blob: ccf3595202fa6a75073370b58b7b864f201384e8 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002 * Copyright (C) 1995 Linus Torvalds
3 *
4 * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
5 *
6 * Memory region support
7 * David Parsons <orc@pell.chi.il.us>, July-August 1999
8 *
9 * Added E820 sanitization routine (removes overlapping memory regions);
10 * Brian Moyle <bmoyle@mvista.com>, February 2001
11 *
12 * Moved CPU detection code to cpu/${cpu}.c
13 * Patrick Mochel <mochel@osdl.org>, March 2002
14 *
15 * Provisions for empty E820 memory regions (reported by certain BIOSes).
16 * Alex Achenbach <xela@slit.de>, December 2002.
17 *
18 */
19
20/*
21 * This file handles the architecture-dependent parts of initialization
22 */
23
24#include <linux/sched.h>
25#include <linux/mm.h>
Andy Whitcroft05b79bd2005-06-23 00:07:57 -070026#include <linux/mmzone.h>
Jon Smirl894673e2006-07-10 04:44:13 -070027#include <linux/screen_info.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070028#include <linux/ioport.h>
29#include <linux/acpi.h>
30#include <linux/apm_bios.h>
31#include <linux/initrd.h>
32#include <linux/bootmem.h>
33#include <linux/seq_file.h>
34#include <linux/console.h>
35#include <linux/mca.h>
36#include <linux/root_dev.h>
37#include <linux/highmem.h>
38#include <linux/module.h>
39#include <linux/efi.h>
40#include <linux/init.h>
41#include <linux/edd.h>
Konrad Rzeszutek138fe4e2008-04-09 19:50:41 -070042#include <linux/iscsi_ibft.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#include <linux/nodemask.h>
Eric W. Biederman1bc3b912005-06-25 14:58:01 -070044#include <linux/kexec.h>
Vivek Goyal2030eae2005-06-25 14:58:20 -070045#include <linux/crash_dump.h>
Andi Kleene9928672006-01-11 22:43:33 +010046#include <linux/dmi.h>
Dave Hansen22a98352006-03-27 01:16:04 -080047#include <linux/pfn.h>
Thomas Gleixner376ff032008-01-30 13:30:16 +010048#include <linux/pci.h>
Bernhard Kaindlf212ec42008-01-30 13:34:11 +010049#include <linux/init_ohci1394_dma.h>
Glauber de Oliveira Costa790c73f2008-02-15 17:52:48 -020050#include <linux/kvm_para.h>
Eric W. Biederman1bc3b912005-06-25 14:58:01 -070051
Linus Torvalds1da177e2005-04-16 15:20:36 -070052#include <video/edid.h>
Eric W. Biederman1bc3b912005-06-25 14:58:01 -070053
Yinghai Lu093af8d2008-01-30 13:33:32 +010054#include <asm/mtrr.h>
Eric W. Biederman9635b472005-06-25 14:57:41 -070055#include <asm/apic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070056#include <asm/e820.h>
57#include <asm/mpspec.h>
keith mannthey91023302006-09-25 23:31:03 -070058#include <asm/mmzone.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070059#include <asm/setup.h>
60#include <asm/arch_hooks.h>
61#include <asm/sections.h>
62#include <asm/io_apic.h>
63#include <asm/ist.h>
64#include <asm/io.h>
Zachary Amsden7ce0bcf2007-02-13 13:26:21 +010065#include <asm/vmi.h>
Jeremy Fitzhardingee75eac32006-06-25 05:46:50 -070066#include <setup_arch.h>
Alexey Starikovskiyce3fe6b2008-03-17 22:08:17 +030067#include <asm/bios_ebda.h>
Bernhard Walle00bf4092007-10-21 16:42:01 -070068#include <asm/cacheflush.h>
Alexander van Heukelum2fde61f2008-03-04 19:57:42 +010069#include <asm/processor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070070
71/* This value is set up by the early boot code to point to the value
72 immediately after the boot time page tables. It contains a *physical*
73 address, and must not be in the .bss segment! */
Yinghai Luf0d43102008-05-29 12:56:36 -070074unsigned long init_pg_tables_start __initdata = ~0UL;
Linus Torvalds1da177e2005-04-16 15:20:36 -070075unsigned long init_pg_tables_end __initdata = ~0UL;
76
Linus Torvalds1da177e2005-04-16 15:20:36 -070077/*
78 * Machine setup..
79 */
Bernhard Wallec9cce832008-01-30 13:30:32 +010080static struct resource data_resource = {
81 .name = "Kernel data",
82 .start = 0,
83 .end = 0,
84 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
85};
86
87static struct resource code_resource = {
88 .name = "Kernel code",
89 .start = 0,
90 .end = 0,
91 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
92};
93
94static struct resource bss_resource = {
95 .name = "Kernel bss",
96 .start = 0,
97 .end = 0,
98 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
99};
100
101static struct resource video_ram_resource = {
102 .name = "Video RAM area",
103 .start = 0xa0000,
104 .end = 0xbffff,
105 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
106};
107
108static struct resource standard_io_resources[] = { {
109 .name = "dma1",
110 .start = 0x0000,
111 .end = 0x001f,
112 .flags = IORESOURCE_BUSY | IORESOURCE_IO
113}, {
114 .name = "pic1",
115 .start = 0x0020,
116 .end = 0x0021,
117 .flags = IORESOURCE_BUSY | IORESOURCE_IO
118}, {
119 .name = "timer0",
120 .start = 0x0040,
121 .end = 0x0043,
122 .flags = IORESOURCE_BUSY | IORESOURCE_IO
123}, {
124 .name = "timer1",
125 .start = 0x0050,
126 .end = 0x0053,
127 .flags = IORESOURCE_BUSY | IORESOURCE_IO
128}, {
129 .name = "keyboard",
130 .start = 0x0060,
Helge Wagner9096bd72008-04-29 14:20:40 +0200131 .end = 0x0060,
132 .flags = IORESOURCE_BUSY | IORESOURCE_IO
133}, {
134 .name = "keyboard",
135 .start = 0x0064,
136 .end = 0x0064,
Bernhard Wallec9cce832008-01-30 13:30:32 +0100137 .flags = IORESOURCE_BUSY | IORESOURCE_IO
138}, {
139 .name = "dma page reg",
140 .start = 0x0080,
141 .end = 0x008f,
142 .flags = IORESOURCE_BUSY | IORESOURCE_IO
143}, {
144 .name = "pic2",
145 .start = 0x00a0,
146 .end = 0x00a1,
147 .flags = IORESOURCE_BUSY | IORESOURCE_IO
148}, {
149 .name = "dma2",
150 .start = 0x00c0,
151 .end = 0x00df,
152 .flags = IORESOURCE_BUSY | IORESOURCE_IO
153}, {
154 .name = "fpu",
155 .start = 0x00f0,
156 .end = 0x00ff,
157 .flags = IORESOURCE_BUSY | IORESOURCE_IO
158} };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159
160/* cpu data as detected by the assembly code in head.S */
Vivek Goyal4a5d1072007-01-11 01:52:44 +0100161struct cpuinfo_x86 new_cpu_data __cpuinitdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162/* common cpu data for all cpus */
Christoph Lameterc3d8c142005-09-06 15:16:33 -0700163struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
Alexey Dobriyan129f6942005-06-23 00:08:33 -0700164EXPORT_SYMBOL(boot_cpu_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165
Alexey Starikovskiy0c254e32008-03-27 23:55:04 +0300166unsigned int def_to_bigsmp;
167
Ian Campbell551889a2008-02-09 23:24:09 +0100168#ifndef CONFIG_X86_PAE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169unsigned long mmu_cr4_features;
Ian Campbell551889a2008-02-09 23:24:09 +0100170#else
171unsigned long mmu_cr4_features = X86_CR4_PAE;
172#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174/* for MCA, but anyone else can use it if they want */
175unsigned int machine_id;
176unsigned int machine_submodel_id;
177unsigned int BIOS_revision;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179/* Boot loader ID as an integer, for the benefit of proc_dointvec */
180int bootloader_type;
181
182/* user-defined highmem size */
183static unsigned int highmem_pages = -1;
184
185/*
186 * Setup options
187 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188struct screen_info screen_info;
Alexey Dobriyan129f6942005-06-23 00:08:33 -0700189EXPORT_SYMBOL(screen_info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190struct apm_info apm_info;
Alexey Dobriyan129f6942005-06-23 00:08:33 -0700191EXPORT_SYMBOL(apm_info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192struct edid_info edid_info;
Antonino A. Daplas5e518d72005-09-09 13:04:34 -0700193EXPORT_SYMBOL_GPL(edid_info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194struct ist_info ist_info;
Alexey Dobriyan129f6942005-06-23 00:08:33 -0700195#if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
196 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
197EXPORT_SYMBOL(ist_info);
198#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199
200extern void early_cpu_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201extern int root_mountflags;
202
Pavel Macheke44b7b72008-04-10 23:28:10 +0200203unsigned long saved_video_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100205#define RAMDISK_IMAGE_START_MASK 0x07FF
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206#define RAMDISK_PROMPT_FLAG 0x8000
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100207#define RAMDISK_LOAD_FLAG 0x4000
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208
Alon Bar-Lev4e498b62007-02-12 00:54:11 -0800209static char __initdata command_line[COMMAND_LINE_SIZE];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210
Huang, Ying6d7d7432008-01-30 13:32:51 +0100211#ifndef CONFIG_DEBUG_BOOT_PARAMS
H. Peter Anvin48c7ae62007-07-11 12:18:35 -0700212struct boot_params __initdata boot_params;
Huang, Ying6d7d7432008-01-30 13:32:51 +0100213#else
214struct boot_params boot_params;
215#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
218struct edd edd;
219#ifdef CONFIG_EDD_MODULE
220EXPORT_SYMBOL(edd);
221#endif
222/**
223 * copy_edd() - Copy the BIOS EDD information
224 * from boot_params into a safe place.
225 *
226 */
227static inline void copy_edd(void)
228{
H. Peter Anvin30c82642007-10-15 17:13:22 -0700229 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
230 sizeof(edd.mbr_signature));
231 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
232 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
233 edd.edd_info_nr = boot_params.eddbuf_entries;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234}
235#else
236static inline void copy_edd(void)
237{
238}
239#endif
240
Rusty Russell1a3f2392006-09-26 10:52:32 +0200241#ifdef CONFIG_PROC_VMCORE
242/* elfcorehdr= specifies the location of elf core header
243 * stored by the crashed kernel.
244 */
245static int __init parse_elfcorehdr(char *arg)
246{
247 if (!arg)
248 return -EINVAL;
249
250 elfcorehdr_addr = memparse(arg, &arg);
251 return 0;
252}
253early_param("elfcorehdr", parse_elfcorehdr);
254#endif /* CONFIG_PROC_VMCORE */
255
256/*
257 * highmem=size forces highmem to be exactly 'size' bytes.
258 * This works even on boxes that have no highmem otherwise.
259 * This also works to reduce highmem size on bigger boxes.
260 */
261static int __init parse_highmem(char *arg)
262{
263 if (!arg)
264 return -EINVAL;
265
266 highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT;
267 return 0;
268}
269early_param("highmem", parse_highmem);
270
271/*
272 * vmalloc=size forces the vmalloc area to be exactly 'size'
273 * bytes. This can be used to increase (or decrease) the
274 * vmalloc area - the default is 128m.
275 */
276static int __init parse_vmalloc(char *arg)
277{
278 if (!arg)
279 return -EINVAL;
280
281 __VMALLOC_RESERVE = memparse(arg, &arg);
282 return 0;
283}
284early_param("vmalloc", parse_vmalloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285
286/*
Zachary Amsden461a9af2006-09-25 23:32:25 -0700287 * reservetop=size reserves a hole at the top of the kernel address space which
288 * a hypervisor can load into later. Needed for dynamically loaded hypervisors,
289 * so relocating the fixmap can be done before paging initialization.
290 */
291static int __init parse_reservetop(char *arg)
292{
293 unsigned long address;
294
295 if (!arg)
296 return -EINVAL;
297
298 address = memparse(arg, &arg);
299 reserve_top_address(address);
300 return 0;
301}
302early_param("reservetop", parse_reservetop);
303
304/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 * Determine low and high memory ranges:
306 */
307unsigned long __init find_max_low_pfn(void)
308{
309 unsigned long max_low_pfn;
310
311 max_low_pfn = max_pfn;
312 if (max_low_pfn > MAXMEM_PFN) {
313 if (highmem_pages == -1)
314 highmem_pages = max_pfn - MAXMEM_PFN;
315 if (highmem_pages + MAXMEM_PFN < max_pfn)
316 max_pfn = MAXMEM_PFN + highmem_pages;
317 if (highmem_pages + MAXMEM_PFN > max_pfn) {
318 printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn - MAXMEM_PFN), pages_to_mb(highmem_pages));
319 highmem_pages = 0;
320 }
321 max_low_pfn = MAXMEM_PFN;
322#ifndef CONFIG_HIGHMEM
323 /* Maximum memory usable is what is directly addressable */
324 printk(KERN_WARNING "Warning only %ldMB will be used.\n",
325 MAXMEM>>20);
326 if (max_pfn > MAX_NONPAE_PFN)
William Lee Irwin IIIc673f1a2007-07-21 17:11:13 +0200327 printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328 else
329 printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n");
330 max_pfn = MAXMEM_PFN;
331#else /* !CONFIG_HIGHMEM */
William Lee Irwin IIIc673f1a2007-07-21 17:11:13 +0200332#ifndef CONFIG_HIGHMEM64G
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 if (max_pfn > MAX_NONPAE_PFN) {
334 max_pfn = MAX_NONPAE_PFN;
335 printk(KERN_WARNING "Warning only 4GB will be used.\n");
William Lee Irwin IIIc673f1a2007-07-21 17:11:13 +0200336 printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 }
William Lee Irwin IIIc673f1a2007-07-21 17:11:13 +0200338#endif /* !CONFIG_HIGHMEM64G */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339#endif /* !CONFIG_HIGHMEM */
340 } else {
341 if (highmem_pages == -1)
342 highmem_pages = 0;
343#ifdef CONFIG_HIGHMEM
344 if (highmem_pages >= max_pfn) {
345 printk(KERN_ERR "highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages), pages_to_mb(max_pfn));
346 highmem_pages = 0;
347 }
348 if (highmem_pages) {
349 if (max_low_pfn-highmem_pages < 64*1024*1024/PAGE_SIZE){
350 printk(KERN_ERR "highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages));
351 highmem_pages = 0;
352 }
353 max_low_pfn -= highmem_pages;
354 }
355#else
356 if (highmem_pages)
357 printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n");
358#endif
359 }
360 return max_low_pfn;
361}
362
Andy Whitcroft05b79bd2005-06-23 00:07:57 -0700363#ifndef CONFIG_NEED_MULTIPLE_NODES
Ingo Molnara4928cf2008-04-23 13:20:56 +0200364static void __init setup_bootmem_allocator(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365static unsigned long __init setup_memory(void)
366{
367 /*
368 * partially used pages are not usable - thus
369 * we are rounding upwards:
370 */
371 min_low_pfn = PFN_UP(init_pg_tables_end);
372
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 max_low_pfn = find_max_low_pfn();
374
375#ifdef CONFIG_HIGHMEM
376 highstart_pfn = highend_pfn = max_pfn;
377 if (max_pfn > max_low_pfn) {
378 highstart_pfn = max_low_pfn;
379 }
Yinghai Luf19dc2f2008-06-03 10:24:49 -0700380 memory_present(0, 0, highend_pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
382 pages_to_mb(highend_pfn - highstart_pfn));
Jan Beulichba9c2312006-09-26 10:52:31 +0200383 num_physpages = highend_pfn;
384 high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
385#else
Yinghai Luf19dc2f2008-06-03 10:24:49 -0700386 memory_present(0, 0, max_low_pfn);
Jan Beulichba9c2312006-09-26 10:52:31 +0200387 num_physpages = max_low_pfn;
388 high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
389#endif
390#ifdef CONFIG_FLATMEM
391 max_mapnr = num_physpages;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392#endif
393 printk(KERN_NOTICE "%ldMB LOWMEM available.\n",
394 pages_to_mb(max_low_pfn));
395
396 setup_bootmem_allocator();
397
398 return max_low_pfn;
399}
400
Ingo Molnara4928cf2008-04-23 13:20:56 +0200401static void __init zone_sizes_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402{
Mel Gorman6391af12006-10-11 01:20:39 -0700403 unsigned long max_zone_pfns[MAX_NR_ZONES];
404 memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
405 max_zone_pfns[ZONE_DMA] =
406 virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
407 max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408#ifdef CONFIG_HIGHMEM
Mel Gorman6391af12006-10-11 01:20:39 -0700409 max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
Mel Gorman4cfee882006-09-27 01:49:51 -0700410 add_active_range(0, 0, highend_pfn);
411#else
Mel Gorman4cfee882006-09-27 01:49:51 -0700412 add_active_range(0, 0, max_low_pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413#endif
Mel Gorman4cfee882006-09-27 01:49:51 -0700414
415 free_area_init_nodes(max_zone_pfns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416}
417#else
Andy Whitcroft05b79bd2005-06-23 00:07:57 -0700418extern unsigned long __init setup_memory(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419extern void zone_sizes_init(void);
Andy Whitcroft05b79bd2005-06-23 00:07:57 -0700420#endif /* !CONFIG_NEED_MULTIPLE_NODES */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421
Bernhard Walled62cc472007-10-18 23:40:59 -0700422static inline unsigned long long get_total_mem(void)
423{
424 unsigned long long total;
425
426 total = max_low_pfn - min_low_pfn;
427#ifdef CONFIG_HIGHMEM
428 total += highend_pfn - highstart_pfn;
429#endif
430
431 return total << PAGE_SHIFT;
432}
433
434#ifdef CONFIG_KEXEC
435static void __init reserve_crashkernel(void)
436{
437 unsigned long long total_mem;
438 unsigned long long crash_size, crash_base;
439 int ret;
440
441 total_mem = get_total_mem();
442
443 ret = parse_crashkernel(boot_command_line, total_mem,
444 &crash_size, &crash_base);
445 if (ret == 0 && crash_size > 0) {
446 if (crash_base > 0) {
447 printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
448 "for crashkernel (System RAM: %ldMB)\n",
449 (unsigned long)(crash_size >> 20),
450 (unsigned long)(crash_base >> 20),
451 (unsigned long)(total_mem >> 20));
452 crashk_res.start = crash_base;
453 crashk_res.end = crash_base + crash_size - 1;
Bernhard Walle72a7fe32008-02-07 00:15:17 -0800454 reserve_bootmem(crash_base, crash_size,
455 BOOTMEM_DEFAULT);
Bernhard Walled62cc472007-10-18 23:40:59 -0700456 } else
457 printk(KERN_INFO "crashkernel reservation failed - "
458 "you have to specify a base address\n");
459 }
460}
461#else
462static inline void __init reserve_crashkernel(void)
463{}
464#endif
465
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100466#ifdef CONFIG_BLK_DEV_INITRD
467
468static bool do_relocate_initrd = false;
469
470static void __init reserve_initrd(void)
471{
Yinghai Luba5b14c2008-05-21 18:40:18 -0700472 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
473 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
474 u64 ramdisk_end = ramdisk_image + ramdisk_size;
475 u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
476 u64 ramdisk_here;
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100477
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100478 if (!boot_params.hdr.type_of_loader ||
479 !ramdisk_image || !ramdisk_size)
480 return; /* No initrd provided by bootloader */
481
Yinghai Lua4c81cf2008-05-18 01:18:57 -0700482 initrd_start = 0;
483
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100484 if (ramdisk_size >= end_of_lowmem/2) {
Yinghai Lu3945e2c2008-05-25 10:00:09 -0700485 free_early(ramdisk_image, ramdisk_end);
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100486 printk(KERN_ERR "initrd too large to handle, "
487 "disabling initrd\n");
488 return;
489 }
Yinghai Lua4c81cf2008-05-18 01:18:57 -0700490
Yinghai Luf0d43102008-05-29 12:56:36 -0700491 printk(KERN_INFO "old RAMDISK: %08llx - %08llx\n", ramdisk_image,
492 ramdisk_end);
493
494
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100495 if (ramdisk_end <= end_of_lowmem) {
496 /* All in lowmem, easy case */
Yinghai Lua4c81cf2008-05-18 01:18:57 -0700497 /*
498 * don't need to reserve again, already reserved early
Yinghai Luba5b14c2008-05-21 18:40:18 -0700499 * in i386_start_kernel
Yinghai Lua4c81cf2008-05-18 01:18:57 -0700500 */
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100501 initrd_start = ramdisk_image + PAGE_OFFSET;
502 initrd_end = initrd_start+ramdisk_size;
503 return;
504 }
505
506 /* We need to move the initrd down into lowmem */
Yinghai Luba5b14c2008-05-21 18:40:18 -0700507 ramdisk_here = find_e820_area(min_low_pfn<<PAGE_SHIFT,
508 end_of_lowmem, ramdisk_size,
509 PAGE_SIZE);
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100510
Yinghai Lu3945e2c2008-05-25 10:00:09 -0700511 if (ramdisk_here == -1ULL)
512 panic("Cannot find place for new RAMDISK of size %lld\n",
513 ramdisk_size);
514
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100515 /* Note: this includes all the lowmem currently occupied by
516 the initrd, we rely on that fact to keep the data intact. */
Yinghai Lu3945e2c2008-05-25 10:00:09 -0700517 reserve_early(ramdisk_here, ramdisk_here + ramdisk_size,
Yinghai Luba5b14c2008-05-21 18:40:18 -0700518 "NEW RAMDISK");
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100519 initrd_start = ramdisk_here + PAGE_OFFSET;
520 initrd_end = initrd_start + ramdisk_size;
Yinghai Luf0d43102008-05-29 12:56:36 -0700521 printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
522 ramdisk_here, ramdisk_here + ramdisk_size);
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100523
524 do_relocate_initrd = true;
525}
526
527#define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT)
528
529static void __init relocate_initrd(void)
530{
Yinghai Luba5b14c2008-05-21 18:40:18 -0700531 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
532 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
533 u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
534 u64 ramdisk_here;
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100535 unsigned long slop, clen, mapaddr;
536 char *p, *q;
537
538 if (!do_relocate_initrd)
539 return;
540
541 ramdisk_here = initrd_start - PAGE_OFFSET;
542
543 q = (char *)initrd_start;
544
545 /* Copy any lowmem portion of the initrd */
546 if (ramdisk_image < end_of_lowmem) {
547 clen = end_of_lowmem - ramdisk_image;
548 p = (char *)__va(ramdisk_image);
549 memcpy(q, p, clen);
550 q += clen;
Yinghai Lua4c81cf2008-05-18 01:18:57 -0700551 /* need to free these low pages...*/
Yinghai Luba5b14c2008-05-21 18:40:18 -0700552 printk(KERN_INFO "Freeing old partial RAMDISK %08llx-%08llx\n",
553 ramdisk_image, ramdisk_image + clen - 1);
Yinghai Lua4c81cf2008-05-18 01:18:57 -0700554 free_bootmem(ramdisk_image, clen);
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100555 ramdisk_image += clen;
556 ramdisk_size -= clen;
557 }
558
559 /* Copy the highmem portion of the initrd */
560 while (ramdisk_size) {
561 slop = ramdisk_image & ~PAGE_MASK;
562 clen = ramdisk_size;
563 if (clen > MAX_MAP_CHUNK-slop)
564 clen = MAX_MAP_CHUNK-slop;
565 mapaddr = ramdisk_image & PAGE_MASK;
Huang, Yingbeacfaa2008-01-30 13:33:44 +0100566 p = early_ioremap(mapaddr, clen+slop);
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100567 memcpy(q, p+slop, clen);
Huang, Yingbeacfaa2008-01-30 13:33:44 +0100568 early_iounmap(p, clen+slop);
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100569 q += clen;
570 ramdisk_image += clen;
571 ramdisk_size -= clen;
572 }
Yinghai Lua4c81cf2008-05-18 01:18:57 -0700573 /* high pages is not converted by early_res_to_bootmem */
Yinghai Luba5b14c2008-05-21 18:40:18 -0700574 ramdisk_image = boot_params.hdr.ramdisk_image;
575 ramdisk_size = boot_params.hdr.ramdisk_size;
576 printk(KERN_INFO "Copied RAMDISK from %016llx - %016llx to %08llx - %08llx\n",
577 ramdisk_image, ramdisk_image + ramdisk_size - 1,
578 ramdisk_here, ramdisk_here + ramdisk_size - 1);
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100579}
580
581#endif /* CONFIG_BLK_DEV_INITRD */
582
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583void __init setup_bootmem_allocator(void)
584{
Yinghai Lua4c81cf2008-05-18 01:18:57 -0700585 unsigned long bootmap_size, bootmap;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 /*
587 * Initialize the boot-time allocator (with low memory only):
588 */
Yinghai Lua4c81cf2008-05-18 01:18:57 -0700589 bootmap_size = bootmem_bootmap_pages(max_low_pfn)<<PAGE_SHIFT;
590 bootmap = find_e820_area(min_low_pfn<<PAGE_SHIFT,
Yinghai Lu6af61a72008-06-01 23:53:50 -0700591 max_pfn_mapped<<PAGE_SHIFT, bootmap_size,
Yinghai Lua4c81cf2008-05-18 01:18:57 -0700592 PAGE_SIZE);
593 if (bootmap == -1L)
594 panic("Cannot find bootmem map of size %ld\n", bootmap_size);
Yinghai Lu3945e2c2008-05-25 10:00:09 -0700595 reserve_early(bootmap, bootmap + bootmap_size, "BOOTMAP");
Yinghai Luba5b14c2008-05-21 18:40:18 -0700596#ifdef CONFIG_BLK_DEV_INITRD
597 reserve_initrd();
598#endif
Yinghai Lua4c81cf2008-05-18 01:18:57 -0700599 bootmap_size = init_bootmem(bootmap >> PAGE_SHIFT, max_low_pfn);
Yinghai Lu6af61a72008-06-01 23:53:50 -0700600 printk(KERN_INFO " mapped low ram: 0 - %08lx\n",
601 max_pfn_mapped<<PAGE_SHIFT);
Yinghai Luba5b14c2008-05-21 18:40:18 -0700602 printk(KERN_INFO " low ram: %08lx - %08lx\n",
603 min_low_pfn<<PAGE_SHIFT, max_low_pfn<<PAGE_SHIFT);
Yinghai Lu3945e2c2008-05-25 10:00:09 -0700604 printk(KERN_INFO " bootmap %08lx - %08lx\n",
605 bootmap, bootmap + bootmap_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606 register_bootmem_low_pages(max_low_pfn);
Yinghai Lua4c81cf2008-05-18 01:18:57 -0700607 early_res_to_bootmem(0, max_low_pfn<<PAGE_SHIFT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608
Len Brown673d5b42007-07-28 03:33:16 -0400609#ifdef CONFIG_ACPI_SLEEP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610 /*
611 * Reserve low memory region for sleep support.
612 */
613 acpi_reserve_bootmem();
614#endif
615#ifdef CONFIG_X86_FIND_SMP_CONFIG
616 /*
617 * Find and reserve possible boot-time SMP configuration:
618 */
619 find_smp_config();
620#endif
Bernhard Walled62cc472007-10-18 23:40:59 -0700621 reserve_crashkernel();
Konrad Rzeszutek138fe4e2008-04-09 19:50:41 -0700622
623 reserve_ibft_region();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624}
625
626/*
627 * The node 0 pgdat is initialized before all of these because
628 * it's needed for bootmem. node>0 pgdats have their virtual
629 * space allocated before the pagetables are in place to access
630 * them, so they can't be cleared then.
631 *
632 * This should all compile down to nothing when NUMA is off.
633 */
Adrian Bunk10079ae2007-07-21 17:10:27 +0200634static void __init remapped_pgdat_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635{
636 int nid;
637
638 for_each_online_node(nid) {
639 if (nid != 0)
640 memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
641 }
642}
643
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644#ifdef CONFIG_MCA
645static void set_mca_bus(int x)
646{
647 MCA_bus = x;
648}
649#else
650static void set_mca_bus(int x) { }
651#endif
652
Glauber de Oliveira Costaccf82082008-03-19 14:25:20 -0300653#ifdef CONFIG_NUMA
654/*
655 * In the golden day, when everything among i386 and x86_64 will be
656 * integrated, this will not live here
657 */
658void *x86_cpu_to_node_map_early_ptr;
659int x86_cpu_to_node_map_init[NR_CPUS] = {
660 [0 ... NR_CPUS-1] = NUMA_NO_NODE
661};
662DEFINE_PER_CPU(int, x86_cpu_to_node_map) = NUMA_NO_NODE;
663#endif
664
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665/*
666 * Determine if we were loaded by an EFI loader. If so, then we have also been
667 * passed the efi memmap, systab, etc., so we should use these data structures
668 * for initialization. Note, the efi init code path is determined by the
669 * global efi_enabled. This allows the same kernel image to be used on existing
670 * systems (with a traditional BIOS) as well as on EFI systems.
671 */
672void __init setup_arch(char **cmdline_p)
673{
674 unsigned long max_low_pfn;
675
676 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
677 pre_setup_arch_hook();
678 early_cpu_init();
Huang, Yingbeacfaa2008-01-30 13:33:44 +0100679 early_ioremap_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681#ifdef CONFIG_EFI
Huang, Yinge4297952008-01-30 13:31:19 +0100682 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
683 "EL32", 4))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 efi_enabled = 1;
685#endif
686
H. Peter Anvin30c82642007-10-15 17:13:22 -0700687 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
688 screen_info = boot_params.screen_info;
689 edid_info = boot_params.edid_info;
690 apm_info.bios = boot_params.apm_bios_info;
691 ist_info = boot_params.ist_info;
Pavel Macheke44b7b72008-04-10 23:28:10 +0200692 saved_video_mode = boot_params.hdr.vid_mode;
H. Peter Anvin30c82642007-10-15 17:13:22 -0700693 if( boot_params.sys_desc_table.length != 0 ) {
694 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
695 machine_id = boot_params.sys_desc_table.table[0];
696 machine_submodel_id = boot_params.sys_desc_table.table[1];
697 BIOS_revision = boot_params.sys_desc_table.table[2];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 }
H. Peter Anvin30c82642007-10-15 17:13:22 -0700699 bootloader_type = boot_params.hdr.type_of_loader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700
701#ifdef CONFIG_BLK_DEV_RAM
H. Peter Anvin30c82642007-10-15 17:13:22 -0700702 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
703 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
704 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705#endif
706 ARCH_SETUP
Huang, Ying2215e692008-01-30 13:31:19 +0100707
Alexander van Heukelum0dbfafa2008-04-23 15:09:05 +0200708 setup_memory_map();
Huang, Ying2215e692008-01-30 13:31:19 +0100709
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 copy_edd();
711
H. Peter Anvin30c82642007-10-15 17:13:22 -0700712 if (!boot_params.hdr.root_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713 root_mountflags &= ~MS_RDONLY;
714 init_mm.start_code = (unsigned long) _text;
715 init_mm.end_code = (unsigned long) _etext;
716 init_mm.end_data = (unsigned long) _edata;
717 init_mm.brk = init_pg_tables_end + PAGE_OFFSET;
718
719 code_resource.start = virt_to_phys(_text);
720 code_resource.end = virt_to_phys(_etext)-1;
721 data_resource.start = virt_to_phys(_etext);
722 data_resource.end = virt_to_phys(_edata)-1;
Bernhard Walle00bf4092007-10-21 16:42:01 -0700723 bss_resource.start = virt_to_phys(&__bss_start);
724 bss_resource.end = virt_to_phys(&__bss_stop)-1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725
Rusty Russell1a3f2392006-09-26 10:52:32 +0200726 parse_early_param();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727
Alexander van Heukelum0dbfafa2008-04-23 15:09:05 +0200728 finish_e820_parsing();
Rusty Russell1a3f2392006-09-26 10:52:32 +0200729
Alon Bar-Lev4e498b62007-02-12 00:54:11 -0800730 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
Rusty Russell1a3f2392006-09-26 10:52:32 +0200731 *cmdline_p = command_line;
Stas Sergeev99b7de32006-03-23 02:59:41 -0800732
Huang, Ying8b2cb7a2008-01-30 13:32:11 +0100733 if (efi_enabled)
734 efi_init();
735
Yinghai Lu093af8d2008-01-30 13:33:32 +0100736 /* update e820 for memory not covered by WB MTRRs */
Yinghai Lu05961522008-05-31 22:52:47 -0700737 find_max_pfn();
Yinghai Lu093af8d2008-01-30 13:33:32 +0100738 mtrr_bp_init();
739 if (mtrr_trim_uncached_memory(max_pfn))
Yinghai Lu05961522008-05-31 22:52:47 -0700740 find_max_pfn();
Yinghai Lu76c324182008-03-23 00:16:49 -0700741
742 max_low_pfn = setup_memory();
Yinghai Lu093af8d2008-01-30 13:33:32 +0100743
Glauber de Oliveira Costa790c73f2008-02-15 17:52:48 -0200744#ifdef CONFIG_KVM_CLOCK
745 kvmclock_init();
746#endif
747
Zachary Amsden7ce0bcf2007-02-13 13:26:21 +0100748#ifdef CONFIG_VMI
749 /*
750 * Must be after max_low_pfn is determined, and before kernel
751 * pagetables are setup.
752 */
753 vmi_init();
754#endif
Marcelo Tosatti0cf1bfd2008-02-22 12:21:36 -0500755 kvm_guest_init();
Zachary Amsden7ce0bcf2007-02-13 13:26:21 +0100756
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 /*
758 * NOTE: before this point _nobody_ is allowed to allocate
759 * any memory using the bootmem allocator. Although the
Simon Arlott27b46d72007-10-20 01:13:56 +0200760 * allocator is now initialised only the first 8Mb of the kernel
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 * virtual address space has been mapped. All allocations before
762 * paging_init() has completed must use the alloc_bootmem_low_pages()
763 * variant (which allocates DMA'able memory) and care must be taken
764 * not to exceed the 8Mb limit.
765 */
766
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 paging_init();
Bernhard Kaindlf212ec42008-01-30 13:34:11 +0100768
769 /*
770 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
771 */
772
773#ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
774 if (init_ohci1394_dma_early)
775 init_ohci1394_dma_on_all_controllers();
776#endif
777
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 remapped_pgdat_init();
Andy Whitcroft05b79bd2005-06-23 00:07:57 -0700779 sparse_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 zone_sizes_init();
781
782 /*
783 * NOTE: at this point the bootmem allocator is fully available.
784 */
785
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100786#ifdef CONFIG_BLK_DEV_INITRD
787 relocate_initrd();
788#endif
789
Jeremy Fitzhardinge6996d3b2007-07-17 18:37:03 -0700790 paravirt_post_allocator_init();
791
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 dmi_scan_machine();
793
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100794 io_delay_init();
Rene Hermanb02aae92008-01-30 13:30:05 +0100795
Glauber de Oliveira Costaccf82082008-03-19 14:25:20 -0300796#ifdef CONFIG_X86_SMP
797 /*
798 * setup to use the early static init tables during kernel startup
799 * X86_SMP will exclude sub-arches that don't deal well with it.
800 */
801 x86_cpu_to_apicid_early_ptr = (void *)x86_cpu_to_apicid_init;
802 x86_bios_cpu_apicid_early_ptr = (void *)x86_bios_cpu_apicid_init;
803#ifdef CONFIG_NUMA
804 x86_cpu_to_node_map_early_ptr = (void *)x86_cpu_to_node_map_init;
805#endif
806#endif
807
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808#ifdef CONFIG_X86_GENERICARCH
Rusty Russell1a3f2392006-09-26 10:52:32 +0200809 generic_apic_probe();
H. Peter Anvincf8fa922008-01-30 13:32:51 +0100810#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811
Len Brown888ba6c2005-08-24 12:07:20 -0400812#ifdef CONFIG_ACPI
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 /*
814 * Parse the ACPI tables for possible boot-time SMP configuration.
815 */
816 acpi_boot_table_init();
Andy Curridd44647b2006-06-08 00:43:38 -0700817#endif
818
Andi Kleen54ef3402007-10-19 20:35:03 +0200819 early_quirks();
Andy Curridd44647b2006-06-08 00:43:38 -0700820
821#ifdef CONFIG_ACPI
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 acpi_boot_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823
Venkatesh Pallipadi911a62d2005-09-03 15:56:31 -0700824#if defined(CONFIG_SMP) && defined(CONFIG_X86_PC)
825 if (def_to_bigsmp)
826 printk(KERN_WARNING "More than 8 CPUs detected and "
827 "CONFIG_X86_PC cannot handle it.\nUse "
828 "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
829#endif
830#endif
Alexey Starikovskiy136ef672008-05-20 00:29:59 +0400831#if defined(CONFIG_X86_MPPARSE) || defined(CONFIG_X86_VISWS)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832 if (smp_found_config)
833 get_smp_config();
834#endif
835
Yinghai Lub79cd8f2008-05-11 00:30:15 -0700836 e820_setup_gap();
Yinghai Lubf62f392008-05-20 20:10:58 -0700837 e820_mark_nosave_regions(max_low_pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838
839#ifdef CONFIG_VT
840#if defined(CONFIG_VGA_CONSOLE)
841 if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
842 conswitchp = &vga_con;
843#elif defined(CONFIG_DUMMY_CONSOLE)
844 conswitchp = &dummy_con;
845#endif
846#endif
847}
Bernhard Wallec9cce832008-01-30 13:30:32 +0100848
849/*
850 * Request address space for all standard resources
851 *
852 * This is called just before pcibios_init(), which is also a
853 * subsys_initcall, but is linked in later (in arch/i386/pci/common.c).
854 */
855static int __init request_standard_resources(void)
856{
857 int i;
858
859 printk(KERN_INFO "Setting up standard PCI resources\n");
Huang, Ying2215e692008-01-30 13:31:19 +0100860 init_iomem_resources(&code_resource, &data_resource, &bss_resource);
Bernhard Wallec9cce832008-01-30 13:30:32 +0100861
Bernhard Wallec9cce832008-01-30 13:30:32 +0100862 request_resource(&iomem_resource, &video_ram_resource);
863
864 /* request I/O space for devices used on all i[345]86 PCs */
865 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
866 request_resource(&ioport_resource, &standard_io_resources[i]);
867 return 0;
868}
869
870subsys_initcall(request_standard_resources);