]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1da177e4 LT |
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> | |
05b79bdc | 26 | #include <linux/mmzone.h> |
894673ee | 27 | #include <linux/screen_info.h> |
1da177e4 LT |
28 | #include <linux/ioport.h> |
29 | #include <linux/acpi.h> | |
efafc8b2 | 30 | #include <linux/sfi.h> |
1da177e4 LT |
31 | #include <linux/apm_bios.h> |
32 | #include <linux/initrd.h> | |
33 | #include <linux/bootmem.h> | |
72d7c3b3 | 34 | #include <linux/memblock.h> |
1da177e4 LT |
35 | #include <linux/seq_file.h> |
36 | #include <linux/console.h> | |
1da177e4 LT |
37 | #include <linux/root_dev.h> |
38 | #include <linux/highmem.h> | |
186f4360 | 39 | #include <linux/export.h> |
1da177e4 LT |
40 | #include <linux/efi.h> |
41 | #include <linux/init.h> | |
42 | #include <linux/edd.h> | |
138fe4e0 | 43 | #include <linux/iscsi_ibft.h> |
1da177e4 | 44 | #include <linux/nodemask.h> |
1bc3b91a | 45 | #include <linux/kexec.h> |
e9928674 | 46 | #include <linux/dmi.h> |
22a9835c | 47 | #include <linux/pfn.h> |
376ff035 | 48 | #include <linux/pci.h> |
46d671b5 | 49 | #include <asm/pci-direct.h> |
f212ec4b | 50 | #include <linux/init_ohci1394_dma.h> |
790c73f6 | 51 | #include <linux/kvm_para.h> |
0a2b9a6e | 52 | #include <linux/dma-contiguous.h> |
1bc3b91a | 53 | |
46d671b5 YL |
54 | #include <linux/errno.h> |
55 | #include <linux/kernel.h> | |
56 | #include <linux/stddef.h> | |
57 | #include <linux/unistd.h> | |
58 | #include <linux/ptrace.h> | |
46d671b5 YL |
59 | #include <linux/user.h> |
60 | #include <linux/delay.h> | |
46d671b5 YL |
61 | |
62 | #include <linux/kallsyms.h> | |
46d671b5 YL |
63 | #include <linux/cpufreq.h> |
64 | #include <linux/dma-mapping.h> | |
65 | #include <linux/ctype.h> | |
66 | #include <linux/uaccess.h> | |
67 | ||
68 | #include <linux/percpu.h> | |
69 | #include <linux/crash_dump.h> | |
69575d38 | 70 | #include <linux/tboot.h> |
b3c869d3 | 71 | #include <linux/jiffies.h> |
46d671b5 | 72 | |
1b5aeebf | 73 | #include <linux/usb/xhci-dbgp.h> |
1da177e4 | 74 | #include <video/edid.h> |
1bc3b91a | 75 | |
093af8d7 | 76 | #include <asm/mtrr.h> |
9635b47d | 77 | #include <asm/apic.h> |
084ee1c6 | 78 | #include <asm/realmode.h> |
66441bd3 | 79 | #include <asm/e820/api.h> |
1da177e4 LT |
80 | #include <asm/mpspec.h> |
81 | #include <asm/setup.h> | |
55f26239 | 82 | #include <asm/efi.h> |
8e6dafd6 IM |
83 | #include <asm/timer.h> |
84 | #include <asm/i8259.h> | |
1da177e4 LT |
85 | #include <asm/sections.h> |
86 | #include <asm/io_apic.h> | |
87 | #include <asm/ist.h> | |
1164dd00 | 88 | #include <asm/setup_arch.h> |
ce3fe6b2 | 89 | #include <asm/bios_ebda.h> |
00bf4098 | 90 | #include <asm/cacheflush.h> |
2fde61fd | 91 | #include <asm/processor.h> |
cc9f7a0c | 92 | #include <asm/bugs.h> |
ef7f0d6a | 93 | #include <asm/kasan.h> |
1da177e4 | 94 | |
46d671b5 | 95 | #include <asm/vsyscall.h> |
6e5385d4 | 96 | #include <asm/cpu.h> |
46d671b5 YL |
97 | #include <asm/desc.h> |
98 | #include <asm/dma.h> | |
46a7fa27 | 99 | #include <asm/iommu.h> |
1d9b16d1 | 100 | #include <asm/gart.h> |
46d671b5 YL |
101 | #include <asm/mmu_context.h> |
102 | #include <asm/proto.h> | |
103 | ||
46d671b5 | 104 | #include <asm/paravirt.h> |
88b094fb | 105 | #include <asm/hypervisor.h> |
fd699c76 | 106 | #include <asm/olpc_ofw.h> |
46d671b5 YL |
107 | |
108 | #include <asm/percpu.h> | |
46d671b5 YL |
109 | #include <asm/topology.h> |
110 | #include <asm/apicdef.h> | |
23ac4ae8 | 111 | #include <asm/amd_nb.h> |
a2202aa2 | 112 | #include <asm/mce.h> |
f49aa448 | 113 | #include <asm/alternative.h> |
da6b737b | 114 | #include <asm/prom.h> |
9a2bc335 | 115 | #include <asm/microcode.h> |
c1192f84 | 116 | #include <asm/mmu_context.h> |
0483e1fa | 117 | #include <asm/kaslr.h> |
46d671b5 | 118 | |
2b72394e | 119 | /* |
66520ebc JS |
120 | * max_low_pfn_mapped: highest direct mapped pfn under 4GB |
121 | * max_pfn_mapped: highest direct mapped pfn over 4GB | |
122 | * | |
09821ff1 | 123 | * The direct mapping only covers E820_TYPE_RAM regions, so the ranges and gaps are |
66520ebc | 124 | * represented by pfn_mapped |
2b72394e PE |
125 | */ |
126 | unsigned long max_low_pfn_mapped; | |
127 | unsigned long max_pfn_mapped; | |
128 | ||
e808bae2 | 129 | #ifdef CONFIG_DMI |
796216a5 | 130 | RESERVE_BRK(dmi_alloc, 65536); |
e808bae2 | 131 | #endif |
796216a5 | 132 | |
c0b5842a | 133 | |
93dbda7c JF |
134 | static __initdata unsigned long _brk_start = (unsigned long)__brk_base; |
135 | unsigned long _brk_end = (unsigned long)__brk_base; | |
136 | ||
c0b5842a IM |
137 | #ifdef CONFIG_X86_64 |
138 | int default_cpu_present_to_apicid(int mps_cpu) | |
139 | { | |
140 | return __default_cpu_present_to_apicid(mps_cpu); | |
141 | } | |
142 | ||
e11dadab | 143 | int default_check_phys_apicid_present(int phys_apicid) |
c0b5842a | 144 | { |
e11dadab | 145 | return __default_check_phys_apicid_present(phys_apicid); |
c0b5842a IM |
146 | } |
147 | #endif | |
148 | ||
217b8ce8 | 149 | struct boot_params boot_params; |
217b8ce8 | 150 | |
4046d6e8 LT |
151 | /* |
152 | * Machine setup.. | |
153 | */ | |
154 | static struct resource data_resource = { | |
155 | .name = "Kernel data", | |
156 | .start = 0, | |
157 | .end = 0, | |
158 | .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM | |
159 | }; | |
160 | ||
161 | static struct resource code_resource = { | |
162 | .name = "Kernel code", | |
163 | .start = 0, | |
164 | .end = 0, | |
165 | .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM | |
166 | }; | |
167 | ||
168 | static struct resource bss_resource = { | |
169 | .name = "Kernel bss", | |
170 | .start = 0, | |
171 | .end = 0, | |
172 | .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM | |
173 | }; | |
174 | ||
175 | ||
7dea23ec | 176 | #ifdef CONFIG_X86_32 |
6415813b MK |
177 | /* cpu data as detected by the assembly code in head_32.S */ |
178 | struct cpuinfo_x86 new_cpu_data; | |
179 | ||
1da177e4 | 180 | /* common cpu data for all cpus */ |
6415813b | 181 | struct cpuinfo_x86 boot_cpu_data __read_mostly; |
129f6946 | 182 | EXPORT_SYMBOL(boot_cpu_data); |
1da177e4 | 183 | |
0c254e38 AS |
184 | unsigned int def_to_bigsmp; |
185 | ||
1da177e4 LT |
186 | /* for MCA, but anyone else can use it if they want */ |
187 | unsigned int machine_id; | |
188 | unsigned int machine_submodel_id; | |
189 | unsigned int BIOS_revision; | |
1da177e4 | 190 | |
7dea23ec YL |
191 | struct apm_info apm_info; |
192 | EXPORT_SYMBOL(apm_info); | |
193 | ||
194 | #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \ | |
195 | defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE) | |
196 | struct ist_info ist_info; | |
197 | EXPORT_SYMBOL(ist_info); | |
198 | #else | |
199 | struct ist_info ist_info; | |
200 | #endif | |
201 | ||
202 | #else | |
ed26dbe5 JF |
203 | struct cpuinfo_x86 boot_cpu_data __read_mostly = { |
204 | .x86_phys_bits = MAX_PHYSMEM_BITS, | |
205 | }; | |
7dea23ec YL |
206 | EXPORT_SYMBOL(boot_cpu_data); |
207 | #endif | |
208 | ||
209 | ||
210 | #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64) | |
404f6aac | 211 | __visible unsigned long mmu_cr4_features __ro_after_init; |
7dea23ec | 212 | #else |
404f6aac | 213 | __visible unsigned long mmu_cr4_features __ro_after_init = X86_CR4_PAE; |
7dea23ec YL |
214 | #endif |
215 | ||
5031296c PA |
216 | /* Boot loader ID and version as integers, for the benefit of proc_dointvec */ |
217 | int bootloader_type, bootloader_version; | |
1da177e4 | 218 | |
1da177e4 LT |
219 | /* |
220 | * Setup options | |
221 | */ | |
1da177e4 | 222 | struct screen_info screen_info; |
129f6946 | 223 | EXPORT_SYMBOL(screen_info); |
1da177e4 | 224 | struct edid_info edid_info; |
5e518d76 | 225 | EXPORT_SYMBOL_GPL(edid_info); |
1da177e4 | 226 | |
1da177e4 LT |
227 | extern int root_mountflags; |
228 | ||
e44b7b75 | 229 | unsigned long saved_video_mode; |
1da177e4 | 230 | |
cf8fa920 | 231 | #define RAMDISK_IMAGE_START_MASK 0x07FF |
1da177e4 | 232 | #define RAMDISK_PROMPT_FLAG 0x8000 |
cf8fa920 | 233 | #define RAMDISK_LOAD_FLAG 0x4000 |
1da177e4 | 234 | |
4e498b66 | 235 | static char __initdata command_line[COMMAND_LINE_SIZE]; |
516cbf37 TB |
236 | #ifdef CONFIG_CMDLINE_BOOL |
237 | static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE; | |
238 | #endif | |
1da177e4 | 239 | |
1da177e4 LT |
240 | #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE) |
241 | struct edd edd; | |
242 | #ifdef CONFIG_EDD_MODULE | |
243 | EXPORT_SYMBOL(edd); | |
244 | #endif | |
245 | /** | |
246 | * copy_edd() - Copy the BIOS EDD information | |
247 | * from boot_params into a safe place. | |
248 | * | |
249 | */ | |
9eaa192d | 250 | static inline void __init copy_edd(void) |
1da177e4 | 251 | { |
30c82645 PA |
252 | memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer, |
253 | sizeof(edd.mbr_signature)); | |
254 | memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info)); | |
255 | edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries; | |
256 | edd.edd_info_nr = boot_params.eddbuf_entries; | |
1da177e4 LT |
257 | } |
258 | #else | |
9eaa192d | 259 | static inline void __init copy_edd(void) |
1da177e4 LT |
260 | { |
261 | } | |
262 | #endif | |
263 | ||
5368a2be PA |
264 | void * __init extend_brk(size_t size, size_t align) |
265 | { | |
266 | size_t mask = align - 1; | |
267 | void *ret; | |
268 | ||
269 | BUG_ON(_brk_start == 0); | |
270 | BUG_ON(align & mask); | |
271 | ||
272 | _brk_end = (_brk_end + mask) & ~mask; | |
273 | BUG_ON((char *)(_brk_end + size) > __brk_limit); | |
274 | ||
275 | ret = (void *)_brk_end; | |
276 | _brk_end += size; | |
277 | ||
278 | memset(ret, 0, size); | |
279 | ||
280 | return ret; | |
281 | } | |
282 | ||
148b2098 | 283 | #ifdef CONFIG_X86_32 |
e5f15b45 | 284 | static void __init cleanup_highmap(void) |
f005fe12 YL |
285 | { |
286 | } | |
854c879f PE |
287 | #endif |
288 | ||
5368a2be PA |
289 | static void __init reserve_brk(void) |
290 | { | |
291 | if (_brk_end > _brk_start) | |
fc8d7826 AD |
292 | memblock_reserve(__pa_symbol(_brk_start), |
293 | _brk_end - _brk_start); | |
5368a2be PA |
294 | |
295 | /* Mark brk area as locked down and no longer taking any | |
296 | new allocations */ | |
297 | _brk_start = 0; | |
298 | } | |
299 | ||
5aa3d718 BP |
300 | u64 relocated_ramdisk; |
301 | ||
cf8fa920 PA |
302 | #ifdef CONFIG_BLK_DEV_INITRD |
303 | ||
a8a51a88 YL |
304 | static u64 __init get_ramdisk_image(void) |
305 | { | |
306 | u64 ramdisk_image = boot_params.hdr.ramdisk_image; | |
307 | ||
ee92d815 YL |
308 | ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32; |
309 | ||
a8a51a88 YL |
310 | return ramdisk_image; |
311 | } | |
312 | static u64 __init get_ramdisk_size(void) | |
313 | { | |
314 | u64 ramdisk_size = boot_params.hdr.ramdisk_size; | |
315 | ||
ee92d815 YL |
316 | ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32; |
317 | ||
a8a51a88 YL |
318 | return ramdisk_size; |
319 | } | |
320 | ||
eb1379cb | 321 | static void __init relocate_initrd(void) |
cf8fa920 | 322 | { |
c967da6a | 323 | /* Assume only end is not page aligned */ |
a8a51a88 YL |
324 | u64 ramdisk_image = get_ramdisk_image(); |
325 | u64 ramdisk_size = get_ramdisk_size(); | |
c967da6a | 326 | u64 area_size = PAGE_ALIGN(ramdisk_size); |
cf8fa920 | 327 | |
e8c57d40 | 328 | /* We need to move the initrd down into directly mapped mem */ |
5aa3d718 BP |
329 | relocated_ramdisk = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped), |
330 | area_size, PAGE_SIZE); | |
cf8fa920 | 331 | |
5aa3d718 | 332 | if (!relocated_ramdisk) |
3945e2c9 | 333 | panic("Cannot find place for new RAMDISK of size %lld\n", |
5aa3d718 | 334 | ramdisk_size); |
3945e2c9 | 335 | |
e8c57d40 | 336 | /* Note: this includes all the mem currently occupied by |
cf8fa920 | 337 | the initrd, we rely on that fact to keep the data intact. */ |
5aa3d718 BP |
338 | memblock_reserve(relocated_ramdisk, area_size); |
339 | initrd_start = relocated_ramdisk + PAGE_OFFSET; | |
cf8fa920 | 340 | initrd_end = initrd_start + ramdisk_size; |
365811d6 | 341 | printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n", |
5aa3d718 | 342 | relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1); |
cf8fa920 | 343 | |
5dd2c4bd MS |
344 | copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size); |
345 | ||
365811d6 BH |
346 | printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to" |
347 | " [mem %#010llx-%#010llx]\n", | |
ba5b14cc | 348 | ramdisk_image, ramdisk_image + ramdisk_size - 1, |
5aa3d718 | 349 | relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1); |
eb1379cb | 350 | } |
9a27f5c5 | 351 | |
1b8c78be YL |
352 | static void __init early_reserve_initrd(void) |
353 | { | |
354 | /* Assume only end is not page aligned */ | |
a8a51a88 YL |
355 | u64 ramdisk_image = get_ramdisk_image(); |
356 | u64 ramdisk_size = get_ramdisk_size(); | |
1b8c78be YL |
357 | u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size); |
358 | ||
359 | if (!boot_params.hdr.type_of_loader || | |
360 | !ramdisk_image || !ramdisk_size) | |
361 | return; /* No initrd provided by bootloader */ | |
362 | ||
363 | memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image); | |
364 | } | |
eb1379cb YL |
365 | static void __init reserve_initrd(void) |
366 | { | |
c967da6a | 367 | /* Assume only end is not page aligned */ |
a8a51a88 YL |
368 | u64 ramdisk_image = get_ramdisk_image(); |
369 | u64 ramdisk_size = get_ramdisk_size(); | |
c967da6a | 370 | u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size); |
e8c57d40 | 371 | u64 mapped_size; |
eb1379cb YL |
372 | |
373 | if (!boot_params.hdr.type_of_loader || | |
374 | !ramdisk_image || !ramdisk_size) | |
375 | return; /* No initrd provided by bootloader */ | |
376 | ||
377 | initrd_start = 0; | |
378 | ||
595ad9af | 379 | mapped_size = memblock_mem_size(max_pfn_mapped); |
e8c57d40 | 380 | if (ramdisk_size >= (mapped_size>>1)) |
ab7b64e9 PJ |
381 | panic("initrd too large to handle, " |
382 | "disabling initrd (%lld needed, %lld available)\n", | |
e8c57d40 | 383 | ramdisk_size, mapped_size>>1); |
eb1379cb | 384 | |
365811d6 BH |
385 | printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image, |
386 | ramdisk_end - 1); | |
eb1379cb | 387 | |
74f27655 | 388 | if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image), |
e8c57d40 YL |
389 | PFN_DOWN(ramdisk_end))) { |
390 | /* All are mapped, easy case */ | |
eb1379cb YL |
391 | initrd_start = ramdisk_image + PAGE_OFFSET; |
392 | initrd_end = initrd_start + ramdisk_size; | |
393 | return; | |
394 | } | |
395 | ||
eb1379cb | 396 | relocate_initrd(); |
8c5dd8f4 | 397 | |
24aa0788 | 398 | memblock_free(ramdisk_image, ramdisk_end - ramdisk_image); |
cf8fa920 | 399 | } |
af06f8b7 | 400 | |
225c37d7 | 401 | #else |
1b8c78be YL |
402 | static void __init early_reserve_initrd(void) |
403 | { | |
404 | } | |
eb1379cb | 405 | static void __init reserve_initrd(void) |
225c37d7 YL |
406 | { |
407 | } | |
cf8fa920 PA |
408 | #endif /* CONFIG_BLK_DEV_INITRD */ |
409 | ||
29f784e3 | 410 | static void __init parse_setup_data(void) |
257b0fde YL |
411 | { |
412 | struct setup_data *data; | |
30e46b57 | 413 | u64 pa_data, pa_next; |
257b0fde | 414 | |
257b0fde YL |
415 | pa_data = boot_params.hdr.setup_data; |
416 | while (pa_data) { | |
7389882c | 417 | u32 data_len, data_type; |
f1c2b357 | 418 | |
7389882c | 419 | data = early_memremap(pa_data, sizeof(*data)); |
f1c2b357 | 420 | data_len = data->len + sizeof(struct setup_data); |
30e46b57 LC |
421 | data_type = data->type; |
422 | pa_next = data->next; | |
8d4a40bc | 423 | early_memunmap(data, sizeof(*data)); |
f1c2b357 | 424 | |
30e46b57 | 425 | switch (data_type) { |
257b0fde | 426 | case SETUP_E820_EXT: |
914053c0 | 427 | e820__memory_setup_extended(pa_data, data_len); |
257b0fde | 428 | break; |
da6b737b SAS |
429 | case SETUP_DTB: |
430 | add_dtb(pa_data); | |
257b0fde | 431 | break; |
1fec0533 DY |
432 | case SETUP_EFI: |
433 | parse_efi_setup(pa_data, data_len); | |
434 | break; | |
257b0fde YL |
435 | default: |
436 | break; | |
437 | } | |
30e46b57 | 438 | pa_data = pa_next; |
257b0fde YL |
439 | } |
440 | } | |
441 | ||
a9ce6bc1 | 442 | static void __init memblock_x86_reserve_range_setup_data(void) |
a0a0becd YL |
443 | { |
444 | struct setup_data *data; | |
445 | u64 pa_data; | |
a0a0becd | 446 | |
a0a0becd YL |
447 | pa_data = boot_params.hdr.setup_data; |
448 | while (pa_data) { | |
88b4c146 | 449 | data = early_memremap(pa_data, sizeof(*data)); |
24aa0788 | 450 | memblock_reserve(pa_data, sizeof(*data) + data->len); |
a0a0becd | 451 | pa_data = data->next; |
8d4a40bc | 452 | early_memunmap(data, sizeof(*data)); |
a0a0becd YL |
453 | } |
454 | } | |
455 | ||
ccb4defa YL |
456 | /* |
457 | * --------- Crashkernel reservation ------------------------------ | |
458 | */ | |
459 | ||
2965faa5 | 460 | #ifdef CONFIG_KEXEC_CORE |
32105f7f | 461 | |
606134f7 BP |
462 | /* 16M alignment for crash kernel regions */ |
463 | #define CRASH_ALIGN (16 << 20) | |
464 | ||
7f8595bf PA |
465 | /* |
466 | * Keep the crash kernel below this limit. On 32 bits earlier kernels | |
467 | * would limit the kernel to the low 512 MiB due to mapping restrictions. | |
55a20ee7 | 468 | * On 64bit, old kexec-tools need to under 896MiB. |
7f8595bf PA |
469 | */ |
470 | #ifdef CONFIG_X86_32 | |
97eac21b BP |
471 | # define CRASH_ADDR_LOW_MAX (512 << 20) |
472 | # define CRASH_ADDR_HIGH_MAX (512 << 20) | |
7f8595bf | 473 | #else |
97eac21b BP |
474 | # define CRASH_ADDR_LOW_MAX (896UL << 20) |
475 | # define CRASH_ADDR_HIGH_MAX MAXMEM | |
7f8595bf PA |
476 | #endif |
477 | ||
eb6db83d | 478 | static int __init reserve_crashkernel_low(void) |
0212f915 YL |
479 | { |
480 | #ifdef CONFIG_X86_64 | |
f56d5578 | 481 | unsigned long long base, low_base = 0, low_size = 0; |
0212f915 | 482 | unsigned long total_low_mem; |
0212f915 YL |
483 | int ret; |
484 | ||
97eac21b BP |
485 | total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT)); |
486 | ||
adbc742b | 487 | /* crashkernel=Y,low */ |
97eac21b | 488 | ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base); |
f56d5578 | 489 | if (ret) { |
c729de8f YL |
490 | /* |
491 | * two parts from lib/swiotlb.c: | |
94fb9334 JR |
492 | * -swiotlb size: user-specified with swiotlb= or default. |
493 | * | |
494 | * -swiotlb overflow buffer: now hardcoded to 32k. We round it | |
495 | * to 8M for other buffers that may need to stay low too. Also | |
496 | * make sure we allocate enough extra low memory so that we | |
497 | * don't run out of DMA buffers for 32-bit devices. | |
c729de8f | 498 | */ |
97eac21b | 499 | low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20); |
c729de8f | 500 | } else { |
adbc742b | 501 | /* passed with crashkernel=0,low ? */ |
c729de8f | 502 | if (!low_size) |
eb6db83d | 503 | return 0; |
c729de8f | 504 | } |
0212f915 | 505 | |
606134f7 | 506 | low_base = memblock_find_in_range(low_size, 1ULL << 32, low_size, CRASH_ALIGN); |
0212f915 | 507 | if (!low_base) { |
eb6db83d BH |
508 | pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n", |
509 | (unsigned long)(low_size >> 20)); | |
510 | return -ENOMEM; | |
0212f915 | 511 | } |
0212f915 | 512 | |
6f376057 BP |
513 | ret = memblock_reserve(low_base, low_size); |
514 | if (ret) { | |
515 | pr_err("%s: Error reserving crashkernel low memblock.\n", __func__); | |
516 | return ret; | |
0212f915 YL |
517 | } |
518 | ||
0212f915 | 519 | pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n", |
97eac21b BP |
520 | (unsigned long)(low_size >> 20), |
521 | (unsigned long)(low_base >> 20), | |
522 | (unsigned long)(total_low_mem >> 20)); | |
523 | ||
0212f915 YL |
524 | crashk_low_res.start = low_base; |
525 | crashk_low_res.end = low_base + low_size - 1; | |
526 | insert_resource(&iomem_resource, &crashk_low_res); | |
7f8595bf | 527 | #endif |
eb6db83d | 528 | return 0; |
0212f915 | 529 | } |
7f8595bf | 530 | |
29f784e3 | 531 | static void __init reserve_crashkernel(void) |
ccb4defa | 532 | { |
f56d5578 | 533 | unsigned long long crash_size, crash_base, total_mem; |
55a20ee7 | 534 | bool high = false; |
ccb4defa YL |
535 | int ret; |
536 | ||
09c71bfd | 537 | total_mem = memblock_phys_mem_size(); |
ccb4defa | 538 | |
55a20ee7 | 539 | /* crashkernel=XM */ |
97eac21b | 540 | ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base); |
55a20ee7 | 541 | if (ret != 0 || crash_size <= 0) { |
adbc742b | 542 | /* crashkernel=X,high */ |
55a20ee7 | 543 | ret = parse_crashkernel_high(boot_command_line, total_mem, |
97eac21b | 544 | &crash_size, &crash_base); |
55a20ee7 YL |
545 | if (ret != 0 || crash_size <= 0) |
546 | return; | |
547 | high = true; | |
548 | } | |
32105f7f BW |
549 | |
550 | /* 0 means: find the address automatically */ | |
551 | if (crash_base <= 0) { | |
9f4c1396 | 552 | /* |
a8d4c824 XP |
553 | * Set CRASH_ADDR_LOW_MAX upper bound for crash memory, |
554 | * as old kexec-tools loads bzImage below that, unless | |
555 | * "crashkernel=size[KMG],high" is specified. | |
9f4c1396 | 556 | */ |
606134f7 | 557 | crash_base = memblock_find_in_range(CRASH_ALIGN, |
97eac21b BP |
558 | high ? CRASH_ADDR_HIGH_MAX |
559 | : CRASH_ADDR_LOW_MAX, | |
606134f7 | 560 | crash_size, CRASH_ALIGN); |
1f5026a7 | 561 | if (!crash_base) { |
44280733 | 562 | pr_info("crashkernel reservation failed - No suitable area found.\n"); |
ccb4defa YL |
563 | return; |
564 | } | |
0212f915 | 565 | |
32105f7f | 566 | } else { |
44280733 YL |
567 | unsigned long long start; |
568 | ||
9f4c1396 | 569 | start = memblock_find_in_range(crash_base, |
97eac21b BP |
570 | crash_base + crash_size, |
571 | crash_size, 1 << 20); | |
44280733 YL |
572 | if (start != crash_base) { |
573 | pr_info("crashkernel reservation failed - memory is in use.\n"); | |
ccb4defa YL |
574 | return; |
575 | } | |
32105f7f | 576 | } |
6f376057 BP |
577 | ret = memblock_reserve(crash_base, crash_size); |
578 | if (ret) { | |
579 | pr_err("%s: Error reserving crashkernel memblock.\n", __func__); | |
580 | return; | |
581 | } | |
ccb4defa | 582 | |
eb6db83d BH |
583 | if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) { |
584 | memblock_free(crash_base, crash_size); | |
585 | return; | |
586 | } | |
ccb4defa | 587 | |
f56d5578 BP |
588 | pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n", |
589 | (unsigned long)(crash_size >> 20), | |
590 | (unsigned long)(crash_base >> 20), | |
591 | (unsigned long)(total_mem >> 20)); | |
ccb4defa | 592 | |
32105f7f BW |
593 | crashk_res.start = crash_base; |
594 | crashk_res.end = crash_base + crash_size - 1; | |
595 | insert_resource(&iomem_resource, &crashk_res); | |
ccb4defa YL |
596 | } |
597 | #else | |
29f784e3 | 598 | static void __init reserve_crashkernel(void) |
ccb4defa YL |
599 | { |
600 | } | |
601 | #endif | |
602 | ||
bdba0e70 YL |
603 | static struct resource standard_io_resources[] = { |
604 | { .name = "dma1", .start = 0x00, .end = 0x1f, | |
605 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
606 | { .name = "pic1", .start = 0x20, .end = 0x21, | |
607 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
608 | { .name = "timer0", .start = 0x40, .end = 0x43, | |
609 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
610 | { .name = "timer1", .start = 0x50, .end = 0x53, | |
611 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
612 | { .name = "keyboard", .start = 0x60, .end = 0x60, | |
613 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
614 | { .name = "keyboard", .start = 0x64, .end = 0x64, | |
615 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
616 | { .name = "dma page reg", .start = 0x80, .end = 0x8f, | |
617 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
618 | { .name = "pic2", .start = 0xa0, .end = 0xa1, | |
619 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
620 | { .name = "dma2", .start = 0xc0, .end = 0xdf, | |
621 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
622 | { .name = "fpu", .start = 0xf0, .end = 0xff, | |
623 | .flags = IORESOURCE_BUSY | IORESOURCE_IO } | |
624 | }; | |
625 | ||
8fee697d | 626 | void __init reserve_standard_io_resources(void) |
bdba0e70 YL |
627 | { |
628 | int i; | |
629 | ||
630 | /* request I/O space for devices used on all i[345]86 PCs */ | |
631 | for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++) | |
632 | request_resource(&ioport_resource, &standard_io_resources[i]); | |
633 | ||
634 | } | |
635 | ||
042be38e YL |
636 | static __init void reserve_ibft_region(void) |
637 | { | |
638 | unsigned long addr, size = 0; | |
639 | ||
640 | addr = find_ibft_region(&size); | |
641 | ||
642 | if (size) | |
24aa0788 | 643 | memblock_reserve(addr, size); |
042be38e YL |
644 | } |
645 | ||
a9acc536 JB |
646 | static bool __init snb_gfx_workaround_needed(void) |
647 | { | |
e43b3cec | 648 | #ifdef CONFIG_PCI |
a9acc536 JB |
649 | int i; |
650 | u16 vendor, devid; | |
ab3cd867 | 651 | static const __initconst u16 snb_ids[] = { |
a9acc536 JB |
652 | 0x0102, |
653 | 0x0112, | |
654 | 0x0122, | |
655 | 0x0106, | |
656 | 0x0116, | |
657 | 0x0126, | |
658 | 0x010a, | |
659 | }; | |
660 | ||
661 | /* Assume no if something weird is going on with PCI */ | |
662 | if (!early_pci_allowed()) | |
663 | return false; | |
664 | ||
665 | vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID); | |
666 | if (vendor != 0x8086) | |
667 | return false; | |
668 | ||
669 | devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID); | |
670 | for (i = 0; i < ARRAY_SIZE(snb_ids); i++) | |
671 | if (devid == snb_ids[i]) | |
672 | return true; | |
e43b3cec | 673 | #endif |
a9acc536 JB |
674 | |
675 | return false; | |
676 | } | |
677 | ||
678 | /* | |
679 | * Sandy Bridge graphics has trouble with certain ranges, exclude | |
680 | * them from allocation. | |
681 | */ | |
682 | static void __init trim_snb_memory(void) | |
683 | { | |
ab3cd867 | 684 | static const __initconst unsigned long bad_pages[] = { |
a9acc536 JB |
685 | 0x20050000, |
686 | 0x20110000, | |
687 | 0x20130000, | |
688 | 0x20138000, | |
689 | 0x40004000, | |
690 | }; | |
691 | int i; | |
692 | ||
693 | if (!snb_gfx_workaround_needed()) | |
694 | return; | |
695 | ||
696 | printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n"); | |
697 | ||
698 | /* | |
699 | * Reserve all memory below the 1 MB mark that has not | |
700 | * already been reserved. | |
701 | */ | |
702 | memblock_reserve(0, 1<<20); | |
703 | ||
704 | for (i = 0; i < ARRAY_SIZE(bad_pages); i++) { | |
705 | if (memblock_reserve(bad_pages[i], PAGE_SIZE)) | |
706 | printk(KERN_WARNING "failed to reserve 0x%08lx\n", | |
707 | bad_pages[i]); | |
708 | } | |
709 | } | |
710 | ||
711 | /* | |
712 | * Here we put platform-specific memory range workarounds, i.e. | |
713 | * memory known to be corrupt or otherwise in need to be reserved on | |
714 | * specific platforms. | |
715 | * | |
716 | * If this gets used more widely it could use a real dispatch mechanism. | |
717 | */ | |
718 | static void __init trim_platform_memory_ranges(void) | |
719 | { | |
720 | trim_snb_memory(); | |
721 | } | |
722 | ||
1b5576e6 YL |
723 | static void __init trim_bios_range(void) |
724 | { | |
725 | /* | |
726 | * A special case is the first 4Kb of memory; | |
727 | * This is a BIOS owned area, not kernel ram, but generally | |
728 | * not listed as such in the E820 table. | |
d0cd7425 PA |
729 | * |
730 | * This typically reserves additional memory (64KiB by default) | |
731 | * since some BIOSes are known to corrupt low memory. See the | |
9ea77bdb | 732 | * Kconfig help text for X86_RESERVE_LOW. |
1b5576e6 | 733 | */ |
09821ff1 | 734 | e820__range_update(0, PAGE_SIZE, E820_TYPE_RAM, E820_TYPE_RESERVED); |
d0cd7425 | 735 | |
1b5576e6 YL |
736 | /* |
737 | * special case: Some BIOSen report the PC BIOS | |
738 | * area (640->1Mb) as ram even though it is not. | |
739 | * take them out. | |
740 | */ | |
09821ff1 | 741 | e820__range_remove(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_TYPE_RAM, 1); |
a9acc536 | 742 | |
f9748fa0 | 743 | e820__update_table(e820_table); |
1b5576e6 YL |
744 | } |
745 | ||
b422a309 YL |
746 | /* called before trim_bios_range() to spare extra sanitize */ |
747 | static void __init e820_add_kernel_range(void) | |
748 | { | |
749 | u64 start = __pa_symbol(_text); | |
750 | u64 size = __pa_symbol(_end) - start; | |
751 | ||
752 | /* | |
09821ff1 | 753 | * Complain if .text .data and .bss are not marked as E820_TYPE_RAM and |
b422a309 YL |
754 | * attempt to fix it by adding the range. We may have a confused BIOS, |
755 | * or the user may have used memmap=exactmap or memmap=xxM$yyM to | |
756 | * exclude kernel range. If we really are running on top non-RAM, | |
757 | * we will crash later anyways. | |
758 | */ | |
09821ff1 | 759 | if (e820__mapped_all(start, start + size, E820_TYPE_RAM)) |
b422a309 YL |
760 | return; |
761 | ||
09821ff1 IM |
762 | pr_warn(".text .data .bss are not marked as E820_TYPE_RAM!\n"); |
763 | e820__range_remove(start, size, E820_TYPE_RAM, 0); | |
764 | e820__range_add(start, size, E820_TYPE_RAM); | |
b422a309 YL |
765 | } |
766 | ||
95c96084 PA |
767 | static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10; |
768 | ||
9ea77bdb PA |
769 | static int __init parse_reservelow(char *p) |
770 | { | |
771 | unsigned long long size; | |
772 | ||
773 | if (!p) | |
774 | return -EINVAL; | |
775 | ||
776 | size = memparse(p, &p); | |
777 | ||
778 | if (size < 4096) | |
779 | size = 4096; | |
780 | ||
781 | if (size > 640*1024) | |
782 | size = 640*1024; | |
783 | ||
784 | reserve_low = size; | |
785 | ||
786 | return 0; | |
787 | } | |
788 | ||
789 | early_param("reservelow", parse_reservelow); | |
790 | ||
95c96084 PA |
791 | static void __init trim_low_memory_range(void) |
792 | { | |
793 | memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE)); | |
794 | } | |
795 | ||
f32360ef KC |
796 | /* |
797 | * Dump out kernel offset information on panic. | |
798 | */ | |
799 | static int | |
800 | dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p) | |
801 | { | |
78cac48c BP |
802 | if (kaslr_enabled()) { |
803 | pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n", | |
4545c898 | 804 | kaslr_offset(), |
78cac48c BP |
805 | __START_KERNEL, |
806 | __START_KERNEL_map, | |
807 | MODULES_VADDR-1); | |
808 | } else { | |
809 | pr_emerg("Kernel Offset: disabled\n"); | |
810 | } | |
f32360ef KC |
811 | |
812 | return 0; | |
813 | } | |
814 | ||
dd759d93 LB |
815 | static void __init simple_udelay_calibration(void) |
816 | { | |
817 | unsigned int tsc_khz, cpu_khz; | |
818 | unsigned long lpj; | |
819 | ||
820 | if (!boot_cpu_has(X86_FEATURE_TSC)) | |
821 | return; | |
822 | ||
823 | cpu_khz = x86_platform.calibrate_cpu(); | |
824 | tsc_khz = x86_platform.calibrate_tsc(); | |
825 | ||
826 | tsc_khz = tsc_khz ? : cpu_khz; | |
827 | if (!tsc_khz) | |
828 | return; | |
829 | ||
830 | lpj = tsc_khz * 1000; | |
831 | do_div(lpj, HZ); | |
832 | loops_per_jiffy = lpj; | |
833 | } | |
834 | ||
1da177e4 LT |
835 | /* |
836 | * Determine if we were loaded by an EFI loader. If so, then we have also been | |
837 | * passed the efi memmap, systab, etc., so we should use these data structures | |
838 | * for initialization. Note, the efi init code path is determined by the | |
839 | * global efi_enabled. This allows the same kernel image to be used on existing | |
840 | * systems (with a traditional BIOS) as well as on EFI systems. | |
841 | */ | |
76934ed4 YL |
842 | /* |
843 | * setup_arch - architecture-specific boot-time initializations | |
844 | * | |
845 | * Note: On x86_64, fixmaps are ready for use even before this is called. | |
846 | */ | |
847 | ||
1da177e4 LT |
848 | void __init setup_arch(char **cmdline_p) |
849 | { | |
6c902b65 YL |
850 | memblock_reserve(__pa_symbol(_text), |
851 | (unsigned long)__bss_stop - (unsigned long)_text); | |
852 | ||
1b8c78be YL |
853 | early_reserve_initrd(); |
854 | ||
6c902b65 YL |
855 | /* |
856 | * At this point everything still needed from the boot loader | |
857 | * or BIOS or kernel text should be early reserved or marked not | |
858 | * RAM in e820. All other memory is free game. | |
859 | */ | |
860 | ||
76934ed4 | 861 | #ifdef CONFIG_X86_32 |
1da177e4 | 862 | memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data)); |
b40827fa BP |
863 | |
864 | /* | |
865 | * copy kernel address range established so far and switch | |
866 | * to the proper swapper page table | |
867 | */ | |
868 | clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY, | |
869 | initial_page_table + KERNEL_PGD_BOUNDARY, | |
870 | KERNEL_PGD_PTRS); | |
871 | ||
872 | load_cr3(swapper_pg_dir); | |
2075244f BD |
873 | /* |
874 | * Note: Quark X1000 CPUs advertise PGE incorrectly and require | |
875 | * a cr3 based tlb flush, so the following __flush_tlb_all() | |
876 | * will not flush anything because the cpu quirk which clears | |
877 | * X86_FEATURE_PGE has not been invoked yet. Though due to the | |
878 | * load_cr3() above the TLB has been flushed already. The | |
879 | * quirk is invoked before subsequent calls to __flush_tlb_all() | |
880 | * so proper operation is guaranteed. | |
881 | */ | |
b40827fa | 882 | __flush_tlb_all(); |
76934ed4 YL |
883 | #else |
884 | printk(KERN_INFO "Command line: %s\n", boot_command_line); | |
885 | #endif | |
1da177e4 | 886 | |
9863c90f AK |
887 | /* |
888 | * If we have OLPC OFW, we might end up relocating the fixmap due to | |
889 | * reserve_top(), so do this before touching the ioremap area. | |
890 | */ | |
fd699c76 AS |
891 | olpc_ofw_detect(); |
892 | ||
29c84391 | 893 | early_trap_init(); |
9e882c92 | 894 | early_cpu_init(); |
1a98fd14 JF |
895 | early_ioremap_init(); |
896 | ||
fd699c76 AS |
897 | setup_olpc_ofw_pgd(); |
898 | ||
30c82645 PA |
899 | ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev); |
900 | screen_info = boot_params.screen_info; | |
901 | edid_info = boot_params.edid_info; | |
76934ed4 | 902 | #ifdef CONFIG_X86_32 |
30c82645 PA |
903 | apm_info.bios = boot_params.apm_bios_info; |
904 | ist_info = boot_params.ist_info; | |
76934ed4 YL |
905 | #endif |
906 | saved_video_mode = boot_params.hdr.vid_mode; | |
30c82645 | 907 | bootloader_type = boot_params.hdr.type_of_loader; |
5031296c PA |
908 | if ((bootloader_type >> 4) == 0xe) { |
909 | bootloader_type &= 0xf; | |
910 | bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4; | |
911 | } | |
912 | bootloader_version = bootloader_type & 0xf; | |
913 | bootloader_version |= boot_params.hdr.ext_loader_ver << 4; | |
1da177e4 LT |
914 | |
915 | #ifdef CONFIG_BLK_DEV_RAM | |
30c82645 PA |
916 | rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK; |
917 | rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0); | |
918 | rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0); | |
1da177e4 | 919 | #endif |
7465252e YL |
920 | #ifdef CONFIG_EFI |
921 | if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature, | |
9402973d | 922 | EFI32_LOADER_SIGNATURE, 4)) { |
3e909599 | 923 | set_bit(EFI_BOOT, &efi.flags); |
1adbfa35 | 924 | } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature, |
9402973d | 925 | EFI64_LOADER_SIGNATURE, 4)) { |
3e909599 MF |
926 | set_bit(EFI_BOOT, &efi.flags); |
927 | set_bit(EFI_64BIT, &efi.flags); | |
7465252e | 928 | } |
83e68189 MF |
929 | |
930 | if (efi_enabled(EFI_BOOT)) | |
931 | efi_memblock_x86_reserve_range(); | |
7465252e YL |
932 | #endif |
933 | ||
42bbdb43 | 934 | x86_init.oem.arch_setup(); |
2215e69d | 935 | |
419afdf5 | 936 | iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1; |
103e2063 | 937 | e820__memory_setup(); |
28bb2237 YL |
938 | parse_setup_data(); |
939 | ||
1da177e4 LT |
940 | copy_edd(); |
941 | ||
30c82645 | 942 | if (!boot_params.hdr.root_flags) |
1da177e4 LT |
943 | root_mountflags &= ~MS_RDONLY; |
944 | init_mm.start_code = (unsigned long) _text; | |
945 | init_mm.end_code = (unsigned long) _etext; | |
946 | init_mm.end_data = (unsigned long) _edata; | |
93dbda7c | 947 | init_mm.brk = _brk_end; |
fe3d197f DH |
948 | |
949 | mpx_mm_init(&init_mm); | |
1da177e4 | 950 | |
4046d6e8 LT |
951 | code_resource.start = __pa_symbol(_text); |
952 | code_resource.end = __pa_symbol(_etext)-1; | |
953 | data_resource.start = __pa_symbol(_etext); | |
954 | data_resource.end = __pa_symbol(_edata)-1; | |
955 | bss_resource.start = __pa_symbol(__bss_start); | |
956 | bss_resource.end = __pa_symbol(__bss_stop)-1; | |
957 | ||
516cbf37 TB |
958 | #ifdef CONFIG_CMDLINE_BOOL |
959 | #ifdef CONFIG_CMDLINE_OVERRIDE | |
960 | strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE); | |
961 | #else | |
962 | if (builtin_cmdline[0]) { | |
963 | /* append boot loader cmdline to builtin */ | |
964 | strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE); | |
965 | strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE); | |
966 | strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE); | |
967 | } | |
968 | #endif | |
969 | #endif | |
970 | ||
eda6da92 YL |
971 | strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE); |
972 | *cmdline_p = command_line; | |
973 | ||
eda6da92 | 974 | /* |
4b0f3b81 KC |
975 | * x86_configure_nx() is called before parse_early_param() to detect |
976 | * whether hardware doesn't support NX (so that the early EHCI debug | |
977 | * console setup can safely call set_fixmap()). It may then be called | |
978 | * again from within noexec_setup() during parsing early parameters | |
979 | * to honor the respective command line option. | |
eda6da92 | 980 | */ |
4763ed4d | 981 | x86_configure_nx(); |
eda6da92 | 982 | |
dd759d93 LB |
983 | simple_udelay_calibration(); |
984 | ||
eda6da92 YL |
985 | parse_early_param(); |
986 | ||
39fa104d RA |
987 | #ifdef CONFIG_MEMORY_HOTPLUG |
988 | /* | |
989 | * Memory used by the kernel cannot be hot-removed because Linux | |
990 | * cannot migrate the kernel pages. When memory hotplug is | |
991 | * enabled, we should prevent memblock from allocating memory | |
992 | * for the kernel. | |
993 | * | |
994 | * ACPI SRAT records all hotpluggable memory ranges. But before | |
995 | * SRAT is parsed, we don't know about it. | |
996 | * | |
997 | * The kernel image is loaded into memory at very early time. We | |
998 | * cannot prevent this anyway. So on NUMA system, we set any | |
999 | * node the kernel resides in as un-hotpluggable. | |
1000 | * | |
1001 | * Since on modern servers, one node could have double-digit | |
1002 | * gigabytes memory, we can assume the memory around the kernel | |
1003 | * image is also un-hotpluggable. So before SRAT is parsed, just | |
1004 | * allocate memory near the kernel image to try the best to keep | |
1005 | * the kernel away from hotpluggable memory. | |
1006 | */ | |
1007 | if (movable_node_is_enabled()) | |
1008 | memblock_set_bottom_up(true); | |
1009 | #endif | |
1010 | ||
4b0f3b81 | 1011 | x86_report_nx(); |
0ad5bce7 | 1012 | |
28bb2237 | 1013 | /* after early param, so could get panic from serial */ |
a9ce6bc1 | 1014 | memblock_x86_reserve_range_setup_data(); |
28bb2237 | 1015 | |
76934ed4 | 1016 | if (acpi_mps_check()) { |
3eb11edc | 1017 | #ifdef CONFIG_X86_LOCAL_APIC |
76934ed4 | 1018 | disable_apic = 1; |
3eb11edc | 1019 | #endif |
988781dc | 1020 | setup_clear_cpu_cap(X86_FEATURE_APIC); |
3c999f14 YL |
1021 | } |
1022 | ||
dc7c65db LT |
1023 | #ifdef CONFIG_PCI |
1024 | if (pci_early_dump_regs) | |
1025 | early_dump_pci_devices(); | |
1026 | #endif | |
1027 | ||
1a127034 | 1028 | e820__reserve_setup_data(); |
9641bdaf | 1029 | e820__finish_early_params(); |
1a3f239d | 1030 | |
83e68189 | 1031 | if (efi_enabled(EFI_BOOT)) |
ff0c0874 BM |
1032 | efi_init(); |
1033 | ||
2216d199 | 1034 | dmi_scan_machine(); |
dd6dad42 | 1035 | dmi_memdev_walk(); |
98e5e1bf | 1036 | dmi_set_dump_stack_arch_desc(); |
2216d199 | 1037 | |
88b094fb AK |
1038 | /* |
1039 | * VMware detection requires dmi to be available, so this | |
1040 | * needs to be done after dmi_scan_machine, for the BP. | |
1041 | */ | |
2d826404 | 1042 | init_hypervisor_platform(); |
88b094fb | 1043 | |
f7cf5a5b | 1044 | x86_init.resources.probe_roms(); |
41c094fd | 1045 | |
4046d6e8 LT |
1046 | /* after parse_early_param, so could debug it */ |
1047 | insert_resource(&iomem_resource, &code_resource); | |
1048 | insert_resource(&iomem_resource, &data_resource); | |
1049 | insert_resource(&iomem_resource, &bss_resource); | |
1050 | ||
b422a309 | 1051 | e820_add_kernel_range(); |
1b5576e6 | 1052 | trim_bios_range(); |
76934ed4 | 1053 | #ifdef CONFIG_X86_32 |
cc9f7a0c | 1054 | if (ppro_with_ram_bug()) { |
09821ff1 IM |
1055 | e820__range_update(0x70000000ULL, 0x40000ULL, E820_TYPE_RAM, |
1056 | E820_TYPE_RESERVED); | |
f9748fa0 | 1057 | e820__update_table(e820_table); |
cc9f7a0c | 1058 | printk(KERN_INFO "fixed physical RAM map:\n"); |
be0c3f0f | 1059 | e820__print_table("bad_ppro"); |
cc9f7a0c | 1060 | } |
76934ed4 YL |
1061 | #else |
1062 | early_gart_iommu_check(); | |
1063 | #endif | |
cc9f7a0c | 1064 | |
7b2a0a6c YL |
1065 | /* |
1066 | * partially used pages are not usable - thus | |
1067 | * we are rounding upwards: | |
1068 | */ | |
0c6fc11a | 1069 | max_pfn = e820__end_of_ram_pfn(); |
7b2a0a6c | 1070 | |
093af8d7 YL |
1071 | /* update e820 for memory not covered by WB MTRRs */ |
1072 | mtrr_bp_init(); | |
2dc807b3 | 1073 | if (mtrr_trim_uncached_memory(max_pfn)) |
0c6fc11a | 1074 | max_pfn = e820__end_of_ram_pfn(); |
76c32418 | 1075 | |
8dd33030 IM |
1076 | max_possible_pfn = max_pfn; |
1077 | ||
c7d2361f TG |
1078 | /* |
1079 | * Define random base addresses for memory sections after max_pfn is | |
1080 | * defined and before each memory section base is used. | |
1081 | */ | |
1082 | kernel_randomize_memory(); | |
1083 | ||
76934ed4 | 1084 | #ifdef CONFIG_X86_32 |
4e29684c | 1085 | /* max_low_pfn get updated here */ |
2ec65f8b | 1086 | find_low_pfn_range(); |
76934ed4 | 1087 | #else |
06cd9a7d | 1088 | check_x2apic(); |
76934ed4 YL |
1089 | |
1090 | /* How many end-of-memory variables you have, grandma! */ | |
1091 | /* need this before calling reserve_initrd */ | |
f361a450 | 1092 | if (max_pfn > (1UL<<(32 - PAGE_SHIFT))) |
0c6fc11a | 1093 | max_low_pfn = e820__end_of_low_ram_pfn(); |
f361a450 YL |
1094 | else |
1095 | max_low_pfn = max_pfn; | |
1096 | ||
76934ed4 | 1097 | high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1; |
5394f80f JF |
1098 | #endif |
1099 | ||
893f38d1 YL |
1100 | /* |
1101 | * Find and reserve possible boot-time SMP configuration: | |
1102 | */ | |
1103 | find_smp_config(); | |
1104 | ||
042be38e YL |
1105 | reserve_ibft_region(); |
1106 | ||
8d57470d YL |
1107 | early_alloc_pgt_buf(); |
1108 | ||
72d7c3b3 | 1109 | /* |
4918e228 | 1110 | * Need to conclude brk, before e820__memblock_setup() |
72d7c3b3 YL |
1111 | * it could use memblock_find_in_range, could overlap with |
1112 | * brk area. | |
1113 | */ | |
1114 | reserve_brk(); | |
1115 | ||
e5f15b45 YL |
1116 | cleanup_highmap(); |
1117 | ||
2449f343 | 1118 | memblock_set_current_limit(ISA_END_ADDRESS); |
4918e228 | 1119 | e820__memblock_setup(); |
72d7c3b3 | 1120 | |
1b5aeebf LB |
1121 | if (!early_xdbc_setup_hardware()) |
1122 | early_xdbc_register_console(); | |
1123 | ||
4971531a MF |
1124 | reserve_bios_regions(); |
1125 | ||
1126 | if (efi_enabled(EFI_MEMMAP)) { | |
0f96a99d | 1127 | efi_fake_memmap(); |
b05b9f5f | 1128 | efi_find_mirror(); |
3dad6f7f | 1129 | efi_esrt_init(); |
007b7560 | 1130 | |
4971531a MF |
1131 | /* |
1132 | * The EFI specification says that boot service code won't be | |
1133 | * called after ExitBootServices(). This is, in fact, a lie. | |
1134 | */ | |
916f676f | 1135 | efi_reserve_boot_services(); |
4971531a | 1136 | } |
916f676f | 1137 | |
72d7c3b3 | 1138 | /* preallocate 4k for mptable mpc */ |
5da217ca | 1139 | e820__memblock_alloc_reserved_mpc_new(); |
72d7c3b3 YL |
1140 | |
1141 | #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION | |
1142 | setup_bios_corruption_check(); | |
1143 | #endif | |
1144 | ||
10054230 | 1145 | #ifdef CONFIG_X86_32 |
365811d6 BH |
1146 | printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n", |
1147 | (max_pfn_mapped<<PAGE_SHIFT) - 1); | |
10054230 | 1148 | #endif |
72d7c3b3 | 1149 | |
4f7b9226 | 1150 | reserve_real_mode(); |
893f38d1 | 1151 | |
a9acc536 | 1152 | trim_platform_memory_ranges(); |
95c96084 | 1153 | trim_low_memory_range(); |
a9acc536 | 1154 | |
22ddfcaa | 1155 | init_mem_mapping(); |
1bbbbe77 | 1156 | |
8170e6be | 1157 | early_trap_pf_init(); |
1bbbbe77 | 1158 | |
18bc7bd5 AL |
1159 | /* |
1160 | * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features) | |
1161 | * with the current CR4 value. This may not be necessary, but | |
1162 | * auditing all the early-boot CR4 manipulation would be needed to | |
1163 | * rule it out. | |
1164 | */ | |
1ef55be1 | 1165 | mmu_cr4_features = __read_cr4(); |
18bc7bd5 | 1166 | |
4ce7a869 | 1167 | memblock_set_current_limit(get_max_mapped()); |
4e29684c | 1168 | |
e7b37895 YL |
1169 | /* |
1170 | * NOTE: On x86-32, only from this point on, fixmaps are ready for use. | |
1171 | */ | |
1172 | ||
1173 | #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT | |
1174 | if (init_ohci1394_dma_early) | |
1175 | init_ohci1394_dma_on_all_controllers(); | |
1176 | #endif | |
162a7e75 MT |
1177 | /* Allocate bigger log buffer */ |
1178 | setup_log_buf(1); | |
e7b37895 | 1179 | |
9661b332 DH |
1180 | if (efi_enabled(EFI_BOOT)) { |
1181 | switch (boot_params.secure_boot) { | |
1182 | case efi_secureboot_mode_disabled: | |
1183 | pr_info("Secure boot disabled\n"); | |
1184 | break; | |
1185 | case efi_secureboot_mode_enabled: | |
1186 | pr_info("Secure boot enabled\n"); | |
1187 | break; | |
1188 | default: | |
1189 | pr_info("Secure boot could not be determined\n"); | |
1190 | break; | |
1191 | } | |
1192 | } | |
1193 | ||
2ec65f8b YL |
1194 | reserve_initrd(); |
1195 | ||
da3d3f98 | 1196 | acpi_table_upgrade(); |
53aac44c | 1197 | |
76934ed4 | 1198 | vsmp_init(); |
76934ed4 | 1199 | |
1c6e5503 YL |
1200 | io_delay_init(); |
1201 | ||
1c6e5503 YL |
1202 | /* |
1203 | * Parse the ACPI tables for possible boot-time SMP configuration. | |
1204 | */ | |
20e6926d YL |
1205 | acpi_boot_table_init(); |
1206 | ||
1207 | early_acpi_boot_init(); | |
1208 | ||
d8fc3afc | 1209 | initmem_init(); |
3c325f82 | 1210 | dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT); |
fa591c4a TC |
1211 | |
1212 | /* | |
1213 | * Reserve memory for crash kernel after SRAT is parsed so that it | |
1214 | * won't consume hotpluggable memory. | |
1215 | */ | |
1216 | reserve_crashkernel(); | |
1217 | ||
6f2a7536 | 1218 | memblock_find_dma_reserve(); |
91467bdf | 1219 | |
90993cdd | 1220 | #ifdef CONFIG_KVM_GUEST |
790c73f6 GOC |
1221 | kvmclock_init(); |
1222 | #endif | |
1223 | ||
7737b215 | 1224 | x86_init.paging.pagetable_init(); |
f212ec4b | 1225 | |
ef7f0d6a AR |
1226 | kasan_init(); |
1227 | ||
d2b6dc61 AL |
1228 | #ifdef CONFIG_X86_32 |
1229 | /* sync back kernel address range */ | |
1230 | clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY, | |
1231 | swapper_pg_dir + KERNEL_PGD_BOUNDARY, | |
1232 | KERNEL_PGD_PTRS); | |
1233 | ||
1234 | /* | |
1235 | * sync back low identity map too. It is used for example | |
1236 | * in the 32-bit EFI stub. | |
1237 | */ | |
1238 | clone_pgd_range(initial_page_table, | |
1239 | swapper_pg_dir + KERNEL_PGD_BOUNDARY, | |
1240 | min(KERNEL_PGD_PTRS, KERNEL_PGD_BOUNDARY)); | |
1241 | #endif | |
1242 | ||
31625340 JC |
1243 | tboot_probe(); |
1244 | ||
76934ed4 | 1245 | map_vsyscall(); |
76934ed4 | 1246 | |
1a3f239d | 1247 | generic_apic_probe(); |
1da177e4 | 1248 | |
54ef3400 | 1249 | early_quirks(); |
d44647b0 | 1250 | |
295deae4 YL |
1251 | /* |
1252 | * Read APIC and some other early information from ACPI tables. | |
1253 | */ | |
1da177e4 | 1254 | acpi_boot_init(); |
efafc8b2 | 1255 | sfi_init(); |
a906fdaa | 1256 | x86_dtb_init(); |
04606618 | 1257 | |
295deae4 YL |
1258 | /* |
1259 | * get boot-time SMP configuration: | |
1260 | */ | |
a91bf718 | 1261 | get_smp_config(); |
76934ed4 | 1262 | |
1e90a13d TG |
1263 | /* |
1264 | * Systems w/o ACPI and mptables might not have it mapped the local | |
1265 | * APIC yet, but prefill_possible_map() might need to access it. | |
1266 | */ | |
1267 | init_apic_mappings(); | |
1268 | ||
329513a3 | 1269 | prefill_possible_map(); |
301e6190 | 1270 | |
5f4765f9 | 1271 | init_cpu_to_node(); |
5f4765f9 | 1272 | |
ca1b8862 | 1273 | io_apic_init_mappings(); |
9d6a4d08 | 1274 | |
295deae4 | 1275 | kvm_guest_init(); |
1da177e4 | 1276 | |
1506c8dc | 1277 | e820__reserve_resources(); |
090d7171 | 1278 | e820__register_nosave_regions(max_low_pfn); |
1da177e4 | 1279 | |
8fee697d | 1280 | x86_init.resources.reserve_resources(); |
41c094fd | 1281 | |
2df908ba | 1282 | e820__setup_pci_gap(); |
41c094fd | 1283 | |
1da177e4 LT |
1284 | #ifdef CONFIG_VT |
1285 | #if defined(CONFIG_VGA_CONSOLE) | |
83e68189 | 1286 | if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY)) |
1da177e4 LT |
1287 | conswitchp = &vga_con; |
1288 | #elif defined(CONFIG_DUMMY_CONSOLE) | |
1289 | conswitchp = &dummy_con; | |
1290 | #endif | |
1291 | #endif | |
6f30c1ac | 1292 | x86_init.oem.banner(); |
a2202aa2 | 1293 | |
6b617e22 FT |
1294 | x86_init.timers.wallclock_init(); |
1295 | ||
a2202aa2 | 1296 | mcheck_init(); |
f49aa448 | 1297 | |
dc326fca | 1298 | arch_init_ideal_nops(); |
b3c869d3 JS |
1299 | |
1300 | register_refined_jiffies(CLOCK_TICK_RATE); | |
5189c2a7 OJ |
1301 | |
1302 | #ifdef CONFIG_EFI | |
a5d90c92 BP |
1303 | if (efi_enabled(EFI_BOOT)) |
1304 | efi_apply_memmap_quirks(); | |
5189c2a7 | 1305 | #endif |
1da177e4 | 1306 | } |
5649b7c3 | 1307 | |
9be1b56a IM |
1308 | #ifdef CONFIG_X86_32 |
1309 | ||
8fee697d TG |
1310 | static struct resource video_ram_resource = { |
1311 | .name = "Video RAM area", | |
1312 | .start = 0xa0000, | |
1313 | .end = 0xbffff, | |
1314 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
9be1b56a IM |
1315 | }; |
1316 | ||
8fee697d | 1317 | void __init i386_reserve_resources(void) |
9be1b56a | 1318 | { |
8fee697d TG |
1319 | request_resource(&iomem_resource, &video_ram_resource); |
1320 | reserve_standard_io_resources(); | |
9be1b56a IM |
1321 | } |
1322 | ||
9be1b56a | 1323 | #endif /* CONFIG_X86_32 */ |
f32360ef KC |
1324 | |
1325 | static struct notifier_block kernel_offset_notifier = { | |
1326 | .notifier_call = dump_kernel_offset | |
1327 | }; | |
1328 | ||
1329 | static int __init register_kernel_offset_dumper(void) | |
1330 | { | |
1331 | atomic_notifier_chain_register(&panic_notifier_list, | |
1332 | &kernel_offset_notifier); | |
1333 | return 0; | |
1334 | } | |
1335 | __initcall(register_kernel_offset_dumper); | |
c1192f84 DH |
1336 | |
1337 | void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma) | |
1338 | { | |
1339 | if (!boot_cpu_has(X86_FEATURE_OSPKE)) | |
1340 | return; | |
1341 | ||
1342 | seq_printf(m, "ProtectionKey: %8u\n", vma_pkey(vma)); | |
1343 | } |