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> | |
42 | #include <linux/nodemask.h> | |
1bc3b91a | 43 | #include <linux/kexec.h> |
2030eae5 | 44 | #include <linux/crash_dump.h> |
e9928674 | 45 | #include <linux/dmi.h> |
22a9835c | 46 | #include <linux/pfn.h> |
376ff035 | 47 | #include <linux/pci.h> |
1bc3b91a | 48 | |
1da177e4 | 49 | #include <video/edid.h> |
1bc3b91a | 50 | |
9635b47d | 51 | #include <asm/apic.h> |
1da177e4 LT |
52 | #include <asm/e820.h> |
53 | #include <asm/mpspec.h> | |
91023300 | 54 | #include <asm/mmzone.h> |
1da177e4 LT |
55 | #include <asm/setup.h> |
56 | #include <asm/arch_hooks.h> | |
57 | #include <asm/sections.h> | |
58 | #include <asm/io_apic.h> | |
59 | #include <asm/ist.h> | |
60 | #include <asm/io.h> | |
7ce0bcfd | 61 | #include <asm/vmi.h> |
e75eac33 | 62 | #include <setup_arch.h> |
1da177e4 | 63 | #include <bios_ebda.h> |
00bf4098 | 64 | #include <asm/cacheflush.h> |
1da177e4 LT |
65 | |
66 | /* This value is set up by the early boot code to point to the value | |
67 | immediately after the boot time page tables. It contains a *physical* | |
68 | address, and must not be in the .bss segment! */ | |
69 | unsigned long init_pg_tables_end __initdata = ~0UL; | |
70 | ||
1da177e4 LT |
71 | /* |
72 | * Machine setup.. | |
73 | */ | |
c9cce83d BW |
74 | static struct resource data_resource = { |
75 | .name = "Kernel data", | |
76 | .start = 0, | |
77 | .end = 0, | |
78 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
79 | }; | |
80 | ||
81 | static struct resource code_resource = { | |
82 | .name = "Kernel code", | |
83 | .start = 0, | |
84 | .end = 0, | |
85 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
86 | }; | |
87 | ||
88 | static struct resource bss_resource = { | |
89 | .name = "Kernel bss", | |
90 | .start = 0, | |
91 | .end = 0, | |
92 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
93 | }; | |
94 | ||
95 | static struct resource video_ram_resource = { | |
96 | .name = "Video RAM area", | |
97 | .start = 0xa0000, | |
98 | .end = 0xbffff, | |
99 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
100 | }; | |
101 | ||
102 | static struct resource standard_io_resources[] = { { | |
103 | .name = "dma1", | |
104 | .start = 0x0000, | |
105 | .end = 0x001f, | |
106 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
107 | }, { | |
108 | .name = "pic1", | |
109 | .start = 0x0020, | |
110 | .end = 0x0021, | |
111 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
112 | }, { | |
113 | .name = "timer0", | |
114 | .start = 0x0040, | |
115 | .end = 0x0043, | |
116 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
117 | }, { | |
118 | .name = "timer1", | |
119 | .start = 0x0050, | |
120 | .end = 0x0053, | |
121 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
122 | }, { | |
123 | .name = "keyboard", | |
124 | .start = 0x0060, | |
125 | .end = 0x006f, | |
126 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
127 | }, { | |
128 | .name = "dma page reg", | |
129 | .start = 0x0080, | |
130 | .end = 0x008f, | |
131 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
132 | }, { | |
133 | .name = "pic2", | |
134 | .start = 0x00a0, | |
135 | .end = 0x00a1, | |
136 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
137 | }, { | |
138 | .name = "dma2", | |
139 | .start = 0x00c0, | |
140 | .end = 0x00df, | |
141 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
142 | }, { | |
143 | .name = "fpu", | |
144 | .start = 0x00f0, | |
145 | .end = 0x00ff, | |
146 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
147 | } }; | |
1da177e4 LT |
148 | |
149 | /* cpu data as detected by the assembly code in head.S */ | |
4a5d107a | 150 | struct cpuinfo_x86 new_cpu_data __cpuinitdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 }; |
1da177e4 | 151 | /* common cpu data for all cpus */ |
c3d8c141 | 152 | struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 }; |
129f6946 | 153 | EXPORT_SYMBOL(boot_cpu_data); |
1da177e4 LT |
154 | |
155 | unsigned long mmu_cr4_features; | |
156 | ||
1da177e4 LT |
157 | /* for MCA, but anyone else can use it if they want */ |
158 | unsigned int machine_id; | |
159 | unsigned int machine_submodel_id; | |
160 | unsigned int BIOS_revision; | |
161 | unsigned int mca_pentium_flag; | |
162 | ||
1da177e4 LT |
163 | /* Boot loader ID as an integer, for the benefit of proc_dointvec */ |
164 | int bootloader_type; | |
165 | ||
166 | /* user-defined highmem size */ | |
167 | static unsigned int highmem_pages = -1; | |
168 | ||
169 | /* | |
170 | * Setup options | |
171 | */ | |
1da177e4 | 172 | struct screen_info screen_info; |
129f6946 | 173 | EXPORT_SYMBOL(screen_info); |
1da177e4 | 174 | struct apm_info apm_info; |
129f6946 | 175 | EXPORT_SYMBOL(apm_info); |
1da177e4 | 176 | struct edid_info edid_info; |
5e518d76 | 177 | EXPORT_SYMBOL_GPL(edid_info); |
1da177e4 | 178 | struct ist_info ist_info; |
129f6946 AD |
179 | #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \ |
180 | defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE) | |
181 | EXPORT_SYMBOL(ist_info); | |
182 | #endif | |
1da177e4 LT |
183 | |
184 | extern void early_cpu_init(void); | |
1da177e4 LT |
185 | extern int root_mountflags; |
186 | ||
187 | unsigned long saved_videomode; | |
188 | ||
cf8fa920 | 189 | #define RAMDISK_IMAGE_START_MASK 0x07FF |
1da177e4 | 190 | #define RAMDISK_PROMPT_FLAG 0x8000 |
cf8fa920 | 191 | #define RAMDISK_LOAD_FLAG 0x4000 |
1da177e4 | 192 | |
4e498b66 | 193 | static char __initdata command_line[COMMAND_LINE_SIZE]; |
1da177e4 | 194 | |
6d7d7433 | 195 | #ifndef CONFIG_DEBUG_BOOT_PARAMS |
48c7ae67 | 196 | struct boot_params __initdata boot_params; |
6d7d7433 HY |
197 | #else |
198 | struct boot_params boot_params; | |
199 | #endif | |
1da177e4 | 200 | |
1da177e4 LT |
201 | #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE) |
202 | struct edd edd; | |
203 | #ifdef CONFIG_EDD_MODULE | |
204 | EXPORT_SYMBOL(edd); | |
205 | #endif | |
206 | /** | |
207 | * copy_edd() - Copy the BIOS EDD information | |
208 | * from boot_params into a safe place. | |
209 | * | |
210 | */ | |
211 | static inline void copy_edd(void) | |
212 | { | |
30c82645 PA |
213 | memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer, |
214 | sizeof(edd.mbr_signature)); | |
215 | memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info)); | |
216 | edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries; | |
217 | edd.edd_info_nr = boot_params.eddbuf_entries; | |
1da177e4 LT |
218 | } |
219 | #else | |
220 | static inline void copy_edd(void) | |
221 | { | |
222 | } | |
223 | #endif | |
224 | ||
cef518e8 | 225 | int __initdata user_defined_memmap = 0; |
1da177e4 | 226 | |
1a3f239d RR |
227 | /* |
228 | * "mem=nopentium" disables the 4MB page tables. | |
229 | * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM | |
230 | * to <mem>, overriding the bios size. | |
231 | * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from | |
232 | * <start> to <start>+<mem>, overriding the bios size. | |
233 | * | |
234 | * HPA tells me bootloaders need to parse mem=, so no new | |
235 | * option should be mem= [also see Documentation/i386/boot.txt] | |
236 | */ | |
237 | static int __init parse_mem(char *arg) | |
238 | { | |
239 | if (!arg) | |
240 | return -EINVAL; | |
1da177e4 | 241 | |
1a3f239d | 242 | if (strcmp(arg, "nopentium") == 0) { |
13530257 | 243 | setup_clear_cpu_cap(X86_FEATURE_PSE); |
1a3f239d RR |
244 | } else { |
245 | /* If the user specifies memory size, we | |
246 | * limit the BIOS-provided memory map to | |
247 | * that size. exactmap can be used to specify | |
248 | * the exact map. mem=number can be used to | |
249 | * trim the existing memory map. | |
1da177e4 | 250 | */ |
1a3f239d | 251 | unsigned long long mem_size; |
cf8fa920 | 252 | |
1a3f239d RR |
253 | mem_size = memparse(arg, &arg); |
254 | limit_regions(mem_size); | |
255 | user_defined_memmap = 1; | |
256 | } | |
257 | return 0; | |
258 | } | |
259 | early_param("mem", parse_mem); | |
1da177e4 | 260 | |
1a3f239d RR |
261 | #ifdef CONFIG_PROC_VMCORE |
262 | /* elfcorehdr= specifies the location of elf core header | |
263 | * stored by the crashed kernel. | |
264 | */ | |
265 | static int __init parse_elfcorehdr(char *arg) | |
266 | { | |
267 | if (!arg) | |
268 | return -EINVAL; | |
66759a01 | 269 | |
1a3f239d RR |
270 | elfcorehdr_addr = memparse(arg, &arg); |
271 | return 0; | |
272 | } | |
273 | early_param("elfcorehdr", parse_elfcorehdr); | |
274 | #endif /* CONFIG_PROC_VMCORE */ | |
1da177e4 | 275 | |
1a3f239d RR |
276 | /* |
277 | * highmem=size forces highmem to be exactly 'size' bytes. | |
278 | * This works even on boxes that have no highmem otherwise. | |
279 | * This also works to reduce highmem size on bigger boxes. | |
280 | */ | |
281 | static int __init parse_highmem(char *arg) | |
282 | { | |
283 | if (!arg) | |
284 | return -EINVAL; | |
9635b47d | 285 | |
1a3f239d RR |
286 | highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT; |
287 | return 0; | |
288 | } | |
289 | early_param("highmem", parse_highmem); | |
9635b47d | 290 | |
1a3f239d RR |
291 | /* |
292 | * vmalloc=size forces the vmalloc area to be exactly 'size' | |
293 | * bytes. This can be used to increase (or decrease) the | |
294 | * vmalloc area - the default is 128m. | |
295 | */ | |
296 | static int __init parse_vmalloc(char *arg) | |
297 | { | |
298 | if (!arg) | |
299 | return -EINVAL; | |
1bc3b91a | 300 | |
1a3f239d RR |
301 | __VMALLOC_RESERVE = memparse(arg, &arg); |
302 | return 0; | |
1da177e4 | 303 | } |
1a3f239d | 304 | early_param("vmalloc", parse_vmalloc); |
1da177e4 | 305 | |
461a9aff ZA |
306 | /* |
307 | * reservetop=size reserves a hole at the top of the kernel address space which | |
308 | * a hypervisor can load into later. Needed for dynamically loaded hypervisors, | |
309 | * so relocating the fixmap can be done before paging initialization. | |
310 | */ | |
311 | static int __init parse_reservetop(char *arg) | |
312 | { | |
313 | unsigned long address; | |
314 | ||
315 | if (!arg) | |
316 | return -EINVAL; | |
317 | ||
318 | address = memparse(arg, &arg); | |
319 | reserve_top_address(address); | |
320 | return 0; | |
321 | } | |
322 | early_param("reservetop", parse_reservetop); | |
323 | ||
1da177e4 LT |
324 | /* |
325 | * Determine low and high memory ranges: | |
326 | */ | |
327 | unsigned long __init find_max_low_pfn(void) | |
328 | { | |
329 | unsigned long max_low_pfn; | |
330 | ||
331 | max_low_pfn = max_pfn; | |
332 | if (max_low_pfn > MAXMEM_PFN) { | |
333 | if (highmem_pages == -1) | |
334 | highmem_pages = max_pfn - MAXMEM_PFN; | |
335 | if (highmem_pages + MAXMEM_PFN < max_pfn) | |
336 | max_pfn = MAXMEM_PFN + highmem_pages; | |
337 | if (highmem_pages + MAXMEM_PFN > max_pfn) { | |
338 | printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn - MAXMEM_PFN), pages_to_mb(highmem_pages)); | |
339 | highmem_pages = 0; | |
340 | } | |
341 | max_low_pfn = MAXMEM_PFN; | |
342 | #ifndef CONFIG_HIGHMEM | |
343 | /* Maximum memory usable is what is directly addressable */ | |
344 | printk(KERN_WARNING "Warning only %ldMB will be used.\n", | |
345 | MAXMEM>>20); | |
346 | if (max_pfn > MAX_NONPAE_PFN) | |
c673f1a9 | 347 | printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n"); |
1da177e4 LT |
348 | else |
349 | printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n"); | |
350 | max_pfn = MAXMEM_PFN; | |
351 | #else /* !CONFIG_HIGHMEM */ | |
c673f1a9 | 352 | #ifndef CONFIG_HIGHMEM64G |
1da177e4 LT |
353 | if (max_pfn > MAX_NONPAE_PFN) { |
354 | max_pfn = MAX_NONPAE_PFN; | |
355 | printk(KERN_WARNING "Warning only 4GB will be used.\n"); | |
c673f1a9 | 356 | printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n"); |
1da177e4 | 357 | } |
c673f1a9 | 358 | #endif /* !CONFIG_HIGHMEM64G */ |
1da177e4 LT |
359 | #endif /* !CONFIG_HIGHMEM */ |
360 | } else { | |
361 | if (highmem_pages == -1) | |
362 | highmem_pages = 0; | |
363 | #ifdef CONFIG_HIGHMEM | |
364 | if (highmem_pages >= max_pfn) { | |
365 | printk(KERN_ERR "highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages), pages_to_mb(max_pfn)); | |
366 | highmem_pages = 0; | |
367 | } | |
368 | if (highmem_pages) { | |
369 | if (max_low_pfn-highmem_pages < 64*1024*1024/PAGE_SIZE){ | |
370 | printk(KERN_ERR "highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages)); | |
371 | highmem_pages = 0; | |
372 | } | |
373 | max_low_pfn -= highmem_pages; | |
374 | } | |
375 | #else | |
376 | if (highmem_pages) | |
377 | printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n"); | |
378 | #endif | |
379 | } | |
380 | return max_low_pfn; | |
381 | } | |
382 | ||
1da177e4 LT |
383 | /* |
384 | * workaround for Dell systems that neglect to reserve EBDA | |
385 | */ | |
386 | static void __init reserve_ebda_region(void) | |
387 | { | |
388 | unsigned int addr; | |
389 | addr = get_bios_ebda(); | |
390 | if (addr) | |
cf8fa920 | 391 | reserve_bootmem(addr, PAGE_SIZE); |
1da177e4 LT |
392 | } |
393 | ||
05b79bdc | 394 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
1da177e4 LT |
395 | void __init setup_bootmem_allocator(void); |
396 | static unsigned long __init setup_memory(void) | |
397 | { | |
398 | /* | |
399 | * partially used pages are not usable - thus | |
400 | * we are rounding upwards: | |
401 | */ | |
402 | min_low_pfn = PFN_UP(init_pg_tables_end); | |
403 | ||
404 | find_max_pfn(); | |
405 | ||
406 | max_low_pfn = find_max_low_pfn(); | |
407 | ||
408 | #ifdef CONFIG_HIGHMEM | |
409 | highstart_pfn = highend_pfn = max_pfn; | |
410 | if (max_pfn > max_low_pfn) { | |
411 | highstart_pfn = max_low_pfn; | |
412 | } | |
413 | printk(KERN_NOTICE "%ldMB HIGHMEM available.\n", | |
414 | pages_to_mb(highend_pfn - highstart_pfn)); | |
ba9c231f JB |
415 | num_physpages = highend_pfn; |
416 | high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1; | |
417 | #else | |
418 | num_physpages = max_low_pfn; | |
419 | high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1; | |
420 | #endif | |
421 | #ifdef CONFIG_FLATMEM | |
422 | max_mapnr = num_physpages; | |
1da177e4 LT |
423 | #endif |
424 | printk(KERN_NOTICE "%ldMB LOWMEM available.\n", | |
425 | pages_to_mb(max_low_pfn)); | |
426 | ||
427 | setup_bootmem_allocator(); | |
428 | ||
429 | return max_low_pfn; | |
430 | } | |
431 | ||
432 | void __init zone_sizes_init(void) | |
433 | { | |
6391af17 MG |
434 | unsigned long max_zone_pfns[MAX_NR_ZONES]; |
435 | memset(max_zone_pfns, 0, sizeof(max_zone_pfns)); | |
436 | max_zone_pfns[ZONE_DMA] = | |
437 | virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT; | |
438 | max_zone_pfns[ZONE_NORMAL] = max_low_pfn; | |
1da177e4 | 439 | #ifdef CONFIG_HIGHMEM |
6391af17 | 440 | max_zone_pfns[ZONE_HIGHMEM] = highend_pfn; |
4cfee88a MG |
441 | add_active_range(0, 0, highend_pfn); |
442 | #else | |
4cfee88a | 443 | add_active_range(0, 0, max_low_pfn); |
1da177e4 | 444 | #endif |
4cfee88a MG |
445 | |
446 | free_area_init_nodes(max_zone_pfns); | |
1da177e4 LT |
447 | } |
448 | #else | |
05b79bdc | 449 | extern unsigned long __init setup_memory(void); |
1da177e4 | 450 | extern void zone_sizes_init(void); |
05b79bdc | 451 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
1da177e4 | 452 | |
d62cc471 BW |
453 | static inline unsigned long long get_total_mem(void) |
454 | { | |
455 | unsigned long long total; | |
456 | ||
457 | total = max_low_pfn - min_low_pfn; | |
458 | #ifdef CONFIG_HIGHMEM | |
459 | total += highend_pfn - highstart_pfn; | |
460 | #endif | |
461 | ||
462 | return total << PAGE_SHIFT; | |
463 | } | |
464 | ||
465 | #ifdef CONFIG_KEXEC | |
466 | static void __init reserve_crashkernel(void) | |
467 | { | |
468 | unsigned long long total_mem; | |
469 | unsigned long long crash_size, crash_base; | |
470 | int ret; | |
471 | ||
472 | total_mem = get_total_mem(); | |
473 | ||
474 | ret = parse_crashkernel(boot_command_line, total_mem, | |
475 | &crash_size, &crash_base); | |
476 | if (ret == 0 && crash_size > 0) { | |
477 | if (crash_base > 0) { | |
478 | printk(KERN_INFO "Reserving %ldMB of memory at %ldMB " | |
479 | "for crashkernel (System RAM: %ldMB)\n", | |
480 | (unsigned long)(crash_size >> 20), | |
481 | (unsigned long)(crash_base >> 20), | |
482 | (unsigned long)(total_mem >> 20)); | |
483 | crashk_res.start = crash_base; | |
484 | crashk_res.end = crash_base + crash_size - 1; | |
485 | reserve_bootmem(crash_base, crash_size); | |
486 | } else | |
487 | printk(KERN_INFO "crashkernel reservation failed - " | |
488 | "you have to specify a base address\n"); | |
489 | } | |
490 | } | |
491 | #else | |
492 | static inline void __init reserve_crashkernel(void) | |
493 | {} | |
494 | #endif | |
495 | ||
cf8fa920 PA |
496 | #ifdef CONFIG_BLK_DEV_INITRD |
497 | ||
498 | static bool do_relocate_initrd = false; | |
499 | ||
500 | static void __init reserve_initrd(void) | |
501 | { | |
502 | unsigned long ramdisk_image = boot_params.hdr.ramdisk_image; | |
503 | unsigned long ramdisk_size = boot_params.hdr.ramdisk_size; | |
504 | unsigned long ramdisk_end = ramdisk_image + ramdisk_size; | |
505 | unsigned long end_of_lowmem = max_low_pfn << PAGE_SHIFT; | |
506 | unsigned long ramdisk_here; | |
507 | ||
508 | initrd_start = 0; | |
509 | ||
510 | if (!boot_params.hdr.type_of_loader || | |
511 | !ramdisk_image || !ramdisk_size) | |
512 | return; /* No initrd provided by bootloader */ | |
513 | ||
514 | if (ramdisk_end < ramdisk_image) { | |
515 | printk(KERN_ERR "initrd wraps around end of memory, " | |
516 | "disabling initrd\n"); | |
517 | return; | |
518 | } | |
519 | if (ramdisk_size >= end_of_lowmem/2) { | |
520 | printk(KERN_ERR "initrd too large to handle, " | |
521 | "disabling initrd\n"); | |
522 | return; | |
523 | } | |
524 | if (ramdisk_end <= end_of_lowmem) { | |
525 | /* All in lowmem, easy case */ | |
526 | reserve_bootmem(ramdisk_image, ramdisk_size); | |
527 | initrd_start = ramdisk_image + PAGE_OFFSET; | |
528 | initrd_end = initrd_start+ramdisk_size; | |
529 | return; | |
530 | } | |
531 | ||
532 | /* We need to move the initrd down into lowmem */ | |
533 | ramdisk_here = (end_of_lowmem - ramdisk_size) & PAGE_MASK; | |
534 | ||
535 | /* Note: this includes all the lowmem currently occupied by | |
536 | the initrd, we rely on that fact to keep the data intact. */ | |
537 | reserve_bootmem(ramdisk_here, ramdisk_size); | |
538 | initrd_start = ramdisk_here + PAGE_OFFSET; | |
539 | initrd_end = initrd_start + ramdisk_size; | |
540 | ||
541 | do_relocate_initrd = true; | |
542 | } | |
543 | ||
544 | #define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT) | |
545 | ||
546 | static void __init relocate_initrd(void) | |
547 | { | |
548 | unsigned long ramdisk_image = boot_params.hdr.ramdisk_image; | |
549 | unsigned long ramdisk_size = boot_params.hdr.ramdisk_size; | |
550 | unsigned long end_of_lowmem = max_low_pfn << PAGE_SHIFT; | |
551 | unsigned long ramdisk_here; | |
552 | unsigned long slop, clen, mapaddr; | |
553 | char *p, *q; | |
554 | ||
555 | if (!do_relocate_initrd) | |
556 | return; | |
557 | ||
558 | ramdisk_here = initrd_start - PAGE_OFFSET; | |
559 | ||
560 | q = (char *)initrd_start; | |
561 | ||
562 | /* Copy any lowmem portion of the initrd */ | |
563 | if (ramdisk_image < end_of_lowmem) { | |
564 | clen = end_of_lowmem - ramdisk_image; | |
565 | p = (char *)__va(ramdisk_image); | |
566 | memcpy(q, p, clen); | |
567 | q += clen; | |
568 | ramdisk_image += clen; | |
569 | ramdisk_size -= clen; | |
570 | } | |
571 | ||
572 | /* Copy the highmem portion of the initrd */ | |
573 | while (ramdisk_size) { | |
574 | slop = ramdisk_image & ~PAGE_MASK; | |
575 | clen = ramdisk_size; | |
576 | if (clen > MAX_MAP_CHUNK-slop) | |
577 | clen = MAX_MAP_CHUNK-slop; | |
578 | mapaddr = ramdisk_image & PAGE_MASK; | |
579 | p = bt_ioremap(mapaddr, clen+slop); | |
580 | memcpy(q, p+slop, clen); | |
581 | bt_iounmap(p, clen+slop); | |
582 | q += clen; | |
583 | ramdisk_image += clen; | |
584 | ramdisk_size -= clen; | |
585 | } | |
586 | } | |
587 | ||
588 | #endif /* CONFIG_BLK_DEV_INITRD */ | |
589 | ||
1da177e4 LT |
590 | void __init setup_bootmem_allocator(void) |
591 | { | |
592 | unsigned long bootmap_size; | |
593 | /* | |
594 | * Initialize the boot-time allocator (with low memory only): | |
595 | */ | |
596 | bootmap_size = init_bootmem(min_low_pfn, max_low_pfn); | |
597 | ||
598 | register_bootmem_low_pages(max_low_pfn); | |
599 | ||
600 | /* | |
601 | * Reserve the bootmem bitmap itself as well. We do this in two | |
602 | * steps (first step was init_bootmem()) because this catches | |
603 | * the (very unlikely) case of us accidentally initializing the | |
604 | * bootmem allocator with an invalid RAM area. | |
605 | */ | |
8621b81c EB |
606 | reserve_bootmem(__pa_symbol(_text), (PFN_PHYS(min_low_pfn) + |
607 | bootmap_size + PAGE_SIZE-1) - __pa_symbol(_text)); | |
1da177e4 LT |
608 | |
609 | /* | |
610 | * reserve physical page 0 - it's a special BIOS page on many boxes, | |
611 | * enabling clean reboots, SMP operation, laptop functions. | |
612 | */ | |
613 | reserve_bootmem(0, PAGE_SIZE); | |
614 | ||
615 | /* reserve EBDA region, it's a 4K region */ | |
616 | reserve_ebda_region(); | |
617 | ||
618 | /* could be an AMD 768MPX chipset. Reserve a page before VGA to prevent | |
619 | PCI prefetch into it (errata #56). Usually the page is reserved anyways, | |
620 | unless you have no PS/2 mouse plugged in. */ | |
621 | if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD && | |
622 | boot_cpu_data.x86 == 6) | |
623 | reserve_bootmem(0xa0000 - 4096, 4096); | |
624 | ||
625 | #ifdef CONFIG_SMP | |
626 | /* | |
627 | * But first pinch a few for the stack/trampoline stuff | |
628 | * FIXME: Don't need the extra page at 4K, but need to fix | |
629 | * trampoline before removing it. (see the GDT stuff) | |
630 | */ | |
631 | reserve_bootmem(PAGE_SIZE, PAGE_SIZE); | |
632 | #endif | |
673d5b43 | 633 | #ifdef CONFIG_ACPI_SLEEP |
1da177e4 LT |
634 | /* |
635 | * Reserve low memory region for sleep support. | |
636 | */ | |
637 | acpi_reserve_bootmem(); | |
638 | #endif | |
639 | #ifdef CONFIG_X86_FIND_SMP_CONFIG | |
640 | /* | |
641 | * Find and reserve possible boot-time SMP configuration: | |
642 | */ | |
643 | find_smp_config(); | |
644 | #endif | |
1da177e4 | 645 | #ifdef CONFIG_BLK_DEV_INITRD |
cf8fa920 | 646 | reserve_initrd(); |
1da177e4 | 647 | #endif |
cf8fa920 | 648 | numa_kva_reserve(); |
d62cc471 | 649 | reserve_crashkernel(); |
1da177e4 LT |
650 | } |
651 | ||
652 | /* | |
653 | * The node 0 pgdat is initialized before all of these because | |
654 | * it's needed for bootmem. node>0 pgdats have their virtual | |
655 | * space allocated before the pagetables are in place to access | |
656 | * them, so they can't be cleared then. | |
657 | * | |
658 | * This should all compile down to nothing when NUMA is off. | |
659 | */ | |
10079ae3 | 660 | static void __init remapped_pgdat_init(void) |
1da177e4 LT |
661 | { |
662 | int nid; | |
663 | ||
664 | for_each_online_node(nid) { | |
665 | if (nid != 0) | |
666 | memset(NODE_DATA(nid), 0, sizeof(struct pglist_data)); | |
667 | } | |
668 | } | |
669 | ||
1da177e4 LT |
670 | #ifdef CONFIG_MCA |
671 | static void set_mca_bus(int x) | |
672 | { | |
673 | MCA_bus = x; | |
674 | } | |
675 | #else | |
676 | static void set_mca_bus(int x) { } | |
677 | #endif | |
678 | ||
d3561b7f | 679 | /* Overridden in paravirt.c if CONFIG_PARAVIRT */ |
88d20328 | 680 | char * __init __attribute__((weak)) memory_setup(void) |
d3561b7f RR |
681 | { |
682 | return machine_specific_memory_setup(); | |
683 | } | |
684 | ||
1da177e4 LT |
685 | /* |
686 | * Determine if we were loaded by an EFI loader. If so, then we have also been | |
687 | * passed the efi memmap, systab, etc., so we should use these data structures | |
688 | * for initialization. Note, the efi init code path is determined by the | |
689 | * global efi_enabled. This allows the same kernel image to be used on existing | |
690 | * systems (with a traditional BIOS) as well as on EFI systems. | |
691 | */ | |
692 | void __init setup_arch(char **cmdline_p) | |
693 | { | |
694 | unsigned long max_low_pfn; | |
695 | ||
696 | memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data)); | |
697 | pre_setup_arch_hook(); | |
698 | early_cpu_init(); | |
699 | ||
1da177e4 | 700 | #ifdef CONFIG_EFI |
e429795c HY |
701 | if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature, |
702 | "EL32", 4)) | |
1da177e4 LT |
703 | efi_enabled = 1; |
704 | #endif | |
705 | ||
30c82645 PA |
706 | ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev); |
707 | screen_info = boot_params.screen_info; | |
708 | edid_info = boot_params.edid_info; | |
709 | apm_info.bios = boot_params.apm_bios_info; | |
710 | ist_info = boot_params.ist_info; | |
711 | saved_videomode = boot_params.hdr.vid_mode; | |
712 | if( boot_params.sys_desc_table.length != 0 ) { | |
713 | set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2); | |
714 | machine_id = boot_params.sys_desc_table.table[0]; | |
715 | machine_submodel_id = boot_params.sys_desc_table.table[1]; | |
716 | BIOS_revision = boot_params.sys_desc_table.table[2]; | |
1da177e4 | 717 | } |
30c82645 | 718 | bootloader_type = boot_params.hdr.type_of_loader; |
1da177e4 LT |
719 | |
720 | #ifdef CONFIG_BLK_DEV_RAM | |
30c82645 PA |
721 | rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK; |
722 | rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0); | |
723 | rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0); | |
1da177e4 LT |
724 | #endif |
725 | ARCH_SETUP | |
2215e69d HY |
726 | |
727 | printk(KERN_INFO "BIOS-provided physical RAM map:\n"); | |
728 | print_memory_map(memory_setup()); | |
729 | ||
1da177e4 LT |
730 | copy_edd(); |
731 | ||
30c82645 | 732 | if (!boot_params.hdr.root_flags) |
1da177e4 LT |
733 | root_mountflags &= ~MS_RDONLY; |
734 | init_mm.start_code = (unsigned long) _text; | |
735 | init_mm.end_code = (unsigned long) _etext; | |
736 | init_mm.end_data = (unsigned long) _edata; | |
737 | init_mm.brk = init_pg_tables_end + PAGE_OFFSET; | |
738 | ||
739 | code_resource.start = virt_to_phys(_text); | |
740 | code_resource.end = virt_to_phys(_etext)-1; | |
741 | data_resource.start = virt_to_phys(_etext); | |
742 | data_resource.end = virt_to_phys(_edata)-1; | |
00bf4098 BW |
743 | bss_resource.start = virt_to_phys(&__bss_start); |
744 | bss_resource.end = virt_to_phys(&__bss_stop)-1; | |
1da177e4 | 745 | |
1a3f239d | 746 | parse_early_param(); |
1da177e4 | 747 | |
1a3f239d RR |
748 | if (user_defined_memmap) { |
749 | printk(KERN_INFO "user-defined physical RAM map:\n"); | |
750 | print_memory_map("user"); | |
99b7de33 | 751 | } |
1a3f239d | 752 | |
4e498b66 | 753 | strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE); |
1a3f239d | 754 | *cmdline_p = command_line; |
99b7de33 | 755 | |
8b2cb7a8 HY |
756 | if (efi_enabled) |
757 | efi_init(); | |
758 | ||
1da177e4 LT |
759 | max_low_pfn = setup_memory(); |
760 | ||
7ce0bcfd ZA |
761 | #ifdef CONFIG_VMI |
762 | /* | |
763 | * Must be after max_low_pfn is determined, and before kernel | |
764 | * pagetables are setup. | |
765 | */ | |
766 | vmi_init(); | |
767 | #endif | |
768 | ||
1da177e4 LT |
769 | /* |
770 | * NOTE: before this point _nobody_ is allowed to allocate | |
771 | * any memory using the bootmem allocator. Although the | |
27b46d76 | 772 | * allocator is now initialised only the first 8Mb of the kernel |
1da177e4 LT |
773 | * virtual address space has been mapped. All allocations before |
774 | * paging_init() has completed must use the alloc_bootmem_low_pages() | |
775 | * variant (which allocates DMA'able memory) and care must be taken | |
776 | * not to exceed the 8Mb limit. | |
777 | */ | |
778 | ||
779 | #ifdef CONFIG_SMP | |
780 | smp_alloc_memory(); /* AP processor realmode stacks in low memory*/ | |
781 | #endif | |
782 | paging_init(); | |
783 | remapped_pgdat_init(); | |
05b79bdc | 784 | sparse_init(); |
1da177e4 LT |
785 | zone_sizes_init(); |
786 | ||
787 | /* | |
788 | * NOTE: at this point the bootmem allocator is fully available. | |
789 | */ | |
790 | ||
cf8fa920 PA |
791 | #ifdef CONFIG_BLK_DEV_INITRD |
792 | relocate_initrd(); | |
793 | #endif | |
794 | ||
6996d3b6 JF |
795 | paravirt_post_allocator_init(); |
796 | ||
1da177e4 LT |
797 | dmi_scan_machine(); |
798 | ||
cf8fa920 | 799 | io_delay_init(); |
b02aae9c | 800 | |
1da177e4 | 801 | #ifdef CONFIG_X86_GENERICARCH |
1a3f239d | 802 | generic_apic_probe(); |
cf8fa920 | 803 | #endif |
1da177e4 LT |
804 | if (efi_enabled) |
805 | efi_map_memmap(); | |
806 | ||
888ba6c6 | 807 | #ifdef CONFIG_ACPI |
1da177e4 LT |
808 | /* |
809 | * Parse the ACPI tables for possible boot-time SMP configuration. | |
810 | */ | |
811 | acpi_boot_table_init(); | |
d44647b0 AC |
812 | #endif |
813 | ||
54ef3400 | 814 | early_quirks(); |
d44647b0 AC |
815 | |
816 | #ifdef CONFIG_ACPI | |
1da177e4 LT |
817 | acpi_boot_init(); |
818 | ||
911a62d4 VP |
819 | #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC) |
820 | if (def_to_bigsmp) | |
821 | printk(KERN_WARNING "More than 8 CPUs detected and " | |
822 | "CONFIG_X86_PC cannot handle it.\nUse " | |
823 | "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n"); | |
824 | #endif | |
825 | #endif | |
1da177e4 LT |
826 | #ifdef CONFIG_X86_LOCAL_APIC |
827 | if (smp_found_config) | |
828 | get_smp_config(); | |
829 | #endif | |
830 | ||
5c95da9f | 831 | e820_register_memory(); |
1c10070a | 832 | e820_mark_nosave_regions(); |
1da177e4 LT |
833 | |
834 | #ifdef CONFIG_VT | |
835 | #if defined(CONFIG_VGA_CONSOLE) | |
836 | if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY)) | |
837 | conswitchp = &vga_con; | |
838 | #elif defined(CONFIG_DUMMY_CONSOLE) | |
839 | conswitchp = &dummy_con; | |
840 | #endif | |
841 | #endif | |
842 | } | |
c9cce83d BW |
843 | |
844 | /* | |
845 | * Request address space for all standard resources | |
846 | * | |
847 | * This is called just before pcibios_init(), which is also a | |
848 | * subsys_initcall, but is linked in later (in arch/i386/pci/common.c). | |
849 | */ | |
850 | static int __init request_standard_resources(void) | |
851 | { | |
852 | int i; | |
853 | ||
854 | printk(KERN_INFO "Setting up standard PCI resources\n"); | |
2215e69d | 855 | init_iomem_resources(&code_resource, &data_resource, &bss_resource); |
c9cce83d | 856 | |
c9cce83d BW |
857 | request_resource(&iomem_resource, &video_ram_resource); |
858 | ||
859 | /* request I/O space for devices used on all i[345]86 PCs */ | |
860 | for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++) | |
861 | request_resource(&ioport_resource, &standard_io_resources[i]); | |
862 | return 0; | |
863 | } | |
864 | ||
865 | subsys_initcall(request_standard_resources); |