]>
Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 | 2 | /* |
1da177e4 LT |
3 | * Copyright (C) 1995 Linus Torvalds |
4 | * | |
12609013 IM |
5 | * This file contains the setup_arch() code, which handles the architecture-dependent |
6 | * parts of early kernel initialization. | |
1da177e4 | 7 | */ |
ea7b4244 | 8 | #include <linux/acpi.h> |
1da177e4 | 9 | #include <linux/console.h> |
12609013 | 10 | #include <linux/crash_dump.h> |
0b1abd1f | 11 | #include <linux/dma-map-ops.h> |
12609013 | 12 | #include <linux/dmi.h> |
1da177e4 | 13 | #include <linux/efi.h> |
12609013 IM |
14 | #include <linux/init_ohci1394_dma.h> |
15 | #include <linux/initrd.h> | |
138fe4e0 | 16 | #include <linux/iscsi_ibft.h> |
12609013 | 17 | #include <linux/memblock.h> |
f39650de | 18 | #include <linux/panic_notifier.h> |
376ff035 | 19 | #include <linux/pci.h> |
12609013 | 20 | #include <linux/root_dev.h> |
cf11e85f | 21 | #include <linux/hugetlb.h> |
69575d38 | 22 | #include <linux/tboot.h> |
438296a5 | 23 | #include <linux/security.h> |
1b5aeebf | 24 | #include <linux/usb/xhci-dbgp.h> |
a945c834 | 25 | #include <linux/static_call.h> |
0a0f0d8b | 26 | #include <linux/swiotlb.h> |
1bc3b91a | 27 | |
12609013 IM |
28 | #include <uapi/linux/mount.h> |
29 | ||
30 | #include <xen/xen.h> | |
31 | ||
9635b47d | 32 | #include <asm/apic.h> |
13c01139 | 33 | #include <asm/numa.h> |
ce3fe6b2 | 34 | #include <asm/bios_ebda.h> |
cc9f7a0c | 35 | #include <asm/bugs.h> |
6e5385d4 | 36 | #include <asm/cpu.h> |
12609013 | 37 | #include <asm/efi.h> |
1d9b16d1 | 38 | #include <asm/gart.h> |
88b094fb | 39 | #include <asm/hypervisor.h> |
12609013 IM |
40 | #include <asm/io_apic.h> |
41 | #include <asm/kasan.h> | |
42 | #include <asm/kaslr.h> | |
a2202aa2 | 43 | #include <asm/mce.h> |
12609013 | 44 | #include <asm/mtrr.h> |
ca947b72 | 45 | #include <asm/realmode.h> |
12609013 IM |
46 | #include <asm/olpc_ofw.h> |
47 | #include <asm/pci-direct.h> | |
da6b737b | 48 | #include <asm/prom.h> |
12609013 | 49 | #include <asm/proto.h> |
9a90ed06 | 50 | #include <asm/thermal.h> |
ee9f8fce | 51 | #include <asm/unwind.h> |
12609013 | 52 | #include <asm/vsyscall.h> |
186525bd | 53 | #include <linux/vmalloc.h> |
46d671b5 | 54 | |
2b72394e | 55 | /* |
360db4ac IM |
56 | * max_low_pfn_mapped: highest directly mapped pfn < 4 GB |
57 | * max_pfn_mapped: highest directly mapped pfn > 4 GB | |
66520ebc | 58 | * |
09821ff1 | 59 | * The direct mapping only covers E820_TYPE_RAM regions, so the ranges and gaps are |
360db4ac | 60 | * represented by pfn_mapped[]. |
2b72394e PE |
61 | */ |
62 | unsigned long max_low_pfn_mapped; | |
63 | unsigned long max_pfn_mapped; | |
64 | ||
e808bae2 | 65 | #ifdef CONFIG_DMI |
796216a5 | 66 | RESERVE_BRK(dmi_alloc, 65536); |
e808bae2 | 67 | #endif |
796216a5 | 68 | |
c0b5842a | 69 | |
360db4ac IM |
70 | /* |
71 | * Range of the BSS area. The size of the BSS area is determined | |
81519f77 | 72 | * at link time, with RESERVE_BRK() facility reserving additional |
360db4ac IM |
73 | * chunks. |
74 | */ | |
360db4ac IM |
75 | unsigned long _brk_start = (unsigned long)__brk_base; |
76 | unsigned long _brk_end = (unsigned long)__brk_base; | |
93dbda7c | 77 | |
217b8ce8 | 78 | struct boot_params boot_params; |
217b8ce8 | 79 | |
4046d6e8 | 80 | /* |
360db4ac IM |
81 | * These are the four main kernel memory regions, we put them into |
82 | * the resource tree so that kdump tools and other debugging tools | |
83 | * recover it: | |
4046d6e8 | 84 | */ |
360db4ac | 85 | |
a3299754 KC |
86 | static struct resource rodata_resource = { |
87 | .name = "Kernel rodata", | |
88 | .start = 0, | |
89 | .end = 0, | |
90 | .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM | |
91 | }; | |
92 | ||
4046d6e8 LT |
93 | static struct resource data_resource = { |
94 | .name = "Kernel data", | |
95 | .start = 0, | |
96 | .end = 0, | |
97 | .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM | |
98 | }; | |
99 | ||
100 | static struct resource code_resource = { | |
101 | .name = "Kernel code", | |
102 | .start = 0, | |
103 | .end = 0, | |
104 | .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM | |
105 | }; | |
106 | ||
107 | static struct resource bss_resource = { | |
108 | .name = "Kernel bss", | |
109 | .start = 0, | |
110 | .end = 0, | |
111 | .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM | |
112 | }; | |
113 | ||
114 | ||
7dea23ec | 115 | #ifdef CONFIG_X86_32 |
360db4ac | 116 | /* CPU data as detected by the assembly code in head_32.S */ |
6415813b MK |
117 | struct cpuinfo_x86 new_cpu_data; |
118 | ||
360db4ac | 119 | /* Common CPU data for all CPUs */ |
6415813b | 120 | struct cpuinfo_x86 boot_cpu_data __read_mostly; |
129f6946 | 121 | EXPORT_SYMBOL(boot_cpu_data); |
1da177e4 | 122 | |
0c254e38 AS |
123 | unsigned int def_to_bigsmp; |
124 | ||
7dea23ec YL |
125 | struct apm_info apm_info; |
126 | EXPORT_SYMBOL(apm_info); | |
127 | ||
128 | #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \ | |
129 | defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE) | |
130 | struct ist_info ist_info; | |
131 | EXPORT_SYMBOL(ist_info); | |
132 | #else | |
133 | struct ist_info ist_info; | |
134 | #endif | |
135 | ||
136 | #else | |
162434e7 | 137 | struct cpuinfo_x86 boot_cpu_data __read_mostly; |
7dea23ec YL |
138 | EXPORT_SYMBOL(boot_cpu_data); |
139 | #endif | |
140 | ||
141 | ||
142 | #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64) | |
404f6aac | 143 | __visible unsigned long mmu_cr4_features __ro_after_init; |
7dea23ec | 144 | #else |
404f6aac | 145 | __visible unsigned long mmu_cr4_features __ro_after_init = X86_CR4_PAE; |
7dea23ec YL |
146 | #endif |
147 | ||
5031296c PA |
148 | /* Boot loader ID and version as integers, for the benefit of proc_dointvec */ |
149 | int bootloader_type, bootloader_version; | |
1da177e4 | 150 | |
1da177e4 LT |
151 | /* |
152 | * Setup options | |
153 | */ | |
1da177e4 | 154 | struct screen_info screen_info; |
129f6946 | 155 | EXPORT_SYMBOL(screen_info); |
1da177e4 | 156 | struct edid_info edid_info; |
5e518d76 | 157 | EXPORT_SYMBOL_GPL(edid_info); |
1da177e4 | 158 | |
1da177e4 LT |
159 | extern int root_mountflags; |
160 | ||
e44b7b75 | 161 | unsigned long saved_video_mode; |
1da177e4 | 162 | |
cf8fa920 | 163 | #define RAMDISK_IMAGE_START_MASK 0x07FF |
1da177e4 | 164 | #define RAMDISK_PROMPT_FLAG 0x8000 |
cf8fa920 | 165 | #define RAMDISK_LOAD_FLAG 0x4000 |
1da177e4 | 166 | |
4e498b66 | 167 | static char __initdata command_line[COMMAND_LINE_SIZE]; |
516cbf37 TB |
168 | #ifdef CONFIG_CMDLINE_BOOL |
169 | static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE; | |
170 | #endif | |
1da177e4 | 171 | |
1da177e4 LT |
172 | #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE) |
173 | struct edd edd; | |
174 | #ifdef CONFIG_EDD_MODULE | |
175 | EXPORT_SYMBOL(edd); | |
176 | #endif | |
177 | /** | |
178 | * copy_edd() - Copy the BIOS EDD information | |
179 | * from boot_params into a safe place. | |
180 | * | |
181 | */ | |
9eaa192d | 182 | static inline void __init copy_edd(void) |
1da177e4 | 183 | { |
30c82645 PA |
184 | memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer, |
185 | sizeof(edd.mbr_signature)); | |
186 | memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info)); | |
187 | edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries; | |
188 | edd.edd_info_nr = boot_params.eddbuf_entries; | |
1da177e4 LT |
189 | } |
190 | #else | |
9eaa192d | 191 | static inline void __init copy_edd(void) |
1da177e4 LT |
192 | { |
193 | } | |
194 | #endif | |
195 | ||
5368a2be PA |
196 | void * __init extend_brk(size_t size, size_t align) |
197 | { | |
198 | size_t mask = align - 1; | |
199 | void *ret; | |
200 | ||
201 | BUG_ON(_brk_start == 0); | |
202 | BUG_ON(align & mask); | |
203 | ||
204 | _brk_end = (_brk_end + mask) & ~mask; | |
205 | BUG_ON((char *)(_brk_end + size) > __brk_limit); | |
206 | ||
207 | ret = (void *)_brk_end; | |
208 | _brk_end += size; | |
209 | ||
210 | memset(ret, 0, size); | |
211 | ||
212 | return ret; | |
213 | } | |
214 | ||
148b2098 | 215 | #ifdef CONFIG_X86_32 |
e5f15b45 | 216 | static void __init cleanup_highmap(void) |
f005fe12 YL |
217 | { |
218 | } | |
854c879f PE |
219 | #endif |
220 | ||
5368a2be PA |
221 | static void __init reserve_brk(void) |
222 | { | |
223 | if (_brk_end > _brk_start) | |
fc8d7826 AD |
224 | memblock_reserve(__pa_symbol(_brk_start), |
225 | _brk_end - _brk_start); | |
5368a2be PA |
226 | |
227 | /* Mark brk area as locked down and no longer taking any | |
228 | new allocations */ | |
229 | _brk_start = 0; | |
230 | } | |
231 | ||
5aa3d718 BP |
232 | u64 relocated_ramdisk; |
233 | ||
cf8fa920 PA |
234 | #ifdef CONFIG_BLK_DEV_INITRD |
235 | ||
a8a51a88 YL |
236 | static u64 __init get_ramdisk_image(void) |
237 | { | |
238 | u64 ramdisk_image = boot_params.hdr.ramdisk_image; | |
239 | ||
ee92d815 YL |
240 | ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32; |
241 | ||
694cfd87 RM |
242 | if (ramdisk_image == 0) |
243 | ramdisk_image = phys_initrd_start; | |
244 | ||
a8a51a88 YL |
245 | return ramdisk_image; |
246 | } | |
247 | static u64 __init get_ramdisk_size(void) | |
248 | { | |
249 | u64 ramdisk_size = boot_params.hdr.ramdisk_size; | |
250 | ||
ee92d815 YL |
251 | ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32; |
252 | ||
694cfd87 RM |
253 | if (ramdisk_size == 0) |
254 | ramdisk_size = phys_initrd_size; | |
255 | ||
a8a51a88 YL |
256 | return ramdisk_size; |
257 | } | |
258 | ||
eb1379cb | 259 | static void __init relocate_initrd(void) |
cf8fa920 | 260 | { |
c967da6a | 261 | /* Assume only end is not page aligned */ |
a8a51a88 YL |
262 | u64 ramdisk_image = get_ramdisk_image(); |
263 | u64 ramdisk_size = get_ramdisk_size(); | |
c967da6a | 264 | u64 area_size = PAGE_ALIGN(ramdisk_size); |
cf8fa920 | 265 | |
e8c57d40 | 266 | /* We need to move the initrd down into directly mapped mem */ |
3c45ee6d MR |
267 | relocated_ramdisk = memblock_phys_alloc_range(area_size, PAGE_SIZE, 0, |
268 | PFN_PHYS(max_pfn_mapped)); | |
5aa3d718 | 269 | if (!relocated_ramdisk) |
3945e2c9 | 270 | panic("Cannot find place for new RAMDISK of size %lld\n", |
5aa3d718 | 271 | ramdisk_size); |
3945e2c9 | 272 | |
5aa3d718 | 273 | initrd_start = relocated_ramdisk + PAGE_OFFSET; |
cf8fa920 | 274 | initrd_end = initrd_start + ramdisk_size; |
365811d6 | 275 | printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n", |
5aa3d718 | 276 | relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1); |
cf8fa920 | 277 | |
5dd2c4bd MS |
278 | copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size); |
279 | ||
365811d6 BH |
280 | printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to" |
281 | " [mem %#010llx-%#010llx]\n", | |
ba5b14cc | 282 | ramdisk_image, ramdisk_image + ramdisk_size - 1, |
5aa3d718 | 283 | relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1); |
eb1379cb | 284 | } |
9a27f5c5 | 285 | |
1b8c78be YL |
286 | static void __init early_reserve_initrd(void) |
287 | { | |
288 | /* Assume only end is not page aligned */ | |
a8a51a88 YL |
289 | u64 ramdisk_image = get_ramdisk_image(); |
290 | u64 ramdisk_size = get_ramdisk_size(); | |
1b8c78be YL |
291 | u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size); |
292 | ||
293 | if (!boot_params.hdr.type_of_loader || | |
294 | !ramdisk_image || !ramdisk_size) | |
295 | return; /* No initrd provided by bootloader */ | |
296 | ||
297 | memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image); | |
298 | } | |
3c45ee6d | 299 | |
eb1379cb YL |
300 | static void __init reserve_initrd(void) |
301 | { | |
c967da6a | 302 | /* Assume only end is not page aligned */ |
a8a51a88 YL |
303 | u64 ramdisk_image = get_ramdisk_image(); |
304 | u64 ramdisk_size = get_ramdisk_size(); | |
c967da6a | 305 | u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size); |
eb1379cb YL |
306 | |
307 | if (!boot_params.hdr.type_of_loader || | |
308 | !ramdisk_image || !ramdisk_size) | |
309 | return; /* No initrd provided by bootloader */ | |
310 | ||
311 | initrd_start = 0; | |
312 | ||
365811d6 BH |
313 | printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image, |
314 | ramdisk_end - 1); | |
eb1379cb | 315 | |
74f27655 | 316 | if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image), |
e8c57d40 YL |
317 | PFN_DOWN(ramdisk_end))) { |
318 | /* All are mapped, easy case */ | |
eb1379cb YL |
319 | initrd_start = ramdisk_image + PAGE_OFFSET; |
320 | initrd_end = initrd_start + ramdisk_size; | |
321 | return; | |
322 | } | |
323 | ||
eb1379cb | 324 | relocate_initrd(); |
8c5dd8f4 | 325 | |
24aa0788 | 326 | memblock_free(ramdisk_image, ramdisk_end - ramdisk_image); |
cf8fa920 | 327 | } |
af06f8b7 | 328 | |
225c37d7 | 329 | #else |
1b8c78be YL |
330 | static void __init early_reserve_initrd(void) |
331 | { | |
332 | } | |
eb1379cb | 333 | static void __init reserve_initrd(void) |
225c37d7 YL |
334 | { |
335 | } | |
cf8fa920 PA |
336 | #endif /* CONFIG_BLK_DEV_INITRD */ |
337 | ||
29f784e3 | 338 | static void __init parse_setup_data(void) |
257b0fde YL |
339 | { |
340 | struct setup_data *data; | |
30e46b57 | 341 | u64 pa_data, pa_next; |
257b0fde | 342 | |
257b0fde YL |
343 | pa_data = boot_params.hdr.setup_data; |
344 | while (pa_data) { | |
7389882c | 345 | u32 data_len, data_type; |
f1c2b357 | 346 | |
7389882c | 347 | data = early_memremap(pa_data, sizeof(*data)); |
f1c2b357 | 348 | data_len = data->len + sizeof(struct setup_data); |
30e46b57 LC |
349 | data_type = data->type; |
350 | pa_next = data->next; | |
8d4a40bc | 351 | early_memunmap(data, sizeof(*data)); |
f1c2b357 | 352 | |
30e46b57 | 353 | switch (data_type) { |
257b0fde | 354 | case SETUP_E820_EXT: |
914053c0 | 355 | e820__memory_setup_extended(pa_data, data_len); |
257b0fde | 356 | break; |
da6b737b SAS |
357 | case SETUP_DTB: |
358 | add_dtb(pa_data); | |
257b0fde | 359 | break; |
1fec0533 DY |
360 | case SETUP_EFI: |
361 | parse_efi_setup(pa_data, data_len); | |
362 | break; | |
257b0fde YL |
363 | default: |
364 | break; | |
365 | } | |
30e46b57 | 366 | pa_data = pa_next; |
257b0fde YL |
367 | } |
368 | } | |
369 | ||
a9ce6bc1 | 370 | static void __init memblock_x86_reserve_range_setup_data(void) |
a0a0becd YL |
371 | { |
372 | struct setup_data *data; | |
373 | u64 pa_data; | |
a0a0becd | 374 | |
a0a0becd YL |
375 | pa_data = boot_params.hdr.setup_data; |
376 | while (pa_data) { | |
88b4c146 | 377 | data = early_memremap(pa_data, sizeof(*data)); |
24aa0788 | 378 | memblock_reserve(pa_data, sizeof(*data) + data->len); |
b3c72fc9 DK |
379 | |
380 | if (data->type == SETUP_INDIRECT && | |
381 | ((struct setup_indirect *)data->data)->type != SETUP_INDIRECT) | |
382 | memblock_reserve(((struct setup_indirect *)data->data)->addr, | |
383 | ((struct setup_indirect *)data->data)->len); | |
384 | ||
a0a0becd | 385 | pa_data = data->next; |
8d4a40bc | 386 | early_memunmap(data, sizeof(*data)); |
a0a0becd YL |
387 | } |
388 | } | |
389 | ||
ccb4defa YL |
390 | /* |
391 | * --------- Crashkernel reservation ------------------------------ | |
392 | */ | |
393 | ||
2965faa5 | 394 | #ifdef CONFIG_KEXEC_CORE |
32105f7f | 395 | |
606134f7 | 396 | /* 16M alignment for crash kernel regions */ |
9ca5c8e6 | 397 | #define CRASH_ALIGN SZ_16M |
606134f7 | 398 | |
7f8595bf | 399 | /* |
8ff80fbe BH |
400 | * Keep the crash kernel below this limit. |
401 | * | |
360db4ac | 402 | * Earlier 32-bits kernels would limit the kernel to the low 512 MB range |
8ff80fbe BH |
403 | * due to mapping restrictions. |
404 | * | |
360db4ac | 405 | * 64-bit kdump kernels need to be restricted to be under 64 TB, which is |
11a98f37 | 406 | * the upper limit of system RAM in 4-level paging mode. Since the kdump |
360db4ac IM |
407 | * jump could be from 5-level paging to 4-level paging, the jump will fail if |
408 | * the kernel is put above 64 TB, and during the 1st kernel bootup there's | |
409 | * no good way to detect the paging mode of the target kernel which will be | |
410 | * loaded for dumping. | |
7f8595bf PA |
411 | */ |
412 | #ifdef CONFIG_X86_32 | |
9ca5c8e6 DY |
413 | # define CRASH_ADDR_LOW_MAX SZ_512M |
414 | # define CRASH_ADDR_HIGH_MAX SZ_512M | |
7f8595bf | 415 | #else |
9ca5c8e6 | 416 | # define CRASH_ADDR_LOW_MAX SZ_4G |
8ff80fbe | 417 | # define CRASH_ADDR_HIGH_MAX SZ_64T |
7f8595bf PA |
418 | #endif |
419 | ||
eb6db83d | 420 | static int __init reserve_crashkernel_low(void) |
0212f915 YL |
421 | { |
422 | #ifdef CONFIG_X86_64 | |
f56d5578 | 423 | unsigned long long base, low_base = 0, low_size = 0; |
6120cdc0 | 424 | unsigned long low_mem_limit; |
0212f915 YL |
425 | int ret; |
426 | ||
6120cdc0 | 427 | low_mem_limit = min(memblock_phys_mem_size(), CRASH_ADDR_LOW_MAX); |
97eac21b | 428 | |
adbc742b | 429 | /* crashkernel=Y,low */ |
6120cdc0 | 430 | ret = parse_crashkernel_low(boot_command_line, low_mem_limit, &low_size, &base); |
f56d5578 | 431 | if (ret) { |
c729de8f | 432 | /* |
392e879a | 433 | * two parts from kernel/dma/swiotlb.c: |
94fb9334 JR |
434 | * -swiotlb size: user-specified with swiotlb= or default. |
435 | * | |
436 | * -swiotlb overflow buffer: now hardcoded to 32k. We round it | |
437 | * to 8M for other buffers that may need to stay low too. Also | |
438 | * make sure we allocate enough extra low memory so that we | |
439 | * don't run out of DMA buffers for 32-bit devices. | |
c729de8f | 440 | */ |
97eac21b | 441 | low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20); |
c729de8f | 442 | } else { |
adbc742b | 443 | /* passed with crashkernel=0,low ? */ |
c729de8f | 444 | if (!low_size) |
eb6db83d | 445 | return 0; |
c729de8f | 446 | } |
0212f915 | 447 | |
6120cdc0 | 448 | low_base = memblock_phys_alloc_range(low_size, CRASH_ALIGN, 0, CRASH_ADDR_LOW_MAX); |
0212f915 | 449 | if (!low_base) { |
eb6db83d BH |
450 | pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n", |
451 | (unsigned long)(low_size >> 20)); | |
452 | return -ENOMEM; | |
0212f915 | 453 | } |
0212f915 | 454 | |
6120cdc0 | 455 | pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (low RAM limit: %ldMB)\n", |
97eac21b BP |
456 | (unsigned long)(low_size >> 20), |
457 | (unsigned long)(low_base >> 20), | |
6120cdc0 | 458 | (unsigned long)(low_mem_limit >> 20)); |
97eac21b | 459 | |
0212f915 YL |
460 | crashk_low_res.start = low_base; |
461 | crashk_low_res.end = low_base + low_size - 1; | |
462 | insert_resource(&iomem_resource, &crashk_low_res); | |
7f8595bf | 463 | #endif |
eb6db83d | 464 | return 0; |
0212f915 | 465 | } |
7f8595bf | 466 | |
29f784e3 | 467 | static void __init reserve_crashkernel(void) |
ccb4defa | 468 | { |
f56d5578 | 469 | unsigned long long crash_size, crash_base, total_mem; |
55a20ee7 | 470 | bool high = false; |
ccb4defa YL |
471 | int ret; |
472 | ||
09c71bfd | 473 | total_mem = memblock_phys_mem_size(); |
ccb4defa | 474 | |
55a20ee7 | 475 | /* crashkernel=XM */ |
97eac21b | 476 | ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base); |
55a20ee7 | 477 | if (ret != 0 || crash_size <= 0) { |
adbc742b | 478 | /* crashkernel=X,high */ |
55a20ee7 | 479 | ret = parse_crashkernel_high(boot_command_line, total_mem, |
97eac21b | 480 | &crash_size, &crash_base); |
55a20ee7 YL |
481 | if (ret != 0 || crash_size <= 0) |
482 | return; | |
483 | high = true; | |
484 | } | |
32105f7f | 485 | |
3db3eb28 PT |
486 | if (xen_pv_domain()) { |
487 | pr_info("Ignoring crashkernel for a Xen PV domain\n"); | |
488 | return; | |
489 | } | |
490 | ||
32105f7f | 491 | /* 0 means: find the address automatically */ |
b9ac3849 | 492 | if (!crash_base) { |
9f4c1396 | 493 | /* |
a8d4c824 | 494 | * Set CRASH_ADDR_LOW_MAX upper bound for crash memory, |
b9ac3849 DY |
495 | * crashkernel=x,high reserves memory over 4G, also allocates |
496 | * 256M extra low memory for DMA buffers and swiotlb. | |
497 | * But the extra memory is not required for all machines. | |
498 | * So try low memory first and fall back to high memory | |
499 | * unless "crashkernel=size[KMG],high" is specified. | |
9f4c1396 | 500 | */ |
b9ac3849 | 501 | if (!high) |
6120cdc0 MR |
502 | crash_base = memblock_phys_alloc_range(crash_size, |
503 | CRASH_ALIGN, CRASH_ALIGN, | |
504 | CRASH_ADDR_LOW_MAX); | |
b9ac3849 | 505 | if (!crash_base) |
6120cdc0 MR |
506 | crash_base = memblock_phys_alloc_range(crash_size, |
507 | CRASH_ALIGN, CRASH_ALIGN, | |
508 | CRASH_ADDR_HIGH_MAX); | |
1f5026a7 | 509 | if (!crash_base) { |
44280733 | 510 | pr_info("crashkernel reservation failed - No suitable area found.\n"); |
ccb4defa YL |
511 | return; |
512 | } | |
32105f7f | 513 | } else { |
44280733 YL |
514 | unsigned long long start; |
515 | ||
6120cdc0 MR |
516 | start = memblock_phys_alloc_range(crash_size, SZ_1M, crash_base, |
517 | crash_base + crash_size); | |
44280733 YL |
518 | if (start != crash_base) { |
519 | pr_info("crashkernel reservation failed - memory is in use.\n"); | |
ccb4defa YL |
520 | return; |
521 | } | |
32105f7f | 522 | } |
ccb4defa | 523 | |
eb6db83d BH |
524 | if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) { |
525 | memblock_free(crash_base, crash_size); | |
526 | return; | |
527 | } | |
ccb4defa | 528 | |
f56d5578 BP |
529 | pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n", |
530 | (unsigned long)(crash_size >> 20), | |
531 | (unsigned long)(crash_base >> 20), | |
532 | (unsigned long)(total_mem >> 20)); | |
ccb4defa | 533 | |
32105f7f BW |
534 | crashk_res.start = crash_base; |
535 | crashk_res.end = crash_base + crash_size - 1; | |
536 | insert_resource(&iomem_resource, &crashk_res); | |
ccb4defa YL |
537 | } |
538 | #else | |
29f784e3 | 539 | static void __init reserve_crashkernel(void) |
ccb4defa YL |
540 | { |
541 | } | |
542 | #endif | |
543 | ||
bdba0e70 YL |
544 | static struct resource standard_io_resources[] = { |
545 | { .name = "dma1", .start = 0x00, .end = 0x1f, | |
546 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
547 | { .name = "pic1", .start = 0x20, .end = 0x21, | |
548 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
549 | { .name = "timer0", .start = 0x40, .end = 0x43, | |
550 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
551 | { .name = "timer1", .start = 0x50, .end = 0x53, | |
552 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
553 | { .name = "keyboard", .start = 0x60, .end = 0x60, | |
554 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
555 | { .name = "keyboard", .start = 0x64, .end = 0x64, | |
556 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
557 | { .name = "dma page reg", .start = 0x80, .end = 0x8f, | |
558 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
559 | { .name = "pic2", .start = 0xa0, .end = 0xa1, | |
560 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
561 | { .name = "dma2", .start = 0xc0, .end = 0xdf, | |
562 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
563 | { .name = "fpu", .start = 0xf0, .end = 0xff, | |
564 | .flags = IORESOURCE_BUSY | IORESOURCE_IO } | |
565 | }; | |
566 | ||
8fee697d | 567 | void __init reserve_standard_io_resources(void) |
bdba0e70 YL |
568 | { |
569 | int i; | |
570 | ||
571 | /* request I/O space for devices used on all i[345]86 PCs */ | |
572 | for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++) | |
573 | request_resource(&ioport_resource, &standard_io_resources[i]); | |
574 | ||
575 | } | |
576 | ||
a9acc536 JB |
577 | static bool __init snb_gfx_workaround_needed(void) |
578 | { | |
e43b3cec | 579 | #ifdef CONFIG_PCI |
a9acc536 JB |
580 | int i; |
581 | u16 vendor, devid; | |
ab3cd867 | 582 | static const __initconst u16 snb_ids[] = { |
a9acc536 JB |
583 | 0x0102, |
584 | 0x0112, | |
585 | 0x0122, | |
586 | 0x0106, | |
587 | 0x0116, | |
588 | 0x0126, | |
589 | 0x010a, | |
590 | }; | |
591 | ||
592 | /* Assume no if something weird is going on with PCI */ | |
593 | if (!early_pci_allowed()) | |
594 | return false; | |
595 | ||
596 | vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID); | |
597 | if (vendor != 0x8086) | |
598 | return false; | |
599 | ||
600 | devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID); | |
601 | for (i = 0; i < ARRAY_SIZE(snb_ids); i++) | |
602 | if (devid == snb_ids[i]) | |
603 | return true; | |
e43b3cec | 604 | #endif |
a9acc536 JB |
605 | |
606 | return false; | |
607 | } | |
608 | ||
609 | /* | |
610 | * Sandy Bridge graphics has trouble with certain ranges, exclude | |
611 | * them from allocation. | |
612 | */ | |
613 | static void __init trim_snb_memory(void) | |
614 | { | |
ab3cd867 | 615 | static const __initconst unsigned long bad_pages[] = { |
a9acc536 JB |
616 | 0x20050000, |
617 | 0x20110000, | |
618 | 0x20130000, | |
619 | 0x20138000, | |
620 | 0x40004000, | |
621 | }; | |
622 | int i; | |
623 | ||
624 | if (!snb_gfx_workaround_needed()) | |
625 | return; | |
626 | ||
627 | printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n"); | |
628 | ||
629 | /* | |
c361e5d4 MR |
630 | * SandyBridge integrated graphics devices have a bug that prevents |
631 | * them from accessing certain memory ranges, namely anything below | |
632 | * 1M and in the pages listed in bad_pages[] above. | |
633 | * | |
f1d4d47c MR |
634 | * To avoid these pages being ever accessed by SNB gfx devices reserve |
635 | * bad_pages that have not already been reserved at boot time. | |
636 | * All memory below the 1 MB mark is anyway reserved later during | |
637 | * setup_arch(), so there is no need to reserve it here. | |
a9acc536 | 638 | */ |
c361e5d4 | 639 | |
a9acc536 JB |
640 | for (i = 0; i < ARRAY_SIZE(bad_pages); i++) { |
641 | if (memblock_reserve(bad_pages[i], PAGE_SIZE)) | |
642 | printk(KERN_WARNING "failed to reserve 0x%08lx\n", | |
643 | bad_pages[i]); | |
644 | } | |
645 | } | |
646 | ||
1b5576e6 YL |
647 | static void __init trim_bios_range(void) |
648 | { | |
5c279c4c MR |
649 | /* |
650 | * A special case is the first 4Kb of memory; | |
651 | * This is a BIOS owned area, not kernel ram, but generally | |
652 | * not listed as such in the E820 table. | |
653 | * | |
654 | * This typically reserves additional memory (64KiB by default) | |
655 | * since some BIOSes are known to corrupt low memory. See the | |
656 | * Kconfig help text for X86_RESERVE_LOW. | |
657 | */ | |
658 | e820__range_update(0, PAGE_SIZE, E820_TYPE_RAM, E820_TYPE_RESERVED); | |
659 | ||
1b5576e6 | 660 | /* |
11a98f37 C |
661 | * special case: Some BIOSes report the PC BIOS |
662 | * area (640Kb -> 1Mb) as RAM even though it is not. | |
1b5576e6 YL |
663 | * take them out. |
664 | */ | |
09821ff1 | 665 | e820__range_remove(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_TYPE_RAM, 1); |
a9acc536 | 666 | |
f9748fa0 | 667 | e820__update_table(e820_table); |
1b5576e6 YL |
668 | } |
669 | ||
b422a309 YL |
670 | /* called before trim_bios_range() to spare extra sanitize */ |
671 | static void __init e820_add_kernel_range(void) | |
672 | { | |
673 | u64 start = __pa_symbol(_text); | |
674 | u64 size = __pa_symbol(_end) - start; | |
675 | ||
676 | /* | |
09821ff1 | 677 | * Complain if .text .data and .bss are not marked as E820_TYPE_RAM and |
b422a309 YL |
678 | * attempt to fix it by adding the range. We may have a confused BIOS, |
679 | * or the user may have used memmap=exactmap or memmap=xxM$yyM to | |
680 | * exclude kernel range. If we really are running on top non-RAM, | |
681 | * we will crash later anyways. | |
682 | */ | |
09821ff1 | 683 | if (e820__mapped_all(start, start + size, E820_TYPE_RAM)) |
b422a309 YL |
684 | return; |
685 | ||
09821ff1 IM |
686 | pr_warn(".text .data .bss are not marked as E820_TYPE_RAM!\n"); |
687 | e820__range_remove(start, size, E820_TYPE_RAM, 0); | |
688 | e820__range_add(start, size, E820_TYPE_RAM); | |
b422a309 YL |
689 | } |
690 | ||
a799c2bd | 691 | static void __init early_reserve_memory(void) |
95c96084 | 692 | { |
a799c2bd MR |
693 | /* |
694 | * Reserve the memory occupied by the kernel between _text and | |
695 | * __end_of_kernel_reserve symbols. Any kernel sections after the | |
696 | * __end_of_kernel_reserve symbol must be explicitly reserved with a | |
697 | * separate memblock_reserve() or they will be discarded. | |
698 | */ | |
699 | memblock_reserve(__pa_symbol(_text), | |
700 | (unsigned long)__end_of_kernel_reserve - (unsigned long)_text); | |
701 | ||
702 | /* | |
4c674481 MR |
703 | * The first 4Kb of memory is a BIOS owned area, but generally it is |
704 | * not listed as such in the E820 table. | |
705 | * | |
f1d4d47c MR |
706 | * Reserve the first 64K of memory since some BIOSes are known to |
707 | * corrupt low memory. After the real mode trampoline is allocated the | |
708 | * rest of the memory below 640k is reserved. | |
4c674481 MR |
709 | * |
710 | * In addition, make sure page 0 is always reserved because on | |
711 | * systems with L1TF its contents can be leaked to user processes. | |
a799c2bd | 712 | */ |
f1d4d47c | 713 | memblock_reserve(0, SZ_64K); |
a799c2bd MR |
714 | |
715 | early_reserve_initrd(); | |
716 | ||
717 | if (efi_enabled(EFI_BOOT)) | |
718 | efi_memblock_x86_reserve_range(); | |
719 | ||
720 | memblock_x86_reserve_range_setup_data(); | |
721 | ||
722 | reserve_ibft_region(); | |
723 | reserve_bios_regions(); | |
f1d4d47c | 724 | trim_snb_memory(); |
95c96084 | 725 | } |
a799c2bd | 726 | |
f32360ef KC |
727 | /* |
728 | * Dump out kernel offset information on panic. | |
729 | */ | |
730 | static int | |
731 | dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p) | |
732 | { | |
78cac48c BP |
733 | if (kaslr_enabled()) { |
734 | pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n", | |
4545c898 | 735 | kaslr_offset(), |
78cac48c BP |
736 | __START_KERNEL, |
737 | __START_KERNEL_map, | |
738 | MODULES_VADDR-1); | |
739 | } else { | |
740 | pr_emerg("Kernel Offset: disabled\n"); | |
741 | } | |
f32360ef KC |
742 | |
743 | return 0; | |
744 | } | |
745 | ||
52bcb8d7 BP |
746 | static char *prepare_command_line(void) |
747 | { | |
748 | #ifdef CONFIG_CMDLINE_BOOL | |
749 | #ifdef CONFIG_CMDLINE_OVERRIDE | |
750 | strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE); | |
751 | #else | |
752 | if (builtin_cmdline[0]) { | |
753 | /* append boot loader cmdline to builtin */ | |
754 | strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE); | |
755 | strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE); | |
756 | strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE); | |
757 | } | |
758 | #endif | |
759 | #endif | |
760 | ||
761 | strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE); | |
762 | ||
763 | parse_early_param(); | |
764 | ||
765 | return command_line; | |
766 | } | |
767 | ||
1da177e4 LT |
768 | /* |
769 | * Determine if we were loaded by an EFI loader. If so, then we have also been | |
770 | * passed the efi memmap, systab, etc., so we should use these data structures | |
771 | * for initialization. Note, the efi init code path is determined by the | |
772 | * global efi_enabled. This allows the same kernel image to be used on existing | |
773 | * systems (with a traditional BIOS) as well as on EFI systems. | |
774 | */ | |
76934ed4 YL |
775 | /* |
776 | * setup_arch - architecture-specific boot-time initializations | |
777 | * | |
778 | * Note: On x86_64, fixmaps are ready for use even before this is called. | |
779 | */ | |
780 | ||
1da177e4 LT |
781 | void __init setup_arch(char **cmdline_p) |
782 | { | |
76934ed4 | 783 | #ifdef CONFIG_X86_32 |
1da177e4 | 784 | memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data)); |
b40827fa BP |
785 | |
786 | /* | |
787 | * copy kernel address range established so far and switch | |
788 | * to the proper swapper page table | |
789 | */ | |
790 | clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY, | |
791 | initial_page_table + KERNEL_PGD_BOUNDARY, | |
792 | KERNEL_PGD_PTRS); | |
793 | ||
794 | load_cr3(swapper_pg_dir); | |
2075244f BD |
795 | /* |
796 | * Note: Quark X1000 CPUs advertise PGE incorrectly and require | |
797 | * a cr3 based tlb flush, so the following __flush_tlb_all() | |
360db4ac | 798 | * will not flush anything because the CPU quirk which clears |
2075244f BD |
799 | * X86_FEATURE_PGE has not been invoked yet. Though due to the |
800 | * load_cr3() above the TLB has been flushed already. The | |
801 | * quirk is invoked before subsequent calls to __flush_tlb_all() | |
802 | * so proper operation is guaranteed. | |
803 | */ | |
b40827fa | 804 | __flush_tlb_all(); |
76934ed4 YL |
805 | #else |
806 | printk(KERN_INFO "Command line: %s\n", boot_command_line); | |
162434e7 | 807 | boot_cpu_data.x86_phys_bits = MAX_PHYSMEM_BITS; |
76934ed4 | 808 | #endif |
1da177e4 | 809 | |
9863c90f AK |
810 | /* |
811 | * If we have OLPC OFW, we might end up relocating the fixmap due to | |
812 | * reserve_top(), so do this before touching the ioremap area. | |
813 | */ | |
fd699c76 AS |
814 | olpc_ofw_detect(); |
815 | ||
433f8924 | 816 | idt_setup_early_traps(); |
9e882c92 | 817 | early_cpu_init(); |
8990cac6 | 818 | jump_label_init(); |
a945c834 | 819 | static_call_init(); |
1a98fd14 JF |
820 | early_ioremap_init(); |
821 | ||
fd699c76 AS |
822 | setup_olpc_ofw_pgd(); |
823 | ||
30c82645 PA |
824 | ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev); |
825 | screen_info = boot_params.screen_info; | |
826 | edid_info = boot_params.edid_info; | |
76934ed4 | 827 | #ifdef CONFIG_X86_32 |
30c82645 PA |
828 | apm_info.bios = boot_params.apm_bios_info; |
829 | ist_info = boot_params.ist_info; | |
76934ed4 YL |
830 | #endif |
831 | saved_video_mode = boot_params.hdr.vid_mode; | |
30c82645 | 832 | bootloader_type = boot_params.hdr.type_of_loader; |
5031296c PA |
833 | if ((bootloader_type >> 4) == 0xe) { |
834 | bootloader_type &= 0xf; | |
835 | bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4; | |
836 | } | |
837 | bootloader_version = bootloader_type & 0xf; | |
838 | bootloader_version |= boot_params.hdr.ext_loader_ver << 4; | |
1da177e4 LT |
839 | |
840 | #ifdef CONFIG_BLK_DEV_RAM | |
30c82645 | 841 | rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK; |
1da177e4 | 842 | #endif |
7465252e YL |
843 | #ifdef CONFIG_EFI |
844 | if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature, | |
9402973d | 845 | EFI32_LOADER_SIGNATURE, 4)) { |
3e909599 | 846 | set_bit(EFI_BOOT, &efi.flags); |
1adbfa35 | 847 | } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature, |
9402973d | 848 | EFI64_LOADER_SIGNATURE, 4)) { |
3e909599 MF |
849 | set_bit(EFI_BOOT, &efi.flags); |
850 | set_bit(EFI_64BIT, &efi.flags); | |
7465252e YL |
851 | } |
852 | #endif | |
853 | ||
42bbdb43 | 854 | x86_init.oem.arch_setup(); |
2215e69d | 855 | |
52bcb8d7 BP |
856 | /* |
857 | * x86_configure_nx() is called before parse_early_param() (called by | |
858 | * prepare_command_line()) to detect whether hardware doesn't support | |
859 | * NX (so that the early EHCI debug console setup can safely call | |
860 | * set_fixmap()). It may then be called again from within noexec_setup() | |
861 | * during parsing early parameters to honor the respective command line | |
862 | * option. | |
863 | */ | |
864 | x86_configure_nx(); | |
865 | ||
866 | /* | |
867 | * This parses early params and it needs to run before | |
868 | * early_reserve_memory() because latter relies on such settings | |
869 | * supplied as early params. | |
870 | */ | |
871 | *cmdline_p = prepare_command_line(); | |
872 | ||
8aa83e63 JG |
873 | /* |
874 | * Do some memory reservations *before* memory is added to memblock, so | |
875 | * memblock allocations won't overwrite it. | |
876 | * | |
877 | * After this point, everything still needed from the boot loader or | |
878 | * firmware or kernel text should be early reserved or marked not RAM in | |
879 | * e820. All other memory is free game. | |
880 | * | |
881 | * This call needs to happen before e820__memory_setup() which calls the | |
882 | * xen_memory_setup() on Xen dom0 which relies on the fact that those | |
883 | * early reservations have happened already. | |
884 | */ | |
885 | early_reserve_memory(); | |
886 | ||
419afdf5 | 887 | iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1; |
103e2063 | 888 | e820__memory_setup(); |
28bb2237 YL |
889 | parse_setup_data(); |
890 | ||
1da177e4 LT |
891 | copy_edd(); |
892 | ||
30c82645 | 893 | if (!boot_params.hdr.root_flags) |
1da177e4 | 894 | root_mountflags &= ~MS_RDONLY; |
30120d72 | 895 | setup_initial_init_mm(_text, _etext, _edata, (void *)_brk_end); |
fe3d197f | 896 | |
4046d6e8 LT |
897 | code_resource.start = __pa_symbol(_text); |
898 | code_resource.end = __pa_symbol(_etext)-1; | |
a3299754 KC |
899 | rodata_resource.start = __pa_symbol(__start_rodata); |
900 | rodata_resource.end = __pa_symbol(__end_rodata)-1; | |
901 | data_resource.start = __pa_symbol(_sdata); | |
4046d6e8 LT |
902 | data_resource.end = __pa_symbol(_edata)-1; |
903 | bss_resource.start = __pa_symbol(__bss_start); | |
904 | bss_resource.end = __pa_symbol(__bss_stop)-1; | |
905 | ||
39fa104d RA |
906 | #ifdef CONFIG_MEMORY_HOTPLUG |
907 | /* | |
908 | * Memory used by the kernel cannot be hot-removed because Linux | |
909 | * cannot migrate the kernel pages. When memory hotplug is | |
910 | * enabled, we should prevent memblock from allocating memory | |
911 | * for the kernel. | |
912 | * | |
913 | * ACPI SRAT records all hotpluggable memory ranges. But before | |
914 | * SRAT is parsed, we don't know about it. | |
915 | * | |
916 | * The kernel image is loaded into memory at very early time. We | |
917 | * cannot prevent this anyway. So on NUMA system, we set any | |
918 | * node the kernel resides in as un-hotpluggable. | |
919 | * | |
920 | * Since on modern servers, one node could have double-digit | |
921 | * gigabytes memory, we can assume the memory around the kernel | |
922 | * image is also un-hotpluggable. So before SRAT is parsed, just | |
923 | * allocate memory near the kernel image to try the best to keep | |
924 | * the kernel away from hotpluggable memory. | |
925 | */ | |
926 | if (movable_node_is_enabled()) | |
927 | memblock_set_bottom_up(true); | |
928 | #endif | |
929 | ||
4b0f3b81 | 930 | x86_report_nx(); |
0ad5bce7 | 931 | |
76934ed4 | 932 | if (acpi_mps_check()) { |
3eb11edc | 933 | #ifdef CONFIG_X86_LOCAL_APIC |
76934ed4 | 934 | disable_apic = 1; |
3eb11edc | 935 | #endif |
988781dc | 936 | setup_clear_cpu_cap(X86_FEATURE_APIC); |
3c999f14 YL |
937 | } |
938 | ||
1a127034 | 939 | e820__reserve_setup_data(); |
9641bdaf | 940 | e820__finish_early_params(); |
1a3f239d | 941 | |
83e68189 | 942 | if (efi_enabled(EFI_BOOT)) |
ff0c0874 BM |
943 | efi_init(); |
944 | ||
438296a5 DH |
945 | efi_set_secure_boot(boot_params.secure_boot); |
946 | ||
01f96e4a | 947 | #ifdef CONFIG_LOCK_DOWN_IN_SECURE_BOOT |
438296a5 DH |
948 | if (efi_enabled(EFI_SECURE_BOOT)) |
949 | security_lock_kernel_down("EFI Secure Boot mode", LOCKDOWN_INTEGRITY_MAX); | |
950 | #endif | |
951 | ||
0fca0812 | 952 | dmi_setup(); |
2216d199 | 953 | |
88b094fb AK |
954 | /* |
955 | * VMware detection requires dmi to be available, so this | |
0fca0812 | 956 | * needs to be done after dmi_setup(), for the boot CPU. |
88b094fb | 957 | */ |
2d826404 | 958 | init_hypervisor_platform(); |
88b094fb | 959 | |
cf7a63ef | 960 | tsc_early_init(); |
f7cf5a5b | 961 | x86_init.resources.probe_roms(); |
41c094fd | 962 | |
4046d6e8 LT |
963 | /* after parse_early_param, so could debug it */ |
964 | insert_resource(&iomem_resource, &code_resource); | |
a3299754 | 965 | insert_resource(&iomem_resource, &rodata_resource); |
4046d6e8 LT |
966 | insert_resource(&iomem_resource, &data_resource); |
967 | insert_resource(&iomem_resource, &bss_resource); | |
968 | ||
b422a309 | 969 | e820_add_kernel_range(); |
1b5576e6 | 970 | trim_bios_range(); |
76934ed4 | 971 | #ifdef CONFIG_X86_32 |
cc9f7a0c | 972 | if (ppro_with_ram_bug()) { |
09821ff1 IM |
973 | e820__range_update(0x70000000ULL, 0x40000ULL, E820_TYPE_RAM, |
974 | E820_TYPE_RESERVED); | |
f9748fa0 | 975 | e820__update_table(e820_table); |
cc9f7a0c | 976 | printk(KERN_INFO "fixed physical RAM map:\n"); |
be0c3f0f | 977 | e820__print_table("bad_ppro"); |
cc9f7a0c | 978 | } |
76934ed4 YL |
979 | #else |
980 | early_gart_iommu_check(); | |
981 | #endif | |
cc9f7a0c | 982 | |
7b2a0a6c YL |
983 | /* |
984 | * partially used pages are not usable - thus | |
985 | * we are rounding upwards: | |
986 | */ | |
0c6fc11a | 987 | max_pfn = e820__end_of_ram_pfn(); |
7b2a0a6c | 988 | |
093af8d7 YL |
989 | /* update e820 for memory not covered by WB MTRRs */ |
990 | mtrr_bp_init(); | |
2dc807b3 | 991 | if (mtrr_trim_uncached_memory(max_pfn)) |
0c6fc11a | 992 | max_pfn = e820__end_of_ram_pfn(); |
76c32418 | 993 | |
8dd33030 IM |
994 | max_possible_pfn = max_pfn; |
995 | ||
99c13b8c MP |
996 | /* |
997 | * This call is required when the CPU does not support PAT. If | |
998 | * mtrr_bp_init() invoked it already via pat_init() the call has no | |
999 | * effect. | |
1000 | */ | |
1001 | init_cache_modes(); | |
1002 | ||
c7d2361f TG |
1003 | /* |
1004 | * Define random base addresses for memory sections after max_pfn is | |
1005 | * defined and before each memory section base is used. | |
1006 | */ | |
1007 | kernel_randomize_memory(); | |
1008 | ||
76934ed4 | 1009 | #ifdef CONFIG_X86_32 |
4e29684c | 1010 | /* max_low_pfn get updated here */ |
2ec65f8b | 1011 | find_low_pfn_range(); |
76934ed4 | 1012 | #else |
06cd9a7d | 1013 | check_x2apic(); |
76934ed4 YL |
1014 | |
1015 | /* How many end-of-memory variables you have, grandma! */ | |
1016 | /* need this before calling reserve_initrd */ | |
f361a450 | 1017 | if (max_pfn > (1UL<<(32 - PAGE_SHIFT))) |
0c6fc11a | 1018 | max_low_pfn = e820__end_of_low_ram_pfn(); |
f361a450 YL |
1019 | else |
1020 | max_low_pfn = max_pfn; | |
1021 | ||
76934ed4 | 1022 | high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1; |
5394f80f JF |
1023 | #endif |
1024 | ||
893f38d1 YL |
1025 | /* |
1026 | * Find and reserve possible boot-time SMP configuration: | |
1027 | */ | |
1028 | find_smp_config(); | |
1029 | ||
8d57470d YL |
1030 | early_alloc_pgt_buf(); |
1031 | ||
72d7c3b3 | 1032 | /* |
4918e228 | 1033 | * Need to conclude brk, before e820__memblock_setup() |
81519f77 C |
1034 | * it could use memblock_find_in_range, could overlap with |
1035 | * brk area. | |
72d7c3b3 YL |
1036 | */ |
1037 | reserve_brk(); | |
1038 | ||
e5f15b45 YL |
1039 | cleanup_highmap(); |
1040 | ||
2449f343 | 1041 | memblock_set_current_limit(ISA_END_ADDRESS); |
4918e228 | 1042 | e820__memblock_setup(); |
72d7c3b3 | 1043 | |
e998879d AK |
1044 | /* |
1045 | * Needs to run after memblock setup because it needs the physical | |
1046 | * memory size. | |
1047 | */ | |
1048 | sev_setup_arch(); | |
1049 | ||
6950e31b DW |
1050 | efi_fake_memmap(); |
1051 | efi_find_mirror(); | |
1052 | efi_esrt_init(); | |
58c90902 | 1053 | efi_mokvar_table_init(); |
007b7560 | 1054 | |
6950e31b DW |
1055 | /* |
1056 | * The EFI specification says that boot service code won't be | |
1057 | * called after ExitBootServices(). This is, in fact, a lie. | |
1058 | */ | |
1059 | efi_reserve_boot_services(); | |
916f676f | 1060 | |
72d7c3b3 | 1061 | /* preallocate 4k for mptable mpc */ |
5da217ca | 1062 | e820__memblock_alloc_reserved_mpc_new(); |
72d7c3b3 YL |
1063 | |
1064 | #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION | |
1065 | setup_bios_corruption_check(); | |
1066 | #endif | |
1067 | ||
10054230 | 1068 | #ifdef CONFIG_X86_32 |
365811d6 BH |
1069 | printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n", |
1070 | (max_pfn_mapped<<PAGE_SHIFT) - 1); | |
10054230 | 1071 | #endif |
72d7c3b3 | 1072 | |
c361e5d4 | 1073 | /* |
ec35d1d9 BP |
1074 | * Find free memory for the real mode trampoline and place it there. If |
1075 | * there is not enough free memory under 1M, on EFI-enabled systems | |
1076 | * there will be additional attempt to reclaim the memory for the real | |
1077 | * mode trampoline at efi_free_boot_services(). | |
f1d4d47c | 1078 | * |
ec35d1d9 BP |
1079 | * Unconditionally reserve the entire first 1M of RAM because BIOSes |
1080 | * are known to corrupt low memory and several hundred kilobytes are not | |
1081 | * worth complex detection what memory gets clobbered. Windows does the | |
1082 | * same thing for very similar reasons. | |
1083 | * | |
1084 | * Moreover, on machines with SandyBridge graphics or in setups that use | |
1085 | * crashkernel the entire 1M is reserved anyway. | |
c361e5d4 | 1086 | */ |
f1d4d47c | 1087 | reserve_real_mode(); |
a9acc536 | 1088 | |
22ddfcaa | 1089 | init_mem_mapping(); |
1bbbbe77 | 1090 | |
433f8924 | 1091 | idt_setup_early_pf(); |
1bbbbe77 | 1092 | |
18bc7bd5 AL |
1093 | /* |
1094 | * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features) | |
1095 | * with the current CR4 value. This may not be necessary, but | |
1096 | * auditing all the early-boot CR4 manipulation would be needed to | |
1097 | * rule it out. | |
c7ad5ad2 AL |
1098 | * |
1099 | * Mask off features that don't work outside long mode (just | |
1100 | * PCIDE for now). | |
18bc7bd5 | 1101 | */ |
c7ad5ad2 | 1102 | mmu_cr4_features = __read_cr4() & ~X86_CR4_PCIDE; |
18bc7bd5 | 1103 | |
4ce7a869 | 1104 | memblock_set_current_limit(get_max_mapped()); |
4e29684c | 1105 | |
e7b37895 YL |
1106 | /* |
1107 | * NOTE: On x86-32, only from this point on, fixmaps are ready for use. | |
1108 | */ | |
1109 | ||
1110 | #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT | |
1111 | if (init_ohci1394_dma_early) | |
1112 | init_ohci1394_dma_on_all_controllers(); | |
1113 | #endif | |
162a7e75 MT |
1114 | /* Allocate bigger log buffer */ |
1115 | setup_log_buf(1); | |
e7b37895 | 1116 | |
fede7320 | 1117 | efi_set_secure_boot(boot_params.secure_boot); |
9661b332 | 1118 | |
2ec65f8b YL |
1119 | reserve_initrd(); |
1120 | ||
da3d3f98 | 1121 | acpi_table_upgrade(); |
6998a880 RW |
1122 | /* Look for ACPI tables and reserve memory occupied by them. */ |
1123 | acpi_boot_table_init(); | |
53aac44c | 1124 | |
76934ed4 | 1125 | vsmp_init(); |
76934ed4 | 1126 | |
1c6e5503 YL |
1127 | io_delay_init(); |
1128 | ||
630b3aff LW |
1129 | early_platform_quirks(); |
1130 | ||
20e6926d YL |
1131 | early_acpi_boot_init(); |
1132 | ||
d8fc3afc | 1133 | initmem_init(); |
3c325f82 | 1134 | dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT); |
fa591c4a | 1135 | |
cf11e85f RG |
1136 | if (boot_cpu_has(X86_FEATURE_GBPAGES)) |
1137 | hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT); | |
1138 | ||
fa591c4a TC |
1139 | /* |
1140 | * Reserve memory for crash kernel after SRAT is parsed so that it | |
1141 | * won't consume hotpluggable memory. | |
1142 | */ | |
1143 | reserve_crashkernel(); | |
1144 | ||
6f2a7536 | 1145 | memblock_find_dma_reserve(); |
91467bdf | 1146 | |
ccb64941 BO |
1147 | if (!early_xdbc_setup_hardware()) |
1148 | early_xdbc_register_console(); | |
1149 | ||
7737b215 | 1150 | x86_init.paging.pagetable_init(); |
f212ec4b | 1151 | |
ef7f0d6a AR |
1152 | kasan_init(); |
1153 | ||
d2b6dc61 | 1154 | /* |
945fd17a TG |
1155 | * Sync back kernel address range. |
1156 | * | |
1157 | * FIXME: Can the later sync in setup_cpu_entry_areas() replace | |
1158 | * this call? | |
d2b6dc61 | 1159 | */ |
945fd17a | 1160 | sync_initial_page_table(); |
d2b6dc61 | 1161 | |
31625340 JC |
1162 | tboot_probe(); |
1163 | ||
76934ed4 | 1164 | map_vsyscall(); |
76934ed4 | 1165 | |
1a3f239d | 1166 | generic_apic_probe(); |
1da177e4 | 1167 | |
54ef3400 | 1168 | early_quirks(); |
d44647b0 | 1169 | |
295deae4 YL |
1170 | /* |
1171 | * Read APIC and some other early information from ACPI tables. | |
1172 | */ | |
1da177e4 | 1173 | acpi_boot_init(); |
a906fdaa | 1174 | x86_dtb_init(); |
04606618 | 1175 | |
295deae4 YL |
1176 | /* |
1177 | * get boot-time SMP configuration: | |
1178 | */ | |
a91bf718 | 1179 | get_smp_config(); |
76934ed4 | 1180 | |
1e90a13d TG |
1181 | /* |
1182 | * Systems w/o ACPI and mptables might not have it mapped the local | |
1183 | * APIC yet, but prefill_possible_map() might need to access it. | |
1184 | */ | |
1185 | init_apic_mappings(); | |
1186 | ||
329513a3 | 1187 | prefill_possible_map(); |
301e6190 | 1188 | |
5f4765f9 | 1189 | init_cpu_to_node(); |
73bf7382 | 1190 | init_gi_nodes(); |
5f4765f9 | 1191 | |
ca1b8862 | 1192 | io_apic_init_mappings(); |
9d6a4d08 | 1193 | |
f3614646 | 1194 | x86_init.hyper.guest_late_init(); |
1da177e4 | 1195 | |
1506c8dc | 1196 | e820__reserve_resources(); |
cc55f753 | 1197 | e820__register_nosave_regions(max_pfn); |
1da177e4 | 1198 | |
8fee697d | 1199 | x86_init.resources.reserve_resources(); |
41c094fd | 1200 | |
2df908ba | 1201 | e820__setup_pci_gap(); |
41c094fd | 1202 | |
1da177e4 LT |
1203 | #ifdef CONFIG_VT |
1204 | #if defined(CONFIG_VGA_CONSOLE) | |
83e68189 | 1205 | if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY)) |
1da177e4 | 1206 | conswitchp = &vga_con; |
1da177e4 LT |
1207 | #endif |
1208 | #endif | |
6f30c1ac | 1209 | x86_init.oem.banner(); |
a2202aa2 | 1210 | |
6b617e22 FT |
1211 | x86_init.timers.wallclock_init(); |
1212 | ||
9a90ed06 BP |
1213 | /* |
1214 | * This needs to run before setup_local_APIC() which soft-disables the | |
1215 | * local APIC temporarily and that masks the thermal LVT interrupt, | |
1216 | * leading to softlockups on machines which have configured SMI | |
1217 | * interrupt delivery. | |
1218 | */ | |
1219 | therm_lvt_init(); | |
1220 | ||
a2202aa2 | 1221 | mcheck_init(); |
f49aa448 | 1222 | |
b3c869d3 | 1223 | register_refined_jiffies(CLOCK_TICK_RATE); |
5189c2a7 OJ |
1224 | |
1225 | #ifdef CONFIG_EFI | |
a5d90c92 BP |
1226 | if (efi_enabled(EFI_BOOT)) |
1227 | efi_apply_memmap_quirks(); | |
5189c2a7 | 1228 | #endif |
ee9f8fce JP |
1229 | |
1230 | unwind_init(); | |
1da177e4 | 1231 | } |
5649b7c3 | 1232 | |
9be1b56a IM |
1233 | #ifdef CONFIG_X86_32 |
1234 | ||
8fee697d TG |
1235 | static struct resource video_ram_resource = { |
1236 | .name = "Video RAM area", | |
1237 | .start = 0xa0000, | |
1238 | .end = 0xbffff, | |
1239 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
9be1b56a IM |
1240 | }; |
1241 | ||
8fee697d | 1242 | void __init i386_reserve_resources(void) |
9be1b56a | 1243 | { |
8fee697d TG |
1244 | request_resource(&iomem_resource, &video_ram_resource); |
1245 | reserve_standard_io_resources(); | |
9be1b56a IM |
1246 | } |
1247 | ||
9be1b56a | 1248 | #endif /* CONFIG_X86_32 */ |
f32360ef KC |
1249 | |
1250 | static struct notifier_block kernel_offset_notifier = { | |
1251 | .notifier_call = dump_kernel_offset | |
1252 | }; | |
1253 | ||
1254 | static int __init register_kernel_offset_dumper(void) | |
1255 | { | |
1256 | atomic_notifier_chain_register(&panic_notifier_list, | |
1257 | &kernel_offset_notifier); | |
1258 | return 0; | |
1259 | } | |
1260 | __initcall(register_kernel_offset_dumper); |