setup.c 28 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181
  1. /*
  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. * This file handles the architecture-dependent parts of initialization
  21. */
  22. #include <linux/sched.h>
  23. #include <linux/mm.h>
  24. #include <linux/mmzone.h>
  25. #include <linux/screen_info.h>
  26. #include <linux/ioport.h>
  27. #include <linux/acpi.h>
  28. #include <linux/sfi.h>
  29. #include <linux/apm_bios.h>
  30. #include <linux/initrd.h>
  31. #include <linux/bootmem.h>
  32. #include <linux/memblock.h>
  33. #include <linux/seq_file.h>
  34. #include <linux/console.h>
  35. #include <linux/root_dev.h>
  36. #include <linux/highmem.h>
  37. #include <linux/module.h>
  38. #include <linux/efi.h>
  39. #include <linux/init.h>
  40. #include <linux/edd.h>
  41. #include <linux/iscsi_ibft.h>
  42. #include <linux/nodemask.h>
  43. #include <linux/kexec.h>
  44. #include <linux/dmi.h>
  45. #include <linux/pfn.h>
  46. #include <linux/pci.h>
  47. #include <asm/pci-direct.h>
  48. #include <linux/init_ohci1394_dma.h>
  49. #include <linux/kvm_para.h>
  50. #include <linux/dma-contiguous.h>
  51. #include <linux/errno.h>
  52. #include <linux/kernel.h>
  53. #include <linux/stddef.h>
  54. #include <linux/unistd.h>
  55. #include <linux/ptrace.h>
  56. #include <linux/user.h>
  57. #include <linux/delay.h>
  58. #include <linux/kallsyms.h>
  59. #include <linux/cpufreq.h>
  60. #include <linux/dma-mapping.h>
  61. #include <linux/ctype.h>
  62. #include <linux/uaccess.h>
  63. #include <linux/percpu.h>
  64. #include <linux/crash_dump.h>
  65. #include <linux/tboot.h>
  66. #include <linux/jiffies.h>
  67. #include <video/edid.h>
  68. #include <asm/mtrr.h>
  69. #include <asm/apic.h>
  70. #include <asm/realmode.h>
  71. #include <asm/e820.h>
  72. #include <asm/mpspec.h>
  73. #include <asm/setup.h>
  74. #include <asm/efi.h>
  75. #include <asm/timer.h>
  76. #include <asm/i8259.h>
  77. #include <asm/sections.h>
  78. #include <asm/dmi.h>
  79. #include <asm/io_apic.h>
  80. #include <asm/ist.h>
  81. #include <asm/setup_arch.h>
  82. #include <asm/bios_ebda.h>
  83. #include <asm/cacheflush.h>
  84. #include <asm/processor.h>
  85. #include <asm/bugs.h>
  86. #include <asm/vsyscall.h>
  87. #include <asm/cpu.h>
  88. #include <asm/desc.h>
  89. #include <asm/dma.h>
  90. #include <asm/iommu.h>
  91. #include <asm/gart.h>
  92. #include <asm/mmu_context.h>
  93. #include <asm/proto.h>
  94. #include <asm/paravirt.h>
  95. #include <asm/hypervisor.h>
  96. #include <asm/olpc_ofw.h>
  97. #include <asm/percpu.h>
  98. #include <asm/topology.h>
  99. #include <asm/apicdef.h>
  100. #include <asm/amd_nb.h>
  101. #include <asm/mce.h>
  102. #include <asm/alternative.h>
  103. #include <asm/prom.h>
  104. /*
  105. * max_low_pfn_mapped: highest direct mapped pfn under 4GB
  106. * max_pfn_mapped: highest direct mapped pfn over 4GB
  107. *
  108. * The direct mapping only covers E820_RAM regions, so the ranges and gaps are
  109. * represented by pfn_mapped
  110. */
  111. unsigned long max_low_pfn_mapped;
  112. unsigned long max_pfn_mapped;
  113. #ifdef CONFIG_DMI
  114. RESERVE_BRK(dmi_alloc, 65536);
  115. #endif
  116. static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
  117. unsigned long _brk_end = (unsigned long)__brk_base;
  118. #ifdef CONFIG_X86_64
  119. int default_cpu_present_to_apicid(int mps_cpu)
  120. {
  121. return __default_cpu_present_to_apicid(mps_cpu);
  122. }
  123. int default_check_phys_apicid_present(int phys_apicid)
  124. {
  125. return __default_check_phys_apicid_present(phys_apicid);
  126. }
  127. #endif
  128. struct boot_params boot_params;
  129. /*
  130. * Machine setup..
  131. */
  132. static struct resource data_resource = {
  133. .name = "Kernel data",
  134. .start = 0,
  135. .end = 0,
  136. .flags = IORESOURCE_BUSY | IORESOURCE_MEM
  137. };
  138. static struct resource code_resource = {
  139. .name = "Kernel code",
  140. .start = 0,
  141. .end = 0,
  142. .flags = IORESOURCE_BUSY | IORESOURCE_MEM
  143. };
  144. static struct resource bss_resource = {
  145. .name = "Kernel bss",
  146. .start = 0,
  147. .end = 0,
  148. .flags = IORESOURCE_BUSY | IORESOURCE_MEM
  149. };
  150. #ifdef CONFIG_X86_32
  151. /* cpu data as detected by the assembly code in head.S */
  152. struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
  153. /* common cpu data for all cpus */
  154. struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
  155. EXPORT_SYMBOL(boot_cpu_data);
  156. unsigned int def_to_bigsmp;
  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. struct apm_info apm_info;
  162. EXPORT_SYMBOL(apm_info);
  163. #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
  164. defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
  165. struct ist_info ist_info;
  166. EXPORT_SYMBOL(ist_info);
  167. #else
  168. struct ist_info ist_info;
  169. #endif
  170. #else
  171. struct cpuinfo_x86 boot_cpu_data __read_mostly = {
  172. .x86_phys_bits = MAX_PHYSMEM_BITS,
  173. };
  174. EXPORT_SYMBOL(boot_cpu_data);
  175. #endif
  176. #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
  177. unsigned long mmu_cr4_features;
  178. #else
  179. unsigned long mmu_cr4_features = X86_CR4_PAE;
  180. #endif
  181. /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
  182. int bootloader_type, bootloader_version;
  183. /*
  184. * Setup options
  185. */
  186. struct screen_info screen_info;
  187. EXPORT_SYMBOL(screen_info);
  188. struct edid_info edid_info;
  189. EXPORT_SYMBOL_GPL(edid_info);
  190. extern int root_mountflags;
  191. unsigned long saved_video_mode;
  192. #define RAMDISK_IMAGE_START_MASK 0x07FF
  193. #define RAMDISK_PROMPT_FLAG 0x8000
  194. #define RAMDISK_LOAD_FLAG 0x4000
  195. static char __initdata command_line[COMMAND_LINE_SIZE];
  196. #ifdef CONFIG_CMDLINE_BOOL
  197. static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
  198. #endif
  199. #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
  200. struct edd edd;
  201. #ifdef CONFIG_EDD_MODULE
  202. EXPORT_SYMBOL(edd);
  203. #endif
  204. /**
  205. * copy_edd() - Copy the BIOS EDD information
  206. * from boot_params into a safe place.
  207. *
  208. */
  209. static inline void __init copy_edd(void)
  210. {
  211. memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
  212. sizeof(edd.mbr_signature));
  213. memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
  214. edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
  215. edd.edd_info_nr = boot_params.eddbuf_entries;
  216. }
  217. #else
  218. static inline void __init copy_edd(void)
  219. {
  220. }
  221. #endif
  222. void * __init extend_brk(size_t size, size_t align)
  223. {
  224. size_t mask = align - 1;
  225. void *ret;
  226. BUG_ON(_brk_start == 0);
  227. BUG_ON(align & mask);
  228. _brk_end = (_brk_end + mask) & ~mask;
  229. BUG_ON((char *)(_brk_end + size) > __brk_limit);
  230. ret = (void *)_brk_end;
  231. _brk_end += size;
  232. memset(ret, 0, size);
  233. return ret;
  234. }
  235. #ifdef CONFIG_X86_32
  236. static void __init cleanup_highmap(void)
  237. {
  238. }
  239. #endif
  240. static void __init reserve_brk(void)
  241. {
  242. if (_brk_end > _brk_start)
  243. memblock_reserve(__pa(_brk_start),
  244. __pa(_brk_end) - __pa(_brk_start));
  245. /* Mark brk area as locked down and no longer taking any
  246. new allocations */
  247. _brk_start = 0;
  248. }
  249. #ifdef CONFIG_BLK_DEV_INITRD
  250. static u64 __init get_ramdisk_image(void)
  251. {
  252. u64 ramdisk_image = boot_params.hdr.ramdisk_image;
  253. ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
  254. return ramdisk_image;
  255. }
  256. static u64 __init get_ramdisk_size(void)
  257. {
  258. u64 ramdisk_size = boot_params.hdr.ramdisk_size;
  259. ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
  260. return ramdisk_size;
  261. }
  262. #define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT)
  263. static void __init relocate_initrd(void)
  264. {
  265. /* Assume only end is not page aligned */
  266. u64 ramdisk_image = get_ramdisk_image();
  267. u64 ramdisk_size = get_ramdisk_size();
  268. u64 area_size = PAGE_ALIGN(ramdisk_size);
  269. u64 ramdisk_here;
  270. unsigned long slop, clen, mapaddr;
  271. char *p, *q;
  272. /* We need to move the initrd down into directly mapped mem */
  273. ramdisk_here = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
  274. area_size, PAGE_SIZE);
  275. if (!ramdisk_here)
  276. panic("Cannot find place for new RAMDISK of size %lld\n",
  277. ramdisk_size);
  278. /* Note: this includes all the mem currently occupied by
  279. the initrd, we rely on that fact to keep the data intact. */
  280. memblock_reserve(ramdisk_here, area_size);
  281. initrd_start = ramdisk_here + PAGE_OFFSET;
  282. initrd_end = initrd_start + ramdisk_size;
  283. printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
  284. ramdisk_here, ramdisk_here + ramdisk_size - 1);
  285. q = (char *)initrd_start;
  286. /* Copy the initrd */
  287. while (ramdisk_size) {
  288. slop = ramdisk_image & ~PAGE_MASK;
  289. clen = ramdisk_size;
  290. if (clen > MAX_MAP_CHUNK-slop)
  291. clen = MAX_MAP_CHUNK-slop;
  292. mapaddr = ramdisk_image & PAGE_MASK;
  293. p = early_memremap(mapaddr, clen+slop);
  294. memcpy(q, p+slop, clen);
  295. early_iounmap(p, clen+slop);
  296. q += clen;
  297. ramdisk_image += clen;
  298. ramdisk_size -= clen;
  299. }
  300. ramdisk_image = get_ramdisk_image();
  301. ramdisk_size = get_ramdisk_size();
  302. printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
  303. " [mem %#010llx-%#010llx]\n",
  304. ramdisk_image, ramdisk_image + ramdisk_size - 1,
  305. ramdisk_here, ramdisk_here + ramdisk_size - 1);
  306. }
  307. static u64 __init get_mem_size(unsigned long limit_pfn)
  308. {
  309. int i;
  310. u64 mapped_pages = 0;
  311. unsigned long start_pfn, end_pfn;
  312. for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn, NULL) {
  313. start_pfn = min_t(unsigned long, start_pfn, limit_pfn);
  314. end_pfn = min_t(unsigned long, end_pfn, limit_pfn);
  315. mapped_pages += end_pfn - start_pfn;
  316. }
  317. return mapped_pages << PAGE_SHIFT;
  318. }
  319. static void __init early_reserve_initrd(void)
  320. {
  321. /* Assume only end is not page aligned */
  322. u64 ramdisk_image = get_ramdisk_image();
  323. u64 ramdisk_size = get_ramdisk_size();
  324. u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
  325. if (!boot_params.hdr.type_of_loader ||
  326. !ramdisk_image || !ramdisk_size)
  327. return; /* No initrd provided by bootloader */
  328. memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
  329. }
  330. static void __init reserve_initrd(void)
  331. {
  332. /* Assume only end is not page aligned */
  333. u64 ramdisk_image = get_ramdisk_image();
  334. u64 ramdisk_size = get_ramdisk_size();
  335. u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
  336. u64 mapped_size;
  337. if (!boot_params.hdr.type_of_loader ||
  338. !ramdisk_image || !ramdisk_size)
  339. return; /* No initrd provided by bootloader */
  340. initrd_start = 0;
  341. mapped_size = get_mem_size(max_pfn_mapped);
  342. if (ramdisk_size >= (mapped_size>>1))
  343. panic("initrd too large to handle, "
  344. "disabling initrd (%lld needed, %lld available)\n",
  345. ramdisk_size, mapped_size>>1);
  346. printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
  347. ramdisk_end - 1);
  348. if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
  349. PFN_DOWN(ramdisk_end))) {
  350. /* All are mapped, easy case */
  351. initrd_start = ramdisk_image + PAGE_OFFSET;
  352. initrd_end = initrd_start + ramdisk_size;
  353. return;
  354. }
  355. relocate_initrd();
  356. memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
  357. }
  358. #else
  359. static void __init early_reserve_initrd(void)
  360. {
  361. }
  362. static void __init reserve_initrd(void)
  363. {
  364. }
  365. #endif /* CONFIG_BLK_DEV_INITRD */
  366. static void __init parse_setup_data(void)
  367. {
  368. struct setup_data *data;
  369. u64 pa_data;
  370. pa_data = boot_params.hdr.setup_data;
  371. while (pa_data) {
  372. u32 data_len, map_len;
  373. map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
  374. (u64)sizeof(struct setup_data));
  375. data = early_memremap(pa_data, map_len);
  376. data_len = data->len + sizeof(struct setup_data);
  377. if (data_len > map_len) {
  378. early_iounmap(data, map_len);
  379. data = early_memremap(pa_data, data_len);
  380. map_len = data_len;
  381. }
  382. switch (data->type) {
  383. case SETUP_E820_EXT:
  384. parse_e820_ext(data);
  385. break;
  386. case SETUP_DTB:
  387. add_dtb(pa_data);
  388. break;
  389. default:
  390. break;
  391. }
  392. pa_data = data->next;
  393. early_iounmap(data, map_len);
  394. }
  395. }
  396. static void __init e820_reserve_setup_data(void)
  397. {
  398. struct setup_data *data;
  399. u64 pa_data;
  400. int found = 0;
  401. pa_data = boot_params.hdr.setup_data;
  402. while (pa_data) {
  403. data = early_memremap(pa_data, sizeof(*data));
  404. e820_update_range(pa_data, sizeof(*data)+data->len,
  405. E820_RAM, E820_RESERVED_KERN);
  406. found = 1;
  407. pa_data = data->next;
  408. early_iounmap(data, sizeof(*data));
  409. }
  410. if (!found)
  411. return;
  412. sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
  413. memcpy(&e820_saved, &e820, sizeof(struct e820map));
  414. printk(KERN_INFO "extended physical RAM map:\n");
  415. e820_print_map("reserve setup_data");
  416. }
  417. static void __init memblock_x86_reserve_range_setup_data(void)
  418. {
  419. struct setup_data *data;
  420. u64 pa_data;
  421. pa_data = boot_params.hdr.setup_data;
  422. while (pa_data) {
  423. data = early_memremap(pa_data, sizeof(*data));
  424. memblock_reserve(pa_data, sizeof(*data) + data->len);
  425. pa_data = data->next;
  426. early_iounmap(data, sizeof(*data));
  427. }
  428. }
  429. /*
  430. * --------- Crashkernel reservation ------------------------------
  431. */
  432. #ifdef CONFIG_KEXEC
  433. /*
  434. * Keep the crash kernel below this limit. On 32 bits earlier kernels
  435. * would limit the kernel to the low 512 MiB due to mapping restrictions.
  436. * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this
  437. * limit once kexec-tools are fixed.
  438. */
  439. #ifdef CONFIG_X86_32
  440. # define CRASH_KERNEL_ADDR_MAX (512 << 20)
  441. #else
  442. # define CRASH_KERNEL_ADDR_MAX (896 << 20)
  443. #endif
  444. static void __init reserve_crashkernel(void)
  445. {
  446. unsigned long long total_mem;
  447. unsigned long long crash_size, crash_base;
  448. int ret;
  449. total_mem = memblock_phys_mem_size();
  450. ret = parse_crashkernel(boot_command_line, total_mem,
  451. &crash_size, &crash_base);
  452. if (ret != 0 || crash_size <= 0)
  453. return;
  454. /* 0 means: find the address automatically */
  455. if (crash_base <= 0) {
  456. const unsigned long long alignment = 16<<20; /* 16M */
  457. /*
  458. * kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
  459. */
  460. crash_base = memblock_find_in_range(alignment,
  461. CRASH_KERNEL_ADDR_MAX, crash_size, alignment);
  462. if (!crash_base) {
  463. pr_info("crashkernel reservation failed - No suitable area found.\n");
  464. return;
  465. }
  466. } else {
  467. unsigned long long start;
  468. start = memblock_find_in_range(crash_base,
  469. crash_base + crash_size, crash_size, 1<<20);
  470. if (start != crash_base) {
  471. pr_info("crashkernel reservation failed - memory is in use.\n");
  472. return;
  473. }
  474. }
  475. memblock_reserve(crash_base, crash_size);
  476. printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
  477. "for crashkernel (System RAM: %ldMB)\n",
  478. (unsigned long)(crash_size >> 20),
  479. (unsigned long)(crash_base >> 20),
  480. (unsigned long)(total_mem >> 20));
  481. crashk_res.start = crash_base;
  482. crashk_res.end = crash_base + crash_size - 1;
  483. insert_resource(&iomem_resource, &crashk_res);
  484. }
  485. #else
  486. static void __init reserve_crashkernel(void)
  487. {
  488. }
  489. #endif
  490. static struct resource standard_io_resources[] = {
  491. { .name = "dma1", .start = 0x00, .end = 0x1f,
  492. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  493. { .name = "pic1", .start = 0x20, .end = 0x21,
  494. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  495. { .name = "timer0", .start = 0x40, .end = 0x43,
  496. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  497. { .name = "timer1", .start = 0x50, .end = 0x53,
  498. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  499. { .name = "keyboard", .start = 0x60, .end = 0x60,
  500. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  501. { .name = "keyboard", .start = 0x64, .end = 0x64,
  502. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  503. { .name = "dma page reg", .start = 0x80, .end = 0x8f,
  504. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  505. { .name = "pic2", .start = 0xa0, .end = 0xa1,
  506. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  507. { .name = "dma2", .start = 0xc0, .end = 0xdf,
  508. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  509. { .name = "fpu", .start = 0xf0, .end = 0xff,
  510. .flags = IORESOURCE_BUSY | IORESOURCE_IO }
  511. };
  512. void __init reserve_standard_io_resources(void)
  513. {
  514. int i;
  515. /* request I/O space for devices used on all i[345]86 PCs */
  516. for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
  517. request_resource(&ioport_resource, &standard_io_resources[i]);
  518. }
  519. static __init void reserve_ibft_region(void)
  520. {
  521. unsigned long addr, size = 0;
  522. addr = find_ibft_region(&size);
  523. if (size)
  524. memblock_reserve(addr, size);
  525. }
  526. static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
  527. static bool __init snb_gfx_workaround_needed(void)
  528. {
  529. #ifdef CONFIG_PCI
  530. int i;
  531. u16 vendor, devid;
  532. static const __initconst u16 snb_ids[] = {
  533. 0x0102,
  534. 0x0112,
  535. 0x0122,
  536. 0x0106,
  537. 0x0116,
  538. 0x0126,
  539. 0x010a,
  540. };
  541. /* Assume no if something weird is going on with PCI */
  542. if (!early_pci_allowed())
  543. return false;
  544. vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
  545. if (vendor != 0x8086)
  546. return false;
  547. devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
  548. for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
  549. if (devid == snb_ids[i])
  550. return true;
  551. #endif
  552. return false;
  553. }
  554. /*
  555. * Sandy Bridge graphics has trouble with certain ranges, exclude
  556. * them from allocation.
  557. */
  558. static void __init trim_snb_memory(void)
  559. {
  560. static const __initconst unsigned long bad_pages[] = {
  561. 0x20050000,
  562. 0x20110000,
  563. 0x20130000,
  564. 0x20138000,
  565. 0x40004000,
  566. };
  567. int i;
  568. if (!snb_gfx_workaround_needed())
  569. return;
  570. printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
  571. /*
  572. * Reserve all memory below the 1 MB mark that has not
  573. * already been reserved.
  574. */
  575. memblock_reserve(0, 1<<20);
  576. for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
  577. if (memblock_reserve(bad_pages[i], PAGE_SIZE))
  578. printk(KERN_WARNING "failed to reserve 0x%08lx\n",
  579. bad_pages[i]);
  580. }
  581. }
  582. /*
  583. * Here we put platform-specific memory range workarounds, i.e.
  584. * memory known to be corrupt or otherwise in need to be reserved on
  585. * specific platforms.
  586. *
  587. * If this gets used more widely it could use a real dispatch mechanism.
  588. */
  589. static void __init trim_platform_memory_ranges(void)
  590. {
  591. trim_snb_memory();
  592. }
  593. static void __init trim_bios_range(void)
  594. {
  595. /*
  596. * A special case is the first 4Kb of memory;
  597. * This is a BIOS owned area, not kernel ram, but generally
  598. * not listed as such in the E820 table.
  599. *
  600. * This typically reserves additional memory (64KiB by default)
  601. * since some BIOSes are known to corrupt low memory. See the
  602. * Kconfig help text for X86_RESERVE_LOW.
  603. */
  604. e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
  605. E820_RAM, E820_RESERVED);
  606. /*
  607. * special case: Some BIOSen report the PC BIOS
  608. * area (640->1Mb) as ram even though it is not.
  609. * take them out.
  610. */
  611. e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
  612. sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
  613. }
  614. /* called before trim_bios_range() to spare extra sanitize */
  615. static void __init e820_add_kernel_range(void)
  616. {
  617. u64 start = __pa_symbol(_text);
  618. u64 size = __pa_symbol(_end) - start;
  619. /*
  620. * Complain if .text .data and .bss are not marked as E820_RAM and
  621. * attempt to fix it by adding the range. We may have a confused BIOS,
  622. * or the user may have used memmap=exactmap or memmap=xxM$yyM to
  623. * exclude kernel range. If we really are running on top non-RAM,
  624. * we will crash later anyways.
  625. */
  626. if (e820_all_mapped(start, start + size, E820_RAM))
  627. return;
  628. pr_warn(".text .data .bss are not marked as E820_RAM!\n");
  629. e820_remove_range(start, size, E820_RAM, 0);
  630. e820_add_region(start, size, E820_RAM);
  631. }
  632. static int __init parse_reservelow(char *p)
  633. {
  634. unsigned long long size;
  635. if (!p)
  636. return -EINVAL;
  637. size = memparse(p, &p);
  638. if (size < 4096)
  639. size = 4096;
  640. if (size > 640*1024)
  641. size = 640*1024;
  642. reserve_low = size;
  643. return 0;
  644. }
  645. early_param("reservelow", parse_reservelow);
  646. /*
  647. * Determine if we were loaded by an EFI loader. If so, then we have also been
  648. * passed the efi memmap, systab, etc., so we should use these data structures
  649. * for initialization. Note, the efi init code path is determined by the
  650. * global efi_enabled. This allows the same kernel image to be used on existing
  651. * systems (with a traditional BIOS) as well as on EFI systems.
  652. */
  653. /*
  654. * setup_arch - architecture-specific boot-time initializations
  655. *
  656. * Note: On x86_64, fixmaps are ready for use even before this is called.
  657. */
  658. void __init setup_arch(char **cmdline_p)
  659. {
  660. early_reserve_initrd();
  661. #ifdef CONFIG_X86_32
  662. memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
  663. visws_early_detect();
  664. /*
  665. * copy kernel address range established so far and switch
  666. * to the proper swapper page table
  667. */
  668. clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
  669. initial_page_table + KERNEL_PGD_BOUNDARY,
  670. KERNEL_PGD_PTRS);
  671. load_cr3(swapper_pg_dir);
  672. __flush_tlb_all();
  673. #else
  674. printk(KERN_INFO "Command line: %s\n", boot_command_line);
  675. #endif
  676. /*
  677. * If we have OLPC OFW, we might end up relocating the fixmap due to
  678. * reserve_top(), so do this before touching the ioremap area.
  679. */
  680. olpc_ofw_detect();
  681. early_trap_init();
  682. early_cpu_init();
  683. early_ioremap_init();
  684. setup_olpc_ofw_pgd();
  685. ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
  686. screen_info = boot_params.screen_info;
  687. edid_info = boot_params.edid_info;
  688. #ifdef CONFIG_X86_32
  689. apm_info.bios = boot_params.apm_bios_info;
  690. ist_info = boot_params.ist_info;
  691. if (boot_params.sys_desc_table.length != 0) {
  692. machine_id = boot_params.sys_desc_table.table[0];
  693. machine_submodel_id = boot_params.sys_desc_table.table[1];
  694. BIOS_revision = boot_params.sys_desc_table.table[2];
  695. }
  696. #endif
  697. saved_video_mode = boot_params.hdr.vid_mode;
  698. bootloader_type = boot_params.hdr.type_of_loader;
  699. if ((bootloader_type >> 4) == 0xe) {
  700. bootloader_type &= 0xf;
  701. bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
  702. }
  703. bootloader_version = bootloader_type & 0xf;
  704. bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
  705. #ifdef CONFIG_BLK_DEV_RAM
  706. rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
  707. rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
  708. rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
  709. #endif
  710. #ifdef CONFIG_EFI
  711. if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
  712. "EL32", 4)) {
  713. efi_enabled = 1;
  714. efi_64bit = false;
  715. } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
  716. "EL64", 4)) {
  717. efi_enabled = 1;
  718. efi_64bit = true;
  719. }
  720. if (efi_enabled && efi_memblock_x86_reserve_range())
  721. efi_enabled = 0;
  722. #endif
  723. x86_init.oem.arch_setup();
  724. iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
  725. setup_memory_map();
  726. parse_setup_data();
  727. /* update the e820_saved too */
  728. e820_reserve_setup_data();
  729. copy_edd();
  730. if (!boot_params.hdr.root_flags)
  731. root_mountflags &= ~MS_RDONLY;
  732. init_mm.start_code = (unsigned long) _text;
  733. init_mm.end_code = (unsigned long) _etext;
  734. init_mm.end_data = (unsigned long) _edata;
  735. init_mm.brk = _brk_end;
  736. code_resource.start = virt_to_phys(_text);
  737. code_resource.end = virt_to_phys(_etext)-1;
  738. data_resource.start = virt_to_phys(_etext);
  739. data_resource.end = virt_to_phys(_edata)-1;
  740. bss_resource.start = virt_to_phys(&__bss_start);
  741. bss_resource.end = virt_to_phys(&__bss_stop)-1;
  742. #ifdef CONFIG_CMDLINE_BOOL
  743. #ifdef CONFIG_CMDLINE_OVERRIDE
  744. strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
  745. #else
  746. if (builtin_cmdline[0]) {
  747. /* append boot loader cmdline to builtin */
  748. strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
  749. strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  750. strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
  751. }
  752. #endif
  753. #endif
  754. strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
  755. *cmdline_p = command_line;
  756. /*
  757. * x86_configure_nx() is called before parse_early_param() to detect
  758. * whether hardware doesn't support NX (so that the early EHCI debug
  759. * console setup can safely call set_fixmap()). It may then be called
  760. * again from within noexec_setup() during parsing early parameters
  761. * to honor the respective command line option.
  762. */
  763. x86_configure_nx();
  764. parse_early_param();
  765. x86_report_nx();
  766. /* after early param, so could get panic from serial */
  767. memblock_x86_reserve_range_setup_data();
  768. if (acpi_mps_check()) {
  769. #ifdef CONFIG_X86_LOCAL_APIC
  770. disable_apic = 1;
  771. #endif
  772. setup_clear_cpu_cap(X86_FEATURE_APIC);
  773. }
  774. #ifdef CONFIG_PCI
  775. if (pci_early_dump_regs)
  776. early_dump_pci_devices();
  777. #endif
  778. finish_e820_parsing();
  779. if (efi_enabled)
  780. efi_init();
  781. dmi_scan_machine();
  782. /*
  783. * VMware detection requires dmi to be available, so this
  784. * needs to be done after dmi_scan_machine, for the BP.
  785. */
  786. init_hypervisor_platform();
  787. x86_init.resources.probe_roms();
  788. /* after parse_early_param, so could debug it */
  789. insert_resource(&iomem_resource, &code_resource);
  790. insert_resource(&iomem_resource, &data_resource);
  791. insert_resource(&iomem_resource, &bss_resource);
  792. e820_add_kernel_range();
  793. trim_bios_range();
  794. #ifdef CONFIG_X86_32
  795. if (ppro_with_ram_bug()) {
  796. e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
  797. E820_RESERVED);
  798. sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
  799. printk(KERN_INFO "fixed physical RAM map:\n");
  800. e820_print_map("bad_ppro");
  801. }
  802. #else
  803. early_gart_iommu_check();
  804. #endif
  805. /*
  806. * partially used pages are not usable - thus
  807. * we are rounding upwards:
  808. */
  809. max_pfn = e820_end_of_ram_pfn();
  810. /* update e820 for memory not covered by WB MTRRs */
  811. mtrr_bp_init();
  812. if (mtrr_trim_uncached_memory(max_pfn))
  813. max_pfn = e820_end_of_ram_pfn();
  814. #ifdef CONFIG_X86_32
  815. /* max_low_pfn get updated here */
  816. find_low_pfn_range();
  817. #else
  818. num_physpages = max_pfn;
  819. check_x2apic();
  820. /* How many end-of-memory variables you have, grandma! */
  821. /* need this before calling reserve_initrd */
  822. if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
  823. max_low_pfn = e820_end_of_low_ram_pfn();
  824. else
  825. max_low_pfn = max_pfn;
  826. high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
  827. #endif
  828. /*
  829. * Find and reserve possible boot-time SMP configuration:
  830. */
  831. find_smp_config();
  832. reserve_ibft_region();
  833. early_alloc_pgt_buf();
  834. /*
  835. * Need to conclude brk, before memblock_x86_fill()
  836. * it could use memblock_find_in_range, could overlap with
  837. * brk area.
  838. */
  839. reserve_brk();
  840. cleanup_highmap();
  841. memblock.current_limit = ISA_END_ADDRESS;
  842. memblock_x86_fill();
  843. /*
  844. * The EFI specification says that boot service code won't be called
  845. * after ExitBootServices(). This is, in fact, a lie.
  846. */
  847. if (efi_enabled)
  848. efi_reserve_boot_services();
  849. /* preallocate 4k for mptable mpc */
  850. early_reserve_e820_mpc_new();
  851. #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
  852. setup_bios_corruption_check();
  853. #endif
  854. #ifdef CONFIG_X86_32
  855. printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
  856. (max_pfn_mapped<<PAGE_SHIFT) - 1);
  857. #endif
  858. reserve_real_mode();
  859. trim_platform_memory_ranges();
  860. init_mem_mapping();
  861. early_trap_pf_init();
  862. setup_real_mode();
  863. memblock.current_limit = get_max_mapped();
  864. dma_contiguous_reserve(0);
  865. /*
  866. * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
  867. */
  868. #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
  869. if (init_ohci1394_dma_early)
  870. init_ohci1394_dma_on_all_controllers();
  871. #endif
  872. /* Allocate bigger log buffer */
  873. setup_log_buf(1);
  874. reserve_initrd();
  875. #if defined(CONFIG_ACPI) && defined(CONFIG_BLK_DEV_INITRD)
  876. acpi_initrd_override((void *)initrd_start, initrd_end - initrd_start);
  877. #endif
  878. reserve_crashkernel();
  879. vsmp_init();
  880. io_delay_init();
  881. /*
  882. * Parse the ACPI tables for possible boot-time SMP configuration.
  883. */
  884. acpi_boot_table_init();
  885. early_acpi_boot_init();
  886. initmem_init();
  887. memblock_find_dma_reserve();
  888. #ifdef CONFIG_KVM_GUEST
  889. kvmclock_init();
  890. #endif
  891. x86_init.paging.pagetable_init();
  892. if (boot_cpu_data.cpuid_level >= 0) {
  893. /* A CPU has %cr4 if and only if it has CPUID */
  894. mmu_cr4_features = read_cr4();
  895. if (trampoline_cr4_features)
  896. *trampoline_cr4_features = mmu_cr4_features;
  897. }
  898. #ifdef CONFIG_X86_32
  899. /* sync back kernel address range */
  900. clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
  901. swapper_pg_dir + KERNEL_PGD_BOUNDARY,
  902. KERNEL_PGD_PTRS);
  903. #endif
  904. tboot_probe();
  905. #ifdef CONFIG_X86_64
  906. map_vsyscall();
  907. #endif
  908. generic_apic_probe();
  909. early_quirks();
  910. /*
  911. * Read APIC and some other early information from ACPI tables.
  912. */
  913. acpi_boot_init();
  914. sfi_init();
  915. x86_dtb_init();
  916. /*
  917. * get boot-time SMP configuration:
  918. */
  919. if (smp_found_config)
  920. get_smp_config();
  921. prefill_possible_map();
  922. init_cpu_to_node();
  923. init_apic_mappings();
  924. if (x86_io_apic_ops.init)
  925. x86_io_apic_ops.init();
  926. kvm_guest_init();
  927. e820_reserve_resources();
  928. e820_mark_nosave_regions(max_low_pfn);
  929. x86_init.resources.reserve_resources();
  930. e820_setup_gap();
  931. #ifdef CONFIG_VT
  932. #if defined(CONFIG_VGA_CONSOLE)
  933. if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
  934. conswitchp = &vga_con;
  935. #elif defined(CONFIG_DUMMY_CONSOLE)
  936. conswitchp = &dummy_con;
  937. #endif
  938. #endif
  939. x86_init.oem.banner();
  940. x86_init.timers.wallclock_init();
  941. mcheck_init();
  942. arch_init_ideal_nops();
  943. register_refined_jiffies(CLOCK_TICK_RATE);
  944. #ifdef CONFIG_EFI
  945. /* Once setup is done above, disable efi_enabled on mismatched
  946. * firmware/kernel archtectures since there is no support for
  947. * runtime services.
  948. */
  949. if (efi_enabled && IS_ENABLED(CONFIG_X86_64) != efi_64bit) {
  950. pr_info("efi: Setup done, disabling due to 32/64-bit mismatch\n");
  951. efi_unmap_memmap();
  952. efi_enabled = 0;
  953. }
  954. #endif
  955. }
  956. #ifdef CONFIG_X86_32
  957. static struct resource video_ram_resource = {
  958. .name = "Video RAM area",
  959. .start = 0xa0000,
  960. .end = 0xbffff,
  961. .flags = IORESOURCE_BUSY | IORESOURCE_MEM
  962. };
  963. void __init i386_reserve_resources(void)
  964. {
  965. request_resource(&iomem_resource, &video_ram_resource);
  966. reserve_standard_io_resources();
  967. }
  968. #endif /* CONFIG_X86_32 */