diff options
Diffstat (limited to 'arch/x86/kernel/setup.c')
-rw-r--r-- | arch/x86/kernel/setup.c | 929 |
1 files changed, 842 insertions, 87 deletions
diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c index 6f80b852a196..531b55b8e81a 100644 --- a/arch/x86/kernel/setup.c +++ b/arch/x86/kernel/setup.c | |||
@@ -1,139 +1,894 @@ | |||
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 | #include <linux/highmem.h> | ||
61 | |||
62 | #include <linux/kallsyms.h> | ||
63 | #include <linux/edd.h> | ||
64 | #include <linux/iscsi_ibft.h> | ||
65 | #include <linux/kexec.h> | ||
66 | #include <linux/cpufreq.h> | ||
67 | #include <linux/dma-mapping.h> | ||
68 | #include <linux/ctype.h> | ||
69 | #include <linux/uaccess.h> | ||
70 | |||
5 | #include <linux/percpu.h> | 71 | #include <linux/percpu.h> |
72 | #include <linux/crash_dump.h> | ||
73 | |||
74 | #include <video/edid.h> | ||
75 | |||
76 | #include <asm/mtrr.h> | ||
77 | #include <asm/apic.h> | ||
78 | #include <asm/e820.h> | ||
79 | #include <asm/mpspec.h> | ||
80 | #include <asm/setup.h> | ||
81 | #include <asm/arch_hooks.h> | ||
82 | #include <asm/efi.h> | ||
83 | #include <asm/sections.h> | ||
84 | #include <asm/dmi.h> | ||
85 | #include <asm/io_apic.h> | ||
86 | #include <asm/ist.h> | ||
87 | #include <asm/vmi.h> | ||
88 | #include <setup_arch.h> | ||
89 | #include <asm/bios_ebda.h> | ||
90 | #include <asm/cacheflush.h> | ||
91 | #include <asm/processor.h> | ||
92 | #include <asm/bugs.h> | ||
93 | |||
94 | #include <asm/system.h> | ||
95 | #include <asm/vsyscall.h> | ||
6 | #include <asm/smp.h> | 96 | #include <asm/smp.h> |
97 | #include <asm/desc.h> | ||
98 | #include <asm/dma.h> | ||
99 | #include <asm/gart.h> | ||
100 | #include <asm/mmu_context.h> | ||
101 | #include <asm/proto.h> | ||
102 | |||
103 | #include <mach_apic.h> | ||
104 | #include <asm/paravirt.h> | ||
105 | |||
7 | #include <asm/percpu.h> | 106 | #include <asm/percpu.h> |
8 | #include <asm/sections.h> | 107 | #include <asm/sections.h> |
9 | #include <asm/processor.h> | ||
10 | #include <asm/setup.h> | ||
11 | #include <asm/topology.h> | 108 | #include <asm/topology.h> |
12 | #include <asm/mpspec.h> | ||
13 | #include <asm/apicdef.h> | 109 | #include <asm/apicdef.h> |
110 | #ifdef CONFIG_X86_64 | ||
111 | #include <asm/numa_64.h> | ||
112 | #endif | ||
14 | 113 | ||
15 | #ifdef CONFIG_X86_LOCAL_APIC | 114 | #ifndef ARCH_SETUP |
16 | unsigned int num_processors; | 115 | #define ARCH_SETUP |
17 | unsigned disabled_cpus __cpuinitdata; | 116 | #endif |
18 | /* Processor that is doing the boot up */ | 117 | |
19 | unsigned int boot_cpu_physical_apicid = -1U; | 118 | #ifndef CONFIG_DEBUG_BOOT_PARAMS |
20 | EXPORT_SYMBOL(boot_cpu_physical_apicid); | 119 | struct boot_params __initdata boot_params; |
120 | #else | ||
121 | struct boot_params boot_params; | ||
122 | #endif | ||
123 | |||
124 | /* | ||
125 | * Machine setup.. | ||
126 | */ | ||
127 | static struct resource data_resource = { | ||
128 | .name = "Kernel data", | ||
129 | .start = 0, | ||
130 | .end = 0, | ||
131 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | ||
132 | }; | ||
133 | |||
134 | static struct resource code_resource = { | ||
135 | .name = "Kernel code", | ||
136 | .start = 0, | ||
137 | .end = 0, | ||
138 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | ||
139 | }; | ||
140 | |||
141 | static struct resource bss_resource = { | ||
142 | .name = "Kernel bss", | ||
143 | .start = 0, | ||
144 | .end = 0, | ||
145 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | ||
146 | }; | ||
147 | |||
148 | |||
149 | #ifdef CONFIG_X86_32 | ||
150 | /* This value is set up by the early boot code to point to the value | ||
151 | immediately after the boot time page tables. It contains a *physical* | ||
152 | address, and must not be in the .bss segment! */ | ||
153 | unsigned long init_pg_tables_start __initdata = ~0UL; | ||
154 | unsigned long init_pg_tables_end __initdata = ~0UL; | ||
155 | |||
156 | static struct resource video_ram_resource = { | ||
157 | .name = "Video RAM area", | ||
158 | .start = 0xa0000, | ||
159 | .end = 0xbffff, | ||
160 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | ||
161 | }; | ||
162 | |||
163 | /* cpu data as detected by the assembly code in head.S */ | ||
164 | struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1}; | ||
165 | /* common cpu data for all cpus */ | ||
166 | struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1}; | ||
167 | EXPORT_SYMBOL(boot_cpu_data); | ||
168 | static void set_mca_bus(int x) | ||
169 | { | ||
170 | #ifdef CONFIG_MCA | ||
171 | MCA_bus = x; | ||
172 | #endif | ||
173 | } | ||
174 | |||
175 | unsigned int def_to_bigsmp; | ||
176 | |||
177 | /* for MCA, but anyone else can use it if they want */ | ||
178 | unsigned int machine_id; | ||
179 | unsigned int machine_submodel_id; | ||
180 | unsigned int BIOS_revision; | ||
181 | |||
182 | struct apm_info apm_info; | ||
183 | EXPORT_SYMBOL(apm_info); | ||
184 | |||
185 | #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \ | ||
186 | defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE) | ||
187 | struct ist_info ist_info; | ||
188 | EXPORT_SYMBOL(ist_info); | ||
189 | #else | ||
190 | struct ist_info ist_info; | ||
191 | #endif | ||
192 | |||
193 | #else | ||
194 | struct cpuinfo_x86 boot_cpu_data __read_mostly; | ||
195 | EXPORT_SYMBOL(boot_cpu_data); | ||
196 | #endif | ||
21 | 197 | ||
22 | DEFINE_PER_CPU(u16, x86_cpu_to_apicid) = BAD_APICID; | ||
23 | EXPORT_PER_CPU_SYMBOL(x86_cpu_to_apicid); | ||
24 | 198 | ||
25 | /* Bitmask of physically existing CPUs */ | 199 | #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64) |
26 | physid_mask_t phys_cpu_present_map; | 200 | unsigned long mmu_cr4_features; |
201 | #else | ||
202 | unsigned long mmu_cr4_features = X86_CR4_PAE; | ||
27 | #endif | 203 | #endif |
28 | 204 | ||
29 | #if defined(CONFIG_HAVE_SETUP_PER_CPU_AREA) && defined(CONFIG_X86_SMP) | 205 | /* Boot loader ID as an integer, for the benefit of proc_dointvec */ |
206 | int bootloader_type; | ||
207 | |||
30 | /* | 208 | /* |
31 | * Copy data used in early init routines from the initial arrays to the | 209 | * Early DMI memory |
32 | * per cpu data areas. These arrays then become expendable and the | 210 | */ |
33 | * *_early_ptr's are zeroed indicating that the static arrays are gone. | 211 | int dmi_alloc_index; |
212 | char dmi_alloc_data[DMI_MAX_DATA]; | ||
213 | |||
214 | /* | ||
215 | * Setup options | ||
216 | */ | ||
217 | struct screen_info screen_info; | ||
218 | EXPORT_SYMBOL(screen_info); | ||
219 | struct edid_info edid_info; | ||
220 | EXPORT_SYMBOL_GPL(edid_info); | ||
221 | |||
222 | extern int root_mountflags; | ||
223 | |||
224 | unsigned long saved_video_mode; | ||
225 | |||
226 | #define RAMDISK_IMAGE_START_MASK 0x07FF | ||
227 | #define RAMDISK_PROMPT_FLAG 0x8000 | ||
228 | #define RAMDISK_LOAD_FLAG 0x4000 | ||
229 | |||
230 | static char __initdata command_line[COMMAND_LINE_SIZE]; | ||
231 | |||
232 | #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE) | ||
233 | struct edd edd; | ||
234 | #ifdef CONFIG_EDD_MODULE | ||
235 | EXPORT_SYMBOL(edd); | ||
236 | #endif | ||
237 | /** | ||
238 | * copy_edd() - Copy the BIOS EDD information | ||
239 | * from boot_params into a safe place. | ||
240 | * | ||
34 | */ | 241 | */ |
35 | static void __init setup_per_cpu_maps(void) | 242 | static inline void copy_edd(void) |
243 | { | ||
244 | memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer, | ||
245 | sizeof(edd.mbr_signature)); | ||
246 | memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info)); | ||
247 | edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries; | ||
248 | edd.edd_info_nr = boot_params.eddbuf_entries; | ||
249 | } | ||
250 | #else | ||
251 | static inline void copy_edd(void) | ||
252 | { | ||
253 | } | ||
254 | #endif | ||
255 | |||
256 | #ifdef CONFIG_BLK_DEV_INITRD | ||
257 | |||
258 | #ifdef CONFIG_X86_32 | ||
259 | |||
260 | #define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT) | ||
261 | static void __init relocate_initrd(void) | ||
36 | { | 262 | { |
37 | int cpu; | ||
38 | 263 | ||
39 | for_each_possible_cpu(cpu) { | 264 | u64 ramdisk_image = boot_params.hdr.ramdisk_image; |
40 | per_cpu(x86_cpu_to_apicid, cpu) = x86_cpu_to_apicid_init[cpu]; | 265 | u64 ramdisk_size = boot_params.hdr.ramdisk_size; |
41 | per_cpu(x86_bios_cpu_apicid, cpu) = | 266 | u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT; |
42 | x86_bios_cpu_apicid_init[cpu]; | 267 | u64 ramdisk_here; |
43 | #ifdef CONFIG_NUMA | 268 | unsigned long slop, clen, mapaddr; |
44 | per_cpu(x86_cpu_to_node_map, cpu) = | 269 | char *p, *q; |
45 | x86_cpu_to_node_map_init[cpu]; | 270 | |
271 | /* We need to move the initrd down into lowmem */ | ||
272 | ramdisk_here = find_e820_area(0, end_of_lowmem, ramdisk_size, | ||
273 | PAGE_SIZE); | ||
274 | |||
275 | if (ramdisk_here == -1ULL) | ||
276 | panic("Cannot find place for new RAMDISK of size %lld\n", | ||
277 | ramdisk_size); | ||
278 | |||
279 | /* Note: this includes all the lowmem currently occupied by | ||
280 | the initrd, we rely on that fact to keep the data intact. */ | ||
281 | reserve_early(ramdisk_here, ramdisk_here + ramdisk_size, | ||
282 | "NEW RAMDISK"); | ||
283 | initrd_start = ramdisk_here + PAGE_OFFSET; | ||
284 | initrd_end = initrd_start + ramdisk_size; | ||
285 | printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n", | ||
286 | ramdisk_here, ramdisk_here + ramdisk_size); | ||
287 | |||
288 | q = (char *)initrd_start; | ||
289 | |||
290 | /* Copy any lowmem portion of the initrd */ | ||
291 | if (ramdisk_image < end_of_lowmem) { | ||
292 | clen = end_of_lowmem - ramdisk_image; | ||
293 | p = (char *)__va(ramdisk_image); | ||
294 | memcpy(q, p, clen); | ||
295 | q += clen; | ||
296 | ramdisk_image += clen; | ||
297 | ramdisk_size -= clen; | ||
298 | } | ||
299 | |||
300 | /* Copy the highmem portion of the initrd */ | ||
301 | while (ramdisk_size) { | ||
302 | slop = ramdisk_image & ~PAGE_MASK; | ||
303 | clen = ramdisk_size; | ||
304 | if (clen > MAX_MAP_CHUNK-slop) | ||
305 | clen = MAX_MAP_CHUNK-slop; | ||
306 | mapaddr = ramdisk_image & PAGE_MASK; | ||
307 | p = early_ioremap(mapaddr, clen+slop); | ||
308 | memcpy(q, p+slop, clen); | ||
309 | early_iounmap(p, clen+slop); | ||
310 | q += clen; | ||
311 | ramdisk_image += clen; | ||
312 | ramdisk_size -= clen; | ||
313 | } | ||
314 | /* high pages is not converted by early_res_to_bootmem */ | ||
315 | ramdisk_image = boot_params.hdr.ramdisk_image; | ||
316 | ramdisk_size = boot_params.hdr.ramdisk_size; | ||
317 | printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to" | ||
318 | " %08llx - %08llx\n", | ||
319 | ramdisk_image, ramdisk_image + ramdisk_size - 1, | ||
320 | ramdisk_here, ramdisk_here + ramdisk_size - 1); | ||
321 | } | ||
46 | #endif | 322 | #endif |
323 | |||
324 | static void __init reserve_initrd(void) | ||
325 | { | ||
326 | u64 ramdisk_image = boot_params.hdr.ramdisk_image; | ||
327 | u64 ramdisk_size = boot_params.hdr.ramdisk_size; | ||
328 | u64 ramdisk_end = ramdisk_image + ramdisk_size; | ||
329 | u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT; | ||
330 | |||
331 | if (!boot_params.hdr.type_of_loader || | ||
332 | !ramdisk_image || !ramdisk_size) | ||
333 | return; /* No initrd provided by bootloader */ | ||
334 | |||
335 | initrd_start = 0; | ||
336 | |||
337 | if (ramdisk_size >= (end_of_lowmem>>1)) { | ||
338 | free_early(ramdisk_image, ramdisk_end); | ||
339 | printk(KERN_ERR "initrd too large to handle, " | ||
340 | "disabling initrd\n"); | ||
341 | return; | ||
342 | } | ||
343 | |||
344 | printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image, | ||
345 | ramdisk_end); | ||
346 | |||
347 | |||
348 | if (ramdisk_end <= end_of_lowmem) { | ||
349 | /* All in lowmem, easy case */ | ||
350 | /* | ||
351 | * don't need to reserve again, already reserved early | ||
352 | * in i386_start_kernel | ||
353 | */ | ||
354 | initrd_start = ramdisk_image + PAGE_OFFSET; | ||
355 | initrd_end = initrd_start + ramdisk_size; | ||
356 | return; | ||
47 | } | 357 | } |
48 | 358 | ||
49 | /* indicate the early static arrays will soon be gone */ | 359 | #ifdef CONFIG_X86_32 |
50 | x86_cpu_to_apicid_early_ptr = NULL; | 360 | relocate_initrd(); |
51 | x86_bios_cpu_apicid_early_ptr = NULL; | 361 | #else |
52 | #ifdef CONFIG_NUMA | 362 | printk(KERN_ERR "initrd extends beyond end of memory " |
53 | x86_cpu_to_node_map_early_ptr = NULL; | 363 | "(0x%08llx > 0x%08llx)\ndisabling initrd\n", |
364 | ramdisk_end, end_of_lowmem); | ||
365 | initrd_start = 0; | ||
54 | #endif | 366 | #endif |
367 | free_early(ramdisk_image, ramdisk_end); | ||
55 | } | 368 | } |
369 | #else | ||
370 | static void __init reserve_initrd(void) | ||
371 | { | ||
372 | } | ||
373 | #endif /* CONFIG_BLK_DEV_INITRD */ | ||
374 | |||
375 | static void __init parse_setup_data(void) | ||
376 | { | ||
377 | struct setup_data *data; | ||
378 | u64 pa_data; | ||
379 | |||
380 | if (boot_params.hdr.version < 0x0209) | ||
381 | return; | ||
382 | pa_data = boot_params.hdr.setup_data; | ||
383 | while (pa_data) { | ||
384 | data = early_ioremap(pa_data, PAGE_SIZE); | ||
385 | switch (data->type) { | ||
386 | case SETUP_E820_EXT: | ||
387 | parse_e820_ext(data, pa_data); | ||
388 | break; | ||
389 | default: | ||
390 | break; | ||
391 | } | ||
392 | pa_data = data->next; | ||
393 | early_iounmap(data, PAGE_SIZE); | ||
394 | } | ||
395 | } | ||
396 | |||
397 | static void __init e820_reserve_setup_data(void) | ||
398 | { | ||
399 | struct setup_data *data; | ||
400 | u64 pa_data; | ||
401 | int found = 0; | ||
402 | |||
403 | if (boot_params.hdr.version < 0x0209) | ||
404 | return; | ||
405 | pa_data = boot_params.hdr.setup_data; | ||
406 | while (pa_data) { | ||
407 | data = early_ioremap(pa_data, sizeof(*data)); | ||
408 | e820_update_range(pa_data, sizeof(*data)+data->len, | ||
409 | E820_RAM, E820_RESERVED_KERN); | ||
410 | found = 1; | ||
411 | pa_data = data->next; | ||
412 | early_iounmap(data, sizeof(*data)); | ||
413 | } | ||
414 | if (!found) | ||
415 | return; | ||
56 | 416 | ||
57 | #ifdef CONFIG_HAVE_CPUMASK_OF_CPU_MAP | 417 | sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); |
58 | cpumask_t *cpumask_of_cpu_map __read_mostly; | 418 | memcpy(&e820_saved, &e820, sizeof(struct e820map)); |
59 | EXPORT_SYMBOL(cpumask_of_cpu_map); | 419 | printk(KERN_INFO "extended physical RAM map:\n"); |
420 | e820_print_map("reserve setup_data"); | ||
421 | } | ||
60 | 422 | ||
61 | /* requires nr_cpu_ids to be initialized */ | 423 | static void __init reserve_early_setup_data(void) |
62 | static void __init setup_cpumask_of_cpu(void) | ||
63 | { | 424 | { |
64 | int i; | 425 | struct setup_data *data; |
426 | u64 pa_data; | ||
427 | char buf[32]; | ||
428 | |||
429 | if (boot_params.hdr.version < 0x0209) | ||
430 | return; | ||
431 | pa_data = boot_params.hdr.setup_data; | ||
432 | while (pa_data) { | ||
433 | data = early_ioremap(pa_data, sizeof(*data)); | ||
434 | sprintf(buf, "setup data %x", data->type); | ||
435 | reserve_early(pa_data, pa_data+sizeof(*data)+data->len, buf); | ||
436 | pa_data = data->next; | ||
437 | early_iounmap(data, sizeof(*data)); | ||
438 | } | ||
439 | } | ||
440 | |||
441 | /* | ||
442 | * --------- Crashkernel reservation ------------------------------ | ||
443 | */ | ||
444 | |||
445 | #ifdef CONFIG_KEXEC | ||
446 | |||
447 | /** | ||
448 | * Reserve @size bytes of crashkernel memory at any suitable offset. | ||
449 | * | ||
450 | * @size: Size of the crashkernel memory to reserve. | ||
451 | * Returns the base address on success, and -1ULL on failure. | ||
452 | */ | ||
453 | unsigned long long find_and_reserve_crashkernel(unsigned long long size) | ||
454 | { | ||
455 | const unsigned long long alignment = 16<<20; /* 16M */ | ||
456 | unsigned long long start = 0LL; | ||
457 | |||
458 | while (1) { | ||
459 | int ret; | ||
460 | |||
461 | start = find_e820_area(start, ULONG_MAX, size, alignment); | ||
462 | if (start == -1ULL) | ||
463 | return start; | ||
464 | |||
465 | /* try to reserve it */ | ||
466 | ret = reserve_bootmem_generic(start, size, BOOTMEM_EXCLUSIVE); | ||
467 | if (ret >= 0) | ||
468 | return start; | ||
469 | |||
470 | start += alignment; | ||
471 | } | ||
472 | } | ||
473 | |||
474 | static inline unsigned long long get_total_mem(void) | ||
475 | { | ||
476 | unsigned long long total; | ||
477 | |||
478 | total = max_low_pfn - min_low_pfn; | ||
479 | #ifdef CONFIG_HIGHMEM | ||
480 | total += highend_pfn - highstart_pfn; | ||
481 | #endif | ||
482 | |||
483 | return total << PAGE_SHIFT; | ||
484 | } | ||
485 | |||
486 | static void __init reserve_crashkernel(void) | ||
487 | { | ||
488 | unsigned long long total_mem; | ||
489 | unsigned long long crash_size, crash_base; | ||
490 | int ret; | ||
491 | |||
492 | total_mem = get_total_mem(); | ||
65 | 493 | ||
66 | /* alloc_bootmem zeroes memory */ | 494 | ret = parse_crashkernel(boot_command_line, total_mem, |
67 | cpumask_of_cpu_map = alloc_bootmem_low(sizeof(cpumask_t) * nr_cpu_ids); | 495 | &crash_size, &crash_base); |
68 | for (i = 0; i < nr_cpu_ids; i++) | 496 | if (ret != 0 || crash_size <= 0) |
69 | cpu_set(i, cpumask_of_cpu_map[i]); | 497 | return; |
498 | |||
499 | /* 0 means: find the address automatically */ | ||
500 | if (crash_base <= 0) { | ||
501 | crash_base = find_and_reserve_crashkernel(crash_size); | ||
502 | if (crash_base == -1ULL) { | ||
503 | pr_info("crashkernel reservation failed. " | ||
504 | "No suitable area found.\n"); | ||
505 | return; | ||
506 | } | ||
507 | } else { | ||
508 | ret = reserve_bootmem_generic(crash_base, crash_size, | ||
509 | BOOTMEM_EXCLUSIVE); | ||
510 | if (ret < 0) { | ||
511 | pr_info("crashkernel reservation failed - " | ||
512 | "memory is in use\n"); | ||
513 | return; | ||
514 | } | ||
515 | } | ||
516 | |||
517 | printk(KERN_INFO "Reserving %ldMB of memory at %ldMB " | ||
518 | "for crashkernel (System RAM: %ldMB)\n", | ||
519 | (unsigned long)(crash_size >> 20), | ||
520 | (unsigned long)(crash_base >> 20), | ||
521 | (unsigned long)(total_mem >> 20)); | ||
522 | |||
523 | crashk_res.start = crash_base; | ||
524 | crashk_res.end = crash_base + crash_size - 1; | ||
525 | insert_resource(&iomem_resource, &crashk_res); | ||
70 | } | 526 | } |
71 | #else | 527 | #else |
72 | static inline void setup_cpumask_of_cpu(void) { } | 528 | static void __init reserve_crashkernel(void) |
529 | { | ||
530 | } | ||
73 | #endif | 531 | #endif |
74 | 532 | ||
75 | #ifdef CONFIG_X86_32 | 533 | static struct resource standard_io_resources[] = { |
76 | /* | 534 | { .name = "dma1", .start = 0x00, .end = 0x1f, |
77 | * Great future not-so-futuristic plan: make i386 and x86_64 do it | 535 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, |
78 | * the same way | 536 | { .name = "pic1", .start = 0x20, .end = 0x21, |
537 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
538 | { .name = "timer0", .start = 0x40, .end = 0x43, | ||
539 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
540 | { .name = "timer1", .start = 0x50, .end = 0x53, | ||
541 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
542 | { .name = "keyboard", .start = 0x60, .end = 0x60, | ||
543 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
544 | { .name = "keyboard", .start = 0x64, .end = 0x64, | ||
545 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
546 | { .name = "dma page reg", .start = 0x80, .end = 0x8f, | ||
547 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
548 | { .name = "pic2", .start = 0xa0, .end = 0xa1, | ||
549 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
550 | { .name = "dma2", .start = 0xc0, .end = 0xdf, | ||
551 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | ||
552 | { .name = "fpu", .start = 0xf0, .end = 0xff, | ||
553 | .flags = IORESOURCE_BUSY | IORESOURCE_IO } | ||
554 | }; | ||
555 | |||
556 | static void __init reserve_standard_io_resources(void) | ||
557 | { | ||
558 | int i; | ||
559 | |||
560 | /* request I/O space for devices used on all i[345]86 PCs */ | ||
561 | for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++) | ||
562 | request_resource(&ioport_resource, &standard_io_resources[i]); | ||
563 | |||
564 | } | ||
565 | |||
566 | #ifdef CONFIG_PROC_VMCORE | ||
567 | /* elfcorehdr= specifies the location of elf core header | ||
568 | * stored by the crashed kernel. This option will be passed | ||
569 | * by kexec loader to the capture kernel. | ||
79 | */ | 570 | */ |
80 | unsigned long __per_cpu_offset[NR_CPUS] __read_mostly; | 571 | static int __init setup_elfcorehdr(char *arg) |
81 | EXPORT_SYMBOL(__per_cpu_offset); | 572 | { |
573 | char *end; | ||
574 | if (!arg) | ||
575 | return -EINVAL; | ||
576 | elfcorehdr_addr = memparse(arg, &end); | ||
577 | return end > arg ? 0 : -EINVAL; | ||
578 | } | ||
579 | early_param("elfcorehdr", setup_elfcorehdr); | ||
82 | #endif | 580 | #endif |
83 | 581 | ||
84 | /* | 582 | /* |
85 | * Great future plan: | 583 | * Determine if we were loaded by an EFI loader. If so, then we have also been |
86 | * Declare PDA itself and support (irqstack,tss,pgd) as per cpu data. | 584 | * passed the efi memmap, systab, etc., so we should use these data structures |
87 | * Always point %gs to its beginning | 585 | * for initialization. Note, the efi init code path is determined by the |
586 | * global efi_enabled. This allows the same kernel image to be used on existing | ||
587 | * systems (with a traditional BIOS) as well as on EFI systems. | ||
588 | */ | ||
589 | /* | ||
590 | * setup_arch - architecture-specific boot-time initializations | ||
591 | * | ||
592 | * Note: On x86_64, fixmaps are ready for use even before this is called. | ||
88 | */ | 593 | */ |
89 | void __init setup_per_cpu_areas(void) | 594 | |
595 | void __init setup_arch(char **cmdline_p) | ||
90 | { | 596 | { |
91 | int i, highest_cpu = 0; | 597 | #ifdef CONFIG_X86_32 |
92 | unsigned long size; | 598 | memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data)); |
599 | visws_early_detect(); | ||
600 | pre_setup_arch_hook(); | ||
601 | early_cpu_init(); | ||
602 | #else | ||
603 | printk(KERN_INFO "Command line: %s\n", boot_command_line); | ||
604 | #endif | ||
93 | 605 | ||
94 | #ifdef CONFIG_HOTPLUG_CPU | 606 | early_ioremap_init(); |
95 | prefill_possible_map(); | 607 | |
608 | ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev); | ||
609 | screen_info = boot_params.screen_info; | ||
610 | edid_info = boot_params.edid_info; | ||
611 | #ifdef CONFIG_X86_32 | ||
612 | apm_info.bios = boot_params.apm_bios_info; | ||
613 | ist_info = boot_params.ist_info; | ||
614 | if (boot_params.sys_desc_table.length != 0) { | ||
615 | set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2); | ||
616 | machine_id = boot_params.sys_desc_table.table[0]; | ||
617 | machine_submodel_id = boot_params.sys_desc_table.table[1]; | ||
618 | BIOS_revision = boot_params.sys_desc_table.table[2]; | ||
619 | } | ||
620 | #endif | ||
621 | saved_video_mode = boot_params.hdr.vid_mode; | ||
622 | bootloader_type = boot_params.hdr.type_of_loader; | ||
623 | |||
624 | #ifdef CONFIG_BLK_DEV_RAM | ||
625 | rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK; | ||
626 | rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0); | ||
627 | rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0); | ||
628 | #endif | ||
629 | #ifdef CONFIG_EFI | ||
630 | if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature, | ||
631 | #ifdef CONFIG_X86_32 | ||
632 | "EL32", | ||
633 | #else | ||
634 | "EL64", | ||
635 | #endif | ||
636 | 4)) { | ||
637 | efi_enabled = 1; | ||
638 | efi_reserve_early(); | ||
639 | } | ||
96 | #endif | 640 | #endif |
97 | 641 | ||
98 | /* Copy section for each CPU (we discard the original) */ | 642 | ARCH_SETUP |
99 | size = PERCPU_ENOUGH_ROOM; | ||
100 | printk(KERN_INFO "PERCPU: Allocating %lu bytes of per cpu data\n", | ||
101 | size); | ||
102 | 643 | ||
103 | for_each_possible_cpu(i) { | 644 | setup_memory_map(); |
104 | char *ptr; | 645 | parse_setup_data(); |
105 | #ifndef CONFIG_NEED_MULTIPLE_NODES | 646 | /* update the e820_saved too */ |
106 | ptr = alloc_bootmem_pages(size); | 647 | e820_reserve_setup_data(); |
648 | |||
649 | copy_edd(); | ||
650 | |||
651 | if (!boot_params.hdr.root_flags) | ||
652 | root_mountflags &= ~MS_RDONLY; | ||
653 | init_mm.start_code = (unsigned long) _text; | ||
654 | init_mm.end_code = (unsigned long) _etext; | ||
655 | init_mm.end_data = (unsigned long) _edata; | ||
656 | #ifdef CONFIG_X86_32 | ||
657 | init_mm.brk = init_pg_tables_end + PAGE_OFFSET; | ||
107 | #else | 658 | #else |
108 | int node = early_cpu_to_node(i); | 659 | 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 | 660 | #endif |
117 | if (!ptr) | 661 | |
118 | panic("Cannot allocate cpu data for CPU %d\n", i); | 662 | code_resource.start = virt_to_phys(_text); |
663 | code_resource.end = virt_to_phys(_etext)-1; | ||
664 | data_resource.start = virt_to_phys(_etext); | ||
665 | data_resource.end = virt_to_phys(_edata)-1; | ||
666 | bss_resource.start = virt_to_phys(&__bss_start); | ||
667 | bss_resource.end = virt_to_phys(&__bss_stop)-1; | ||
668 | |||
119 | #ifdef CONFIG_X86_64 | 669 | #ifdef CONFIG_X86_64 |
120 | cpu_pda(i)->data_offset = ptr - __per_cpu_start; | 670 | early_cpu_init(); |
671 | #endif | ||
672 | strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE); | ||
673 | *cmdline_p = command_line; | ||
674 | |||
675 | parse_early_param(); | ||
676 | |||
677 | /* after early param, so could get panic from serial */ | ||
678 | reserve_early_setup_data(); | ||
679 | |||
680 | if (acpi_mps_check()) { | ||
681 | #ifdef CONFIG_X86_LOCAL_APIC | ||
682 | disable_apic = 1; | ||
683 | #endif | ||
684 | clear_cpu_cap(&boot_cpu_data, X86_FEATURE_APIC); | ||
685 | } | ||
686 | |||
687 | #ifdef CONFIG_PCI | ||
688 | if (pci_early_dump_regs) | ||
689 | early_dump_pci_devices(); | ||
690 | #endif | ||
691 | |||
692 | finish_e820_parsing(); | ||
693 | |||
694 | #ifdef CONFIG_X86_32 | ||
695 | probe_roms(); | ||
696 | #endif | ||
697 | |||
698 | /* after parse_early_param, so could debug it */ | ||
699 | insert_resource(&iomem_resource, &code_resource); | ||
700 | insert_resource(&iomem_resource, &data_resource); | ||
701 | insert_resource(&iomem_resource, &bss_resource); | ||
702 | |||
703 | if (efi_enabled) | ||
704 | efi_init(); | ||
705 | |||
706 | #ifdef CONFIG_X86_32 | ||
707 | if (ppro_with_ram_bug()) { | ||
708 | e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM, | ||
709 | E820_RESERVED); | ||
710 | sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); | ||
711 | printk(KERN_INFO "fixed physical RAM map:\n"); | ||
712 | e820_print_map("bad_ppro"); | ||
713 | } | ||
714 | #else | ||
715 | early_gart_iommu_check(); | ||
716 | #endif | ||
717 | |||
718 | /* | ||
719 | * partially used pages are not usable - thus | ||
720 | * we are rounding upwards: | ||
721 | */ | ||
722 | max_pfn = e820_end_of_ram_pfn(); | ||
723 | |||
724 | /* preallocate 4k for mptable mpc */ | ||
725 | early_reserve_e820_mpc_new(); | ||
726 | /* update e820 for memory not covered by WB MTRRs */ | ||
727 | mtrr_bp_init(); | ||
728 | if (mtrr_trim_uncached_memory(max_pfn)) | ||
729 | max_pfn = e820_end_of_ram_pfn(); | ||
730 | |||
731 | #ifdef CONFIG_X86_32 | ||
732 | /* max_low_pfn get updated here */ | ||
733 | find_low_pfn_range(); | ||
121 | #else | 734 | #else |
122 | __per_cpu_offset[i] = ptr - __per_cpu_start; | 735 | num_physpages = max_pfn; |
736 | |||
737 | check_efer(); | ||
738 | |||
739 | /* How many end-of-memory variables you have, grandma! */ | ||
740 | /* need this before calling reserve_initrd */ | ||
741 | if (max_pfn > (1UL<<(32 - PAGE_SHIFT))) | ||
742 | max_low_pfn = e820_end_of_low_ram_pfn(); | ||
743 | else | ||
744 | max_low_pfn = max_pfn; | ||
745 | |||
746 | high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1; | ||
123 | #endif | 747 | #endif |
124 | memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start); | ||
125 | 748 | ||
126 | highest_cpu = i; | 749 | /* max_pfn_mapped is updated here */ |
750 | max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT); | ||
751 | max_pfn_mapped = max_low_pfn_mapped; | ||
752 | |||
753 | #ifdef CONFIG_X86_64 | ||
754 | if (max_pfn > max_low_pfn) { | ||
755 | max_pfn_mapped = init_memory_mapping(1UL<<32, | ||
756 | max_pfn<<PAGE_SHIFT); | ||
757 | /* can we preseve max_low_pfn ?*/ | ||
758 | max_low_pfn = max_pfn; | ||
127 | } | 759 | } |
760 | #endif | ||
128 | 761 | ||
129 | nr_cpu_ids = highest_cpu + 1; | 762 | /* |
130 | printk(KERN_DEBUG "NR_CPUS: %d, nr_cpu_ids: %d\n", NR_CPUS, nr_cpu_ids); | 763 | * NOTE: On x86-32, only from this point on, fixmaps are ready for use. |
764 | */ | ||
131 | 765 | ||
132 | /* Setup percpu data maps */ | 766 | #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT |
133 | setup_per_cpu_maps(); | 767 | if (init_ohci1394_dma_early) |
768 | init_ohci1394_dma_on_all_controllers(); | ||
769 | #endif | ||
134 | 770 | ||
135 | /* Setup cpumask_of_cpu map */ | 771 | reserve_initrd(); |
136 | setup_cpumask_of_cpu(); | 772 | |
137 | } | 773 | #ifdef CONFIG_X86_64 |
774 | vsmp_init(); | ||
775 | #endif | ||
776 | |||
777 | dmi_scan_machine(); | ||
778 | |||
779 | io_delay_init(); | ||
780 | |||
781 | /* | ||
782 | * Parse the ACPI tables for possible boot-time SMP configuration. | ||
783 | */ | ||
784 | acpi_boot_table_init(); | ||
785 | |||
786 | #ifdef CONFIG_ACPI_NUMA | ||
787 | /* | ||
788 | * Parse SRAT to discover nodes. | ||
789 | */ | ||
790 | acpi_numa_init(); | ||
791 | #endif | ||
792 | |||
793 | initmem_init(0, max_pfn); | ||
794 | |||
795 | #ifdef CONFIG_X86_64 | ||
796 | dma32_reserve_bootmem(); | ||
797 | #endif | ||
798 | |||
799 | #ifdef CONFIG_ACPI_SLEEP | ||
800 | /* | ||
801 | * Reserve low memory region for sleep support. | ||
802 | */ | ||
803 | acpi_reserve_bootmem(); | ||
804 | #endif | ||
805 | #ifdef CONFIG_X86_FIND_SMP_CONFIG | ||
806 | /* | ||
807 | * Find and reserve possible boot-time SMP configuration: | ||
808 | */ | ||
809 | find_smp_config(); | ||
810 | #endif | ||
811 | reserve_crashkernel(); | ||
812 | |||
813 | reserve_ibft_region(); | ||
814 | |||
815 | #ifdef CONFIG_KVM_CLOCK | ||
816 | kvmclock_init(); | ||
817 | #endif | ||
818 | |||
819 | #if defined(CONFIG_VMI) && defined(CONFIG_X86_32) | ||
820 | /* | ||
821 | * Must be after max_low_pfn is determined, and before kernel | ||
822 | * pagetables are setup. | ||
823 | */ | ||
824 | vmi_init(); | ||
825 | #endif | ||
826 | |||
827 | paging_init(); | ||
828 | |||
829 | #ifdef CONFIG_X86_64 | ||
830 | map_vsyscall(); | ||
831 | #endif | ||
832 | |||
833 | #ifdef CONFIG_X86_GENERICARCH | ||
834 | generic_apic_probe(); | ||
835 | #endif | ||
836 | |||
837 | early_quirks(); | ||
838 | |||
839 | /* | ||
840 | * Read APIC and some other early information from ACPI tables. | ||
841 | */ | ||
842 | acpi_boot_init(); | ||
843 | |||
844 | #if defined(CONFIG_X86_MPPARSE) || defined(CONFIG_X86_VISWS) | ||
845 | /* | ||
846 | * get boot-time SMP configuration: | ||
847 | */ | ||
848 | if (smp_found_config) | ||
849 | get_smp_config(); | ||
850 | #endif | ||
138 | 851 | ||
852 | prefill_possible_map(); | ||
853 | #ifdef CONFIG_X86_64 | ||
854 | init_cpu_to_node(); | ||
855 | #endif | ||
856 | |||
857 | #ifdef CONFIG_X86_NUMAQ | ||
858 | /* | ||
859 | * need to check online nodes num, call it | ||
860 | * here before time_init/tsc_init | ||
861 | */ | ||
862 | numaq_tsc_disable(); | ||
863 | #endif | ||
864 | |||
865 | init_apic_mappings(); | ||
866 | ioapic_init_mappings(); | ||
867 | |||
868 | #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC) && defined(CONFIG_X86_32) | ||
869 | if (def_to_bigsmp) | ||
870 | printk(KERN_WARNING "More than 8 CPUs detected and " | ||
871 | "CONFIG_X86_PC cannot handle it.\nUse " | ||
872 | "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n"); | ||
873 | #endif | ||
874 | kvm_guest_init(); | ||
875 | |||
876 | e820_reserve_resources(); | ||
877 | e820_mark_nosave_regions(max_low_pfn); | ||
878 | |||
879 | #ifdef CONFIG_X86_32 | ||
880 | request_resource(&iomem_resource, &video_ram_resource); | ||
881 | #endif | ||
882 | reserve_standard_io_resources(); | ||
883 | |||
884 | e820_setup_gap(); | ||
885 | |||
886 | #ifdef CONFIG_VT | ||
887 | #if defined(CONFIG_VGA_CONSOLE) | ||
888 | if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY)) | ||
889 | conswitchp = &vga_con; | ||
890 | #elif defined(CONFIG_DUMMY_CONSOLE) | ||
891 | conswitchp = &dummy_con; | ||
139 | #endif | 892 | #endif |
893 | #endif | ||
894 | } | ||