2 * linux/arch/i386/kernel/setup.c
4 * Copyright (C) 1995 Linus Torvalds
6 * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
8 * Memory region support
9 * David Parsons <orc@pell.chi.il.us>, July-August 1999
11 * Added E820 sanitization routine (removes overlapping memory regions);
12 * Brian Moyle <bmoyle@mvista.com>, February 2001
14 * Moved CPU detection code to cpu/${cpu}.c
15 * Patrick Mochel <mochel@osdl.org>, March 2002
17 * Provisions for empty E820 memory regions (reported by certain BIOSes).
18 * Alex Achenbach <xela@slit.de>, December 2002.
23 * This file handles the architecture-dependent parts of initialization
26 #include <linux/sched.h>
28 #include <linux/mmzone.h>
29 #include <linux/screen_info.h>
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/platform_device.h>
37 #include <linux/console.h>
38 #include <linux/mca.h>
39 #include <linux/root_dev.h>
40 #include <linux/highmem.h>
41 #include <linux/module.h>
42 #include <linux/efi.h>
43 #include <linux/init.h>
44 #include <linux/edd.h>
45 #include <linux/nodemask.h>
46 #include <linux/kexec.h>
47 #include <linux/crash_dump.h>
48 #include <linux/dmi.h>
49 #include <linux/pfn.h>
51 #include <video/edid.h>
55 #include <asm/mpspec.h>
56 #include <asm/mmzone.h>
57 #include <asm/setup.h>
58 #include <asm/arch_hooks.h>
59 #include <asm/sections.h>
60 #include <asm/io_apic.h>
63 #include <setup_arch.h>
64 #include <bios_ebda.h>
66 /* Forward Declaration. */
67 void __init find_max_pfn(void);
69 /* This value is set up by the early boot code to point to the value
70 immediately after the boot time page tables. It contains a *physical*
71 address, and must not be in the .bss segment! */
72 unsigned long init_pg_tables_end __initdata = ~0UL;
74 int disable_pse __devinitdata = 0;
82 EXPORT_SYMBOL(efi_enabled);
85 /* cpu data as detected by the assembly code in head.S */
86 struct cpuinfo_x86 new_cpu_data __initdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
87 /* common cpu data for all cpus */
88 struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
89 EXPORT_SYMBOL(boot_cpu_data);
91 unsigned long mmu_cr4_features;
93 /* for MCA, but anyone else can use it if they want */
94 unsigned int machine_id;
96 EXPORT_SYMBOL(machine_id);
98 unsigned int machine_submodel_id;
99 unsigned int BIOS_revision;
100 unsigned int mca_pentium_flag;
102 /* For PCI or other memory-mapped resources */
103 unsigned long pci_mem_start = 0x10000000;
105 EXPORT_SYMBOL(pci_mem_start);
108 /* Boot loader ID as an integer, for the benefit of proc_dointvec */
111 /* user-defined highmem size */
112 static unsigned int highmem_pages = -1;
117 struct drive_info_struct { char dummy[32]; } drive_info;
118 #if defined(CONFIG_BLK_DEV_IDE) || defined(CONFIG_BLK_DEV_HD) || \
119 defined(CONFIG_BLK_DEV_IDE_MODULE) || defined(CONFIG_BLK_DEV_HD_MODULE)
120 EXPORT_SYMBOL(drive_info);
122 struct screen_info screen_info;
123 EXPORT_SYMBOL(screen_info);
124 struct apm_info apm_info;
125 EXPORT_SYMBOL(apm_info);
126 struct sys_desc_table_struct {
127 unsigned short length;
128 unsigned char table[0];
130 struct edid_info edid_info;
131 EXPORT_SYMBOL_GPL(edid_info);
132 struct ist_info ist_info;
133 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
134 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
135 EXPORT_SYMBOL(ist_info);
139 extern void early_cpu_init(void);
140 extern int root_mountflags;
142 unsigned long saved_videomode;
144 #define RAMDISK_IMAGE_START_MASK 0x07FF
145 #define RAMDISK_PROMPT_FLAG 0x8000
146 #define RAMDISK_LOAD_FLAG 0x4000
148 static char command_line[COMMAND_LINE_SIZE];
150 unsigned char __initdata boot_params[PARAM_SIZE];
152 static struct resource data_resource = {
153 .name = "Kernel data",
156 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
159 static struct resource code_resource = {
160 .name = "Kernel code",
163 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
166 static struct resource system_rom_resource = {
167 .name = "System ROM",
170 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
173 static struct resource extension_rom_resource = {
174 .name = "Extension ROM",
177 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
180 static struct resource adapter_rom_resources[] = { {
181 .name = "Adapter ROM",
184 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
186 .name = "Adapter ROM",
189 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
191 .name = "Adapter ROM",
194 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
196 .name = "Adapter ROM",
199 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
201 .name = "Adapter ROM",
204 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
206 .name = "Adapter ROM",
209 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
212 #define ADAPTER_ROM_RESOURCES \
213 (sizeof adapter_rom_resources / sizeof adapter_rom_resources[0])
215 static struct resource video_rom_resource = {
219 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
222 static struct resource video_ram_resource = {
223 .name = "Video RAM area",
226 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
229 static struct resource standard_io_resources[] = { {
233 .flags = IORESOURCE_BUSY | IORESOURCE_IO
238 .flags = IORESOURCE_BUSY | IORESOURCE_IO
243 .flags = IORESOURCE_BUSY | IORESOURCE_IO
248 .flags = IORESOURCE_BUSY | IORESOURCE_IO
253 .flags = IORESOURCE_BUSY | IORESOURCE_IO
255 .name = "dma page reg",
258 .flags = IORESOURCE_BUSY | IORESOURCE_IO
263 .flags = IORESOURCE_BUSY | IORESOURCE_IO
268 .flags = IORESOURCE_BUSY | IORESOURCE_IO
273 .flags = IORESOURCE_BUSY | IORESOURCE_IO
276 #define STANDARD_IO_RESOURCES \
277 (sizeof standard_io_resources / sizeof standard_io_resources[0])
279 #define romsignature(x) (*(unsigned short *)(x) == 0xaa55)
281 static int __init romchecksum(unsigned char *rom, unsigned long length)
283 unsigned char *p, sum = 0;
285 for (p = rom; p < rom + length; p++)
290 static void __init probe_roms(void)
292 unsigned long start, length, upper;
297 upper = adapter_rom_resources[0].start;
298 for (start = video_rom_resource.start; start < upper; start += 2048) {
299 rom = isa_bus_to_virt(start);
300 if (!romsignature(rom))
303 video_rom_resource.start = start;
305 /* 0 < length <= 0x7f * 512, historically */
306 length = rom[2] * 512;
308 /* if checksum okay, trust length byte */
309 if (length && romchecksum(rom, length))
310 video_rom_resource.end = start + length - 1;
312 request_resource(&iomem_resource, &video_rom_resource);
316 start = (video_rom_resource.end + 1 + 2047) & ~2047UL;
321 request_resource(&iomem_resource, &system_rom_resource);
322 upper = system_rom_resource.start;
324 /* check for extension rom (ignore length byte!) */
325 rom = isa_bus_to_virt(extension_rom_resource.start);
326 if (romsignature(rom)) {
327 length = extension_rom_resource.end - extension_rom_resource.start + 1;
328 if (romchecksum(rom, length)) {
329 request_resource(&iomem_resource, &extension_rom_resource);
330 upper = extension_rom_resource.start;
334 /* check for adapter roms on 2k boundaries */
335 for (i = 0; i < ADAPTER_ROM_RESOURCES && start < upper; start += 2048) {
336 rom = isa_bus_to_virt(start);
337 if (!romsignature(rom))
340 /* 0 < length <= 0x7f * 512, historically */
341 length = rom[2] * 512;
343 /* but accept any length that fits if checksum okay */
344 if (!length || start + length > upper || !romchecksum(rom, length))
347 adapter_rom_resources[i].start = start;
348 adapter_rom_resources[i].end = start + length - 1;
349 request_resource(&iomem_resource, &adapter_rom_resources[i]);
351 start = adapter_rom_resources[i++].end & ~2047UL;
355 static void __init limit_regions(unsigned long long size)
357 unsigned long long current_addr = 0;
361 efi_memory_desc_t *md;
364 for (p = memmap.map, i = 0; p < memmap.map_end;
365 p += memmap.desc_size, i++) {
367 current_addr = md->phys_addr + (md->num_pages << 12);
368 if (md->type == EFI_CONVENTIONAL_MEMORY) {
369 if (current_addr >= size) {
371 (((current_addr-size) + PAGE_SIZE-1) >> PAGE_SHIFT);
372 memmap.nr_map = i + 1;
378 for (i = 0; i < e820.nr_map; i++) {
379 current_addr = e820.map[i].addr + e820.map[i].size;
380 if (current_addr < size)
383 if (e820.map[i].type != E820_RAM)
386 if (e820.map[i].addr >= size) {
388 * This region starts past the end of the
389 * requested size, skip it completely.
394 e820.map[i].size -= current_addr - size;
400 void __init add_memory_region(unsigned long long start,
401 unsigned long long size, int type)
409 printk(KERN_ERR "Ooops! Too many entries in the memory map!\n");
413 e820.map[x].addr = start;
414 e820.map[x].size = size;
415 e820.map[x].type = type;
418 } /* add_memory_region */
422 static void __init print_memory_map(char *who)
426 for (i = 0; i < e820.nr_map; i++) {
427 printk(" %s: %016Lx - %016Lx ", who,
429 e820.map[i].addr + e820.map[i].size);
430 switch (e820.map[i].type) {
431 case E820_RAM: printk("(usable)\n");
434 printk("(reserved)\n");
437 printk("(ACPI data)\n");
440 printk("(ACPI NVS)\n");
442 default: printk("type %lu\n", e820.map[i].type);
449 * Sanitize the BIOS e820 map.
451 * Some e820 responses include overlapping entries. The following
452 * replaces the original e820 map with a new one, removing overlaps.
455 struct change_member {
456 struct e820entry *pbios; /* pointer to original bios entry */
457 unsigned long long addr; /* address for this change point */
459 static struct change_member change_point_list[2*E820MAX] __initdata;
460 static struct change_member *change_point[2*E820MAX] __initdata;
461 static struct e820entry *overlap_list[E820MAX] __initdata;
462 static struct e820entry new_bios[E820MAX] __initdata;
464 int __init sanitize_e820_map(struct e820entry * biosmap, char * pnr_map)
466 struct change_member *change_tmp;
467 unsigned long current_type, last_type;
468 unsigned long long last_addr;
469 int chgidx, still_changing;
472 int old_nr, new_nr, chg_nr;
476 Visually we're performing the following (1,2,3,4 = memory types)...
478 Sample memory map (w/overlaps):
479 ____22__________________
480 ______________________4_
481 ____1111________________
482 _44_____________________
483 11111111________________
484 ____________________33__
485 ___________44___________
486 __________33333_________
487 ______________22________
488 ___________________2222_
489 _________111111111______
490 _____________________11_
491 _________________4______
493 Sanitized equivalent (no overlap):
494 1_______________________
495 _44_____________________
496 ___1____________________
497 ____22__________________
498 ______11________________
499 _________1______________
500 __________3_____________
501 ___________44___________
502 _____________33_________
503 _______________2________
504 ________________1_______
505 _________________4______
506 ___________________2____
507 ____________________33__
508 ______________________4_
511 /* if there's only one memory region, don't bother */
517 /* bail out if we find any unreasonable addresses in bios map */
518 for (i=0; i<old_nr; i++)
519 if (biosmap[i].addr + biosmap[i].size < biosmap[i].addr)
522 /* create pointers for initial change-point information (for sorting) */
523 for (i=0; i < 2*old_nr; i++)
524 change_point[i] = &change_point_list[i];
526 /* record all known change-points (starting and ending addresses),
527 omitting those that are for empty memory regions */
529 for (i=0; i < old_nr; i++) {
530 if (biosmap[i].size != 0) {
531 change_point[chgidx]->addr = biosmap[i].addr;
532 change_point[chgidx++]->pbios = &biosmap[i];
533 change_point[chgidx]->addr = biosmap[i].addr + biosmap[i].size;
534 change_point[chgidx++]->pbios = &biosmap[i];
537 chg_nr = chgidx; /* true number of change-points */
539 /* sort change-point list by memory addresses (low -> high) */
541 while (still_changing) {
543 for (i=1; i < chg_nr; i++) {
544 /* if <current_addr> > <last_addr>, swap */
545 /* or, if current=<start_addr> & last=<end_addr>, swap */
546 if ((change_point[i]->addr < change_point[i-1]->addr) ||
547 ((change_point[i]->addr == change_point[i-1]->addr) &&
548 (change_point[i]->addr == change_point[i]->pbios->addr) &&
549 (change_point[i-1]->addr != change_point[i-1]->pbios->addr))
552 change_tmp = change_point[i];
553 change_point[i] = change_point[i-1];
554 change_point[i-1] = change_tmp;
560 /* create a new bios memory map, removing overlaps */
561 overlap_entries=0; /* number of entries in the overlap table */
562 new_bios_entry=0; /* index for creating new bios map entries */
563 last_type = 0; /* start with undefined memory type */
564 last_addr = 0; /* start with 0 as last starting address */
565 /* loop through change-points, determining affect on the new bios map */
566 for (chgidx=0; chgidx < chg_nr; chgidx++)
568 /* keep track of all overlapping bios entries */
569 if (change_point[chgidx]->addr == change_point[chgidx]->pbios->addr)
571 /* add map entry to overlap list (> 1 entry implies an overlap) */
572 overlap_list[overlap_entries++]=change_point[chgidx]->pbios;
576 /* remove entry from list (order independent, so swap with last) */
577 for (i=0; i<overlap_entries; i++)
579 if (overlap_list[i] == change_point[chgidx]->pbios)
580 overlap_list[i] = overlap_list[overlap_entries-1];
584 /* if there are overlapping entries, decide which "type" to use */
585 /* (larger value takes precedence -- 1=usable, 2,3,4,4+=unusable) */
587 for (i=0; i<overlap_entries; i++)
588 if (overlap_list[i]->type > current_type)
589 current_type = overlap_list[i]->type;
590 /* continue building up new bios map based on this information */
591 if (current_type != last_type) {
592 if (last_type != 0) {
593 new_bios[new_bios_entry].size =
594 change_point[chgidx]->addr - last_addr;
595 /* move forward only if the new size was non-zero */
596 if (new_bios[new_bios_entry].size != 0)
597 if (++new_bios_entry >= E820MAX)
598 break; /* no more space left for new bios entries */
600 if (current_type != 0) {
601 new_bios[new_bios_entry].addr = change_point[chgidx]->addr;
602 new_bios[new_bios_entry].type = current_type;
603 last_addr=change_point[chgidx]->addr;
605 last_type = current_type;
608 new_nr = new_bios_entry; /* retain count for new bios entries */
610 /* copy new bios mapping into original location */
611 memcpy(biosmap, new_bios, new_nr*sizeof(struct e820entry));
618 * Copy the BIOS e820 map into a safe place.
620 * Sanity-check it while we're at it..
622 * If we're lucky and live on a modern system, the setup code
623 * will have given us a memory map that we can use to properly
624 * set up memory. If we aren't, we'll fake a memory map.
626 * We check to see that the memory map contains at least 2 elements
627 * before we'll use it, because the detection code in setup.S may
628 * not be perfect and most every PC known to man has two memory
629 * regions: one from 0 to 640k, and one from 1mb up. (The IBM
630 * thinkpad 560x, for example, does not cooperate with the memory
633 int __init copy_e820_map(struct e820entry * biosmap, int nr_map)
635 /* Only one memory region (or negative)? Ignore it */
640 unsigned long long start = biosmap->addr;
641 unsigned long long size = biosmap->size;
642 unsigned long long end = start + size;
643 unsigned long type = biosmap->type;
645 /* Overflow in 64 bits? Ignore the memory map. */
650 * Some BIOSes claim RAM in the 640k - 1M region.
651 * Not right. Fix it up.
653 if (type == E820_RAM) {
654 if (start < 0x100000ULL && end > 0xA0000ULL) {
655 if (start < 0xA0000ULL)
656 add_memory_region(start, 0xA0000ULL-start, type);
657 if (end <= 0x100000ULL)
663 add_memory_region(start, size, type);
664 } while (biosmap++,--nr_map);
668 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
670 #ifdef CONFIG_EDD_MODULE
674 * copy_edd() - Copy the BIOS EDD information
675 * from boot_params into a safe place.
678 static inline void copy_edd(void)
680 memcpy(edd.mbr_signature, EDD_MBR_SIGNATURE, sizeof(edd.mbr_signature));
681 memcpy(edd.edd_info, EDD_BUF, sizeof(edd.edd_info));
682 edd.mbr_signature_nr = EDD_MBR_SIG_NR;
683 edd.edd_info_nr = EDD_NR;
686 static inline void copy_edd(void)
691 static int __initdata user_defined_memmap = 0;
694 * "mem=nopentium" disables the 4MB page tables.
695 * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM
696 * to <mem>, overriding the bios size.
697 * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from
698 * <start> to <start>+<mem>, overriding the bios size.
700 * HPA tells me bootloaders need to parse mem=, so no new
701 * option should be mem= [also see Documentation/i386/boot.txt]
703 static int __init parse_mem(char *arg)
708 if (strcmp(arg, "nopentium") == 0) {
709 clear_bit(X86_FEATURE_PSE, boot_cpu_data.x86_capability);
712 /* If the user specifies memory size, we
713 * limit the BIOS-provided memory map to
714 * that size. exactmap can be used to specify
715 * the exact map. mem=number can be used to
716 * trim the existing memory map.
718 unsigned long long mem_size;
720 mem_size = memparse(arg, &arg);
721 limit_regions(mem_size);
722 user_defined_memmap = 1;
726 early_param("mem", parse_mem);
728 static int __init parse_memmap(char *arg)
733 if (strcmp(arg, "exactmap") == 0) {
734 #ifdef CONFIG_CRASH_DUMP
735 /* If we are doing a crash dump, we
736 * still need to know the real mem
737 * size before original memory map is
741 saved_max_pfn = max_pfn;
744 user_defined_memmap = 1;
746 /* If the user specifies memory size, we
747 * limit the BIOS-provided memory map to
748 * that size. exactmap can be used to specify
749 * the exact map. mem=number can be used to
750 * trim the existing memory map.
752 unsigned long long start_at, mem_size;
754 mem_size = memparse(arg, &arg);
756 start_at = memparse(arg+1, &arg);
757 add_memory_region(start_at, mem_size, E820_RAM);
758 } else if (*arg == '#') {
759 start_at = memparse(arg+1, &arg);
760 add_memory_region(start_at, mem_size, E820_ACPI);
761 } else if (*arg == '$') {
762 start_at = memparse(arg+1, &arg);
763 add_memory_region(start_at, mem_size, E820_RESERVED);
765 limit_regions(mem_size);
766 user_defined_memmap = 1;
771 early_param("memmap", parse_memmap);
773 #ifdef CONFIG_PROC_VMCORE
774 /* elfcorehdr= specifies the location of elf core header
775 * stored by the crashed kernel.
777 static int __init parse_elfcorehdr(char *arg)
782 elfcorehdr_addr = memparse(arg, &arg);
785 early_param("elfcorehdr", parse_elfcorehdr);
786 #endif /* CONFIG_PROC_VMCORE */
789 * highmem=size forces highmem to be exactly 'size' bytes.
790 * This works even on boxes that have no highmem otherwise.
791 * This also works to reduce highmem size on bigger boxes.
793 static int __init parse_highmem(char *arg)
798 highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT;
801 early_param("highmem", parse_highmem);
804 * vmalloc=size forces the vmalloc area to be exactly 'size'
805 * bytes. This can be used to increase (or decrease) the
806 * vmalloc area - the default is 128m.
808 static int __init parse_vmalloc(char *arg)
813 __VMALLOC_RESERVE = memparse(arg, &arg);
816 early_param("vmalloc", parse_vmalloc);
819 * reservetop=size reserves a hole at the top of the kernel address space which
820 * a hypervisor can load into later. Needed for dynamically loaded hypervisors,
821 * so relocating the fixmap can be done before paging initialization.
823 static int __init parse_reservetop(char *arg)
825 unsigned long address;
830 address = memparse(arg, &arg);
831 reserve_top_address(address);
834 early_param("reservetop", parse_reservetop);
837 * Callback for efi_memory_walk.
840 efi_find_max_pfn(unsigned long start, unsigned long end, void *arg)
842 unsigned long *max_pfn = arg, pfn;
845 pfn = PFN_UP(end -1);
853 efi_memory_present_wrapper(unsigned long start, unsigned long end, void *arg)
855 memory_present(0, start, end);
860 * This function checks if the entire range <start,end> is mapped with type.
862 * Note: this function only works correct if the e820 table is sorted and
863 * not-overlapping, which is the case
866 e820_all_mapped(unsigned long s, unsigned long e, unsigned type)
871 for (i = 0; i < e820.nr_map; i++) {
872 struct e820entry *ei = &e820.map[i];
873 if (type && ei->type != type)
875 /* is the region (part) in overlap with the current region ?*/
876 if (ei->addr >= end || ei->addr + ei->size <= start)
878 /* if the region is at the beginning of <start,end> we move
879 * start to the end of the region since it's ok until there
881 if (ei->addr <= start)
882 start = ei->addr + ei->size;
883 /* if start is now at or beyond end, we're done, full
886 return 1; /* we're done */
892 * Find the highest page frame number we have available
894 void __init find_max_pfn(void)
900 efi_memmap_walk(efi_find_max_pfn, &max_pfn);
901 efi_memmap_walk(efi_memory_present_wrapper, NULL);
905 for (i = 0; i < e820.nr_map; i++) {
906 unsigned long start, end;
908 if (e820.map[i].type != E820_RAM)
910 start = PFN_UP(e820.map[i].addr);
911 end = PFN_DOWN(e820.map[i].addr + e820.map[i].size);
916 memory_present(0, start, end);
921 * Determine low and high memory ranges:
923 unsigned long __init find_max_low_pfn(void)
925 unsigned long max_low_pfn;
927 max_low_pfn = max_pfn;
928 if (max_low_pfn > MAXMEM_PFN) {
929 if (highmem_pages == -1)
930 highmem_pages = max_pfn - MAXMEM_PFN;
931 if (highmem_pages + MAXMEM_PFN < max_pfn)
932 max_pfn = MAXMEM_PFN + highmem_pages;
933 if (highmem_pages + MAXMEM_PFN > max_pfn) {
934 printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn - MAXMEM_PFN), pages_to_mb(highmem_pages));
937 max_low_pfn = MAXMEM_PFN;
938 #ifndef CONFIG_HIGHMEM
939 /* Maximum memory usable is what is directly addressable */
940 printk(KERN_WARNING "Warning only %ldMB will be used.\n",
942 if (max_pfn > MAX_NONPAE_PFN)
943 printk(KERN_WARNING "Use a PAE enabled kernel.\n");
945 printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n");
946 max_pfn = MAXMEM_PFN;
947 #else /* !CONFIG_HIGHMEM */
948 #ifndef CONFIG_X86_PAE
949 if (max_pfn > MAX_NONPAE_PFN) {
950 max_pfn = MAX_NONPAE_PFN;
951 printk(KERN_WARNING "Warning only 4GB will be used.\n");
952 printk(KERN_WARNING "Use a PAE enabled kernel.\n");
954 #endif /* !CONFIG_X86_PAE */
955 #endif /* !CONFIG_HIGHMEM */
957 if (highmem_pages == -1)
959 #ifdef CONFIG_HIGHMEM
960 if (highmem_pages >= max_pfn) {
961 printk(KERN_ERR "highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages), pages_to_mb(max_pfn));
965 if (max_low_pfn-highmem_pages < 64*1024*1024/PAGE_SIZE){
966 printk(KERN_ERR "highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages));
969 max_low_pfn -= highmem_pages;
973 printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n");
980 * Free all available memory for boot time allocation. Used
981 * as a callback function by efi_memory_walk()
985 free_available_memory(unsigned long start, unsigned long end, void *arg)
987 /* check max_low_pfn */
988 if (start >= (max_low_pfn << PAGE_SHIFT))
990 if (end >= (max_low_pfn << PAGE_SHIFT))
991 end = max_low_pfn << PAGE_SHIFT;
993 free_bootmem(start, end - start);
998 * Register fully available low RAM pages with the bootmem allocator.
1000 static void __init register_bootmem_low_pages(unsigned long max_low_pfn)
1005 efi_memmap_walk(free_available_memory, NULL);
1008 for (i = 0; i < e820.nr_map; i++) {
1009 unsigned long curr_pfn, last_pfn, size;
1011 * Reserve usable low memory
1013 if (e820.map[i].type != E820_RAM)
1016 * We are rounding up the start address of usable memory:
1018 curr_pfn = PFN_UP(e820.map[i].addr);
1019 if (curr_pfn >= max_low_pfn)
1022 * ... and at the end of the usable range downwards:
1024 last_pfn = PFN_DOWN(e820.map[i].addr + e820.map[i].size);
1026 if (last_pfn > max_low_pfn)
1027 last_pfn = max_low_pfn;
1030 * .. finally, did all the rounding and playing
1031 * around just make the area go away?
1033 if (last_pfn <= curr_pfn)
1036 size = last_pfn - curr_pfn;
1037 free_bootmem(PFN_PHYS(curr_pfn), PFN_PHYS(size));
1042 * workaround for Dell systems that neglect to reserve EBDA
1044 static void __init reserve_ebda_region(void)
1047 addr = get_bios_ebda();
1049 reserve_bootmem(addr, PAGE_SIZE);
1052 #ifndef CONFIG_NEED_MULTIPLE_NODES
1053 void __init setup_bootmem_allocator(void);
1054 static unsigned long __init setup_memory(void)
1057 * partially used pages are not usable - thus
1058 * we are rounding upwards:
1060 min_low_pfn = PFN_UP(init_pg_tables_end);
1064 max_low_pfn = find_max_low_pfn();
1066 #ifdef CONFIG_HIGHMEM
1067 highstart_pfn = highend_pfn = max_pfn;
1068 if (max_pfn > max_low_pfn) {
1069 highstart_pfn = max_low_pfn;
1071 printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
1072 pages_to_mb(highend_pfn - highstart_pfn));
1073 num_physpages = highend_pfn;
1074 high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
1076 num_physpages = max_low_pfn;
1077 high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
1079 #ifdef CONFIG_FLATMEM
1080 max_mapnr = num_physpages;
1082 printk(KERN_NOTICE "%ldMB LOWMEM available.\n",
1083 pages_to_mb(max_low_pfn));
1085 setup_bootmem_allocator();
1090 void __init zone_sizes_init(void)
1092 #ifdef CONFIG_HIGHMEM
1093 unsigned long max_zone_pfns[MAX_NR_ZONES] = {
1094 virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT,
1097 add_active_range(0, 0, highend_pfn);
1099 unsigned long max_zone_pfns[MAX_NR_ZONES] = {
1100 virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT,
1102 add_active_range(0, 0, max_low_pfn);
1105 free_area_init_nodes(max_zone_pfns);
1108 extern unsigned long __init setup_memory(void);
1109 extern void zone_sizes_init(void);
1110 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
1112 void __init setup_bootmem_allocator(void)
1114 unsigned long bootmap_size;
1116 * Initialize the boot-time allocator (with low memory only):
1118 bootmap_size = init_bootmem(min_low_pfn, max_low_pfn);
1120 register_bootmem_low_pages(max_low_pfn);
1123 * Reserve the bootmem bitmap itself as well. We do this in two
1124 * steps (first step was init_bootmem()) because this catches
1125 * the (very unlikely) case of us accidentally initializing the
1126 * bootmem allocator with an invalid RAM area.
1128 reserve_bootmem(__PHYSICAL_START, (PFN_PHYS(min_low_pfn) +
1129 bootmap_size + PAGE_SIZE-1) - (__PHYSICAL_START));
1132 * reserve physical page 0 - it's a special BIOS page on many boxes,
1133 * enabling clean reboots, SMP operation, laptop functions.
1135 reserve_bootmem(0, PAGE_SIZE);
1137 /* reserve EBDA region, it's a 4K region */
1138 reserve_ebda_region();
1140 /* could be an AMD 768MPX chipset. Reserve a page before VGA to prevent
1141 PCI prefetch into it (errata #56). Usually the page is reserved anyways,
1142 unless you have no PS/2 mouse plugged in. */
1143 if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
1144 boot_cpu_data.x86 == 6)
1145 reserve_bootmem(0xa0000 - 4096, 4096);
1149 * But first pinch a few for the stack/trampoline stuff
1150 * FIXME: Don't need the extra page at 4K, but need to fix
1151 * trampoline before removing it. (see the GDT stuff)
1153 reserve_bootmem(PAGE_SIZE, PAGE_SIZE);
1155 #ifdef CONFIG_ACPI_SLEEP
1157 * Reserve low memory region for sleep support.
1159 acpi_reserve_bootmem();
1161 #ifdef CONFIG_X86_FIND_SMP_CONFIG
1163 * Find and reserve possible boot-time SMP configuration:
1168 #ifdef CONFIG_BLK_DEV_INITRD
1169 if (LOADER_TYPE && INITRD_START) {
1170 if (INITRD_START + INITRD_SIZE <= (max_low_pfn << PAGE_SHIFT)) {
1171 reserve_bootmem(INITRD_START, INITRD_SIZE);
1173 INITRD_START ? INITRD_START + PAGE_OFFSET : 0;
1174 initrd_end = initrd_start+INITRD_SIZE;
1177 printk(KERN_ERR "initrd extends beyond end of memory "
1178 "(0x%08lx > 0x%08lx)\ndisabling initrd\n",
1179 INITRD_START + INITRD_SIZE,
1180 max_low_pfn << PAGE_SHIFT);
1186 if (crashk_res.start != crashk_res.end)
1187 reserve_bootmem(crashk_res.start,
1188 crashk_res.end - crashk_res.start + 1);
1193 * The node 0 pgdat is initialized before all of these because
1194 * it's needed for bootmem. node>0 pgdats have their virtual
1195 * space allocated before the pagetables are in place to access
1196 * them, so they can't be cleared then.
1198 * This should all compile down to nothing when NUMA is off.
1200 void __init remapped_pgdat_init(void)
1204 for_each_online_node(nid) {
1206 memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
1211 * Request address space for all standard RAM and ROM resources
1212 * and also for regions reported as reserved by the e820.
1215 legacy_init_iomem_resources(struct resource *code_resource, struct resource *data_resource)
1220 for (i = 0; i < e820.nr_map; i++) {
1221 struct resource *res;
1222 #ifndef CONFIG_RESOURCES_64BIT
1223 if (e820.map[i].addr + e820.map[i].size > 0x100000000ULL)
1226 res = kzalloc(sizeof(struct resource), GFP_ATOMIC);
1227 switch (e820.map[i].type) {
1228 case E820_RAM: res->name = "System RAM"; break;
1229 case E820_ACPI: res->name = "ACPI Tables"; break;
1230 case E820_NVS: res->name = "ACPI Non-volatile Storage"; break;
1231 default: res->name = "reserved";
1233 res->start = e820.map[i].addr;
1234 res->end = res->start + e820.map[i].size - 1;
1235 res->flags = IORESOURCE_MEM | IORESOURCE_BUSY;
1236 if (request_resource(&iomem_resource, res)) {
1240 if (e820.map[i].type == E820_RAM) {
1242 * We don't know which RAM region contains kernel data,
1243 * so we try it repeatedly and let the resource manager
1246 request_resource(res, code_resource);
1247 request_resource(res, data_resource);
1249 request_resource(res, &crashk_res);
1256 * Request address space for all standard resources
1258 * This is called just before pcibios_init(), which is also a
1259 * subsys_initcall, but is linked in later (in arch/i386/pci/common.c).
1261 static int __init request_standard_resources(void)
1265 printk("Setting up standard PCI resources\n");
1267 efi_initialize_iomem_resources(&code_resource, &data_resource);
1269 legacy_init_iomem_resources(&code_resource, &data_resource);
1271 /* EFI systems may still have VGA */
1272 request_resource(&iomem_resource, &video_ram_resource);
1274 /* request I/O space for devices used on all i[345]86 PCs */
1275 for (i = 0; i < STANDARD_IO_RESOURCES; i++)
1276 request_resource(&ioport_resource, &standard_io_resources[i]);
1280 subsys_initcall(request_standard_resources);
1282 static void __init register_memory(void)
1284 unsigned long gapstart, gapsize, round;
1285 unsigned long long last;
1289 * Search for the bigest gap in the low 32 bits of the e820
1292 last = 0x100000000ull;
1293 gapstart = 0x10000000;
1297 unsigned long long start = e820.map[i].addr;
1298 unsigned long long end = start + e820.map[i].size;
1301 * Since "last" is at most 4GB, we know we'll
1302 * fit in 32 bits if this condition is true
1305 unsigned long gap = last - end;
1307 if (gap > gapsize) {
1317 * See how much we want to round up: start off with
1318 * rounding to the next 1MB area.
1321 while ((gapsize >> 4) > round)
1323 /* Fun with two's complement */
1324 pci_mem_start = (gapstart + round) & -round;
1326 printk("Allocating PCI resources starting at %08lx (gap: %08lx:%08lx)\n",
1327 pci_mem_start, gapstart, gapsize);
1331 static void set_mca_bus(int x)
1336 static void set_mca_bus(int x) { }
1340 * Determine if we were loaded by an EFI loader. If so, then we have also been
1341 * passed the efi memmap, systab, etc., so we should use these data structures
1342 * for initialization. Note, the efi init code path is determined by the
1343 * global efi_enabled. This allows the same kernel image to be used on existing
1344 * systems (with a traditional BIOS) as well as on EFI systems.
1346 void __init setup_arch(char **cmdline_p)
1348 unsigned long max_low_pfn;
1350 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
1351 pre_setup_arch_hook();
1355 * FIXME: This isn't an official loader_type right
1356 * now but does currently work with elilo.
1357 * If we were configured as an EFI kernel, check to make
1358 * sure that we were loaded correctly from elilo and that
1359 * the system table is valid. If not, then initialize normally.
1362 if ((LOADER_TYPE == 0x50) && EFI_SYSTAB)
1366 ROOT_DEV = old_decode_dev(ORIG_ROOT_DEV);
1367 drive_info = DRIVE_INFO;
1368 screen_info = SCREEN_INFO;
1369 edid_info = EDID_INFO;
1370 apm_info.bios = APM_BIOS_INFO;
1371 ist_info = IST_INFO;
1372 saved_videomode = VIDEO_MODE;
1373 if( SYS_DESC_TABLE.length != 0 ) {
1374 set_mca_bus(SYS_DESC_TABLE.table[3] & 0x2);
1375 machine_id = SYS_DESC_TABLE.table[0];
1376 machine_submodel_id = SYS_DESC_TABLE.table[1];
1377 BIOS_revision = SYS_DESC_TABLE.table[2];
1379 bootloader_type = LOADER_TYPE;
1381 #ifdef CONFIG_BLK_DEV_RAM
1382 rd_image_start = RAMDISK_FLAGS & RAMDISK_IMAGE_START_MASK;
1383 rd_prompt = ((RAMDISK_FLAGS & RAMDISK_PROMPT_FLAG) != 0);
1384 rd_doload = ((RAMDISK_FLAGS & RAMDISK_LOAD_FLAG) != 0);
1390 printk(KERN_INFO "BIOS-provided physical RAM map:\n");
1391 print_memory_map(machine_specific_memory_setup());
1396 if (!MOUNT_ROOT_RDONLY)
1397 root_mountflags &= ~MS_RDONLY;
1398 init_mm.start_code = (unsigned long) _text;
1399 init_mm.end_code = (unsigned long) _etext;
1400 init_mm.end_data = (unsigned long) _edata;
1401 init_mm.brk = init_pg_tables_end + PAGE_OFFSET;
1403 code_resource.start = virt_to_phys(_text);
1404 code_resource.end = virt_to_phys(_etext)-1;
1405 data_resource.start = virt_to_phys(_etext);
1406 data_resource.end = virt_to_phys(_edata)-1;
1408 parse_early_param();
1410 if (user_defined_memmap) {
1411 printk(KERN_INFO "user-defined physical RAM map:\n");
1412 print_memory_map("user");
1415 strlcpy(command_line, saved_command_line, COMMAND_LINE_SIZE);
1416 *cmdline_p = command_line;
1418 max_low_pfn = setup_memory();
1421 * NOTE: before this point _nobody_ is allowed to allocate
1422 * any memory using the bootmem allocator. Although the
1423 * alloctor is now initialised only the first 8Mb of the kernel
1424 * virtual address space has been mapped. All allocations before
1425 * paging_init() has completed must use the alloc_bootmem_low_pages()
1426 * variant (which allocates DMA'able memory) and care must be taken
1427 * not to exceed the 8Mb limit.
1431 smp_alloc_memory(); /* AP processor realmode stacks in low memory*/
1434 remapped_pgdat_init();
1439 * NOTE: at this point the bootmem allocator is fully available.
1444 #ifdef CONFIG_X86_GENERICARCH
1445 generic_apic_probe();
1452 * Parse the ACPI tables for possible boot-time SMP configuration.
1454 acpi_boot_table_init();
1458 #ifdef CONFIG_X86_IO_APIC
1459 check_acpi_pci(); /* Checks more than just ACPI actually */
1466 #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC)
1468 printk(KERN_WARNING "More than 8 CPUs detected and "
1469 "CONFIG_X86_PC cannot handle it.\nUse "
1470 "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
1473 #ifdef CONFIG_X86_LOCAL_APIC
1474 if (smp_found_config)
1481 #if defined(CONFIG_VGA_CONSOLE)
1482 if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1483 conswitchp = &vga_con;
1484 #elif defined(CONFIG_DUMMY_CONSOLE)
1485 conswitchp = &dummy_con;
1491 static __init int add_pcspkr(void)
1493 struct platform_device *pd;
1496 pd = platform_device_alloc("pcspkr", -1);
1500 ret = platform_device_add(pd);
1502 platform_device_put(pd);
1506 device_initcall(add_pcspkr);
1511 * c-file-style:"k&r"