setup.c 28 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169
  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/mca.h>
  36. #include <linux/root_dev.h>
  37. #include <linux/highmem.h>
  38. #include <linux/module.h>
  39. #include <linux/efi.h>
  40. #include <linux/init.h>
  41. #include <linux/edd.h>
  42. #include <linux/iscsi_ibft.h>
  43. #include <linux/nodemask.h>
  44. #include <linux/kexec.h>
  45. #include <linux/dmi.h>
  46. #include <linux/pfn.h>
  47. #include <linux/pci.h>
  48. #include <asm/pci-direct.h>
  49. #include <linux/init_ohci1394_dma.h>
  50. #include <linux/kvm_para.h>
  51. #include <linux/dma-contiguous.h>
  52. #include <linux/errno.h>
  53. #include <linux/kernel.h>
  54. #include <linux/stddef.h>
  55. #include <linux/unistd.h>
  56. #include <linux/ptrace.h>
  57. #include <linux/user.h>
  58. #include <linux/delay.h>
  59. #include <linux/kallsyms.h>
  60. #include <linux/cpufreq.h>
  61. #include <linux/dma-mapping.h>
  62. #include <linux/ctype.h>
  63. #include <linux/uaccess.h>
  64. #include <linux/percpu.h>
  65. #include <linux/crash_dump.h>
  66. #include <linux/tboot.h>
  67. #include <video/edid.h>
  68. #include <asm/mtrr.h>
  69. #include <asm/apic.h>
  70. #include <asm/trampoline.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. #ifdef CONFIG_X86_64
  102. #include <asm/numa_64.h>
  103. #endif
  104. #include <asm/mce.h>
  105. #include <asm/alternative.h>
  106. #include <asm/prom.h>
  107. /*
  108. * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
  109. * The direct mapping extends to max_pfn_mapped, so that we can directly access
  110. * apertures, ACPI and other tables without having to play with fixmaps.
  111. */
  112. unsigned long max_low_pfn_mapped;
  113. unsigned long max_pfn_mapped;
  114. #ifdef CONFIG_DMI
  115. RESERVE_BRK(dmi_alloc, 65536);
  116. #endif
  117. static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
  118. unsigned long _brk_end = (unsigned long)__brk_base;
  119. #ifdef CONFIG_X86_64
  120. int default_cpu_present_to_apicid(int mps_cpu)
  121. {
  122. return __default_cpu_present_to_apicid(mps_cpu);
  123. }
  124. int default_check_phys_apicid_present(int phys_apicid)
  125. {
  126. return __default_check_phys_apicid_present(phys_apicid);
  127. }
  128. #endif
  129. #ifndef CONFIG_DEBUG_BOOT_PARAMS
  130. struct boot_params __initdata boot_params;
  131. #else
  132. struct boot_params boot_params;
  133. #endif
  134. /*
  135. * Machine setup..
  136. */
  137. static struct resource data_resource = {
  138. .name = "Kernel data",
  139. .start = 0,
  140. .end = 0,
  141. .flags = IORESOURCE_BUSY | IORESOURCE_MEM
  142. };
  143. static struct resource code_resource = {
  144. .name = "Kernel code",
  145. .start = 0,
  146. .end = 0,
  147. .flags = IORESOURCE_BUSY | IORESOURCE_MEM
  148. };
  149. static struct resource bss_resource = {
  150. .name = "Kernel bss",
  151. .start = 0,
  152. .end = 0,
  153. .flags = IORESOURCE_BUSY | IORESOURCE_MEM
  154. };
  155. #ifdef CONFIG_X86_32
  156. /* cpu data as detected by the assembly code in head.S */
  157. struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
  158. /* common cpu data for all cpus */
  159. struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
  160. EXPORT_SYMBOL(boot_cpu_data);
  161. static void set_mca_bus(int x)
  162. {
  163. #ifdef CONFIG_MCA
  164. MCA_bus = x;
  165. #endif
  166. }
  167. unsigned int def_to_bigsmp;
  168. /* for MCA, but anyone else can use it if they want */
  169. unsigned int machine_id;
  170. unsigned int machine_submodel_id;
  171. unsigned int BIOS_revision;
  172. struct apm_info apm_info;
  173. EXPORT_SYMBOL(apm_info);
  174. #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
  175. defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
  176. struct ist_info ist_info;
  177. EXPORT_SYMBOL(ist_info);
  178. #else
  179. struct ist_info ist_info;
  180. #endif
  181. #else
  182. struct cpuinfo_x86 boot_cpu_data __read_mostly = {
  183. .x86_phys_bits = MAX_PHYSMEM_BITS,
  184. };
  185. EXPORT_SYMBOL(boot_cpu_data);
  186. #endif
  187. #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
  188. unsigned long mmu_cr4_features;
  189. #else
  190. unsigned long mmu_cr4_features = X86_CR4_PAE;
  191. #endif
  192. /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
  193. int bootloader_type, bootloader_version;
  194. /*
  195. * Setup options
  196. */
  197. struct screen_info screen_info;
  198. EXPORT_SYMBOL(screen_info);
  199. struct edid_info edid_info;
  200. EXPORT_SYMBOL_GPL(edid_info);
  201. extern int root_mountflags;
  202. unsigned long saved_video_mode;
  203. #define RAMDISK_IMAGE_START_MASK 0x07FF
  204. #define RAMDISK_PROMPT_FLAG 0x8000
  205. #define RAMDISK_LOAD_FLAG 0x4000
  206. static char __initdata command_line[COMMAND_LINE_SIZE];
  207. #ifdef CONFIG_CMDLINE_BOOL
  208. static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
  209. #endif
  210. #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
  211. struct edd edd;
  212. #ifdef CONFIG_EDD_MODULE
  213. EXPORT_SYMBOL(edd);
  214. #endif
  215. /**
  216. * copy_edd() - Copy the BIOS EDD information
  217. * from boot_params into a safe place.
  218. *
  219. */
  220. static inline void __init copy_edd(void)
  221. {
  222. memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
  223. sizeof(edd.mbr_signature));
  224. memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
  225. edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
  226. edd.edd_info_nr = boot_params.eddbuf_entries;
  227. }
  228. #else
  229. static inline void __init copy_edd(void)
  230. {
  231. }
  232. #endif
  233. void * __init extend_brk(size_t size, size_t align)
  234. {
  235. size_t mask = align - 1;
  236. void *ret;
  237. BUG_ON(_brk_start == 0);
  238. BUG_ON(align & mask);
  239. _brk_end = (_brk_end + mask) & ~mask;
  240. BUG_ON((char *)(_brk_end + size) > __brk_limit);
  241. ret = (void *)_brk_end;
  242. _brk_end += size;
  243. memset(ret, 0, size);
  244. return ret;
  245. }
  246. #ifdef CONFIG_X86_64
  247. static void __init init_gbpages(void)
  248. {
  249. if (direct_gbpages && cpu_has_gbpages)
  250. printk(KERN_INFO "Using GB pages for direct mapping\n");
  251. else
  252. direct_gbpages = 0;
  253. }
  254. #else
  255. static inline void init_gbpages(void)
  256. {
  257. }
  258. static void __init cleanup_highmap(void)
  259. {
  260. }
  261. #endif
  262. static void __init reserve_brk(void)
  263. {
  264. if (_brk_end > _brk_start)
  265. memblock_reserve(__pa(_brk_start),
  266. __pa(_brk_end) - __pa(_brk_start));
  267. /* Mark brk area as locked down and no longer taking any
  268. new allocations */
  269. _brk_start = 0;
  270. }
  271. #ifdef CONFIG_BLK_DEV_INITRD
  272. #define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT)
  273. static void __init relocate_initrd(void)
  274. {
  275. /* Assume only end is not page aligned */
  276. u64 ramdisk_image = boot_params.hdr.ramdisk_image;
  277. u64 ramdisk_size = boot_params.hdr.ramdisk_size;
  278. u64 area_size = PAGE_ALIGN(ramdisk_size);
  279. u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
  280. u64 ramdisk_here;
  281. unsigned long slop, clen, mapaddr;
  282. char *p, *q;
  283. /* We need to move the initrd down into lowmem */
  284. ramdisk_here = memblock_find_in_range(0, end_of_lowmem, area_size,
  285. PAGE_SIZE);
  286. if (!ramdisk_here)
  287. panic("Cannot find place for new RAMDISK of size %lld\n",
  288. ramdisk_size);
  289. /* Note: this includes all the lowmem currently occupied by
  290. the initrd, we rely on that fact to keep the data intact. */
  291. memblock_reserve(ramdisk_here, area_size);
  292. initrd_start = ramdisk_here + PAGE_OFFSET;
  293. initrd_end = initrd_start + ramdisk_size;
  294. printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
  295. ramdisk_here, ramdisk_here + ramdisk_size);
  296. q = (char *)initrd_start;
  297. /* Copy any lowmem portion of the initrd */
  298. if (ramdisk_image < end_of_lowmem) {
  299. clen = end_of_lowmem - ramdisk_image;
  300. p = (char *)__va(ramdisk_image);
  301. memcpy(q, p, clen);
  302. q += clen;
  303. ramdisk_image += clen;
  304. ramdisk_size -= clen;
  305. }
  306. /* Copy the highmem portion of the initrd */
  307. while (ramdisk_size) {
  308. slop = ramdisk_image & ~PAGE_MASK;
  309. clen = ramdisk_size;
  310. if (clen > MAX_MAP_CHUNK-slop)
  311. clen = MAX_MAP_CHUNK-slop;
  312. mapaddr = ramdisk_image & PAGE_MASK;
  313. p = early_memremap(mapaddr, clen+slop);
  314. memcpy(q, p+slop, clen);
  315. early_iounmap(p, clen+slop);
  316. q += clen;
  317. ramdisk_image += clen;
  318. ramdisk_size -= clen;
  319. }
  320. /* high pages is not converted by early_res_to_bootmem */
  321. ramdisk_image = boot_params.hdr.ramdisk_image;
  322. ramdisk_size = boot_params.hdr.ramdisk_size;
  323. printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to"
  324. " %08llx - %08llx\n",
  325. ramdisk_image, ramdisk_image + ramdisk_size - 1,
  326. ramdisk_here, ramdisk_here + ramdisk_size - 1);
  327. }
  328. static void __init reserve_initrd(void)
  329. {
  330. /* Assume only end is not page aligned */
  331. u64 ramdisk_image = boot_params.hdr.ramdisk_image;
  332. u64 ramdisk_size = boot_params.hdr.ramdisk_size;
  333. u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
  334. u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
  335. if (!boot_params.hdr.type_of_loader ||
  336. !ramdisk_image || !ramdisk_size)
  337. return; /* No initrd provided by bootloader */
  338. initrd_start = 0;
  339. if (ramdisk_size >= (end_of_lowmem>>1)) {
  340. memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
  341. printk(KERN_ERR "initrd too large to handle, "
  342. "disabling initrd\n");
  343. return;
  344. }
  345. printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image,
  346. ramdisk_end);
  347. if (ramdisk_end <= end_of_lowmem) {
  348. /* All in lowmem, easy case */
  349. /*
  350. * don't need to reserve again, already reserved early
  351. * in i386_start_kernel
  352. */
  353. initrd_start = ramdisk_image + PAGE_OFFSET;
  354. initrd_end = initrd_start + ramdisk_size;
  355. return;
  356. }
  357. relocate_initrd();
  358. memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
  359. }
  360. #else
  361. static void __init reserve_initrd(void)
  362. {
  363. }
  364. #endif /* CONFIG_BLK_DEV_INITRD */
  365. static void __init parse_setup_data(void)
  366. {
  367. struct setup_data *data;
  368. u64 pa_data;
  369. if (boot_params.hdr.version < 0x0209)
  370. return;
  371. pa_data = boot_params.hdr.setup_data;
  372. while (pa_data) {
  373. u32 data_len, map_len;
  374. map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
  375. (u64)sizeof(struct setup_data));
  376. data = early_memremap(pa_data, map_len);
  377. data_len = data->len + sizeof(struct setup_data);
  378. if (data_len > map_len) {
  379. early_iounmap(data, map_len);
  380. data = early_memremap(pa_data, data_len);
  381. map_len = data_len;
  382. }
  383. switch (data->type) {
  384. case SETUP_E820_EXT:
  385. parse_e820_ext(data);
  386. break;
  387. case SETUP_DTB:
  388. add_dtb(pa_data);
  389. break;
  390. default:
  391. break;
  392. }
  393. pa_data = data->next;
  394. early_iounmap(data, map_len);
  395. }
  396. }
  397. static void __init e820_reserve_setup_data(void)
  398. {
  399. struct setup_data *data;
  400. u64 pa_data;
  401. int found = 0;
  402. if (boot_params.hdr.version < 0x0209)
  403. return;
  404. pa_data = boot_params.hdr.setup_data;
  405. while (pa_data) {
  406. data = early_memremap(pa_data, sizeof(*data));
  407. e820_update_range(pa_data, sizeof(*data)+data->len,
  408. E820_RAM, E820_RESERVED_KERN);
  409. found = 1;
  410. pa_data = data->next;
  411. early_iounmap(data, sizeof(*data));
  412. }
  413. if (!found)
  414. return;
  415. sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
  416. memcpy(&e820_saved, &e820, sizeof(struct e820map));
  417. printk(KERN_INFO "extended physical RAM map:\n");
  418. e820_print_map("reserve setup_data");
  419. }
  420. static void __init memblock_x86_reserve_range_setup_data(void)
  421. {
  422. struct setup_data *data;
  423. u64 pa_data;
  424. if (boot_params.hdr.version < 0x0209)
  425. return;
  426. pa_data = boot_params.hdr.setup_data;
  427. while (pa_data) {
  428. data = early_memremap(pa_data, sizeof(*data));
  429. memblock_reserve(pa_data, sizeof(*data) + data->len);
  430. pa_data = data->next;
  431. early_iounmap(data, sizeof(*data));
  432. }
  433. }
  434. /*
  435. * --------- Crashkernel reservation ------------------------------
  436. */
  437. #ifdef CONFIG_KEXEC
  438. /*
  439. * Keep the crash kernel below this limit. On 32 bits earlier kernels
  440. * would limit the kernel to the low 512 MiB due to mapping restrictions.
  441. * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this
  442. * limit once kexec-tools are fixed.
  443. */
  444. #ifdef CONFIG_X86_32
  445. # define CRASH_KERNEL_ADDR_MAX (512 << 20)
  446. #else
  447. # define CRASH_KERNEL_ADDR_MAX (896 << 20)
  448. #endif
  449. static void __init reserve_crashkernel(void)
  450. {
  451. unsigned long long total_mem;
  452. unsigned long long crash_size, crash_base;
  453. int ret;
  454. total_mem = memblock_phys_mem_size();
  455. ret = parse_crashkernel(boot_command_line, total_mem,
  456. &crash_size, &crash_base);
  457. if (ret != 0 || crash_size <= 0)
  458. return;
  459. /* 0 means: find the address automatically */
  460. if (crash_base <= 0) {
  461. const unsigned long long alignment = 16<<20; /* 16M */
  462. /*
  463. * kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
  464. */
  465. crash_base = memblock_find_in_range(alignment,
  466. CRASH_KERNEL_ADDR_MAX, crash_size, alignment);
  467. if (!crash_base) {
  468. pr_info("crashkernel reservation failed - No suitable area found.\n");
  469. return;
  470. }
  471. } else {
  472. unsigned long long start;
  473. start = memblock_find_in_range(crash_base,
  474. crash_base + crash_size, crash_size, 1<<20);
  475. if (start != crash_base) {
  476. pr_info("crashkernel reservation failed - memory is in use.\n");
  477. return;
  478. }
  479. }
  480. memblock_reserve(crash_base, crash_size);
  481. printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
  482. "for crashkernel (System RAM: %ldMB)\n",
  483. (unsigned long)(crash_size >> 20),
  484. (unsigned long)(crash_base >> 20),
  485. (unsigned long)(total_mem >> 20));
  486. crashk_res.start = crash_base;
  487. crashk_res.end = crash_base + crash_size - 1;
  488. insert_resource(&iomem_resource, &crashk_res);
  489. }
  490. #else
  491. static void __init reserve_crashkernel(void)
  492. {
  493. }
  494. #endif
  495. static struct resource standard_io_resources[] = {
  496. { .name = "dma1", .start = 0x00, .end = 0x1f,
  497. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  498. { .name = "pic1", .start = 0x20, .end = 0x21,
  499. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  500. { .name = "timer0", .start = 0x40, .end = 0x43,
  501. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  502. { .name = "timer1", .start = 0x50, .end = 0x53,
  503. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  504. { .name = "keyboard", .start = 0x60, .end = 0x60,
  505. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  506. { .name = "keyboard", .start = 0x64, .end = 0x64,
  507. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  508. { .name = "dma page reg", .start = 0x80, .end = 0x8f,
  509. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  510. { .name = "pic2", .start = 0xa0, .end = 0xa1,
  511. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  512. { .name = "dma2", .start = 0xc0, .end = 0xdf,
  513. .flags = IORESOURCE_BUSY | IORESOURCE_IO },
  514. { .name = "fpu", .start = 0xf0, .end = 0xff,
  515. .flags = IORESOURCE_BUSY | IORESOURCE_IO }
  516. };
  517. void __init reserve_standard_io_resources(void)
  518. {
  519. int i;
  520. /* request I/O space for devices used on all i[345]86 PCs */
  521. for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
  522. request_resource(&ioport_resource, &standard_io_resources[i]);
  523. }
  524. static __init void reserve_ibft_region(void)
  525. {
  526. unsigned long addr, size = 0;
  527. addr = find_ibft_region(&size);
  528. if (size)
  529. memblock_reserve(addr, size);
  530. }
  531. static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
  532. static bool __init snb_gfx_workaround_needed(void)
  533. {
  534. #ifdef CONFIG_PCI
  535. int i;
  536. u16 vendor, devid;
  537. static const __initconst u16 snb_ids[] = {
  538. 0x0102,
  539. 0x0112,
  540. 0x0122,
  541. 0x0106,
  542. 0x0116,
  543. 0x0126,
  544. 0x010a,
  545. };
  546. /* Assume no if something weird is going on with PCI */
  547. if (!early_pci_allowed())
  548. return false;
  549. vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
  550. if (vendor != 0x8086)
  551. return false;
  552. devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
  553. for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
  554. if (devid == snb_ids[i])
  555. return true;
  556. #endif
  557. return false;
  558. }
  559. /*
  560. * Sandy Bridge graphics has trouble with certain ranges, exclude
  561. * them from allocation.
  562. */
  563. static void __init trim_snb_memory(void)
  564. {
  565. static const __initconst unsigned long bad_pages[] = {
  566. 0x20050000,
  567. 0x20110000,
  568. 0x20130000,
  569. 0x20138000,
  570. 0x40004000,
  571. };
  572. int i;
  573. if (!snb_gfx_workaround_needed())
  574. return;
  575. printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
  576. /*
  577. * Reserve all memory below the 1 MB mark that has not
  578. * already been reserved.
  579. */
  580. memblock_reserve(0, 1<<20);
  581. for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
  582. if (memblock_reserve(bad_pages[i], PAGE_SIZE))
  583. printk(KERN_WARNING "failed to reserve 0x%08lx\n",
  584. bad_pages[i]);
  585. }
  586. }
  587. /*
  588. * Here we put platform-specific memory range workarounds, i.e.
  589. * memory known to be corrupt or otherwise in need to be reserved on
  590. * specific platforms.
  591. *
  592. * If this gets used more widely it could use a real dispatch mechanism.
  593. */
  594. static void __init trim_platform_memory_ranges(void)
  595. {
  596. trim_snb_memory();
  597. }
  598. static void __init trim_bios_range(void)
  599. {
  600. /*
  601. * A special case is the first 4Kb of memory;
  602. * This is a BIOS owned area, not kernel ram, but generally
  603. * not listed as such in the E820 table.
  604. *
  605. * This typically reserves additional memory (64KiB by default)
  606. * since some BIOSes are known to corrupt low memory. See the
  607. * Kconfig help text for X86_RESERVE_LOW.
  608. */
  609. e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
  610. E820_RAM, E820_RESERVED);
  611. /*
  612. * special case: Some BIOSen report the PC BIOS
  613. * area (640->1Mb) as ram even though it is not.
  614. * take them out.
  615. */
  616. e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
  617. sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
  618. }
  619. static int __init parse_reservelow(char *p)
  620. {
  621. unsigned long long size;
  622. if (!p)
  623. return -EINVAL;
  624. size = memparse(p, &p);
  625. if (size < 4096)
  626. size = 4096;
  627. if (size > 640*1024)
  628. size = 640*1024;
  629. reserve_low = size;
  630. return 0;
  631. }
  632. early_param("reservelow", parse_reservelow);
  633. /*
  634. * Determine if we were loaded by an EFI loader. If so, then we have also been
  635. * passed the efi memmap, systab, etc., so we should use these data structures
  636. * for initialization. Note, the efi init code path is determined by the
  637. * global efi_enabled. This allows the same kernel image to be used on existing
  638. * systems (with a traditional BIOS) as well as on EFI systems.
  639. */
  640. /*
  641. * setup_arch - architecture-specific boot-time initializations
  642. *
  643. * Note: On x86_64, fixmaps are ready for use even before this is called.
  644. */
  645. void __init setup_arch(char **cmdline_p)
  646. {
  647. #ifdef CONFIG_X86_32
  648. memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
  649. visws_early_detect();
  650. /*
  651. * copy kernel address range established so far and switch
  652. * to the proper swapper page table
  653. */
  654. clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
  655. initial_page_table + KERNEL_PGD_BOUNDARY,
  656. KERNEL_PGD_PTRS);
  657. load_cr3(swapper_pg_dir);
  658. __flush_tlb_all();
  659. #else
  660. printk(KERN_INFO "Command line: %s\n", boot_command_line);
  661. #endif
  662. /*
  663. * If we have OLPC OFW, we might end up relocating the fixmap due to
  664. * reserve_top(), so do this before touching the ioremap area.
  665. */
  666. olpc_ofw_detect();
  667. early_trap_init();
  668. early_cpu_init();
  669. early_ioremap_init();
  670. setup_olpc_ofw_pgd();
  671. ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
  672. screen_info = boot_params.screen_info;
  673. edid_info = boot_params.edid_info;
  674. #ifdef CONFIG_X86_32
  675. apm_info.bios = boot_params.apm_bios_info;
  676. ist_info = boot_params.ist_info;
  677. if (boot_params.sys_desc_table.length != 0) {
  678. set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
  679. machine_id = boot_params.sys_desc_table.table[0];
  680. machine_submodel_id = boot_params.sys_desc_table.table[1];
  681. BIOS_revision = boot_params.sys_desc_table.table[2];
  682. }
  683. #endif
  684. saved_video_mode = boot_params.hdr.vid_mode;
  685. bootloader_type = boot_params.hdr.type_of_loader;
  686. if ((bootloader_type >> 4) == 0xe) {
  687. bootloader_type &= 0xf;
  688. bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
  689. }
  690. bootloader_version = bootloader_type & 0xf;
  691. bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
  692. #ifdef CONFIG_BLK_DEV_RAM
  693. rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
  694. rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
  695. rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
  696. #endif
  697. #ifdef CONFIG_EFI
  698. if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
  699. "EL32", 4)) {
  700. set_bit(EFI_BOOT, &x86_efi_facility);
  701. } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
  702. "EL64", 4)) {
  703. set_bit(EFI_BOOT, &x86_efi_facility);
  704. set_bit(EFI_64BIT, &x86_efi_facility);
  705. }
  706. if (efi_enabled(EFI_BOOT))
  707. efi_memblock_x86_reserve_range();
  708. #endif
  709. x86_init.oem.arch_setup();
  710. iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
  711. setup_memory_map();
  712. parse_setup_data();
  713. /* update the e820_saved too */
  714. e820_reserve_setup_data();
  715. copy_edd();
  716. if (!boot_params.hdr.root_flags)
  717. root_mountflags &= ~MS_RDONLY;
  718. init_mm.start_code = (unsigned long) _text;
  719. init_mm.end_code = (unsigned long) _etext;
  720. init_mm.end_data = (unsigned long) _edata;
  721. init_mm.brk = _brk_end;
  722. code_resource.start = virt_to_phys(_text);
  723. code_resource.end = virt_to_phys(_etext)-1;
  724. data_resource.start = virt_to_phys(_etext);
  725. data_resource.end = virt_to_phys(_edata)-1;
  726. bss_resource.start = virt_to_phys(&__bss_start);
  727. bss_resource.end = virt_to_phys(&__bss_stop)-1;
  728. #ifdef CONFIG_CMDLINE_BOOL
  729. #ifdef CONFIG_CMDLINE_OVERRIDE
  730. strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
  731. #else
  732. if (builtin_cmdline[0]) {
  733. /* append boot loader cmdline to builtin */
  734. strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
  735. strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
  736. strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
  737. }
  738. #endif
  739. #endif
  740. strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
  741. *cmdline_p = command_line;
  742. /*
  743. * x86_configure_nx() is called before parse_early_param() to detect
  744. * whether hardware doesn't support NX (so that the early EHCI debug
  745. * console setup can safely call set_fixmap()). It may then be called
  746. * again from within noexec_setup() during parsing early parameters
  747. * to honor the respective command line option.
  748. */
  749. x86_configure_nx();
  750. parse_early_param();
  751. x86_report_nx();
  752. /* after early param, so could get panic from serial */
  753. memblock_x86_reserve_range_setup_data();
  754. if (acpi_mps_check()) {
  755. #ifdef CONFIG_X86_LOCAL_APIC
  756. disable_apic = 1;
  757. #endif
  758. setup_clear_cpu_cap(X86_FEATURE_APIC);
  759. }
  760. #ifdef CONFIG_PCI
  761. if (pci_early_dump_regs)
  762. early_dump_pci_devices();
  763. #endif
  764. finish_e820_parsing();
  765. if (efi_enabled(EFI_BOOT))
  766. efi_init();
  767. dmi_scan_machine();
  768. /*
  769. * VMware detection requires dmi to be available, so this
  770. * needs to be done after dmi_scan_machine, for the BP.
  771. */
  772. init_hypervisor_platform();
  773. x86_init.resources.probe_roms();
  774. /* after parse_early_param, so could debug it */
  775. insert_resource(&iomem_resource, &code_resource);
  776. insert_resource(&iomem_resource, &data_resource);
  777. insert_resource(&iomem_resource, &bss_resource);
  778. trim_bios_range();
  779. #ifdef CONFIG_X86_32
  780. if (ppro_with_ram_bug()) {
  781. e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
  782. E820_RESERVED);
  783. sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
  784. printk(KERN_INFO "fixed physical RAM map:\n");
  785. e820_print_map("bad_ppro");
  786. }
  787. #else
  788. early_gart_iommu_check();
  789. #endif
  790. /*
  791. * partially used pages are not usable - thus
  792. * we are rounding upwards:
  793. */
  794. max_pfn = e820_end_of_ram_pfn();
  795. /* update e820 for memory not covered by WB MTRRs */
  796. mtrr_bp_init();
  797. if (mtrr_trim_uncached_memory(max_pfn))
  798. max_pfn = e820_end_of_ram_pfn();
  799. #ifdef CONFIG_X86_32
  800. /* max_low_pfn get updated here */
  801. find_low_pfn_range();
  802. #else
  803. num_physpages = max_pfn;
  804. check_x2apic();
  805. /* How many end-of-memory variables you have, grandma! */
  806. /* need this before calling reserve_initrd */
  807. if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
  808. max_low_pfn = e820_end_of_low_ram_pfn();
  809. else
  810. max_low_pfn = max_pfn;
  811. high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
  812. #endif
  813. /*
  814. * Find and reserve possible boot-time SMP configuration:
  815. */
  816. find_smp_config();
  817. reserve_ibft_region();
  818. /*
  819. * Need to conclude brk, before memblock_x86_fill()
  820. * it could use memblock_find_in_range, could overlap with
  821. * brk area.
  822. */
  823. reserve_brk();
  824. cleanup_highmap();
  825. memblock.current_limit = get_max_mapped();
  826. memblock_x86_fill();
  827. /*
  828. * The EFI specification says that boot service code won't be called
  829. * after ExitBootServices(). This is, in fact, a lie.
  830. */
  831. if (efi_enabled(EFI_MEMMAP))
  832. efi_reserve_boot_services();
  833. /* preallocate 4k for mptable mpc */
  834. early_reserve_e820_mpc_new();
  835. #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
  836. setup_bios_corruption_check();
  837. #endif
  838. printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n",
  839. max_pfn_mapped<<PAGE_SHIFT);
  840. setup_trampolines();
  841. trim_platform_memory_ranges();
  842. init_gbpages();
  843. /* max_pfn_mapped is updated here */
  844. max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
  845. max_pfn_mapped = max_low_pfn_mapped;
  846. #ifdef CONFIG_X86_64
  847. if (max_pfn > max_low_pfn) {
  848. int i;
  849. unsigned long start, end;
  850. unsigned long start_pfn, end_pfn;
  851. for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn,
  852. NULL) {
  853. end = PFN_PHYS(end_pfn);
  854. if (end <= (1UL<<32))
  855. continue;
  856. start = PFN_PHYS(start_pfn);
  857. max_pfn_mapped = init_memory_mapping(
  858. max((1UL<<32), start), end);
  859. }
  860. /* can we preseve max_low_pfn ?*/
  861. max_low_pfn = max_pfn;
  862. }
  863. #endif
  864. memblock.current_limit = get_max_mapped();
  865. dma_contiguous_reserve(0);
  866. /*
  867. * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
  868. */
  869. #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
  870. if (init_ohci1394_dma_early)
  871. init_ohci1394_dma_on_all_controllers();
  872. #endif
  873. /* Allocate bigger log buffer */
  874. setup_log_buf(1);
  875. reserve_initrd();
  876. reserve_crashkernel();
  877. vsmp_init();
  878. io_delay_init();
  879. /*
  880. * Parse the ACPI tables for possible boot-time SMP configuration.
  881. */
  882. acpi_boot_table_init();
  883. early_acpi_boot_init();
  884. initmem_init();
  885. memblock_find_dma_reserve();
  886. #ifdef CONFIG_KVM_CLOCK
  887. kvmclock_init();
  888. #endif
  889. x86_init.paging.pagetable_setup_start(swapper_pg_dir);
  890. paging_init();
  891. x86_init.paging.pagetable_setup_done(swapper_pg_dir);
  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. }
  896. #ifdef CONFIG_X86_32
  897. /* sync back kernel address range */
  898. clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
  899. swapper_pg_dir + KERNEL_PGD_BOUNDARY,
  900. KERNEL_PGD_PTRS);
  901. #endif
  902. tboot_probe();
  903. #ifdef CONFIG_X86_64
  904. map_vsyscall();
  905. #endif
  906. generic_apic_probe();
  907. early_quirks();
  908. /*
  909. * Read APIC and some other early information from ACPI tables.
  910. */
  911. acpi_boot_init();
  912. sfi_init();
  913. x86_dtb_init();
  914. /*
  915. * get boot-time SMP configuration:
  916. */
  917. if (smp_found_config)
  918. get_smp_config();
  919. prefill_possible_map();
  920. init_cpu_to_node();
  921. init_apic_mappings();
  922. ioapic_and_gsi_init();
  923. kvm_guest_init();
  924. e820_reserve_resources();
  925. e820_mark_nosave_regions(max_low_pfn);
  926. x86_init.resources.reserve_resources();
  927. e820_setup_gap();
  928. #ifdef CONFIG_VT
  929. #if defined(CONFIG_VGA_CONSOLE)
  930. if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
  931. conswitchp = &vga_con;
  932. #elif defined(CONFIG_DUMMY_CONSOLE)
  933. conswitchp = &dummy_con;
  934. #endif
  935. #endif
  936. x86_init.oem.banner();
  937. x86_init.timers.wallclock_init();
  938. x86_platform.wallclock_init();
  939. mcheck_init();
  940. arch_init_ideal_nops();
  941. #ifdef CONFIG_EFI
  942. /* Once setup is done above, unmap the EFI memory map on
  943. * mismatched firmware/kernel archtectures since there is no
  944. * support for runtime services.
  945. */
  946. if (efi_enabled(EFI_BOOT) &&
  947. IS_ENABLED(CONFIG_X86_64) != efi_enabled(EFI_64BIT)) {
  948. pr_info("efi: Setup done, disabling due to 32/64-bit mismatch\n");
  949. efi_unmap_memmap();
  950. }
  951. #endif
  952. }
  953. #ifdef CONFIG_X86_32
  954. static struct resource video_ram_resource = {
  955. .name = "Video RAM area",
  956. .start = 0xa0000,
  957. .end = 0xbffff,
  958. .flags = IORESOURCE_BUSY | IORESOURCE_MEM
  959. };
  960. void __init i386_reserve_resources(void)
  961. {
  962. request_resource(&iomem_resource, &video_ram_resource);
  963. reserve_standard_io_resources();
  964. }
  965. #endif /* CONFIG_X86_32 */