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/config.h>
27 #include <linux/sched.h>
29 #include <linux/mmzone.h>
30 #include <linux/tty.h>
31 #include <linux/ioport.h>
32 #include <linux/acpi.h>
33 #include <linux/apm_bios.h>
34 #include <linux/initrd.h>
35 #include <linux/bootmem.h>
36 #include <linux/seq_file.h>
37 #include <linux/platform_device.h>
38 #include <linux/console.h>
39 #include <linux/mca.h>
40 #include <linux/root_dev.h>
41 #include <linux/highmem.h>
42 #include <linux/module.h>
43 #include <linux/efi.h>
44 #include <linux/init.h>
45 #include <linux/edd.h>
46 #include <linux/nodemask.h>
47 #include <linux/kexec.h>
48 #include <linux/crash_dump.h>
49 #include <linux/dmi.h>
50 #include <linux/pfn.h>
52 #include <video/edid.h>
56 #include <asm/mpspec.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
;
94 int acpi_disabled
= 0;
96 int acpi_disabled
= 1;
98 EXPORT_SYMBOL(acpi_disabled
);
101 int __initdata acpi_force
= 0;
102 extern acpi_interrupt_flags acpi_sci_flags
;
105 /* for MCA, but anyone else can use it if they want */
106 unsigned int machine_id
;
108 EXPORT_SYMBOL(machine_id
);
110 unsigned int machine_submodel_id
;
111 unsigned int BIOS_revision
;
112 unsigned int mca_pentium_flag
;
114 /* For PCI or other memory-mapped resources */
115 unsigned long pci_mem_start
= 0x10000000;
117 EXPORT_SYMBOL(pci_mem_start
);
120 /* Boot loader ID as an integer, for the benefit of proc_dointvec */
123 /* user-defined highmem size */
124 static unsigned int highmem_pages
= -1;
129 struct drive_info_struct
{ char dummy
[32]; } drive_info
;
130 #if defined(CONFIG_BLK_DEV_IDE) || defined(CONFIG_BLK_DEV_HD) || \
131 defined(CONFIG_BLK_DEV_IDE_MODULE) || defined(CONFIG_BLK_DEV_HD_MODULE)
132 EXPORT_SYMBOL(drive_info
);
134 struct screen_info screen_info
;
135 EXPORT_SYMBOL(screen_info
);
136 struct apm_info apm_info
;
137 EXPORT_SYMBOL(apm_info
);
138 struct sys_desc_table_struct
{
139 unsigned short length
;
140 unsigned char table
[0];
142 struct edid_info edid_info
;
143 EXPORT_SYMBOL_GPL(edid_info
);
144 struct ist_info ist_info
;
145 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
146 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
147 EXPORT_SYMBOL(ist_info
);
151 extern void early_cpu_init(void);
152 extern void generic_apic_probe(char *);
153 extern int root_mountflags
;
155 unsigned long saved_videomode
;
157 #define RAMDISK_IMAGE_START_MASK 0x07FF
158 #define RAMDISK_PROMPT_FLAG 0x8000
159 #define RAMDISK_LOAD_FLAG 0x4000
161 static char command_line
[COMMAND_LINE_SIZE
];
163 unsigned char __initdata boot_params
[PARAM_SIZE
];
165 static struct resource data_resource
= {
166 .name
= "Kernel data",
169 .flags
= IORESOURCE_BUSY
| IORESOURCE_MEM
172 static struct resource code_resource
= {
173 .name
= "Kernel code",
176 .flags
= IORESOURCE_BUSY
| IORESOURCE_MEM
179 static struct resource system_rom_resource
= {
180 .name
= "System ROM",
183 .flags
= IORESOURCE_BUSY
| IORESOURCE_READONLY
| IORESOURCE_MEM
186 static struct resource extension_rom_resource
= {
187 .name
= "Extension ROM",
190 .flags
= IORESOURCE_BUSY
| IORESOURCE_READONLY
| IORESOURCE_MEM
193 static struct resource adapter_rom_resources
[] = { {
194 .name
= "Adapter ROM",
197 .flags
= IORESOURCE_BUSY
| IORESOURCE_READONLY
| IORESOURCE_MEM
199 .name
= "Adapter ROM",
202 .flags
= IORESOURCE_BUSY
| IORESOURCE_READONLY
| IORESOURCE_MEM
204 .name
= "Adapter ROM",
207 .flags
= IORESOURCE_BUSY
| IORESOURCE_READONLY
| IORESOURCE_MEM
209 .name
= "Adapter ROM",
212 .flags
= IORESOURCE_BUSY
| IORESOURCE_READONLY
| IORESOURCE_MEM
214 .name
= "Adapter ROM",
217 .flags
= IORESOURCE_BUSY
| IORESOURCE_READONLY
| IORESOURCE_MEM
219 .name
= "Adapter ROM",
222 .flags
= IORESOURCE_BUSY
| IORESOURCE_READONLY
| IORESOURCE_MEM
225 #define ADAPTER_ROM_RESOURCES \
226 (sizeof adapter_rom_resources / sizeof adapter_rom_resources[0])
228 static struct resource video_rom_resource
= {
232 .flags
= IORESOURCE_BUSY
| IORESOURCE_READONLY
| IORESOURCE_MEM
235 static struct resource video_ram_resource
= {
236 .name
= "Video RAM area",
239 .flags
= IORESOURCE_BUSY
| IORESOURCE_MEM
242 static struct resource standard_io_resources
[] = { {
246 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
251 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
256 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
261 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
266 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
268 .name
= "dma page reg",
271 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
276 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
281 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
286 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
289 #define STANDARD_IO_RESOURCES \
290 (sizeof standard_io_resources / sizeof standard_io_resources[0])
292 #define romsignature(x) (*(unsigned short *)(x) == 0xaa55)
294 static int __init
romchecksum(unsigned char *rom
, unsigned long length
)
296 unsigned char *p
, sum
= 0;
298 for (p
= rom
; p
< rom
+ length
; p
++)
303 static void __init
probe_roms(void)
305 unsigned long start
, length
, upper
;
310 upper
= adapter_rom_resources
[0].start
;
311 for (start
= video_rom_resource
.start
; start
< upper
; start
+= 2048) {
312 rom
= isa_bus_to_virt(start
);
313 if (!romsignature(rom
))
316 video_rom_resource
.start
= start
;
318 /* 0 < length <= 0x7f * 512, historically */
319 length
= rom
[2] * 512;
321 /* if checksum okay, trust length byte */
322 if (length
&& romchecksum(rom
, length
))
323 video_rom_resource
.end
= start
+ length
- 1;
325 request_resource(&iomem_resource
, &video_rom_resource
);
329 start
= (video_rom_resource
.end
+ 1 + 2047) & ~2047UL;
334 request_resource(&iomem_resource
, &system_rom_resource
);
335 upper
= system_rom_resource
.start
;
337 /* check for extension rom (ignore length byte!) */
338 rom
= isa_bus_to_virt(extension_rom_resource
.start
);
339 if (romsignature(rom
)) {
340 length
= extension_rom_resource
.end
- extension_rom_resource
.start
+ 1;
341 if (romchecksum(rom
, length
)) {
342 request_resource(&iomem_resource
, &extension_rom_resource
);
343 upper
= extension_rom_resource
.start
;
347 /* check for adapter roms on 2k boundaries */
348 for (i
= 0; i
< ADAPTER_ROM_RESOURCES
&& start
< upper
; start
+= 2048) {
349 rom
= isa_bus_to_virt(start
);
350 if (!romsignature(rom
))
353 /* 0 < length <= 0x7f * 512, historically */
354 length
= rom
[2] * 512;
356 /* but accept any length that fits if checksum okay */
357 if (!length
|| start
+ length
> upper
|| !romchecksum(rom
, length
))
360 adapter_rom_resources
[i
].start
= start
;
361 adapter_rom_resources
[i
].end
= start
+ length
- 1;
362 request_resource(&iomem_resource
, &adapter_rom_resources
[i
]);
364 start
= adapter_rom_resources
[i
++].end
& ~2047UL;
368 static void __init
limit_regions(unsigned long long size
)
370 unsigned long long current_addr
= 0;
374 efi_memory_desc_t
*md
;
377 for (p
= memmap
.map
, i
= 0; p
< memmap
.map_end
;
378 p
+= memmap
.desc_size
, i
++) {
380 current_addr
= md
->phys_addr
+ (md
->num_pages
<< 12);
381 if (md
->type
== EFI_CONVENTIONAL_MEMORY
) {
382 if (current_addr
>= size
) {
384 (((current_addr
-size
) + PAGE_SIZE
-1) >> PAGE_SHIFT
);
385 memmap
.nr_map
= i
+ 1;
391 for (i
= 0; i
< e820
.nr_map
; i
++) {
392 current_addr
= e820
.map
[i
].addr
+ e820
.map
[i
].size
;
393 if (current_addr
< size
)
396 if (e820
.map
[i
].type
!= E820_RAM
)
399 if (e820
.map
[i
].addr
>= size
) {
401 * This region starts past the end of the
402 * requested size, skip it completely.
407 e820
.map
[i
].size
-= current_addr
- size
;
413 void __init
add_memory_region(unsigned long long start
,
414 unsigned long long size
, int type
)
422 printk(KERN_ERR
"Ooops! Too many entries in the memory map!\n");
426 e820
.map
[x
].addr
= start
;
427 e820
.map
[x
].size
= size
;
428 e820
.map
[x
].type
= type
;
431 } /* add_memory_region */
435 static void __init
print_memory_map(char *who
)
439 for (i
= 0; i
< e820
.nr_map
; i
++) {
440 printk(" %s: %016Lx - %016Lx ", who
,
442 e820
.map
[i
].addr
+ e820
.map
[i
].size
);
443 switch (e820
.map
[i
].type
) {
444 case E820_RAM
: printk("(usable)\n");
447 printk("(reserved)\n");
450 printk("(ACPI data)\n");
453 printk("(ACPI NVS)\n");
455 default: printk("type %lu\n", e820
.map
[i
].type
);
462 * Sanitize the BIOS e820 map.
464 * Some e820 responses include overlapping entries. The following
465 * replaces the original e820 map with a new one, removing overlaps.
468 struct change_member
{
469 struct e820entry
*pbios
; /* pointer to original bios entry */
470 unsigned long long addr
; /* address for this change point */
472 static struct change_member change_point_list
[2*E820MAX
] __initdata
;
473 static struct change_member
*change_point
[2*E820MAX
] __initdata
;
474 static struct e820entry
*overlap_list
[E820MAX
] __initdata
;
475 static struct e820entry new_bios
[E820MAX
] __initdata
;
477 int __init
sanitize_e820_map(struct e820entry
* biosmap
, char * pnr_map
)
479 struct change_member
*change_tmp
;
480 unsigned long current_type
, last_type
;
481 unsigned long long last_addr
;
482 int chgidx
, still_changing
;
485 int old_nr
, new_nr
, chg_nr
;
489 Visually we're performing the following (1,2,3,4 = memory types)...
491 Sample memory map (w/overlaps):
492 ____22__________________
493 ______________________4_
494 ____1111________________
495 _44_____________________
496 11111111________________
497 ____________________33__
498 ___________44___________
499 __________33333_________
500 ______________22________
501 ___________________2222_
502 _________111111111______
503 _____________________11_
504 _________________4______
506 Sanitized equivalent (no overlap):
507 1_______________________
508 _44_____________________
509 ___1____________________
510 ____22__________________
511 ______11________________
512 _________1______________
513 __________3_____________
514 ___________44___________
515 _____________33_________
516 _______________2________
517 ________________1_______
518 _________________4______
519 ___________________2____
520 ____________________33__
521 ______________________4_
524 /* if there's only one memory region, don't bother */
530 /* bail out if we find any unreasonable addresses in bios map */
531 for (i
=0; i
<old_nr
; i
++)
532 if (biosmap
[i
].addr
+ biosmap
[i
].size
< biosmap
[i
].addr
)
535 /* create pointers for initial change-point information (for sorting) */
536 for (i
=0; i
< 2*old_nr
; i
++)
537 change_point
[i
] = &change_point_list
[i
];
539 /* record all known change-points (starting and ending addresses),
540 omitting those that are for empty memory regions */
542 for (i
=0; i
< old_nr
; i
++) {
543 if (biosmap
[i
].size
!= 0) {
544 change_point
[chgidx
]->addr
= biosmap
[i
].addr
;
545 change_point
[chgidx
++]->pbios
= &biosmap
[i
];
546 change_point
[chgidx
]->addr
= biosmap
[i
].addr
+ biosmap
[i
].size
;
547 change_point
[chgidx
++]->pbios
= &biosmap
[i
];
550 chg_nr
= chgidx
; /* true number of change-points */
552 /* sort change-point list by memory addresses (low -> high) */
554 while (still_changing
) {
556 for (i
=1; i
< chg_nr
; i
++) {
557 /* if <current_addr> > <last_addr>, swap */
558 /* or, if current=<start_addr> & last=<end_addr>, swap */
559 if ((change_point
[i
]->addr
< change_point
[i
-1]->addr
) ||
560 ((change_point
[i
]->addr
== change_point
[i
-1]->addr
) &&
561 (change_point
[i
]->addr
== change_point
[i
]->pbios
->addr
) &&
562 (change_point
[i
-1]->addr
!= change_point
[i
-1]->pbios
->addr
))
565 change_tmp
= change_point
[i
];
566 change_point
[i
] = change_point
[i
-1];
567 change_point
[i
-1] = change_tmp
;
573 /* create a new bios memory map, removing overlaps */
574 overlap_entries
=0; /* number of entries in the overlap table */
575 new_bios_entry
=0; /* index for creating new bios map entries */
576 last_type
= 0; /* start with undefined memory type */
577 last_addr
= 0; /* start with 0 as last starting address */
578 /* loop through change-points, determining affect on the new bios map */
579 for (chgidx
=0; chgidx
< chg_nr
; chgidx
++)
581 /* keep track of all overlapping bios entries */
582 if (change_point
[chgidx
]->addr
== change_point
[chgidx
]->pbios
->addr
)
584 /* add map entry to overlap list (> 1 entry implies an overlap) */
585 overlap_list
[overlap_entries
++]=change_point
[chgidx
]->pbios
;
589 /* remove entry from list (order independent, so swap with last) */
590 for (i
=0; i
<overlap_entries
; i
++)
592 if (overlap_list
[i
] == change_point
[chgidx
]->pbios
)
593 overlap_list
[i
] = overlap_list
[overlap_entries
-1];
597 /* if there are overlapping entries, decide which "type" to use */
598 /* (larger value takes precedence -- 1=usable, 2,3,4,4+=unusable) */
600 for (i
=0; i
<overlap_entries
; i
++)
601 if (overlap_list
[i
]->type
> current_type
)
602 current_type
= overlap_list
[i
]->type
;
603 /* continue building up new bios map based on this information */
604 if (current_type
!= last_type
) {
605 if (last_type
!= 0) {
606 new_bios
[new_bios_entry
].size
=
607 change_point
[chgidx
]->addr
- last_addr
;
608 /* move forward only if the new size was non-zero */
609 if (new_bios
[new_bios_entry
].size
!= 0)
610 if (++new_bios_entry
>= E820MAX
)
611 break; /* no more space left for new bios entries */
613 if (current_type
!= 0) {
614 new_bios
[new_bios_entry
].addr
= change_point
[chgidx
]->addr
;
615 new_bios
[new_bios_entry
].type
= current_type
;
616 last_addr
=change_point
[chgidx
]->addr
;
618 last_type
= current_type
;
621 new_nr
= new_bios_entry
; /* retain count for new bios entries */
623 /* copy new bios mapping into original location */
624 memcpy(biosmap
, new_bios
, new_nr
*sizeof(struct e820entry
));
631 * Copy the BIOS e820 map into a safe place.
633 * Sanity-check it while we're at it..
635 * If we're lucky and live on a modern system, the setup code
636 * will have given us a memory map that we can use to properly
637 * set up memory. If we aren't, we'll fake a memory map.
639 * We check to see that the memory map contains at least 2 elements
640 * before we'll use it, because the detection code in setup.S may
641 * not be perfect and most every PC known to man has two memory
642 * regions: one from 0 to 640k, and one from 1mb up. (The IBM
643 * thinkpad 560x, for example, does not cooperate with the memory
646 int __init
copy_e820_map(struct e820entry
* biosmap
, int nr_map
)
648 /* Only one memory region (or negative)? Ignore it */
653 unsigned long long start
= biosmap
->addr
;
654 unsigned long long size
= biosmap
->size
;
655 unsigned long long end
= start
+ size
;
656 unsigned long type
= biosmap
->type
;
658 /* Overflow in 64 bits? Ignore the memory map. */
663 * Some BIOSes claim RAM in the 640k - 1M region.
664 * Not right. Fix it up.
666 if (type
== E820_RAM
) {
667 if (start
< 0x100000ULL
&& end
> 0xA0000ULL
) {
668 if (start
< 0xA0000ULL
)
669 add_memory_region(start
, 0xA0000ULL
-start
, type
);
670 if (end
<= 0x100000ULL
)
676 add_memory_region(start
, size
, type
);
677 } while (biosmap
++,--nr_map
);
681 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
683 #ifdef CONFIG_EDD_MODULE
687 * copy_edd() - Copy the BIOS EDD information
688 * from boot_params into a safe place.
691 static inline void copy_edd(void)
693 memcpy(edd
.mbr_signature
, EDD_MBR_SIGNATURE
, sizeof(edd
.mbr_signature
));
694 memcpy(edd
.edd_info
, EDD_BUF
, sizeof(edd
.edd_info
));
695 edd
.mbr_signature_nr
= EDD_MBR_SIG_NR
;
696 edd
.edd_info_nr
= EDD_NR
;
699 static inline void copy_edd(void)
704 static void __init
parse_cmdline_early (char ** cmdline_p
)
706 char c
= ' ', *to
= command_line
, *from
= saved_command_line
;
710 /* Save unparsed command line copy for /proc/cmdline */
711 saved_command_line
[COMMAND_LINE_SIZE
-1] = '\0';
717 * "mem=nopentium" disables the 4MB page tables.
718 * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM
719 * to <mem>, overriding the bios size.
720 * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from
721 * <start> to <start>+<mem>, overriding the bios size.
723 * HPA tells me bootloaders need to parse mem=, so no new
724 * option should be mem= [also see Documentation/i386/boot.txt]
726 if (!memcmp(from
, "mem=", 4)) {
727 if (to
!= command_line
)
729 if (!memcmp(from
+4, "nopentium", 9)) {
731 clear_bit(X86_FEATURE_PSE
, boot_cpu_data
.x86_capability
);
734 /* If the user specifies memory size, we
735 * limit the BIOS-provided memory map to
736 * that size. exactmap can be used to specify
737 * the exact map. mem=number can be used to
738 * trim the existing memory map.
740 unsigned long long mem_size
;
742 mem_size
= memparse(from
+4, &from
);
743 limit_regions(mem_size
);
748 else if (!memcmp(from
, "memmap=", 7)) {
749 if (to
!= command_line
)
751 if (!memcmp(from
+7, "exactmap", 8)) {
752 #ifdef CONFIG_CRASH_DUMP
753 /* If we are doing a crash dump, we
754 * still need to know the real mem
755 * size before original memory map is
759 saved_max_pfn
= max_pfn
;
765 /* If the user specifies memory size, we
766 * limit the BIOS-provided memory map to
767 * that size. exactmap can be used to specify
768 * the exact map. mem=number can be used to
769 * trim the existing memory map.
771 unsigned long long start_at
, mem_size
;
773 mem_size
= memparse(from
+7, &from
);
775 start_at
= memparse(from
+1, &from
);
776 add_memory_region(start_at
, mem_size
, E820_RAM
);
777 } else if (*from
== '#') {
778 start_at
= memparse(from
+1, &from
);
779 add_memory_region(start_at
, mem_size
, E820_ACPI
);
780 } else if (*from
== '$') {
781 start_at
= memparse(from
+1, &from
);
782 add_memory_region(start_at
, mem_size
, E820_RESERVED
);
784 limit_regions(mem_size
);
790 else if (!memcmp(from
, "noexec=", 7))
791 noexec_setup(from
+ 7);
794 #ifdef CONFIG_X86_SMP
796 * If the BIOS enumerates physical processors before logical,
797 * maxcpus=N at enumeration-time can be used to disable HT.
799 else if (!memcmp(from
, "maxcpus=", 8)) {
800 extern unsigned int maxcpus
;
802 maxcpus
= simple_strtoul(from
+ 8, NULL
, 0);
807 /* "acpi=off" disables both ACPI table parsing and interpreter */
808 else if (!memcmp(from
, "acpi=off", 8)) {
812 /* acpi=force to over-ride black-list */
813 else if (!memcmp(from
, "acpi=force", 10)) {
819 /* acpi=strict disables out-of-spec workarounds */
820 else if (!memcmp(from
, "acpi=strict", 11)) {
824 /* Limit ACPI just to boot-time to enable HT */
825 else if (!memcmp(from
, "acpi=ht", 7)) {
831 /* "pci=noacpi" disable ACPI IRQ routing and PCI scan */
832 else if (!memcmp(from
, "pci=noacpi", 10)) {
835 /* "acpi=noirq" disables ACPI interrupt routing */
836 else if (!memcmp(from
, "acpi=noirq", 10)) {
840 else if (!memcmp(from
, "acpi_sci=edge", 13))
841 acpi_sci_flags
.trigger
= 1;
843 else if (!memcmp(from
, "acpi_sci=level", 14))
844 acpi_sci_flags
.trigger
= 3;
846 else if (!memcmp(from
, "acpi_sci=high", 13))
847 acpi_sci_flags
.polarity
= 1;
849 else if (!memcmp(from
, "acpi_sci=low", 12))
850 acpi_sci_flags
.polarity
= 3;
852 #ifdef CONFIG_X86_IO_APIC
853 else if (!memcmp(from
, "acpi_skip_timer_override", 24))
854 acpi_skip_timer_override
= 1;
856 if (!memcmp(from
, "disable_timer_pin_1", 19))
857 disable_timer_pin_1
= 1;
858 if (!memcmp(from
, "enable_timer_pin_1", 18))
859 disable_timer_pin_1
= -1;
861 /* disable IO-APIC */
862 else if (!memcmp(from
, "noapic", 6))
863 disable_ioapic_setup();
864 #endif /* CONFIG_X86_IO_APIC */
865 #endif /* CONFIG_ACPI */
867 #ifdef CONFIG_X86_LOCAL_APIC
868 /* enable local APIC */
869 else if (!memcmp(from
, "lapic", 5))
872 /* disable local APIC */
873 else if (!memcmp(from
, "nolapic", 6))
875 #endif /* CONFIG_X86_LOCAL_APIC */
878 /* crashkernel=size@addr specifies the location to reserve for
879 * a crash kernel. By reserving this memory we guarantee
880 * that linux never set's it up as a DMA target.
881 * Useful for holding code to do something appropriate
882 * after a kernel panic.
884 else if (!memcmp(from
, "crashkernel=", 12)) {
885 unsigned long size
, base
;
886 size
= memparse(from
+12, &from
);
888 base
= memparse(from
+1, &from
);
889 /* FIXME: Do I want a sanity check
890 * to validate the memory range?
892 crashk_res
.start
= base
;
893 crashk_res
.end
= base
+ size
- 1;
897 #ifdef CONFIG_PROC_VMCORE
898 /* elfcorehdr= specifies the location of elf core header
899 * stored by the crashed kernel.
901 else if (!memcmp(from
, "elfcorehdr=", 11))
902 elfcorehdr_addr
= memparse(from
+11, &from
);
906 * highmem=size forces highmem to be exactly 'size' bytes.
907 * This works even on boxes that have no highmem otherwise.
908 * This also works to reduce highmem size on bigger boxes.
910 else if (!memcmp(from
, "highmem=", 8))
911 highmem_pages
= memparse(from
+8, &from
) >> PAGE_SHIFT
;
914 * vmalloc=size forces the vmalloc area to be exactly 'size'
915 * bytes. This can be used to increase (or decrease) the
916 * vmalloc area - the default is 128m.
918 else if (!memcmp(from
, "vmalloc=", 8))
919 __VMALLOC_RESERVE
= memparse(from
+8, &from
);
925 if (COMMAND_LINE_SIZE
<= ++len
)
930 *cmdline_p
= command_line
;
932 printk(KERN_INFO
"user-defined physical RAM map:\n");
933 print_memory_map("user");
938 * Callback for efi_memory_walk.
941 efi_find_max_pfn(unsigned long start
, unsigned long end
, void *arg
)
943 unsigned long *max_pfn
= arg
, pfn
;
946 pfn
= PFN_UP(end
-1);
954 efi_memory_present_wrapper(unsigned long start
, unsigned long end
, void *arg
)
956 memory_present(0, start
, end
);
961 * This function checks if the entire range <start,end> is mapped with type.
963 * Note: this function only works correct if the e820 table is sorted and
964 * not-overlapping, which is the case
967 e820_all_mapped(unsigned long s
, unsigned long e
, unsigned type
)
972 for (i
= 0; i
< e820
.nr_map
; i
++) {
973 struct e820entry
*ei
= &e820
.map
[i
];
974 if (type
&& ei
->type
!= type
)
976 /* is the region (part) in overlap with the current region ?*/
977 if (ei
->addr
>= end
|| ei
->addr
+ ei
->size
<= start
)
979 /* if the region is at the beginning of <start,end> we move
980 * start to the end of the region since it's ok until there
982 if (ei
->addr
<= start
)
983 start
= ei
->addr
+ ei
->size
;
984 /* if start is now at or beyond end, we're done, full
987 return 1; /* we're done */
993 * Find the highest page frame number we have available
995 void __init
find_max_pfn(void)
1001 efi_memmap_walk(efi_find_max_pfn
, &max_pfn
);
1002 efi_memmap_walk(efi_memory_present_wrapper
, NULL
);
1006 for (i
= 0; i
< e820
.nr_map
; i
++) {
1007 unsigned long start
, end
;
1009 if (e820
.map
[i
].type
!= E820_RAM
)
1011 start
= PFN_UP(e820
.map
[i
].addr
);
1012 end
= PFN_DOWN(e820
.map
[i
].addr
+ e820
.map
[i
].size
);
1017 memory_present(0, start
, end
);
1022 * Determine low and high memory ranges:
1024 unsigned long __init
find_max_low_pfn(void)
1026 unsigned long max_low_pfn
;
1028 max_low_pfn
= max_pfn
;
1029 if (max_low_pfn
> MAXMEM_PFN
) {
1030 if (highmem_pages
== -1)
1031 highmem_pages
= max_pfn
- MAXMEM_PFN
;
1032 if (highmem_pages
+ MAXMEM_PFN
< max_pfn
)
1033 max_pfn
= MAXMEM_PFN
+ highmem_pages
;
1034 if (highmem_pages
+ MAXMEM_PFN
> max_pfn
) {
1035 printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn
- MAXMEM_PFN
), pages_to_mb(highmem_pages
));
1038 max_low_pfn
= MAXMEM_PFN
;
1039 #ifndef CONFIG_HIGHMEM
1040 /* Maximum memory usable is what is directly addressable */
1041 printk(KERN_WARNING
"Warning only %ldMB will be used.\n",
1043 if (max_pfn
> MAX_NONPAE_PFN
)
1044 printk(KERN_WARNING
"Use a PAE enabled kernel.\n");
1046 printk(KERN_WARNING
"Use a HIGHMEM enabled kernel.\n");
1047 max_pfn
= MAXMEM_PFN
;
1048 #else /* !CONFIG_HIGHMEM */
1049 #ifndef CONFIG_X86_PAE
1050 if (max_pfn
> MAX_NONPAE_PFN
) {
1051 max_pfn
= MAX_NONPAE_PFN
;
1052 printk(KERN_WARNING
"Warning only 4GB will be used.\n");
1053 printk(KERN_WARNING
"Use a PAE enabled kernel.\n");
1055 #endif /* !CONFIG_X86_PAE */
1056 #endif /* !CONFIG_HIGHMEM */
1058 if (highmem_pages
== -1)
1060 #ifdef CONFIG_HIGHMEM
1061 if (highmem_pages
>= max_pfn
) {
1062 printk(KERN_ERR
"highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages
), pages_to_mb(max_pfn
));
1065 if (highmem_pages
) {
1066 if (max_low_pfn
-highmem_pages
< 64*1024*1024/PAGE_SIZE
){
1067 printk(KERN_ERR
"highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages
));
1070 max_low_pfn
-= highmem_pages
;
1074 printk(KERN_ERR
"ignoring highmem size on non-highmem kernel!\n");
1081 * Free all available memory for boot time allocation. Used
1082 * as a callback function by efi_memory_walk()
1086 free_available_memory(unsigned long start
, unsigned long end
, void *arg
)
1088 /* check max_low_pfn */
1089 if (start
>= (max_low_pfn
<< PAGE_SHIFT
))
1091 if (end
>= (max_low_pfn
<< PAGE_SHIFT
))
1092 end
= max_low_pfn
<< PAGE_SHIFT
;
1094 free_bootmem(start
, end
- start
);
1099 * Register fully available low RAM pages with the bootmem allocator.
1101 static void __init
register_bootmem_low_pages(unsigned long max_low_pfn
)
1106 efi_memmap_walk(free_available_memory
, NULL
);
1109 for (i
= 0; i
< e820
.nr_map
; i
++) {
1110 unsigned long curr_pfn
, last_pfn
, size
;
1112 * Reserve usable low memory
1114 if (e820
.map
[i
].type
!= E820_RAM
)
1117 * We are rounding up the start address of usable memory:
1119 curr_pfn
= PFN_UP(e820
.map
[i
].addr
);
1120 if (curr_pfn
>= max_low_pfn
)
1123 * ... and at the end of the usable range downwards:
1125 last_pfn
= PFN_DOWN(e820
.map
[i
].addr
+ e820
.map
[i
].size
);
1127 if (last_pfn
> max_low_pfn
)
1128 last_pfn
= max_low_pfn
;
1131 * .. finally, did all the rounding and playing
1132 * around just make the area go away?
1134 if (last_pfn
<= curr_pfn
)
1137 size
= last_pfn
- curr_pfn
;
1138 free_bootmem(PFN_PHYS(curr_pfn
), PFN_PHYS(size
));
1143 * workaround for Dell systems that neglect to reserve EBDA
1145 static void __init
reserve_ebda_region(void)
1148 addr
= get_bios_ebda();
1150 reserve_bootmem(addr
, PAGE_SIZE
);
1153 #ifndef CONFIG_NEED_MULTIPLE_NODES
1154 void __init
setup_bootmem_allocator(void);
1155 static unsigned long __init
setup_memory(void)
1158 * partially used pages are not usable - thus
1159 * we are rounding upwards:
1161 min_low_pfn
= PFN_UP(init_pg_tables_end
);
1165 max_low_pfn
= find_max_low_pfn();
1167 #ifdef CONFIG_HIGHMEM
1168 highstart_pfn
= highend_pfn
= max_pfn
;
1169 if (max_pfn
> max_low_pfn
) {
1170 highstart_pfn
= max_low_pfn
;
1172 printk(KERN_NOTICE
"%ldMB HIGHMEM available.\n",
1173 pages_to_mb(highend_pfn
- highstart_pfn
));
1175 printk(KERN_NOTICE
"%ldMB LOWMEM available.\n",
1176 pages_to_mb(max_low_pfn
));
1178 setup_bootmem_allocator();
1183 void __init
zone_sizes_init(void)
1185 unsigned long zones_size
[MAX_NR_ZONES
] = {0, 0, 0};
1186 unsigned int max_dma
, low
;
1188 max_dma
= virt_to_phys((char *)MAX_DMA_ADDRESS
) >> PAGE_SHIFT
;
1192 zones_size
[ZONE_DMA
] = low
;
1194 zones_size
[ZONE_DMA
] = max_dma
;
1195 zones_size
[ZONE_NORMAL
] = low
- max_dma
;
1196 #ifdef CONFIG_HIGHMEM
1197 zones_size
[ZONE_HIGHMEM
] = highend_pfn
- low
;
1200 free_area_init(zones_size
);
1203 extern unsigned long __init
setup_memory(void);
1204 extern void zone_sizes_init(void);
1205 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
1207 void __init
setup_bootmem_allocator(void)
1209 unsigned long bootmap_size
;
1211 * Initialize the boot-time allocator (with low memory only):
1213 bootmap_size
= init_bootmem(min_low_pfn
, max_low_pfn
);
1215 register_bootmem_low_pages(max_low_pfn
);
1218 * Reserve the bootmem bitmap itself as well. We do this in two
1219 * steps (first step was init_bootmem()) because this catches
1220 * the (very unlikely) case of us accidentally initializing the
1221 * bootmem allocator with an invalid RAM area.
1223 reserve_bootmem(__PHYSICAL_START
, (PFN_PHYS(min_low_pfn
) +
1224 bootmap_size
+ PAGE_SIZE
-1) - (__PHYSICAL_START
));
1227 * reserve physical page 0 - it's a special BIOS page on many boxes,
1228 * enabling clean reboots, SMP operation, laptop functions.
1230 reserve_bootmem(0, PAGE_SIZE
);
1232 /* reserve EBDA region, it's a 4K region */
1233 reserve_ebda_region();
1235 /* could be an AMD 768MPX chipset. Reserve a page before VGA to prevent
1236 PCI prefetch into it (errata #56). Usually the page is reserved anyways,
1237 unless you have no PS/2 mouse plugged in. */
1238 if (boot_cpu_data
.x86_vendor
== X86_VENDOR_AMD
&&
1239 boot_cpu_data
.x86
== 6)
1240 reserve_bootmem(0xa0000 - 4096, 4096);
1244 * But first pinch a few for the stack/trampoline stuff
1245 * FIXME: Don't need the extra page at 4K, but need to fix
1246 * trampoline before removing it. (see the GDT stuff)
1248 reserve_bootmem(PAGE_SIZE
, PAGE_SIZE
);
1250 #ifdef CONFIG_ACPI_SLEEP
1252 * Reserve low memory region for sleep support.
1254 acpi_reserve_bootmem();
1256 #ifdef CONFIG_X86_FIND_SMP_CONFIG
1258 * Find and reserve possible boot-time SMP configuration:
1263 #ifdef CONFIG_BLK_DEV_INITRD
1264 if (LOADER_TYPE
&& INITRD_START
) {
1265 if (INITRD_START
+ INITRD_SIZE
<= (max_low_pfn
<< PAGE_SHIFT
)) {
1266 reserve_bootmem(INITRD_START
, INITRD_SIZE
);
1268 INITRD_START
? INITRD_START
+ PAGE_OFFSET
: 0;
1269 initrd_end
= initrd_start
+INITRD_SIZE
;
1272 printk(KERN_ERR
"initrd extends beyond end of memory "
1273 "(0x%08lx > 0x%08lx)\ndisabling initrd\n",
1274 INITRD_START
+ INITRD_SIZE
,
1275 max_low_pfn
<< PAGE_SHIFT
);
1281 if (crashk_res
.start
!= crashk_res
.end
)
1282 reserve_bootmem(crashk_res
.start
,
1283 crashk_res
.end
- crashk_res
.start
+ 1);
1288 * The node 0 pgdat is initialized before all of these because
1289 * it's needed for bootmem. node>0 pgdats have their virtual
1290 * space allocated before the pagetables are in place to access
1291 * them, so they can't be cleared then.
1293 * This should all compile down to nothing when NUMA is off.
1295 void __init
remapped_pgdat_init(void)
1299 for_each_online_node(nid
) {
1301 memset(NODE_DATA(nid
), 0, sizeof(struct pglist_data
));
1306 * Request address space for all standard RAM and ROM resources
1307 * and also for regions reported as reserved by the e820.
1310 legacy_init_iomem_resources(struct resource
*code_resource
, struct resource
*data_resource
)
1315 for (i
= 0; i
< e820
.nr_map
; i
++) {
1316 struct resource
*res
;
1317 if (e820
.map
[i
].addr
+ e820
.map
[i
].size
> 0x100000000ULL
)
1319 res
= kzalloc(sizeof(struct resource
), GFP_ATOMIC
);
1320 switch (e820
.map
[i
].type
) {
1321 case E820_RAM
: res
->name
= "System RAM"; break;
1322 case E820_ACPI
: res
->name
= "ACPI Tables"; break;
1323 case E820_NVS
: res
->name
= "ACPI Non-volatile Storage"; break;
1324 default: res
->name
= "reserved";
1326 res
->start
= e820
.map
[i
].addr
;
1327 res
->end
= res
->start
+ e820
.map
[i
].size
- 1;
1328 res
->flags
= IORESOURCE_MEM
| IORESOURCE_BUSY
;
1329 request_resource(&iomem_resource
, res
);
1330 if (e820
.map
[i
].type
== E820_RAM
) {
1332 * We don't know which RAM region contains kernel data,
1333 * so we try it repeatedly and let the resource manager
1336 request_resource(res
, code_resource
);
1337 request_resource(res
, data_resource
);
1339 request_resource(res
, &crashk_res
);
1346 * Request address space for all standard resources
1348 * This is called just before pcibios_init(), which is also a
1349 * subsys_initcall, but is linked in later (in arch/i386/pci/common.c).
1351 static int __init
request_standard_resources(void)
1355 printk("Setting up standard PCI resources\n");
1357 efi_initialize_iomem_resources(&code_resource
, &data_resource
);
1359 legacy_init_iomem_resources(&code_resource
, &data_resource
);
1361 /* EFI systems may still have VGA */
1362 request_resource(&iomem_resource
, &video_ram_resource
);
1364 /* request I/O space for devices used on all i[345]86 PCs */
1365 for (i
= 0; i
< STANDARD_IO_RESOURCES
; i
++)
1366 request_resource(&ioport_resource
, &standard_io_resources
[i
]);
1370 subsys_initcall(request_standard_resources
);
1372 static void __init
register_memory(void)
1374 unsigned long gapstart
, gapsize
, round
;
1375 unsigned long long last
;
1379 * Search for the bigest gap in the low 32 bits of the e820
1382 last
= 0x100000000ull
;
1383 gapstart
= 0x10000000;
1387 unsigned long long start
= e820
.map
[i
].addr
;
1388 unsigned long long end
= start
+ e820
.map
[i
].size
;
1391 * Since "last" is at most 4GB, we know we'll
1392 * fit in 32 bits if this condition is true
1395 unsigned long gap
= last
- end
;
1397 if (gap
> gapsize
) {
1407 * See how much we want to round up: start off with
1408 * rounding to the next 1MB area.
1411 while ((gapsize
>> 4) > round
)
1413 /* Fun with two's complement */
1414 pci_mem_start
= (gapstart
+ round
) & -round
;
1416 printk("Allocating PCI resources starting at %08lx (gap: %08lx:%08lx)\n",
1417 pci_mem_start
, gapstart
, gapsize
);
1421 static void set_mca_bus(int x
)
1426 static void set_mca_bus(int x
) { }
1430 * Determine if we were loaded by an EFI loader. If so, then we have also been
1431 * passed the efi memmap, systab, etc., so we should use these data structures
1432 * for initialization. Note, the efi init code path is determined by the
1433 * global efi_enabled. This allows the same kernel image to be used on existing
1434 * systems (with a traditional BIOS) as well as on EFI systems.
1436 void __init
setup_arch(char **cmdline_p
)
1438 unsigned long max_low_pfn
;
1440 memcpy(&boot_cpu_data
, &new_cpu_data
, sizeof(new_cpu_data
));
1441 pre_setup_arch_hook();
1445 * FIXME: This isn't an official loader_type right
1446 * now but does currently work with elilo.
1447 * If we were configured as an EFI kernel, check to make
1448 * sure that we were loaded correctly from elilo and that
1449 * the system table is valid. If not, then initialize normally.
1452 if ((LOADER_TYPE
== 0x50) && EFI_SYSTAB
)
1456 ROOT_DEV
= old_decode_dev(ORIG_ROOT_DEV
);
1457 drive_info
= DRIVE_INFO
;
1458 screen_info
= SCREEN_INFO
;
1459 edid_info
= EDID_INFO
;
1460 apm_info
.bios
= APM_BIOS_INFO
;
1461 ist_info
= IST_INFO
;
1462 saved_videomode
= VIDEO_MODE
;
1463 if( SYS_DESC_TABLE
.length
!= 0 ) {
1464 set_mca_bus(SYS_DESC_TABLE
.table
[3] & 0x2);
1465 machine_id
= SYS_DESC_TABLE
.table
[0];
1466 machine_submodel_id
= SYS_DESC_TABLE
.table
[1];
1467 BIOS_revision
= SYS_DESC_TABLE
.table
[2];
1469 bootloader_type
= LOADER_TYPE
;
1471 #ifdef CONFIG_BLK_DEV_RAM
1472 rd_image_start
= RAMDISK_FLAGS
& RAMDISK_IMAGE_START_MASK
;
1473 rd_prompt
= ((RAMDISK_FLAGS
& RAMDISK_PROMPT_FLAG
) != 0);
1474 rd_doload
= ((RAMDISK_FLAGS
& RAMDISK_LOAD_FLAG
) != 0);
1480 printk(KERN_INFO
"BIOS-provided physical RAM map:\n");
1481 print_memory_map(machine_specific_memory_setup());
1486 if (!MOUNT_ROOT_RDONLY
)
1487 root_mountflags
&= ~MS_RDONLY
;
1488 init_mm
.start_code
= (unsigned long) _text
;
1489 init_mm
.end_code
= (unsigned long) _etext
;
1490 init_mm
.end_data
= (unsigned long) _edata
;
1491 init_mm
.brk
= init_pg_tables_end
+ PAGE_OFFSET
;
1493 code_resource
.start
= virt_to_phys(_text
);
1494 code_resource
.end
= virt_to_phys(_etext
)-1;
1495 data_resource
.start
= virt_to_phys(_etext
);
1496 data_resource
.end
= virt_to_phys(_edata
)-1;
1498 parse_cmdline_early(cmdline_p
);
1500 #ifdef CONFIG_EARLY_PRINTK
1502 char *s
= strstr(*cmdline_p
, "earlyprintk=");
1504 setup_early_printk(strchr(s
, '=') + 1);
1505 printk("early console enabled\n");
1510 max_low_pfn
= setup_memory();
1513 * NOTE: before this point _nobody_ is allowed to allocate
1514 * any memory using the bootmem allocator. Although the
1515 * alloctor is now initialised only the first 8Mb of the kernel
1516 * virtual address space has been mapped. All allocations before
1517 * paging_init() has completed must use the alloc_bootmem_low_pages()
1518 * variant (which allocates DMA'able memory) and care must be taken
1519 * not to exceed the 8Mb limit.
1523 smp_alloc_memory(); /* AP processor realmode stacks in low memory*/
1526 remapped_pgdat_init();
1531 * NOTE: at this point the bootmem allocator is fully available.
1536 #ifdef CONFIG_X86_GENERICARCH
1537 generic_apic_probe(*cmdline_p
);
1544 * Parse the ACPI tables for possible boot-time SMP configuration.
1546 acpi_boot_table_init();
1549 #ifdef CONFIG_X86_IO_APIC
1550 check_acpi_pci(); /* Checks more than just ACPI actually */
1556 #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC)
1558 printk(KERN_WARNING
"More than 8 CPUs detected and "
1559 "CONFIG_X86_PC cannot handle it.\nUse "
1560 "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
1563 #ifdef CONFIG_X86_LOCAL_APIC
1564 if (smp_found_config
)
1571 #if defined(CONFIG_VGA_CONSOLE)
1572 if (!efi_enabled
|| (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY
))
1573 conswitchp
= &vga_con
;
1574 #elif defined(CONFIG_DUMMY_CONSOLE)
1575 conswitchp
= &dummy_con
;
1580 static __init
int add_pcspkr(void)
1582 struct platform_device
*pd
;
1585 pd
= platform_device_alloc("pcspkr", -1);
1589 ret
= platform_device_add(pd
);
1591 platform_device_put(pd
);
1595 device_initcall(add_pcspkr
);
1600 * c-file-style:"k&r"