diff options
Diffstat (limited to 'arch/x86/kernel/setup.c')
-rw-r--r-- | arch/x86/kernel/setup.c | 925 |
1 files changed, 837 insertions, 88 deletions
diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c index 6f80b852a19..ec952aa5394 100644 --- a/arch/x86/kernel/setup.c +++ b/arch/x86/kernel/setup.c | |||
@@ -1,139 +1,888 @@ | |||
1 | #include <linux/kernel.h> | 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> | ||
2 | #include <linux/module.h> | 38 | #include <linux/module.h> |
39 | #include <linux/efi.h> | ||
3 | #include <linux/init.h> | 40 | #include <linux/init.h> |
4 | #include <linux/bootmem.h> | 41 | #include <linux/edd.h> |
42 | #include <linux/iscsi_ibft.h> | ||
43 | #include <linux/nodemask.h> | ||
44 | #include <linux/kexec.h> | ||
45 | #include <linux/dmi.h> | ||
46 | #include <linux/pfn.h> | ||
47 | #include <linux/pci.h> | ||
48 | #include <asm/pci-direct.h> | ||
49 | #include <linux/init_ohci1394_dma.h> | ||
50 | #include <linux/kvm_para.h> | ||
51 | |||
52 | #include <linux/errno.h> | ||
53 | #include <linux/kernel.h> | ||
54 | #include <linux/stddef.h> | ||
55 | #include <linux/unistd.h> | ||
56 | #include <linux/ptrace.h> | ||
57 | #include <linux/slab.h> | ||
58 | #include <linux/user.h> | ||
59 | #include <linux/delay.h> | ||
60 | |||
61 | #include <linux/kallsyms.h> | ||
62 | #include <linux/cpufreq.h> | ||
63 | #include <linux/dma-mapping.h> | ||
64 | #include <linux/ctype.h> | ||
65 | #include <linux/uaccess.h> | ||
66 | |||
5 | #include <linux/percpu.h> | 67 | #include <linux/percpu.h> |
6 | #include <asm/smp.h> | 68 | #include <linux/crash_dump.h> |
7 | #include <asm/percpu.h> | 69 | |
70 | #include <video/edid.h> | ||
71 | |||
72 | #include <asm/mtrr.h> | ||
73 | #include <asm/apic.h> | ||
74 | #include <asm/e820.h> | ||
75 | #include <asm/mpspec.h> | ||
76 | #include <asm/setup.h> | ||
77 | #include <asm/arch_hooks.h> | ||
78 | #include <asm/efi.h> | ||
8 | #include <asm/sections.h> | 79 | #include <asm/sections.h> |
80 | #include <asm/dmi.h> | ||
81 | #include <asm/io_apic.h> | ||
82 | #include <asm/ist.h> | ||
83 | #include <asm/vmi.h> | ||
84 | #include <setup_arch.h> | ||
85 | #include <asm/bios_ebda.h> | ||
86 | #include <asm/cacheflush.h> | ||
9 | #include <asm/processor.h> | 87 | #include <asm/processor.h> |
10 | #include <asm/setup.h> | 88 | #include <asm/bugs.h> |
89 | |||
90 | #include <asm/system.h> | ||
91 | #include <asm/vsyscall.h> | ||
92 | #include <asm/smp.h> | ||
93 | #include <asm/desc.h> | ||
94 | #include <asm/dma.h> | ||
95 | #include <asm/iommu.h> | ||
96 | #include <asm/mmu_context.h> | ||
97 | #include <asm/proto.h> | ||
98 | |||
99 | #include <mach_apic.h> | ||
100 | #include <asm/paravirt.h> | ||
101 | |||
102 | #include <asm/percpu.h> | ||
11 | #include <asm/topology.h> | 103 | #include <asm/topology.h> |
12 | #include <asm/mpspec.h> | ||
13 | #include <asm/apicdef.h> | 104 | #include <asm/apicdef.h> |
105 | #ifdef CONFIG_X86_64 | ||
106 | #include <asm/numa_64.h> | ||
107 | #endif | ||
14 | 108 | ||
15 | #ifdef CONFIG_X86_LOCAL_APIC | 109 | #ifndef ARCH_SETUP |
16 | unsigned int num_processors; | 110 | #define ARCH_SETUP |
17 | unsigned disabled_cpus __cpuinitdata; | 111 | #endif |
18 | /* Processor that is doing the boot up */ | ||
19 | unsigned int boot_cpu_physical_apicid = -1U; | ||
20 | EXPORT_SYMBOL(boot_cpu_physical_apicid); | ||
21 | 112 | ||
22 | DEFINE_PER_CPU(u16, x86_cpu_to_apicid) = BAD_APICID; | 113 | #ifndef CONFIG_DEBUG_BOOT_PARAMS |
23 | EXPORT_PER_CPU_SYMBOL(x86_cpu_to_apicid); | 114 | struct boot_params __initdata boot_params; |
115 | #else | ||
116 | struct boot_params boot_params; | ||
117 | #endif | ||
24 | 118 | ||
25 | /* Bitmask of physically existing CPUs */ | 119 | /* |
26 | physid_mask_t phys_cpu_present_map; | 120 | * Machine setup.. |
121 | */ | ||
122 | static struct resource data_resource = { | ||
123 | .name = "Kernel data", | ||
124 | .start = 0, | ||
125 | .end = 0, | ||
126 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | ||
127 | }; | ||
128 | |||
129 | static struct resource code_resource = { | ||
130 | .name = "Kernel code", | ||
131 | .start = 0, | ||
132 | .end = 0, | ||
133 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | ||
134 | }; | ||
135 | |||
136 | static struct resource bss_resource = { | ||
137 | .name = "Kernel bss", | ||
138 | .start = 0, | ||
139 | .end = 0, | ||
140 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | ||
141 | }; | ||
142 | |||
143 | |||
144 | #ifdef CONFIG_X86_32 | ||
145 | /* This value is set up by the early boot code to point to the value | ||
146 | immediately after the boot time page tables. It contains a *physical* | ||
147 | address, and must not be in the .bss segment! */ | ||
148 | unsigned long init_pg_tables_start __initdata = ~0UL; | ||
149 | unsigned long init_pg_tables_end __initdata = ~0UL; | ||
150 | |||
151 | static struct resource video_ram_resource = { | ||
152 | .name = "Video RAM area", | ||
153 | .start = 0xa0000, | ||
154 | .end = 0xbffff, | ||
155 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | ||
156 | }; | ||
157 | |||
158 | /* cpu data as detected by the assembly code in head.S */ | ||
159 | struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1}; | ||
160 | /* common cpu data for all cpus */ | ||
161 | struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1}; | ||
162 | EXPORT_SYMBOL(boot_cpu_data); | ||
163 | static void set_mca_bus(int x) | ||
164 | { | ||
165 | #ifdef CONFIG_MCA | ||
166 | MCA_bus = x; | ||
167 | #endif | ||
168 | } | ||
169 | |||
170 | unsigned int def_to_bigsmp; | ||
171 | |||
172 | /* for MCA, but anyone else can use it if they want */ | ||
173 | unsigned int machine_id; | ||
174 | unsigned int machine_submodel_id; | ||
175 | unsigned int BIOS_revision; | ||
176 | |||
177 | struct apm_info apm_info; | ||
178 | EXPORT_SYMBOL(apm_info); | ||
179 | |||
180 | #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \ | ||
181 | defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE) | ||
182 | struct ist_info ist_info; | ||
183 | EXPORT_SYMBOL(ist_info); | ||
184 | #else | ||
185 | struct ist_info ist_info; | ||
27 | #endif | 186 | #endif |
28 | 187 | ||
29 | #if defined(CONFIG_HAVE_SETUP_PER_CPU_AREA) && defined(CONFIG_X86_SMP) | 188 | #else |
189 | struct cpuinfo_x86 boot_cpu_data __read_mostly; | ||
190 | EXPORT_SYMBOL(boot_cpu_data); | ||
191 | #endif | ||
192 | |||
193 | |||
194 | #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64) | ||
195 | unsigned long mmu_cr4_features; | ||
196 | #else | ||
197 | unsigned long mmu_cr4_features = X86_CR4_PAE; | ||
198 | #endif | ||
199 | |||
200 | /* Boot loader ID as an integer, for the benefit of proc_dointvec */ | ||
201 | int bootloader_type; | ||
202 | |||
30 | /* | 203 | /* |
31 | * Copy data used in early init routines from the initial arrays to the | 204 | * Early DMI memory |
32 | * per cpu data areas. These arrays then become expendable and the | ||
33 | * *_early_ptr's are zeroed indicating that the static arrays are gone. | ||
34 | */ | 205 | */ |
35 | static void __init setup_per_cpu_maps(void) | 206 | int dmi_alloc_index; |
207 | char dmi_alloc_data[DMI_MAX_DATA]; | ||
208 | |||
209 | /* | ||
210 | * Setup options | ||
211 | */ | ||
212 | struct screen_info screen_info; | ||
213 | EXPORT_SYMBOL(screen_info); | ||
214 | struct edid_info edid_info; | ||
215 | EXPORT_SYMBOL_GPL(edid_info); | ||
216 | |||
217 | extern int root_mountflags; | ||
218 | |||
219 | unsigned long saved_video_mode; | ||
220 | |||
221 | #define RAMDISK_IMAGE_START_MASK 0x07FF | ||
222 | #define RAMDISK_PROMPT_FLAG 0x8000 | ||
223 | #define RAMDISK_LOAD_FLAG 0x4000 | ||
224 | |||
225 | static char __initdata command_line[COMMAND_LINE_SIZE]; | ||
226 | |||
227 | #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE) | ||
228 | struct edd edd; | ||
229 | #ifdef CONFIG_EDD_MODULE | ||
230 | EXPORT_SYMBOL(edd); | ||
231 | #endif | ||
232 | /** | ||
233 | * copy_edd() - Copy the BIOS EDD information | ||
234 | * from boot_params into a safe place. | ||
235 | * | ||
236 | */ | ||
237 | static inline void copy_edd(void) | ||
238 | { | ||
239 | memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer, | ||
240 | sizeof(edd.mbr_signature)); | ||
241 | memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info)); | ||
242 | edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries; | ||
243 | edd.edd_info_nr = boot_params.eddbuf_entries; | ||
244 | } | ||
245 | #else | ||
246 | static inline void copy_edd(void) | ||
247 | { | ||
248 | } | ||
249 | #endif | ||
250 | |||
251 | #ifdef CONFIG_BLK_DEV_INITRD | ||
252 | |||
253 | #ifdef CONFIG_X86_32 | ||
254 | |||
255 | #define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT) | ||
256 | static void __init relocate_initrd(void) | ||
36 | { | 257 | { |
37 | int cpu; | ||
38 | 258 | ||
39 | for_each_possible_cpu(cpu) { | 259 | u64 ramdisk_image = boot_params.hdr.ramdisk_image; |
40 | per_cpu(x86_cpu_to_apicid, cpu) = x86_cpu_to_apicid_init[cpu]; | 260 | u64 ramdisk_size = boot_params.hdr.ramdisk_size; |
41 | per_cpu(x86_bios_cpu_apicid, cpu) = | 261 | u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT; |
42 | x86_bios_cpu_apicid_init[cpu]; | 262 | u64 ramdisk_here; |
43 | #ifdef CONFIG_NUMA | 263 | unsigned long slop, clen, mapaddr; |
44 | per_cpu(x86_cpu_to_node_map, cpu) = | 264 | char *p, *q; |
45 | x86_cpu_to_node_map_init[cpu]; | 265 | |
266 | /* We need to move the initrd down into lowmem */ | ||
267 | ramdisk_here = find_e820_area(0, end_of_lowmem, ramdisk_size, | ||
268 | PAGE_SIZE); | ||
269 | |||
270 | if (ramdisk_here == -1ULL) | ||
271 | panic("Cannot find place for new RAMDISK of size %lld\n", | ||
272 | ramdisk_size); | ||
273 | |||
274 | /* Note: this includes all the lowmem currently occupied by | ||
275 | the initrd, we rely on that fact to keep the data intact. */ | ||
276 | reserve_early(ramdisk_here, ramdisk_here + ramdisk_size, | ||
277 | "NEW RAMDISK"); | ||
278 | initrd_start = ramdisk_here + PAGE_OFFSET; | ||
279 | initrd_end = initrd_start + ramdisk_size; | ||
280 | printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n", | ||
281 | ramdisk_here, ramdisk_here + ramdisk_size); | ||
282 | |||
283 | q = (char *)initrd_start; | ||
284 | |||
285 | /* Copy any lowmem portion of the initrd */ | ||
286 | if (ramdisk_image < end_of_lowmem) { | ||
287 | clen = end_of_lowmem - ramdisk_image; | ||
288 | p = (char *)__va(ramdisk_image); | ||
289 | memcpy(q, p, clen); | ||
290 | q += clen; | ||
291 | ramdisk_image += clen; | ||
292 | ramdisk_size -= clen; | ||
293 | } | ||
294 | |||
295 | /* Copy the highmem portion of the initrd */ | ||
296 | while (ramdisk_size) { | ||
297 | slop = ramdisk_image & ~PAGE_MASK; | ||
298 | clen = ramdisk_size; | ||
299 | if (clen > MAX_MAP_CHUNK-slop) | ||
300 | clen = MAX_MAP_CHUNK-slop; | ||
301 | mapaddr = ramdisk_image & PAGE_MASK; | ||
302 | p = early_ioremap(mapaddr, clen+slop); | ||
303 | memcpy(q, p+slop, clen); | ||
304 | early_iounmap(p, clen+slop); | ||
305 | q += clen; | ||
306 | ramdisk_image += clen; | ||
307 | ramdisk_size -= clen; | ||
308 | } | ||
309 | /* high pages is not converted by early_res_to_bootmem */ | ||
310 | ramdisk_image = boot_params.hdr.ramdisk_image; | ||
311 | ramdisk_size = boot_params.hdr.ramdisk_size; | ||
312 | printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to" | ||
313 | " %08llx - %08llx\n", | ||
314 | ramdisk_image, ramdisk_image + ramdisk_size - 1, | ||
315 | ramdisk_here, ramdisk_here + ramdisk_size - 1); | ||
316 | } | ||
46 | #endif | 317 | #endif |
318 | |||
319 | static void __init reserve_initrd(void) | ||
320 | { | ||
321 | u64 ramdisk_image = boot_params.hdr.ramdisk_image; | ||
322 | u64 ramdisk_size = boot_params.hdr.ramdisk_size; | ||
323 | u64 ramdisk_end = ramdisk_image + ramdisk_size; | ||
324 | u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT; | ||
325 | |||
326 | if (!boot_params.hdr.type_of_loader || | ||
327 | !ramdisk_image || !ramdisk_size) | ||
328 | return; /* No initrd provided by bootloader */ | ||
329 | |||
330 | initrd_start = 0; | ||
331 | |||
332 | if (ramdisk_size >= (end_of_lowmem>>1)) { | ||
333 | free_early(ramdisk_image, ramdisk_end); | ||
334 | printk(KERN_ERR "initrd too large to handle, " | ||
335 | "disabling initrd\n"); | ||
336 | return; | ||
337 | } | ||
338 | |||
339 | printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image, | ||
340 | ramdisk_end); | ||
341 | |||
342 | |||
343 | if (ramdisk_end <= end_of_lowmem) { | ||
344 | /* All in lowmem, easy case */ | ||
345 | /* | ||
346 | * don't need to reserve again, already reserved early | ||
347 | * in i386_start_kernel | ||
348 | */ | ||
349 | initrd_start = ramdisk_image + PAGE_OFFSET; | ||
350 | initrd_end = initrd_start + ramdisk_size; | ||
351 | return; | ||
47 | } | 352 | } |
48 | 353 | ||
49 | /* indicate the early static arrays will soon be gone */ | 354 | #ifdef CONFIG_X86_32 |
50 | x86_cpu_to_apicid_early_ptr = NULL; | 355 | relocate_initrd(); |
51 | x86_bios_cpu_apicid_early_ptr = NULL; | 356 | #else |
52 | #ifdef CONFIG_NUMA | 357 | printk(KERN_ERR "initrd extends beyond end of memory " |
53 | x86_cpu_to_node_map_early_ptr = NULL; | 358 | "(0x%08llx > 0x%08llx)\ndisabling initrd\n", |
359 | ramdisk_end, end_of_lowmem); | ||
360 | initrd_start = 0; | ||
54 | #endif | 361 | #endif |
362 | free_early(ramdisk_image, ramdisk_end); | ||
55 | } | 363 | } |
364 | #else | ||
365 | static void __init reserve_initrd(void) | ||
366 | { | ||
367 | } | ||
368 | #endif /* CONFIG_BLK_DEV_INITRD */ | ||
369 | |||
370 | static void __init parse_setup_data(void) | ||
371 | { | ||
372 | struct setup_data *data; | ||
373 | u64 pa_data; | ||
374 | |||
375 | if (boot_params.hdr.version < 0x0209) | ||
376 | return; | ||
377 | pa_data = boot_params.hdr.setup_data; | ||
378 | while (pa_data) { | ||
379 | data = early_ioremap(pa_data, PAGE_SIZE); | ||
380 | switch (data->type) { | ||
381 | case SETUP_E820_EXT: | ||
382 | parse_e820_ext(data, pa_data); | ||
383 | break; | ||
384 | default: | ||
385 | break; | ||
386 | } | ||
387 | pa_data = data->next; | ||
388 | early_iounmap(data, PAGE_SIZE); | ||
389 | } | ||
390 | } | ||
391 | |||
392 | static void __init e820_reserve_setup_data(void) | ||
393 | { | ||
394 | struct setup_data *data; | ||
395 | u64 pa_data; | ||
396 | int found = 0; | ||
397 | |||
398 | if (boot_params.hdr.version < 0x0209) | ||
399 | return; | ||
400 | pa_data = boot_params.hdr.setup_data; | ||
401 | while (pa_data) { | ||
402 | data = early_ioremap(pa_data, sizeof(*data)); | ||
403 | e820_update_range(pa_data, sizeof(*data)+data->len, | ||
404 | E820_RAM, E820_RESERVED_KERN); | ||
405 | found = 1; | ||
406 | pa_data = data->next; | ||
407 | early_iounmap(data, sizeof(*data)); | ||
408 | } | ||
409 | if (!found) | ||
410 | return; | ||
56 | 411 | ||
57 | #ifdef CONFIG_HAVE_CPUMASK_OF_CPU_MAP | 412 | sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); |
58 | cpumask_t *cpumask_of_cpu_map __read_mostly; | 413 | memcpy(&e820_saved, &e820, sizeof(struct e820map)); |
59 | EXPORT_SYMBOL(cpumask_of_cpu_map); | 414 | printk(KERN_INFO "extended physical RAM map:\n"); |
415 | e820_print_map("reserve setup_data"); | ||
416 | } | ||
60 | 417 | ||
61 | /* requires nr_cpu_ids to be initialized */ | 418 | static void __init reserve_early_setup_data(void) |
62 | static void __init setup_cpumask_of_cpu(void) | ||
63 | { | 419 | { |
64 | int i; | 420 | struct setup_data *data; |
421 | u64 pa_data; | ||
422 | char buf[32]; | ||
423 | |||
424 | if (boot_params.hdr.version < 0x0209) | ||
425 | return; | ||
426 | pa_data = boot_params.hdr.setup_data; | ||
427 | while (pa_data) { | ||
428 | data = early_ioremap(pa_data, sizeof(*data)); | ||
429 | sprintf(buf, "setup data %x", data->type); | ||
430 | reserve_early(pa_data, pa_data+sizeof(*data)+data->len, buf); | ||
431 | pa_data = data->next; | ||
432 | early_iounmap(data, sizeof(*data)); | ||
433 | } | ||
434 | } | ||
435 | |||
436 | /* | ||
437 | * --------- Crashkernel reservation ------------------------------ | ||
438 | */ | ||
439 | |||
440 | #ifdef CONFIG_KEXEC | ||
441 | |||
442 | /** | ||
443 | * Reserve @size bytes of crashkernel memory at any suitable offset. | ||
444 | * | ||
445 | * @size: Size of the crashkernel memory to reserve. | ||
446 | * Returns the base address on success, and -1ULL on failure. | ||
447 | */ | ||
448 | unsigned long long find_and_reserve_crashkernel(unsigned long long size) | ||
449 | { | ||
450 | const unsigned long long alignment = 16<<20; /* 16M */ | ||
451 | unsigned long long start = 0LL; | ||
452 | |||
453 | while (1) { | ||
454 | int ret; | ||
455 | |||
456 | start = find_e820_area(start, ULONG_MAX, size, alignment); | ||
457 | if (start == -1ULL) | ||
458 | return start; | ||
459 | |||
460 | /* try to reserve it */ | ||
461 | ret = reserve_bootmem_generic(start, size, BOOTMEM_EXCLUSIVE); | ||
462 | if (ret >= 0) | ||
463 | return start; | ||
65 | 464 | ||
66 | /* alloc_bootmem zeroes memory */ | 465 | start += alignment; |
67 | cpumask_of_cpu_map = alloc_bootmem_low(sizeof(cpumask_t) * nr_cpu_ids); | 466 | } |
68 | for (i = 0; i < nr_cpu_ids; i++) | 467 | } |
69 | cpu_set(i, cpumask_of_cpu_map[i]); | 468 | |
469 | static inline unsigned long long get_total_mem(void) | ||
470 | { | ||
471 | unsigned long long total; | ||
472 | |||
473 | total = max_low_pfn - min_low_pfn; | ||
474 | #ifdef CONFIG_HIGHMEM | ||
475 | total += highend_pfn - highstart_pfn; | ||
476 | #endif | ||
477 | |||
478 | return total << PAGE_SHIFT; | ||
479 | } | ||
480 | |||
481 | static void __init reserve_crashkernel(void) | ||
482 | { | ||
483 | unsigned long long total_mem; | ||
484 | unsigned long long crash_size, crash_base; | ||
485 | int ret; | ||
486 | |||
487 | total_mem = get_total_mem(); | ||
488 | |||
489 | ret = parse_crashkernel(boot_command_line, total_mem, | ||
490 | &crash_size, &crash_base); | ||
491 | if (ret != 0 || crash_size <= 0) | ||
492 | return; | ||
493 | |||
494 | /* 0 means: find the address automatically */ | ||
495 | if (crash_base <= 0) { | ||
496 | crash_base = find_and_reserve_crashkernel(crash_size); | ||
497 | if (crash_base == -1ULL) { | ||
498 | pr_info("crashkernel reservation failed. " | ||
499 | "No suitable area found.\n"); | ||
500 | return; | ||
501 | } | ||
502 | } else { | ||
503 | ret = reserve_bootmem_generic(crash_base, crash_size, | ||
504 | BOOTMEM_EXCLUSIVE); | ||
505 | if (ret < 0) { | ||
506 | pr_info("crashkernel reservation failed - " | ||
507 | "memory is in use\n"); | ||
508 | return; | ||
509 | } | ||
510 | } | ||
511 | |||
512 | printk(KERN_INFO "Reserving %ldMB of memory at %ldMB " | ||
513 | "for crashkernel (System RAM: %ldMB)\n", | ||
514 | (unsigned long)(crash_size >> 20), | ||
515 | (unsigned long)(crash_base >> 20), | ||
516 | (unsigned long)(total_mem >> 20)); | ||
517 | |||
518 | crashk_res.start = crash_base; | ||
519 | crashk_res.end = crash_base + crash_size - 1; | ||
520 | insert_resource(&iomem_resource, &crashk_res); | ||
70 | } | 521 | } |
71 | #else | 522 | #else |
72 | static inline void setup_cpumask_of_cpu(void) { } | 523 | static void __init reserve_crashkernel(void) |
524 | { | ||
525 | } | ||
73 | #endif | 526 | #endif |
74 | 527 | ||
75 | #ifdef CONFIG_X86_32 | 528 | static struct resource standard_io_resources[] = { |
76 | /* | 529 | { .name = "dma1", .start = 0x00, .end = 0x1f, |
77 | * Great future not-so-futuristic plan: make i386 and x86_64 do it | 530 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, |
78 | * the same way | 531 | { .name = "pic1", .start = 0x20, .end = 0x21, |
532 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
533 | { .name = "timer0", .start = 0x40, .end = 0x43, | ||
534 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
535 | { .name = "timer1", .start = 0x50, .end = 0x53, | ||
536 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
537 | { .name = "keyboard", .start = 0x60, .end = 0x60, | ||
538 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
539 | { .name = "keyboard", .start = 0x64, .end = 0x64, | ||
540 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
541 | { .name = "dma page reg", .start = 0x80, .end = 0x8f, | ||
542 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
543 | { .name = "pic2", .start = 0xa0, .end = 0xa1, | ||
544 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
545 | { .name = "dma2", .start = 0xc0, .end = 0xdf, | ||
546 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
547 | { .name = "fpu", .start = 0xf0, .end = 0xff, | ||
548 | .flags = IORESOURCE_BUSY | IORESOURCE_IO } | ||
549 | }; | ||
550 | |||
551 | static void __init reserve_standard_io_resources(void) | ||
552 | { | ||
553 | int i; | ||
554 | |||
555 | /* request I/O space for devices used on all i[345]86 PCs */ | ||
556 | for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++) | ||
557 | request_resource(&ioport_resource, &standard_io_resources[i]); | ||
558 | |||
559 | } | ||
560 | |||
561 | #ifdef CONFIG_PROC_VMCORE | ||
562 | /* elfcorehdr= specifies the location of elf core header | ||
563 | * stored by the crashed kernel. This option will be passed | ||
564 | * by kexec loader to the capture kernel. | ||
79 | */ | 565 | */ |
80 | unsigned long __per_cpu_offset[NR_CPUS] __read_mostly; | 566 | static int __init setup_elfcorehdr(char *arg) |
81 | EXPORT_SYMBOL(__per_cpu_offset); | 567 | { |
568 | char *end; | ||
569 | if (!arg) | ||
570 | return -EINVAL; | ||
571 | elfcorehdr_addr = memparse(arg, &end); | ||
572 | return end > arg ? 0 : -EINVAL; | ||
573 | } | ||
574 | early_param("elfcorehdr", setup_elfcorehdr); | ||
82 | #endif | 575 | #endif |
83 | 576 | ||
577 | static struct x86_quirks default_x86_quirks __initdata; | ||
578 | |||
579 | struct x86_quirks *x86_quirks __initdata = &default_x86_quirks; | ||
580 | |||
581 | /* | ||
582 | * Determine if we were loaded by an EFI loader. If so, then we have also been | ||
583 | * passed the efi memmap, systab, etc., so we should use these data structures | ||
584 | * for initialization. Note, the efi init code path is determined by the | ||
585 | * global efi_enabled. This allows the same kernel image to be used on existing | ||
586 | * systems (with a traditional BIOS) as well as on EFI systems. | ||
587 | */ | ||
84 | /* | 588 | /* |
85 | * Great future plan: | 589 | * setup_arch - architecture-specific boot-time initializations |
86 | * Declare PDA itself and support (irqstack,tss,pgd) as per cpu data. | 590 | * |
87 | * Always point %gs to its beginning | 591 | * Note: On x86_64, fixmaps are ready for use even before this is called. |
88 | */ | 592 | */ |
89 | void __init setup_per_cpu_areas(void) | 593 | |
594 | void __init setup_arch(char **cmdline_p) | ||
90 | { | 595 | { |
91 | int i, highest_cpu = 0; | 596 | #ifdef CONFIG_X86_32 |
92 | unsigned long size; | 597 | memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data)); |
598 | visws_early_detect(); | ||
599 | pre_setup_arch_hook(); | ||
600 | early_cpu_init(); | ||
601 | #else | ||
602 | printk(KERN_INFO "Command line: %s\n", boot_command_line); | ||
603 | #endif | ||
93 | 604 | ||
94 | #ifdef CONFIG_HOTPLUG_CPU | 605 | early_ioremap_init(); |
95 | prefill_possible_map(); | 606 | |
607 | ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev); | ||
608 | screen_info = boot_params.screen_info; | ||
609 | edid_info = boot_params.edid_info; | ||
610 | #ifdef CONFIG_X86_32 | ||
611 | apm_info.bios = boot_params.apm_bios_info; | ||
612 | ist_info = boot_params.ist_info; | ||
613 | if (boot_params.sys_desc_table.length != 0) { | ||
614 | set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2); | ||
615 | machine_id = boot_params.sys_desc_table.table[0]; | ||
616 | machine_submodel_id = boot_params.sys_desc_table.table[1]; | ||
617 | BIOS_revision = boot_params.sys_desc_table.table[2]; | ||
618 | } | ||
619 | #endif | ||
620 | saved_video_mode = boot_params.hdr.vid_mode; | ||
621 | bootloader_type = boot_params.hdr.type_of_loader; | ||
622 | |||
623 | #ifdef CONFIG_BLK_DEV_RAM | ||
624 | rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK; | ||
625 | rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0); | ||
626 | rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0); | ||
627 | #endif | ||
628 | #ifdef CONFIG_EFI | ||
629 | if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature, | ||
630 | #ifdef CONFIG_X86_32 | ||
631 | "EL32", | ||
632 | #else | ||
633 | "EL64", | ||
96 | #endif | 634 | #endif |
635 | 4)) { | ||
636 | efi_enabled = 1; | ||
637 | efi_reserve_early(); | ||
638 | } | ||
639 | #endif | ||
640 | |||
641 | ARCH_SETUP | ||
642 | |||
643 | setup_memory_map(); | ||
644 | parse_setup_data(); | ||
645 | /* update the e820_saved too */ | ||
646 | e820_reserve_setup_data(); | ||
97 | 647 | ||
98 | /* Copy section for each CPU (we discard the original) */ | 648 | copy_edd(); |
99 | size = PERCPU_ENOUGH_ROOM; | ||
100 | printk(KERN_INFO "PERCPU: Allocating %lu bytes of per cpu data\n", | ||
101 | size); | ||
102 | 649 | ||
103 | for_each_possible_cpu(i) { | 650 | if (!boot_params.hdr.root_flags) |
104 | char *ptr; | 651 | root_mountflags &= ~MS_RDONLY; |
105 | #ifndef CONFIG_NEED_MULTIPLE_NODES | 652 | init_mm.start_code = (unsigned long) _text; |
106 | ptr = alloc_bootmem_pages(size); | 653 | init_mm.end_code = (unsigned long) _etext; |
654 | init_mm.end_data = (unsigned long) _edata; | ||
655 | #ifdef CONFIG_X86_32 | ||
656 | init_mm.brk = init_pg_tables_end + PAGE_OFFSET; | ||
107 | #else | 657 | #else |
108 | int node = early_cpu_to_node(i); | 658 | init_mm.brk = (unsigned long) &_end; |
109 | if (!node_online(node) || !NODE_DATA(node)) { | ||
110 | ptr = alloc_bootmem_pages(size); | ||
111 | printk(KERN_INFO | ||
112 | "cpu %d has no node or node-local memory\n", i); | ||
113 | } | ||
114 | else | ||
115 | ptr = alloc_bootmem_pages_node(NODE_DATA(node), size); | ||
116 | #endif | 659 | #endif |
117 | if (!ptr) | 660 | |
118 | panic("Cannot allocate cpu data for CPU %d\n", i); | 661 | code_resource.start = virt_to_phys(_text); |
662 | code_resource.end = virt_to_phys(_etext)-1; | ||
663 | data_resource.start = virt_to_phys(_etext); | ||
664 | data_resource.end = virt_to_phys(_edata)-1; | ||
665 | bss_resource.start = virt_to_phys(&__bss_start); | ||
666 | bss_resource.end = virt_to_phys(&__bss_stop)-1; | ||
667 | |||
119 | #ifdef CONFIG_X86_64 | 668 | #ifdef CONFIG_X86_64 |
120 | cpu_pda(i)->data_offset = ptr - __per_cpu_start; | 669 | early_cpu_init(); |
670 | #endif | ||
671 | strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE); | ||
672 | *cmdline_p = command_line; | ||
673 | |||
674 | parse_early_param(); | ||
675 | |||
676 | /* after early param, so could get panic from serial */ | ||
677 | reserve_early_setup_data(); | ||
678 | |||
679 | if (acpi_mps_check()) { | ||
680 | #ifdef CONFIG_X86_LOCAL_APIC | ||
681 | disable_apic = 1; | ||
682 | #endif | ||
683 | clear_cpu_cap(&boot_cpu_data, X86_FEATURE_APIC); | ||
684 | } | ||
685 | |||
686 | #ifdef CONFIG_PCI | ||
687 | if (pci_early_dump_regs) | ||
688 | early_dump_pci_devices(); | ||
689 | #endif | ||
690 | |||
691 | finish_e820_parsing(); | ||
692 | |||
693 | #ifdef CONFIG_X86_32 | ||
694 | probe_roms(); | ||
695 | #endif | ||
696 | |||
697 | /* after parse_early_param, so could debug it */ | ||
698 | insert_resource(&iomem_resource, &code_resource); | ||
699 | insert_resource(&iomem_resource, &data_resource); | ||
700 | insert_resource(&iomem_resource, &bss_resource); | ||
701 | |||
702 | if (efi_enabled) | ||
703 | efi_init(); | ||
704 | |||
705 | #ifdef CONFIG_X86_32 | ||
706 | if (ppro_with_ram_bug()) { | ||
707 | e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM, | ||
708 | E820_RESERVED); | ||
709 | sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); | ||
710 | printk(KERN_INFO "fixed physical RAM map:\n"); | ||
711 | e820_print_map("bad_ppro"); | ||
712 | } | ||
713 | #else | ||
714 | early_gart_iommu_check(); | ||
715 | #endif | ||
716 | |||
717 | /* | ||
718 | * partially used pages are not usable - thus | ||
719 | * we are rounding upwards: | ||
720 | */ | ||
721 | max_pfn = e820_end_of_ram_pfn(); | ||
722 | |||
723 | /* preallocate 4k for mptable mpc */ | ||
724 | early_reserve_e820_mpc_new(); | ||
725 | /* update e820 for memory not covered by WB MTRRs */ | ||
726 | mtrr_bp_init(); | ||
727 | if (mtrr_trim_uncached_memory(max_pfn)) | ||
728 | max_pfn = e820_end_of_ram_pfn(); | ||
729 | |||
730 | #ifdef CONFIG_X86_32 | ||
731 | /* max_low_pfn get updated here */ | ||
732 | find_low_pfn_range(); | ||
121 | #else | 733 | #else |
122 | __per_cpu_offset[i] = ptr - __per_cpu_start; | 734 | num_physpages = max_pfn; |
735 | |||
736 | check_efer(); | ||
737 | |||
738 | /* How many end-of-memory variables you have, grandma! */ | ||
739 | /* need this before calling reserve_initrd */ | ||
740 | if (max_pfn > (1UL<<(32 - PAGE_SHIFT))) | ||
741 | max_low_pfn = e820_end_of_low_ram_pfn(); | ||
742 | else | ||
743 | max_low_pfn = max_pfn; | ||
744 | |||
745 | high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1; | ||
123 | #endif | 746 | #endif |
124 | memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start); | ||
125 | 747 | ||
126 | highest_cpu = i; | 748 | /* max_pfn_mapped is updated here */ |
749 | max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT); | ||
750 | max_pfn_mapped = max_low_pfn_mapped; | ||
751 | |||
752 | #ifdef CONFIG_X86_64 | ||
753 | if (max_pfn > max_low_pfn) { | ||
754 | max_pfn_mapped = init_memory_mapping(1UL<<32, | ||
755 | max_pfn<<PAGE_SHIFT); | ||
756 | /* can we preseve max_low_pfn ?*/ | ||
757 | max_low_pfn = max_pfn; | ||
127 | } | 758 | } |
759 | #endif | ||
128 | 760 | ||
129 | nr_cpu_ids = highest_cpu + 1; | 761 | /* |
130 | printk(KERN_DEBUG "NR_CPUS: %d, nr_cpu_ids: %d\n", NR_CPUS, nr_cpu_ids); | 762 | * NOTE: On x86-32, only from this point on, fixmaps are ready for use. |
763 | */ | ||
131 | 764 | ||
132 | /* Setup percpu data maps */ | 765 | #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT |
133 | setup_per_cpu_maps(); | 766 | if (init_ohci1394_dma_early) |
767 | init_ohci1394_dma_on_all_controllers(); | ||
768 | #endif | ||
134 | 769 | ||
135 | /* Setup cpumask_of_cpu map */ | 770 | reserve_initrd(); |
136 | setup_cpumask_of_cpu(); | 771 | |
137 | } | 772 | #ifdef CONFIG_X86_64 |
773 | vsmp_init(); | ||
774 | #endif | ||
775 | |||
776 | dmi_scan_machine(); | ||
777 | |||
778 | io_delay_init(); | ||
779 | |||
780 | /* | ||
781 | * Parse the ACPI tables for possible boot-time SMP configuration. | ||
782 | */ | ||
783 | acpi_boot_table_init(); | ||
784 | |||
785 | #ifdef CONFIG_ACPI_NUMA | ||
786 | /* | ||
787 | * Parse SRAT to discover nodes. | ||
788 | */ | ||
789 | acpi_numa_init(); | ||
790 | #endif | ||
791 | |||
792 | initmem_init(0, max_pfn); | ||
793 | |||
794 | #ifdef CONFIG_X86_64 | ||
795 | dma32_reserve_bootmem(); | ||
796 | #endif | ||
138 | 797 | ||
798 | #ifdef CONFIG_ACPI_SLEEP | ||
799 | /* | ||
800 | * Reserve low memory region for sleep support. | ||
801 | */ | ||
802 | acpi_reserve_bootmem(); | ||
139 | #endif | 803 | #endif |
804 | #ifdef CONFIG_X86_FIND_SMP_CONFIG | ||
805 | /* | ||
806 | * Find and reserve possible boot-time SMP configuration: | ||
807 | */ | ||
808 | find_smp_config(); | ||
809 | #endif | ||
810 | reserve_crashkernel(); | ||
811 | |||
812 | reserve_ibft_region(); | ||
813 | |||
814 | #ifdef CONFIG_KVM_CLOCK | ||
815 | kvmclock_init(); | ||
816 | #endif | ||
817 | |||
818 | #if defined(CONFIG_VMI) && defined(CONFIG_X86_32) | ||
819 | /* | ||
820 | * Must be after max_low_pfn is determined, and before kernel | ||
821 | * pagetables are setup. | ||
822 | */ | ||
823 | vmi_init(); | ||
824 | #endif | ||
825 | |||
826 | paravirt_pagetable_setup_start(swapper_pg_dir); | ||
827 | paging_init(); | ||
828 | paravirt_pagetable_setup_done(swapper_pg_dir); | ||
829 | paravirt_post_allocator_init(); | ||
830 | |||
831 | #ifdef CONFIG_X86_64 | ||
832 | map_vsyscall(); | ||
833 | #endif | ||
834 | |||
835 | #ifdef CONFIG_X86_GENERICARCH | ||
836 | generic_apic_probe(); | ||
837 | #endif | ||
838 | |||
839 | early_quirks(); | ||
840 | |||
841 | /* | ||
842 | * Read APIC and some other early information from ACPI tables. | ||
843 | */ | ||
844 | acpi_boot_init(); | ||
845 | |||
846 | #if defined(CONFIG_X86_MPPARSE) || defined(CONFIG_X86_VISWS) | ||
847 | /* | ||
848 | * get boot-time SMP configuration: | ||
849 | */ | ||
850 | if (smp_found_config) | ||
851 | get_smp_config(); | ||
852 | #endif | ||
853 | |||
854 | prefill_possible_map(); | ||
855 | #ifdef CONFIG_X86_64 | ||
856 | init_cpu_to_node(); | ||
857 | #endif | ||
858 | |||
859 | init_apic_mappings(); | ||
860 | ioapic_init_mappings(); | ||
861 | |||
862 | #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC) && defined(CONFIG_X86_32) | ||
863 | if (def_to_bigsmp) | ||
864 | printk(KERN_WARNING "More than 8 CPUs detected and " | ||
865 | "CONFIG_X86_PC cannot handle it.\nUse " | ||
866 | "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n"); | ||
867 | #endif | ||
868 | kvm_guest_init(); | ||
869 | |||
870 | e820_reserve_resources(); | ||
871 | e820_mark_nosave_regions(max_low_pfn); | ||
872 | |||
873 | #ifdef CONFIG_X86_32 | ||
874 | request_resource(&iomem_resource, &video_ram_resource); | ||
875 | #endif | ||
876 | reserve_standard_io_resources(); | ||
877 | |||
878 | e820_setup_gap(); | ||
879 | |||
880 | #ifdef CONFIG_VT | ||
881 | #if defined(CONFIG_VGA_CONSOLE) | ||
882 | if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY)) | ||
883 | conswitchp = &vga_con; | ||
884 | #elif defined(CONFIG_DUMMY_CONSOLE) | ||
885 | conswitchp = &dummy_con; | ||
886 | #endif | ||
887 | #endif | ||
888 | } | ||