Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * linux/arch/i386/kernel/setup.c |
| 3 | * |
| 4 | * Copyright (C) 1995 Linus Torvalds |
| 5 | * |
| 6 | * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999 |
| 7 | * |
| 8 | * Memory region support |
| 9 | * David Parsons <orc@pell.chi.il.us>, July-August 1999 |
| 10 | * |
| 11 | * Added E820 sanitization routine (removes overlapping memory regions); |
| 12 | * Brian Moyle <bmoyle@mvista.com>, February 2001 |
| 13 | * |
| 14 | * Moved CPU detection code to cpu/${cpu}.c |
| 15 | * Patrick Mochel <mochel@osdl.org>, March 2002 |
| 16 | * |
| 17 | * Provisions for empty E820 memory regions (reported by certain BIOSes). |
| 18 | * Alex Achenbach <xela@slit.de>, December 2002. |
| 19 | * |
| 20 | */ |
| 21 | |
| 22 | /* |
| 23 | * This file handles the architecture-dependent parts of initialization |
| 24 | */ |
| 25 | |
| 26 | #include <linux/sched.h> |
| 27 | #include <linux/mm.h> |
Andy Whitcroft | 05b79bd | 2005-06-23 00:07:57 -0700 | [diff] [blame] | 28 | #include <linux/mmzone.h> |
Jon Smirl | 894673e | 2006-07-10 04:44:13 -0700 | [diff] [blame] | 29 | #include <linux/screen_info.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 30 | #include <linux/ioport.h> |
| 31 | #include <linux/acpi.h> |
| 32 | #include <linux/apm_bios.h> |
| 33 | #include <linux/initrd.h> |
| 34 | #include <linux/bootmem.h> |
| 35 | #include <linux/seq_file.h> |
| 36 | #include <linux/console.h> |
| 37 | #include <linux/mca.h> |
| 38 | #include <linux/root_dev.h> |
| 39 | #include <linux/highmem.h> |
| 40 | #include <linux/module.h> |
| 41 | #include <linux/efi.h> |
| 42 | #include <linux/init.h> |
| 43 | #include <linux/edd.h> |
| 44 | #include <linux/nodemask.h> |
Eric W. Biederman | 1bc3b91 | 2005-06-25 14:58:01 -0700 | [diff] [blame] | 45 | #include <linux/kexec.h> |
Vivek Goyal | 2030eae | 2005-06-25 14:58:20 -0700 | [diff] [blame] | 46 | #include <linux/crash_dump.h> |
Andi Kleen | e992867 | 2006-01-11 22:43:33 +0100 | [diff] [blame] | 47 | #include <linux/dmi.h> |
Dave Hansen | 22a9835 | 2006-03-27 01:16:04 -0800 | [diff] [blame] | 48 | #include <linux/pfn.h> |
Eric W. Biederman | 1bc3b91 | 2005-06-25 14:58:01 -0700 | [diff] [blame] | 49 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 50 | #include <video/edid.h> |
Eric W. Biederman | 1bc3b91 | 2005-06-25 14:58:01 -0700 | [diff] [blame] | 51 | |
Eric W. Biederman | 9635b47 | 2005-06-25 14:57:41 -0700 | [diff] [blame] | 52 | #include <asm/apic.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 53 | #include <asm/e820.h> |
| 54 | #include <asm/mpspec.h> |
keith mannthey | 9102330 | 2006-09-25 23:31:03 -0700 | [diff] [blame] | 55 | #include <asm/mmzone.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 56 | #include <asm/setup.h> |
| 57 | #include <asm/arch_hooks.h> |
| 58 | #include <asm/sections.h> |
| 59 | #include <asm/io_apic.h> |
| 60 | #include <asm/ist.h> |
| 61 | #include <asm/io.h> |
Zachary Amsden | 7ce0bcf | 2007-02-13 13:26:21 +0100 | [diff] [blame] | 62 | #include <asm/vmi.h> |
Jeremy Fitzhardinge | e75eac3 | 2006-06-25 05:46:50 -0700 | [diff] [blame] | 63 | #include <setup_arch.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 64 | #include <bios_ebda.h> |
| 65 | |
| 66 | /* This value is set up by the early boot code to point to the value |
| 67 | immediately after the boot time page tables. It contains a *physical* |
| 68 | address, and must not be in the .bss segment! */ |
| 69 | unsigned long init_pg_tables_end __initdata = ~0UL; |
| 70 | |
Li Shaohua | 0bb3184 | 2005-06-25 14:54:55 -0700 | [diff] [blame] | 71 | int disable_pse __devinitdata = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 72 | |
| 73 | /* |
| 74 | * Machine setup.. |
| 75 | */ |
bibo,mao | 269c2d8 | 2006-12-07 02:14:06 +0100 | [diff] [blame] | 76 | extern struct resource code_resource; |
| 77 | extern struct resource data_resource; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 78 | |
| 79 | /* cpu data as detected by the assembly code in head.S */ |
Vivek Goyal | 4a5d107 | 2007-01-11 01:52:44 +0100 | [diff] [blame] | 80 | struct cpuinfo_x86 new_cpu_data __cpuinitdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 81 | /* common cpu data for all cpus */ |
Christoph Lameter | c3d8c14 | 2005-09-06 15:16:33 -0700 | [diff] [blame] | 82 | struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 }; |
Alexey Dobriyan | 129f694 | 2005-06-23 00:08:33 -0700 | [diff] [blame] | 83 | EXPORT_SYMBOL(boot_cpu_data); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 84 | |
| 85 | unsigned long mmu_cr4_features; |
| 86 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 87 | /* for MCA, but anyone else can use it if they want */ |
| 88 | unsigned int machine_id; |
Alexey Dobriyan | 129f694 | 2005-06-23 00:08:33 -0700 | [diff] [blame] | 89 | #ifdef CONFIG_MCA |
| 90 | EXPORT_SYMBOL(machine_id); |
| 91 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 92 | unsigned int machine_submodel_id; |
| 93 | unsigned int BIOS_revision; |
| 94 | unsigned int mca_pentium_flag; |
| 95 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 96 | /* Boot loader ID as an integer, for the benefit of proc_dointvec */ |
| 97 | int bootloader_type; |
| 98 | |
| 99 | /* user-defined highmem size */ |
| 100 | static unsigned int highmem_pages = -1; |
| 101 | |
| 102 | /* |
| 103 | * Setup options |
| 104 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 105 | struct screen_info screen_info; |
Alexey Dobriyan | 129f694 | 2005-06-23 00:08:33 -0700 | [diff] [blame] | 106 | EXPORT_SYMBOL(screen_info); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 107 | struct apm_info apm_info; |
Alexey Dobriyan | 129f694 | 2005-06-23 00:08:33 -0700 | [diff] [blame] | 108 | EXPORT_SYMBOL(apm_info); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 109 | struct edid_info edid_info; |
Antonino A. Daplas | 5e518d7 | 2005-09-09 13:04:34 -0700 | [diff] [blame] | 110 | EXPORT_SYMBOL_GPL(edid_info); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 111 | struct ist_info ist_info; |
Alexey Dobriyan | 129f694 | 2005-06-23 00:08:33 -0700 | [diff] [blame] | 112 | #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \ |
| 113 | defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE) |
| 114 | EXPORT_SYMBOL(ist_info); |
| 115 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 116 | |
| 117 | extern void early_cpu_init(void); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 118 | extern int root_mountflags; |
| 119 | |
| 120 | unsigned long saved_videomode; |
| 121 | |
| 122 | #define RAMDISK_IMAGE_START_MASK 0x07FF |
| 123 | #define RAMDISK_PROMPT_FLAG 0x8000 |
| 124 | #define RAMDISK_LOAD_FLAG 0x4000 |
| 125 | |
Alon Bar-Lev | 4e498b6 | 2007-02-12 00:54:11 -0800 | [diff] [blame] | 126 | static char __initdata command_line[COMMAND_LINE_SIZE]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 127 | |
H. Peter Anvin | 48c7ae6 | 2007-07-11 12:18:35 -0700 | [diff] [blame] | 128 | struct boot_params __initdata boot_params; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 129 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 130 | #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE) |
| 131 | struct edd edd; |
| 132 | #ifdef CONFIG_EDD_MODULE |
| 133 | EXPORT_SYMBOL(edd); |
| 134 | #endif |
| 135 | /** |
| 136 | * copy_edd() - Copy the BIOS EDD information |
| 137 | * from boot_params into a safe place. |
| 138 | * |
| 139 | */ |
| 140 | static inline void copy_edd(void) |
| 141 | { |
| 142 | memcpy(edd.mbr_signature, EDD_MBR_SIGNATURE, sizeof(edd.mbr_signature)); |
| 143 | memcpy(edd.edd_info, EDD_BUF, sizeof(edd.edd_info)); |
| 144 | edd.mbr_signature_nr = EDD_MBR_SIG_NR; |
| 145 | edd.edd_info_nr = EDD_NR; |
| 146 | } |
| 147 | #else |
| 148 | static inline void copy_edd(void) |
| 149 | { |
| 150 | } |
| 151 | #endif |
| 152 | |
bibo,mao | cef518e | 2006-12-07 02:14:06 +0100 | [diff] [blame] | 153 | int __initdata user_defined_memmap = 0; |
Rusty Russell | 1a3f239 | 2006-09-26 10:52:32 +0200 | [diff] [blame] | 154 | |
| 155 | /* |
| 156 | * "mem=nopentium" disables the 4MB page tables. |
| 157 | * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM |
| 158 | * to <mem>, overriding the bios size. |
| 159 | * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from |
| 160 | * <start> to <start>+<mem>, overriding the bios size. |
| 161 | * |
| 162 | * HPA tells me bootloaders need to parse mem=, so no new |
| 163 | * option should be mem= [also see Documentation/i386/boot.txt] |
| 164 | */ |
| 165 | static int __init parse_mem(char *arg) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 166 | { |
Rusty Russell | 1a3f239 | 2006-09-26 10:52:32 +0200 | [diff] [blame] | 167 | if (!arg) |
| 168 | return -EINVAL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 169 | |
Rusty Russell | 1a3f239 | 2006-09-26 10:52:32 +0200 | [diff] [blame] | 170 | if (strcmp(arg, "nopentium") == 0) { |
| 171 | clear_bit(X86_FEATURE_PSE, boot_cpu_data.x86_capability); |
| 172 | disable_pse = 1; |
| 173 | } else { |
| 174 | /* If the user specifies memory size, we |
| 175 | * limit the BIOS-provided memory map to |
| 176 | * that size. exactmap can be used to specify |
| 177 | * the exact map. mem=number can be used to |
| 178 | * trim the existing memory map. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 179 | */ |
Rusty Russell | 1a3f239 | 2006-09-26 10:52:32 +0200 | [diff] [blame] | 180 | unsigned long long mem_size; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 181 | |
Rusty Russell | 1a3f239 | 2006-09-26 10:52:32 +0200 | [diff] [blame] | 182 | mem_size = memparse(arg, &arg); |
| 183 | limit_regions(mem_size); |
| 184 | user_defined_memmap = 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 185 | } |
Rusty Russell | 1a3f239 | 2006-09-26 10:52:32 +0200 | [diff] [blame] | 186 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 187 | } |
Rusty Russell | 1a3f239 | 2006-09-26 10:52:32 +0200 | [diff] [blame] | 188 | early_param("mem", parse_mem); |
| 189 | |
Rusty Russell | 1a3f239 | 2006-09-26 10:52:32 +0200 | [diff] [blame] | 190 | #ifdef CONFIG_PROC_VMCORE |
| 191 | /* elfcorehdr= specifies the location of elf core header |
| 192 | * stored by the crashed kernel. |
| 193 | */ |
| 194 | static int __init parse_elfcorehdr(char *arg) |
| 195 | { |
| 196 | if (!arg) |
| 197 | return -EINVAL; |
| 198 | |
| 199 | elfcorehdr_addr = memparse(arg, &arg); |
| 200 | return 0; |
| 201 | } |
| 202 | early_param("elfcorehdr", parse_elfcorehdr); |
| 203 | #endif /* CONFIG_PROC_VMCORE */ |
| 204 | |
| 205 | /* |
| 206 | * highmem=size forces highmem to be exactly 'size' bytes. |
| 207 | * This works even on boxes that have no highmem otherwise. |
| 208 | * This also works to reduce highmem size on bigger boxes. |
| 209 | */ |
| 210 | static int __init parse_highmem(char *arg) |
| 211 | { |
| 212 | if (!arg) |
| 213 | return -EINVAL; |
| 214 | |
| 215 | highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT; |
| 216 | return 0; |
| 217 | } |
| 218 | early_param("highmem", parse_highmem); |
| 219 | |
| 220 | /* |
| 221 | * vmalloc=size forces the vmalloc area to be exactly 'size' |
| 222 | * bytes. This can be used to increase (or decrease) the |
| 223 | * vmalloc area - the default is 128m. |
| 224 | */ |
| 225 | static int __init parse_vmalloc(char *arg) |
| 226 | { |
| 227 | if (!arg) |
| 228 | return -EINVAL; |
| 229 | |
| 230 | __VMALLOC_RESERVE = memparse(arg, &arg); |
| 231 | return 0; |
| 232 | } |
| 233 | early_param("vmalloc", parse_vmalloc); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 234 | |
| 235 | /* |
Zachary Amsden | 461a9af | 2006-09-25 23:32:25 -0700 | [diff] [blame] | 236 | * reservetop=size reserves a hole at the top of the kernel address space which |
| 237 | * a hypervisor can load into later. Needed for dynamically loaded hypervisors, |
| 238 | * so relocating the fixmap can be done before paging initialization. |
| 239 | */ |
| 240 | static int __init parse_reservetop(char *arg) |
| 241 | { |
| 242 | unsigned long address; |
| 243 | |
| 244 | if (!arg) |
| 245 | return -EINVAL; |
| 246 | |
| 247 | address = memparse(arg, &arg); |
| 248 | reserve_top_address(address); |
| 249 | return 0; |
| 250 | } |
| 251 | early_param("reservetop", parse_reservetop); |
| 252 | |
| 253 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 254 | * Determine low and high memory ranges: |
| 255 | */ |
| 256 | unsigned long __init find_max_low_pfn(void) |
| 257 | { |
| 258 | unsigned long max_low_pfn; |
| 259 | |
| 260 | max_low_pfn = max_pfn; |
| 261 | if (max_low_pfn > MAXMEM_PFN) { |
| 262 | if (highmem_pages == -1) |
| 263 | highmem_pages = max_pfn - MAXMEM_PFN; |
| 264 | if (highmem_pages + MAXMEM_PFN < max_pfn) |
| 265 | max_pfn = MAXMEM_PFN + highmem_pages; |
| 266 | if (highmem_pages + MAXMEM_PFN > max_pfn) { |
| 267 | printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn - MAXMEM_PFN), pages_to_mb(highmem_pages)); |
| 268 | highmem_pages = 0; |
| 269 | } |
| 270 | max_low_pfn = MAXMEM_PFN; |
| 271 | #ifndef CONFIG_HIGHMEM |
| 272 | /* Maximum memory usable is what is directly addressable */ |
| 273 | printk(KERN_WARNING "Warning only %ldMB will be used.\n", |
| 274 | MAXMEM>>20); |
| 275 | if (max_pfn > MAX_NONPAE_PFN) |
William Lee Irwin III | c673f1a | 2007-07-21 17:11:13 +0200 | [diff] [blame] | 276 | printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 277 | else |
| 278 | printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n"); |
| 279 | max_pfn = MAXMEM_PFN; |
| 280 | #else /* !CONFIG_HIGHMEM */ |
William Lee Irwin III | c673f1a | 2007-07-21 17:11:13 +0200 | [diff] [blame] | 281 | #ifndef CONFIG_HIGHMEM64G |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 282 | if (max_pfn > MAX_NONPAE_PFN) { |
| 283 | max_pfn = MAX_NONPAE_PFN; |
| 284 | printk(KERN_WARNING "Warning only 4GB will be used.\n"); |
William Lee Irwin III | c673f1a | 2007-07-21 17:11:13 +0200 | [diff] [blame] | 285 | printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 286 | } |
William Lee Irwin III | c673f1a | 2007-07-21 17:11:13 +0200 | [diff] [blame] | 287 | #endif /* !CONFIG_HIGHMEM64G */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 288 | #endif /* !CONFIG_HIGHMEM */ |
| 289 | } else { |
| 290 | if (highmem_pages == -1) |
| 291 | highmem_pages = 0; |
| 292 | #ifdef CONFIG_HIGHMEM |
| 293 | if (highmem_pages >= max_pfn) { |
| 294 | printk(KERN_ERR "highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages), pages_to_mb(max_pfn)); |
| 295 | highmem_pages = 0; |
| 296 | } |
| 297 | if (highmem_pages) { |
| 298 | if (max_low_pfn-highmem_pages < 64*1024*1024/PAGE_SIZE){ |
| 299 | printk(KERN_ERR "highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages)); |
| 300 | highmem_pages = 0; |
| 301 | } |
| 302 | max_low_pfn -= highmem_pages; |
| 303 | } |
| 304 | #else |
| 305 | if (highmem_pages) |
| 306 | printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n"); |
| 307 | #endif |
| 308 | } |
| 309 | return max_low_pfn; |
| 310 | } |
| 311 | |
| 312 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 313 | * workaround for Dell systems that neglect to reserve EBDA |
| 314 | */ |
| 315 | static void __init reserve_ebda_region(void) |
| 316 | { |
| 317 | unsigned int addr; |
| 318 | addr = get_bios_ebda(); |
| 319 | if (addr) |
| 320 | reserve_bootmem(addr, PAGE_SIZE); |
| 321 | } |
| 322 | |
Andy Whitcroft | 05b79bd | 2005-06-23 00:07:57 -0700 | [diff] [blame] | 323 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 324 | void __init setup_bootmem_allocator(void); |
| 325 | static unsigned long __init setup_memory(void) |
| 326 | { |
| 327 | /* |
| 328 | * partially used pages are not usable - thus |
| 329 | * we are rounding upwards: |
| 330 | */ |
| 331 | min_low_pfn = PFN_UP(init_pg_tables_end); |
| 332 | |
| 333 | find_max_pfn(); |
| 334 | |
| 335 | max_low_pfn = find_max_low_pfn(); |
| 336 | |
| 337 | #ifdef CONFIG_HIGHMEM |
| 338 | highstart_pfn = highend_pfn = max_pfn; |
| 339 | if (max_pfn > max_low_pfn) { |
| 340 | highstart_pfn = max_low_pfn; |
| 341 | } |
| 342 | printk(KERN_NOTICE "%ldMB HIGHMEM available.\n", |
| 343 | pages_to_mb(highend_pfn - highstart_pfn)); |
Jan Beulich | ba9c231 | 2006-09-26 10:52:31 +0200 | [diff] [blame] | 344 | num_physpages = highend_pfn; |
| 345 | high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1; |
| 346 | #else |
| 347 | num_physpages = max_low_pfn; |
| 348 | high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1; |
| 349 | #endif |
| 350 | #ifdef CONFIG_FLATMEM |
| 351 | max_mapnr = num_physpages; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 352 | #endif |
| 353 | printk(KERN_NOTICE "%ldMB LOWMEM available.\n", |
| 354 | pages_to_mb(max_low_pfn)); |
| 355 | |
| 356 | setup_bootmem_allocator(); |
| 357 | |
| 358 | return max_low_pfn; |
| 359 | } |
| 360 | |
| 361 | void __init zone_sizes_init(void) |
| 362 | { |
Mel Gorman | 6391af1 | 2006-10-11 01:20:39 -0700 | [diff] [blame] | 363 | unsigned long max_zone_pfns[MAX_NR_ZONES]; |
| 364 | memset(max_zone_pfns, 0, sizeof(max_zone_pfns)); |
| 365 | max_zone_pfns[ZONE_DMA] = |
| 366 | virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT; |
| 367 | max_zone_pfns[ZONE_NORMAL] = max_low_pfn; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 368 | #ifdef CONFIG_HIGHMEM |
Mel Gorman | 6391af1 | 2006-10-11 01:20:39 -0700 | [diff] [blame] | 369 | max_zone_pfns[ZONE_HIGHMEM] = highend_pfn; |
Mel Gorman | 4cfee88 | 2006-09-27 01:49:51 -0700 | [diff] [blame] | 370 | add_active_range(0, 0, highend_pfn); |
| 371 | #else |
Mel Gorman | 4cfee88 | 2006-09-27 01:49:51 -0700 | [diff] [blame] | 372 | add_active_range(0, 0, max_low_pfn); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 373 | #endif |
Mel Gorman | 4cfee88 | 2006-09-27 01:49:51 -0700 | [diff] [blame] | 374 | |
| 375 | free_area_init_nodes(max_zone_pfns); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 376 | } |
| 377 | #else |
Andy Whitcroft | 05b79bd | 2005-06-23 00:07:57 -0700 | [diff] [blame] | 378 | extern unsigned long __init setup_memory(void); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 379 | extern void zone_sizes_init(void); |
Andy Whitcroft | 05b79bd | 2005-06-23 00:07:57 -0700 | [diff] [blame] | 380 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 381 | |
| 382 | void __init setup_bootmem_allocator(void) |
| 383 | { |
| 384 | unsigned long bootmap_size; |
| 385 | /* |
| 386 | * Initialize the boot-time allocator (with low memory only): |
| 387 | */ |
| 388 | bootmap_size = init_bootmem(min_low_pfn, max_low_pfn); |
| 389 | |
| 390 | register_bootmem_low_pages(max_low_pfn); |
| 391 | |
| 392 | /* |
| 393 | * Reserve the bootmem bitmap itself as well. We do this in two |
| 394 | * steps (first step was init_bootmem()) because this catches |
| 395 | * the (very unlikely) case of us accidentally initializing the |
| 396 | * bootmem allocator with an invalid RAM area. |
| 397 | */ |
Eric W. Biederman | 8621b81 | 2006-12-07 02:14:03 +0100 | [diff] [blame] | 398 | reserve_bootmem(__pa_symbol(_text), (PFN_PHYS(min_low_pfn) + |
| 399 | bootmap_size + PAGE_SIZE-1) - __pa_symbol(_text)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 400 | |
| 401 | /* |
| 402 | * reserve physical page 0 - it's a special BIOS page on many boxes, |
| 403 | * enabling clean reboots, SMP operation, laptop functions. |
| 404 | */ |
| 405 | reserve_bootmem(0, PAGE_SIZE); |
| 406 | |
| 407 | /* reserve EBDA region, it's a 4K region */ |
| 408 | reserve_ebda_region(); |
| 409 | |
| 410 | /* could be an AMD 768MPX chipset. Reserve a page before VGA to prevent |
| 411 | PCI prefetch into it (errata #56). Usually the page is reserved anyways, |
| 412 | unless you have no PS/2 mouse plugged in. */ |
| 413 | if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD && |
| 414 | boot_cpu_data.x86 == 6) |
| 415 | reserve_bootmem(0xa0000 - 4096, 4096); |
| 416 | |
| 417 | #ifdef CONFIG_SMP |
| 418 | /* |
| 419 | * But first pinch a few for the stack/trampoline stuff |
| 420 | * FIXME: Don't need the extra page at 4K, but need to fix |
| 421 | * trampoline before removing it. (see the GDT stuff) |
| 422 | */ |
| 423 | reserve_bootmem(PAGE_SIZE, PAGE_SIZE); |
| 424 | #endif |
Len Brown | 673d5b4 | 2007-07-28 03:33:16 -0400 | [diff] [blame] | 425 | #ifdef CONFIG_ACPI_SLEEP |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 426 | /* |
| 427 | * Reserve low memory region for sleep support. |
| 428 | */ |
| 429 | acpi_reserve_bootmem(); |
| 430 | #endif |
| 431 | #ifdef CONFIG_X86_FIND_SMP_CONFIG |
| 432 | /* |
| 433 | * Find and reserve possible boot-time SMP configuration: |
| 434 | */ |
| 435 | find_smp_config(); |
| 436 | #endif |
keith mannthey | 9102330 | 2006-09-25 23:31:03 -0700 | [diff] [blame] | 437 | numa_kva_reserve(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 438 | #ifdef CONFIG_BLK_DEV_INITRD |
| 439 | if (LOADER_TYPE && INITRD_START) { |
| 440 | if (INITRD_START + INITRD_SIZE <= (max_low_pfn << PAGE_SHIFT)) { |
| 441 | reserve_bootmem(INITRD_START, INITRD_SIZE); |
Henry Nestler | 19e5d9c | 2006-12-06 20:37:45 -0800 | [diff] [blame] | 442 | initrd_start = INITRD_START + PAGE_OFFSET; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 443 | initrd_end = initrd_start+INITRD_SIZE; |
| 444 | } |
| 445 | else { |
| 446 | printk(KERN_ERR "initrd extends beyond end of memory " |
| 447 | "(0x%08lx > 0x%08lx)\ndisabling initrd\n", |
| 448 | INITRD_START + INITRD_SIZE, |
| 449 | max_low_pfn << PAGE_SHIFT); |
| 450 | initrd_start = 0; |
| 451 | } |
| 452 | } |
| 453 | #endif |
Eric W. Biederman | 1bc3b91 | 2005-06-25 14:58:01 -0700 | [diff] [blame] | 454 | #ifdef CONFIG_KEXEC |
| 455 | if (crashk_res.start != crashk_res.end) |
| 456 | reserve_bootmem(crashk_res.start, |
| 457 | crashk_res.end - crashk_res.start + 1); |
| 458 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 459 | } |
| 460 | |
| 461 | /* |
| 462 | * The node 0 pgdat is initialized before all of these because |
| 463 | * it's needed for bootmem. node>0 pgdats have their virtual |
| 464 | * space allocated before the pagetables are in place to access |
| 465 | * them, so they can't be cleared then. |
| 466 | * |
| 467 | * This should all compile down to nothing when NUMA is off. |
| 468 | */ |
Adrian Bunk | 10079ae | 2007-07-21 17:10:27 +0200 | [diff] [blame] | 469 | static void __init remapped_pgdat_init(void) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 470 | { |
| 471 | int nid; |
| 472 | |
| 473 | for_each_online_node(nid) { |
| 474 | if (nid != 0) |
| 475 | memset(NODE_DATA(nid), 0, sizeof(struct pglist_data)); |
| 476 | } |
| 477 | } |
| 478 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 479 | #ifdef CONFIG_MCA |
| 480 | static void set_mca_bus(int x) |
| 481 | { |
| 482 | MCA_bus = x; |
| 483 | } |
| 484 | #else |
| 485 | static void set_mca_bus(int x) { } |
| 486 | #endif |
| 487 | |
Rusty Russell | d3561b7 | 2006-12-07 02:14:07 +0100 | [diff] [blame] | 488 | /* Overridden in paravirt.c if CONFIG_PARAVIRT */ |
Vivek Goyal | 88d2032 | 2007-01-11 01:52:44 +0100 | [diff] [blame] | 489 | char * __init __attribute__((weak)) memory_setup(void) |
Rusty Russell | d3561b7 | 2006-12-07 02:14:07 +0100 | [diff] [blame] | 490 | { |
| 491 | return machine_specific_memory_setup(); |
| 492 | } |
| 493 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 494 | /* |
| 495 | * Determine if we were loaded by an EFI loader. If so, then we have also been |
| 496 | * passed the efi memmap, systab, etc., so we should use these data structures |
| 497 | * for initialization. Note, the efi init code path is determined by the |
| 498 | * global efi_enabled. This allows the same kernel image to be used on existing |
| 499 | * systems (with a traditional BIOS) as well as on EFI systems. |
| 500 | */ |
| 501 | void __init setup_arch(char **cmdline_p) |
| 502 | { |
| 503 | unsigned long max_low_pfn; |
| 504 | |
| 505 | memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data)); |
| 506 | pre_setup_arch_hook(); |
| 507 | early_cpu_init(); |
| 508 | |
| 509 | /* |
| 510 | * FIXME: This isn't an official loader_type right |
| 511 | * now but does currently work with elilo. |
| 512 | * If we were configured as an EFI kernel, check to make |
| 513 | * sure that we were loaded correctly from elilo and that |
| 514 | * the system table is valid. If not, then initialize normally. |
| 515 | */ |
| 516 | #ifdef CONFIG_EFI |
| 517 | if ((LOADER_TYPE == 0x50) && EFI_SYSTAB) |
| 518 | efi_enabled = 1; |
| 519 | #endif |
| 520 | |
| 521 | ROOT_DEV = old_decode_dev(ORIG_ROOT_DEV); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 522 | screen_info = SCREEN_INFO; |
| 523 | edid_info = EDID_INFO; |
| 524 | apm_info.bios = APM_BIOS_INFO; |
| 525 | ist_info = IST_INFO; |
| 526 | saved_videomode = VIDEO_MODE; |
| 527 | if( SYS_DESC_TABLE.length != 0 ) { |
| 528 | set_mca_bus(SYS_DESC_TABLE.table[3] & 0x2); |
| 529 | machine_id = SYS_DESC_TABLE.table[0]; |
| 530 | machine_submodel_id = SYS_DESC_TABLE.table[1]; |
| 531 | BIOS_revision = SYS_DESC_TABLE.table[2]; |
| 532 | } |
| 533 | bootloader_type = LOADER_TYPE; |
| 534 | |
| 535 | #ifdef CONFIG_BLK_DEV_RAM |
| 536 | rd_image_start = RAMDISK_FLAGS & RAMDISK_IMAGE_START_MASK; |
| 537 | rd_prompt = ((RAMDISK_FLAGS & RAMDISK_PROMPT_FLAG) != 0); |
| 538 | rd_doload = ((RAMDISK_FLAGS & RAMDISK_LOAD_FLAG) != 0); |
| 539 | #endif |
| 540 | ARCH_SETUP |
| 541 | if (efi_enabled) |
| 542 | efi_init(); |
| 543 | else { |
| 544 | printk(KERN_INFO "BIOS-provided physical RAM map:\n"); |
Rusty Russell | d3561b7 | 2006-12-07 02:14:07 +0100 | [diff] [blame] | 545 | print_memory_map(memory_setup()); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 546 | } |
| 547 | |
| 548 | copy_edd(); |
| 549 | |
| 550 | if (!MOUNT_ROOT_RDONLY) |
| 551 | root_mountflags &= ~MS_RDONLY; |
| 552 | init_mm.start_code = (unsigned long) _text; |
| 553 | init_mm.end_code = (unsigned long) _etext; |
| 554 | init_mm.end_data = (unsigned long) _edata; |
| 555 | init_mm.brk = init_pg_tables_end + PAGE_OFFSET; |
| 556 | |
| 557 | code_resource.start = virt_to_phys(_text); |
| 558 | code_resource.end = virt_to_phys(_etext)-1; |
| 559 | data_resource.start = virt_to_phys(_etext); |
| 560 | data_resource.end = virt_to_phys(_edata)-1; |
| 561 | |
Rusty Russell | 1a3f239 | 2006-09-26 10:52:32 +0200 | [diff] [blame] | 562 | parse_early_param(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 563 | |
Rusty Russell | 1a3f239 | 2006-09-26 10:52:32 +0200 | [diff] [blame] | 564 | if (user_defined_memmap) { |
| 565 | printk(KERN_INFO "user-defined physical RAM map:\n"); |
| 566 | print_memory_map("user"); |
Stas Sergeev | 99b7de3 | 2006-03-23 02:59:41 -0800 | [diff] [blame] | 567 | } |
Rusty Russell | 1a3f239 | 2006-09-26 10:52:32 +0200 | [diff] [blame] | 568 | |
Alon Bar-Lev | 4e498b6 | 2007-02-12 00:54:11 -0800 | [diff] [blame] | 569 | strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE); |
Rusty Russell | 1a3f239 | 2006-09-26 10:52:32 +0200 | [diff] [blame] | 570 | *cmdline_p = command_line; |
Stas Sergeev | 99b7de3 | 2006-03-23 02:59:41 -0800 | [diff] [blame] | 571 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 572 | max_low_pfn = setup_memory(); |
| 573 | |
Zachary Amsden | 7ce0bcf | 2007-02-13 13:26:21 +0100 | [diff] [blame] | 574 | #ifdef CONFIG_VMI |
| 575 | /* |
| 576 | * Must be after max_low_pfn is determined, and before kernel |
| 577 | * pagetables are setup. |
| 578 | */ |
| 579 | vmi_init(); |
| 580 | #endif |
| 581 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 582 | /* |
| 583 | * NOTE: before this point _nobody_ is allowed to allocate |
| 584 | * any memory using the bootmem allocator. Although the |
| 585 | * alloctor is now initialised only the first 8Mb of the kernel |
| 586 | * virtual address space has been mapped. All allocations before |
| 587 | * paging_init() has completed must use the alloc_bootmem_low_pages() |
| 588 | * variant (which allocates DMA'able memory) and care must be taken |
| 589 | * not to exceed the 8Mb limit. |
| 590 | */ |
| 591 | |
| 592 | #ifdef CONFIG_SMP |
| 593 | smp_alloc_memory(); /* AP processor realmode stacks in low memory*/ |
| 594 | #endif |
| 595 | paging_init(); |
| 596 | remapped_pgdat_init(); |
Andy Whitcroft | 05b79bd | 2005-06-23 00:07:57 -0700 | [diff] [blame] | 597 | sparse_init(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 598 | zone_sizes_init(); |
| 599 | |
| 600 | /* |
| 601 | * NOTE: at this point the bootmem allocator is fully available. |
| 602 | */ |
| 603 | |
Jeremy Fitzhardinge | 6996d3b | 2007-07-17 18:37:03 -0700 | [diff] [blame] | 604 | paravirt_post_allocator_init(); |
| 605 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 606 | dmi_scan_machine(); |
| 607 | |
| 608 | #ifdef CONFIG_X86_GENERICARCH |
Rusty Russell | 1a3f239 | 2006-09-26 10:52:32 +0200 | [diff] [blame] | 609 | generic_apic_probe(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 610 | #endif |
| 611 | if (efi_enabled) |
| 612 | efi_map_memmap(); |
| 613 | |
Len Brown | 888ba6c | 2005-08-24 12:07:20 -0400 | [diff] [blame] | 614 | #ifdef CONFIG_ACPI |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 615 | /* |
| 616 | * Parse the ACPI tables for possible boot-time SMP configuration. |
| 617 | */ |
| 618 | acpi_boot_table_init(); |
Andy Currid | d44647b | 2006-06-08 00:43:38 -0700 | [diff] [blame] | 619 | #endif |
| 620 | |
Andi Kleen | f157cbb | 2006-09-26 10:52:41 +0200 | [diff] [blame] | 621 | #ifdef CONFIG_PCI |
Andy Currid | d44647b | 2006-06-08 00:43:38 -0700 | [diff] [blame] | 622 | #ifdef CONFIG_X86_IO_APIC |
| 623 | check_acpi_pci(); /* Checks more than just ACPI actually */ |
| 624 | #endif |
Andi Kleen | f157cbb | 2006-09-26 10:52:41 +0200 | [diff] [blame] | 625 | #endif |
Andy Currid | d44647b | 2006-06-08 00:43:38 -0700 | [diff] [blame] | 626 | |
| 627 | #ifdef CONFIG_ACPI |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 628 | acpi_boot_init(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 629 | |
Venkatesh Pallipadi | 911a62d | 2005-09-03 15:56:31 -0700 | [diff] [blame] | 630 | #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC) |
| 631 | if (def_to_bigsmp) |
| 632 | printk(KERN_WARNING "More than 8 CPUs detected and " |
| 633 | "CONFIG_X86_PC cannot handle it.\nUse " |
| 634 | "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n"); |
| 635 | #endif |
| 636 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 637 | #ifdef CONFIG_X86_LOCAL_APIC |
| 638 | if (smp_found_config) |
| 639 | get_smp_config(); |
| 640 | #endif |
| 641 | |
Yasunori Goto | 5c95da9 | 2006-12-22 01:09:54 -0800 | [diff] [blame] | 642 | e820_register_memory(); |
Rafael J. Wysocki | 1c10070 | 2007-07-21 17:11:09 +0200 | [diff] [blame] | 643 | e820_mark_nosave_regions(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 644 | |
| 645 | #ifdef CONFIG_VT |
| 646 | #if defined(CONFIG_VGA_CONSOLE) |
| 647 | if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY)) |
| 648 | conswitchp = &vga_con; |
| 649 | #elif defined(CONFIG_DUMMY_CONSOLE) |
| 650 | conswitchp = &dummy_con; |
| 651 | #endif |
| 652 | #endif |
| 653 | } |