[PATCH] i386: vMI backend for paravirt-ops
[linux-2.6/linux-2.6-openrd.git] / arch / i386 / kernel / setup.c
blobbd8c218d94af5e1cdc6de06c4ce2fc29941e04b0
1 /*
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>
27 #include <linux/mm.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>
53 #include <asm/apic.h>
54 #include <asm/e820.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>
61 #include <asm/ist.h>
62 #include <asm/io.h>
63 #include <asm/vmi.h>
64 #include <setup_arch.h>
65 #include <bios_ebda.h>
67 /* This value is set up by the early boot code to point to the value
68 immediately after the boot time page tables. It contains a *physical*
69 address, and must not be in the .bss segment! */
70 unsigned long init_pg_tables_end __initdata = ~0UL;
72 int disable_pse __devinitdata = 0;
75 * Machine setup..
77 extern struct resource code_resource;
78 extern struct resource data_resource;
80 /* cpu data as detected by the assembly code in head.S */
81 struct cpuinfo_x86 new_cpu_data __cpuinitdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
82 /* common cpu data for all cpus */
83 struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
84 EXPORT_SYMBOL(boot_cpu_data);
86 unsigned long mmu_cr4_features;
88 /* for MCA, but anyone else can use it if they want */
89 unsigned int machine_id;
90 #ifdef CONFIG_MCA
91 EXPORT_SYMBOL(machine_id);
92 #endif
93 unsigned int machine_submodel_id;
94 unsigned int BIOS_revision;
95 unsigned int mca_pentium_flag;
97 /* Boot loader ID as an integer, for the benefit of proc_dointvec */
98 int bootloader_type;
100 /* user-defined highmem size */
101 static unsigned int highmem_pages = -1;
104 * Setup options
106 struct drive_info_struct { char dummy[32]; } drive_info;
107 #if defined(CONFIG_BLK_DEV_IDE) || defined(CONFIG_BLK_DEV_HD) || \
108 defined(CONFIG_BLK_DEV_IDE_MODULE) || defined(CONFIG_BLK_DEV_HD_MODULE)
109 EXPORT_SYMBOL(drive_info);
110 #endif
111 struct screen_info screen_info;
112 EXPORT_SYMBOL(screen_info);
113 struct apm_info apm_info;
114 EXPORT_SYMBOL(apm_info);
115 struct sys_desc_table_struct {
116 unsigned short length;
117 unsigned char table[0];
119 struct edid_info edid_info;
120 EXPORT_SYMBOL_GPL(edid_info);
121 struct ist_info ist_info;
122 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
123 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
124 EXPORT_SYMBOL(ist_info);
125 #endif
127 extern void early_cpu_init(void);
128 extern int root_mountflags;
130 unsigned long saved_videomode;
132 #define RAMDISK_IMAGE_START_MASK 0x07FF
133 #define RAMDISK_PROMPT_FLAG 0x8000
134 #define RAMDISK_LOAD_FLAG 0x4000
136 static char __initdata command_line[COMMAND_LINE_SIZE];
138 unsigned char __initdata boot_params[PARAM_SIZE];
140 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
141 struct edd edd;
142 #ifdef CONFIG_EDD_MODULE
143 EXPORT_SYMBOL(edd);
144 #endif
146 * copy_edd() - Copy the BIOS EDD information
147 * from boot_params into a safe place.
150 static inline void copy_edd(void)
152 memcpy(edd.mbr_signature, EDD_MBR_SIGNATURE, sizeof(edd.mbr_signature));
153 memcpy(edd.edd_info, EDD_BUF, sizeof(edd.edd_info));
154 edd.mbr_signature_nr = EDD_MBR_SIG_NR;
155 edd.edd_info_nr = EDD_NR;
157 #else
158 static inline void copy_edd(void)
161 #endif
163 int __initdata user_defined_memmap = 0;
166 * "mem=nopentium" disables the 4MB page tables.
167 * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM
168 * to <mem>, overriding the bios size.
169 * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from
170 * <start> to <start>+<mem>, overriding the bios size.
172 * HPA tells me bootloaders need to parse mem=, so no new
173 * option should be mem= [also see Documentation/i386/boot.txt]
175 static int __init parse_mem(char *arg)
177 if (!arg)
178 return -EINVAL;
180 if (strcmp(arg, "nopentium") == 0) {
181 clear_bit(X86_FEATURE_PSE, boot_cpu_data.x86_capability);
182 disable_pse = 1;
183 } else {
184 /* If the user specifies memory size, we
185 * limit the BIOS-provided memory map to
186 * that size. exactmap can be used to specify
187 * the exact map. mem=number can be used to
188 * trim the existing memory map.
190 unsigned long long mem_size;
192 mem_size = memparse(arg, &arg);
193 limit_regions(mem_size);
194 user_defined_memmap = 1;
196 return 0;
198 early_param("mem", parse_mem);
200 #ifdef CONFIG_PROC_VMCORE
201 /* elfcorehdr= specifies the location of elf core header
202 * stored by the crashed kernel.
204 static int __init parse_elfcorehdr(char *arg)
206 if (!arg)
207 return -EINVAL;
209 elfcorehdr_addr = memparse(arg, &arg);
210 return 0;
212 early_param("elfcorehdr", parse_elfcorehdr);
213 #endif /* CONFIG_PROC_VMCORE */
216 * highmem=size forces highmem to be exactly 'size' bytes.
217 * This works even on boxes that have no highmem otherwise.
218 * This also works to reduce highmem size on bigger boxes.
220 static int __init parse_highmem(char *arg)
222 if (!arg)
223 return -EINVAL;
225 highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT;
226 return 0;
228 early_param("highmem", parse_highmem);
231 * vmalloc=size forces the vmalloc area to be exactly 'size'
232 * bytes. This can be used to increase (or decrease) the
233 * vmalloc area - the default is 128m.
235 static int __init parse_vmalloc(char *arg)
237 if (!arg)
238 return -EINVAL;
240 __VMALLOC_RESERVE = memparse(arg, &arg);
241 return 0;
243 early_param("vmalloc", parse_vmalloc);
246 * reservetop=size reserves a hole at the top of the kernel address space which
247 * a hypervisor can load into later. Needed for dynamically loaded hypervisors,
248 * so relocating the fixmap can be done before paging initialization.
250 static int __init parse_reservetop(char *arg)
252 unsigned long address;
254 if (!arg)
255 return -EINVAL;
257 address = memparse(arg, &arg);
258 reserve_top_address(address);
259 return 0;
261 early_param("reservetop", parse_reservetop);
264 * Determine low and high memory ranges:
266 unsigned long __init find_max_low_pfn(void)
268 unsigned long max_low_pfn;
270 max_low_pfn = max_pfn;
271 if (max_low_pfn > MAXMEM_PFN) {
272 if (highmem_pages == -1)
273 highmem_pages = max_pfn - MAXMEM_PFN;
274 if (highmem_pages + MAXMEM_PFN < max_pfn)
275 max_pfn = MAXMEM_PFN + highmem_pages;
276 if (highmem_pages + MAXMEM_PFN > max_pfn) {
277 printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn - MAXMEM_PFN), pages_to_mb(highmem_pages));
278 highmem_pages = 0;
280 max_low_pfn = MAXMEM_PFN;
281 #ifndef CONFIG_HIGHMEM
282 /* Maximum memory usable is what is directly addressable */
283 printk(KERN_WARNING "Warning only %ldMB will be used.\n",
284 MAXMEM>>20);
285 if (max_pfn > MAX_NONPAE_PFN)
286 printk(KERN_WARNING "Use a PAE enabled kernel.\n");
287 else
288 printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n");
289 max_pfn = MAXMEM_PFN;
290 #else /* !CONFIG_HIGHMEM */
291 #ifndef CONFIG_X86_PAE
292 if (max_pfn > MAX_NONPAE_PFN) {
293 max_pfn = MAX_NONPAE_PFN;
294 printk(KERN_WARNING "Warning only 4GB will be used.\n");
295 printk(KERN_WARNING "Use a PAE enabled kernel.\n");
297 #endif /* !CONFIG_X86_PAE */
298 #endif /* !CONFIG_HIGHMEM */
299 } else {
300 if (highmem_pages == -1)
301 highmem_pages = 0;
302 #ifdef CONFIG_HIGHMEM
303 if (highmem_pages >= max_pfn) {
304 printk(KERN_ERR "highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages), pages_to_mb(max_pfn));
305 highmem_pages = 0;
307 if (highmem_pages) {
308 if (max_low_pfn-highmem_pages < 64*1024*1024/PAGE_SIZE){
309 printk(KERN_ERR "highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages));
310 highmem_pages = 0;
312 max_low_pfn -= highmem_pages;
314 #else
315 if (highmem_pages)
316 printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n");
317 #endif
319 return max_low_pfn;
323 * workaround for Dell systems that neglect to reserve EBDA
325 static void __init reserve_ebda_region(void)
327 unsigned int addr;
328 addr = get_bios_ebda();
329 if (addr)
330 reserve_bootmem(addr, PAGE_SIZE);
333 #ifndef CONFIG_NEED_MULTIPLE_NODES
334 void __init setup_bootmem_allocator(void);
335 static unsigned long __init setup_memory(void)
338 * partially used pages are not usable - thus
339 * we are rounding upwards:
341 min_low_pfn = PFN_UP(init_pg_tables_end);
343 find_max_pfn();
345 max_low_pfn = find_max_low_pfn();
347 #ifdef CONFIG_HIGHMEM
348 highstart_pfn = highend_pfn = max_pfn;
349 if (max_pfn > max_low_pfn) {
350 highstart_pfn = max_low_pfn;
352 printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
353 pages_to_mb(highend_pfn - highstart_pfn));
354 num_physpages = highend_pfn;
355 high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
356 #else
357 num_physpages = max_low_pfn;
358 high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
359 #endif
360 #ifdef CONFIG_FLATMEM
361 max_mapnr = num_physpages;
362 #endif
363 printk(KERN_NOTICE "%ldMB LOWMEM available.\n",
364 pages_to_mb(max_low_pfn));
366 setup_bootmem_allocator();
368 return max_low_pfn;
371 void __init zone_sizes_init(void)
373 unsigned long max_zone_pfns[MAX_NR_ZONES];
374 memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
375 max_zone_pfns[ZONE_DMA] =
376 virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
377 max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
378 #ifdef CONFIG_HIGHMEM
379 max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
380 add_active_range(0, 0, highend_pfn);
381 #else
382 add_active_range(0, 0, max_low_pfn);
383 #endif
385 free_area_init_nodes(max_zone_pfns);
387 #else
388 extern unsigned long __init setup_memory(void);
389 extern void zone_sizes_init(void);
390 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
392 void __init setup_bootmem_allocator(void)
394 unsigned long bootmap_size;
396 * Initialize the boot-time allocator (with low memory only):
398 bootmap_size = init_bootmem(min_low_pfn, max_low_pfn);
400 register_bootmem_low_pages(max_low_pfn);
403 * Reserve the bootmem bitmap itself as well. We do this in two
404 * steps (first step was init_bootmem()) because this catches
405 * the (very unlikely) case of us accidentally initializing the
406 * bootmem allocator with an invalid RAM area.
408 reserve_bootmem(__pa_symbol(_text), (PFN_PHYS(min_low_pfn) +
409 bootmap_size + PAGE_SIZE-1) - __pa_symbol(_text));
412 * reserve physical page 0 - it's a special BIOS page on many boxes,
413 * enabling clean reboots, SMP operation, laptop functions.
415 reserve_bootmem(0, PAGE_SIZE);
417 /* reserve EBDA region, it's a 4K region */
418 reserve_ebda_region();
420 /* could be an AMD 768MPX chipset. Reserve a page before VGA to prevent
421 PCI prefetch into it (errata #56). Usually the page is reserved anyways,
422 unless you have no PS/2 mouse plugged in. */
423 if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
424 boot_cpu_data.x86 == 6)
425 reserve_bootmem(0xa0000 - 4096, 4096);
427 #ifdef CONFIG_SMP
429 * But first pinch a few for the stack/trampoline stuff
430 * FIXME: Don't need the extra page at 4K, but need to fix
431 * trampoline before removing it. (see the GDT stuff)
433 reserve_bootmem(PAGE_SIZE, PAGE_SIZE);
434 #endif
435 #ifdef CONFIG_ACPI_SLEEP
437 * Reserve low memory region for sleep support.
439 acpi_reserve_bootmem();
440 #endif
441 #ifdef CONFIG_X86_FIND_SMP_CONFIG
443 * Find and reserve possible boot-time SMP configuration:
445 find_smp_config();
446 #endif
447 numa_kva_reserve();
448 #ifdef CONFIG_BLK_DEV_INITRD
449 if (LOADER_TYPE && INITRD_START) {
450 if (INITRD_START + INITRD_SIZE <= (max_low_pfn << PAGE_SHIFT)) {
451 reserve_bootmem(INITRD_START, INITRD_SIZE);
452 initrd_start = INITRD_START + PAGE_OFFSET;
453 initrd_end = initrd_start+INITRD_SIZE;
455 else {
456 printk(KERN_ERR "initrd extends beyond end of memory "
457 "(0x%08lx > 0x%08lx)\ndisabling initrd\n",
458 INITRD_START + INITRD_SIZE,
459 max_low_pfn << PAGE_SHIFT);
460 initrd_start = 0;
463 #endif
464 #ifdef CONFIG_KEXEC
465 if (crashk_res.start != crashk_res.end)
466 reserve_bootmem(crashk_res.start,
467 crashk_res.end - crashk_res.start + 1);
468 #endif
472 * The node 0 pgdat is initialized before all of these because
473 * it's needed for bootmem. node>0 pgdats have their virtual
474 * space allocated before the pagetables are in place to access
475 * them, so they can't be cleared then.
477 * This should all compile down to nothing when NUMA is off.
479 void __init remapped_pgdat_init(void)
481 int nid;
483 for_each_online_node(nid) {
484 if (nid != 0)
485 memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
489 #ifdef CONFIG_MCA
490 static void set_mca_bus(int x)
492 MCA_bus = x;
494 #else
495 static void set_mca_bus(int x) { }
496 #endif
498 /* Overridden in paravirt.c if CONFIG_PARAVIRT */
499 char * __init __attribute__((weak)) memory_setup(void)
501 return machine_specific_memory_setup();
505 * Determine if we were loaded by an EFI loader. If so, then we have also been
506 * passed the efi memmap, systab, etc., so we should use these data structures
507 * for initialization. Note, the efi init code path is determined by the
508 * global efi_enabled. This allows the same kernel image to be used on existing
509 * systems (with a traditional BIOS) as well as on EFI systems.
511 void __init setup_arch(char **cmdline_p)
513 unsigned long max_low_pfn;
515 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
516 pre_setup_arch_hook();
517 early_cpu_init();
520 * FIXME: This isn't an official loader_type right
521 * now but does currently work with elilo.
522 * If we were configured as an EFI kernel, check to make
523 * sure that we were loaded correctly from elilo and that
524 * the system table is valid. If not, then initialize normally.
526 #ifdef CONFIG_EFI
527 if ((LOADER_TYPE == 0x50) && EFI_SYSTAB)
528 efi_enabled = 1;
529 #endif
531 ROOT_DEV = old_decode_dev(ORIG_ROOT_DEV);
532 drive_info = DRIVE_INFO;
533 screen_info = SCREEN_INFO;
534 edid_info = EDID_INFO;
535 apm_info.bios = APM_BIOS_INFO;
536 ist_info = IST_INFO;
537 saved_videomode = VIDEO_MODE;
538 if( SYS_DESC_TABLE.length != 0 ) {
539 set_mca_bus(SYS_DESC_TABLE.table[3] & 0x2);
540 machine_id = SYS_DESC_TABLE.table[0];
541 machine_submodel_id = SYS_DESC_TABLE.table[1];
542 BIOS_revision = SYS_DESC_TABLE.table[2];
544 bootloader_type = LOADER_TYPE;
546 #ifdef CONFIG_BLK_DEV_RAM
547 rd_image_start = RAMDISK_FLAGS & RAMDISK_IMAGE_START_MASK;
548 rd_prompt = ((RAMDISK_FLAGS & RAMDISK_PROMPT_FLAG) != 0);
549 rd_doload = ((RAMDISK_FLAGS & RAMDISK_LOAD_FLAG) != 0);
550 #endif
551 ARCH_SETUP
552 if (efi_enabled)
553 efi_init();
554 else {
555 printk(KERN_INFO "BIOS-provided physical RAM map:\n");
556 print_memory_map(memory_setup());
559 copy_edd();
561 if (!MOUNT_ROOT_RDONLY)
562 root_mountflags &= ~MS_RDONLY;
563 init_mm.start_code = (unsigned long) _text;
564 init_mm.end_code = (unsigned long) _etext;
565 init_mm.end_data = (unsigned long) _edata;
566 init_mm.brk = init_pg_tables_end + PAGE_OFFSET;
568 code_resource.start = virt_to_phys(_text);
569 code_resource.end = virt_to_phys(_etext)-1;
570 data_resource.start = virt_to_phys(_etext);
571 data_resource.end = virt_to_phys(_edata)-1;
573 parse_early_param();
575 if (user_defined_memmap) {
576 printk(KERN_INFO "user-defined physical RAM map:\n");
577 print_memory_map("user");
580 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
581 *cmdline_p = command_line;
583 max_low_pfn = setup_memory();
585 #ifdef CONFIG_VMI
587 * Must be after max_low_pfn is determined, and before kernel
588 * pagetables are setup.
590 vmi_init();
591 #endif
594 * NOTE: before this point _nobody_ is allowed to allocate
595 * any memory using the bootmem allocator. Although the
596 * alloctor is now initialised only the first 8Mb of the kernel
597 * virtual address space has been mapped. All allocations before
598 * paging_init() has completed must use the alloc_bootmem_low_pages()
599 * variant (which allocates DMA'able memory) and care must be taken
600 * not to exceed the 8Mb limit.
603 #ifdef CONFIG_SMP
604 smp_alloc_memory(); /* AP processor realmode stacks in low memory*/
605 #endif
606 paging_init();
607 remapped_pgdat_init();
608 sparse_init();
609 zone_sizes_init();
612 * NOTE: at this point the bootmem allocator is fully available.
615 dmi_scan_machine();
617 #ifdef CONFIG_X86_GENERICARCH
618 generic_apic_probe();
619 #endif
620 if (efi_enabled)
621 efi_map_memmap();
623 #ifdef CONFIG_ACPI
625 * Parse the ACPI tables for possible boot-time SMP configuration.
627 acpi_boot_table_init();
628 #endif
630 #ifdef CONFIG_PCI
631 #ifdef CONFIG_X86_IO_APIC
632 check_acpi_pci(); /* Checks more than just ACPI actually */
633 #endif
634 #endif
636 #ifdef CONFIG_ACPI
637 acpi_boot_init();
639 #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC)
640 if (def_to_bigsmp)
641 printk(KERN_WARNING "More than 8 CPUs detected and "
642 "CONFIG_X86_PC cannot handle it.\nUse "
643 "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
644 #endif
645 #endif
646 #ifdef CONFIG_X86_LOCAL_APIC
647 if (smp_found_config)
648 get_smp_config();
649 #endif
651 e820_register_memory();
653 #ifdef CONFIG_VT
654 #if defined(CONFIG_VGA_CONSOLE)
655 if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
656 conswitchp = &vga_con;
657 #elif defined(CONFIG_DUMMY_CONSOLE)
658 conswitchp = &dummy_con;
659 #endif
660 #endif
661 tsc_init();
664 static __init int add_pcspkr(void)
666 struct platform_device *pd;
667 int ret;
669 pd = platform_device_alloc("pcspkr", -1);
670 if (!pd)
671 return -ENOMEM;
673 ret = platform_device_add(pd);
674 if (ret)
675 platform_device_put(pd);
677 return ret;
679 device_initcall(add_pcspkr);
682 * Local Variables:
683 * mode:c
684 * c-file-style:"k&r"
685 * c-basic-offset:8
686 * End: