]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/arch/i386/kernel/setup.c | |
3 | * | |
4 | * Copyright (C) 1995 Linus Torvalds | |
5 | * | |
6 | * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999 | |
7 | * | |
8 | * Memory region support | |
9 | * David Parsons <orc@pell.chi.il.us>, July-August 1999 | |
10 | * | |
11 | * Added E820 sanitization routine (removes overlapping memory regions); | |
12 | * Brian Moyle <bmoyle@mvista.com>, February 2001 | |
13 | * | |
14 | * Moved CPU detection code to cpu/${cpu}.c | |
15 | * Patrick Mochel <mochel@osdl.org>, March 2002 | |
16 | * | |
17 | * Provisions for empty E820 memory regions (reported by certain BIOSes). | |
18 | * Alex Achenbach <xela@slit.de>, December 2002. | |
19 | * | |
20 | */ | |
21 | ||
22 | /* | |
23 | * This file handles the architecture-dependent parts of initialization | |
24 | */ | |
25 | ||
26 | #include <linux/sched.h> | |
27 | #include <linux/mm.h> | |
05b79bdc | 28 | #include <linux/mmzone.h> |
894673ee | 29 | #include <linux/screen_info.h> |
1da177e4 LT |
30 | #include <linux/ioport.h> |
31 | #include <linux/acpi.h> | |
32 | #include <linux/apm_bios.h> | |
33 | #include <linux/initrd.h> | |
34 | #include <linux/bootmem.h> | |
35 | #include <linux/seq_file.h> | |
e5c6c8e4 | 36 | #include <linux/platform_device.h> |
1da177e4 LT |
37 | #include <linux/console.h> |
38 | #include <linux/mca.h> | |
39 | #include <linux/root_dev.h> | |
40 | #include <linux/highmem.h> | |
41 | #include <linux/module.h> | |
42 | #include <linux/efi.h> | |
43 | #include <linux/init.h> | |
44 | #include <linux/edd.h> | |
45 | #include <linux/nodemask.h> | |
1bc3b91a | 46 | #include <linux/kexec.h> |
2030eae5 | 47 | #include <linux/crash_dump.h> |
e9928674 | 48 | #include <linux/dmi.h> |
22a9835c | 49 | #include <linux/pfn.h> |
1bc3b91a | 50 | |
1da177e4 | 51 | #include <video/edid.h> |
1bc3b91a | 52 | |
9635b47d | 53 | #include <asm/apic.h> |
1da177e4 LT |
54 | #include <asm/e820.h> |
55 | #include <asm/mpspec.h> | |
91023300 | 56 | #include <asm/mmzone.h> |
1da177e4 LT |
57 | #include <asm/setup.h> |
58 | #include <asm/arch_hooks.h> | |
59 | #include <asm/sections.h> | |
60 | #include <asm/io_apic.h> | |
61 | #include <asm/ist.h> | |
62 | #include <asm/io.h> | |
e75eac33 | 63 | #include <setup_arch.h> |
1da177e4 LT |
64 | #include <bios_ebda.h> |
65 | ||
92aa63a5 VG |
66 | /* Forward Declaration. */ |
67 | void __init find_max_pfn(void); | |
68 | ||
1da177e4 LT |
69 | /* This value is set up by the early boot code to point to the value |
70 | immediately after the boot time page tables. It contains a *physical* | |
71 | address, and must not be in the .bss segment! */ | |
72 | unsigned long init_pg_tables_end __initdata = ~0UL; | |
73 | ||
0bb3184d | 74 | int disable_pse __devinitdata = 0; |
1da177e4 LT |
75 | |
76 | /* | |
77 | * Machine setup.. | |
78 | */ | |
79 | ||
80 | #ifdef CONFIG_EFI | |
81 | int efi_enabled = 0; | |
82 | EXPORT_SYMBOL(efi_enabled); | |
83 | #endif | |
84 | ||
85 | /* cpu data as detected by the assembly code in head.S */ | |
86 | struct cpuinfo_x86 new_cpu_data __initdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 }; | |
87 | /* common cpu data for all cpus */ | |
c3d8c141 | 88 | struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 }; |
129f6946 | 89 | EXPORT_SYMBOL(boot_cpu_data); |
1da177e4 LT |
90 | |
91 | unsigned long mmu_cr4_features; | |
92 | ||
1da177e4 LT |
93 | /* for MCA, but anyone else can use it if they want */ |
94 | unsigned int machine_id; | |
129f6946 AD |
95 | #ifdef CONFIG_MCA |
96 | EXPORT_SYMBOL(machine_id); | |
97 | #endif | |
1da177e4 LT |
98 | unsigned int machine_submodel_id; |
99 | unsigned int BIOS_revision; | |
100 | unsigned int mca_pentium_flag; | |
101 | ||
102 | /* For PCI or other memory-mapped resources */ | |
103 | unsigned long pci_mem_start = 0x10000000; | |
129f6946 AD |
104 | #ifdef CONFIG_PCI |
105 | EXPORT_SYMBOL(pci_mem_start); | |
106 | #endif | |
1da177e4 LT |
107 | |
108 | /* Boot loader ID as an integer, for the benefit of proc_dointvec */ | |
109 | int bootloader_type; | |
110 | ||
111 | /* user-defined highmem size */ | |
112 | static unsigned int highmem_pages = -1; | |
113 | ||
114 | /* | |
115 | * Setup options | |
116 | */ | |
117 | struct drive_info_struct { char dummy[32]; } drive_info; | |
129f6946 AD |
118 | #if defined(CONFIG_BLK_DEV_IDE) || defined(CONFIG_BLK_DEV_HD) || \ |
119 | defined(CONFIG_BLK_DEV_IDE_MODULE) || defined(CONFIG_BLK_DEV_HD_MODULE) | |
120 | EXPORT_SYMBOL(drive_info); | |
121 | #endif | |
1da177e4 | 122 | struct screen_info screen_info; |
129f6946 | 123 | EXPORT_SYMBOL(screen_info); |
1da177e4 | 124 | struct apm_info apm_info; |
129f6946 | 125 | EXPORT_SYMBOL(apm_info); |
1da177e4 LT |
126 | struct sys_desc_table_struct { |
127 | unsigned short length; | |
128 | unsigned char table[0]; | |
129 | }; | |
130 | struct edid_info edid_info; | |
5e518d76 | 131 | EXPORT_SYMBOL_GPL(edid_info); |
1da177e4 | 132 | struct ist_info ist_info; |
129f6946 AD |
133 | #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \ |
134 | defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE) | |
135 | EXPORT_SYMBOL(ist_info); | |
136 | #endif | |
1da177e4 LT |
137 | struct e820map e820; |
138 | ||
139 | extern void early_cpu_init(void); | |
1da177e4 LT |
140 | extern int root_mountflags; |
141 | ||
142 | unsigned long saved_videomode; | |
143 | ||
144 | #define RAMDISK_IMAGE_START_MASK 0x07FF | |
145 | #define RAMDISK_PROMPT_FLAG 0x8000 | |
146 | #define RAMDISK_LOAD_FLAG 0x4000 | |
147 | ||
148 | static char command_line[COMMAND_LINE_SIZE]; | |
149 | ||
150 | unsigned char __initdata boot_params[PARAM_SIZE]; | |
151 | ||
152 | static struct resource data_resource = { | |
153 | .name = "Kernel data", | |
154 | .start = 0, | |
155 | .end = 0, | |
156 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
157 | }; | |
158 | ||
159 | static struct resource code_resource = { | |
160 | .name = "Kernel code", | |
161 | .start = 0, | |
162 | .end = 0, | |
163 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
164 | }; | |
165 | ||
166 | static struct resource system_rom_resource = { | |
167 | .name = "System ROM", | |
168 | .start = 0xf0000, | |
169 | .end = 0xfffff, | |
170 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
171 | }; | |
172 | ||
173 | static struct resource extension_rom_resource = { | |
174 | .name = "Extension ROM", | |
175 | .start = 0xe0000, | |
176 | .end = 0xeffff, | |
177 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
178 | }; | |
179 | ||
180 | static struct resource adapter_rom_resources[] = { { | |
181 | .name = "Adapter ROM", | |
182 | .start = 0xc8000, | |
183 | .end = 0, | |
184 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
185 | }, { | |
186 | .name = "Adapter ROM", | |
187 | .start = 0, | |
188 | .end = 0, | |
189 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
190 | }, { | |
191 | .name = "Adapter ROM", | |
192 | .start = 0, | |
193 | .end = 0, | |
194 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
195 | }, { | |
196 | .name = "Adapter ROM", | |
197 | .start = 0, | |
198 | .end = 0, | |
199 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
200 | }, { | |
201 | .name = "Adapter ROM", | |
202 | .start = 0, | |
203 | .end = 0, | |
204 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
205 | }, { | |
206 | .name = "Adapter ROM", | |
207 | .start = 0, | |
208 | .end = 0, | |
209 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
210 | } }; | |
211 | ||
212 | #define ADAPTER_ROM_RESOURCES \ | |
213 | (sizeof adapter_rom_resources / sizeof adapter_rom_resources[0]) | |
214 | ||
215 | static struct resource video_rom_resource = { | |
216 | .name = "Video ROM", | |
217 | .start = 0xc0000, | |
218 | .end = 0xc7fff, | |
219 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
220 | }; | |
221 | ||
222 | static struct resource video_ram_resource = { | |
223 | .name = "Video RAM area", | |
224 | .start = 0xa0000, | |
225 | .end = 0xbffff, | |
226 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
227 | }; | |
228 | ||
229 | static struct resource standard_io_resources[] = { { | |
230 | .name = "dma1", | |
231 | .start = 0x0000, | |
232 | .end = 0x001f, | |
233 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
234 | }, { | |
235 | .name = "pic1", | |
236 | .start = 0x0020, | |
237 | .end = 0x0021, | |
238 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
239 | }, { | |
240 | .name = "timer0", | |
241 | .start = 0x0040, | |
242 | .end = 0x0043, | |
243 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
244 | }, { | |
245 | .name = "timer1", | |
246 | .start = 0x0050, | |
247 | .end = 0x0053, | |
248 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
249 | }, { | |
250 | .name = "keyboard", | |
251 | .start = 0x0060, | |
252 | .end = 0x006f, | |
253 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
254 | }, { | |
255 | .name = "dma page reg", | |
256 | .start = 0x0080, | |
257 | .end = 0x008f, | |
258 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
259 | }, { | |
260 | .name = "pic2", | |
261 | .start = 0x00a0, | |
262 | .end = 0x00a1, | |
263 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
264 | }, { | |
265 | .name = "dma2", | |
266 | .start = 0x00c0, | |
267 | .end = 0x00df, | |
268 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
269 | }, { | |
270 | .name = "fpu", | |
271 | .start = 0x00f0, | |
272 | .end = 0x00ff, | |
273 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
274 | } }; | |
275 | ||
276 | #define STANDARD_IO_RESOURCES \ | |
277 | (sizeof standard_io_resources / sizeof standard_io_resources[0]) | |
278 | ||
279 | #define romsignature(x) (*(unsigned short *)(x) == 0xaa55) | |
280 | ||
281 | static int __init romchecksum(unsigned char *rom, unsigned long length) | |
282 | { | |
283 | unsigned char *p, sum = 0; | |
284 | ||
285 | for (p = rom; p < rom + length; p++) | |
286 | sum += *p; | |
287 | return sum == 0; | |
288 | } | |
289 | ||
290 | static void __init probe_roms(void) | |
291 | { | |
292 | unsigned long start, length, upper; | |
293 | unsigned char *rom; | |
294 | int i; | |
295 | ||
296 | /* video rom */ | |
297 | upper = adapter_rom_resources[0].start; | |
298 | for (start = video_rom_resource.start; start < upper; start += 2048) { | |
299 | rom = isa_bus_to_virt(start); | |
300 | if (!romsignature(rom)) | |
301 | continue; | |
302 | ||
303 | video_rom_resource.start = start; | |
304 | ||
305 | /* 0 < length <= 0x7f * 512, historically */ | |
306 | length = rom[2] * 512; | |
307 | ||
308 | /* if checksum okay, trust length byte */ | |
309 | if (length && romchecksum(rom, length)) | |
310 | video_rom_resource.end = start + length - 1; | |
311 | ||
312 | request_resource(&iomem_resource, &video_rom_resource); | |
313 | break; | |
314 | } | |
315 | ||
316 | start = (video_rom_resource.end + 1 + 2047) & ~2047UL; | |
317 | if (start < upper) | |
318 | start = upper; | |
319 | ||
320 | /* system rom */ | |
321 | request_resource(&iomem_resource, &system_rom_resource); | |
322 | upper = system_rom_resource.start; | |
323 | ||
324 | /* check for extension rom (ignore length byte!) */ | |
325 | rom = isa_bus_to_virt(extension_rom_resource.start); | |
326 | if (romsignature(rom)) { | |
327 | length = extension_rom_resource.end - extension_rom_resource.start + 1; | |
328 | if (romchecksum(rom, length)) { | |
329 | request_resource(&iomem_resource, &extension_rom_resource); | |
330 | upper = extension_rom_resource.start; | |
331 | } | |
332 | } | |
333 | ||
334 | /* check for adapter roms on 2k boundaries */ | |
335 | for (i = 0; i < ADAPTER_ROM_RESOURCES && start < upper; start += 2048) { | |
336 | rom = isa_bus_to_virt(start); | |
337 | if (!romsignature(rom)) | |
338 | continue; | |
339 | ||
340 | /* 0 < length <= 0x7f * 512, historically */ | |
341 | length = rom[2] * 512; | |
342 | ||
343 | /* but accept any length that fits if checksum okay */ | |
344 | if (!length || start + length > upper || !romchecksum(rom, length)) | |
345 | continue; | |
346 | ||
347 | adapter_rom_resources[i].start = start; | |
348 | adapter_rom_resources[i].end = start + length - 1; | |
349 | request_resource(&iomem_resource, &adapter_rom_resources[i]); | |
350 | ||
351 | start = adapter_rom_resources[i++].end & ~2047UL; | |
352 | } | |
353 | } | |
354 | ||
355 | static void __init limit_regions(unsigned long long size) | |
356 | { | |
357 | unsigned long long current_addr = 0; | |
358 | int i; | |
359 | ||
360 | if (efi_enabled) { | |
7ae65fd3 MT |
361 | efi_memory_desc_t *md; |
362 | void *p; | |
363 | ||
364 | for (p = memmap.map, i = 0; p < memmap.map_end; | |
365 | p += memmap.desc_size, i++) { | |
366 | md = p; | |
367 | current_addr = md->phys_addr + (md->num_pages << 12); | |
368 | if (md->type == EFI_CONVENTIONAL_MEMORY) { | |
1da177e4 | 369 | if (current_addr >= size) { |
7ae65fd3 | 370 | md->num_pages -= |
1da177e4 LT |
371 | (((current_addr-size) + PAGE_SIZE-1) >> PAGE_SHIFT); |
372 | memmap.nr_map = i + 1; | |
373 | return; | |
374 | } | |
375 | } | |
376 | } | |
377 | } | |
378 | for (i = 0; i < e820.nr_map; i++) { | |
f014a556 DH |
379 | current_addr = e820.map[i].addr + e820.map[i].size; |
380 | if (current_addr < size) | |
381 | continue; | |
382 | ||
383 | if (e820.map[i].type != E820_RAM) | |
384 | continue; | |
385 | ||
386 | if (e820.map[i].addr >= size) { | |
387 | /* | |
388 | * This region starts past the end of the | |
389 | * requested size, skip it completely. | |
390 | */ | |
391 | e820.nr_map = i; | |
392 | } else { | |
393 | e820.nr_map = i + 1; | |
394 | e820.map[i].size -= current_addr - size; | |
1da177e4 | 395 | } |
f014a556 | 396 | return; |
1da177e4 LT |
397 | } |
398 | } | |
399 | ||
e75eac33 JF |
400 | void __init add_memory_region(unsigned long long start, |
401 | unsigned long long size, int type) | |
1da177e4 LT |
402 | { |
403 | int x; | |
404 | ||
405 | if (!efi_enabled) { | |
406 | x = e820.nr_map; | |
407 | ||
408 | if (x == E820MAX) { | |
409 | printk(KERN_ERR "Ooops! Too many entries in the memory map!\n"); | |
410 | return; | |
411 | } | |
412 | ||
413 | e820.map[x].addr = start; | |
414 | e820.map[x].size = size; | |
415 | e820.map[x].type = type; | |
416 | e820.nr_map++; | |
417 | } | |
418 | } /* add_memory_region */ | |
419 | ||
420 | #define E820_DEBUG 1 | |
421 | ||
422 | static void __init print_memory_map(char *who) | |
423 | { | |
424 | int i; | |
425 | ||
426 | for (i = 0; i < e820.nr_map; i++) { | |
427 | printk(" %s: %016Lx - %016Lx ", who, | |
428 | e820.map[i].addr, | |
429 | e820.map[i].addr + e820.map[i].size); | |
430 | switch (e820.map[i].type) { | |
431 | case E820_RAM: printk("(usable)\n"); | |
432 | break; | |
433 | case E820_RESERVED: | |
434 | printk("(reserved)\n"); | |
435 | break; | |
436 | case E820_ACPI: | |
437 | printk("(ACPI data)\n"); | |
438 | break; | |
439 | case E820_NVS: | |
440 | printk("(ACPI NVS)\n"); | |
441 | break; | |
442 | default: printk("type %lu\n", e820.map[i].type); | |
443 | break; | |
444 | } | |
445 | } | |
446 | } | |
447 | ||
448 | /* | |
449 | * Sanitize the BIOS e820 map. | |
450 | * | |
451 | * Some e820 responses include overlapping entries. The following | |
452 | * replaces the original e820 map with a new one, removing overlaps. | |
453 | * | |
454 | */ | |
455 | struct change_member { | |
456 | struct e820entry *pbios; /* pointer to original bios entry */ | |
457 | unsigned long long addr; /* address for this change point */ | |
458 | }; | |
459 | static struct change_member change_point_list[2*E820MAX] __initdata; | |
460 | static struct change_member *change_point[2*E820MAX] __initdata; | |
461 | static struct e820entry *overlap_list[E820MAX] __initdata; | |
462 | static struct e820entry new_bios[E820MAX] __initdata; | |
463 | ||
e75eac33 | 464 | int __init sanitize_e820_map(struct e820entry * biosmap, char * pnr_map) |
1da177e4 LT |
465 | { |
466 | struct change_member *change_tmp; | |
467 | unsigned long current_type, last_type; | |
468 | unsigned long long last_addr; | |
469 | int chgidx, still_changing; | |
470 | int overlap_entries; | |
471 | int new_bios_entry; | |
472 | int old_nr, new_nr, chg_nr; | |
473 | int i; | |
474 | ||
475 | /* | |
476 | Visually we're performing the following (1,2,3,4 = memory types)... | |
477 | ||
478 | Sample memory map (w/overlaps): | |
479 | ____22__________________ | |
480 | ______________________4_ | |
481 | ____1111________________ | |
482 | _44_____________________ | |
483 | 11111111________________ | |
484 | ____________________33__ | |
485 | ___________44___________ | |
486 | __________33333_________ | |
487 | ______________22________ | |
488 | ___________________2222_ | |
489 | _________111111111______ | |
490 | _____________________11_ | |
491 | _________________4______ | |
492 | ||
493 | Sanitized equivalent (no overlap): | |
494 | 1_______________________ | |
495 | _44_____________________ | |
496 | ___1____________________ | |
497 | ____22__________________ | |
498 | ______11________________ | |
499 | _________1______________ | |
500 | __________3_____________ | |
501 | ___________44___________ | |
502 | _____________33_________ | |
503 | _______________2________ | |
504 | ________________1_______ | |
505 | _________________4______ | |
506 | ___________________2____ | |
507 | ____________________33__ | |
508 | ______________________4_ | |
509 | */ | |
510 | ||
511 | /* if there's only one memory region, don't bother */ | |
512 | if (*pnr_map < 2) | |
513 | return -1; | |
514 | ||
515 | old_nr = *pnr_map; | |
516 | ||
517 | /* bail out if we find any unreasonable addresses in bios map */ | |
518 | for (i=0; i<old_nr; i++) | |
519 | if (biosmap[i].addr + biosmap[i].size < biosmap[i].addr) | |
520 | return -1; | |
521 | ||
522 | /* create pointers for initial change-point information (for sorting) */ | |
523 | for (i=0; i < 2*old_nr; i++) | |
524 | change_point[i] = &change_point_list[i]; | |
525 | ||
526 | /* record all known change-points (starting and ending addresses), | |
527 | omitting those that are for empty memory regions */ | |
528 | chgidx = 0; | |
529 | for (i=0; i < old_nr; i++) { | |
530 | if (biosmap[i].size != 0) { | |
531 | change_point[chgidx]->addr = biosmap[i].addr; | |
532 | change_point[chgidx++]->pbios = &biosmap[i]; | |
533 | change_point[chgidx]->addr = biosmap[i].addr + biosmap[i].size; | |
534 | change_point[chgidx++]->pbios = &biosmap[i]; | |
535 | } | |
536 | } | |
537 | chg_nr = chgidx; /* true number of change-points */ | |
538 | ||
539 | /* sort change-point list by memory addresses (low -> high) */ | |
540 | still_changing = 1; | |
541 | while (still_changing) { | |
542 | still_changing = 0; | |
543 | for (i=1; i < chg_nr; i++) { | |
544 | /* if <current_addr> > <last_addr>, swap */ | |
545 | /* or, if current=<start_addr> & last=<end_addr>, swap */ | |
546 | if ((change_point[i]->addr < change_point[i-1]->addr) || | |
547 | ((change_point[i]->addr == change_point[i-1]->addr) && | |
548 | (change_point[i]->addr == change_point[i]->pbios->addr) && | |
549 | (change_point[i-1]->addr != change_point[i-1]->pbios->addr)) | |
550 | ) | |
551 | { | |
552 | change_tmp = change_point[i]; | |
553 | change_point[i] = change_point[i-1]; | |
554 | change_point[i-1] = change_tmp; | |
555 | still_changing=1; | |
556 | } | |
557 | } | |
558 | } | |
559 | ||
560 | /* create a new bios memory map, removing overlaps */ | |
561 | overlap_entries=0; /* number of entries in the overlap table */ | |
562 | new_bios_entry=0; /* index for creating new bios map entries */ | |
563 | last_type = 0; /* start with undefined memory type */ | |
564 | last_addr = 0; /* start with 0 as last starting address */ | |
565 | /* loop through change-points, determining affect on the new bios map */ | |
566 | for (chgidx=0; chgidx < chg_nr; chgidx++) | |
567 | { | |
568 | /* keep track of all overlapping bios entries */ | |
569 | if (change_point[chgidx]->addr == change_point[chgidx]->pbios->addr) | |
570 | { | |
571 | /* add map entry to overlap list (> 1 entry implies an overlap) */ | |
572 | overlap_list[overlap_entries++]=change_point[chgidx]->pbios; | |
573 | } | |
574 | else | |
575 | { | |
576 | /* remove entry from list (order independent, so swap with last) */ | |
577 | for (i=0; i<overlap_entries; i++) | |
578 | { | |
579 | if (overlap_list[i] == change_point[chgidx]->pbios) | |
580 | overlap_list[i] = overlap_list[overlap_entries-1]; | |
581 | } | |
582 | overlap_entries--; | |
583 | } | |
584 | /* if there are overlapping entries, decide which "type" to use */ | |
585 | /* (larger value takes precedence -- 1=usable, 2,3,4,4+=unusable) */ | |
586 | current_type = 0; | |
587 | for (i=0; i<overlap_entries; i++) | |
588 | if (overlap_list[i]->type > current_type) | |
589 | current_type = overlap_list[i]->type; | |
590 | /* continue building up new bios map based on this information */ | |
591 | if (current_type != last_type) { | |
592 | if (last_type != 0) { | |
593 | new_bios[new_bios_entry].size = | |
594 | change_point[chgidx]->addr - last_addr; | |
595 | /* move forward only if the new size was non-zero */ | |
596 | if (new_bios[new_bios_entry].size != 0) | |
597 | if (++new_bios_entry >= E820MAX) | |
598 | break; /* no more space left for new bios entries */ | |
599 | } | |
600 | if (current_type != 0) { | |
601 | new_bios[new_bios_entry].addr = change_point[chgidx]->addr; | |
602 | new_bios[new_bios_entry].type = current_type; | |
603 | last_addr=change_point[chgidx]->addr; | |
604 | } | |
605 | last_type = current_type; | |
606 | } | |
607 | } | |
608 | new_nr = new_bios_entry; /* retain count for new bios entries */ | |
609 | ||
610 | /* copy new bios mapping into original location */ | |
611 | memcpy(biosmap, new_bios, new_nr*sizeof(struct e820entry)); | |
612 | *pnr_map = new_nr; | |
613 | ||
614 | return 0; | |
615 | } | |
616 | ||
617 | /* | |
618 | * Copy the BIOS e820 map into a safe place. | |
619 | * | |
620 | * Sanity-check it while we're at it.. | |
621 | * | |
622 | * If we're lucky and live on a modern system, the setup code | |
623 | * will have given us a memory map that we can use to properly | |
624 | * set up memory. If we aren't, we'll fake a memory map. | |
625 | * | |
626 | * We check to see that the memory map contains at least 2 elements | |
627 | * before we'll use it, because the detection code in setup.S may | |
628 | * not be perfect and most every PC known to man has two memory | |
629 | * regions: one from 0 to 640k, and one from 1mb up. (The IBM | |
630 | * thinkpad 560x, for example, does not cooperate with the memory | |
631 | * detection code.) | |
632 | */ | |
e75eac33 | 633 | int __init copy_e820_map(struct e820entry * biosmap, int nr_map) |
1da177e4 LT |
634 | { |
635 | /* Only one memory region (or negative)? Ignore it */ | |
636 | if (nr_map < 2) | |
637 | return -1; | |
638 | ||
639 | do { | |
640 | unsigned long long start = biosmap->addr; | |
641 | unsigned long long size = biosmap->size; | |
642 | unsigned long long end = start + size; | |
643 | unsigned long type = biosmap->type; | |
644 | ||
645 | /* Overflow in 64 bits? Ignore the memory map. */ | |
646 | if (start > end) | |
647 | return -1; | |
648 | ||
649 | /* | |
650 | * Some BIOSes claim RAM in the 640k - 1M region. | |
651 | * Not right. Fix it up. | |
652 | */ | |
653 | if (type == E820_RAM) { | |
654 | if (start < 0x100000ULL && end > 0xA0000ULL) { | |
655 | if (start < 0xA0000ULL) | |
656 | add_memory_region(start, 0xA0000ULL-start, type); | |
657 | if (end <= 0x100000ULL) | |
658 | continue; | |
659 | start = 0x100000ULL; | |
660 | size = end - start; | |
661 | } | |
662 | } | |
663 | add_memory_region(start, size, type); | |
664 | } while (biosmap++,--nr_map); | |
665 | return 0; | |
666 | } | |
667 | ||
668 | #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE) | |
669 | struct edd edd; | |
670 | #ifdef CONFIG_EDD_MODULE | |
671 | EXPORT_SYMBOL(edd); | |
672 | #endif | |
673 | /** | |
674 | * copy_edd() - Copy the BIOS EDD information | |
675 | * from boot_params into a safe place. | |
676 | * | |
677 | */ | |
678 | static inline void copy_edd(void) | |
679 | { | |
680 | memcpy(edd.mbr_signature, EDD_MBR_SIGNATURE, sizeof(edd.mbr_signature)); | |
681 | memcpy(edd.edd_info, EDD_BUF, sizeof(edd.edd_info)); | |
682 | edd.mbr_signature_nr = EDD_MBR_SIG_NR; | |
683 | edd.edd_info_nr = EDD_NR; | |
684 | } | |
685 | #else | |
686 | static inline void copy_edd(void) | |
687 | { | |
688 | } | |
689 | #endif | |
690 | ||
1a3f239d | 691 | static int __initdata user_defined_memmap = 0; |
1da177e4 | 692 | |
1a3f239d RR |
693 | /* |
694 | * "mem=nopentium" disables the 4MB page tables. | |
695 | * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM | |
696 | * to <mem>, overriding the bios size. | |
697 | * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from | |
698 | * <start> to <start>+<mem>, overriding the bios size. | |
699 | * | |
700 | * HPA tells me bootloaders need to parse mem=, so no new | |
701 | * option should be mem= [also see Documentation/i386/boot.txt] | |
702 | */ | |
703 | static int __init parse_mem(char *arg) | |
704 | { | |
705 | if (!arg) | |
706 | return -EINVAL; | |
1da177e4 | 707 | |
1a3f239d RR |
708 | if (strcmp(arg, "nopentium") == 0) { |
709 | clear_bit(X86_FEATURE_PSE, boot_cpu_data.x86_capability); | |
710 | disable_pse = 1; | |
711 | } else { | |
712 | /* If the user specifies memory size, we | |
713 | * limit the BIOS-provided memory map to | |
714 | * that size. exactmap can be used to specify | |
715 | * the exact map. mem=number can be used to | |
716 | * trim the existing memory map. | |
1da177e4 | 717 | */ |
1a3f239d | 718 | unsigned long long mem_size; |
1da177e4 | 719 | |
1a3f239d RR |
720 | mem_size = memparse(arg, &arg); |
721 | limit_regions(mem_size); | |
722 | user_defined_memmap = 1; | |
723 | } | |
724 | return 0; | |
725 | } | |
726 | early_param("mem", parse_mem); | |
1da177e4 | 727 | |
1a3f239d RR |
728 | static int __init parse_memmap(char *arg) |
729 | { | |
730 | if (!arg) | |
731 | return -EINVAL; | |
1da177e4 | 732 | |
1a3f239d RR |
733 | if (strcmp(arg, "exactmap") == 0) { |
734 | #ifdef CONFIG_CRASH_DUMP | |
735 | /* If we are doing a crash dump, we | |
736 | * still need to know the real mem | |
737 | * size before original memory map is | |
738 | * reset. | |
1da177e4 | 739 | */ |
1a3f239d RR |
740 | find_max_pfn(); |
741 | saved_max_pfn = max_pfn; | |
1da177e4 | 742 | #endif |
1a3f239d RR |
743 | e820.nr_map = 0; |
744 | user_defined_memmap = 1; | |
745 | } else { | |
746 | /* If the user specifies memory size, we | |
747 | * limit the BIOS-provided memory map to | |
748 | * that size. exactmap can be used to specify | |
749 | * the exact map. mem=number can be used to | |
750 | * trim the existing memory map. | |
751 | */ | |
752 | unsigned long long start_at, mem_size; | |
753 | ||
754 | mem_size = memparse(arg, &arg); | |
755 | if (*arg == '@') { | |
756 | start_at = memparse(arg+1, &arg); | |
757 | add_memory_region(start_at, mem_size, E820_RAM); | |
758 | } else if (*arg == '#') { | |
759 | start_at = memparse(arg+1, &arg); | |
760 | add_memory_region(start_at, mem_size, E820_ACPI); | |
761 | } else if (*arg == '$') { | |
762 | start_at = memparse(arg+1, &arg); | |
763 | add_memory_region(start_at, mem_size, E820_RESERVED); | |
764 | } else { | |
765 | limit_regions(mem_size); | |
766 | user_defined_memmap = 1; | |
1da177e4 | 767 | } |
1a3f239d RR |
768 | } |
769 | return 0; | |
770 | } | |
771 | early_param("memmap", parse_memmap); | |
1da177e4 | 772 | |
1a3f239d RR |
773 | #ifdef CONFIG_PROC_VMCORE |
774 | /* elfcorehdr= specifies the location of elf core header | |
775 | * stored by the crashed kernel. | |
776 | */ | |
777 | static int __init parse_elfcorehdr(char *arg) | |
778 | { | |
779 | if (!arg) | |
780 | return -EINVAL; | |
66759a01 | 781 | |
1a3f239d RR |
782 | elfcorehdr_addr = memparse(arg, &arg); |
783 | return 0; | |
784 | } | |
785 | early_param("elfcorehdr", parse_elfcorehdr); | |
786 | #endif /* CONFIG_PROC_VMCORE */ | |
1da177e4 | 787 | |
1a3f239d RR |
788 | /* |
789 | * highmem=size forces highmem to be exactly 'size' bytes. | |
790 | * This works even on boxes that have no highmem otherwise. | |
791 | * This also works to reduce highmem size on bigger boxes. | |
792 | */ | |
793 | static int __init parse_highmem(char *arg) | |
794 | { | |
795 | if (!arg) | |
796 | return -EINVAL; | |
9635b47d | 797 | |
1a3f239d RR |
798 | highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT; |
799 | return 0; | |
800 | } | |
801 | early_param("highmem", parse_highmem); | |
9635b47d | 802 | |
1a3f239d RR |
803 | /* |
804 | * vmalloc=size forces the vmalloc area to be exactly 'size' | |
805 | * bytes. This can be used to increase (or decrease) the | |
806 | * vmalloc area - the default is 128m. | |
807 | */ | |
808 | static int __init parse_vmalloc(char *arg) | |
809 | { | |
810 | if (!arg) | |
811 | return -EINVAL; | |
1bc3b91a | 812 | |
1a3f239d RR |
813 | __VMALLOC_RESERVE = memparse(arg, &arg); |
814 | return 0; | |
1da177e4 | 815 | } |
1a3f239d | 816 | early_param("vmalloc", parse_vmalloc); |
1da177e4 | 817 | |
461a9aff ZA |
818 | /* |
819 | * reservetop=size reserves a hole at the top of the kernel address space which | |
820 | * a hypervisor can load into later. Needed for dynamically loaded hypervisors, | |
821 | * so relocating the fixmap can be done before paging initialization. | |
822 | */ | |
823 | static int __init parse_reservetop(char *arg) | |
824 | { | |
825 | unsigned long address; | |
826 | ||
827 | if (!arg) | |
828 | return -EINVAL; | |
829 | ||
830 | address = memparse(arg, &arg); | |
831 | reserve_top_address(address); | |
832 | return 0; | |
833 | } | |
834 | early_param("reservetop", parse_reservetop); | |
835 | ||
1da177e4 LT |
836 | /* |
837 | * Callback for efi_memory_walk. | |
838 | */ | |
839 | static int __init | |
840 | efi_find_max_pfn(unsigned long start, unsigned long end, void *arg) | |
841 | { | |
842 | unsigned long *max_pfn = arg, pfn; | |
843 | ||
844 | if (start < end) { | |
845 | pfn = PFN_UP(end -1); | |
846 | if (pfn > *max_pfn) | |
847 | *max_pfn = pfn; | |
848 | } | |
849 | return 0; | |
850 | } | |
851 | ||
215c3409 AW |
852 | static int __init |
853 | efi_memory_present_wrapper(unsigned long start, unsigned long end, void *arg) | |
854 | { | |
855 | memory_present(0, start, end); | |
856 | return 0; | |
857 | } | |
1da177e4 | 858 | |
79e453d4 LT |
859 | /* |
860 | * This function checks if the entire range <start,end> is mapped with type. | |
861 | * | |
862 | * Note: this function only works correct if the e820 table is sorted and | |
863 | * not-overlapping, which is the case | |
864 | */ | |
865 | int __init | |
866 | e820_all_mapped(unsigned long s, unsigned long e, unsigned type) | |
867 | { | |
868 | u64 start = s; | |
869 | u64 end = e; | |
870 | int i; | |
871 | for (i = 0; i < e820.nr_map; i++) { | |
872 | struct e820entry *ei = &e820.map[i]; | |
873 | if (type && ei->type != type) | |
874 | continue; | |
875 | /* is the region (part) in overlap with the current region ?*/ | |
876 | if (ei->addr >= end || ei->addr + ei->size <= start) | |
877 | continue; | |
878 | /* if the region is at the beginning of <start,end> we move | |
879 | * start to the end of the region since it's ok until there | |
880 | */ | |
881 | if (ei->addr <= start) | |
882 | start = ei->addr + ei->size; | |
883 | /* if start is now at or beyond end, we're done, full | |
884 | * coverage */ | |
885 | if (start >= end) | |
886 | return 1; /* we're done */ | |
887 | } | |
888 | return 0; | |
889 | } | |
890 | ||
1da177e4 LT |
891 | /* |
892 | * Find the highest page frame number we have available | |
893 | */ | |
894 | void __init find_max_pfn(void) | |
895 | { | |
896 | int i; | |
897 | ||
898 | max_pfn = 0; | |
899 | if (efi_enabled) { | |
900 | efi_memmap_walk(efi_find_max_pfn, &max_pfn); | |
215c3409 | 901 | efi_memmap_walk(efi_memory_present_wrapper, NULL); |
1da177e4 LT |
902 | return; |
903 | } | |
904 | ||
905 | for (i = 0; i < e820.nr_map; i++) { | |
906 | unsigned long start, end; | |
907 | /* RAM? */ | |
908 | if (e820.map[i].type != E820_RAM) | |
909 | continue; | |
910 | start = PFN_UP(e820.map[i].addr); | |
911 | end = PFN_DOWN(e820.map[i].addr + e820.map[i].size); | |
912 | if (start >= end) | |
913 | continue; | |
914 | if (end > max_pfn) | |
915 | max_pfn = end; | |
215c3409 | 916 | memory_present(0, start, end); |
1da177e4 LT |
917 | } |
918 | } | |
919 | ||
920 | /* | |
921 | * Determine low and high memory ranges: | |
922 | */ | |
923 | unsigned long __init find_max_low_pfn(void) | |
924 | { | |
925 | unsigned long max_low_pfn; | |
926 | ||
927 | max_low_pfn = max_pfn; | |
928 | if (max_low_pfn > MAXMEM_PFN) { | |
929 | if (highmem_pages == -1) | |
930 | highmem_pages = max_pfn - MAXMEM_PFN; | |
931 | if (highmem_pages + MAXMEM_PFN < max_pfn) | |
932 | max_pfn = MAXMEM_PFN + highmem_pages; | |
933 | if (highmem_pages + MAXMEM_PFN > max_pfn) { | |
934 | printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn - MAXMEM_PFN), pages_to_mb(highmem_pages)); | |
935 | highmem_pages = 0; | |
936 | } | |
937 | max_low_pfn = MAXMEM_PFN; | |
938 | #ifndef CONFIG_HIGHMEM | |
939 | /* Maximum memory usable is what is directly addressable */ | |
940 | printk(KERN_WARNING "Warning only %ldMB will be used.\n", | |
941 | MAXMEM>>20); | |
942 | if (max_pfn > MAX_NONPAE_PFN) | |
943 | printk(KERN_WARNING "Use a PAE enabled kernel.\n"); | |
944 | else | |
945 | printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n"); | |
946 | max_pfn = MAXMEM_PFN; | |
947 | #else /* !CONFIG_HIGHMEM */ | |
948 | #ifndef CONFIG_X86_PAE | |
949 | if (max_pfn > MAX_NONPAE_PFN) { | |
950 | max_pfn = MAX_NONPAE_PFN; | |
951 | printk(KERN_WARNING "Warning only 4GB will be used.\n"); | |
952 | printk(KERN_WARNING "Use a PAE enabled kernel.\n"); | |
953 | } | |
954 | #endif /* !CONFIG_X86_PAE */ | |
955 | #endif /* !CONFIG_HIGHMEM */ | |
956 | } else { | |
957 | if (highmem_pages == -1) | |
958 | highmem_pages = 0; | |
959 | #ifdef CONFIG_HIGHMEM | |
960 | if (highmem_pages >= max_pfn) { | |
961 | printk(KERN_ERR "highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages), pages_to_mb(max_pfn)); | |
962 | highmem_pages = 0; | |
963 | } | |
964 | if (highmem_pages) { | |
965 | if (max_low_pfn-highmem_pages < 64*1024*1024/PAGE_SIZE){ | |
966 | printk(KERN_ERR "highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages)); | |
967 | highmem_pages = 0; | |
968 | } | |
969 | max_low_pfn -= highmem_pages; | |
970 | } | |
971 | #else | |
972 | if (highmem_pages) | |
973 | printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n"); | |
974 | #endif | |
975 | } | |
976 | return max_low_pfn; | |
977 | } | |
978 | ||
979 | /* | |
980 | * Free all available memory for boot time allocation. Used | |
981 | * as a callback function by efi_memory_walk() | |
982 | */ | |
983 | ||
984 | static int __init | |
985 | free_available_memory(unsigned long start, unsigned long end, void *arg) | |
986 | { | |
987 | /* check max_low_pfn */ | |
23dd842c | 988 | if (start >= (max_low_pfn << PAGE_SHIFT)) |
1da177e4 | 989 | return 0; |
23dd842c TM |
990 | if (end >= (max_low_pfn << PAGE_SHIFT)) |
991 | end = max_low_pfn << PAGE_SHIFT; | |
1da177e4 LT |
992 | if (start < end) |
993 | free_bootmem(start, end - start); | |
994 | ||
995 | return 0; | |
996 | } | |
997 | /* | |
998 | * Register fully available low RAM pages with the bootmem allocator. | |
999 | */ | |
1000 | static void __init register_bootmem_low_pages(unsigned long max_low_pfn) | |
1001 | { | |
1002 | int i; | |
1003 | ||
1004 | if (efi_enabled) { | |
1005 | efi_memmap_walk(free_available_memory, NULL); | |
1006 | return; | |
1007 | } | |
1008 | for (i = 0; i < e820.nr_map; i++) { | |
1009 | unsigned long curr_pfn, last_pfn, size; | |
1010 | /* | |
1011 | * Reserve usable low memory | |
1012 | */ | |
1013 | if (e820.map[i].type != E820_RAM) | |
1014 | continue; | |
1015 | /* | |
1016 | * We are rounding up the start address of usable memory: | |
1017 | */ | |
1018 | curr_pfn = PFN_UP(e820.map[i].addr); | |
1019 | if (curr_pfn >= max_low_pfn) | |
1020 | continue; | |
1021 | /* | |
1022 | * ... and at the end of the usable range downwards: | |
1023 | */ | |
1024 | last_pfn = PFN_DOWN(e820.map[i].addr + e820.map[i].size); | |
1025 | ||
1026 | if (last_pfn > max_low_pfn) | |
1027 | last_pfn = max_low_pfn; | |
1028 | ||
1029 | /* | |
1030 | * .. finally, did all the rounding and playing | |
1031 | * around just make the area go away? | |
1032 | */ | |
1033 | if (last_pfn <= curr_pfn) | |
1034 | continue; | |
1035 | ||
1036 | size = last_pfn - curr_pfn; | |
1037 | free_bootmem(PFN_PHYS(curr_pfn), PFN_PHYS(size)); | |
1038 | } | |
1039 | } | |
1040 | ||
1041 | /* | |
1042 | * workaround for Dell systems that neglect to reserve EBDA | |
1043 | */ | |
1044 | static void __init reserve_ebda_region(void) | |
1045 | { | |
1046 | unsigned int addr; | |
1047 | addr = get_bios_ebda(); | |
1048 | if (addr) | |
1049 | reserve_bootmem(addr, PAGE_SIZE); | |
1050 | } | |
1051 | ||
05b79bdc | 1052 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
1da177e4 LT |
1053 | void __init setup_bootmem_allocator(void); |
1054 | static unsigned long __init setup_memory(void) | |
1055 | { | |
1056 | /* | |
1057 | * partially used pages are not usable - thus | |
1058 | * we are rounding upwards: | |
1059 | */ | |
1060 | min_low_pfn = PFN_UP(init_pg_tables_end); | |
1061 | ||
1062 | find_max_pfn(); | |
1063 | ||
1064 | max_low_pfn = find_max_low_pfn(); | |
1065 | ||
1066 | #ifdef CONFIG_HIGHMEM | |
1067 | highstart_pfn = highend_pfn = max_pfn; | |
1068 | if (max_pfn > max_low_pfn) { | |
1069 | highstart_pfn = max_low_pfn; | |
1070 | } | |
1071 | printk(KERN_NOTICE "%ldMB HIGHMEM available.\n", | |
1072 | pages_to_mb(highend_pfn - highstart_pfn)); | |
ba9c231f JB |
1073 | num_physpages = highend_pfn; |
1074 | high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1; | |
1075 | #else | |
1076 | num_physpages = max_low_pfn; | |
1077 | high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1; | |
1078 | #endif | |
1079 | #ifdef CONFIG_FLATMEM | |
1080 | max_mapnr = num_physpages; | |
1da177e4 LT |
1081 | #endif |
1082 | printk(KERN_NOTICE "%ldMB LOWMEM available.\n", | |
1083 | pages_to_mb(max_low_pfn)); | |
1084 | ||
1085 | setup_bootmem_allocator(); | |
1086 | ||
1087 | return max_low_pfn; | |
1088 | } | |
1089 | ||
1090 | void __init zone_sizes_init(void) | |
1091 | { | |
f06a9684 | 1092 | unsigned long zones_size[MAX_NR_ZONES] = { 0, }; |
1da177e4 LT |
1093 | unsigned int max_dma, low; |
1094 | ||
1095 | max_dma = virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT; | |
1096 | low = max_low_pfn; | |
1097 | ||
1098 | if (low < max_dma) | |
1099 | zones_size[ZONE_DMA] = low; | |
1100 | else { | |
1101 | zones_size[ZONE_DMA] = max_dma; | |
1102 | zones_size[ZONE_NORMAL] = low - max_dma; | |
1103 | #ifdef CONFIG_HIGHMEM | |
1104 | zones_size[ZONE_HIGHMEM] = highend_pfn - low; | |
1105 | #endif | |
1106 | } | |
1107 | free_area_init(zones_size); | |
1108 | } | |
1109 | #else | |
05b79bdc | 1110 | extern unsigned long __init setup_memory(void); |
1da177e4 | 1111 | extern void zone_sizes_init(void); |
05b79bdc | 1112 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
1da177e4 LT |
1113 | |
1114 | void __init setup_bootmem_allocator(void) | |
1115 | { | |
1116 | unsigned long bootmap_size; | |
1117 | /* | |
1118 | * Initialize the boot-time allocator (with low memory only): | |
1119 | */ | |
1120 | bootmap_size = init_bootmem(min_low_pfn, max_low_pfn); | |
1121 | ||
1122 | register_bootmem_low_pages(max_low_pfn); | |
1123 | ||
1124 | /* | |
1125 | * Reserve the bootmem bitmap itself as well. We do this in two | |
1126 | * steps (first step was init_bootmem()) because this catches | |
1127 | * the (very unlikely) case of us accidentally initializing the | |
1128 | * bootmem allocator with an invalid RAM area. | |
1129 | */ | |
8a919085 VG |
1130 | reserve_bootmem(__PHYSICAL_START, (PFN_PHYS(min_low_pfn) + |
1131 | bootmap_size + PAGE_SIZE-1) - (__PHYSICAL_START)); | |
1da177e4 LT |
1132 | |
1133 | /* | |
1134 | * reserve physical page 0 - it's a special BIOS page on many boxes, | |
1135 | * enabling clean reboots, SMP operation, laptop functions. | |
1136 | */ | |
1137 | reserve_bootmem(0, PAGE_SIZE); | |
1138 | ||
1139 | /* reserve EBDA region, it's a 4K region */ | |
1140 | reserve_ebda_region(); | |
1141 | ||
1142 | /* could be an AMD 768MPX chipset. Reserve a page before VGA to prevent | |
1143 | PCI prefetch into it (errata #56). Usually the page is reserved anyways, | |
1144 | unless you have no PS/2 mouse plugged in. */ | |
1145 | if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD && | |
1146 | boot_cpu_data.x86 == 6) | |
1147 | reserve_bootmem(0xa0000 - 4096, 4096); | |
1148 | ||
1149 | #ifdef CONFIG_SMP | |
1150 | /* | |
1151 | * But first pinch a few for the stack/trampoline stuff | |
1152 | * FIXME: Don't need the extra page at 4K, but need to fix | |
1153 | * trampoline before removing it. (see the GDT stuff) | |
1154 | */ | |
1155 | reserve_bootmem(PAGE_SIZE, PAGE_SIZE); | |
1156 | #endif | |
1157 | #ifdef CONFIG_ACPI_SLEEP | |
1158 | /* | |
1159 | * Reserve low memory region for sleep support. | |
1160 | */ | |
1161 | acpi_reserve_bootmem(); | |
1162 | #endif | |
1163 | #ifdef CONFIG_X86_FIND_SMP_CONFIG | |
1164 | /* | |
1165 | * Find and reserve possible boot-time SMP configuration: | |
1166 | */ | |
1167 | find_smp_config(); | |
1168 | #endif | |
91023300 | 1169 | numa_kva_reserve(); |
1da177e4 LT |
1170 | #ifdef CONFIG_BLK_DEV_INITRD |
1171 | if (LOADER_TYPE && INITRD_START) { | |
1172 | if (INITRD_START + INITRD_SIZE <= (max_low_pfn << PAGE_SHIFT)) { | |
1173 | reserve_bootmem(INITRD_START, INITRD_SIZE); | |
1174 | initrd_start = | |
1175 | INITRD_START ? INITRD_START + PAGE_OFFSET : 0; | |
1176 | initrd_end = initrd_start+INITRD_SIZE; | |
1177 | } | |
1178 | else { | |
1179 | printk(KERN_ERR "initrd extends beyond end of memory " | |
1180 | "(0x%08lx > 0x%08lx)\ndisabling initrd\n", | |
1181 | INITRD_START + INITRD_SIZE, | |
1182 | max_low_pfn << PAGE_SHIFT); | |
1183 | initrd_start = 0; | |
1184 | } | |
1185 | } | |
1186 | #endif | |
1bc3b91a EB |
1187 | #ifdef CONFIG_KEXEC |
1188 | if (crashk_res.start != crashk_res.end) | |
1189 | reserve_bootmem(crashk_res.start, | |
1190 | crashk_res.end - crashk_res.start + 1); | |
1191 | #endif | |
1da177e4 LT |
1192 | } |
1193 | ||
1194 | /* | |
1195 | * The node 0 pgdat is initialized before all of these because | |
1196 | * it's needed for bootmem. node>0 pgdats have their virtual | |
1197 | * space allocated before the pagetables are in place to access | |
1198 | * them, so they can't be cleared then. | |
1199 | * | |
1200 | * This should all compile down to nothing when NUMA is off. | |
1201 | */ | |
1202 | void __init remapped_pgdat_init(void) | |
1203 | { | |
1204 | int nid; | |
1205 | ||
1206 | for_each_online_node(nid) { | |
1207 | if (nid != 0) | |
1208 | memset(NODE_DATA(nid), 0, sizeof(struct pglist_data)); | |
1209 | } | |
1210 | } | |
1211 | ||
1212 | /* | |
1213 | * Request address space for all standard RAM and ROM resources | |
1214 | * and also for regions reported as reserved by the e820. | |
1215 | */ | |
1216 | static void __init | |
1217 | legacy_init_iomem_resources(struct resource *code_resource, struct resource *data_resource) | |
1218 | { | |
1219 | int i; | |
1220 | ||
1221 | probe_roms(); | |
1222 | for (i = 0; i < e820.nr_map; i++) { | |
1223 | struct resource *res; | |
87937472 | 1224 | #ifndef CONFIG_RESOURCES_64BIT |
9be2f7c3 LT |
1225 | if (e820.map[i].addr + e820.map[i].size > 0x100000000ULL) |
1226 | continue; | |
87937472 | 1227 | #endif |
b408cbc7 | 1228 | res = kzalloc(sizeof(struct resource), GFP_ATOMIC); |
1da177e4 LT |
1229 | switch (e820.map[i].type) { |
1230 | case E820_RAM: res->name = "System RAM"; break; | |
1231 | case E820_ACPI: res->name = "ACPI Tables"; break; | |
1232 | case E820_NVS: res->name = "ACPI Non-volatile Storage"; break; | |
1233 | default: res->name = "reserved"; | |
1234 | } | |
1235 | res->start = e820.map[i].addr; | |
1236 | res->end = res->start + e820.map[i].size - 1; | |
1237 | res->flags = IORESOURCE_MEM | IORESOURCE_BUSY; | |
a7546075 CM |
1238 | if (request_resource(&iomem_resource, res)) { |
1239 | kfree(res); | |
1240 | continue; | |
1241 | } | |
1da177e4 LT |
1242 | if (e820.map[i].type == E820_RAM) { |
1243 | /* | |
1244 | * We don't know which RAM region contains kernel data, | |
1245 | * so we try it repeatedly and let the resource manager | |
1246 | * test it. | |
1247 | */ | |
1248 | request_resource(res, code_resource); | |
1249 | request_resource(res, data_resource); | |
1bc3b91a EB |
1250 | #ifdef CONFIG_KEXEC |
1251 | request_resource(res, &crashk_res); | |
1252 | #endif | |
1da177e4 LT |
1253 | } |
1254 | } | |
1255 | } | |
1256 | ||
1257 | /* | |
1258 | * Request address space for all standard resources | |
b408cbc7 | 1259 | * |
66004a6c LT |
1260 | * This is called just before pcibios_init(), which is also a |
1261 | * subsys_initcall, but is linked in later (in arch/i386/pci/common.c). | |
1da177e4 | 1262 | */ |
b408cbc7 | 1263 | static int __init request_standard_resources(void) |
1da177e4 | 1264 | { |
b408cbc7 | 1265 | int i; |
1da177e4 | 1266 | |
b408cbc7 | 1267 | printk("Setting up standard PCI resources\n"); |
1da177e4 LT |
1268 | if (efi_enabled) |
1269 | efi_initialize_iomem_resources(&code_resource, &data_resource); | |
1270 | else | |
1271 | legacy_init_iomem_resources(&code_resource, &data_resource); | |
1272 | ||
1273 | /* EFI systems may still have VGA */ | |
1274 | request_resource(&iomem_resource, &video_ram_resource); | |
1275 | ||
1276 | /* request I/O space for devices used on all i[345]86 PCs */ | |
1277 | for (i = 0; i < STANDARD_IO_RESOURCES; i++) | |
1278 | request_resource(&ioport_resource, &standard_io_resources[i]); | |
b408cbc7 LT |
1279 | return 0; |
1280 | } | |
1281 | ||
66004a6c | 1282 | subsys_initcall(request_standard_resources); |
b408cbc7 LT |
1283 | |
1284 | static void __init register_memory(void) | |
1285 | { | |
1286 | unsigned long gapstart, gapsize, round; | |
1287 | unsigned long long last; | |
1288 | int i; | |
1da177e4 LT |
1289 | |
1290 | /* | |
1291 | * Search for the bigest gap in the low 32 bits of the e820 | |
1292 | * memory space. | |
1293 | */ | |
1294 | last = 0x100000000ull; | |
1295 | gapstart = 0x10000000; | |
1296 | gapsize = 0x400000; | |
1297 | i = e820.nr_map; | |
1298 | while (--i >= 0) { | |
1299 | unsigned long long start = e820.map[i].addr; | |
1300 | unsigned long long end = start + e820.map[i].size; | |
1301 | ||
1302 | /* | |
1303 | * Since "last" is at most 4GB, we know we'll | |
1304 | * fit in 32 bits if this condition is true | |
1305 | */ | |
1306 | if (last > end) { | |
1307 | unsigned long gap = last - end; | |
1308 | ||
1309 | if (gap > gapsize) { | |
1310 | gapsize = gap; | |
1311 | gapstart = end; | |
1312 | } | |
1313 | } | |
1314 | if (start < last) | |
1315 | last = start; | |
1316 | } | |
1317 | ||
1318 | /* | |
f0eca962 DR |
1319 | * See how much we want to round up: start off with |
1320 | * rounding to the next 1MB area. | |
1da177e4 | 1321 | */ |
f0eca962 DR |
1322 | round = 0x100000; |
1323 | while ((gapsize >> 4) > round) | |
1324 | round += round; | |
1325 | /* Fun with two's complement */ | |
1326 | pci_mem_start = (gapstart + round) & -round; | |
1da177e4 LT |
1327 | |
1328 | printk("Allocating PCI resources starting at %08lx (gap: %08lx:%08lx)\n", | |
1329 | pci_mem_start, gapstart, gapsize); | |
1330 | } | |
1331 | ||
1da177e4 LT |
1332 | #ifdef CONFIG_MCA |
1333 | static void set_mca_bus(int x) | |
1334 | { | |
1335 | MCA_bus = x; | |
1336 | } | |
1337 | #else | |
1338 | static void set_mca_bus(int x) { } | |
1339 | #endif | |
1340 | ||
1341 | /* | |
1342 | * Determine if we were loaded by an EFI loader. If so, then we have also been | |
1343 | * passed the efi memmap, systab, etc., so we should use these data structures | |
1344 | * for initialization. Note, the efi init code path is determined by the | |
1345 | * global efi_enabled. This allows the same kernel image to be used on existing | |
1346 | * systems (with a traditional BIOS) as well as on EFI systems. | |
1347 | */ | |
1348 | void __init setup_arch(char **cmdline_p) | |
1349 | { | |
1350 | unsigned long max_low_pfn; | |
1351 | ||
1352 | memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data)); | |
1353 | pre_setup_arch_hook(); | |
1354 | early_cpu_init(); | |
1355 | ||
1356 | /* | |
1357 | * FIXME: This isn't an official loader_type right | |
1358 | * now but does currently work with elilo. | |
1359 | * If we were configured as an EFI kernel, check to make | |
1360 | * sure that we were loaded correctly from elilo and that | |
1361 | * the system table is valid. If not, then initialize normally. | |
1362 | */ | |
1363 | #ifdef CONFIG_EFI | |
1364 | if ((LOADER_TYPE == 0x50) && EFI_SYSTAB) | |
1365 | efi_enabled = 1; | |
1366 | #endif | |
1367 | ||
1368 | ROOT_DEV = old_decode_dev(ORIG_ROOT_DEV); | |
1369 | drive_info = DRIVE_INFO; | |
1370 | screen_info = SCREEN_INFO; | |
1371 | edid_info = EDID_INFO; | |
1372 | apm_info.bios = APM_BIOS_INFO; | |
1373 | ist_info = IST_INFO; | |
1374 | saved_videomode = VIDEO_MODE; | |
1375 | if( SYS_DESC_TABLE.length != 0 ) { | |
1376 | set_mca_bus(SYS_DESC_TABLE.table[3] & 0x2); | |
1377 | machine_id = SYS_DESC_TABLE.table[0]; | |
1378 | machine_submodel_id = SYS_DESC_TABLE.table[1]; | |
1379 | BIOS_revision = SYS_DESC_TABLE.table[2]; | |
1380 | } | |
1381 | bootloader_type = LOADER_TYPE; | |
1382 | ||
1383 | #ifdef CONFIG_BLK_DEV_RAM | |
1384 | rd_image_start = RAMDISK_FLAGS & RAMDISK_IMAGE_START_MASK; | |
1385 | rd_prompt = ((RAMDISK_FLAGS & RAMDISK_PROMPT_FLAG) != 0); | |
1386 | rd_doload = ((RAMDISK_FLAGS & RAMDISK_LOAD_FLAG) != 0); | |
1387 | #endif | |
1388 | ARCH_SETUP | |
1389 | if (efi_enabled) | |
1390 | efi_init(); | |
1391 | else { | |
1392 | printk(KERN_INFO "BIOS-provided physical RAM map:\n"); | |
1393 | print_memory_map(machine_specific_memory_setup()); | |
1394 | } | |
1395 | ||
1396 | copy_edd(); | |
1397 | ||
1398 | if (!MOUNT_ROOT_RDONLY) | |
1399 | root_mountflags &= ~MS_RDONLY; | |
1400 | init_mm.start_code = (unsigned long) _text; | |
1401 | init_mm.end_code = (unsigned long) _etext; | |
1402 | init_mm.end_data = (unsigned long) _edata; | |
1403 | init_mm.brk = init_pg_tables_end + PAGE_OFFSET; | |
1404 | ||
1405 | code_resource.start = virt_to_phys(_text); | |
1406 | code_resource.end = virt_to_phys(_etext)-1; | |
1407 | data_resource.start = virt_to_phys(_etext); | |
1408 | data_resource.end = virt_to_phys(_edata)-1; | |
1409 | ||
1a3f239d | 1410 | parse_early_param(); |
1da177e4 | 1411 | |
1a3f239d RR |
1412 | if (user_defined_memmap) { |
1413 | printk(KERN_INFO "user-defined physical RAM map:\n"); | |
1414 | print_memory_map("user"); | |
99b7de33 | 1415 | } |
1a3f239d RR |
1416 | |
1417 | strlcpy(command_line, saved_command_line, COMMAND_LINE_SIZE); | |
1418 | *cmdline_p = command_line; | |
99b7de33 | 1419 | |
1da177e4 LT |
1420 | max_low_pfn = setup_memory(); |
1421 | ||
1422 | /* | |
1423 | * NOTE: before this point _nobody_ is allowed to allocate | |
1424 | * any memory using the bootmem allocator. Although the | |
1425 | * alloctor is now initialised only the first 8Mb of the kernel | |
1426 | * virtual address space has been mapped. All allocations before | |
1427 | * paging_init() has completed must use the alloc_bootmem_low_pages() | |
1428 | * variant (which allocates DMA'able memory) and care must be taken | |
1429 | * not to exceed the 8Mb limit. | |
1430 | */ | |
1431 | ||
1432 | #ifdef CONFIG_SMP | |
1433 | smp_alloc_memory(); /* AP processor realmode stacks in low memory*/ | |
1434 | #endif | |
1435 | paging_init(); | |
1436 | remapped_pgdat_init(); | |
05b79bdc | 1437 | sparse_init(); |
1da177e4 LT |
1438 | zone_sizes_init(); |
1439 | ||
1440 | /* | |
1441 | * NOTE: at this point the bootmem allocator is fully available. | |
1442 | */ | |
1443 | ||
1da177e4 LT |
1444 | dmi_scan_machine(); |
1445 | ||
1446 | #ifdef CONFIG_X86_GENERICARCH | |
1a3f239d | 1447 | generic_apic_probe(); |
1da177e4 LT |
1448 | #endif |
1449 | if (efi_enabled) | |
1450 | efi_map_memmap(); | |
1451 | ||
888ba6c6 | 1452 | #ifdef CONFIG_ACPI |
1da177e4 LT |
1453 | /* |
1454 | * Parse the ACPI tables for possible boot-time SMP configuration. | |
1455 | */ | |
1456 | acpi_boot_table_init(); | |
d44647b0 AC |
1457 | #endif |
1458 | ||
f157cbb1 | 1459 | #ifdef CONFIG_PCI |
d44647b0 AC |
1460 | #ifdef CONFIG_X86_IO_APIC |
1461 | check_acpi_pci(); /* Checks more than just ACPI actually */ | |
1462 | #endif | |
f157cbb1 | 1463 | #endif |
d44647b0 AC |
1464 | |
1465 | #ifdef CONFIG_ACPI | |
1da177e4 LT |
1466 | acpi_boot_init(); |
1467 | ||
911a62d4 VP |
1468 | #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC) |
1469 | if (def_to_bigsmp) | |
1470 | printk(KERN_WARNING "More than 8 CPUs detected and " | |
1471 | "CONFIG_X86_PC cannot handle it.\nUse " | |
1472 | "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n"); | |
1473 | #endif | |
1474 | #endif | |
1da177e4 LT |
1475 | #ifdef CONFIG_X86_LOCAL_APIC |
1476 | if (smp_found_config) | |
1477 | get_smp_config(); | |
1478 | #endif | |
1479 | ||
1480 | register_memory(); | |
1481 | ||
1482 | #ifdef CONFIG_VT | |
1483 | #if defined(CONFIG_VGA_CONSOLE) | |
1484 | if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY)) | |
1485 | conswitchp = &vga_con; | |
1486 | #elif defined(CONFIG_DUMMY_CONSOLE) | |
1487 | conswitchp = &dummy_con; | |
1488 | #endif | |
1489 | #endif | |
539eb11e | 1490 | tsc_init(); |
1da177e4 LT |
1491 | } |
1492 | ||
e5c6c8e4 MN |
1493 | static __init int add_pcspkr(void) |
1494 | { | |
1495 | struct platform_device *pd; | |
1496 | int ret; | |
1497 | ||
1498 | pd = platform_device_alloc("pcspkr", -1); | |
1499 | if (!pd) | |
1500 | return -ENOMEM; | |
1501 | ||
1502 | ret = platform_device_add(pd); | |
1503 | if (ret) | |
1504 | platform_device_put(pd); | |
1505 | ||
1506 | return ret; | |
1507 | } | |
1508 | device_initcall(add_pcspkr); | |
1509 | ||
1da177e4 LT |
1510 | /* |
1511 | * Local Variables: | |
1512 | * mode:c | |
1513 | * c-file-style:"k&r" | |
1514 | * c-basic-offset:8 | |
1515 | * End: | |
1516 | */ |