Merge git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild
[deliverable/linux.git] / arch / x86 / kernel / setup_32.c
1 /*
2 * Copyright (C) 1995 Linus Torvalds
3 *
4 * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
5 *
6 * Memory region support
7 * David Parsons <orc@pell.chi.il.us>, July-August 1999
8 *
9 * Added E820 sanitization routine (removes overlapping memory regions);
10 * Brian Moyle <bmoyle@mvista.com>, February 2001
11 *
12 * Moved CPU detection code to cpu/${cpu}.c
13 * Patrick Mochel <mochel@osdl.org>, March 2002
14 *
15 * Provisions for empty E820 memory regions (reported by certain BIOSes).
16 * Alex Achenbach <xela@slit.de>, December 2002.
17 *
18 */
19
20 /*
21 * This file handles the architecture-dependent parts of initialization
22 */
23
24 #include <linux/sched.h>
25 #include <linux/mm.h>
26 #include <linux/mmzone.h>
27 #include <linux/screen_info.h>
28 #include <linux/ioport.h>
29 #include <linux/acpi.h>
30 #include <linux/apm_bios.h>
31 #include <linux/initrd.h>
32 #include <linux/bootmem.h>
33 #include <linux/seq_file.h>
34 #include <linux/console.h>
35 #include <linux/mca.h>
36 #include <linux/root_dev.h>
37 #include <linux/highmem.h>
38 #include <linux/module.h>
39 #include <linux/efi.h>
40 #include <linux/init.h>
41 #include <linux/edd.h>
42 #include <linux/nodemask.h>
43 #include <linux/kexec.h>
44 #include <linux/crash_dump.h>
45 #include <linux/dmi.h>
46 #include <linux/pfn.h>
47
48 #include <video/edid.h>
49
50 #include <asm/apic.h>
51 #include <asm/e820.h>
52 #include <asm/mpspec.h>
53 #include <asm/mmzone.h>
54 #include <asm/setup.h>
55 #include <asm/arch_hooks.h>
56 #include <asm/sections.h>
57 #include <asm/io_apic.h>
58 #include <asm/ist.h>
59 #include <asm/io.h>
60 #include <asm/vmi.h>
61 #include <setup_arch.h>
62 #include <bios_ebda.h>
63
64 /* This value is set up by the early boot code to point to the value
65 immediately after the boot time page tables. It contains a *physical*
66 address, and must not be in the .bss segment! */
67 unsigned long init_pg_tables_end __initdata = ~0UL;
68
69 int disable_pse __devinitdata = 0;
70
71 /*
72 * Machine setup..
73 */
74 extern struct resource code_resource;
75 extern struct resource data_resource;
76
77 /* cpu data as detected by the assembly code in head.S */
78 struct cpuinfo_x86 new_cpu_data __cpuinitdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
79 /* common cpu data for all cpus */
80 struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
81 EXPORT_SYMBOL(boot_cpu_data);
82
83 unsigned long mmu_cr4_features;
84
85 /* for MCA, but anyone else can use it if they want */
86 unsigned int machine_id;
87 #ifdef CONFIG_MCA
88 EXPORT_SYMBOL(machine_id);
89 #endif
90 unsigned int machine_submodel_id;
91 unsigned int BIOS_revision;
92 unsigned int mca_pentium_flag;
93
94 /* Boot loader ID as an integer, for the benefit of proc_dointvec */
95 int bootloader_type;
96
97 /* user-defined highmem size */
98 static unsigned int highmem_pages = -1;
99
100 /*
101 * Setup options
102 */
103 struct screen_info screen_info;
104 EXPORT_SYMBOL(screen_info);
105 struct apm_info apm_info;
106 EXPORT_SYMBOL(apm_info);
107 struct edid_info edid_info;
108 EXPORT_SYMBOL_GPL(edid_info);
109 struct ist_info ist_info;
110 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
111 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
112 EXPORT_SYMBOL(ist_info);
113 #endif
114
115 extern void early_cpu_init(void);
116 extern int root_mountflags;
117
118 unsigned long saved_videomode;
119
120 #define RAMDISK_IMAGE_START_MASK 0x07FF
121 #define RAMDISK_PROMPT_FLAG 0x8000
122 #define RAMDISK_LOAD_FLAG 0x4000
123
124 static char __initdata command_line[COMMAND_LINE_SIZE];
125
126 struct boot_params __initdata boot_params;
127
128 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
129 struct edd edd;
130 #ifdef CONFIG_EDD_MODULE
131 EXPORT_SYMBOL(edd);
132 #endif
133 /**
134 * copy_edd() - Copy the BIOS EDD information
135 * from boot_params into a safe place.
136 *
137 */
138 static inline void copy_edd(void)
139 {
140 memcpy(edd.mbr_signature, EDD_MBR_SIGNATURE, sizeof(edd.mbr_signature));
141 memcpy(edd.edd_info, EDD_BUF, sizeof(edd.edd_info));
142 edd.mbr_signature_nr = EDD_MBR_SIG_NR;
143 edd.edd_info_nr = EDD_NR;
144 }
145 #else
146 static inline void copy_edd(void)
147 {
148 }
149 #endif
150
151 int __initdata user_defined_memmap = 0;
152
153 /*
154 * "mem=nopentium" disables the 4MB page tables.
155 * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM
156 * to <mem>, overriding the bios size.
157 * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from
158 * <start> to <start>+<mem>, overriding the bios size.
159 *
160 * HPA tells me bootloaders need to parse mem=, so no new
161 * option should be mem= [also see Documentation/i386/boot.txt]
162 */
163 static int __init parse_mem(char *arg)
164 {
165 if (!arg)
166 return -EINVAL;
167
168 if (strcmp(arg, "nopentium") == 0) {
169 clear_bit(X86_FEATURE_PSE, boot_cpu_data.x86_capability);
170 disable_pse = 1;
171 } else {
172 /* If the user specifies memory size, we
173 * limit the BIOS-provided memory map to
174 * that size. exactmap can be used to specify
175 * the exact map. mem=number can be used to
176 * trim the existing memory map.
177 */
178 unsigned long long mem_size;
179
180 mem_size = memparse(arg, &arg);
181 limit_regions(mem_size);
182 user_defined_memmap = 1;
183 }
184 return 0;
185 }
186 early_param("mem", parse_mem);
187
188 #ifdef CONFIG_PROC_VMCORE
189 /* elfcorehdr= specifies the location of elf core header
190 * stored by the crashed kernel.
191 */
192 static int __init parse_elfcorehdr(char *arg)
193 {
194 if (!arg)
195 return -EINVAL;
196
197 elfcorehdr_addr = memparse(arg, &arg);
198 return 0;
199 }
200 early_param("elfcorehdr", parse_elfcorehdr);
201 #endif /* CONFIG_PROC_VMCORE */
202
203 /*
204 * highmem=size forces highmem to be exactly 'size' bytes.
205 * This works even on boxes that have no highmem otherwise.
206 * This also works to reduce highmem size on bigger boxes.
207 */
208 static int __init parse_highmem(char *arg)
209 {
210 if (!arg)
211 return -EINVAL;
212
213 highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT;
214 return 0;
215 }
216 early_param("highmem", parse_highmem);
217
218 /*
219 * vmalloc=size forces the vmalloc area to be exactly 'size'
220 * bytes. This can be used to increase (or decrease) the
221 * vmalloc area - the default is 128m.
222 */
223 static int __init parse_vmalloc(char *arg)
224 {
225 if (!arg)
226 return -EINVAL;
227
228 __VMALLOC_RESERVE = memparse(arg, &arg);
229 return 0;
230 }
231 early_param("vmalloc", parse_vmalloc);
232
233 /*
234 * reservetop=size reserves a hole at the top of the kernel address space which
235 * a hypervisor can load into later. Needed for dynamically loaded hypervisors,
236 * so relocating the fixmap can be done before paging initialization.
237 */
238 static int __init parse_reservetop(char *arg)
239 {
240 unsigned long address;
241
242 if (!arg)
243 return -EINVAL;
244
245 address = memparse(arg, &arg);
246 reserve_top_address(address);
247 return 0;
248 }
249 early_param("reservetop", parse_reservetop);
250
251 /*
252 * Determine low and high memory ranges:
253 */
254 unsigned long __init find_max_low_pfn(void)
255 {
256 unsigned long max_low_pfn;
257
258 max_low_pfn = max_pfn;
259 if (max_low_pfn > MAXMEM_PFN) {
260 if (highmem_pages == -1)
261 highmem_pages = max_pfn - MAXMEM_PFN;
262 if (highmem_pages + MAXMEM_PFN < max_pfn)
263 max_pfn = MAXMEM_PFN + highmem_pages;
264 if (highmem_pages + MAXMEM_PFN > max_pfn) {
265 printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn - MAXMEM_PFN), pages_to_mb(highmem_pages));
266 highmem_pages = 0;
267 }
268 max_low_pfn = MAXMEM_PFN;
269 #ifndef CONFIG_HIGHMEM
270 /* Maximum memory usable is what is directly addressable */
271 printk(KERN_WARNING "Warning only %ldMB will be used.\n",
272 MAXMEM>>20);
273 if (max_pfn > MAX_NONPAE_PFN)
274 printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n");
275 else
276 printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n");
277 max_pfn = MAXMEM_PFN;
278 #else /* !CONFIG_HIGHMEM */
279 #ifndef CONFIG_HIGHMEM64G
280 if (max_pfn > MAX_NONPAE_PFN) {
281 max_pfn = MAX_NONPAE_PFN;
282 printk(KERN_WARNING "Warning only 4GB will be used.\n");
283 printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n");
284 }
285 #endif /* !CONFIG_HIGHMEM64G */
286 #endif /* !CONFIG_HIGHMEM */
287 } else {
288 if (highmem_pages == -1)
289 highmem_pages = 0;
290 #ifdef CONFIG_HIGHMEM
291 if (highmem_pages >= max_pfn) {
292 printk(KERN_ERR "highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages), pages_to_mb(max_pfn));
293 highmem_pages = 0;
294 }
295 if (highmem_pages) {
296 if (max_low_pfn-highmem_pages < 64*1024*1024/PAGE_SIZE){
297 printk(KERN_ERR "highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages));
298 highmem_pages = 0;
299 }
300 max_low_pfn -= highmem_pages;
301 }
302 #else
303 if (highmem_pages)
304 printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n");
305 #endif
306 }
307 return max_low_pfn;
308 }
309
310 /*
311 * workaround for Dell systems that neglect to reserve EBDA
312 */
313 static void __init reserve_ebda_region(void)
314 {
315 unsigned int addr;
316 addr = get_bios_ebda();
317 if (addr)
318 reserve_bootmem(addr, PAGE_SIZE);
319 }
320
321 #ifndef CONFIG_NEED_MULTIPLE_NODES
322 void __init setup_bootmem_allocator(void);
323 static unsigned long __init setup_memory(void)
324 {
325 /*
326 * partially used pages are not usable - thus
327 * we are rounding upwards:
328 */
329 min_low_pfn = PFN_UP(init_pg_tables_end);
330
331 find_max_pfn();
332
333 max_low_pfn = find_max_low_pfn();
334
335 #ifdef CONFIG_HIGHMEM
336 highstart_pfn = highend_pfn = max_pfn;
337 if (max_pfn > max_low_pfn) {
338 highstart_pfn = max_low_pfn;
339 }
340 printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
341 pages_to_mb(highend_pfn - highstart_pfn));
342 num_physpages = highend_pfn;
343 high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
344 #else
345 num_physpages = max_low_pfn;
346 high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
347 #endif
348 #ifdef CONFIG_FLATMEM
349 max_mapnr = num_physpages;
350 #endif
351 printk(KERN_NOTICE "%ldMB LOWMEM available.\n",
352 pages_to_mb(max_low_pfn));
353
354 setup_bootmem_allocator();
355
356 return max_low_pfn;
357 }
358
359 void __init zone_sizes_init(void)
360 {
361 unsigned long max_zone_pfns[MAX_NR_ZONES];
362 memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
363 max_zone_pfns[ZONE_DMA] =
364 virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
365 max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
366 #ifdef CONFIG_HIGHMEM
367 max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
368 add_active_range(0, 0, highend_pfn);
369 #else
370 add_active_range(0, 0, max_low_pfn);
371 #endif
372
373 free_area_init_nodes(max_zone_pfns);
374 }
375 #else
376 extern unsigned long __init setup_memory(void);
377 extern void zone_sizes_init(void);
378 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
379
380 void __init setup_bootmem_allocator(void)
381 {
382 unsigned long bootmap_size;
383 /*
384 * Initialize the boot-time allocator (with low memory only):
385 */
386 bootmap_size = init_bootmem(min_low_pfn, max_low_pfn);
387
388 register_bootmem_low_pages(max_low_pfn);
389
390 /*
391 * Reserve the bootmem bitmap itself as well. We do this in two
392 * steps (first step was init_bootmem()) because this catches
393 * the (very unlikely) case of us accidentally initializing the
394 * bootmem allocator with an invalid RAM area.
395 */
396 reserve_bootmem(__pa_symbol(_text), (PFN_PHYS(min_low_pfn) +
397 bootmap_size + PAGE_SIZE-1) - __pa_symbol(_text));
398
399 /*
400 * reserve physical page 0 - it's a special BIOS page on many boxes,
401 * enabling clean reboots, SMP operation, laptop functions.
402 */
403 reserve_bootmem(0, PAGE_SIZE);
404
405 /* reserve EBDA region, it's a 4K region */
406 reserve_ebda_region();
407
408 /* could be an AMD 768MPX chipset. Reserve a page before VGA to prevent
409 PCI prefetch into it (errata #56). Usually the page is reserved anyways,
410 unless you have no PS/2 mouse plugged in. */
411 if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
412 boot_cpu_data.x86 == 6)
413 reserve_bootmem(0xa0000 - 4096, 4096);
414
415 #ifdef CONFIG_SMP
416 /*
417 * But first pinch a few for the stack/trampoline stuff
418 * FIXME: Don't need the extra page at 4K, but need to fix
419 * trampoline before removing it. (see the GDT stuff)
420 */
421 reserve_bootmem(PAGE_SIZE, PAGE_SIZE);
422 #endif
423 #ifdef CONFIG_ACPI_SLEEP
424 /*
425 * Reserve low memory region for sleep support.
426 */
427 acpi_reserve_bootmem();
428 #endif
429 #ifdef CONFIG_X86_FIND_SMP_CONFIG
430 /*
431 * Find and reserve possible boot-time SMP configuration:
432 */
433 find_smp_config();
434 #endif
435 numa_kva_reserve();
436 #ifdef CONFIG_BLK_DEV_INITRD
437 if (LOADER_TYPE && INITRD_START) {
438 if (INITRD_START + INITRD_SIZE <= (max_low_pfn << PAGE_SHIFT)) {
439 reserve_bootmem(INITRD_START, INITRD_SIZE);
440 initrd_start = INITRD_START + PAGE_OFFSET;
441 initrd_end = initrd_start+INITRD_SIZE;
442 }
443 else {
444 printk(KERN_ERR "initrd extends beyond end of memory "
445 "(0x%08lx > 0x%08lx)\ndisabling initrd\n",
446 INITRD_START + INITRD_SIZE,
447 max_low_pfn << PAGE_SHIFT);
448 initrd_start = 0;
449 }
450 }
451 #endif
452 #ifdef CONFIG_KEXEC
453 if (crashk_res.start != crashk_res.end)
454 reserve_bootmem(crashk_res.start,
455 crashk_res.end - crashk_res.start + 1);
456 #endif
457 }
458
459 /*
460 * The node 0 pgdat is initialized before all of these because
461 * it's needed for bootmem. node>0 pgdats have their virtual
462 * space allocated before the pagetables are in place to access
463 * them, so they can't be cleared then.
464 *
465 * This should all compile down to nothing when NUMA is off.
466 */
467 static void __init remapped_pgdat_init(void)
468 {
469 int nid;
470
471 for_each_online_node(nid) {
472 if (nid != 0)
473 memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
474 }
475 }
476
477 #ifdef CONFIG_MCA
478 static void set_mca_bus(int x)
479 {
480 MCA_bus = x;
481 }
482 #else
483 static void set_mca_bus(int x) { }
484 #endif
485
486 /* Overridden in paravirt.c if CONFIG_PARAVIRT */
487 char * __init __attribute__((weak)) memory_setup(void)
488 {
489 return machine_specific_memory_setup();
490 }
491
492 /*
493 * Determine if we were loaded by an EFI loader. If so, then we have also been
494 * passed the efi memmap, systab, etc., so we should use these data structures
495 * for initialization. Note, the efi init code path is determined by the
496 * global efi_enabled. This allows the same kernel image to be used on existing
497 * systems (with a traditional BIOS) as well as on EFI systems.
498 */
499 void __init setup_arch(char **cmdline_p)
500 {
501 unsigned long max_low_pfn;
502
503 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
504 pre_setup_arch_hook();
505 early_cpu_init();
506
507 /*
508 * FIXME: This isn't an official loader_type right
509 * now but does currently work with elilo.
510 * If we were configured as an EFI kernel, check to make
511 * sure that we were loaded correctly from elilo and that
512 * the system table is valid. If not, then initialize normally.
513 */
514 #ifdef CONFIG_EFI
515 if ((LOADER_TYPE == 0x50) && EFI_SYSTAB)
516 efi_enabled = 1;
517 #endif
518
519 ROOT_DEV = old_decode_dev(ORIG_ROOT_DEV);
520 screen_info = SCREEN_INFO;
521 edid_info = EDID_INFO;
522 apm_info.bios = APM_BIOS_INFO;
523 ist_info = IST_INFO;
524 saved_videomode = VIDEO_MODE;
525 if( SYS_DESC_TABLE.length != 0 ) {
526 set_mca_bus(SYS_DESC_TABLE.table[3] & 0x2);
527 machine_id = SYS_DESC_TABLE.table[0];
528 machine_submodel_id = SYS_DESC_TABLE.table[1];
529 BIOS_revision = SYS_DESC_TABLE.table[2];
530 }
531 bootloader_type = LOADER_TYPE;
532
533 #ifdef CONFIG_BLK_DEV_RAM
534 rd_image_start = RAMDISK_FLAGS & RAMDISK_IMAGE_START_MASK;
535 rd_prompt = ((RAMDISK_FLAGS & RAMDISK_PROMPT_FLAG) != 0);
536 rd_doload = ((RAMDISK_FLAGS & RAMDISK_LOAD_FLAG) != 0);
537 #endif
538 ARCH_SETUP
539 if (efi_enabled)
540 efi_init();
541 else {
542 printk(KERN_INFO "BIOS-provided physical RAM map:\n");
543 print_memory_map(memory_setup());
544 }
545
546 copy_edd();
547
548 if (!MOUNT_ROOT_RDONLY)
549 root_mountflags &= ~MS_RDONLY;
550 init_mm.start_code = (unsigned long) _text;
551 init_mm.end_code = (unsigned long) _etext;
552 init_mm.end_data = (unsigned long) _edata;
553 init_mm.brk = init_pg_tables_end + PAGE_OFFSET;
554
555 code_resource.start = virt_to_phys(_text);
556 code_resource.end = virt_to_phys(_etext)-1;
557 data_resource.start = virt_to_phys(_etext);
558 data_resource.end = virt_to_phys(_edata)-1;
559
560 parse_early_param();
561
562 if (user_defined_memmap) {
563 printk(KERN_INFO "user-defined physical RAM map:\n");
564 print_memory_map("user");
565 }
566
567 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
568 *cmdline_p = command_line;
569
570 max_low_pfn = setup_memory();
571
572 #ifdef CONFIG_VMI
573 /*
574 * Must be after max_low_pfn is determined, and before kernel
575 * pagetables are setup.
576 */
577 vmi_init();
578 #endif
579
580 /*
581 * NOTE: before this point _nobody_ is allowed to allocate
582 * any memory using the bootmem allocator. Although the
583 * alloctor is now initialised only the first 8Mb of the kernel
584 * virtual address space has been mapped. All allocations before
585 * paging_init() has completed must use the alloc_bootmem_low_pages()
586 * variant (which allocates DMA'able memory) and care must be taken
587 * not to exceed the 8Mb limit.
588 */
589
590 #ifdef CONFIG_SMP
591 smp_alloc_memory(); /* AP processor realmode stacks in low memory*/
592 #endif
593 paging_init();
594 remapped_pgdat_init();
595 sparse_init();
596 zone_sizes_init();
597
598 /*
599 * NOTE: at this point the bootmem allocator is fully available.
600 */
601
602 paravirt_post_allocator_init();
603
604 dmi_scan_machine();
605
606 #ifdef CONFIG_X86_GENERICARCH
607 generic_apic_probe();
608 #endif
609 if (efi_enabled)
610 efi_map_memmap();
611
612 #ifdef CONFIG_ACPI
613 /*
614 * Parse the ACPI tables for possible boot-time SMP configuration.
615 */
616 acpi_boot_table_init();
617 #endif
618
619 #ifdef CONFIG_PCI
620 #ifdef CONFIG_X86_IO_APIC
621 check_acpi_pci(); /* Checks more than just ACPI actually */
622 #endif
623 #endif
624
625 #ifdef CONFIG_ACPI
626 acpi_boot_init();
627
628 #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC)
629 if (def_to_bigsmp)
630 printk(KERN_WARNING "More than 8 CPUs detected and "
631 "CONFIG_X86_PC cannot handle it.\nUse "
632 "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
633 #endif
634 #endif
635 #ifdef CONFIG_X86_LOCAL_APIC
636 if (smp_found_config)
637 get_smp_config();
638 #endif
639
640 e820_register_memory();
641 e820_mark_nosave_regions();
642
643 #ifdef CONFIG_VT
644 #if defined(CONFIG_VGA_CONSOLE)
645 if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
646 conswitchp = &vga_con;
647 #elif defined(CONFIG_DUMMY_CONSOLE)
648 conswitchp = &dummy_con;
649 #endif
650 #endif
651 }
This page took 0.170829 seconds and 6 git commands to generate.