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> |
b69a2afd | 14 | #include <linux/ima.h> |
12609013 IM |
15 | #include <linux/init_ohci1394_dma.h> |
16 | #include <linux/initrd.h> | |
138fe4e0 | 17 | #include <linux/iscsi_ibft.h> |
12609013 | 18 | #include <linux/memblock.h> |
f39650de | 19 | #include <linux/panic_notifier.h> |
376ff035 | 20 | #include <linux/pci.h> |
12609013 | 21 | #include <linux/root_dev.h> |
cf11e85f | 22 | #include <linux/hugetlb.h> |
69575d38 | 23 | #include <linux/tboot.h> |
1b5aeebf | 24 | #include <linux/usb/xhci-dbgp.h> |
a945c834 | 25 | #include <linux/static_call.h> |
0a0f0d8b | 26 | #include <linux/swiotlb.h> |
68b8e971 | 27 | #include <linux/random.h> |
1bc3b91a | 28 | |
12609013 IM |
29 | #include <uapi/linux/mount.h> |
30 | ||
31 | #include <xen/xen.h> | |
32 | ||
9635b47d | 33 | #include <asm/apic.h> |
4059ba65 | 34 | #include <asm/efi.h> |
13c01139 | 35 | #include <asm/numa.h> |
ce3fe6b2 | 36 | #include <asm/bios_ebda.h> |
cc9f7a0c | 37 | #include <asm/bugs.h> |
0b9a6a8b | 38 | #include <asm/cacheinfo.h> |
6e5385d4 | 39 | #include <asm/cpu.h> |
12609013 | 40 | #include <asm/efi.h> |
1d9b16d1 | 41 | #include <asm/gart.h> |
88b094fb | 42 | #include <asm/hypervisor.h> |
12609013 IM |
43 | #include <asm/io_apic.h> |
44 | #include <asm/kasan.h> | |
45 | #include <asm/kaslr.h> | |
a2202aa2 | 46 | #include <asm/mce.h> |
4d5cff69 | 47 | #include <asm/memtype.h> |
12609013 | 48 | #include <asm/mtrr.h> |
ca947b72 | 49 | #include <asm/realmode.h> |
12609013 IM |
50 | #include <asm/olpc_ofw.h> |
51 | #include <asm/pci-direct.h> | |
da6b737b | 52 | #include <asm/prom.h> |
12609013 | 53 | #include <asm/proto.h> |
9a90ed06 | 54 | #include <asm/thermal.h> |
ee9f8fce | 55 | #include <asm/unwind.h> |
12609013 | 56 | #include <asm/vsyscall.h> |
186525bd | 57 | #include <linux/vmalloc.h> |
46d671b5 | 58 | |
2b72394e | 59 | /* |
360db4ac IM |
60 | * max_low_pfn_mapped: highest directly mapped pfn < 4 GB |
61 | * max_pfn_mapped: highest directly mapped pfn > 4 GB | |
66520ebc | 62 | * |
09821ff1 | 63 | * The direct mapping only covers E820_TYPE_RAM regions, so the ranges and gaps are |
360db4ac | 64 | * represented by pfn_mapped[]. |
2b72394e PE |
65 | */ |
66 | unsigned long max_low_pfn_mapped; | |
67 | unsigned long max_pfn_mapped; | |
68 | ||
e808bae2 | 69 | #ifdef CONFIG_DMI |
796216a5 | 70 | RESERVE_BRK(dmi_alloc, 65536); |
e808bae2 | 71 | #endif |
796216a5 | 72 | |
c0b5842a | 73 | |
360db4ac IM |
74 | unsigned long _brk_start = (unsigned long)__brk_base; |
75 | unsigned long _brk_end = (unsigned long)__brk_base; | |
93dbda7c | 76 | |
217b8ce8 | 77 | struct boot_params boot_params; |
217b8ce8 | 78 | |
4046d6e8 | 79 | /* |
360db4ac IM |
80 | * These are the four main kernel memory regions, we put them into |
81 | * the resource tree so that kdump tools and other debugging tools | |
82 | * recover it: | |
4046d6e8 | 83 | */ |
360db4ac | 84 | |
a3299754 KC |
85 | static struct resource rodata_resource = { |
86 | .name = "Kernel rodata", | |
87 | .start = 0, | |
88 | .end = 0, | |
89 | .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM | |
90 | }; | |
91 | ||
4046d6e8 LT |
92 | static struct resource data_resource = { |
93 | .name = "Kernel data", | |
94 | .start = 0, | |
95 | .end = 0, | |
96 | .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM | |
97 | }; | |
98 | ||
99 | static struct resource code_resource = { | |
100 | .name = "Kernel code", | |
101 | .start = 0, | |
102 | .end = 0, | |
103 | .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM | |
104 | }; | |
105 | ||
106 | static struct resource bss_resource = { | |
107 | .name = "Kernel bss", | |
108 | .start = 0, | |
109 | .end = 0, | |
110 | .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM | |
111 | }; | |
112 | ||
113 | ||
7dea23ec | 114 | #ifdef CONFIG_X86_32 |
360db4ac | 115 | /* CPU data as detected by the assembly code in head_32.S */ |
6415813b | 116 | struct cpuinfo_x86 new_cpu_data; |
0c254e38 AS |
117 | unsigned int def_to_bigsmp; |
118 | ||
7dea23ec YL |
119 | struct apm_info apm_info; |
120 | EXPORT_SYMBOL(apm_info); | |
121 | ||
122 | #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \ | |
123 | defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE) | |
124 | struct ist_info ist_info; | |
125 | EXPORT_SYMBOL(ist_info); | |
126 | #else | |
127 | struct ist_info ist_info; | |
128 | #endif | |
129 | ||
7dea23ec YL |
130 | #endif |
131 | ||
50c66d7b YW |
132 | struct cpuinfo_x86 boot_cpu_data __read_mostly; |
133 | EXPORT_SYMBOL(boot_cpu_data); | |
7dea23ec YL |
134 | |
135 | #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64) | |
404f6aac | 136 | __visible unsigned long mmu_cr4_features __ro_after_init; |
7dea23ec | 137 | #else |
404f6aac | 138 | __visible unsigned long mmu_cr4_features __ro_after_init = X86_CR4_PAE; |
7dea23ec YL |
139 | #endif |
140 | ||
b69a2afd JM |
141 | #ifdef CONFIG_IMA |
142 | static phys_addr_t ima_kexec_buffer_phys; | |
143 | static size_t ima_kexec_buffer_size; | |
144 | #endif | |
145 | ||
5031296c PA |
146 | /* Boot loader ID and version as integers, for the benefit of proc_dointvec */ |
147 | int bootloader_type, bootloader_version; | |
1da177e4 | 148 | |
1da177e4 LT |
149 | /* |
150 | * Setup options | |
151 | */ | |
1da177e4 | 152 | struct screen_info screen_info; |
129f6946 | 153 | EXPORT_SYMBOL(screen_info); |
1da177e4 | 154 | struct edid_info edid_info; |
5e518d76 | 155 | EXPORT_SYMBOL_GPL(edid_info); |
1da177e4 | 156 | |
1da177e4 LT |
157 | extern int root_mountflags; |
158 | ||
e44b7b75 | 159 | unsigned long saved_video_mode; |
1da177e4 | 160 | |
cf8fa920 | 161 | #define RAMDISK_IMAGE_START_MASK 0x07FF |
1da177e4 | 162 | #define RAMDISK_PROMPT_FLAG 0x8000 |
cf8fa920 | 163 | #define RAMDISK_LOAD_FLAG 0x4000 |
1da177e4 | 164 | |
4e498b66 | 165 | static char __initdata command_line[COMMAND_LINE_SIZE]; |
516cbf37 TB |
166 | #ifdef CONFIG_CMDLINE_BOOL |
167 | static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE; | |
168 | #endif | |
1da177e4 | 169 | |
1da177e4 LT |
170 | #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE) |
171 | struct edd edd; | |
172 | #ifdef CONFIG_EDD_MODULE | |
173 | EXPORT_SYMBOL(edd); | |
174 | #endif | |
175 | /** | |
176 | * copy_edd() - Copy the BIOS EDD information | |
177 | * from boot_params into a safe place. | |
178 | * | |
179 | */ | |
9eaa192d | 180 | static inline void __init copy_edd(void) |
1da177e4 | 181 | { |
30c82645 PA |
182 | memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer, |
183 | sizeof(edd.mbr_signature)); | |
184 | memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info)); | |
185 | edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries; | |
186 | edd.edd_info_nr = boot_params.eddbuf_entries; | |
1da177e4 LT |
187 | } |
188 | #else | |
9eaa192d | 189 | static inline void __init copy_edd(void) |
1da177e4 LT |
190 | { |
191 | } | |
192 | #endif | |
193 | ||
5368a2be PA |
194 | void * __init extend_brk(size_t size, size_t align) |
195 | { | |
196 | size_t mask = align - 1; | |
197 | void *ret; | |
198 | ||
199 | BUG_ON(_brk_start == 0); | |
200 | BUG_ON(align & mask); | |
201 | ||
202 | _brk_end = (_brk_end + mask) & ~mask; | |
203 | BUG_ON((char *)(_brk_end + size) > __brk_limit); | |
204 | ||
205 | ret = (void *)_brk_end; | |
206 | _brk_end += size; | |
207 | ||
208 | memset(ret, 0, size); | |
209 | ||
210 | return ret; | |
211 | } | |
212 | ||
148b2098 | 213 | #ifdef CONFIG_X86_32 |
e5f15b45 | 214 | static void __init cleanup_highmap(void) |
f005fe12 YL |
215 | { |
216 | } | |
854c879f PE |
217 | #endif |
218 | ||
5368a2be PA |
219 | static void __init reserve_brk(void) |
220 | { | |
221 | if (_brk_end > _brk_start) | |
fc8d7826 AD |
222 | memblock_reserve(__pa_symbol(_brk_start), |
223 | _brk_end - _brk_start); | |
5368a2be PA |
224 | |
225 | /* Mark brk area as locked down and no longer taking any | |
226 | new allocations */ | |
227 | _brk_start = 0; | |
228 | } | |
229 | ||
5aa3d718 BP |
230 | u64 relocated_ramdisk; |
231 | ||
cf8fa920 PA |
232 | #ifdef CONFIG_BLK_DEV_INITRD |
233 | ||
a8a51a88 YL |
234 | static u64 __init get_ramdisk_image(void) |
235 | { | |
236 | u64 ramdisk_image = boot_params.hdr.ramdisk_image; | |
237 | ||
ee92d815 YL |
238 | ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32; |
239 | ||
694cfd87 RM |
240 | if (ramdisk_image == 0) |
241 | ramdisk_image = phys_initrd_start; | |
242 | ||
a8a51a88 YL |
243 | return ramdisk_image; |
244 | } | |
245 | static u64 __init get_ramdisk_size(void) | |
246 | { | |
247 | u64 ramdisk_size = boot_params.hdr.ramdisk_size; | |
248 | ||
ee92d815 YL |
249 | ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32; |
250 | ||
694cfd87 RM |
251 | if (ramdisk_size == 0) |
252 | ramdisk_size = phys_initrd_size; | |
253 | ||
a8a51a88 YL |
254 | return ramdisk_size; |
255 | } | |
256 | ||
eb1379cb | 257 | static void __init relocate_initrd(void) |
cf8fa920 | 258 | { |
c967da6a | 259 | /* Assume only end is not page aligned */ |
a8a51a88 YL |
260 | u64 ramdisk_image = get_ramdisk_image(); |
261 | u64 ramdisk_size = get_ramdisk_size(); | |
c967da6a | 262 | u64 area_size = PAGE_ALIGN(ramdisk_size); |
cf8fa920 | 263 | |
e8c57d40 | 264 | /* We need to move the initrd down into directly mapped mem */ |
3c45ee6d MR |
265 | relocated_ramdisk = memblock_phys_alloc_range(area_size, PAGE_SIZE, 0, |
266 | PFN_PHYS(max_pfn_mapped)); | |
5aa3d718 | 267 | if (!relocated_ramdisk) |
3945e2c9 | 268 | panic("Cannot find place for new RAMDISK of size %lld\n", |
5aa3d718 | 269 | ramdisk_size); |
3945e2c9 | 270 | |
5aa3d718 | 271 | initrd_start = relocated_ramdisk + PAGE_OFFSET; |
cf8fa920 | 272 | initrd_end = initrd_start + ramdisk_size; |
365811d6 | 273 | printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n", |
5aa3d718 | 274 | relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1); |
cf8fa920 | 275 | |
5dd2c4bd MS |
276 | copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size); |
277 | ||
365811d6 BH |
278 | printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to" |
279 | " [mem %#010llx-%#010llx]\n", | |
ba5b14cc | 280 | ramdisk_image, ramdisk_image + ramdisk_size - 1, |
5aa3d718 | 281 | relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1); |
eb1379cb | 282 | } |
9a27f5c5 | 283 | |
1b8c78be YL |
284 | static void __init early_reserve_initrd(void) |
285 | { | |
286 | /* Assume only end is not page aligned */ | |
a8a51a88 YL |
287 | u64 ramdisk_image = get_ramdisk_image(); |
288 | u64 ramdisk_size = get_ramdisk_size(); | |
1b8c78be YL |
289 | u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size); |
290 | ||
291 | if (!boot_params.hdr.type_of_loader || | |
292 | !ramdisk_image || !ramdisk_size) | |
293 | return; /* No initrd provided by bootloader */ | |
294 | ||
295 | memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image); | |
296 | } | |
3c45ee6d | 297 | |
eb1379cb YL |
298 | static void __init reserve_initrd(void) |
299 | { | |
c967da6a | 300 | /* Assume only end is not page aligned */ |
a8a51a88 YL |
301 | u64 ramdisk_image = get_ramdisk_image(); |
302 | u64 ramdisk_size = get_ramdisk_size(); | |
c967da6a | 303 | u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size); |
eb1379cb YL |
304 | |
305 | if (!boot_params.hdr.type_of_loader || | |
306 | !ramdisk_image || !ramdisk_size) | |
307 | return; /* No initrd provided by bootloader */ | |
308 | ||
309 | initrd_start = 0; | |
310 | ||
365811d6 BH |
311 | printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image, |
312 | ramdisk_end - 1); | |
eb1379cb | 313 | |
74f27655 | 314 | if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image), |
e8c57d40 YL |
315 | PFN_DOWN(ramdisk_end))) { |
316 | /* All are mapped, easy case */ | |
eb1379cb YL |
317 | initrd_start = ramdisk_image + PAGE_OFFSET; |
318 | initrd_end = initrd_start + ramdisk_size; | |
319 | return; | |
320 | } | |
321 | ||
eb1379cb | 322 | relocate_initrd(); |
8c5dd8f4 | 323 | |
3ecc6834 | 324 | memblock_phys_free(ramdisk_image, ramdisk_end - ramdisk_image); |
cf8fa920 | 325 | } |
af06f8b7 | 326 | |
225c37d7 | 327 | #else |
1b8c78be YL |
328 | static void __init early_reserve_initrd(void) |
329 | { | |
330 | } | |
eb1379cb | 331 | static void __init reserve_initrd(void) |
225c37d7 YL |
332 | { |
333 | } | |
cf8fa920 PA |
334 | #endif /* CONFIG_BLK_DEV_INITRD */ |
335 | ||
b69a2afd JM |
336 | static void __init add_early_ima_buffer(u64 phys_addr) |
337 | { | |
338 | #ifdef CONFIG_IMA | |
339 | struct ima_setup_data *data; | |
340 | ||
341 | data = early_memremap(phys_addr + sizeof(struct setup_data), sizeof(*data)); | |
342 | if (!data) { | |
343 | pr_warn("setup: failed to memremap ima_setup_data entry\n"); | |
344 | return; | |
345 | } | |
346 | ||
347 | if (data->size) { | |
348 | memblock_reserve(data->addr, data->size); | |
349 | ima_kexec_buffer_phys = data->addr; | |
350 | ima_kexec_buffer_size = data->size; | |
351 | } | |
352 | ||
353 | early_memunmap(data, sizeof(*data)); | |
354 | #else | |
355 | pr_warn("Passed IMA kexec data, but CONFIG_IMA not set. Ignoring.\n"); | |
356 | #endif | |
357 | } | |
358 | ||
359 | #if defined(CONFIG_HAVE_IMA_KEXEC) && !defined(CONFIG_OF_FLATTREE) | |
360 | int __init ima_free_kexec_buffer(void) | |
361 | { | |
362 | int rc; | |
363 | ||
364 | if (!ima_kexec_buffer_size) | |
365 | return -ENOENT; | |
366 | ||
367 | rc = memblock_phys_free(ima_kexec_buffer_phys, | |
368 | ima_kexec_buffer_size); | |
369 | if (rc) | |
370 | return rc; | |
371 | ||
372 | ima_kexec_buffer_phys = 0; | |
373 | ima_kexec_buffer_size = 0; | |
374 | ||
375 | return 0; | |
376 | } | |
377 | ||
378 | int __init ima_get_kexec_buffer(void **addr, size_t *size) | |
379 | { | |
380 | if (!ima_kexec_buffer_size) | |
381 | return -ENOENT; | |
382 | ||
383 | *addr = __va(ima_kexec_buffer_phys); | |
384 | *size = ima_kexec_buffer_size; | |
385 | ||
386 | return 0; | |
387 | } | |
388 | #endif | |
389 | ||
29f784e3 | 390 | static void __init parse_setup_data(void) |
257b0fde YL |
391 | { |
392 | struct setup_data *data; | |
30e46b57 | 393 | u64 pa_data, pa_next; |
257b0fde | 394 | |
257b0fde YL |
395 | pa_data = boot_params.hdr.setup_data; |
396 | while (pa_data) { | |
7389882c | 397 | u32 data_len, data_type; |
f1c2b357 | 398 | |
7389882c | 399 | data = early_memremap(pa_data, sizeof(*data)); |
f1c2b357 | 400 | data_len = data->len + sizeof(struct setup_data); |
30e46b57 LC |
401 | data_type = data->type; |
402 | pa_next = data->next; | |
8d4a40bc | 403 | early_memunmap(data, sizeof(*data)); |
f1c2b357 | 404 | |
30e46b57 | 405 | switch (data_type) { |
257b0fde | 406 | case SETUP_E820_EXT: |
914053c0 | 407 | e820__memory_setup_extended(pa_data, data_len); |
257b0fde | 408 | break; |
da6b737b SAS |
409 | case SETUP_DTB: |
410 | add_dtb(pa_data); | |
257b0fde | 411 | break; |
1fec0533 DY |
412 | case SETUP_EFI: |
413 | parse_efi_setup(pa_data, data_len); | |
414 | break; | |
b69a2afd JM |
415 | case SETUP_IMA: |
416 | add_early_ima_buffer(pa_data); | |
417 | break; | |
68b8e971 JD |
418 | case SETUP_RNG_SEED: |
419 | data = early_memremap(pa_data, data_len); | |
420 | add_bootloader_randomness(data->data, data->len); | |
421 | /* Zero seed for forward secrecy. */ | |
422 | memzero_explicit(data->data, data->len); | |
423 | /* Zero length in case we find ourselves back here by accident. */ | |
424 | memzero_explicit(&data->len, sizeof(data->len)); | |
425 | early_memunmap(data, data_len); | |
426 | break; | |
257b0fde YL |
427 | default: |
428 | break; | |
429 | } | |
30e46b57 | 430 | pa_data = pa_next; |
257b0fde YL |
431 | } |
432 | } | |
433 | ||
a9ce6bc1 | 434 | static void __init memblock_x86_reserve_range_setup_data(void) |
a0a0becd | 435 | { |
7228918b | 436 | struct setup_indirect *indirect; |
a0a0becd | 437 | struct setup_data *data; |
7228918b RP |
438 | u64 pa_data, pa_next; |
439 | u32 len; | |
a0a0becd | 440 | |
a0a0becd YL |
441 | pa_data = boot_params.hdr.setup_data; |
442 | while (pa_data) { | |
88b4c146 | 443 | data = early_memremap(pa_data, sizeof(*data)); |
7228918b RP |
444 | if (!data) { |
445 | pr_warn("setup: failed to memremap setup_data entry\n"); | |
446 | return; | |
447 | } | |
448 | ||
449 | len = sizeof(*data); | |
450 | pa_next = data->next; | |
451 | ||
24aa0788 | 452 | memblock_reserve(pa_data, sizeof(*data) + data->len); |
b3c72fc9 | 453 | |
7228918b RP |
454 | if (data->type == SETUP_INDIRECT) { |
455 | len += data->len; | |
456 | early_memunmap(data, sizeof(*data)); | |
457 | data = early_memremap(pa_data, len); | |
458 | if (!data) { | |
459 | pr_warn("setup: failed to memremap indirect setup_data\n"); | |
460 | return; | |
461 | } | |
b3c72fc9 | 462 | |
7228918b RP |
463 | indirect = (struct setup_indirect *)data->data; |
464 | ||
465 | if (indirect->type != SETUP_INDIRECT) | |
466 | memblock_reserve(indirect->addr, indirect->len); | |
467 | } | |
468 | ||
469 | pa_data = pa_next; | |
470 | early_memunmap(data, len); | |
a0a0becd YL |
471 | } |
472 | } | |
473 | ||
ccb4defa YL |
474 | /* |
475 | * --------- Crashkernel reservation ------------------------------ | |
476 | */ | |
477 | ||
606134f7 | 478 | /* 16M alignment for crash kernel regions */ |
9ca5c8e6 | 479 | #define CRASH_ALIGN SZ_16M |
606134f7 | 480 | |
7f8595bf | 481 | /* |
8ff80fbe BH |
482 | * Keep the crash kernel below this limit. |
483 | * | |
360db4ac | 484 | * Earlier 32-bits kernels would limit the kernel to the low 512 MB range |
8ff80fbe BH |
485 | * due to mapping restrictions. |
486 | * | |
360db4ac | 487 | * 64-bit kdump kernels need to be restricted to be under 64 TB, which is |
11a98f37 | 488 | * the upper limit of system RAM in 4-level paging mode. Since the kdump |
360db4ac IM |
489 | * jump could be from 5-level paging to 4-level paging, the jump will fail if |
490 | * the kernel is put above 64 TB, and during the 1st kernel bootup there's | |
491 | * no good way to detect the paging mode of the target kernel which will be | |
492 | * loaded for dumping. | |
7f8595bf PA |
493 | */ |
494 | #ifdef CONFIG_X86_32 | |
9ca5c8e6 DY |
495 | # define CRASH_ADDR_LOW_MAX SZ_512M |
496 | # define CRASH_ADDR_HIGH_MAX SZ_512M | |
7f8595bf | 497 | #else |
9ca5c8e6 | 498 | # define CRASH_ADDR_LOW_MAX SZ_4G |
8ff80fbe | 499 | # define CRASH_ADDR_HIGH_MAX SZ_64T |
7f8595bf PA |
500 | #endif |
501 | ||
eb6db83d | 502 | static int __init reserve_crashkernel_low(void) |
0212f915 YL |
503 | { |
504 | #ifdef CONFIG_X86_64 | |
f56d5578 | 505 | unsigned long long base, low_base = 0, low_size = 0; |
6120cdc0 | 506 | unsigned long low_mem_limit; |
0212f915 YL |
507 | int ret; |
508 | ||
6120cdc0 | 509 | low_mem_limit = min(memblock_phys_mem_size(), CRASH_ADDR_LOW_MAX); |
97eac21b | 510 | |
adbc742b | 511 | /* crashkernel=Y,low */ |
6120cdc0 | 512 | ret = parse_crashkernel_low(boot_command_line, low_mem_limit, &low_size, &base); |
f56d5578 | 513 | if (ret) { |
c729de8f | 514 | /* |
392e879a | 515 | * two parts from kernel/dma/swiotlb.c: |
94fb9334 JR |
516 | * -swiotlb size: user-specified with swiotlb= or default. |
517 | * | |
518 | * -swiotlb overflow buffer: now hardcoded to 32k. We round it | |
519 | * to 8M for other buffers that may need to stay low too. Also | |
520 | * make sure we allocate enough extra low memory so that we | |
521 | * don't run out of DMA buffers for 32-bit devices. | |
c729de8f | 522 | */ |
97eac21b | 523 | low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20); |
c729de8f | 524 | } else { |
adbc742b | 525 | /* passed with crashkernel=0,low ? */ |
c729de8f | 526 | if (!low_size) |
eb6db83d | 527 | return 0; |
c729de8f | 528 | } |
0212f915 | 529 | |
6120cdc0 | 530 | low_base = memblock_phys_alloc_range(low_size, CRASH_ALIGN, 0, CRASH_ADDR_LOW_MAX); |
0212f915 | 531 | if (!low_base) { |
eb6db83d BH |
532 | pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n", |
533 | (unsigned long)(low_size >> 20)); | |
534 | return -ENOMEM; | |
0212f915 | 535 | } |
0212f915 | 536 | |
6120cdc0 | 537 | pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (low RAM limit: %ldMB)\n", |
97eac21b BP |
538 | (unsigned long)(low_size >> 20), |
539 | (unsigned long)(low_base >> 20), | |
6120cdc0 | 540 | (unsigned long)(low_mem_limit >> 20)); |
97eac21b | 541 | |
0212f915 YL |
542 | crashk_low_res.start = low_base; |
543 | crashk_low_res.end = low_base + low_size - 1; | |
544 | insert_resource(&iomem_resource, &crashk_low_res); | |
7f8595bf | 545 | #endif |
eb6db83d | 546 | return 0; |
0212f915 | 547 | } |
7f8595bf | 548 | |
29f784e3 | 549 | static void __init reserve_crashkernel(void) |
ccb4defa | 550 | { |
f56d5578 | 551 | unsigned long long crash_size, crash_base, total_mem; |
55a20ee7 | 552 | bool high = false; |
ccb4defa YL |
553 | int ret; |
554 | ||
4ece09be JZ |
555 | if (!IS_ENABLED(CONFIG_KEXEC_CORE)) |
556 | return; | |
557 | ||
09c71bfd | 558 | total_mem = memblock_phys_mem_size(); |
ccb4defa | 559 | |
55a20ee7 | 560 | /* crashkernel=XM */ |
97eac21b | 561 | ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base); |
55a20ee7 | 562 | if (ret != 0 || crash_size <= 0) { |
adbc742b | 563 | /* crashkernel=X,high */ |
55a20ee7 | 564 | ret = parse_crashkernel_high(boot_command_line, total_mem, |
97eac21b | 565 | &crash_size, &crash_base); |
55a20ee7 YL |
566 | if (ret != 0 || crash_size <= 0) |
567 | return; | |
568 | high = true; | |
569 | } | |
32105f7f | 570 | |
3db3eb28 PT |
571 | if (xen_pv_domain()) { |
572 | pr_info("Ignoring crashkernel for a Xen PV domain\n"); | |
573 | return; | |
574 | } | |
575 | ||
32105f7f | 576 | /* 0 means: find the address automatically */ |
b9ac3849 | 577 | if (!crash_base) { |
9f4c1396 | 578 | /* |
a8d4c824 | 579 | * Set CRASH_ADDR_LOW_MAX upper bound for crash memory, |
b9ac3849 DY |
580 | * crashkernel=x,high reserves memory over 4G, also allocates |
581 | * 256M extra low memory for DMA buffers and swiotlb. | |
582 | * But the extra memory is not required for all machines. | |
583 | * So try low memory first and fall back to high memory | |
584 | * unless "crashkernel=size[KMG],high" is specified. | |
9f4c1396 | 585 | */ |
b9ac3849 | 586 | if (!high) |
6120cdc0 MR |
587 | crash_base = memblock_phys_alloc_range(crash_size, |
588 | CRASH_ALIGN, CRASH_ALIGN, | |
589 | CRASH_ADDR_LOW_MAX); | |
b9ac3849 | 590 | if (!crash_base) |
6120cdc0 MR |
591 | crash_base = memblock_phys_alloc_range(crash_size, |
592 | CRASH_ALIGN, CRASH_ALIGN, | |
593 | CRASH_ADDR_HIGH_MAX); | |
1f5026a7 | 594 | if (!crash_base) { |
44280733 | 595 | pr_info("crashkernel reservation failed - No suitable area found.\n"); |
ccb4defa YL |
596 | return; |
597 | } | |
32105f7f | 598 | } else { |
44280733 YL |
599 | unsigned long long start; |
600 | ||
6120cdc0 MR |
601 | start = memblock_phys_alloc_range(crash_size, SZ_1M, crash_base, |
602 | crash_base + crash_size); | |
44280733 YL |
603 | if (start != crash_base) { |
604 | pr_info("crashkernel reservation failed - memory is in use.\n"); | |
ccb4defa YL |
605 | return; |
606 | } | |
32105f7f | 607 | } |
ccb4defa | 608 | |
eb6db83d | 609 | if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) { |
3ecc6834 | 610 | memblock_phys_free(crash_base, crash_size); |
eb6db83d BH |
611 | return; |
612 | } | |
ccb4defa | 613 | |
f56d5578 BP |
614 | pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n", |
615 | (unsigned long)(crash_size >> 20), | |
616 | (unsigned long)(crash_base >> 20), | |
617 | (unsigned long)(total_mem >> 20)); | |
ccb4defa | 618 | |
32105f7f BW |
619 | crashk_res.start = crash_base; |
620 | crashk_res.end = crash_base + crash_size - 1; | |
621 | insert_resource(&iomem_resource, &crashk_res); | |
ccb4defa | 622 | } |
ccb4defa | 623 | |
bdba0e70 YL |
624 | static struct resource standard_io_resources[] = { |
625 | { .name = "dma1", .start = 0x00, .end = 0x1f, | |
626 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
627 | { .name = "pic1", .start = 0x20, .end = 0x21, | |
628 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
629 | { .name = "timer0", .start = 0x40, .end = 0x43, | |
630 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
631 | { .name = "timer1", .start = 0x50, .end = 0x53, | |
632 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
633 | { .name = "keyboard", .start = 0x60, .end = 0x60, | |
634 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
635 | { .name = "keyboard", .start = 0x64, .end = 0x64, | |
636 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
637 | { .name = "dma page reg", .start = 0x80, .end = 0x8f, | |
638 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
639 | { .name = "pic2", .start = 0xa0, .end = 0xa1, | |
640 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
641 | { .name = "dma2", .start = 0xc0, .end = 0xdf, | |
642 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, | |
643 | { .name = "fpu", .start = 0xf0, .end = 0xff, | |
644 | .flags = IORESOURCE_BUSY | IORESOURCE_IO } | |
645 | }; | |
646 | ||
8fee697d | 647 | void __init reserve_standard_io_resources(void) |
bdba0e70 YL |
648 | { |
649 | int i; | |
650 | ||
651 | /* request I/O space for devices used on all i[345]86 PCs */ | |
652 | for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++) | |
653 | request_resource(&ioport_resource, &standard_io_resources[i]); | |
654 | ||
655 | } | |
656 | ||
a9acc536 JB |
657 | static bool __init snb_gfx_workaround_needed(void) |
658 | { | |
e43b3cec | 659 | #ifdef CONFIG_PCI |
a9acc536 JB |
660 | int i; |
661 | u16 vendor, devid; | |
ab3cd867 | 662 | static const __initconst u16 snb_ids[] = { |
a9acc536 JB |
663 | 0x0102, |
664 | 0x0112, | |
665 | 0x0122, | |
666 | 0x0106, | |
667 | 0x0116, | |
668 | 0x0126, | |
669 | 0x010a, | |
670 | }; | |
671 | ||
672 | /* Assume no if something weird is going on with PCI */ | |
673 | if (!early_pci_allowed()) | |
674 | return false; | |
675 | ||
676 | vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID); | |
677 | if (vendor != 0x8086) | |
678 | return false; | |
679 | ||
680 | devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID); | |
681 | for (i = 0; i < ARRAY_SIZE(snb_ids); i++) | |
682 | if (devid == snb_ids[i]) | |
683 | return true; | |
e43b3cec | 684 | #endif |
a9acc536 JB |
685 | |
686 | return false; | |
687 | } | |
688 | ||
689 | /* | |
690 | * Sandy Bridge graphics has trouble with certain ranges, exclude | |
691 | * them from allocation. | |
692 | */ | |
693 | static void __init trim_snb_memory(void) | |
694 | { | |
ab3cd867 | 695 | static const __initconst unsigned long bad_pages[] = { |
a9acc536 JB |
696 | 0x20050000, |
697 | 0x20110000, | |
698 | 0x20130000, | |
699 | 0x20138000, | |
700 | 0x40004000, | |
701 | }; | |
702 | int i; | |
703 | ||
704 | if (!snb_gfx_workaround_needed()) | |
705 | return; | |
706 | ||
707 | printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n"); | |
708 | ||
709 | /* | |
c361e5d4 MR |
710 | * SandyBridge integrated graphics devices have a bug that prevents |
711 | * them from accessing certain memory ranges, namely anything below | |
712 | * 1M and in the pages listed in bad_pages[] above. | |
713 | * | |
f1d4d47c MR |
714 | * To avoid these pages being ever accessed by SNB gfx devices reserve |
715 | * bad_pages that have not already been reserved at boot time. | |
716 | * All memory below the 1 MB mark is anyway reserved later during | |
717 | * setup_arch(), so there is no need to reserve it here. | |
a9acc536 | 718 | */ |
c361e5d4 | 719 | |
a9acc536 JB |
720 | for (i = 0; i < ARRAY_SIZE(bad_pages); i++) { |
721 | if (memblock_reserve(bad_pages[i], PAGE_SIZE)) | |
722 | printk(KERN_WARNING "failed to reserve 0x%08lx\n", | |
723 | bad_pages[i]); | |
724 | } | |
725 | } | |
726 | ||
1b5576e6 YL |
727 | static void __init trim_bios_range(void) |
728 | { | |
5c279c4c MR |
729 | /* |
730 | * A special case is the first 4Kb of memory; | |
731 | * This is a BIOS owned area, not kernel ram, but generally | |
732 | * not listed as such in the E820 table. | |
733 | * | |
734 | * This typically reserves additional memory (64KiB by default) | |
735 | * since some BIOSes are known to corrupt low memory. See the | |
736 | * Kconfig help text for X86_RESERVE_LOW. | |
737 | */ | |
738 | e820__range_update(0, PAGE_SIZE, E820_TYPE_RAM, E820_TYPE_RESERVED); | |
739 | ||
1b5576e6 | 740 | /* |
11a98f37 C |
741 | * special case: Some BIOSes report the PC BIOS |
742 | * area (640Kb -> 1Mb) as RAM even though it is not. | |
1b5576e6 YL |
743 | * take them out. |
744 | */ | |
09821ff1 | 745 | e820__range_remove(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_TYPE_RAM, 1); |
a9acc536 | 746 | |
f9748fa0 | 747 | e820__update_table(e820_table); |
1b5576e6 YL |
748 | } |
749 | ||
b422a309 YL |
750 | /* called before trim_bios_range() to spare extra sanitize */ |
751 | static void __init e820_add_kernel_range(void) | |
752 | { | |
753 | u64 start = __pa_symbol(_text); | |
754 | u64 size = __pa_symbol(_end) - start; | |
755 | ||
756 | /* | |
09821ff1 | 757 | * Complain if .text .data and .bss are not marked as E820_TYPE_RAM and |
b422a309 YL |
758 | * attempt to fix it by adding the range. We may have a confused BIOS, |
759 | * or the user may have used memmap=exactmap or memmap=xxM$yyM to | |
760 | * exclude kernel range. If we really are running on top non-RAM, | |
761 | * we will crash later anyways. | |
762 | */ | |
09821ff1 | 763 | if (e820__mapped_all(start, start + size, E820_TYPE_RAM)) |
b422a309 YL |
764 | return; |
765 | ||
09821ff1 IM |
766 | pr_warn(".text .data .bss are not marked as E820_TYPE_RAM!\n"); |
767 | e820__range_remove(start, size, E820_TYPE_RAM, 0); | |
768 | e820__range_add(start, size, E820_TYPE_RAM); | |
b422a309 YL |
769 | } |
770 | ||
a799c2bd | 771 | static void __init early_reserve_memory(void) |
95c96084 | 772 | { |
a799c2bd MR |
773 | /* |
774 | * Reserve the memory occupied by the kernel between _text and | |
775 | * __end_of_kernel_reserve symbols. Any kernel sections after the | |
776 | * __end_of_kernel_reserve symbol must be explicitly reserved with a | |
777 | * separate memblock_reserve() or they will be discarded. | |
778 | */ | |
779 | memblock_reserve(__pa_symbol(_text), | |
780 | (unsigned long)__end_of_kernel_reserve - (unsigned long)_text); | |
781 | ||
782 | /* | |
4c674481 MR |
783 | * The first 4Kb of memory is a BIOS owned area, but generally it is |
784 | * not listed as such in the E820 table. | |
785 | * | |
f1d4d47c MR |
786 | * Reserve the first 64K of memory since some BIOSes are known to |
787 | * corrupt low memory. After the real mode trampoline is allocated the | |
788 | * rest of the memory below 640k is reserved. | |
4c674481 MR |
789 | * |
790 | * In addition, make sure page 0 is always reserved because on | |
791 | * systems with L1TF its contents can be leaked to user processes. | |
a799c2bd | 792 | */ |
f1d4d47c | 793 | memblock_reserve(0, SZ_64K); |
a799c2bd MR |
794 | |
795 | early_reserve_initrd(); | |
796 | ||
a799c2bd MR |
797 | memblock_x86_reserve_range_setup_data(); |
798 | ||
799 | reserve_ibft_region(); | |
800 | reserve_bios_regions(); | |
f1d4d47c | 801 | trim_snb_memory(); |
95c96084 | 802 | } |
a799c2bd | 803 | |
f32360ef KC |
804 | /* |
805 | * Dump out kernel offset information on panic. | |
806 | */ | |
807 | static int | |
808 | dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p) | |
809 | { | |
78cac48c BP |
810 | if (kaslr_enabled()) { |
811 | pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n", | |
4545c898 | 812 | kaslr_offset(), |
78cac48c BP |
813 | __START_KERNEL, |
814 | __START_KERNEL_map, | |
815 | MODULES_VADDR-1); | |
816 | } else { | |
817 | pr_emerg("Kernel Offset: disabled\n"); | |
818 | } | |
f32360ef KC |
819 | |
820 | return 0; | |
821 | } | |
822 | ||
76ea0025 BP |
823 | void x86_configure_nx(void) |
824 | { | |
825 | if (boot_cpu_has(X86_FEATURE_NX)) | |
826 | __supported_pte_mask |= _PAGE_NX; | |
827 | else | |
828 | __supported_pte_mask &= ~_PAGE_NX; | |
829 | } | |
830 | ||
831 | static void __init x86_report_nx(void) | |
832 | { | |
833 | if (!boot_cpu_has(X86_FEATURE_NX)) { | |
834 | printk(KERN_NOTICE "Notice: NX (Execute Disable) protection " | |
835 | "missing in CPU!\n"); | |
836 | } else { | |
837 | #if defined(CONFIG_X86_64) || defined(CONFIG_X86_PAE) | |
838 | printk(KERN_INFO "NX (Execute Disable) protection: active\n"); | |
839 | #else | |
840 | /* 32bit non-PAE kernel, NX cannot be used */ | |
841 | printk(KERN_NOTICE "Notice: NX (Execute Disable) protection " | |
842 | "cannot be enabled: non-PAE kernel!\n"); | |
843 | #endif | |
844 | } | |
845 | } | |
846 | ||
1da177e4 LT |
847 | /* |
848 | * Determine if we were loaded by an EFI loader. If so, then we have also been | |
849 | * passed the efi memmap, systab, etc., so we should use these data structures | |
850 | * for initialization. Note, the efi init code path is determined by the | |
851 | * global efi_enabled. This allows the same kernel image to be used on existing | |
852 | * systems (with a traditional BIOS) as well as on EFI systems. | |
853 | */ | |
76934ed4 YL |
854 | /* |
855 | * setup_arch - architecture-specific boot-time initializations | |
856 | * | |
857 | * Note: On x86_64, fixmaps are ready for use even before this is called. | |
858 | */ | |
859 | ||
1da177e4 LT |
860 | void __init setup_arch(char **cmdline_p) |
861 | { | |
76934ed4 | 862 | #ifdef CONFIG_X86_32 |
1da177e4 | 863 | memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data)); |
b40827fa BP |
864 | |
865 | /* | |
866 | * copy kernel address range established so far and switch | |
867 | * to the proper swapper page table | |
868 | */ | |
869 | clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY, | |
870 | initial_page_table + KERNEL_PGD_BOUNDARY, | |
871 | KERNEL_PGD_PTRS); | |
872 | ||
873 | load_cr3(swapper_pg_dir); | |
2075244f BD |
874 | /* |
875 | * Note: Quark X1000 CPUs advertise PGE incorrectly and require | |
876 | * a cr3 based tlb flush, so the following __flush_tlb_all() | |
360db4ac | 877 | * will not flush anything because the CPU quirk which clears |
2075244f BD |
878 | * X86_FEATURE_PGE has not been invoked yet. Though due to the |
879 | * load_cr3() above the TLB has been flushed already. The | |
880 | * quirk is invoked before subsequent calls to __flush_tlb_all() | |
881 | * so proper operation is guaranteed. | |
882 | */ | |
b40827fa | 883 | __flush_tlb_all(); |
76934ed4 YL |
884 | #else |
885 | printk(KERN_INFO "Command line: %s\n", boot_command_line); | |
162434e7 | 886 | boot_cpu_data.x86_phys_bits = MAX_PHYSMEM_BITS; |
76934ed4 | 887 | #endif |
1da177e4 | 888 | |
9863c90f AK |
889 | /* |
890 | * If we have OLPC OFW, we might end up relocating the fixmap due to | |
891 | * reserve_top(), so do this before touching the ioremap area. | |
892 | */ | |
fd699c76 AS |
893 | olpc_ofw_detect(); |
894 | ||
433f8924 | 895 | idt_setup_early_traps(); |
9e882c92 | 896 | early_cpu_init(); |
8990cac6 | 897 | jump_label_init(); |
a945c834 | 898 | static_call_init(); |
1a98fd14 JF |
899 | early_ioremap_init(); |
900 | ||
fd699c76 AS |
901 | setup_olpc_ofw_pgd(); |
902 | ||
30c82645 PA |
903 | ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev); |
904 | screen_info = boot_params.screen_info; | |
905 | edid_info = boot_params.edid_info; | |
76934ed4 | 906 | #ifdef CONFIG_X86_32 |
30c82645 PA |
907 | apm_info.bios = boot_params.apm_bios_info; |
908 | ist_info = boot_params.ist_info; | |
76934ed4 YL |
909 | #endif |
910 | saved_video_mode = boot_params.hdr.vid_mode; | |
30c82645 | 911 | bootloader_type = boot_params.hdr.type_of_loader; |
5031296c PA |
912 | if ((bootloader_type >> 4) == 0xe) { |
913 | bootloader_type &= 0xf; | |
914 | bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4; | |
915 | } | |
916 | bootloader_version = bootloader_type & 0xf; | |
917 | bootloader_version |= boot_params.hdr.ext_loader_ver << 4; | |
1da177e4 LT |
918 | |
919 | #ifdef CONFIG_BLK_DEV_RAM | |
30c82645 | 920 | rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK; |
1da177e4 | 921 | #endif |
7465252e YL |
922 | #ifdef CONFIG_EFI |
923 | if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature, | |
9402973d | 924 | EFI32_LOADER_SIGNATURE, 4)) { |
3e909599 | 925 | set_bit(EFI_BOOT, &efi.flags); |
1adbfa35 | 926 | } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature, |
9402973d | 927 | EFI64_LOADER_SIGNATURE, 4)) { |
3e909599 MF |
928 | set_bit(EFI_BOOT, &efi.flags); |
929 | set_bit(EFI_64BIT, &efi.flags); | |
7465252e YL |
930 | } |
931 | #endif | |
932 | ||
42bbdb43 | 933 | x86_init.oem.arch_setup(); |
2215e69d | 934 | |
8aa83e63 JG |
935 | /* |
936 | * Do some memory reservations *before* memory is added to memblock, so | |
937 | * memblock allocations won't overwrite it. | |
938 | * | |
939 | * After this point, everything still needed from the boot loader or | |
940 | * firmware or kernel text should be early reserved or marked not RAM in | |
941 | * e820. All other memory is free game. | |
942 | * | |
943 | * This call needs to happen before e820__memory_setup() which calls the | |
944 | * xen_memory_setup() on Xen dom0 which relies on the fact that those | |
945 | * early reservations have happened already. | |
946 | */ | |
947 | early_reserve_memory(); | |
948 | ||
419afdf5 | 949 | iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1; |
103e2063 | 950 | e820__memory_setup(); |
28bb2237 YL |
951 | parse_setup_data(); |
952 | ||
1da177e4 LT |
953 | copy_edd(); |
954 | ||
30c82645 | 955 | if (!boot_params.hdr.root_flags) |
1da177e4 | 956 | root_mountflags &= ~MS_RDONLY; |
30120d72 | 957 | setup_initial_init_mm(_text, _etext, _edata, (void *)_brk_end); |
fe3d197f | 958 | |
4046d6e8 LT |
959 | code_resource.start = __pa_symbol(_text); |
960 | code_resource.end = __pa_symbol(_etext)-1; | |
a3299754 KC |
961 | rodata_resource.start = __pa_symbol(__start_rodata); |
962 | rodata_resource.end = __pa_symbol(__end_rodata)-1; | |
963 | data_resource.start = __pa_symbol(_sdata); | |
4046d6e8 LT |
964 | data_resource.end = __pa_symbol(_edata)-1; |
965 | bss_resource.start = __pa_symbol(__bss_start); | |
966 | bss_resource.end = __pa_symbol(__bss_stop)-1; | |
967 | ||
fbe61839 BP |
968 | #ifdef CONFIG_CMDLINE_BOOL |
969 | #ifdef CONFIG_CMDLINE_OVERRIDE | |
8a33d96b | 970 | strscpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE); |
fbe61839 BP |
971 | #else |
972 | if (builtin_cmdline[0]) { | |
973 | /* append boot loader cmdline to builtin */ | |
974 | strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE); | |
975 | strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE); | |
8a33d96b | 976 | strscpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE); |
fbe61839 BP |
977 | } |
978 | #endif | |
979 | #endif | |
980 | ||
8a33d96b | 981 | strscpy(command_line, boot_command_line, COMMAND_LINE_SIZE); |
fbe61839 BP |
982 | *cmdline_p = command_line; |
983 | ||
984 | /* | |
985 | * x86_configure_nx() is called before parse_early_param() to detect | |
986 | * whether hardware doesn't support NX (so that the early EHCI debug | |
76ea0025 | 987 | * console setup can safely call set_fixmap()). |
fbe61839 BP |
988 | */ |
989 | x86_configure_nx(); | |
990 | ||
991 | parse_early_param(); | |
992 | ||
2f5b3514 MR |
993 | if (efi_enabled(EFI_BOOT)) |
994 | efi_memblock_x86_reserve_range(); | |
995 | ||
39fa104d RA |
996 | #ifdef CONFIG_MEMORY_HOTPLUG |
997 | /* | |
998 | * Memory used by the kernel cannot be hot-removed because Linux | |
999 | * cannot migrate the kernel pages. When memory hotplug is | |
1000 | * enabled, we should prevent memblock from allocating memory | |
1001 | * for the kernel. | |
1002 | * | |
1003 | * ACPI SRAT records all hotpluggable memory ranges. But before | |
1004 | * SRAT is parsed, we don't know about it. | |
1005 | * | |
1006 | * The kernel image is loaded into memory at very early time. We | |
1007 | * cannot prevent this anyway. So on NUMA system, we set any | |
1008 | * node the kernel resides in as un-hotpluggable. | |
1009 | * | |
1010 | * Since on modern servers, one node could have double-digit | |
1011 | * gigabytes memory, we can assume the memory around the kernel | |
1012 | * image is also un-hotpluggable. So before SRAT is parsed, just | |
1013 | * allocate memory near the kernel image to try the best to keep | |
1014 | * the kernel away from hotpluggable memory. | |
1015 | */ | |
1016 | if (movable_node_is_enabled()) | |
1017 | memblock_set_bottom_up(true); | |
1018 | #endif | |
1019 | ||
4b0f3b81 | 1020 | x86_report_nx(); |
0ad5bce7 | 1021 | |
76934ed4 | 1022 | if (acpi_mps_check()) { |
3eb11edc | 1023 | #ifdef CONFIG_X86_LOCAL_APIC |
76934ed4 | 1024 | disable_apic = 1; |
3eb11edc | 1025 | #endif |
988781dc | 1026 | setup_clear_cpu_cap(X86_FEATURE_APIC); |
3c999f14 YL |
1027 | } |
1028 | ||
1a127034 | 1029 | e820__reserve_setup_data(); |
9641bdaf | 1030 | e820__finish_early_params(); |
1a3f239d | 1031 | |
83e68189 | 1032 | if (efi_enabled(EFI_BOOT)) |
ff0c0874 BM |
1033 | efi_init(); |
1034 | ||
0fca0812 | 1035 | dmi_setup(); |
2216d199 | 1036 | |
88b094fb AK |
1037 | /* |
1038 | * VMware detection requires dmi to be available, so this | |
0fca0812 | 1039 | * needs to be done after dmi_setup(), for the boot CPU. |
88b094fb | 1040 | */ |
2d826404 | 1041 | init_hypervisor_platform(); |
88b094fb | 1042 | |
cf7a63ef | 1043 | tsc_early_init(); |
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); | |
a3299754 | 1048 | insert_resource(&iomem_resource, &rodata_resource); |
4046d6e8 LT |
1049 | insert_resource(&iomem_resource, &data_resource); |
1050 | insert_resource(&iomem_resource, &bss_resource); | |
1051 | ||
b422a309 | 1052 | e820_add_kernel_range(); |
1b5576e6 | 1053 | trim_bios_range(); |
76934ed4 | 1054 | #ifdef CONFIG_X86_32 |
cc9f7a0c | 1055 | if (ppro_with_ram_bug()) { |
09821ff1 IM |
1056 | e820__range_update(0x70000000ULL, 0x40000ULL, E820_TYPE_RAM, |
1057 | E820_TYPE_RESERVED); | |
f9748fa0 | 1058 | e820__update_table(e820_table); |
cc9f7a0c | 1059 | printk(KERN_INFO "fixed physical RAM map:\n"); |
be0c3f0f | 1060 | e820__print_table("bad_ppro"); |
cc9f7a0c | 1061 | } |
76934ed4 YL |
1062 | #else |
1063 | early_gart_iommu_check(); | |
1064 | #endif | |
cc9f7a0c | 1065 | |
7b2a0a6c YL |
1066 | /* |
1067 | * partially used pages are not usable - thus | |
1068 | * we are rounding upwards: | |
1069 | */ | |
0c6fc11a | 1070 | max_pfn = e820__end_of_ram_pfn(); |
7b2a0a6c | 1071 | |
093af8d7 | 1072 | /* update e820 for memory not covered by WB MTRRs */ |
adfe7512 | 1073 | cache_bp_init(); |
2dc807b3 | 1074 | if (mtrr_trim_uncached_memory(max_pfn)) |
0c6fc11a | 1075 | max_pfn = e820__end_of_ram_pfn(); |
76c32418 | 1076 | |
8dd33030 IM |
1077 | max_possible_pfn = max_pfn; |
1078 | ||
c7d2361f TG |
1079 | /* |
1080 | * Define random base addresses for memory sections after max_pfn is | |
1081 | * defined and before each memory section base is used. | |
1082 | */ | |
1083 | kernel_randomize_memory(); | |
1084 | ||
76934ed4 | 1085 | #ifdef CONFIG_X86_32 |
4e29684c | 1086 | /* max_low_pfn get updated here */ |
2ec65f8b | 1087 | find_low_pfn_range(); |
76934ed4 | 1088 | #else |
06cd9a7d | 1089 | check_x2apic(); |
76934ed4 YL |
1090 | |
1091 | /* How many end-of-memory variables you have, grandma! */ | |
1092 | /* need this before calling reserve_initrd */ | |
f361a450 | 1093 | if (max_pfn > (1UL<<(32 - PAGE_SHIFT))) |
0c6fc11a | 1094 | max_low_pfn = e820__end_of_low_ram_pfn(); |
f361a450 YL |
1095 | else |
1096 | max_low_pfn = max_pfn; | |
1097 | ||
76934ed4 | 1098 | high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1; |
5394f80f JF |
1099 | #endif |
1100 | ||
893f38d1 YL |
1101 | /* |
1102 | * Find and reserve possible boot-time SMP configuration: | |
1103 | */ | |
1104 | find_smp_config(); | |
1105 | ||
8d57470d YL |
1106 | early_alloc_pgt_buf(); |
1107 | ||
72d7c3b3 | 1108 | /* |
4918e228 | 1109 | * Need to conclude brk, before e820__memblock_setup() |
81519f77 C |
1110 | * it could use memblock_find_in_range, could overlap with |
1111 | * brk area. | |
72d7c3b3 YL |
1112 | */ |
1113 | reserve_brk(); | |
1114 | ||
e5f15b45 YL |
1115 | cleanup_highmap(); |
1116 | ||
2449f343 | 1117 | memblock_set_current_limit(ISA_END_ADDRESS); |
4918e228 | 1118 | e820__memblock_setup(); |
72d7c3b3 | 1119 | |
e998879d AK |
1120 | /* |
1121 | * Needs to run after memblock setup because it needs the physical | |
1122 | * memory size. | |
1123 | */ | |
1124 | sev_setup_arch(); | |
1125 | ||
6950e31b DW |
1126 | efi_fake_memmap(); |
1127 | efi_find_mirror(); | |
1128 | efi_esrt_init(); | |
58c90902 | 1129 | efi_mokvar_table_init(); |
007b7560 | 1130 | |
6950e31b DW |
1131 | /* |
1132 | * The EFI specification says that boot service code won't be | |
1133 | * called after ExitBootServices(). This is, in fact, a lie. | |
1134 | */ | |
1135 | efi_reserve_boot_services(); | |
916f676f | 1136 | |
72d7c3b3 | 1137 | /* preallocate 4k for mptable mpc */ |
5da217ca | 1138 | e820__memblock_alloc_reserved_mpc_new(); |
72d7c3b3 YL |
1139 | |
1140 | #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION | |
1141 | setup_bios_corruption_check(); | |
1142 | #endif | |
1143 | ||
10054230 | 1144 | #ifdef CONFIG_X86_32 |
365811d6 BH |
1145 | printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n", |
1146 | (max_pfn_mapped<<PAGE_SHIFT) - 1); | |
10054230 | 1147 | #endif |
72d7c3b3 | 1148 | |
c361e5d4 | 1149 | /* |
ec35d1d9 BP |
1150 | * Find free memory for the real mode trampoline and place it there. If |
1151 | * there is not enough free memory under 1M, on EFI-enabled systems | |
1152 | * there will be additional attempt to reclaim the memory for the real | |
1153 | * mode trampoline at efi_free_boot_services(). | |
f1d4d47c | 1154 | * |
ec35d1d9 BP |
1155 | * Unconditionally reserve the entire first 1M of RAM because BIOSes |
1156 | * are known to corrupt low memory and several hundred kilobytes are not | |
1157 | * worth complex detection what memory gets clobbered. Windows does the | |
1158 | * same thing for very similar reasons. | |
1159 | * | |
1160 | * Moreover, on machines with SandyBridge graphics or in setups that use | |
1161 | * crashkernel the entire 1M is reserved anyway. | |
c361e5d4 | 1162 | */ |
f1e52500 | 1163 | x86_platform.realmode_reserve(); |
a9acc536 | 1164 | |
22ddfcaa | 1165 | init_mem_mapping(); |
1bbbbe77 | 1166 | |
433f8924 | 1167 | idt_setup_early_pf(); |
1bbbbe77 | 1168 | |
18bc7bd5 AL |
1169 | /* |
1170 | * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features) | |
1171 | * with the current CR4 value. This may not be necessary, but | |
1172 | * auditing all the early-boot CR4 manipulation would be needed to | |
1173 | * rule it out. | |
c7ad5ad2 AL |
1174 | * |
1175 | * Mask off features that don't work outside long mode (just | |
1176 | * PCIDE for now). | |
18bc7bd5 | 1177 | */ |
c7ad5ad2 | 1178 | mmu_cr4_features = __read_cr4() & ~X86_CR4_PCIDE; |
18bc7bd5 | 1179 | |
4ce7a869 | 1180 | memblock_set_current_limit(get_max_mapped()); |
4e29684c | 1181 | |
e7b37895 YL |
1182 | /* |
1183 | * NOTE: On x86-32, only from this point on, fixmaps are ready for use. | |
1184 | */ | |
1185 | ||
1186 | #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT | |
1187 | if (init_ohci1394_dma_early) | |
1188 | init_ohci1394_dma_on_all_controllers(); | |
1189 | #endif | |
162a7e75 MT |
1190 | /* Allocate bigger log buffer */ |
1191 | setup_log_buf(1); | |
e7b37895 | 1192 | |
9661b332 DH |
1193 | if (efi_enabled(EFI_BOOT)) { |
1194 | switch (boot_params.secure_boot) { | |
1195 | case efi_secureboot_mode_disabled: | |
1196 | pr_info("Secure boot disabled\n"); | |
1197 | break; | |
1198 | case efi_secureboot_mode_enabled: | |
1199 | pr_info("Secure boot enabled\n"); | |
1200 | break; | |
1201 | default: | |
1202 | pr_info("Secure boot could not be determined\n"); | |
1203 | break; | |
1204 | } | |
1205 | } | |
1206 | ||
2ec65f8b YL |
1207 | reserve_initrd(); |
1208 | ||
da3d3f98 | 1209 | acpi_table_upgrade(); |
6998a880 RW |
1210 | /* Look for ACPI tables and reserve memory occupied by them. */ |
1211 | acpi_boot_table_init(); | |
53aac44c | 1212 | |
76934ed4 | 1213 | vsmp_init(); |
76934ed4 | 1214 | |
1c6e5503 YL |
1215 | io_delay_init(); |
1216 | ||
630b3aff LW |
1217 | early_platform_quirks(); |
1218 | ||
20e6926d YL |
1219 | early_acpi_boot_init(); |
1220 | ||
d8fc3afc | 1221 | initmem_init(); |
3c325f82 | 1222 | dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT); |
fa591c4a | 1223 | |
cf11e85f RG |
1224 | if (boot_cpu_has(X86_FEATURE_GBPAGES)) |
1225 | hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT); | |
1226 | ||
fa591c4a TC |
1227 | /* |
1228 | * Reserve memory for crash kernel after SRAT is parsed so that it | |
1229 | * won't consume hotpluggable memory. | |
1230 | */ | |
1231 | reserve_crashkernel(); | |
1232 | ||
6f2a7536 | 1233 | memblock_find_dma_reserve(); |
91467bdf | 1234 | |
ccb64941 BO |
1235 | if (!early_xdbc_setup_hardware()) |
1236 | early_xdbc_register_console(); | |
1237 | ||
7737b215 | 1238 | x86_init.paging.pagetable_init(); |
f212ec4b | 1239 | |
ef7f0d6a AR |
1240 | kasan_init(); |
1241 | ||
d2b6dc61 | 1242 | /* |
945fd17a TG |
1243 | * Sync back kernel address range. |
1244 | * | |
1245 | * FIXME: Can the later sync in setup_cpu_entry_areas() replace | |
1246 | * this call? | |
d2b6dc61 | 1247 | */ |
945fd17a | 1248 | sync_initial_page_table(); |
d2b6dc61 | 1249 | |
31625340 JC |
1250 | tboot_probe(); |
1251 | ||
76934ed4 | 1252 | map_vsyscall(); |
76934ed4 | 1253 | |
1a3f239d | 1254 | generic_apic_probe(); |
1da177e4 | 1255 | |
54ef3400 | 1256 | early_quirks(); |
d44647b0 | 1257 | |
295deae4 YL |
1258 | /* |
1259 | * Read APIC and some other early information from ACPI tables. | |
1260 | */ | |
1da177e4 | 1261 | acpi_boot_init(); |
a906fdaa | 1262 | x86_dtb_init(); |
04606618 | 1263 | |
295deae4 YL |
1264 | /* |
1265 | * get boot-time SMP configuration: | |
1266 | */ | |
a91bf718 | 1267 | get_smp_config(); |
76934ed4 | 1268 | |
1e90a13d TG |
1269 | /* |
1270 | * Systems w/o ACPI and mptables might not have it mapped the local | |
1271 | * APIC yet, but prefill_possible_map() might need to access it. | |
1272 | */ | |
1273 | init_apic_mappings(); | |
1274 | ||
329513a3 | 1275 | prefill_possible_map(); |
301e6190 | 1276 | |
5f4765f9 | 1277 | init_cpu_to_node(); |
73bf7382 | 1278 | init_gi_nodes(); |
5f4765f9 | 1279 | |
ca1b8862 | 1280 | io_apic_init_mappings(); |
9d6a4d08 | 1281 | |
f3614646 | 1282 | x86_init.hyper.guest_late_init(); |
1da177e4 | 1283 | |
1506c8dc | 1284 | e820__reserve_resources(); |
cc55f753 | 1285 | e820__register_nosave_regions(max_pfn); |
1da177e4 | 1286 | |
8fee697d | 1287 | x86_init.resources.reserve_resources(); |
41c094fd | 1288 | |
2df908ba | 1289 | e820__setup_pci_gap(); |
41c094fd | 1290 | |
1da177e4 LT |
1291 | #ifdef CONFIG_VT |
1292 | #if defined(CONFIG_VGA_CONSOLE) | |
83e68189 | 1293 | if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY)) |
1da177e4 | 1294 | conswitchp = &vga_con; |
1da177e4 LT |
1295 | #endif |
1296 | #endif | |
6f30c1ac | 1297 | x86_init.oem.banner(); |
a2202aa2 | 1298 | |
6b617e22 FT |
1299 | x86_init.timers.wallclock_init(); |
1300 | ||
9a90ed06 BP |
1301 | /* |
1302 | * This needs to run before setup_local_APIC() which soft-disables the | |
1303 | * local APIC temporarily and that masks the thermal LVT interrupt, | |
1304 | * leading to softlockups on machines which have configured SMI | |
1305 | * interrupt delivery. | |
1306 | */ | |
1307 | therm_lvt_init(); | |
1308 | ||
a2202aa2 | 1309 | mcheck_init(); |
f49aa448 | 1310 | |
b3c869d3 | 1311 | register_refined_jiffies(CLOCK_TICK_RATE); |
5189c2a7 OJ |
1312 | |
1313 | #ifdef CONFIG_EFI | |
a5d90c92 BP |
1314 | if (efi_enabled(EFI_BOOT)) |
1315 | efi_apply_memmap_quirks(); | |
5189c2a7 | 1316 | #endif |
ee9f8fce JP |
1317 | |
1318 | unwind_init(); | |
1da177e4 | 1319 | } |
5649b7c3 | 1320 | |
9be1b56a IM |
1321 | #ifdef CONFIG_X86_32 |
1322 | ||
8fee697d TG |
1323 | static struct resource video_ram_resource = { |
1324 | .name = "Video RAM area", | |
1325 | .start = 0xa0000, | |
1326 | .end = 0xbffff, | |
1327 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
9be1b56a IM |
1328 | }; |
1329 | ||
8fee697d | 1330 | void __init i386_reserve_resources(void) |
9be1b56a | 1331 | { |
8fee697d TG |
1332 | request_resource(&iomem_resource, &video_ram_resource); |
1333 | reserve_standard_io_resources(); | |
9be1b56a IM |
1334 | } |
1335 | ||
9be1b56a | 1336 | #endif /* CONFIG_X86_32 */ |
f32360ef KC |
1337 | |
1338 | static struct notifier_block kernel_offset_notifier = { | |
1339 | .notifier_call = dump_kernel_offset | |
1340 | }; | |
1341 | ||
1342 | static int __init register_kernel_offset_dumper(void) | |
1343 | { | |
1344 | atomic_notifier_chain_register(&panic_notifier_list, | |
1345 | &kernel_offset_notifier); | |
1346 | return 0; | |
1347 | } | |
1348 | __initcall(register_kernel_offset_dumper); |