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/setup.h>
57 #include <asm/arch_hooks.h>
58 #include <asm/sections.h>
59 #include <asm/io_apic.h>
62 #include <setup_arch.h>
63 #include <bios_ebda.h>
65 /* Forward Declaration. */
66 void __init find_max_pfn(void);
68 /* This value is set up by the early boot code to point to the value
69 immediately after the boot time page tables. It contains a *physical*
70 address, and must not be in the .bss segment! */
71 unsigned long init_pg_tables_end __initdata = ~0UL;
73 int disable_pse __devinitdata = 0;
81 EXPORT_SYMBOL(efi_enabled);
84 /* cpu data as detected by the assembly code in head.S */
85 struct cpuinfo_x86 new_cpu_data __initdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
86 /* common cpu data for all cpus */
87 struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
88 EXPORT_SYMBOL(boot_cpu_data);
90 unsigned long mmu_cr4_features;
93 int acpi_disabled = 0;
95 int acpi_disabled = 1;
97 EXPORT_SYMBOL(acpi_disabled);
100 int __initdata acpi_force = 0;
101 extern acpi_interrupt_flags acpi_sci_flags;
104 /* for MCA, but anyone else can use it if they want */
105 unsigned int machine_id;
107 EXPORT_SYMBOL(machine_id);
109 unsigned int machine_submodel_id;
110 unsigned int BIOS_revision;
111 unsigned int mca_pentium_flag;
113 /* For PCI or other memory-mapped resources */
114 unsigned long pci_mem_start = 0x10000000;
116 EXPORT_SYMBOL(pci_mem_start);
119 /* Boot loader ID as an integer, for the benefit of proc_dointvec */
122 /* user-defined highmem size */
123 static unsigned int highmem_pages = -1;
128 struct drive_info_struct { char dummy[32]; } drive_info;
129 #if defined(CONFIG_BLK_DEV_IDE) || defined(CONFIG_BLK_DEV_HD) || \
130 defined(CONFIG_BLK_DEV_IDE_MODULE) || defined(CONFIG_BLK_DEV_HD_MODULE)
131 EXPORT_SYMBOL(drive_info);
133 struct screen_info screen_info;
134 EXPORT_SYMBOL(screen_info);
135 struct apm_info apm_info;
136 EXPORT_SYMBOL(apm_info);
137 struct sys_desc_table_struct {
138 unsigned short length;
139 unsigned char table[0];
141 struct edid_info edid_info;
142 EXPORT_SYMBOL_GPL(edid_info);
143 struct ist_info ist_info;
144 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
145 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
146 EXPORT_SYMBOL(ist_info);
150 extern void early_cpu_init(void);
151 extern void generic_apic_probe(char *);
152 extern int root_mountflags;
154 unsigned long saved_videomode;
156 #define RAMDISK_IMAGE_START_MASK 0x07FF
157 #define RAMDISK_PROMPT_FLAG 0x8000
158 #define RAMDISK_LOAD_FLAG 0x4000
160 static char command_line[COMMAND_LINE_SIZE];
162 unsigned char __initdata boot_params[PARAM_SIZE];
164 static struct resource data_resource = {
165 .name = "Kernel data",
168 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
171 static struct resource code_resource = {
172 .name = "Kernel code",
175 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
178 static struct resource system_rom_resource = {
179 .name = "System ROM",
182 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
185 static struct resource extension_rom_resource = {
186 .name = "Extension ROM",
189 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
192 static struct resource adapter_rom_resources[] = { {
193 .name = "Adapter ROM",
196 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
198 .name = "Adapter ROM",
201 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
203 .name = "Adapter ROM",
206 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
208 .name = "Adapter ROM",
211 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
213 .name = "Adapter ROM",
216 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
218 .name = "Adapter ROM",
221 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
224 #define ADAPTER_ROM_RESOURCES \
225 (sizeof adapter_rom_resources / sizeof adapter_rom_resources[0])
227 static struct resource video_rom_resource = {
231 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
234 static struct resource video_ram_resource = {
235 .name = "Video RAM area",
238 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
241 static struct resource standard_io_resources[] = { {
245 .flags = IORESOURCE_BUSY | IORESOURCE_IO
250 .flags = IORESOURCE_BUSY | IORESOURCE_IO
255 .flags = IORESOURCE_BUSY | IORESOURCE_IO
260 .flags = IORESOURCE_BUSY | IORESOURCE_IO
265 .flags = IORESOURCE_BUSY | IORESOURCE_IO
267 .name = "dma page reg",
270 .flags = IORESOURCE_BUSY | IORESOURCE_IO
275 .flags = IORESOURCE_BUSY | IORESOURCE_IO
280 .flags = IORESOURCE_BUSY | IORESOURCE_IO
285 .flags = IORESOURCE_BUSY | IORESOURCE_IO
288 #define STANDARD_IO_RESOURCES \
289 (sizeof standard_io_resources / sizeof standard_io_resources[0])
291 #define romsignature(x) (*(unsigned short *)(x) == 0xaa55)
293 static int __init romchecksum(unsigned char *rom, unsigned long length)
295 unsigned char *p, sum = 0;
297 for (p = rom; p < rom + length; p++)
302 static void __init probe_roms(void)
304 unsigned long start, length, upper;
309 upper = adapter_rom_resources[0].start;
310 for (start = video_rom_resource.start; start < upper; start += 2048) {
311 rom = isa_bus_to_virt(start);
312 if (!romsignature(rom))
315 video_rom_resource.start = start;
317 /* 0 < length <= 0x7f * 512, historically */
318 length = rom[2] * 512;
320 /* if checksum okay, trust length byte */
321 if (length && romchecksum(rom, length))
322 video_rom_resource.end = start + length - 1;
324 request_resource(&iomem_resource, &video_rom_resource);
328 start = (video_rom_resource.end + 1 + 2047) & ~2047UL;
333 request_resource(&iomem_resource, &system_rom_resource);
334 upper = system_rom_resource.start;
336 /* check for extension rom (ignore length byte!) */
337 rom = isa_bus_to_virt(extension_rom_resource.start);
338 if (romsignature(rom)) {
339 length = extension_rom_resource.end - extension_rom_resource.start + 1;
340 if (romchecksum(rom, length)) {
341 request_resource(&iomem_resource, &extension_rom_resource);
342 upper = extension_rom_resource.start;
346 /* check for adapter roms on 2k boundaries */
347 for (i = 0; i < ADAPTER_ROM_RESOURCES && start < upper; start += 2048) {
348 rom = isa_bus_to_virt(start);
349 if (!romsignature(rom))
352 /* 0 < length <= 0x7f * 512, historically */
353 length = rom[2] * 512;
355 /* but accept any length that fits if checksum okay */
356 if (!length || start + length > upper || !romchecksum(rom, length))
359 adapter_rom_resources[i].start = start;
360 adapter_rom_resources[i].end = start + length - 1;
361 request_resource(&iomem_resource, &adapter_rom_resources[i]);
363 start = adapter_rom_resources[i++].end & ~2047UL;
367 static void __init limit_regions(unsigned long long size)
369 unsigned long long current_addr = 0;
373 efi_memory_desc_t *md;
376 for (p = memmap.map, i = 0; p < memmap.map_end;
377 p += memmap.desc_size, i++) {
379 current_addr = md->phys_addr + (md->num_pages << 12);
380 if (md->type == EFI_CONVENTIONAL_MEMORY) {
381 if (current_addr >= size) {
383 (((current_addr-size) + PAGE_SIZE-1) >> PAGE_SHIFT);
384 memmap.nr_map = i + 1;
390 for (i = 0; i < e820.nr_map; i++) {
391 current_addr = e820.map[i].addr + e820.map[i].size;
392 if (current_addr < size)
395 if (e820.map[i].type != E820_RAM)
398 if (e820.map[i].addr >= size) {
400 * This region starts past the end of the
401 * requested size, skip it completely.
406 e820.map[i].size -= current_addr - size;
412 void __init add_memory_region(unsigned long long start,
413 unsigned long long size, int type)
421 printk(KERN_ERR "Ooops! Too many entries in the memory map!\n");
425 e820.map[x].addr = start;
426 e820.map[x].size = size;
427 e820.map[x].type = type;
430 } /* add_memory_region */
434 static void __init print_memory_map(char *who)
438 for (i = 0; i < e820.nr_map; i++) {
439 printk(" %s: %016Lx - %016Lx ", who,
441 e820.map[i].addr + e820.map[i].size);
442 switch (e820.map[i].type) {
443 case E820_RAM: printk("(usable)\n");
446 printk("(reserved)\n");
449 printk("(ACPI data)\n");
452 printk("(ACPI NVS)\n");
454 default: printk("type %lu\n", e820.map[i].type);
461 * Sanitize the BIOS e820 map.
463 * Some e820 responses include overlapping entries. The following
464 * replaces the original e820 map with a new one, removing overlaps.
467 struct change_member {
468 struct e820entry *pbios; /* pointer to original bios entry */
469 unsigned long long addr; /* address for this change point */
471 static struct change_member change_point_list[2*E820MAX] __initdata;
472 static struct change_member *change_point[2*E820MAX] __initdata;
473 static struct e820entry *overlap_list[E820MAX] __initdata;
474 static struct e820entry new_bios[E820MAX] __initdata;
476 int __init sanitize_e820_map(struct e820entry * biosmap, char * pnr_map)
478 struct change_member *change_tmp;
479 unsigned long current_type, last_type;
480 unsigned long long last_addr;
481 int chgidx, still_changing;
484 int old_nr, new_nr, chg_nr;
488 Visually we're performing the following (1,2,3,4 = memory types)...
490 Sample memory map (w/overlaps):
491 ____22__________________
492 ______________________4_
493 ____1111________________
494 _44_____________________
495 11111111________________
496 ____________________33__
497 ___________44___________
498 __________33333_________
499 ______________22________
500 ___________________2222_
501 _________111111111______
502 _____________________11_
503 _________________4______
505 Sanitized equivalent (no overlap):
506 1_______________________
507 _44_____________________
508 ___1____________________
509 ____22__________________
510 ______11________________
511 _________1______________
512 __________3_____________
513 ___________44___________
514 _____________33_________
515 _______________2________
516 ________________1_______
517 _________________4______
518 ___________________2____
519 ____________________33__
520 ______________________4_
523 /* if there's only one memory region, don't bother */
529 /* bail out if we find any unreasonable addresses in bios map */
530 for (i=0; i<old_nr; i++)
531 if (biosmap[i].addr + biosmap[i].size < biosmap[i].addr)
534 /* create pointers for initial change-point information (for sorting) */
535 for (i=0; i < 2*old_nr; i++)
536 change_point[i] = &change_point_list[i];
538 /* record all known change-points (starting and ending addresses),
539 omitting those that are for empty memory regions */
541 for (i=0; i < old_nr; i++) {
542 if (biosmap[i].size != 0) {
543 change_point[chgidx]->addr = biosmap[i].addr;
544 change_point[chgidx++]->pbios = &biosmap[i];
545 change_point[chgidx]->addr = biosmap[i].addr + biosmap[i].size;
546 change_point[chgidx++]->pbios = &biosmap[i];
549 chg_nr = chgidx; /* true number of change-points */
551 /* sort change-point list by memory addresses (low -> high) */
553 while (still_changing) {
555 for (i=1; i < chg_nr; i++) {
556 /* if <current_addr> > <last_addr>, swap */
557 /* or, if current=<start_addr> & last=<end_addr>, swap */
558 if ((change_point[i]->addr < change_point[i-1]->addr) ||
559 ((change_point[i]->addr == change_point[i-1]->addr) &&
560 (change_point[i]->addr == change_point[i]->pbios->addr) &&
561 (change_point[i-1]->addr != change_point[i-1]->pbios->addr))
564 change_tmp = change_point[i];
565 change_point[i] = change_point[i-1];
566 change_point[i-1] = change_tmp;
572 /* create a new bios memory map, removing overlaps */
573 overlap_entries=0; /* number of entries in the overlap table */
574 new_bios_entry=0; /* index for creating new bios map entries */
575 last_type = 0; /* start with undefined memory type */
576 last_addr = 0; /* start with 0 as last starting address */
577 /* loop through change-points, determining affect on the new bios map */
578 for (chgidx=0; chgidx < chg_nr; chgidx++)
580 /* keep track of all overlapping bios entries */
581 if (change_point[chgidx]->addr == change_point[chgidx]->pbios->addr)
583 /* add map entry to overlap list (> 1 entry implies an overlap) */
584 overlap_list[overlap_entries++]=change_point[chgidx]->pbios;
588 /* remove entry from list (order independent, so swap with last) */
589 for (i=0; i<overlap_entries; i++)
591 if (overlap_list[i] == change_point[chgidx]->pbios)
592 overlap_list[i] = overlap_list[overlap_entries-1];
596 /* if there are overlapping entries, decide which "type" to use */
597 /* (larger value takes precedence -- 1=usable, 2,3,4,4+=unusable) */
599 for (i=0; i<overlap_entries; i++)
600 if (overlap_list[i]->type > current_type)
601 current_type = overlap_list[i]->type;
602 /* continue building up new bios map based on this information */
603 if (current_type != last_type) {
604 if (last_type != 0) {
605 new_bios[new_bios_entry].size =
606 change_point[chgidx]->addr - last_addr;
607 /* move forward only if the new size was non-zero */
608 if (new_bios[new_bios_entry].size != 0)
609 if (++new_bios_entry >= E820MAX)
610 break; /* no more space left for new bios entries */
612 if (current_type != 0) {
613 new_bios[new_bios_entry].addr = change_point[chgidx]->addr;
614 new_bios[new_bios_entry].type = current_type;
615 last_addr=change_point[chgidx]->addr;
617 last_type = current_type;
620 new_nr = new_bios_entry; /* retain count for new bios entries */
622 /* copy new bios mapping into original location */
623 memcpy(biosmap, new_bios, new_nr*sizeof(struct e820entry));
630 * Copy the BIOS e820 map into a safe place.
632 * Sanity-check it while we're at it..
634 * If we're lucky and live on a modern system, the setup code
635 * will have given us a memory map that we can use to properly
636 * set up memory. If we aren't, we'll fake a memory map.
638 * We check to see that the memory map contains at least 2 elements
639 * before we'll use it, because the detection code in setup.S may
640 * not be perfect and most every PC known to man has two memory
641 * regions: one from 0 to 640k, and one from 1mb up. (The IBM
642 * thinkpad 560x, for example, does not cooperate with the memory
645 int __init copy_e820_map(struct e820entry * biosmap, int nr_map)
647 /* Only one memory region (or negative)? Ignore it */
652 unsigned long long start = biosmap->addr;
653 unsigned long long size = biosmap->size;
654 unsigned long long end = start + size;
655 unsigned long type = biosmap->type;
657 /* Overflow in 64 bits? Ignore the memory map. */
662 * Some BIOSes claim RAM in the 640k - 1M region.
663 * Not right. Fix it up.
665 if (type == E820_RAM) {
666 if (start < 0x100000ULL && end > 0xA0000ULL) {
667 if (start < 0xA0000ULL)
668 add_memory_region(start, 0xA0000ULL-start, type);
669 if (end <= 0x100000ULL)
675 add_memory_region(start, size, type);
676 } while (biosmap++,--nr_map);
680 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
682 #ifdef CONFIG_EDD_MODULE
686 * copy_edd() - Copy the BIOS EDD information
687 * from boot_params into a safe place.
690 static inline void copy_edd(void)
692 memcpy(edd.mbr_signature, EDD_MBR_SIGNATURE, sizeof(edd.mbr_signature));
693 memcpy(edd.edd_info, EDD_BUF, sizeof(edd.edd_info));
694 edd.mbr_signature_nr = EDD_MBR_SIG_NR;
695 edd.edd_info_nr = EDD_NR;
698 static inline void copy_edd(void)
703 static void __init parse_cmdline_early (char ** cmdline_p)
705 char c = ' ', *to = command_line, *from = saved_command_line;
709 /* Save unparsed command line copy for /proc/cmdline */
710 saved_command_line[COMMAND_LINE_SIZE-1] = '\0';
716 * "mem=nopentium" disables the 4MB page tables.
717 * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM
718 * to <mem>, overriding the bios size.
719 * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from
720 * <start> to <start>+<mem>, overriding the bios size.
722 * HPA tells me bootloaders need to parse mem=, so no new
723 * option should be mem= [also see Documentation/i386/boot.txt]
725 if (!memcmp(from, "mem=", 4)) {
726 if (to != command_line)
728 if (!memcmp(from+4, "nopentium", 9)) {
730 clear_bit(X86_FEATURE_PSE, boot_cpu_data.x86_capability);
733 /* If the user specifies memory size, we
734 * limit the BIOS-provided memory map to
735 * that size. exactmap can be used to specify
736 * the exact map. mem=number can be used to
737 * trim the existing memory map.
739 unsigned long long mem_size;
741 mem_size = memparse(from+4, &from);
742 limit_regions(mem_size);
747 else if (!memcmp(from, "memmap=", 7)) {
748 if (to != command_line)
750 if (!memcmp(from+7, "exactmap", 8)) {
751 #ifdef CONFIG_CRASH_DUMP
752 /* If we are doing a crash dump, we
753 * still need to know the real mem
754 * size before original memory map is
758 saved_max_pfn = max_pfn;
764 /* If the user specifies memory size, we
765 * limit the BIOS-provided memory map to
766 * that size. exactmap can be used to specify
767 * the exact map. mem=number can be used to
768 * trim the existing memory map.
770 unsigned long long start_at, mem_size;
772 mem_size = memparse(from+7, &from);
774 start_at = memparse(from+1, &from);
775 add_memory_region(start_at, mem_size, E820_RAM);
776 } else if (*from == '#') {
777 start_at = memparse(from+1, &from);
778 add_memory_region(start_at, mem_size, E820_ACPI);
779 } else if (*from == '$') {
780 start_at = memparse(from+1, &from);
781 add_memory_region(start_at, mem_size, E820_RESERVED);
783 limit_regions(mem_size);
789 else if (!memcmp(from, "noexec=", 7))
790 noexec_setup(from + 7);
793 #ifdef CONFIG_X86_SMP
795 * If the BIOS enumerates physical processors before logical,
796 * maxcpus=N at enumeration-time can be used to disable HT.
798 else if (!memcmp(from, "maxcpus=", 8)) {
799 extern unsigned int maxcpus;
801 maxcpus = simple_strtoul(from + 8, NULL, 0);
806 /* "acpi=off" disables both ACPI table parsing and interpreter */
807 else if (!memcmp(from, "acpi=off", 8)) {
811 /* acpi=force to over-ride black-list */
812 else if (!memcmp(from, "acpi=force", 10)) {
818 /* acpi=strict disables out-of-spec workarounds */
819 else if (!memcmp(from, "acpi=strict", 11)) {
823 /* Limit ACPI just to boot-time to enable HT */
824 else if (!memcmp(from, "acpi=ht", 7)) {
830 /* "pci=noacpi" disable ACPI IRQ routing and PCI scan */
831 else if (!memcmp(from, "pci=noacpi", 10)) {
834 /* "acpi=noirq" disables ACPI interrupt routing */
835 else if (!memcmp(from, "acpi=noirq", 10)) {
839 else if (!memcmp(from, "acpi_sci=edge", 13))
840 acpi_sci_flags.trigger = 1;
842 else if (!memcmp(from, "acpi_sci=level", 14))
843 acpi_sci_flags.trigger = 3;
845 else if (!memcmp(from, "acpi_sci=high", 13))
846 acpi_sci_flags.polarity = 1;
848 else if (!memcmp(from, "acpi_sci=low", 12))
849 acpi_sci_flags.polarity = 3;
851 #ifdef CONFIG_X86_IO_APIC
852 else if (!memcmp(from, "acpi_skip_timer_override", 24))
853 acpi_skip_timer_override = 1;
855 if (!memcmp(from, "disable_timer_pin_1", 19))
856 disable_timer_pin_1 = 1;
857 if (!memcmp(from, "enable_timer_pin_1", 18))
858 disable_timer_pin_1 = -1;
860 /* disable IO-APIC */
861 else if (!memcmp(from, "noapic", 6))
862 disable_ioapic_setup();
863 #endif /* CONFIG_X86_IO_APIC */
864 #endif /* CONFIG_ACPI */
866 #ifdef CONFIG_X86_LOCAL_APIC
867 /* enable local APIC */
868 else if (!memcmp(from, "lapic", 5))
871 /* disable local APIC */
872 else if (!memcmp(from, "nolapic", 6))
874 #endif /* CONFIG_X86_LOCAL_APIC */
877 /* crashkernel=size@addr specifies the location to reserve for
878 * a crash kernel. By reserving this memory we guarantee
879 * that linux never set's it up as a DMA target.
880 * Useful for holding code to do something appropriate
881 * after a kernel panic.
883 else if (!memcmp(from, "crashkernel=", 12)) {
884 unsigned long size, base;
885 size = memparse(from+12, &from);
887 base = memparse(from+1, &from);
888 /* FIXME: Do I want a sanity check
889 * to validate the memory range?
891 crashk_res.start = base;
892 crashk_res.end = base + size - 1;
896 #ifdef CONFIG_PROC_VMCORE
897 /* elfcorehdr= specifies the location of elf core header
898 * stored by the crashed kernel.
900 else if (!memcmp(from, "elfcorehdr=", 11))
901 elfcorehdr_addr = memparse(from+11, &from);
905 * highmem=size forces highmem to be exactly 'size' bytes.
906 * This works even on boxes that have no highmem otherwise.
907 * This also works to reduce highmem size on bigger boxes.
909 else if (!memcmp(from, "highmem=", 8))
910 highmem_pages = memparse(from+8, &from) >> PAGE_SHIFT;
913 * vmalloc=size forces the vmalloc area to be exactly 'size'
914 * bytes. This can be used to increase (or decrease) the
915 * vmalloc area - the default is 128m.
917 else if (!memcmp(from, "vmalloc=", 8))
918 __VMALLOC_RESERVE = memparse(from+8, &from);
924 if (COMMAND_LINE_SIZE <= ++len)
929 *cmdline_p = command_line;
931 printk(KERN_INFO "user-defined physical RAM map:\n");
932 print_memory_map("user");
937 * Callback for efi_memory_walk.
940 efi_find_max_pfn(unsigned long start, unsigned long end, void *arg)
942 unsigned long *max_pfn = arg, pfn;
945 pfn = PFN_UP(end -1);
953 efi_memory_present_wrapper(unsigned long start, unsigned long end, void *arg)
955 memory_present(0, start, end);
960 * Find the highest page frame number we have available
962 void __init find_max_pfn(void)
968 efi_memmap_walk(efi_find_max_pfn, &max_pfn);
969 efi_memmap_walk(efi_memory_present_wrapper, NULL);
973 for (i = 0; i < e820.nr_map; i++) {
974 unsigned long start, end;
976 if (e820.map[i].type != E820_RAM)
978 start = PFN_UP(e820.map[i].addr);
979 end = PFN_DOWN(e820.map[i].addr + e820.map[i].size);
984 memory_present(0, start, end);
989 * Determine low and high memory ranges:
991 unsigned long __init find_max_low_pfn(void)
993 unsigned long max_low_pfn;
995 max_low_pfn = max_pfn;
996 if (max_low_pfn > MAXMEM_PFN) {
997 if (highmem_pages == -1)
998 highmem_pages = max_pfn - MAXMEM_PFN;
999 if (highmem_pages + MAXMEM_PFN < max_pfn)
1000 max_pfn = MAXMEM_PFN + highmem_pages;
1001 if (highmem_pages + MAXMEM_PFN > max_pfn) {
1002 printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn - MAXMEM_PFN), pages_to_mb(highmem_pages));
1005 max_low_pfn = MAXMEM_PFN;
1006 #ifndef CONFIG_HIGHMEM
1007 /* Maximum memory usable is what is directly addressable */
1008 printk(KERN_WARNING "Warning only %ldMB will be used.\n",
1010 if (max_pfn > MAX_NONPAE_PFN)
1011 printk(KERN_WARNING "Use a PAE enabled kernel.\n");
1013 printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n");
1014 max_pfn = MAXMEM_PFN;
1015 #else /* !CONFIG_HIGHMEM */
1016 #ifndef CONFIG_X86_PAE
1017 if (max_pfn > MAX_NONPAE_PFN) {
1018 max_pfn = MAX_NONPAE_PFN;
1019 printk(KERN_WARNING "Warning only 4GB will be used.\n");
1020 printk(KERN_WARNING "Use a PAE enabled kernel.\n");
1022 #endif /* !CONFIG_X86_PAE */
1023 #endif /* !CONFIG_HIGHMEM */
1025 if (highmem_pages == -1)
1027 #ifdef CONFIG_HIGHMEM
1028 if (highmem_pages >= max_pfn) {
1029 printk(KERN_ERR "highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages), pages_to_mb(max_pfn));
1032 if (highmem_pages) {
1033 if (max_low_pfn-highmem_pages < 64*1024*1024/PAGE_SIZE){
1034 printk(KERN_ERR "highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages));
1037 max_low_pfn -= highmem_pages;
1041 printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n");
1048 * Free all available memory for boot time allocation. Used
1049 * as a callback function by efi_memory_walk()
1053 free_available_memory(unsigned long start, unsigned long end, void *arg)
1055 /* check max_low_pfn */
1056 if (start >= (max_low_pfn << PAGE_SHIFT))
1058 if (end >= (max_low_pfn << PAGE_SHIFT))
1059 end = max_low_pfn << PAGE_SHIFT;
1061 free_bootmem(start, end - start);
1066 * Register fully available low RAM pages with the bootmem allocator.
1068 static void __init register_bootmem_low_pages(unsigned long max_low_pfn)
1073 efi_memmap_walk(free_available_memory, NULL);
1076 for (i = 0; i < e820.nr_map; i++) {
1077 unsigned long curr_pfn, last_pfn, size;
1079 * Reserve usable low memory
1081 if (e820.map[i].type != E820_RAM)
1084 * We are rounding up the start address of usable memory:
1086 curr_pfn = PFN_UP(e820.map[i].addr);
1087 if (curr_pfn >= max_low_pfn)
1090 * ... and at the end of the usable range downwards:
1092 last_pfn = PFN_DOWN(e820.map[i].addr + e820.map[i].size);
1094 if (last_pfn > max_low_pfn)
1095 last_pfn = max_low_pfn;
1098 * .. finally, did all the rounding and playing
1099 * around just make the area go away?
1101 if (last_pfn <= curr_pfn)
1104 size = last_pfn - curr_pfn;
1105 free_bootmem(PFN_PHYS(curr_pfn), PFN_PHYS(size));
1110 * workaround for Dell systems that neglect to reserve EBDA
1112 static void __init reserve_ebda_region(void)
1115 addr = get_bios_ebda();
1117 reserve_bootmem(addr, PAGE_SIZE);
1120 #ifndef CONFIG_NEED_MULTIPLE_NODES
1121 void __init setup_bootmem_allocator(void);
1122 static unsigned long __init setup_memory(void)
1125 * partially used pages are not usable - thus
1126 * we are rounding upwards:
1128 min_low_pfn = PFN_UP(init_pg_tables_end);
1132 max_low_pfn = find_max_low_pfn();
1134 #ifdef CONFIG_HIGHMEM
1135 highstart_pfn = highend_pfn = max_pfn;
1136 if (max_pfn > max_low_pfn) {
1137 highstart_pfn = max_low_pfn;
1139 printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
1140 pages_to_mb(highend_pfn - highstart_pfn));
1142 printk(KERN_NOTICE "%ldMB LOWMEM available.\n",
1143 pages_to_mb(max_low_pfn));
1145 setup_bootmem_allocator();
1150 void __init zone_sizes_init(void)
1152 unsigned long zones_size[MAX_NR_ZONES] = {0, 0, 0};
1153 unsigned int max_dma, low;
1155 max_dma = virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
1159 zones_size[ZONE_DMA] = low;
1161 zones_size[ZONE_DMA] = max_dma;
1162 zones_size[ZONE_NORMAL] = low - max_dma;
1163 #ifdef CONFIG_HIGHMEM
1164 zones_size[ZONE_HIGHMEM] = highend_pfn - low;
1167 free_area_init(zones_size);
1170 extern unsigned long __init setup_memory(void);
1171 extern void zone_sizes_init(void);
1172 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
1174 void __init setup_bootmem_allocator(void)
1176 unsigned long bootmap_size;
1178 * Initialize the boot-time allocator (with low memory only):
1180 bootmap_size = init_bootmem(min_low_pfn, max_low_pfn);
1182 register_bootmem_low_pages(max_low_pfn);
1185 * Reserve the bootmem bitmap itself as well. We do this in two
1186 * steps (first step was init_bootmem()) because this catches
1187 * the (very unlikely) case of us accidentally initializing the
1188 * bootmem allocator with an invalid RAM area.
1190 reserve_bootmem(__PHYSICAL_START, (PFN_PHYS(min_low_pfn) +
1191 bootmap_size + PAGE_SIZE-1) - (__PHYSICAL_START));
1194 * reserve physical page 0 - it's a special BIOS page on many boxes,
1195 * enabling clean reboots, SMP operation, laptop functions.
1197 reserve_bootmem(0, PAGE_SIZE);
1199 /* reserve EBDA region, it's a 4K region */
1200 reserve_ebda_region();
1202 /* could be an AMD 768MPX chipset. Reserve a page before VGA to prevent
1203 PCI prefetch into it (errata #56). Usually the page is reserved anyways,
1204 unless you have no PS/2 mouse plugged in. */
1205 if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
1206 boot_cpu_data.x86 == 6)
1207 reserve_bootmem(0xa0000 - 4096, 4096);
1211 * But first pinch a few for the stack/trampoline stuff
1212 * FIXME: Don't need the extra page at 4K, but need to fix
1213 * trampoline before removing it. (see the GDT stuff)
1215 reserve_bootmem(PAGE_SIZE, PAGE_SIZE);
1217 #ifdef CONFIG_ACPI_SLEEP
1219 * Reserve low memory region for sleep support.
1221 acpi_reserve_bootmem();
1223 #ifdef CONFIG_X86_FIND_SMP_CONFIG
1225 * Find and reserve possible boot-time SMP configuration:
1230 #ifdef CONFIG_BLK_DEV_INITRD
1231 if (LOADER_TYPE && INITRD_START) {
1232 if (INITRD_START + INITRD_SIZE <= (max_low_pfn << PAGE_SHIFT)) {
1233 reserve_bootmem(INITRD_START, INITRD_SIZE);
1235 INITRD_START ? INITRD_START + PAGE_OFFSET : 0;
1236 initrd_end = initrd_start+INITRD_SIZE;
1239 printk(KERN_ERR "initrd extends beyond end of memory "
1240 "(0x%08lx > 0x%08lx)\ndisabling initrd\n",
1241 INITRD_START + INITRD_SIZE,
1242 max_low_pfn << PAGE_SHIFT);
1248 if (crashk_res.start != crashk_res.end)
1249 reserve_bootmem(crashk_res.start,
1250 crashk_res.end - crashk_res.start + 1);
1255 * The node 0 pgdat is initialized before all of these because
1256 * it's needed for bootmem. node>0 pgdats have their virtual
1257 * space allocated before the pagetables are in place to access
1258 * them, so they can't be cleared then.
1260 * This should all compile down to nothing when NUMA is off.
1262 void __init remapped_pgdat_init(void)
1266 for_each_online_node(nid) {
1268 memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
1273 * Request address space for all standard RAM and ROM resources
1274 * and also for regions reported as reserved by the e820.
1277 legacy_init_iomem_resources(struct resource *code_resource, struct resource *data_resource)
1282 for (i = 0; i < e820.nr_map; i++) {
1283 struct resource *res;
1284 #ifndef CONFIG_RESOURCES_64BIT
1285 if (e820.map[i].addr + e820.map[i].size > 0x100000000ULL)
1288 res = kzalloc(sizeof(struct resource), GFP_ATOMIC);
1289 switch (e820.map[i].type) {
1290 case E820_RAM: res->name = "System RAM"; break;
1291 case E820_ACPI: res->name = "ACPI Tables"; break;
1292 case E820_NVS: res->name = "ACPI Non-volatile Storage"; break;
1293 default: res->name = "reserved";
1295 res->start = e820.map[i].addr;
1296 res->end = res->start + e820.map[i].size - 1;
1297 res->flags = IORESOURCE_MEM | IORESOURCE_BUSY;
1298 if (request_resource(&iomem_resource, res)) {
1302 if (e820.map[i].type == E820_RAM) {
1304 * We don't know which RAM region contains kernel data,
1305 * so we try it repeatedly and let the resource manager
1308 request_resource(res, code_resource);
1309 request_resource(res, data_resource);
1311 request_resource(res, &crashk_res);
1318 * Request address space for all standard resources
1320 * This is called just before pcibios_init(), which is also a
1321 * subsys_initcall, but is linked in later (in arch/i386/pci/common.c).
1323 static int __init request_standard_resources(void)
1327 printk("Setting up standard PCI resources\n");
1329 efi_initialize_iomem_resources(&code_resource, &data_resource);
1331 legacy_init_iomem_resources(&code_resource, &data_resource);
1333 /* EFI systems may still have VGA */
1334 request_resource(&iomem_resource, &video_ram_resource);
1336 /* request I/O space for devices used on all i[345]86 PCs */
1337 for (i = 0; i < STANDARD_IO_RESOURCES; i++)
1338 request_resource(&ioport_resource, &standard_io_resources[i]);
1342 subsys_initcall(request_standard_resources);
1344 static void __init register_memory(void)
1346 unsigned long gapstart, gapsize, round;
1347 unsigned long long last;
1351 * Search for the bigest gap in the low 32 bits of the e820
1354 last = 0x100000000ull;
1355 gapstart = 0x10000000;
1359 unsigned long long start = e820.map[i].addr;
1360 unsigned long long end = start + e820.map[i].size;
1363 * Since "last" is at most 4GB, we know we'll
1364 * fit in 32 bits if this condition is true
1367 unsigned long gap = last - end;
1369 if (gap > gapsize) {
1379 * See how much we want to round up: start off with
1380 * rounding to the next 1MB area.
1383 while ((gapsize >> 4) > round)
1385 /* Fun with two's complement */
1386 pci_mem_start = (gapstart + round) & -round;
1388 printk("Allocating PCI resources starting at %08lx (gap: %08lx:%08lx)\n",
1389 pci_mem_start, gapstart, gapsize);
1393 static void set_mca_bus(int x)
1398 static void set_mca_bus(int x) { }
1402 * Determine if we were loaded by an EFI loader. If so, then we have also been
1403 * passed the efi memmap, systab, etc., so we should use these data structures
1404 * for initialization. Note, the efi init code path is determined by the
1405 * global efi_enabled. This allows the same kernel image to be used on existing
1406 * systems (with a traditional BIOS) as well as on EFI systems.
1408 void __init setup_arch(char **cmdline_p)
1410 unsigned long max_low_pfn;
1412 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
1413 pre_setup_arch_hook();
1417 * FIXME: This isn't an official loader_type right
1418 * now but does currently work with elilo.
1419 * If we were configured as an EFI kernel, check to make
1420 * sure that we were loaded correctly from elilo and that
1421 * the system table is valid. If not, then initialize normally.
1424 if ((LOADER_TYPE == 0x50) && EFI_SYSTAB)
1428 ROOT_DEV = old_decode_dev(ORIG_ROOT_DEV);
1429 drive_info = DRIVE_INFO;
1430 screen_info = SCREEN_INFO;
1431 edid_info = EDID_INFO;
1432 apm_info.bios = APM_BIOS_INFO;
1433 ist_info = IST_INFO;
1434 saved_videomode = VIDEO_MODE;
1435 if( SYS_DESC_TABLE.length != 0 ) {
1436 set_mca_bus(SYS_DESC_TABLE.table[3] & 0x2);
1437 machine_id = SYS_DESC_TABLE.table[0];
1438 machine_submodel_id = SYS_DESC_TABLE.table[1];
1439 BIOS_revision = SYS_DESC_TABLE.table[2];
1441 bootloader_type = LOADER_TYPE;
1443 #ifdef CONFIG_BLK_DEV_RAM
1444 rd_image_start = RAMDISK_FLAGS & RAMDISK_IMAGE_START_MASK;
1445 rd_prompt = ((RAMDISK_FLAGS & RAMDISK_PROMPT_FLAG) != 0);
1446 rd_doload = ((RAMDISK_FLAGS & RAMDISK_LOAD_FLAG) != 0);
1452 printk(KERN_INFO "BIOS-provided physical RAM map:\n");
1453 print_memory_map(machine_specific_memory_setup());
1458 if (!MOUNT_ROOT_RDONLY)
1459 root_mountflags &= ~MS_RDONLY;
1460 init_mm.start_code = (unsigned long) _text;
1461 init_mm.end_code = (unsigned long) _etext;
1462 init_mm.end_data = (unsigned long) _edata;
1463 init_mm.brk = init_pg_tables_end + PAGE_OFFSET;
1465 code_resource.start = virt_to_phys(_text);
1466 code_resource.end = virt_to_phys(_etext)-1;
1467 data_resource.start = virt_to_phys(_etext);
1468 data_resource.end = virt_to_phys(_edata)-1;
1470 parse_cmdline_early(cmdline_p);
1472 #ifdef CONFIG_EARLY_PRINTK
1474 char *s = strstr(*cmdline_p, "earlyprintk=");
1476 setup_early_printk(strchr(s, '=') + 1);
1477 printk("early console enabled\n");
1482 max_low_pfn = setup_memory();
1485 * NOTE: before this point _nobody_ is allowed to allocate
1486 * any memory using the bootmem allocator. Although the
1487 * alloctor is now initialised only the first 8Mb of the kernel
1488 * virtual address space has been mapped. All allocations before
1489 * paging_init() has completed must use the alloc_bootmem_low_pages()
1490 * variant (which allocates DMA'able memory) and care must be taken
1491 * not to exceed the 8Mb limit.
1495 smp_alloc_memory(); /* AP processor realmode stacks in low memory*/
1498 remapped_pgdat_init();
1503 * NOTE: at this point the bootmem allocator is fully available.
1508 #ifdef CONFIG_X86_GENERICARCH
1509 generic_apic_probe(*cmdline_p);
1516 * Parse the ACPI tables for possible boot-time SMP configuration.
1518 acpi_boot_table_init();
1521 #ifdef CONFIG_X86_IO_APIC
1522 check_acpi_pci(); /* Checks more than just ACPI actually */
1528 #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC)
1530 printk(KERN_WARNING "More than 8 CPUs detected and "
1531 "CONFIG_X86_PC cannot handle it.\nUse "
1532 "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
1535 #ifdef CONFIG_X86_LOCAL_APIC
1536 if (smp_found_config)
1543 #if defined(CONFIG_VGA_CONSOLE)
1544 if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1545 conswitchp = &vga_con;
1546 #elif defined(CONFIG_DUMMY_CONSOLE)
1547 conswitchp = &dummy_con;
1553 static __init int add_pcspkr(void)
1555 struct platform_device *pd;
1558 pd = platform_device_alloc("pcspkr", -1);
1562 ret = platform_device_add(pd);
1564 platform_device_put(pd);
1568 device_initcall(add_pcspkr);
1573 * c-file-style:"k&r"