda(void)
 {
-       unsigned int addr;
-       /** 
+       /*
         * there is a real-mode segmented pointer pointing to the 
         * 4K EBDA area at 0x40E
         */
-       addr = *(unsigned short *)phys_to_virt(EBDA_ADDR_POINTER);
-       addr <<= 4;
-       if (addr)
-               reserve_bootmem_generic(addr, PAGE_SIZE);
+       ebda_addr = *(unsigned short *)EBDA_ADDR_POINTER;
+       ebda_addr <<= 4;
+
+       ebda_size = *(unsigned short *)(unsigned long)ebda_addr;
+
+       /* Round EBDA up to pages */
+       if (ebda_size == 0)
+               ebda_size = 1;
+       ebda_size <<= 10;
+       ebda_size = round_up(ebda_size + (ebda_addr & ~PAGE_MASK), PAGE_SIZE);
+       if (ebda_size > 64*1024)
+               ebda_size = 64*1024;
 }
 
 void __init setup_arch(char **cmdline_p)
@@ -499,7 +600,6 @@ void __init setup_arch(char **cmdline_p)
        unsigned long kernel_end;
 
        ROOT_DEV = old_decode_dev(ORIG_ROOT_DEV);
-       drive_info = DRIVE_INFO;
        screen_info = SCREEN_INFO;
        edid_info = EDID_INFO;
        saved_video_mode = SAVED_VIDEO_MODE;
@@ -534,12 +634,19 @@ void __init setup_arch(char **cmdline_p)
         * we are rounding upwards:
         */
        end_pfn = e820_end_of_ram();
+       num_physpages = end_pfn;                /* for pfn_valid */
 
        check_efer();
 
+       discover_ebda();
+
        init_memory_mapping(0, (end_pfn_map << PAGE_SHIFT));
 
-#ifdef CONFIG_ACPI_BOOT
+       dmi_scan_machine();
+
+       zap_low_mappings(0);
+
+#ifdef CONFIG_ACPI
        /*
         * Initialize the ACPI boot-time table parser (gets the RSDP and SDT).
         * Call this early for SRAT node setup.
@@ -554,10 +661,10 @@ void __init setup_arch(char **cmdline_p)
        acpi_numa_init();
 #endif
 
-#ifdef CONFIG_DISCONTIGMEM
+#ifdef CONFIG_NUMA
        numa_initmem_init(0, end_pfn); 
 #else
-       contig_initmem_init(); 
+       contig_initmem_init(0, end_pfn);
 #endif
 
        /* Reserve direct mapping */
@@ -575,7 +682,8 @@ void __init setup_arch(char **cmdline_p)
        reserve_bootmem_generic(0, PAGE_SIZE);
 
        /* reserve ebda region */
-       reserve_ebda_region();
+       if (ebda_addr)
+               reserve_bootmem_generic(ebda_addr, ebda_size);
 
 #ifdef CONFIG_SMP
        /*
@@ -618,17 +726,32 @@ void __init setup_arch(char **cmdline_p)
                }
        }
 #endif
+#ifdef CONFIG_KEXEC
+       if (crashk_res.start != crashk_res.end) {
+               reserve_bootmem(crashk_res.start,
+                       crashk_res.end - crashk_res.start + 1);
+       }
+#endif
+
        paging_init();
 
        check_ioapic();
 
-#ifdef CONFIG_ACPI_BOOT
+       /*
+        * set this early, so we dont allocate cpu0
+        * if MADT list doesnt list BSP first
+        * mpparse.c/MP_processor_info() allocates logical cpu numbers.
+        */
+       cpu_set(0, cpu_present_map);
+#ifdef CONFIG_ACPI
        /*
         * Read APIC and some other early information from ACPI tables.
         */
        acpi_boot_init();
 #endif
 
+       init_cpu_to_node();
+
 #ifdef CONFIG_X86_LOCAL_APIC
        /*
         * get boot-time SMP configuration:
@@ -657,7 +780,7 @@ void __init setup_arch(char **cmdline_p)
        e820_setup_gap();
 
 #ifdef CONFIG_GART_IOMMU
-       iommu_hole_init();
+       iommu_hole_init();
 #endif
 
 #ifdef CONFIG_VT
@@ -669,7 +792,7 @@ void __init setup_arch(char **cmdline_p)
 #endif
 }
 
-static int __init get_model_name(struct cpuinfo_x86 *c)
+static int __cpuinit get_model_name(struct cpuinfo_x86 *c)
 {
        unsigned int *v;
 
@@ -685,7 +808,7 @@ static int __init get_model_name(struct cpuinfo_x86 *c)
 }
 
 
-static void __init display_cacheinfo(struct cpuinfo_x86 *c)
+static void __cpuinit display_cacheinfo(struct cpuinfo_x86 *c)
 {
        unsigned int n, dummy, eax, ebx, ecx, edx;
 
@@ -719,6 +842,24 @@ static void __init display_cacheinfo(struct cpuinfo_x86 *c)
        }
 }
 
+#ifdef CONFIG_NUMA
+static int nearby_node(int apicid)
+{
+       int i;
+       for (i = apicid - 1; i >= 0; i--) {
+               int node = apicid_to_node[i];
+               if (node != NUMA_NO_NODE && node_online(node))
+                       return node;
+       }
+       for (i = apicid + 1; i < MAX_LOCAL_APIC; i++) {
+               int node = apicid_to_node[i];
+               if (node != NUMA_NO_NODE && node_online(node))
+                       return node;
+       }
+       return first_node(node_online_map); /* Shouldn't happen */
+}
+#endif
+
 /*
  * On a AMD dual core setup the lower bits of the APIC id distingush the cores.
  * Assumes number of cores is a power of two.
@@ -727,51 +868,85 @@ static void __init amd_detect_cmp(struct cpuinfo_x86 *c)
 {
 #ifdef CONFIG_SMP
        int cpu = smp_processor_id();
-       int node = 0;
        unsigned bits;
-       if (c->x86_num_cores == 1)
-               return;
+#ifdef CONFIG_NUMA
+       int node = 0;
+       unsigned apicid = hard_smp_processor_id();
+#endif
 
        bits = 0;
-       while ((1 << bits) < c->x86_num_cores)
+       while ((1 << bits) < c->x86_max_cores)
                bits++;
 
        /* Low order bits define the core id (index of core in socket) */
        cpu_core_id[cpu] = phys_proc_id[cpu] & ((1 << bits)-1);
        /* Convert the APIC ID into the socket ID */
-       phys_proc_id[cpu] >>= bits;
+       phys_proc_id[cpu] = phys_pkg_id(bits);
 
 #ifdef CONFIG_NUMA
-       /* When an ACPI SRAT table is available use the mappings from SRAT
-          instead. */
-       if (acpi_numa <= 0) {
-               node = phys_proc_id[cpu];
-               if (!node_online(node))
-                       node = first_node(node_online_map);
-               cpu_to_node[cpu] = node;
-       } else {
-               node = cpu_to_node[cpu];
-       }
+       node = phys_proc_id[cpu];
+       if (apicid_to_node[apicid] != NUMA_NO_NODE)
+               node = apicid_to_node[apicid];
+       if (!node_online(node)) {
+               /* Two possibilities here:
+                  - The CPU is missing memory and no node was created.
+                  In that case try picking one from a nearby CPU
+                  - The APIC IDs differ from the HyperTransport node IDs
+                  which the K8 northbridge parsing fills in.
+                  Assume they are all increased by a constant offset,
+                  but in the same order as the HT nodeids.
+                  If that doesn't result in a usable node fall back to the
+                  path for the previous case.  */
+               int ht_nodeid = apicid - (phys_proc_id[0] << bits);
+               if (ht_nodeid >= 0 &&
+                   apicid_to_node[ht_nodeid] != NUMA_NO_NODE)
+                       node = apicid_to_node[ht_nodeid];
+               /* Pick a nearby node */
+               if (!node_online(node))
+                       node = nearby_node(apicid);
+       }
+       numa_set_node(cpu, node);
+
+       printk(KERN_INFO "CPU %d/%x(%d) -> Node %d -> Core %d\n",
+                       cpu, apicid, c->x86_max_cores, node, cpu_core_id[cpu]);
 #endif
-
-       printk(KERN_INFO "CPU %d(%d) -> Node %d -> Core %d\n",
-                       cpu, c->x86_num_cores, node, cpu_core_id[cpu]);
 #endif
 }
 
 static int __init init_amd(struct cpuinfo_x86 *c)
 {
        int r;
-       int level;
+       unsigned level;
+
+#ifdef CONFIG_SMP
+       unsigned long value;
+
+       /*
+        * Disable TLB flush filter by setting HWCR.FFDIS on K8
+        * bit 6 of msr C001_0015
+        *
+        * Errata 63 for SH-B3 steppings
+        * Errata 122 for all steppings (F+ have it disabled by default)
+        */
+       if (c->x86 == 15) {
+               rdmsrl(MSR_K8_HWCR, value);
+               value |= 1 << 6;
+               wrmsrl(MSR_K8_HWCR, value);
+       }
+#endif
 
        /* Bit 31 in normal CPUID used for nonstandard 3DNow ID;
           3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway */
        clear_bit(0*32+31, &c->x86_capability);
        
-       /* C-stepping K8? */
+       /* On C+ stepping K8 rep microcode works well for copy/memset */
        level = cpuid_eax(1);
-       if ((level >= 0x0f48 && level < 0x0f50) || level >= 0x0f58)
-               set_bit(X86_FEATURE_K8_C, &c->x86_capability);
+       if (c->x86 == 15 && ((level >= 0x0f48 && level < 0x0f50) || level >= 0x0f58))
+               set_bit(X86_FEATURE_REP_GOOD, &c->x86_capability);
+
+       /* Enable workaround for FXSAVE leak */
+       if (c->x86 >= 6)
+               set_bit(X86_FEATURE_FXSAVE_LEAK, &c->x86_capability);
 
        r = get_model_name(c);
        if (!r) { 
@@ -785,10 +960,12 @@ static int __init init_amd(struct cpuinfo_x86 *c)
        } 
        display_cacheinfo(c);
 
+       /* c->x86_power is 8000_0007 edx. Bit 8 is constant TSC */
+       if (c->x86_power & (1<<8))
+               set_bit(X86_FEATURE_CONSTANT_TSC, &c->x86_capability);
+
        if (c->extended_cpuid_level >= 0x80000008) {
-               c->x86_num_cores = (cpuid_ecx(0x80000008) & 0xff) + 1;
-               if (c->x86_num_cores & (c->x86_num_cores - 1))
-                       c->x86_num_cores = 1;
+               c->x86_max_cores = (cpuid_ecx(0x80000008) & 0xff) + 1;
 
                amd_detect_cmp(c);
        }
@@ -796,58 +973,47 @@ static int __init init_amd(struct cpuinfo_x86 *c)
        return r;
 }
 
-static void __init detect_ht(struct cpuinfo_x86 *c)
+static void __cpuinit detect_ht(struct cpuinfo_x86 *c)
 {
 #ifdef CONFIG_SMP
        u32     eax, ebx, ecx, edx;
-       int     index_msb, tmp;
+       int     index_msb, core_bits;
        int     cpu = smp_processor_id();
-       
+
+       cpuid(1, &eax, &ebx, &ecx, &edx);
+
+
        if (!cpu_has(c, X86_FEATURE_HT) || cpu_has(c, X86_FEATURE_CMP_LEGACY))
                return;
 
-       cpuid(1, &eax, &ebx, &ecx, &edx);
        smp_num_siblings = (ebx & 0xff0000) >> 16;
-       
+
        if (smp_num_siblings == 1) {
                printk(KERN_INFO  "CPU: Hyper-Threading is disabled\n");
-       } else if (smp_num_siblings > 1) {
-               index_msb = 31;
-               /*
-                * At this point we only support two siblings per
-                * processor package.
-                */
+       } else if (smp_num_siblings > 1 ) {
+
                if (smp_num_siblings > NR_CPUS) {
                        printk(KERN_WARNING "CPU: Unsupported number of the siblings %d", smp_num_siblings);
                        smp_num_siblings = 1;
                        return;
                }
-               tmp = smp_num_siblings;
-               while ((tmp & 0x80000000 ) == 0) {
-                       tmp <<=1 ;
-                       index_msb--;
-               }
-               if (smp_num_siblings & (smp_num_siblings - 1))
-                       index_msb++;
+
+               index_msb = get_count_order(smp_num_siblings);
                phys_proc_id[cpu] = phys_pkg_id(index_msb);
-               
+
                printk(KERN_INFO  "CPU: Physical Processor ID: %d\n",
                       phys_proc_id[cpu]);
 
-               smp_num_siblings = smp_num_siblings / c->x86_num_cores;
+               smp_num_siblings = smp_num_siblings / c->x86_max_cores;
 
-               tmp = smp_num_siblings;
-               index_msb = 31;
-               while ((tmp & 0x80000000) == 0) {
-                       tmp <<=1 ;
-                       index_msb--;
-               }
-               if (smp_num_siblings & (smp_num_siblings - 1))
-                       index_msb++;
+               index_msb = get_count_order(smp_num_siblings) ;
+
+               core_bits = get_count_order(c->x86_max_cores);
 
-               cpu_core_id[cpu] = phys_pkg_id(index_msb);
+               cpu_core_id[cpu] = phys_pkg_id(index_msb) &
+                                              ((1 << core_bits) - 1);
 
-               if (c->x86_num_cores > 1)
+               if (c->x86_max_cores > 1)
                        printk(KERN_INFO  "CPU: Processor Core ID: %d\n",
                               cpu_core_id[cpu]);
        }
@@ -857,7 +1023,7 @@ static void __init detect_ht(struct cpuinfo_x86 *c)
 /*
  * find out the number of processor cores on the die
  */
-static int __init intel_num_cpu_cores(struct cpuinfo_x86 *c)
+static int __cpuinit intel_num_cpu_cores(struct cpuinfo_x86 *c)
 {
        unsigned int eax;
 
@@ -875,7 +1041,25 @@ static int __init intel_num_cpu_cores(struct cpuinfo_x86 *c)
                return 1;
 }
 
-static void __init init_intel(struct cpuinfo_x86 *c)
+static void srat_detect_node(void)
+{
+#ifdef CONFIG_NUMA
+       unsigned node;
+       int cpu = smp_processor_id();
+
+       /* Don't do the funky fallback heuristics the AMD version employs
+          for now. */
+       node = apicid_to_node[hard_smp_processor_id()];
+       if (node == NUMA_NO_NODE)
+               node = first_node(node_online_map);
+       numa_set_node(cpu, node);
+
+       if (acpi_numa > 0)
+               printk(KERN_INFO "CPU %d -> Node %d\n", cpu, node);
+#endif
+}
+
+static void __cpuinit init_intel(struct cpuinfo_x86 *c)
 {
        /* Cache sizes */
        unsigned n;
@@ -886,16 +1070,25 @@ static void __init init_intel(struct cpuinfo_x86 *c)
                unsigned eax = cpuid_eax(0x80000008);
                c->x86_virt_bits = (eax >> 8) & 0xff;
                c->x86_phys_bits = eax & 0xff;
+               /* CPUID workaround for Intel 0F34 CPU */
+               if (c->x86_vendor == X86_VENDOR_INTEL &&
+                   c->x86 == 0xF && c->x86_model == 0x3 &&
+                   c->x86_mask == 0x4)
+                       c->x86_phys_bits = 36;
        }
 
        if (c->x86 == 15)
                c->x86_cache_alignment = c->x86_clflush_size * 2;
-       if (c->x86 >= 15)
+       if ((c->x86 == 0xf && c->x86_model >= 0x03) ||
+           (c->x86 == 0x6 && c->x86_model >= 0x0e))
                set_bit(X86_FEATURE_CONSTANT_TSC, &c->x86_capability);
-       c->x86_num_cores = intel_num_cpu_cores(c);
+       set_bit(X86_FEATURE_SYNC_RDTSC, &c->x86_capability);
+       c->x86_max_cores = intel_num_cpu_cores(c);
+
+       srat_detect_node();
 }
 
-void __init get_cpu_vendor(struct cpuinfo_x86 *c)
+static void __cpuinit get_cpu_vendor(struct cpuinfo_x86 *c)
 {
        char *v = c->x86_vendor_id;
 
@@ -916,7 +1109,7 @@ struct cpu_model_info {
 /* Do some early cpuid on the boot CPU to get some parameter that are
    needed before check_bugs. Everything advanced is in identify_cpu
    below. */
-void __init early_identify_cpu(struct cpuinfo_x86 *c)
+void __cpuinit early_identify_cpu(struct cpuinfo_x86 *c)
 {
        u32 tfms;
 
@@ -928,7 +1121,7 @@ void __init early_identify_cpu(struct cpuinfo_x86 *c)
        c->x86_model_id[0] = '\0';  /* Unset */
        c->x86_clflush_size = 64;
        c->x86_cache_alignment = c->x86_clflush_size;
-       c->x86_num_cores = 1;
+       c->x86_max_cores = 1;
        c->extended_cpuid_level = 0;
        memset(&c->x86_capability, 0, sizeof c->x86_capability);
 
@@ -951,10 +1144,10 @@ void __init early_identify_cpu(struct cpuinfo_x86 *c)
                c->x86 = (tfms >> 8) & 0xf;
                c->x86_model = (tfms >> 4) & 0xf;
                c->x86_mask = tfms & 0xf;
-               if (c->x86 == 0xf) {
+               if (c->x86 == 0xf)
                        c->x86 += (tfms >> 20) & 0xff;
+               if (c->x86 >= 0x6)
                        c->x86_model += ((tfms >> 16) & 0xF) << 4;
-               } 
                if (c->x86_capability[0] & (1<<19)) 
                        c->x86_clflush_size = ((misc >> 8) & 0xff) * 8;
        } else {
@@ -970,7 +1163,7 @@ void __init early_identify_cpu(struct cpuinfo_x86 *c)
 /*
  * This does the hard work of actually picking apart the CPU stuff...
  */
-void __init identify_cpu(struct cpuinfo_x86 *c)
+void __cpuinit identify_cpu(struct cpuinfo_x86 *c)
 {
        int i;
        u32 xlvl;
@@ -997,6 +1190,8 @@ void __init identify_cpu(struct cpuinfo_x86 *c)
                        c->x86_capability[2] = cpuid_edx(0x80860001);
        }
 
+       c->apicid = phys_pkg_id(0);
+
        /*
         * Vendor-specific initialization.  In this section we
         * canonicalize the feature flags, meaning if there are
@@ -1040,14 +1235,17 @@ void __init identify_cpu(struct cpuinfo_x86 *c)
 #ifdef CONFIG_X86_MCE
        mcheck_init(c);
 #endif
+       if (c == &boot_cpu_data)
+               mtrr_bp_init();
+       else
+               mtrr_ap_init();
 #ifdef CONFIG_NUMA
-       if (c != &boot_cpu_data)
-               numa_add_cpu(c - cpu_data);
+       numa_add_cpu(smp_processor_id());
 #endif
 }
  
 
-void __init print_cpu_info(struct cpuinfo_x86 *c)
+void __cpuinit print_cpu_info(struct cpuinfo_x86 *c)
 {
        if (c->x86_model_id[0])
                printk("%s", c->x86_model_id);
@@ -1085,7 +1283,7 @@ static int show_cpuinfo(struct seq_file *m, void *v)
                NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
                NULL, NULL, NULL, "syscall", NULL, NULL, NULL, NULL,
                NULL, NULL, NULL, NULL, "nx", NULL, "mmxext", NULL,
-               NULL, "fxsr_opt", NULL, NULL, NULL, "lm", "3dnowext", "3dnow",
+               NULL, "fxsr_opt", "rdtscp", NULL, NULL, "lm", "3dnowext", "3dnow",
 
                /* Transmeta-defined */
                "recovery", "longrun", NULL, "lrti", NULL, NULL, NULL, NULL,
@@ -1101,7 +1299,7 @@ static int show_cpuinfo(struct seq_file *m, void *v)
                NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
 
                /* Intel-defined (#2) */
-               "pni", NULL, NULL, "monitor", "ds_cpl", NULL, NULL, "est",
+               "pni", NULL, NULL, "monitor", "ds_cpl", "vmx", "smx", "est",
                "tm2", NULL, "cid", NULL, NULL, "cx16", "xtpr", NULL,
                NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
                NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
@@ -1113,7 +1311,7 @@ static int show_cpuinfo(struct seq_file *m, void *v)
                NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
 
                /* AMD-defined (#2) */
-               "lahf_lm", "cmp_legacy", NULL, NULL, NULL, NULL, NULL, NULL,
+               "lahf_lm", "cmp_legacy", "svm", NULL, "cr8_legacy", NULL, NULL, NULL,
                NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
                NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
                NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
@@ -1124,7 +1322,9 @@ static int show_cpuinfo(struct seq_file *m, void *v)
                "vid",  /* voltage id control */
                "ttp",  /* thermal trip */
                "tm",
-               "stc"
+               "stc",
+               NULL,
+               /* nothing */   /* constant_tsc - moved to flags */
        };
 
 
@@ -1150,8 +1350,11 @@ static int show_cpuinfo(struct seq_file *m, void *v)
                seq_printf(m, "stepping\t: unknown\n");
        
        if (cpu_has(c,X86_FEATURE_TSC)) {
+               unsigned int freq = cpufreq_quick_get((unsigned)(c-cpu_data));
+               if (!freq)
+                       freq = cpu_khz;
                seq_printf(m, "cpu MHz\t\t: %u.%03u\n",
-                            cpu_khz / 1000, (cpu_khz % 1000));
+                            freq / 1000, (freq % 1000));
        }
 
        /* Cache size */
@@ -1159,13 +1362,12 @@ static int show_cpuinfo(struct seq_file *m, void *v)
                seq_printf(m, "cache size\t: %d KB\n", c->x86_cache_size);
        
 #ifdef CONFIG_SMP
-       if (smp_num_siblings * c->x86_num_cores > 1) {
+       if (smp_num_siblings * c->x86_max_cores > 1) {
                int cpu = c - cpu_data;
                seq_printf(m, "physical id\t: %d\n", phys_proc_id[cpu]);
-               seq_printf(m, "siblings\t: %d\n",
-                               c->x86_num_cores * smp_num_siblings);
+               seq_printf(m, "siblings\t: %d\n", cpus_weight(cpu_core_map[cpu]));
                seq_printf(m, "core id\t\t: %d\n", cpu_core_id[cpu]);
-               seq_printf(m, "cpu cores\t: %d\n", c->x86_num_cores);
+               seq_printf(m, "cpu cores\t: %d\n", c->booted_cores);
        }
 #endif 
 
@@ -1180,8 +1382,7 @@ static int show_cpuinfo(struct seq_file *m, void *v)
        { 
                int i; 
                for ( i = 0 ; i < 32*NCAPINTS ; i++ )
-                       if ( test_bit(i, &c->x86_capability) &&
-                            x86_cap_flags[i] != NULL )
+                       if (cpu_has(c, i) && x86_cap_flags[i] != NULL)
                                seq_printf(m, " %s", x86_cap_flags[i]);
        }
                
@@ -1202,8 +1403,11 @@ static int show_cpuinfo(struct seq_file *m, void *v)
                unsigned i;
                for (i = 0; i < 32; i++) 
                        if (c->x86_power & (1 << i)) {
-                               if (i < ARRAY_SIZE(x86_power_flags))
-                                       seq_printf(m, " %s", x86_power_flags[i]);
+                               if (i < ARRAY_SIZE(x86_power_flags) &&
+                                       x86_power_flags[i])
+                                       seq_printf(m, "%s%s",
+                                               x86_power_flags[i][0]?" ":"",
+                                               x86_power_flags[i]);
                                else
                                        seq_printf(m, " [%d]", i);
                        }
@@ -1235,3 +1439,23 @@ struct seq_operations cpuinfo_op = {
        .stop = c_stop,
        .show = show_cpuinfo,
 };
+
+#ifdef CONFIG_INPUT_PCSPKR
+#include <linux/platform_device.h>
+static __init int add_pcspkr(void)
+{
+       struct platform_device *pd;
+       int ret;
+
+       pd = platform_device_alloc("pcspkr", -1);
+       if (!pd)
+               return -ENOMEM;
+
+       ret = platform_device_add(pd);
+       if (ret)
+               platform_device_put(pd);
+
+       return ret;
+}
+device_initcall(add_pcspkr);
+#endif