2 * Copyright (C) 1995 Linus Torvalds
4 * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
6 * Memory region support
7 * David Parsons <orc@pell.chi.il.us>, July-August 1999
9 * Added E820 sanitization routine (removes overlapping memory regions);
10 * Brian Moyle <bmoyle@mvista.com>, February 2001
12 * Moved CPU detection code to cpu/${cpu}.c
13 * Patrick Mochel <mochel@osdl.org>, March 2002
15 * Provisions for empty E820 memory regions (reported by certain BIOSes).
16 * Alex Achenbach <xela@slit.de>, December 2002.
21 * This file handles the architecture-dependent parts of initialization
24 #include <linux/sched.h>
26 #include <linux/mmzone.h>
27 #include <linux/screen_info.h>
28 #include <linux/ioport.h>
29 #include <linux/acpi.h>
30 #include <linux/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>
42 #include <linux/iscsi_ibft.h>
43 #include <linux/nodemask.h>
44 #include <linux/kexec.h>
45 #include <linux/dmi.h>
46 #include <linux/pfn.h>
47 #include <linux/pci.h>
48 #include <asm/pci-direct.h>
49 #include <linux/init_ohci1394_dma.h>
50 #include <linux/kvm_para.h>
52 #include <linux/errno.h>
53 #include <linux/kernel.h>
54 #include <linux/stddef.h>
55 #include <linux/unistd.h>
56 #include <linux/ptrace.h>
57 #include <linux/slab.h>
58 #include <linux/user.h>
59 #include <linux/delay.h>
60 #include <linux/highmem.h>
62 #include <linux/kallsyms.h>
63 #include <linux/edd.h>
64 #include <linux/iscsi_ibft.h>
65 #include <linux/kexec.h>
66 #include <linux/cpufreq.h>
67 #include <linux/dma-mapping.h>
68 #include <linux/ctype.h>
69 #include <linux/uaccess.h>
71 #include <linux/percpu.h>
72 #include <linux/crash_dump.h>
74 #include <video/edid.h>
79 #include <asm/mpspec.h>
80 #include <asm/setup.h>
81 #include <asm/arch_hooks.h>
83 #include <asm/sections.h>
85 #include <asm/io_apic.h>
88 #include <setup_arch.h>
89 #include <asm/bios_ebda.h>
90 #include <asm/cacheflush.h>
91 #include <asm/processor.h>
94 #include <asm/system.h>
95 #include <asm/vsyscall.h>
100 #include <asm/mmu_context.h>
101 #include <asm/proto.h>
103 #include <mach_apic.h>
104 #ifdef CONFIG_PARAVIRT
105 #include <asm/paravirt.h>
110 #include <asm/percpu.h>
111 #include <asm/sections.h>
112 #include <asm/topology.h>
113 #include <asm/apicdef.h>
115 #include <asm/numa_64.h>
118 #include <asm/highmem.h>
121 /* This value is set up by the early boot code to point to the value
122 immediately after the boot time page tables. It contains a *physical*
123 address, and must not be in the .bss segment! */
124 unsigned long init_pg_tables_start __initdata = ~0UL;
125 unsigned long init_pg_tables_end __initdata = ~0UL;
130 static struct resource data_resource = {
131 .name = "Kernel data",
134 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
137 static struct resource code_resource = {
138 .name = "Kernel code",
141 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
144 static struct resource bss_resource = {
145 .name = "Kernel bss",
148 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
153 static struct resource video_ram_resource = {
154 .name = "Video RAM area",
157 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
160 /* cpu data as detected by the assembly code in head.S */
161 struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
162 /* common cpu data for all cpus */
163 struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
164 EXPORT_SYMBOL(boot_cpu_data);
165 static void set_mca_bus(int x)
172 unsigned int def_to_bigsmp;
174 /* for MCA, but anyone else can use it if they want */
175 unsigned int machine_id;
176 unsigned int machine_submodel_id;
177 unsigned int BIOS_revision;
179 struct apm_info apm_info;
180 EXPORT_SYMBOL(apm_info);
182 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
183 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
184 struct ist_info ist_info;
185 EXPORT_SYMBOL(ist_info);
187 struct ist_info ist_info;
191 struct cpuinfo_x86 boot_cpu_data __read_mostly;
192 EXPORT_SYMBOL(boot_cpu_data);
196 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
197 unsigned long mmu_cr4_features;
199 unsigned long mmu_cr4_features = X86_CR4_PAE;
202 /* Boot loader ID as an integer, for the benefit of proc_dointvec */
209 char dmi_alloc_data[DMI_MAX_DATA];
214 struct screen_info screen_info;
215 EXPORT_SYMBOL(screen_info);
216 struct edid_info edid_info;
217 EXPORT_SYMBOL_GPL(edid_info);
219 extern int root_mountflags;
221 unsigned long saved_video_mode;
223 #define RAMDISK_IMAGE_START_MASK 0x07FF
224 #define RAMDISK_PROMPT_FLAG 0x8000
225 #define RAMDISK_LOAD_FLAG 0x4000
227 static char __initdata command_line[COMMAND_LINE_SIZE];
229 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
231 #ifdef CONFIG_EDD_MODULE
235 * copy_edd() - Copy the BIOS EDD information
236 * from boot_params into a safe place.
239 static inline void copy_edd(void)
241 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
242 sizeof(edd.mbr_signature));
243 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
244 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
245 edd.edd_info_nr = boot_params.eddbuf_entries;
248 static inline void copy_edd(void)
253 #ifdef CONFIG_BLK_DEV_INITRD
257 #define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT)
258 static void __init relocate_initrd(void)
261 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
262 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
263 u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
265 unsigned long slop, clen, mapaddr;
268 /* We need to move the initrd down into lowmem */
269 ramdisk_here = find_e820_area(0, end_of_lowmem, ramdisk_size,
272 if (ramdisk_here == -1ULL)
273 panic("Cannot find place for new RAMDISK of size %lld\n",
276 /* Note: this includes all the lowmem currently occupied by
277 the initrd, we rely on that fact to keep the data intact. */
278 reserve_early(ramdisk_here, ramdisk_here + ramdisk_size,
280 initrd_start = ramdisk_here + PAGE_OFFSET;
281 initrd_end = initrd_start + ramdisk_size;
282 printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
283 ramdisk_here, ramdisk_here + ramdisk_size);
285 q = (char *)initrd_start;
287 /* Copy any lowmem portion of the initrd */
288 if (ramdisk_image < end_of_lowmem) {
289 clen = end_of_lowmem - ramdisk_image;
290 p = (char *)__va(ramdisk_image);
293 ramdisk_image += clen;
294 ramdisk_size -= clen;
297 /* Copy the highmem portion of the initrd */
298 while (ramdisk_size) {
299 slop = ramdisk_image & ~PAGE_MASK;
301 if (clen > MAX_MAP_CHUNK-slop)
302 clen = MAX_MAP_CHUNK-slop;
303 mapaddr = ramdisk_image & PAGE_MASK;
304 p = early_ioremap(mapaddr, clen+slop);
305 memcpy(q, p+slop, clen);
306 early_iounmap(p, clen+slop);
308 ramdisk_image += clen;
309 ramdisk_size -= clen;
311 /* high pages is not converted by early_res_to_bootmem */
312 ramdisk_image = boot_params.hdr.ramdisk_image;
313 ramdisk_size = boot_params.hdr.ramdisk_size;
314 printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to"
315 " %08llx - %08llx\n",
316 ramdisk_image, ramdisk_image + ramdisk_size - 1,
317 ramdisk_here, ramdisk_here + ramdisk_size - 1);
321 static void __init reserve_initrd(void)
323 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
324 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
325 u64 ramdisk_end = ramdisk_image + ramdisk_size;
326 u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
328 if (!boot_params.hdr.type_of_loader ||
329 !ramdisk_image || !ramdisk_size)
330 return; /* No initrd provided by bootloader */
334 if (ramdisk_size >= (end_of_lowmem>>1)) {
335 free_early(ramdisk_image, ramdisk_end);
336 printk(KERN_ERR "initrd too large to handle, "
337 "disabling initrd\n");
341 printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image,
345 if (ramdisk_end <= end_of_lowmem) {
346 /* All in lowmem, easy case */
348 * don't need to reserve again, already reserved early
349 * in i386_start_kernel
351 initrd_start = ramdisk_image + PAGE_OFFSET;
352 initrd_end = initrd_start + ramdisk_size;
359 printk(KERN_ERR "initrd extends beyond end of memory "
360 "(0x%08llx > 0x%08llx)\ndisabling initrd\n",
361 ramdisk_end, end_of_lowmem);
364 free_early(ramdisk_image, ramdisk_end);
367 static void __init reserve_initrd(void)
370 #endif /* CONFIG_BLK_DEV_INITRD */
373 * Determine if we were loaded by an EFI loader. If so, then we have also been
374 * passed the efi memmap, systab, etc., so we should use these data structures
375 * for initialization. Note, the efi init code path is determined by the
376 * global efi_enabled. This allows the same kernel image to be used on existing
377 * systems (with a traditional BIOS) as well as on EFI systems.
380 * setup_arch - architecture-specific boot-time initializations
382 * Note: On x86_64, fixmaps are ready for use even before this is called.
385 void __init setup_arch(char **cmdline_p)
388 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
389 pre_setup_arch_hook();
391 early_ioremap_init();
392 reserve_setup_data();
394 printk(KERN_INFO "Command line: %s\n", boot_command_line);
397 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
398 screen_info = boot_params.screen_info;
399 edid_info = boot_params.edid_info;
401 apm_info.bios = boot_params.apm_bios_info;
402 ist_info = boot_params.ist_info;
403 if (boot_params.sys_desc_table.length != 0) {
404 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
405 machine_id = boot_params.sys_desc_table.table[0];
406 machine_submodel_id = boot_params.sys_desc_table.table[1];
407 BIOS_revision = boot_params.sys_desc_table.table[2];
410 saved_video_mode = boot_params.hdr.vid_mode;
411 bootloader_type = boot_params.hdr.type_of_loader;
413 #ifdef CONFIG_BLK_DEV_RAM
414 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
415 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
416 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
419 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
436 if (!boot_params.hdr.root_flags)
437 root_mountflags &= ~MS_RDONLY;
438 init_mm.start_code = (unsigned long) _text;
439 init_mm.end_code = (unsigned long) _etext;
440 init_mm.end_data = (unsigned long) _edata;
442 init_mm.brk = init_pg_tables_end + PAGE_OFFSET;
444 init_mm.brk = (unsigned long) &_end;
447 code_resource.start = virt_to_phys(_text);
448 code_resource.end = virt_to_phys(_etext)-1;
449 data_resource.start = virt_to_phys(_etext);
450 data_resource.end = virt_to_phys(_edata)-1;
451 bss_resource.start = virt_to_phys(&__bss_start);
452 bss_resource.end = virt_to_phys(&__bss_stop)-1;
457 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
458 *cmdline_p = command_line;
464 if (acpi_mps_check()) {
465 #ifdef CONFIG_X86_LOCAL_APIC
467 enable_local_apic = -1;
472 clear_cpu_cap(&boot_cpu_data, X86_FEATURE_APIC);
475 finish_e820_parsing();
480 # ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
481 if (init_ohci1394_dma_early)
482 init_ohci1394_dma_on_all_controllers();
486 /* after parse_early_param, so could debug it */
487 insert_resource(&iomem_resource, &code_resource);
488 insert_resource(&iomem_resource, &data_resource);
489 insert_resource(&iomem_resource, &bss_resource);
495 if (ppro_with_ram_bug()) {
496 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
498 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
499 printk(KERN_INFO "fixed physical RAM map:\n");
500 e820_print_map("bad_ppro");
503 early_gart_iommu_check();
506 e820_register_active_regions(0, 0, -1UL);
508 * partially used pages are not usable - thus
509 * we are rounding upwards:
511 max_pfn = e820_end_of_ram();
513 /* preallocate 4k for mptable mpc */
514 early_reserve_e820_mpc_new();
515 /* update e820 for memory not covered by WB MTRRs */
517 if (mtrr_trim_uncached_memory(max_pfn)) {
518 remove_all_active_ranges();
519 e820_register_active_regions(0, 0, -1UL);
520 max_pfn = e820_end_of_ram();
524 /* max_low_pfn get updated here */
525 find_low_pfn_range();
527 num_physpages = max_pfn;
531 /* How many end-of-memory variables you have, grandma! */
532 /* need this before calling reserve_initrd */
533 max_low_pfn = max_pfn;
534 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
537 /* max_pfn_mapped is updated here */
538 max_pfn_mapped = init_memory_mapping(0, (max_low_pfn << PAGE_SHIFT));
551 * Parse the ACPI tables for possible boot-time SMP configuration.
553 acpi_boot_table_init();
556 /* Remove active ranges so rediscovery with NUMA-awareness happens */
557 remove_all_active_ranges();
560 #ifdef CONFIG_ACPI_NUMA
562 * Parse SRAT to discover nodes.
567 initmem_init(0, max_pfn);
570 dma32_reserve_bootmem();
573 #ifdef CONFIG_ACPI_SLEEP
575 * Reserve low memory region for sleep support.
577 acpi_reserve_bootmem();
579 #if defined(CONFIG_X86_FIND_SMP_CONFIG) && defined(CONFIG_X86_32) || \
580 defined(CONFIG_X86_MPPARSE) && defined(CONFIG_X86_64)
582 * Find and reserve possible boot-time SMP configuration:
586 reserve_crashkernel();
588 reserve_ibft_region();
590 #ifdef CONFIG_KVM_CLOCK
594 #if defined(CONFIG_VMI) && defined(CONFIG_X86_32)
596 * Must be after max_low_pfn is determined, and before kernel
597 * pagetables are setup.
609 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
612 #if defined(CONFIG_PROVIDE_OHCI1394_DMA_INIT) && defined(CONFIG_X86_32)
613 if (init_ohci1394_dma_early)
614 init_ohci1394_dma_on_all_controllers();
617 #ifdef CONFIG_X86_GENERICARCH
618 generic_apic_probe();
624 * Read APIC and some other early information from ACPI tables.
632 #if defined(CONFIG_X86_MPPARSE) || defined(CONFIG_X86_VISWS)
634 * get boot-time SMP configuration:
636 if (smp_found_config)
641 init_apic_mappings();
642 ioapic_init_mappings();
644 # if defined(CONFIG_SMP) && defined(CONFIG_X86_PC)
646 printk(KERN_WARNING "More than 8 CPUs detected and "
647 "CONFIG_X86_PC cannot handle it.\nUse "
648 "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
653 e820_reserve_resources();
654 e820_mark_nosave_regions(max_low_pfn);
657 request_resource(&iomem_resource, &video_ram_resource);
659 reserve_standard_io_resources();
664 #if defined(CONFIG_VGA_CONSOLE)
665 if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
666 conswitchp = &vga_con;
667 #elif defined(CONFIG_DUMMY_CONSOLE)
668 conswitchp = &dummy_con;