Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1da177e4 LT |
2 | * Copyright (C) 1995 Linus Torvalds |
3 | * | |
4 | * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999 | |
5 | * | |
6 | * Memory region support | |
7 | * David Parsons <orc@pell.chi.il.us>, July-August 1999 | |
8 | * | |
9 | * Added E820 sanitization routine (removes overlapping memory regions); | |
10 | * Brian Moyle <bmoyle@mvista.com>, February 2001 | |
11 | * | |
12 | * Moved CPU detection code to cpu/${cpu}.c | |
13 | * Patrick Mochel <mochel@osdl.org>, March 2002 | |
14 | * | |
15 | * Provisions for empty E820 memory regions (reported by certain BIOSes). | |
16 | * Alex Achenbach <xela@slit.de>, December 2002. | |
17 | * | |
18 | */ | |
19 | ||
20 | /* | |
21 | * This file handles the architecture-dependent parts of initialization | |
22 | */ | |
23 | ||
24 | #include <linux/sched.h> | |
25 | #include <linux/mm.h> | |
05b79bdc | 26 | #include <linux/mmzone.h> |
894673ee | 27 | #include <linux/screen_info.h> |
1da177e4 LT |
28 | #include <linux/ioport.h> |
29 | #include <linux/acpi.h> | |
30 | #include <linux/apm_bios.h> | |
31 | #include <linux/initrd.h> | |
32 | #include <linux/bootmem.h> | |
33 | #include <linux/seq_file.h> | |
34 | #include <linux/console.h> | |
35 | #include <linux/mca.h> | |
36 | #include <linux/root_dev.h> | |
37 | #include <linux/highmem.h> | |
38 | #include <linux/module.h> | |
39 | #include <linux/efi.h> | |
40 | #include <linux/init.h> | |
41 | #include <linux/edd.h> | |
138fe4e0 | 42 | #include <linux/iscsi_ibft.h> |
1da177e4 | 43 | #include <linux/nodemask.h> |
1bc3b91a | 44 | #include <linux/kexec.h> |
2030eae5 | 45 | #include <linux/crash_dump.h> |
e9928674 | 46 | #include <linux/dmi.h> |
22a9835c | 47 | #include <linux/pfn.h> |
376ff035 | 48 | #include <linux/pci.h> |
f212ec4b | 49 | #include <linux/init_ohci1394_dma.h> |
1bc3b91a | 50 | |
1da177e4 | 51 | #include <video/edid.h> |
1bc3b91a | 52 | |
093af8d7 | 53 | #include <asm/mtrr.h> |
9635b47d | 54 | #include <asm/apic.h> |
1da177e4 LT |
55 | #include <asm/e820.h> |
56 | #include <asm/mpspec.h> | |
91023300 | 57 | #include <asm/mmzone.h> |
1da177e4 LT |
58 | #include <asm/setup.h> |
59 | #include <asm/arch_hooks.h> | |
60 | #include <asm/sections.h> | |
61 | #include <asm/io_apic.h> | |
62 | #include <asm/ist.h> | |
63 | #include <asm/io.h> | |
7ce0bcfd | 64 | #include <asm/vmi.h> |
e75eac33 | 65 | #include <setup_arch.h> |
ce3fe6b2 | 66 | #include <asm/bios_ebda.h> |
00bf4098 | 67 | #include <asm/cacheflush.h> |
2fde61fd | 68 | #include <asm/processor.h> |
1da177e4 LT |
69 | |
70 | /* This value is set up by the early boot code to point to the value | |
71 | immediately after the boot time page tables. It contains a *physical* | |
72 | address, and must not be in the .bss segment! */ | |
73 | unsigned long init_pg_tables_end __initdata = ~0UL; | |
74 | ||
1da177e4 LT |
75 | /* |
76 | * Machine setup.. | |
77 | */ | |
c9cce83d BW |
78 | static struct resource data_resource = { |
79 | .name = "Kernel data", | |
80 | .start = 0, | |
81 | .end = 0, | |
82 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
83 | }; | |
84 | ||
85 | static struct resource code_resource = { | |
86 | .name = "Kernel code", | |
87 | .start = 0, | |
88 | .end = 0, | |
89 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
90 | }; | |
91 | ||
92 | static struct resource bss_resource = { | |
93 | .name = "Kernel bss", | |
94 | .start = 0, | |
95 | .end = 0, | |
96 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
97 | }; | |
98 | ||
99 | static struct resource video_ram_resource = { | |
100 | .name = "Video RAM area", | |
101 | .start = 0xa0000, | |
102 | .end = 0xbffff, | |
103 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
104 | }; | |
105 | ||
106 | static struct resource standard_io_resources[] = { { | |
107 | .name = "dma1", | |
108 | .start = 0x0000, | |
109 | .end = 0x001f, | |
110 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
111 | }, { | |
112 | .name = "pic1", | |
113 | .start = 0x0020, | |
114 | .end = 0x0021, | |
115 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
116 | }, { | |
117 | .name = "timer0", | |
118 | .start = 0x0040, | |
119 | .end = 0x0043, | |
120 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
121 | }, { | |
122 | .name = "timer1", | |
123 | .start = 0x0050, | |
124 | .end = 0x0053, | |
125 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
126 | }, { | |
127 | .name = "keyboard", | |
128 | .start = 0x0060, | |
129 | .end = 0x006f, | |
130 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
131 | }, { | |
132 | .name = "dma page reg", | |
133 | .start = 0x0080, | |
134 | .end = 0x008f, | |
135 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
136 | }, { | |
137 | .name = "pic2", | |
138 | .start = 0x00a0, | |
139 | .end = 0x00a1, | |
140 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
141 | }, { | |
142 | .name = "dma2", | |
143 | .start = 0x00c0, | |
144 | .end = 0x00df, | |
145 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
146 | }, { | |
147 | .name = "fpu", | |
148 | .start = 0x00f0, | |
149 | .end = 0x00ff, | |
150 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
151 | } }; | |
1da177e4 LT |
152 | |
153 | /* cpu data as detected by the assembly code in head.S */ | |
4a5d107a | 154 | struct cpuinfo_x86 new_cpu_data __cpuinitdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 }; |
1da177e4 | 155 | /* common cpu data for all cpus */ |
c3d8c141 | 156 | struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 }; |
129f6946 | 157 | EXPORT_SYMBOL(boot_cpu_data); |
1da177e4 | 158 | |
0c254e38 AS |
159 | unsigned int def_to_bigsmp; |
160 | ||
551889a6 | 161 | #ifndef CONFIG_X86_PAE |
1da177e4 | 162 | unsigned long mmu_cr4_features; |
551889a6 IC |
163 | #else |
164 | unsigned long mmu_cr4_features = X86_CR4_PAE; | |
165 | #endif | |
1da177e4 | 166 | |
1da177e4 LT |
167 | /* for MCA, but anyone else can use it if they want */ |
168 | unsigned int machine_id; | |
169 | unsigned int machine_submodel_id; | |
170 | unsigned int BIOS_revision; | |
1da177e4 | 171 | |
1da177e4 LT |
172 | /* Boot loader ID as an integer, for the benefit of proc_dointvec */ |
173 | int bootloader_type; | |
174 | ||
175 | /* user-defined highmem size */ | |
176 | static unsigned int highmem_pages = -1; | |
177 | ||
178 | /* | |
179 | * Setup options | |
180 | */ | |
1da177e4 | 181 | struct screen_info screen_info; |
129f6946 | 182 | EXPORT_SYMBOL(screen_info); |
1da177e4 | 183 | struct apm_info apm_info; |
129f6946 | 184 | EXPORT_SYMBOL(apm_info); |
1da177e4 | 185 | struct edid_info edid_info; |
5e518d76 | 186 | EXPORT_SYMBOL_GPL(edid_info); |
1da177e4 | 187 | struct ist_info ist_info; |
129f6946 AD |
188 | #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \ |
189 | defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE) | |
190 | EXPORT_SYMBOL(ist_info); | |
191 | #endif | |
1da177e4 LT |
192 | |
193 | extern void early_cpu_init(void); | |
1da177e4 LT |
194 | extern int root_mountflags; |
195 | ||
e44b7b75 | 196 | unsigned long saved_video_mode; |
1da177e4 | 197 | |
cf8fa920 | 198 | #define RAMDISK_IMAGE_START_MASK 0x07FF |
1da177e4 | 199 | #define RAMDISK_PROMPT_FLAG 0x8000 |
cf8fa920 | 200 | #define RAMDISK_LOAD_FLAG 0x4000 |
1da177e4 | 201 | |
4e498b66 | 202 | static char __initdata command_line[COMMAND_LINE_SIZE]; |
1da177e4 | 203 | |
6d7d7433 | 204 | #ifndef CONFIG_DEBUG_BOOT_PARAMS |
48c7ae67 | 205 | struct boot_params __initdata boot_params; |
6d7d7433 HY |
206 | #else |
207 | struct boot_params boot_params; | |
208 | #endif | |
1da177e4 | 209 | |
1da177e4 LT |
210 | #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE) |
211 | struct edd edd; | |
212 | #ifdef CONFIG_EDD_MODULE | |
213 | EXPORT_SYMBOL(edd); | |
214 | #endif | |
215 | /** | |
216 | * copy_edd() - Copy the BIOS EDD information | |
217 | * from boot_params into a safe place. | |
218 | * | |
219 | */ | |
220 | static inline void copy_edd(void) | |
221 | { | |
30c82645 PA |
222 | memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer, |
223 | sizeof(edd.mbr_signature)); | |
224 | memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info)); | |
225 | edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries; | |
226 | edd.edd_info_nr = boot_params.eddbuf_entries; | |
1da177e4 LT |
227 | } |
228 | #else | |
229 | static inline void copy_edd(void) | |
230 | { | |
231 | } | |
232 | #endif | |
233 | ||
537d9160 | 234 | int __initdata user_defined_memmap; |
1da177e4 | 235 | |
1a3f239d RR |
236 | /* |
237 | * "mem=nopentium" disables the 4MB page tables. | |
238 | * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM | |
239 | * to <mem>, overriding the bios size. | |
240 | * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from | |
241 | * <start> to <start>+<mem>, overriding the bios size. | |
242 | * | |
243 | * HPA tells me bootloaders need to parse mem=, so no new | |
244 | * option should be mem= [also see Documentation/i386/boot.txt] | |
245 | */ | |
246 | static int __init parse_mem(char *arg) | |
247 | { | |
248 | if (!arg) | |
249 | return -EINVAL; | |
1da177e4 | 250 | |
1a3f239d | 251 | if (strcmp(arg, "nopentium") == 0) { |
13530257 | 252 | setup_clear_cpu_cap(X86_FEATURE_PSE); |
1a3f239d RR |
253 | } else { |
254 | /* If the user specifies memory size, we | |
255 | * limit the BIOS-provided memory map to | |
256 | * that size. exactmap can be used to specify | |
257 | * the exact map. mem=number can be used to | |
258 | * trim the existing memory map. | |
1da177e4 | 259 | */ |
1a3f239d | 260 | unsigned long long mem_size; |
cf8fa920 | 261 | |
1a3f239d RR |
262 | mem_size = memparse(arg, &arg); |
263 | limit_regions(mem_size); | |
264 | user_defined_memmap = 1; | |
265 | } | |
266 | return 0; | |
267 | } | |
268 | early_param("mem", parse_mem); | |
1da177e4 | 269 | |
1a3f239d RR |
270 | #ifdef CONFIG_PROC_VMCORE |
271 | /* elfcorehdr= specifies the location of elf core header | |
272 | * stored by the crashed kernel. | |
273 | */ | |
274 | static int __init parse_elfcorehdr(char *arg) | |
275 | { | |
276 | if (!arg) | |
277 | return -EINVAL; | |
66759a01 | 278 | |
1a3f239d RR |
279 | elfcorehdr_addr = memparse(arg, &arg); |
280 | return 0; | |
281 | } | |
282 | early_param("elfcorehdr", parse_elfcorehdr); | |
283 | #endif /* CONFIG_PROC_VMCORE */ | |
1da177e4 | 284 | |
1a3f239d RR |
285 | /* |
286 | * highmem=size forces highmem to be exactly 'size' bytes. | |
287 | * This works even on boxes that have no highmem otherwise. | |
288 | * This also works to reduce highmem size on bigger boxes. | |
289 | */ | |
290 | static int __init parse_highmem(char *arg) | |
291 | { | |
292 | if (!arg) | |
293 | return -EINVAL; | |
9635b47d | 294 | |
1a3f239d RR |
295 | highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT; |
296 | return 0; | |
297 | } | |
298 | early_param("highmem", parse_highmem); | |
9635b47d | 299 | |
1a3f239d RR |
300 | /* |
301 | * vmalloc=size forces the vmalloc area to be exactly 'size' | |
302 | * bytes. This can be used to increase (or decrease) the | |
303 | * vmalloc area - the default is 128m. | |
304 | */ | |
305 | static int __init parse_vmalloc(char *arg) | |
306 | { | |
307 | if (!arg) | |
308 | return -EINVAL; | |
1bc3b91a | 309 | |
1a3f239d RR |
310 | __VMALLOC_RESERVE = memparse(arg, &arg); |
311 | return 0; | |
1da177e4 | 312 | } |
1a3f239d | 313 | early_param("vmalloc", parse_vmalloc); |
1da177e4 | 314 | |
461a9aff ZA |
315 | /* |
316 | * reservetop=size reserves a hole at the top of the kernel address space which | |
317 | * a hypervisor can load into later. Needed for dynamically loaded hypervisors, | |
318 | * so relocating the fixmap can be done before paging initialization. | |
319 | */ | |
320 | static int __init parse_reservetop(char *arg) | |
321 | { | |
322 | unsigned long address; | |
323 | ||
324 | if (!arg) | |
325 | return -EINVAL; | |
326 | ||
327 | address = memparse(arg, &arg); | |
328 | reserve_top_address(address); | |
329 | return 0; | |
330 | } | |
331 | early_param("reservetop", parse_reservetop); | |
332 | ||
1da177e4 LT |
333 | /* |
334 | * Determine low and high memory ranges: | |
335 | */ | |
336 | unsigned long __init find_max_low_pfn(void) | |
337 | { | |
338 | unsigned long max_low_pfn; | |
339 | ||
340 | max_low_pfn = max_pfn; | |
341 | if (max_low_pfn > MAXMEM_PFN) { | |
342 | if (highmem_pages == -1) | |
343 | highmem_pages = max_pfn - MAXMEM_PFN; | |
344 | if (highmem_pages + MAXMEM_PFN < max_pfn) | |
345 | max_pfn = MAXMEM_PFN + highmem_pages; | |
346 | if (highmem_pages + MAXMEM_PFN > max_pfn) { | |
347 | printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn - MAXMEM_PFN), pages_to_mb(highmem_pages)); | |
348 | highmem_pages = 0; | |
349 | } | |
350 | max_low_pfn = MAXMEM_PFN; | |
351 | #ifndef CONFIG_HIGHMEM | |
352 | /* Maximum memory usable is what is directly addressable */ | |
353 | printk(KERN_WARNING "Warning only %ldMB will be used.\n", | |
354 | MAXMEM>>20); | |
355 | if (max_pfn > MAX_NONPAE_PFN) | |
c673f1a9 | 356 | printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n"); |
1da177e4 LT |
357 | else |
358 | printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n"); | |
359 | max_pfn = MAXMEM_PFN; | |
360 | #else /* !CONFIG_HIGHMEM */ | |
c673f1a9 | 361 | #ifndef CONFIG_HIGHMEM64G |
1da177e4 LT |
362 | if (max_pfn > MAX_NONPAE_PFN) { |
363 | max_pfn = MAX_NONPAE_PFN; | |
364 | printk(KERN_WARNING "Warning only 4GB will be used.\n"); | |
c673f1a9 | 365 | printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n"); |
1da177e4 | 366 | } |
c673f1a9 | 367 | #endif /* !CONFIG_HIGHMEM64G */ |
1da177e4 LT |
368 | #endif /* !CONFIG_HIGHMEM */ |
369 | } else { | |
370 | if (highmem_pages == -1) | |
371 | highmem_pages = 0; | |
372 | #ifdef CONFIG_HIGHMEM | |
373 | if (highmem_pages >= max_pfn) { | |
374 | printk(KERN_ERR "highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages), pages_to_mb(max_pfn)); | |
375 | highmem_pages = 0; | |
376 | } | |
377 | if (highmem_pages) { | |
378 | if (max_low_pfn-highmem_pages < 64*1024*1024/PAGE_SIZE){ | |
379 | printk(KERN_ERR "highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages)); | |
380 | highmem_pages = 0; | |
381 | } | |
382 | max_low_pfn -= highmem_pages; | |
383 | } | |
384 | #else | |
385 | if (highmem_pages) | |
386 | printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n"); | |
387 | #endif | |
388 | } | |
389 | return max_low_pfn; | |
390 | } | |
391 | ||
dedd04be AH |
392 | #define BIOS_LOWMEM_KILOBYTES 0x413 |
393 | ||
1da177e4 | 394 | /* |
dedd04be AH |
395 | * The BIOS places the EBDA/XBDA at the top of conventional |
396 | * memory, and usually decreases the reported amount of | |
397 | * conventional memory (int 0x12) too. This also contains a | |
398 | * workaround for Dell systems that neglect to reserve EBDA. | |
399 | * The same workaround also avoids a problem with the AMD768MPX | |
400 | * chipset: reserve a page before VGA to prevent PCI prefetch | |
401 | * into it (errata #56). Usually the page is reserved anyways, | |
402 | * unless you have no PS/2 mouse plugged in. | |
1da177e4 LT |
403 | */ |
404 | static void __init reserve_ebda_region(void) | |
405 | { | |
dedd04be AH |
406 | unsigned int lowmem, ebda_addr; |
407 | ||
408 | /* To determine the position of the EBDA and the */ | |
409 | /* end of conventional memory, we need to look at */ | |
410 | /* the BIOS data area. In a paravirtual environment */ | |
411 | /* that area is absent. We'll just have to assume */ | |
412 | /* that the paravirt case can handle memory setup */ | |
413 | /* correctly, without our help. */ | |
2fde61fd | 414 | if (paravirt_enabled()) |
dedd04be | 415 | return; |
dedd04be AH |
416 | |
417 | /* end of low (conventional) memory */ | |
418 | lowmem = *(unsigned short *)__va(BIOS_LOWMEM_KILOBYTES); | |
419 | lowmem <<= 10; | |
420 | ||
421 | /* start of EBDA area */ | |
356fa0c6 | 422 | ebda_addr = get_bios_ebda(); |
dedd04be AH |
423 | |
424 | /* Fixup: bios puts an EBDA in the top 64K segment */ | |
425 | /* of conventional memory, but does not adjust lowmem. */ | |
426 | if ((lowmem - ebda_addr) <= 0x10000) | |
427 | lowmem = ebda_addr; | |
428 | ||
429 | /* Fixup: bios does not report an EBDA at all. */ | |
430 | /* Some old Dells seem to need 4k anyhow (bugzilla 2990) */ | |
431 | if ((ebda_addr == 0) && (lowmem >= 0x9f000)) | |
432 | lowmem = 0x9f000; | |
433 | ||
434 | /* Paranoia: should never happen, but... */ | |
435 | if ((lowmem == 0) || (lowmem >= 0x100000)) | |
436 | lowmem = 0x9f000; | |
437 | ||
438 | /* reserve all memory between lowmem and the 1MB mark */ | |
439 | reserve_bootmem(lowmem, 0x100000 - lowmem, BOOTMEM_DEFAULT); | |
1da177e4 LT |
440 | } |
441 | ||
05b79bdc | 442 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
a4928cff | 443 | static void __init setup_bootmem_allocator(void); |
1da177e4 LT |
444 | static unsigned long __init setup_memory(void) |
445 | { | |
446 | /* | |
447 | * partially used pages are not usable - thus | |
448 | * we are rounding upwards: | |
449 | */ | |
450 | min_low_pfn = PFN_UP(init_pg_tables_end); | |
451 | ||
1da177e4 LT |
452 | max_low_pfn = find_max_low_pfn(); |
453 | ||
454 | #ifdef CONFIG_HIGHMEM | |
455 | highstart_pfn = highend_pfn = max_pfn; | |
456 | if (max_pfn > max_low_pfn) { | |
457 | highstart_pfn = max_low_pfn; | |
458 | } | |
459 | printk(KERN_NOTICE "%ldMB HIGHMEM available.\n", | |
460 | pages_to_mb(highend_pfn - highstart_pfn)); | |
ba9c231f JB |
461 | num_physpages = highend_pfn; |
462 | high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1; | |
463 | #else | |
464 | num_physpages = max_low_pfn; | |
465 | high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1; | |
466 | #endif | |
467 | #ifdef CONFIG_FLATMEM | |
468 | max_mapnr = num_physpages; | |
1da177e4 LT |
469 | #endif |
470 | printk(KERN_NOTICE "%ldMB LOWMEM available.\n", | |
471 | pages_to_mb(max_low_pfn)); | |
472 | ||
473 | setup_bootmem_allocator(); | |
474 | ||
475 | return max_low_pfn; | |
476 | } | |
477 | ||
a4928cff | 478 | static void __init zone_sizes_init(void) |
1da177e4 | 479 | { |
6391af17 MG |
480 | unsigned long max_zone_pfns[MAX_NR_ZONES]; |
481 | memset(max_zone_pfns, 0, sizeof(max_zone_pfns)); | |
482 | max_zone_pfns[ZONE_DMA] = | |
483 | virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT; | |
484 | max_zone_pfns[ZONE_NORMAL] = max_low_pfn; | |
1da177e4 | 485 | #ifdef CONFIG_HIGHMEM |
6391af17 | 486 | max_zone_pfns[ZONE_HIGHMEM] = highend_pfn; |
4cfee88a MG |
487 | add_active_range(0, 0, highend_pfn); |
488 | #else | |
4cfee88a | 489 | add_active_range(0, 0, max_low_pfn); |
1da177e4 | 490 | #endif |
4cfee88a MG |
491 | |
492 | free_area_init_nodes(max_zone_pfns); | |
1da177e4 LT |
493 | } |
494 | #else | |
05b79bdc | 495 | extern unsigned long __init setup_memory(void); |
1da177e4 | 496 | extern void zone_sizes_init(void); |
05b79bdc | 497 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
1da177e4 | 498 | |
d62cc471 BW |
499 | static inline unsigned long long get_total_mem(void) |
500 | { | |
501 | unsigned long long total; | |
502 | ||
503 | total = max_low_pfn - min_low_pfn; | |
504 | #ifdef CONFIG_HIGHMEM | |
505 | total += highend_pfn - highstart_pfn; | |
506 | #endif | |
507 | ||
508 | return total << PAGE_SHIFT; | |
509 | } | |
510 | ||
511 | #ifdef CONFIG_KEXEC | |
512 | static void __init reserve_crashkernel(void) | |
513 | { | |
514 | unsigned long long total_mem; | |
515 | unsigned long long crash_size, crash_base; | |
516 | int ret; | |
517 | ||
518 | total_mem = get_total_mem(); | |
519 | ||
520 | ret = parse_crashkernel(boot_command_line, total_mem, | |
521 | &crash_size, &crash_base); | |
522 | if (ret == 0 && crash_size > 0) { | |
523 | if (crash_base > 0) { | |
524 | printk(KERN_INFO "Reserving %ldMB of memory at %ldMB " | |
525 | "for crashkernel (System RAM: %ldMB)\n", | |
526 | (unsigned long)(crash_size >> 20), | |
527 | (unsigned long)(crash_base >> 20), | |
528 | (unsigned long)(total_mem >> 20)); | |
529 | crashk_res.start = crash_base; | |
530 | crashk_res.end = crash_base + crash_size - 1; | |
72a7fe39 BW |
531 | reserve_bootmem(crash_base, crash_size, |
532 | BOOTMEM_DEFAULT); | |
d62cc471 BW |
533 | } else |
534 | printk(KERN_INFO "crashkernel reservation failed - " | |
535 | "you have to specify a base address\n"); | |
536 | } | |
537 | } | |
538 | #else | |
539 | static inline void __init reserve_crashkernel(void) | |
540 | {} | |
541 | #endif | |
542 | ||
cf8fa920 PA |
543 | #ifdef CONFIG_BLK_DEV_INITRD |
544 | ||
545 | static bool do_relocate_initrd = false; | |
546 | ||
547 | static void __init reserve_initrd(void) | |
548 | { | |
549 | unsigned long ramdisk_image = boot_params.hdr.ramdisk_image; | |
550 | unsigned long ramdisk_size = boot_params.hdr.ramdisk_size; | |
551 | unsigned long ramdisk_end = ramdisk_image + ramdisk_size; | |
552 | unsigned long end_of_lowmem = max_low_pfn << PAGE_SHIFT; | |
553 | unsigned long ramdisk_here; | |
554 | ||
555 | initrd_start = 0; | |
556 | ||
557 | if (!boot_params.hdr.type_of_loader || | |
558 | !ramdisk_image || !ramdisk_size) | |
559 | return; /* No initrd provided by bootloader */ | |
560 | ||
561 | if (ramdisk_end < ramdisk_image) { | |
562 | printk(KERN_ERR "initrd wraps around end of memory, " | |
563 | "disabling initrd\n"); | |
564 | return; | |
565 | } | |
566 | if (ramdisk_size >= end_of_lowmem/2) { | |
567 | printk(KERN_ERR "initrd too large to handle, " | |
568 | "disabling initrd\n"); | |
569 | return; | |
570 | } | |
571 | if (ramdisk_end <= end_of_lowmem) { | |
572 | /* All in lowmem, easy case */ | |
72a7fe39 | 573 | reserve_bootmem(ramdisk_image, ramdisk_size, BOOTMEM_DEFAULT); |
cf8fa920 PA |
574 | initrd_start = ramdisk_image + PAGE_OFFSET; |
575 | initrd_end = initrd_start+ramdisk_size; | |
576 | return; | |
577 | } | |
578 | ||
579 | /* We need to move the initrd down into lowmem */ | |
580 | ramdisk_here = (end_of_lowmem - ramdisk_size) & PAGE_MASK; | |
581 | ||
582 | /* Note: this includes all the lowmem currently occupied by | |
583 | the initrd, we rely on that fact to keep the data intact. */ | |
72a7fe39 | 584 | reserve_bootmem(ramdisk_here, ramdisk_size, BOOTMEM_DEFAULT); |
cf8fa920 PA |
585 | initrd_start = ramdisk_here + PAGE_OFFSET; |
586 | initrd_end = initrd_start + ramdisk_size; | |
587 | ||
588 | do_relocate_initrd = true; | |
589 | } | |
590 | ||
591 | #define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT) | |
592 | ||
593 | static void __init relocate_initrd(void) | |
594 | { | |
595 | unsigned long ramdisk_image = boot_params.hdr.ramdisk_image; | |
596 | unsigned long ramdisk_size = boot_params.hdr.ramdisk_size; | |
597 | unsigned long end_of_lowmem = max_low_pfn << PAGE_SHIFT; | |
598 | unsigned long ramdisk_here; | |
599 | unsigned long slop, clen, mapaddr; | |
600 | char *p, *q; | |
601 | ||
602 | if (!do_relocate_initrd) | |
603 | return; | |
604 | ||
605 | ramdisk_here = initrd_start - PAGE_OFFSET; | |
606 | ||
607 | q = (char *)initrd_start; | |
608 | ||
609 | /* Copy any lowmem portion of the initrd */ | |
610 | if (ramdisk_image < end_of_lowmem) { | |
611 | clen = end_of_lowmem - ramdisk_image; | |
612 | p = (char *)__va(ramdisk_image); | |
613 | memcpy(q, p, clen); | |
614 | q += clen; | |
615 | ramdisk_image += clen; | |
616 | ramdisk_size -= clen; | |
617 | } | |
618 | ||
619 | /* Copy the highmem portion of the initrd */ | |
620 | while (ramdisk_size) { | |
621 | slop = ramdisk_image & ~PAGE_MASK; | |
622 | clen = ramdisk_size; | |
623 | if (clen > MAX_MAP_CHUNK-slop) | |
624 | clen = MAX_MAP_CHUNK-slop; | |
625 | mapaddr = ramdisk_image & PAGE_MASK; | |
beacfaac | 626 | p = early_ioremap(mapaddr, clen+slop); |
cf8fa920 | 627 | memcpy(q, p+slop, clen); |
beacfaac | 628 | early_iounmap(p, clen+slop); |
cf8fa920 PA |
629 | q += clen; |
630 | ramdisk_image += clen; | |
631 | ramdisk_size -= clen; | |
632 | } | |
633 | } | |
634 | ||
635 | #endif /* CONFIG_BLK_DEV_INITRD */ | |
636 | ||
1da177e4 LT |
637 | void __init setup_bootmem_allocator(void) |
638 | { | |
639 | unsigned long bootmap_size; | |
640 | /* | |
641 | * Initialize the boot-time allocator (with low memory only): | |
642 | */ | |
643 | bootmap_size = init_bootmem(min_low_pfn, max_low_pfn); | |
644 | ||
645 | register_bootmem_low_pages(max_low_pfn); | |
646 | ||
647 | /* | |
648 | * Reserve the bootmem bitmap itself as well. We do this in two | |
649 | * steps (first step was init_bootmem()) because this catches | |
650 | * the (very unlikely) case of us accidentally initializing the | |
651 | * bootmem allocator with an invalid RAM area. | |
652 | */ | |
8621b81c | 653 | reserve_bootmem(__pa_symbol(_text), (PFN_PHYS(min_low_pfn) + |
72a7fe39 BW |
654 | bootmap_size + PAGE_SIZE-1) - __pa_symbol(_text), |
655 | BOOTMEM_DEFAULT); | |
1da177e4 LT |
656 | |
657 | /* | |
658 | * reserve physical page 0 - it's a special BIOS page on many boxes, | |
659 | * enabling clean reboots, SMP operation, laptop functions. | |
660 | */ | |
72a7fe39 | 661 | reserve_bootmem(0, PAGE_SIZE, BOOTMEM_DEFAULT); |
1da177e4 | 662 | |
dedd04be | 663 | /* reserve EBDA region */ |
1da177e4 LT |
664 | reserve_ebda_region(); |
665 | ||
1da177e4 LT |
666 | #ifdef CONFIG_SMP |
667 | /* | |
668 | * But first pinch a few for the stack/trampoline stuff | |
669 | * FIXME: Don't need the extra page at 4K, but need to fix | |
670 | * trampoline before removing it. (see the GDT stuff) | |
671 | */ | |
72a7fe39 | 672 | reserve_bootmem(PAGE_SIZE, PAGE_SIZE, BOOTMEM_DEFAULT); |
1da177e4 | 673 | #endif |
673d5b43 | 674 | #ifdef CONFIG_ACPI_SLEEP |
1da177e4 LT |
675 | /* |
676 | * Reserve low memory region for sleep support. | |
677 | */ | |
678 | acpi_reserve_bootmem(); | |
679 | #endif | |
680 | #ifdef CONFIG_X86_FIND_SMP_CONFIG | |
681 | /* | |
682 | * Find and reserve possible boot-time SMP configuration: | |
683 | */ | |
684 | find_smp_config(); | |
685 | #endif | |
1da177e4 | 686 | #ifdef CONFIG_BLK_DEV_INITRD |
cf8fa920 | 687 | reserve_initrd(); |
1da177e4 | 688 | #endif |
cf8fa920 | 689 | numa_kva_reserve(); |
d62cc471 | 690 | reserve_crashkernel(); |
138fe4e0 KR |
691 | |
692 | reserve_ibft_region(); | |
1da177e4 LT |
693 | } |
694 | ||
695 | /* | |
696 | * The node 0 pgdat is initialized before all of these because | |
697 | * it's needed for bootmem. node>0 pgdats have their virtual | |
698 | * space allocated before the pagetables are in place to access | |
699 | * them, so they can't be cleared then. | |
700 | * | |
701 | * This should all compile down to nothing when NUMA is off. | |
702 | */ | |
10079ae3 | 703 | static void __init remapped_pgdat_init(void) |
1da177e4 LT |
704 | { |
705 | int nid; | |
706 | ||
707 | for_each_online_node(nid) { | |
708 | if (nid != 0) | |
709 | memset(NODE_DATA(nid), 0, sizeof(struct pglist_data)); | |
710 | } | |
711 | } | |
712 | ||
1da177e4 LT |
713 | #ifdef CONFIG_MCA |
714 | static void set_mca_bus(int x) | |
715 | { | |
716 | MCA_bus = x; | |
717 | } | |
718 | #else | |
719 | static void set_mca_bus(int x) { } | |
720 | #endif | |
721 | ||
d3561b7f | 722 | /* Overridden in paravirt.c if CONFIG_PARAVIRT */ |
88d20328 | 723 | char * __init __attribute__((weak)) memory_setup(void) |
d3561b7f RR |
724 | { |
725 | return machine_specific_memory_setup(); | |
726 | } | |
727 | ||
ccf82085 GOC |
728 | #ifdef CONFIG_NUMA |
729 | /* | |
730 | * In the golden day, when everything among i386 and x86_64 will be | |
731 | * integrated, this will not live here | |
732 | */ | |
733 | void *x86_cpu_to_node_map_early_ptr; | |
734 | int x86_cpu_to_node_map_init[NR_CPUS] = { | |
735 | [0 ... NR_CPUS-1] = NUMA_NO_NODE | |
736 | }; | |
737 | DEFINE_PER_CPU(int, x86_cpu_to_node_map) = NUMA_NO_NODE; | |
738 | #endif | |
739 | ||
1da177e4 LT |
740 | /* |
741 | * Determine if we were loaded by an EFI loader. If so, then we have also been | |
742 | * passed the efi memmap, systab, etc., so we should use these data structures | |
743 | * for initialization. Note, the efi init code path is determined by the | |
744 | * global efi_enabled. This allows the same kernel image to be used on existing | |
745 | * systems (with a traditional BIOS) as well as on EFI systems. | |
746 | */ | |
747 | void __init setup_arch(char **cmdline_p) | |
748 | { | |
749 | unsigned long max_low_pfn; | |
750 | ||
751 | memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data)); | |
752 | pre_setup_arch_hook(); | |
753 | early_cpu_init(); | |
beacfaac | 754 | early_ioremap_init(); |
1da177e4 | 755 | |
1da177e4 | 756 | #ifdef CONFIG_EFI |
e429795c HY |
757 | if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature, |
758 | "EL32", 4)) | |
1da177e4 LT |
759 | efi_enabled = 1; |
760 | #endif | |
761 | ||
30c82645 PA |
762 | ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev); |
763 | screen_info = boot_params.screen_info; | |
764 | edid_info = boot_params.edid_info; | |
765 | apm_info.bios = boot_params.apm_bios_info; | |
766 | ist_info = boot_params.ist_info; | |
e44b7b75 | 767 | saved_video_mode = boot_params.hdr.vid_mode; |
30c82645 PA |
768 | if( boot_params.sys_desc_table.length != 0 ) { |
769 | set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2); | |
770 | machine_id = boot_params.sys_desc_table.table[0]; | |
771 | machine_submodel_id = boot_params.sys_desc_table.table[1]; | |
772 | BIOS_revision = boot_params.sys_desc_table.table[2]; | |
1da177e4 | 773 | } |
30c82645 | 774 | bootloader_type = boot_params.hdr.type_of_loader; |
1da177e4 LT |
775 | |
776 | #ifdef CONFIG_BLK_DEV_RAM | |
30c82645 PA |
777 | rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK; |
778 | rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0); | |
779 | rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0); | |
1da177e4 LT |
780 | #endif |
781 | ARCH_SETUP | |
2215e69d HY |
782 | |
783 | printk(KERN_INFO "BIOS-provided physical RAM map:\n"); | |
784 | print_memory_map(memory_setup()); | |
785 | ||
1da177e4 LT |
786 | copy_edd(); |
787 | ||
30c82645 | 788 | if (!boot_params.hdr.root_flags) |
1da177e4 LT |
789 | root_mountflags &= ~MS_RDONLY; |
790 | init_mm.start_code = (unsigned long) _text; | |
791 | init_mm.end_code = (unsigned long) _etext; | |
792 | init_mm.end_data = (unsigned long) _edata; | |
793 | init_mm.brk = init_pg_tables_end + PAGE_OFFSET; | |
794 | ||
795 | code_resource.start = virt_to_phys(_text); | |
796 | code_resource.end = virt_to_phys(_etext)-1; | |
797 | data_resource.start = virt_to_phys(_etext); | |
798 | data_resource.end = virt_to_phys(_edata)-1; | |
00bf4098 BW |
799 | bss_resource.start = virt_to_phys(&__bss_start); |
800 | bss_resource.end = virt_to_phys(&__bss_stop)-1; | |
1da177e4 | 801 | |
1a3f239d | 802 | parse_early_param(); |
1da177e4 | 803 | |
1a3f239d RR |
804 | if (user_defined_memmap) { |
805 | printk(KERN_INFO "user-defined physical RAM map:\n"); | |
806 | print_memory_map("user"); | |
99b7de33 | 807 | } |
1a3f239d | 808 | |
4e498b66 | 809 | strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE); |
1a3f239d | 810 | *cmdline_p = command_line; |
99b7de33 | 811 | |
8b2cb7a8 HY |
812 | if (efi_enabled) |
813 | efi_init(); | |
814 | ||
093af8d7 | 815 | /* update e820 for memory not covered by WB MTRRs */ |
fa5c4639 | 816 | propagate_e820_map(); |
093af8d7 YL |
817 | mtrr_bp_init(); |
818 | if (mtrr_trim_uncached_memory(max_pfn)) | |
fa5c4639 | 819 | propagate_e820_map(); |
76c32418 YL |
820 | |
821 | max_low_pfn = setup_memory(); | |
093af8d7 | 822 | |
7ce0bcfd ZA |
823 | #ifdef CONFIG_VMI |
824 | /* | |
825 | * Must be after max_low_pfn is determined, and before kernel | |
826 | * pagetables are setup. | |
827 | */ | |
828 | vmi_init(); | |
829 | #endif | |
830 | ||
1da177e4 LT |
831 | /* |
832 | * NOTE: before this point _nobody_ is allowed to allocate | |
833 | * any memory using the bootmem allocator. Although the | |
27b46d76 | 834 | * allocator is now initialised only the first 8Mb of the kernel |
1da177e4 LT |
835 | * virtual address space has been mapped. All allocations before |
836 | * paging_init() has completed must use the alloc_bootmem_low_pages() | |
837 | * variant (which allocates DMA'able memory) and care must be taken | |
838 | * not to exceed the 8Mb limit. | |
839 | */ | |
840 | ||
841 | #ifdef CONFIG_SMP | |
842 | smp_alloc_memory(); /* AP processor realmode stacks in low memory*/ | |
843 | #endif | |
844 | paging_init(); | |
f212ec4b BK |
845 | |
846 | /* | |
847 | * NOTE: On x86-32, only from this point on, fixmaps are ready for use. | |
848 | */ | |
849 | ||
850 | #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT | |
851 | if (init_ohci1394_dma_early) | |
852 | init_ohci1394_dma_on_all_controllers(); | |
853 | #endif | |
854 | ||
1da177e4 | 855 | remapped_pgdat_init(); |
05b79bdc | 856 | sparse_init(); |
1da177e4 LT |
857 | zone_sizes_init(); |
858 | ||
859 | /* | |
860 | * NOTE: at this point the bootmem allocator is fully available. | |
861 | */ | |
862 | ||
cf8fa920 PA |
863 | #ifdef CONFIG_BLK_DEV_INITRD |
864 | relocate_initrd(); | |
865 | #endif | |
866 | ||
6996d3b6 JF |
867 | paravirt_post_allocator_init(); |
868 | ||
1da177e4 LT |
869 | dmi_scan_machine(); |
870 | ||
cf8fa920 | 871 | io_delay_init(); |
b02aae9c | 872 | |
ccf82085 GOC |
873 | #ifdef CONFIG_X86_SMP |
874 | /* | |
875 | * setup to use the early static init tables during kernel startup | |
876 | * X86_SMP will exclude sub-arches that don't deal well with it. | |
877 | */ | |
878 | x86_cpu_to_apicid_early_ptr = (void *)x86_cpu_to_apicid_init; | |
879 | x86_bios_cpu_apicid_early_ptr = (void *)x86_bios_cpu_apicid_init; | |
880 | #ifdef CONFIG_NUMA | |
881 | x86_cpu_to_node_map_early_ptr = (void *)x86_cpu_to_node_map_init; | |
882 | #endif | |
883 | #endif | |
884 | ||
1da177e4 | 885 | #ifdef CONFIG_X86_GENERICARCH |
1a3f239d | 886 | generic_apic_probe(); |
cf8fa920 | 887 | #endif |
1da177e4 | 888 | |
888ba6c6 | 889 | #ifdef CONFIG_ACPI |
1da177e4 LT |
890 | /* |
891 | * Parse the ACPI tables for possible boot-time SMP configuration. | |
892 | */ | |
893 | acpi_boot_table_init(); | |
d44647b0 AC |
894 | #endif |
895 | ||
54ef3400 | 896 | early_quirks(); |
d44647b0 AC |
897 | |
898 | #ifdef CONFIG_ACPI | |
1da177e4 LT |
899 | acpi_boot_init(); |
900 | ||
911a62d4 VP |
901 | #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC) |
902 | if (def_to_bigsmp) | |
903 | printk(KERN_WARNING "More than 8 CPUs detected and " | |
904 | "CONFIG_X86_PC cannot handle it.\nUse " | |
905 | "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n"); | |
906 | #endif | |
907 | #endif | |
1da177e4 LT |
908 | #ifdef CONFIG_X86_LOCAL_APIC |
909 | if (smp_found_config) | |
910 | get_smp_config(); | |
911 | #endif | |
912 | ||
5c95da9f | 913 | e820_register_memory(); |
1c10070a | 914 | e820_mark_nosave_regions(); |
1da177e4 LT |
915 | |
916 | #ifdef CONFIG_VT | |
917 | #if defined(CONFIG_VGA_CONSOLE) | |
918 | if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY)) | |
919 | conswitchp = &vga_con; | |
920 | #elif defined(CONFIG_DUMMY_CONSOLE) | |
921 | conswitchp = &dummy_con; | |
922 | #endif | |
923 | #endif | |
924 | } | |
c9cce83d BW |
925 | |
926 | /* | |
927 | * Request address space for all standard resources | |
928 | * | |
929 | * This is called just before pcibios_init(), which is also a | |
930 | * subsys_initcall, but is linked in later (in arch/i386/pci/common.c). | |
931 | */ | |
932 | static int __init request_standard_resources(void) | |
933 | { | |
934 | int i; | |
935 | ||
936 | printk(KERN_INFO "Setting up standard PCI resources\n"); | |
2215e69d | 937 | init_iomem_resources(&code_resource, &data_resource, &bss_resource); |
c9cce83d | 938 | |
c9cce83d BW |
939 | request_resource(&iomem_resource, &video_ram_resource); |
940 | ||
941 | /* request I/O space for devices used on all i[345]86 PCs */ | |
942 | for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++) | |
943 | request_resource(&ioport_resource, &standard_io_resources[i]); | |
944 | return 0; | |
945 | } | |
946 | ||
947 | subsys_initcall(request_standard_resources); |